article
stringlengths 0
745k
| abstract
stringlengths 0
92.7k
|
|---|---|
in many recent developments involving cosmological models and brane collisions in higher - dimensional gravity theories @xcite , the dynamical @xmath0-branes carrying charges have played important roles . in the classical solution of a single @xmath0-brane , the coupling of the dilaton to the field strength includes the parameter @xmath3 .
since these brane solutions with @xmath4 are related to well - known d - branes and m - branes in supergravity theories , they certainly exhibit many attractive properties in the higher - dimensional spacetime .
some static solutions with @xmath5 also have supersymmetry after dimensional reductions to lower - dimensional theory @xcite .
the time - dependent generalizations of these solutions are thus important examples of higher - dimensional gravity theories .
the dynamical brane solution with the cosmological constant can be obtained by choosing the coupling constant appropriately @xcite . for a single 2-form field strength and a nontrivial dilaton
, we have found that the dynamical single 0-brane solution describes the milne universe @xcite .
the field equations give an asymptotically de sitter solution if the scalar field is trivial @xcite , which is a generalization of the kastor - traschen solution in the four - dimensional einstein - maxwell theory @xcite .
the construction of intersecting branes with a cosmological constant is a natural generalization of the single cosmological brane solutions .
the time - dependent intersecting branes we have mainly discussed are localized only along the relative or overall transverse directions in a higher - dimensional background , which are delocalized intersecting brane systems .
however , in the higher - dimensional gravity theories , one of the branes is localized along the relative transverse directions but delocalized along the overall transverse directions , which are partially localized branes solutions .
if the background has the cosmological constant , there is little known about the dynamics of the intersecting brane system for not only the delocalized case but also the partially localized one . in the present paper
, we will explore the possible generalization of these solutions to the case of the intersecting brane systems with cosmological constants , although similar single brane solutions have been analyzed in ref .
we recall these arguments for constructions of the solution and modify the ansatz of the fields .
a brane configuration has to satisfy an intersection rule which is an algebraic equation that relates the coupling of the dilaton to the dimensionality of the branes .
the intersection rule implies that only the 0-brane can depend on time and the dynamical 0-brane commutes with the static @xmath0-branes .
we will study the dynamical intersecting brane solutions for not only the delocalized case but also the partially localized one .
the paper is constructed as follows : in secs .
[ sec : pl ] and [ sec : cn ] , we derive the dynamical intersecting brane solutions with cosmological constants in a @xmath6-dimensional theory following the approach developed in ref .
we then illustrate how the dynamical solution of two or @xmath7 intersecting branes arise under the condition of @xmath5 in the @xmath6-dimensional theory .
the spacetime starts with the structure of the combined 0-branes . if they do not have the same charges , a singularity hypersurface appears before they meet as the time decreases for @xmath8 .
we then discuss the dynamics of two 0-branes with static @xmath0-branes ( or the dynamics of two black holes ) in sec .
[ sec : c ] .
if there exists one uncompactified extra dimension [ 0@xmath18-brane system or 0@xmath9-brane systems ( @xmath10 ) ] and two brane systems have the same brane charges , the solution describes a collision of two branes ( or two black holes ) , which is similar to the result in refs .
@xcite . in sec .
[ sec : sug ] , applications of these solutions to five- or six - dimensional supergravity models are discussed .
we consider in detail the construction yielding the dynamical 0- or 1-brane in the nishino - salam - sezgin model .
we also provide brief discussions for a time - dependent brane system in romans supergravity model .
we describe how our universe could be represented in the present formulation via an appropriate compactification and give the application to cosmology .
we show that there exists no accelerating expansion of our universe , although the conventional power - law expansion of the universe is possible . we then discuss the dynamics of two 0- or 1-branes with smeared branes .
if two brane systems have the same brane charges with smearing some dimensions , the solution describes a collision of two brane backgrounds .
there is a curvature singularity in the dynamical brane background if we set a particular value for the constant parameters .
then the solution implies that the presence of the singularities is signaling possible instabilities , making the solutions sick or unphysical .
we study the classical stability of the solutions in sec .
[ sec : in ] .
our preliminary analysis will present that the energy of klein - gordon scalar fields in the dynamical brane background grows with time for inertial observers approaching the singularity . in terms of using the preliminary analysis
performed in refs .
@xcite , the klein - gordon modes will be studied , arriving at the preliminary conclusion of instability .
section [ sec : cd ] will be devoted to the summary and conclusions .
in this section , we will construct the partially localized time - dependent brane systems in @xmath6 dimensions with cosmological constants .
we consider a @xmath6-dimensional theory composed of the metric @xmath11 , the scalar field @xmath12 , cosmological constants @xmath13 , and two antisymmetric tensor field strengths of rank @xmath14 and @xmath15 .
the action in @xmath6 dimensions is given by where @xmath16 denotes the ricci scalar constructed from the @xmath6-dimensional metric @xmath11 , @xmath17 are constants , @xmath18 denotes the @xmath6-dimensional gravitational constant , @xmath19 is the hodge operator in the @xmath6-dimensional spacetime , and @xmath20 and @xmath21 are @xmath22 and @xmath23-form field strengths , respectively . the constant parameters @xmath24 and @xmath25 are defined by respectively .
here @xmath26 is constant .
the @xmath14-form , @xmath15-form field strengths @xmath27 , @xmath28 are given by the @xmath29-form , @xmath30-form gauge potentials @xmath31 , @xmath32 , respectively : for the @xmath6-dimensional action ( [ pl : action : eq ] ) , the field equations read where @xmath33 denotes the laplace operator with respect to the @xmath6-dimensional metric @xmath11 .
the @xmath6-dimensional metric involving the intersecting branes with cosmological constant can be put in the general form where @xmath34 is the @xmath35-dimensional metric depending only on the @xmath35-dimensional coordinates @xmath36 , @xmath37 is the @xmath38-dimensional metric depending only on the @xmath38-dimensional coordinates @xmath39 , @xmath40 is the @xmath41-dimensional metric depending only on the @xmath41-dimensional coordinates @xmath42 and @xmath43 is the @xmath44-dimensional metric depending only on the @xmath44-dimensional coordinates @xmath45 .
here we assume that the parameters @xmath46 and @xmath47 in the metric ( [ pl : metric : eq ] ) are given by the brane configuration is illustrated in table [ table_a ] . .
intersections of @xmath48-branes in the metric ( [ pl : metric : eq ] ) , where @xmath49 . [ cols="^,^,^,^,^,^,^,^,^,^,^,^,^,^,^",options="header " , ] [ table_a ] the dynamical brane solutions are characterized by two warp factors , @xmath50 and @xmath51 , depending on the @xmath44-dimensional coordinates transverse to the corresponding brane as well as the @xmath35-dimensional world - volume coordinate . in the case of intersection involving two branes ,
the powers of warp factors have to obey the intersection rule , and then split the coordinates in four parts @xcite .
one is coordinates of the world - volume spacetime , @xmath36 , which are common to the @xmath52- , @xmath53-branes .
the others are coordinates of the overall transverse space , @xmath45 , and coordinates of the relative transverse space , @xmath39 and @xmath42 , which are transverse to only one of the @xmath52- , @xmath53-branes . in this section ,
we consider the intersections of a @xmath52- and a @xmath53-brane with the following conditions in @xmath6-dimension .
we assume that the functions @xmath50 and @xmath51 depend not only on overall transverse coordinates but also on the corresponding relative coordinates and world - volume coordinates .
we therefore may write @xmath54 .
we give the expression for the field strengths @xmath20 , @xmath21 and scalar field @xmath12 of a @xmath52-brane intersecting with a @xmath53-brane over a @xmath0-brane configuration : where @xmath55 , @xmath56 , and @xmath57 are the volume @xmath58-form , @xmath59-form , and @xmath60-form , respectively here , @xmath61 , @xmath62 , and @xmath63 denote the determinants of the metrics @xmath34 , @xmath37 , and @xmath40 , respectively . in this subsection , we consider the field eqs .
( [ pl : ansatz : eq ] ) with @xmath64 .
the parameters @xmath17 are assumed to be let us first consider the gauge field eqs .
( [ pl : gauge - r : eq ] ) , ( [ pl : gauge - s : eq ] ) . using the assumptions ( [ pl : metric : eq ] ) and ( [ pl : ansatz : eq ] ) , we have where @xmath65 , @xmath66 , and @xmath67 denote the hodge operator on @xmath68 , @xmath69 , and z , respectively , and @xmath70 is given by in the following , we discuss the case of @xmath71 because the relation @xmath71 is consistent with the intersection rule which has been found in @xcite .
setting @xmath71 , the eq .
( [ pl : gauge2-r : eq ] ) gives where @xmath72 , and @xmath73 are the laplace operators on the space of @xmath68 , and z , respectively . on the other hand , eq .
( [ pl : gauge2-s : eq ] ) leads to where we used eq .
( [ pl : chi : eq ] ) and @xmath74 is the laplace operators on the space of @xmath69 , now we consider the einstein eq .
( [ pl : einstein : eq ] ) .
using the ansatz ( [ pl : metric : eq ] ) , ( [ pl : ansatz : eq ] ) , and intersection rule @xmath71 , the einstein equations become where @xmath75 is the covariant derivative constructed from the metric @xmath34 , @xmath76 , @xmath77 , @xmath74 , and @xmath78 are the laplace operators on @xmath79 , @xmath68 , @xmath69 , and @xmath80 , respectively , and @xmath81 , @xmath82 , @xmath83 , and @xmath84 are the ricci tensors with respect to the metrics @xmath85 , @xmath86 , @xmath87 , and @xmath88 , respectively . from eqs .
( [ pl : ceinstein - mm : eq ] ) and ( [ pl : ceinstein - ma : eq ] ) , the warp factors @xmath50 and @xmath51 can be expressed as if we require that the background satisfies the einstein equations ( [ pl : ceinstein : eq ] ) reduce to note that eq .
( [ pl : ceinstein - mn : eq ] ) becomes trivial for @xmath89 . by combining the above equations and setting @xmath89 , the einstein equations for @xmath90 lead to finally
, we check the scalar field equation for the case of @xmath89 . substituting eqs .
( [ pl : ansatz : eq ] ) , ( [ pl : warp : eq ] ) , and ( [ pl : condition : eq ] ) and the intersection rule @xmath71 into eq .
( [ pl : scalar : eq ] ) , we have hence the scalar field equation ( [ pl : scalar - e : eq ] ) reads these are consistent with the einstein equations ( [ pl : solution1:eq ] ) .
the function @xmath50 can depend on the coordinate @xmath91 only if @xmath90 . for @xmath92
, the scalar field equation leads to @xmath93 .
we can find the solution in which the @xmath53-brane part depends on @xmath36 .
for @xmath94 , @xmath93 , and @xmath95 , we have it is clear that there is a solution for @xmath96 such as @xmath97 unless @xmath98 . for @xmath92 ,
the field equations lead to @xmath93 . if @xmath99 and @xmath100 , the warp factors @xmath101 and @xmath102 are trivial functions
. then the @xmath6-dimensional spacetime is no longer warped @xcite .
moreover , eqs .
( [ pl : solution1:eq ] ) and ( [ pl : solution3:eq ] ) imply the two cases .
first , @xmath52- , @xmath53-branes are delocalized .
these are localized only along the overall transverse directions .
second , the 0-brane is completely localized on the @xmath53- ( or @xmath52- ) brane which is localized only along the overall transverse directions , which is a partially localized @xmath103 ( or @xmath104 ) brane system . as an example
, we set where @xmath105 and @xmath106 are the @xmath53- and @xmath107-dimensional euclidean metrics , respectively .
equation ( [ pl : warp1 - 1:eq ] ) gives where @xmath108 and @xmath109 are constants .
hence , solutions exist for @xmath110 if @xmath111 and vice versa . if the functions @xmath101 and @xmath51 satisfy the coupled partial differential equations the harmonic function @xmath51 that satisfies the equation in takes the form where @xmath112 is the location of the @xmath113 th @xmath53-brane and @xmath114 is constant . since we consider the case in which the @xmath53-branes coincide at the same location in the overall transverse directions , the harmonic function @xmath51 can be written by the following form @xcite : where @xmath115 is constant and the stack of @xmath53-branes is located at the same points @xmath116 along the @xmath117 directions
. we can find solutions for the harmonic function @xmath101 in the case where each of the @xmath53-branes does not coincide at the same location in the overall transverse directions . if we set @xmath118 and @xmath119 for the overall transverse space , eq .
can be solved as where @xmath114 is constant and @xmath120 is given by hence , the functions @xmath50 and @xmath51 can be expressed as where @xmath108 , @xmath109 , @xmath114 , and @xmath115 are constant parameters and @xmath121 and @xmath116 are constants representing the positions of the branes .
the curvature singularities appear at @xmath122 in the @xmath6-dimensional metric ( [ pl : metric : eq ] ) .
moreover , there is also a singularity at @xmath123 unless the scalar field is trivial . upon setting @xmath124 and @xmath98 , the solutions of eq .
( [ pl : hrhs : eq ] ) are given by in the case of @xmath125 and @xmath98 , the functions @xmath50 and @xmath51 can be written by the solutions and have a singular hypersurface at infinity as well as at @xmath122 , because the @xmath6-dimensional metric depends on the logarithmic function of the transverse coordinates .
these solutions also give a singularity at @xmath123 if the dilaton is nontrivial .
it is possible to find the solution for @xmath95 and @xmath126 if the roles of @xmath68 and @xmath69 are exchanged .
the solution of the field equations for @xmath127 and @xmath128 can be written by where @xmath129 is given by if we set @xmath130 , @xmath131 , and @xmath92 , the harmonic functions @xmath50 and @xmath51 have logarithmic spatial dependence like and . assuming @xmath111 and introducing a new time coordinate @xmath132 by we find the @xmath6-dimensional metric ( [ pl : metric : eq ] ) as since @xmath51 does not approach constant in any region , the whole spacetime can not be homogeneous and isotropic .
but on each @xmath133 slice the spacetime becomes a homogeneous and isotropic universe . in the limit @xmath134 , the function @xmath101 can be negligible in the warp factor .
this is guaranteed by a scalar field with the exponential potential .
the accelerating universe is obtained on each @xmath133 slice if @xmath135 , which corresponds to the case of a positive cosmological constant . for @xmath136 ,
the solution provides a power - law inflationary universe , and for @xmath137 , the scale factor diverges at @xmath138 , taking the involution @xmath139 . finally , for @xmath140 , we obtain a de sitter universe which will be discussed in the next subsection .
next , we consider the solution with a dilaton which is the case of @xmath141 . in terms of @xmath142 , eq . ( [ pl : c : eq ] ) gives if we assume the field equations reduce to then eq .
( [ ds : warp1:eq ] ) gives where @xmath109 is an integration constant and @xmath108 is given by thus there is no solution for @xmath143 .
if the metric @xmath144 is assumed to be eq .
( [ pl : flat metric : eq ] ) , the function @xmath145 is given by eq .
( [ pl : h_r solution2:eq ] ) .
now we introduce a new time coordinate @xmath132 by where we have set @xmath146 for simplicity .
then the @xmath6-dimensional metric ( [ pl : metric : eq ] ) can be expressed as the function @xmath51 does not become constant in any region .
then , the @xmath6-dimensional spacetime can not be de sitter spacetime
. however , the spacetime gives a homogeneous and isotropic universe on each @xmath147 , @xmath148 slice .
if we set @xmath149const and @xmath150 , eq . ( [ ds : metric : eq ] ) becomes the solution which has been discussed by refs .
@xcite ( see also @xcite ) .
furthermore , for @xmath151 and by setting @xmath152 , the solution is the kastor - traschen one @xcite .
the construction that we have analyzed in sec .
[ sec : pl ] is a special case of a more general construction of intersecting branes with a cosmological constant . in effect , we have been studying the special case of intersections involving a two - brane . the time - dependent brane with a cosmological constant property is a 0-brane , represented by a 2-form . to describe more general intersections on a time - dependent background ,
one simply incorporates additional branes in a dynamical background . without loss of the time dependence ,
it is possible to also add @xmath7 delocalized branes .
this also has one important further refinement . instead of power - law expansion
, the support of a 0-brane might be accelerated expansion , where @xmath6-dimensional geometry is an asymptotically de sitter spacetime .
the @xmath7 intersection allows the time dependence of only 0-branes but not of the @xmath0-branes ( @xmath153 ) .
the reason for this is that the time - dependent brane we have obtained can be performed in the case of @xmath71 , where @xmath70 is defined by .
so the coefficient of the time dependence is simply proportional to the cosmological constant that we have explored in sec [ sec : pl ] : the einstein equations give @xmath154 . in this section ,
we discuss the intersection of the delocalized @xmath7 branes in the higher - dimensional gravity theory with the cosmological constants .
the general action describing the intersection involving the @xmath7 brane system is given by where @xmath18 denotes the @xmath6-dimensional gravitational constant , @xmath16 is the @xmath6-dimensional ricci scalar constructed from the @xmath6-dimensional metric @xmath11 , @xmath12 is a scalar field , @xmath155 is the antisymmetric tensor fields of rank @xmath156 , @xmath157 is the hodge dual operator in the @xmath6-dimensional spacetime , and @xmath24 and @xmath158 are constants defined by here @xmath159 denotes the type of the corresponding branes .
the @xmath6-dimensional action ( [ cn : action : eq ] ) give the field equations where @xmath33 denotes the laplace operator with respect to the @xmath6-dimensional metric @xmath11 .
we adopt the ansatz that @xmath6-dimensional metric can be written by where @xmath88 denotes the metric of the @xmath160-dimensional @xmath80 space which depends only on the @xmath160-dimensional coordinates @xmath45 . concerning the functions @xmath161 , @xmath162 and @xmath163 , we assume where the constants @xmath164 , @xmath165 and @xmath166 are given by the function @xmath167
is a straightforward generalization of the static solution associated with the brane @xmath159 in an intersecting brane system @xcite to the dynamical one .
we further require that the dilaton @xmath12 and the form fields @xmath168 satisfy the following conditions where @xmath169 is the space associated with the brane @xmath159 , and the volume @xmath170-form @xmath171 is written by firstly , we consider the einstein equations ( [ cn : einstein : eq ] ) with @xmath172 .
we assume that the parameters @xmath173 are given by we impose the condition with respect to the components of @xmath6-dimensional metric @xcite the einstein equations become where @xmath84 is the ricci tensor with respect to the metric @xmath88 , and @xmath174 is defined by the eq .
( [ cn : ceinstein - ti : eq ] ) can be rewritten as one can find that the equation ( [ cn : ceinstein - ti2:eq ] ) is equivalent to satisfying that then we have eqs .
( [ cn : c : eq ] ) , ( [ cn : parameter : eq ] ) and ( [ cn : ceinstein - ti2:eq ] ) give combining the with ( [ cn : ceinstein - ti4:eq ] ) the constant @xmath175 in is @xmath176 , which implies taking account of these results , the equation ( [ cn : ceinstein - ti : eq ] ) yields hence we find for @xmath177 , provides the intersection rule on the dimension @xmath178 of the intersection for each pair of branes @xmath159 and @xmath179 @xmath180 @xcite : under the assumptions given above , we next reduce the gauge field equations . in terms of the ansatz ,
the bianchi identity @xmath181 is automatically satisfied ; utilising , the gauge field equation becomes where we used eqs .
( [ cn : extremal : eq ] ) , ( [ cn : warp : eq ] ) , and @xmath182 , @xmath183 are the hodge dual operators on @xmath184 and @xmath80 , respectively .
hence , gives ( [ cn : warp : eq ] ) and we find the roles of the bianchi identity and field equations are interchanged for magnetic ansatz @xcite
. then the net result is the same . in order to complete the system of equations
, we must also consider the scalar field equation .
substituting the ansatz for fields , and the metric ( [ cn : metric : eq ] ) , ( [ cn : warp : eq ] ) , the equation of motion for the scalar field reduces to furthermore , reads from eq .
( [ cn : ki : eq ] ) , we obtain where @xmath185 and @xmath186 are constants .
let us first consider the case that all cosmological constants become nonvanishing . if we set @xmath187 , the field equations imply that all functions are equal : we can find the solutions if the function @xmath188 and @xmath3 satisfy where @xmath189 denotes a constant .
then the remaining einstein equations ( [ cn : ceinstein : eq ] ) are now we set where @xmath106 is the @xmath160-dimensional euclidean metric . in this case
, the solution for @xmath190 can be obtained explicitly as where @xmath191 s are constant parameters and @xmath192 represents the positions of the branes in z space .
if the functions @xmath190 coincide , the locations of the @xmath193-brane will also coincide in this case , all branes have the same total amount of charge at the same position .
let us consider the intersection rule in the @xmath6-dimensional gravity theory .
if we set @xmath194 for all @xmath193 , the intersection rule ( [ cn : rule : eq ] ) leads to then , we find the intersection involving two @xmath195-branes : since the number of intersections for @xmath196 is negative , there is no solution in these brane backgrounds .
if we choose @xmath197 ) , the metric describes the known static and extremal multi - black - hole solution with black hole charges @xmath191 @xcite . in the following ,
we consider the case that there is only one function @xmath190 which depends on both @xmath45 and @xmath91 .
we denote it with the subscript @xmath198 , while other functions of @xmath199 are either dependent on @xmath45 or constant . if we assume @xmath200 , we have we can find the solutions if the function @xmath201 and @xmath202 satisfy where @xmath203 is constant
. then the remaining einstein equations are now we set where @xmath106 is the @xmath160-dimensional euclidean metric . in this case
, the solution for @xmath190 can be written explicitly as where @xmath204 , @xmath205 , @xmath206 , and @xmath207 are constant parameters and @xmath192 and @xmath208 denote the positions of the branes in z space .
@xmath209 leads to @xmath210 and vice versa . since the functions @xmath190
coincide , the locations of the @xmath193-brane will also coincide this physically means that all branes have the same total amount of charge at the same position .
here we have discussed the solution without compactification of @xmath80 space .
if we consider the case that @xmath61 dimensions of @xmath80 space are smeared , we can find the different power of harmonics , i.e. @xmath211 ( @xmath212 ) . for @xmath213 )
, the solution describes the known static and extremal multi - black - hole solution with black hole charges @xmath206 @xcite .
we can find the solution ( [ cn : hi : eq ] ) for any @xmath214 .
if we choose @xmath215 , the solutions have already discussed in ref .
@xcite .
let us consider the intersection rule in the @xmath6-dimensional gravity theory .
if we choose @xmath216 for all @xmath217 , the intersection rule ( [ cn : rule : eq ] ) leads to now we discuss the application of the time - dependent solutions to study the cosmology .
we assume an isotropic and homogeneous three - space in the friedmann - robertson - walker ( frw ) universe after compactification .
we set the @xmath160-dimensional euclidean space with @xmath218 and consider the case that there is only one function @xmath190 depending on both @xmath45 and @xmath91 , which we denote it with the subscript @xmath198 , and other functions are either dependent on @xmath45 or constant .
if we assume @xmath200 , the @xmath6-dimensional metric can be expressed as @xmath219^{a_{\tilde{i}}}d\tau^2{\nonumber}\\ & & + \sum_{\alpha=1}^p\prod_{i\ne \tilde{i}}h_i^{\delta_i^{(\alpha ) } } \left[1+\left(\frac{\tau}{\tau_0}\right ) ^{-\frac{2}{a_{\tilde{i}}+2}}h_{\tilde{i}}\right]^ { \delta_{\tilde{i}}^{(\alpha ) } } \left(\frac{\tau}{\tau_0}\right)^{\frac{2\delta_{\tilde{i}}^{(\alpha ) } } { a_{\tilde i}+2 } } \left(dx^{\alpha}\right)^2{\nonumber}\\ & & + \prod_{i\ne \tilde{i}}h_i^{b_i } \left[1+\left(\frac{\tau}{\tau_0}\right)^{-\frac{2}{a_{\tilde{i}}+2 } } h_{\tilde{i}}\right]^{b_{\tilde{i } } } \left(\frac{\tau}{\tau_0}\right)^{\frac{2b_{\tilde{i}}}{a_{\tilde{i}}+2 } } \delta_{ab}(\zsp)dz^adz^b , \label{cn : metric - j : eq}\end{aligned}\ ] ] where the function @xmath220 is given by and the cosmic time @xmath132 defined by if we can regard the three - dimensional part of the overall transverse space z as our universe , the power of the scale factor in the fastest expanding case is expressed as where we used the @xmath6-dimensional metric ( [ cn : metric - j : eq ] ) .
hence , we can not find the cosmological model which exhibits an accelerating expansion of our universe . on the other hand , if our three - space is given by a three - dimensional subspace in relative transverse space , the power of the scale factor in the fastest expanding case is also given by . by taking @xmath139 , where @xmath221 is constant ,
we have accelerated expansion for @xmath222 and @xmath223 .
this is equivalent to for @xmath224 .
however , the scale factor of our universe diverges at @xmath225 . on the other hand ,
the power of the scale factor in the fastest expanding case is automatically positive for @xmath226 and @xmath227 .
next we discuss the cosmological solution in the lower - dimensional effective theories .
we compactify @xmath228 dimensions to give our universe , where @xmath229 and @xmath230 denote the compactified dimensions with respect to the relative and overall transverse space , respectively .
the @xmath6-dimensional metric ( [ cn : metric : eq ] ) is written by where @xmath231 is a @xmath232-dimensional metric and @xmath233 is a metric of compactified dimensions . in order to discuss the dynamics of the @xmath232-dimensional universe in the einstein frame
, we use the conformal transformation where @xmath234 and @xmath235 are expressed , respectively , as the @xmath232-dimensional metric in the einstein frame is thus given by @xmath236 , \label{cn : metric - e : eq}\end{aligned}\ ] ] where @xmath237 denotes the coordinate of @xmath238-dimensional relative transverse space and @xmath239 is @xmath240-dimensional spaces . if we set @xmath241 , the @xmath232-dimensional metric ( [ cn : metric : eq ] ) in the einstein frame can be expressed as @xmath242 , \label{cn : metric - e2:eq}\end{aligned}\ ] ] where @xmath243 is given by @xmath244 and we define the cosmic time @xmath132 : hence , in the einstein frame , the power of the scale factor in the fastest expanding case is given by if the physical parameters satisfy , the solutions do not give an accelerating expansion in our universe .
these are the similar results with the case of the other partially localized and delocalized intersecting brane backgrounds .
although we find the exact time - dependent brane solution , the power exponent of the scale factor is too small . furthermore , in order to discuss a de sitter solution in an intersecting brane background , one has to consider the trivial dilaton , which will be discussed in the next subsection .
in this subsection , we consider the einstein equations ( [ cn : einstein : eq ] ) with @xmath245 . equation ( [ cn : c : eq ] ) gives if we assume the field equations reduce to then eq .
( [ ds2:warp1:eq ] ) gives where @xmath246 is an integration constant and @xmath108 is given by thus , there is no solution for @xmath247 . if the metric @xmath144 is assumed to be eq .
( [ cn : flat metric : eq ] ) , the function @xmath220 is given by eq .
( [ cn : hi : eq ] ) .
now we introduce a new time coordinate @xmath132 by the @xmath6-dimensional metric ( [ cn : metric : eq ] ) is then rewritten as the @xmath6-dimensional metric ( [ ds2:metric : eq ] ) implies that the spacetime describes an isotropic and homogeneous universe if @xmath248 . in the region where the terms with @xmath220 are negligible and @xmath249
approaches a constant , which is realized in the limit @xmath134 and for @xmath146 , the @xmath6-dimensional spacetime becomes de sitter universe . if we set @xmath250const and @xmath251 , eq .
( [ ds2:metric : eq ] ) becomes the solution which has been discussed by ref .
@xcite ( see also @xcite ) .
furthermore , for @xmath151 and by setting all @xmath252 , the solution is the kastor - traschen one @xcite .
now we will study the spacetime structure .
the metric has singularities at @xmath253 or @xmath254 .
the spacetime is thus not singular when it is restricted inside the domain specified by the conditions where the function @xmath255 is defined in ( [ cn : hi : eq ] ) .
the @xmath6-dimensional spacetime can not be extended beyond this region , because a curvature singularity appears in the @xmath6-dimensional spacetime .
the regular spacetime with branes ends up with the singularities .
since the system with @xmath256 has the time reversal one of @xmath257 , the dynamics of the spacetime depends on the signature of @xmath258 . here
we will consider the case with @xmath256 .
then the function @xmath201 is positive everywhere for @xmath259 and the spacetime is nonsingular . in the limit of @xmath260 and
apart from a position of the branes , near which the geometry takes a cylindrical form of an infinite throat , the solution is approximately described by a time - dependent uniform spacetime .
now we discuss the time evolution for @xmath261 .
the spacetime is regular everywhere and has a cylindrical topology near each brane at @xmath262 .
as time slightly decreases , a curvature singularity appears as @xmath263 .
the singular hypersurface cuts off more and more of the space as time decreases further .
when @xmath91 continues to decrease , the singular hypersurface eventually splits and surrounds each of the @xmath0-brane throats individually . the spatial surface is finally composed of two isolated throats . for @xmath259 ,
the time evolution of the @xmath6-dimensional spacetime is the time reversal of @xmath264 . for any values of fixed @xmath45 in the regular domain in the @xmath6-dimensional spacetime ( [ cn : metric : eq ] ) , the overall transverse space tends to expand asymptotically like @xmath265 .
thus , the solutions describe static intersecting brane systems composed of @xmath0-branes near the positions of the branes , while , in the far region as @xmath266 , the solutions approach de sitter or frw universes with the power - law expansion @xmath265 .
the emergence of time - dependent universes is an important feature of the dynamical brane solutions .
we study the asymptotic behavior of the solutions .
the solution describes a charged black hole in the frw or de sitter universe in the limit of @xmath267 , and @xmath220 vanishes .
first we consider the case of a power - law expanding universe .
the function @xmath201 depends only on time @xmath91 in the far region from branes , and the resulting metric ( [ cn : metric - j : eq ] ) can be expressed as the scale factor of the relative transverse space is given by @xmath268 , while the expansion low for the overall transverse space is written by @xmath269 . on the other hand , for @xmath245 corresponding to de sitter universe ( [ ds2:metric : eq ] ) , the metric of @xmath6-dimensional spacetime in the far region from branes becomes figure [ fig : p ] depicts the conformal diagrams of the frw and de sitter universes . 0.5 cm + ( a ) ( b )
next we discuss the near - horizon geometry of the solutions .
we set the metric of @xmath160-dimensional overall transverse space : where @xmath106 denotes the metric of @xmath160-dimensional flat space and the line elements of a @xmath270-sphere ( @xmath271 ) are given by @xmath272 .
the harmonic function @xmath255 dominates in the limit of @xmath273 , and the time dependence can be ignored .
thus the system becomes static near a position of branes .
when all of the branes are located at the origin of the z spaces , the solutions are rewritten as here @xmath274 and @xmath275 are the mass of @xmath276- , and @xmath277-branes , respectively . in the near - horizon region
@xmath273 , the dependence on @xmath91 in ( [ nh : h : eq ] ) is negligible . then the metric is reduced to the following form : thus the metric ( [ nh : nh : eq ] ) describes a warped product of @xmath278-dimensional spacetime m@xmath279 and @xmath270-dimensional sphere @xmath271 .
hence , the near - brane geometry has the same metric form as the static one .
if it has a horizon geometry , we can obtain a black hole solution in the time - dependent background .
in fact , some solutions , for instance , the m2@xmath1m2@xmath1m2 , m2@xmath1m2@xmath1m5@xmath1m5 intersecting solution in eleven dimensions , give regular black hole spacetimes in the static limit @xcite .
our solution approaches asymptotically the dynamical universe with the scale factor @xmath280 , while the static solution gives a black hole .
then we can regard the present solution as a black hole in the expanding universe .
in this section , we apply our dynamical intersecting brane solutions found in the previous section to brane collisions .
the functions @xmath201 and @xmath249 are assumed to be here @xmath108 and @xmath204 are constants , and the function @xmath220 and @xmath249 are expressed , respectively , as where @xmath281 is constant , @xmath282 and @xmath283 denote the number of smeared dimension for @xmath284-brane and @xmath277-brane , respectively , we assume @xmath285 and @xmath286 , and @xmath287 and @xmath288 are mass constants of 0-brane and @xmath277-branes located at @xmath289 and @xmath290 , respectively . since @xmath249 is the harmonic function on the @xmath291-dimensional euclidean subspace in @xmath80 , we define the metric , scalar , and gauge fields are given by eqs .
( [ cn : metric : eq ] ) and ( [ cn : fields : eq ] ) , respectively . for @xmath292 and @xmath293 ,
these become since the time dependence allows only for the 0-brane , we see that the @xmath294-brane background is critical case .
if we consider the @xmath295-brane , the functions @xmath201 and @xmath249 are written by the sum of linear functions of @xmath117 .
the possibility of brane collisions comes from the difference in the overall transverse dimension . from the solution ( [ cp : h2:eq ] ) , there are curvature singularities at @xmath253 or at @xmath254 in the @xmath6-dimensional background .
note that the regular @xmath6-dimensional spacetime is restricted to the region of @xmath296 and @xmath297 , which is bounded by curvature singularities .
hence , the @xmath6-dimensional metric ( [ cn : metric : eq ] ) is regular if and only if @xmath296 and @xmath297 .
the solution with @xmath298 branes takes the form ( [ cn : metric - j : eq ] ) , where we set @xmath299 and the function @xmath220 is given by ( [ cn : hi : eq ] ) .
we classify the behavior of the harmonic function @xmath249 into two classes : @xmath300 and @xmath301 .
since these depend on the dimensions of the @xmath277-brane , we discuss them below separately . in the case of the @xmath294-brane ,
the harmonic function @xmath249 diverges both at infinity and near @xmath294-branes . since there is no regular spacetime region near branes due to @xmath302
, these solutions are not physically relevant . in the following ,
we discuss the collision involving the @xmath298 brane in @xmath6-dimensional spacetime .
the harmonic function @xmath220 becomes dominant in the limit of @xmath304 , while the function @xmath201 depends only on time @xmath132 in the limit of @xmath267 .
hence , we find a static structure of the @xmath298-brane system near branes . in the far region from branes , the function @xmath220 vanishes .
therefore , the metric can be written by where @xmath305 is defined by in order to analyze the brane collision , we consider a concrete example , in which two @xmath298 branes are located at @xmath306 .
we will discuss the time evolution separately with respect to the signature of a constant @xmath307 , because the behavior of spacetime strongly depends on it .
since the metric function is singular at @xmath308 and @xmath254 , one can note that the regular spacetime exists inside the domain restricted by where the functions @xmath220 and @xmath249 are defined in ( [ cp : h : eq ] ) .
the brane background evolves into a curvature singularity , because the dilaton @xmath12 diverges .
since the @xmath6-dimensional spacetime can not be extended beyond this region , the regular spacetime with two 0-branes ( @xmath309 ) ends on these singular hypersurfaces .
the solution with @xmath310 is the time reversal one of @xmath311 , because the time dependence appears only in the form of @xmath312 . in the following ,
we consider the case with @xmath311 . for @xmath313 , the @xmath6-dimensional spacetime is nonsingular , because the function @xmath201 is positive everywhere . in the limit of @xmath314 ,
the @xmath6-dimensional spacetime becomes asymptotically a time - dependent uniform background , while the cylindrical forms of infinite throats exist near branes . for @xmath315 ,
the spatial metric is initially regular everywhere .
the @xmath6-dimensional spacetime has a cylindrical topology near each brane .
as @xmath132 increases slightly , a singular hypersurface appears from the spatial infinity ( @xmath316 ) .
as @xmath132 increases further , the singularity cuts the space off more and more . since the singular hypersurface eventually splits and surrounds each of the brane throats , the spatial surface is finally composed of two isolated throats .
one notes that the transverse dimensions in the metric ( [ co : sufrace2:eq ] ) expand asymptotically as @xmath317 for fixed spatial coordinates @xmath45 .
the @xmath6-dimensional spacetime becomes static near branes , while the background approaches a frw universe in the far region ( @xmath318 ) .
hence , the time evolution of the four - dimensional universe depends on the position of the observer . for @xmath319 ,
the behavior of @xmath6-dimensional spacetime is the time reversal of the period of @xmath320 .
now we define by using the above equation , the proper distance at @xmath321 between two branes can be written by @xmath322^ { b_{\tilde{i}}/2 } { \nonumber}\\ & & \times \left(1+\frac{q_1}{|z^1+l|^{d-3-p_{i'}-d } } + \frac{q_2}{|z^1-l|^{d-3-p_{i'}-d}}\right)^{b_{i'}/2 } \ , .
\label{cp : distance : eq}\end{aligned}\ ] ] the proper distance is a monotonically increasing function of @xmath132 . we illustrate @xmath323 for the case of the @xmath298 brane system in fig . [ fig : mil ] .
we consider the case of @xmath324 , @xmath325 , @xmath326 , @xmath327 , and @xmath328 or @xmath329 .
it shows that two 0-branes are initially ( @xmath330 ) approaching , the distance @xmath323 takes the minimum finite value at @xmath331 , and then two 0-branes segregate each other .
thus they will never collide .
hence , we can not discuss a brane collision in this case . in the asymptotically power - law expanding universe , the proper distance between two dynamical 0-branes given in ( [ cp : distance : eq ] )
is depicted .
we fix @xmath324 , @xmath328 , @xmath325 , @xmath332 , @xmath333 , and @xmath327 for the @xmath334- ( bold curve ) and @xmath335- ( solid curve ) branes .
the distance decreases initially ( @xmath336 ) but turns to increase at @xmath331 , and then two 0-branes segregate each other .
( b ) we also show the proper distance between two dynamical 0-branes for @xmath334- ( bold curve ) and @xmath335- ( solid curve ) brane systems from the bottom in the case of @xmath324 , @xmath337 , @xmath338 , @xmath333 , @xmath327 , and @xmath328 in the asymptotically power - law expanding universe .
although the proper distance initially decreases as @xmath339 increases , the distance increases as @xmath340 increases.,title="fig : " ] ( -100 , 135)@xmath323 ( 10,10)@xmath132 in the asymptotically power - law expanding universe , the proper distance between two dynamical 0-branes given in ( [ cp : distance : eq ] ) is depicted .
we fix @xmath324 , @xmath328 , @xmath325 , @xmath332 , @xmath333 , and @xmath327 for the @xmath334- ( bold curve ) and @xmath335- ( solid curve ) branes .
the distance decreases initially ( @xmath336 ) but turns to increase at @xmath331 , and then two 0-branes segregate each other .
( b ) we also show the proper distance between two dynamical 0-branes for @xmath334- ( bold curve ) and @xmath335- ( solid curve ) brane systems from the bottom in the case of @xmath324 , @xmath337 , @xmath338 , @xmath333 , @xmath327 , and @xmath328 in the asymptotically power - law expanding universe . although the proper distance initially decreases as @xmath339 increases , the distance increases as @xmath340 increases.,title="fig : " ] ( -105,135)@xmath323 ( 10,10)@xmath132 + ( a ) ( b ) let us next discuss the collision in the @xmath298-brane with a trivial dilaton system .
we consider the case that the harmonic function @xmath220 and @xmath249 are linear in @xmath117 and discuss in detail the @xmath341-brane in @xmath6 dimensions as a example . in this case , we have one extra dimension @xmath117 in @xmath80 space .
the @xmath6-dimensional metric ( [ ds2:metric : eq ] ) can be rewritten by where the function @xmath342 is written by we consider the collision in the @xmath298-brane system with charges @xmath343 and @xmath344 at @xmath345 and the other with charges @xmath346 and @xmath347 at @xmath348 .
the proper distance at @xmath321 between the two @xmath284-branes can be expressed as in the period of @xmath349 , the proper distance increases as @xmath132 increases .
if @xmath350 , a singular hypersurface appears at @xmath351c_0^{-1}<0 $ ] when the distance is still finite .
however , in the case of the equal charges @xmath352 , the situation is completely different , because the proper distance finally vanishes at @xmath353 as then two branes can collide .
a singularity is formed at the same point and time .
let us consider the case @xmath354 .
the @xmath6-dimensional metric ( [ ds2:metric : eq ] ) can be written as since the proper distance at @xmath321 between two branes is given by the distance increases monotonically with respect to @xmath132 . in the case of @xmath349 , initially ( @xmath331 ) , @xmath6-dimensional space is regular except at @xmath355 , while this is an asymptotically time - dependent spacetime and has the cylindrical form of an infinite throat near the 0-brane . at @xmath356 ,
a singularity appears from the spatial infinity ( @xmath316 ) .
as time decreases ( @xmath336 ) , the singular hypersurface erodes the region with the large values of @xmath357 . since
only the region near 0-branes remains regular , eventually it splits and each fragment surrounds each 0-brane individually .
figure [ fig : ds ] shows that this singularity appears before the proper distance @xmath323 vanishes .
hence , the @xmath6-dimensional spacetime has the singularity before two branes collide .
although two 0-branes approach very slowly , a singularity suddenly appears at a finite distance .
then , the spacetime splits into two isolated 0-brane throats . in the asymptotically de sitter universe
, we show the proper distance between two dynamical 0-branes given in ( [ ds : length : eq ] ) .
we set @xmath328 , @xmath358 , @xmath332 , @xmath333 , and @xmath327 for the @xmath334-brane .
the proper distance rapidly vanishes near where two branes collide .
( b ) we also show the proper distance between two dynamical 0-branes for the @xmath334- ( bold curve ) and @xmath335- ( solid curve ) brane systems from the bottom in the case of @xmath337 , @xmath338 , @xmath333 , and @xmath328 in the asymptotically de sitter universe .
the proper distance initially decreases as @xmath132 decreases and remains still finite when a singularity appears.,title="fig : " ] ( -165 , 125)@xmath323 ( 10,10)@xmath132 in the asymptotically de sitter universe , we show the proper distance between two dynamical 0-branes given in ( [ ds : length : eq ] ) .
we set @xmath328 , @xmath358 , @xmath332 , @xmath333 , and @xmath327 for the @xmath334-brane .
the proper distance rapidly vanishes near where two branes collide .
( b ) we also show the proper distance between two dynamical 0-branes for the @xmath334- ( bold curve ) and @xmath335- ( solid curve ) brane systems from the bottom in the case of @xmath337 , @xmath338 , @xmath333 , and @xmath328 in the asymptotically de sitter universe .
the proper distance initially decreases as @xmath132 decreases and remains still finite when a singularity appears.,title="fig : " ] ( -165,125)@xmath323 ( 10,10)@xmath132 + ( a ) ( b ) we show @xmath323 integrated numerically in fig .
[ fig : ds ] for the case of @xmath349 . in the future direction ,
the proper distance @xmath282 increases .
then for @xmath315 , each brane gradually separates as @xmath132 increases .
in the case of ten or eleven dimensions with @xmath4 and @xmath359 , eq .
( [ pl : action : eq ] ) gives the action of supergravities .
for instance , the bosonic part of the action of @xmath360 supergravity includes only 4-form field strength , while , for @xmath328 , the constant @xmath361 is precisely the dilaton coupling for the ramond - ramond @xmath362-form in the type ii supergravities .
the dynamical solutions for the case of @xmath4 have been already discussed in ref .
@xcite . in this section
, we will discuss the time - dependent solution in six - dimensional nishino - salam - sezgin ( nss ) gauged supergravity and romans gauged supergravity models .
the bosonic part of the six - dimensional nss model @xcite is given by the expression ( [ pl : action : eq ] ) with @xmath111 , @xmath363 , while romans six - dimensional gauged supergravity @xcite is expressed by the action ( [ pl : action : eq ] ) with @xmath143 , @xmath363 .
now we consider the nss model among the theories of @xmath364 .
the couplings of the 2-form ( @xmath365 ) and the 3-form ( @xmath366 ) field strengths to the dilaton are given by @xmath367 and @xmath368 , respectively : where @xmath16 denotes the ricci scalar constructed from the six - dimensional metric @xmath11 , @xmath18 is the six - dimensional gravitational constant , @xmath19 is the hodge operator in the six - dimensional spacetime , @xmath12 denotes the scalar field , @xmath369 is the cosmological constant , and @xmath370 and @xmath371 are 2-form and 3-form field strengths , respectively . from eq .
( [ pl : c : eq ] ) , the nss model is realized by choosing @xmath372 , @xmath363 , @xmath373 , and @xmath98 .
the six - dimensional action ( [ nss : action : eq ] ) gives the field equations where @xmath33 denotes the laplace operator with respect to the six - dimensional metric @xmath11 .
we look for solutions whose spacetime metric has the form where @xmath88 is the four - dimensional metric which depends only on the four - dimensional coordinates @xmath45 .
the scalar field @xmath12 and field strengths @xmath370 , and @xmath371 are written , respectively , by first we consider the einstein equation ( [ nss : einstein : eq ] ) . by using the ansatz ( [ nss : metric : eq ] ) and ( [ nss : ansatz : eq ] ) ,
the einstein equations become where @xmath78 denotes the laplace operator on @xmath80 space and @xmath84 is the ricci tensor constructed from the metric @xmath88 .
we next consider the gauge field equations ( [ nss : gauge - r : eq ] ) and ( [ nss : gauge - s : eq ] ) . under the assumption ( [ nss : ansatz : eq ] ) , the gauge field equations are written by where @xmath374 denotes the hodge operator on z and @xmath375 is the volume 4-form on z space : here , @xmath376 is the determinant of the metric @xmath43 .
finally we consider the equation of motion for the scalar field .
substituting the ansatz ( [ nss : ansatz : eq ] ) into eq .
( [ nss : scalar : eq ] ) , we have now we consider the two cases .
one is @xmath377 and @xmath378 .
the other is @xmath379 and @xmath380 . upon setting @xmath381 ,
the field equations reduce to we can also choose the solution in which the 0-brane part depends on @xmath91 .
then , we have where @xmath382 denotes laplace operator with respect to the metric @xmath40 : here , @xmath40 is the five - dimensional metric , and @xmath42 denotes the five - dimensional coordinate . as a special example
, we consider the case where @xmath106 the four - dimensional euclidean metric .
then , the solution for @xmath383 can be obtained explicitly as @xcite where @xmath384 and @xmath290 are constants and the parameter @xmath385 is the mass constant of 1-branes , which is located at @xmath386 .
we can obtain the solution for @xmath387 and @xmath377 if the roles of @xmath388 and @xmath383 are exchanged .
the solution of the field equations is then written as where @xmath389 , @xmath390 , and @xmath391 are constants and @xmath392 .
the delocalized brane solutions in the six - dimensional nss supergravity @xcite have been investigated in refs .
@xcite , including applications to cosmological models .
according to the intersection rule , the number of the intersections dimensions involving the 0-brane and 1-brane is @xmath393 .
although meaningless in ordinary spacetime , these configurations are relevant in the euclidean space , for instance , representing instantons . in the following , we consider cosmological aspects of the solution describing time - dependent branes .
we first study the time dependence of the scale factors in the 0-brane solutions after compactifying the extra directions , and our universe is discussed .
next we discuss the dynamical 1-brane solution and apply it to the cosmology .
for the solution ( [ nss : solution3:eq ] ) , we introduce a new time coordinate @xmath132 as the six - dimensional metric is thus given by where @xmath394 is the five - dimensional euclidean metric and @xmath395 is defined by here @xmath396 denotes the number of smeared dimensions and should satisfy @xmath397 .
the six - dimensional spacetime implies @xmath398 in the limit @xmath134 .
then the scale factor of the six - dimensional space is proportional to @xmath132 .
although the dynamical 0-brane solutions can not give a realistic universe such as an accelerating expansion , a matter- or a radiation - dominated era , there is a possibility that appropriate compactification and smearing of the transverse space to the 0-brane may lead to a realistic expansion .
now we will discuss this possibility .
we consider some compactification and smearing of the extra directions of the solutions .
our universe has to be described by the 0-brane solution with six directions .
since the time direction is expressed as @xmath91 , the remaining task is to identify the three spatial directions from the coordinates @xmath42 . in an approach such as the construction of the cosmological scenario on the basis of a dynamical brane background ,
three spatial directions are supposed to be on the overall transverse space to branes .
if the spatial directions are specified with @xmath42 , it also works in the present case .
then space is isotropic from the expression of the metric .
now we look for a way to realize an isotropic and homogeneous three - dimensional space in the 0-brane solutions .
since we set the coordinates @xmath399 which describes our universe , it is convenient to decompose the six - dimensional metric of the solutions into the following form : where each part of the six - dimensional metric is given by here @xmath400 is the metric of the four - dimensional spacetime with @xmath401 @xmath402 , while @xmath403 denotes the metric of the internal space .
we can obtain the compactifications of the solutions depending on the internal space .
the internal space is described by the coordinates @xmath404 , and the spatial part of our universe @xmath405 is three - dimensional with @xmath406 .
then @xmath405 and @xmath105 are the three- and two - dimensional euclidean metrics , respectively .
now we derive the lower - dimensional effective theory by compactifying the extra directions . in order to find a realistic universe , we compactify the @xmath282-dimensional space to be a @xmath282-dimensional torus , where @xmath282 is the compactified dimensions for the direction of internal space .
the remaining noncompact space is the external space .
the range of @xmath282 is given by @xmath407 , because the @xmath408 direction is preserved to measure the position of the universe in the overall transverse space .
hence the @xmath409 direction will be compactified , where the compactified direction has to be smeared out before the compactification .
then the metric ( [ nss : metric : eq ] ) with @xmath387 is recast into the following form : where @xmath410 is the metric of @xmath411-dimensional external spacetime and @xmath403 is the metric of compactified dimensions . upon setting @xmath412 ,
the compactified metric in the einstein frame is where @xmath413 is the five - dimensional metric in the einstein frame and the constant parameters @xmath307 and the cosmic time @xmath132 are defined , respectively , as since the power exponent of the scale factor is given by 1 , the metric of four - dimensional spacetime in the einstein frame implies that the solutions gives rise to a milne universe . to construct a realistic cosmological model such as in the inflationary scenario
, it would be necessary to add some new ingredients in the background .
figure [ fig : milne ] depicts the conformal diagrams of the five - dimensional spacetime in the limit @xmath134 .
hence , the asymptotic regions of the present spacetime ( [ nss : emetric : eq ] ) resemble the five - dimensional milne universe .
0.5 cm + ( a ) ( b ) finally , we discuss the near - horizon geometry of the 0-brane solution . when all of the 0-branes are located at the origin of the overall transverse space , the solution can be expressed as where @xmath414 is the total mass of 0-branes in the near - horizon limit @xmath273 , the dependence on @xmath91 in ( [ nh : h2:eq ] ) is negligible .
the six - dimensional metric is thus reduced to the following form : where @xmath415 has been performed .
the line elements of a two - dimensional ads space ( ads@xmath416 ) and a four - sphere with the unit radius ( @xmath417 ) are given by @xmath418 and @xmath419 , respectively .
then the six - dimensional metric ( [ nh : nh2-metric : eq ] ) in the near - horizon limit of the 0-brane system describes a warped product of ads@xmath420 and @xmath417 .
figure [ fig : milne ] shows the geometry of the ads@xmath420 and @xmath417 .
now we discuss the cosmological evolution for the time - dependent 1-brane solution ( [ nss : solution : eq ] ) .
we define the cosmic time @xmath132 , which is given by the six - dimensional metric is expressed as where @xmath421 is defined by here @xmath396 is the number of smeared dimensions and should satisfy @xmath422 . in order to fix the location of our universe in the transverse space , let us assume that at least one direction of @xmath423 is not smeared .
now we apply the 1-brane solution to the lower - dimensional effective theory .
let us consider a compactification and smearing of the extra directions of the 1-brane solution .
first of all , our universe is described by the solutions with the six - dimensional coordinates @xmath424 .
the time direction is written by @xmath91 .
our choice is to take the three - dimensional from the overall transverse space with @xmath45 .
the four - dimensional universe is spanned by @xmath91 , @xmath425 , @xmath426 , and @xmath427 , for instance .
the @xmath428 direction is preserved to measure the position of our universe in the overall transverse space of the 1-brane .
since the metric depends on @xmath45 explicitly , we have to smear out @xmath429 , @xmath430 , and @xmath431 so as to define our universe . then the number of the smeared directions @xmath396 should satisfy the condition @xmath432 .
it is necessary to take that @xmath433 and @xmath434 in ( [ nss : metric : eq ] ) to compactify the @xmath435 direction .
we compactify the @xmath435 direction to fit our universe , where @xmath435 denotes the compactified dimensions with respect to the world volume of the 1-brane .
the metric ( [ nss : metric : eq ] ) with @xmath381 is then described by ( [ nss : c - metric3:eq ] ) . in terms of the conformal transformation we can rewrite the @xmath411-dimensional metric in the einstein frame .
if we set @xmath412 , the five - dimensional metric in the einstein frame is where @xmath413 is the metric of five - dimensional external spacetime in the einstein frame .
for @xmath436 , the metric ( [ nss : metric5:eq ] ) is thus rewritten as where the spatial part of our universe @xmath405 is three - dimensional with @xmath437 , and the constant parameters @xmath307 and the cosmic time @xmath132 are defined , respectively , as unfortunately , the power exponent of the four - dimensional universe in the einstein frame becomes 1/4 .
hence , we have to conclude that , in order to obtain a realistic expansion of the universe in this type of models , one has to include additional fields on the background .
let us finally consider the case of the near - horizon limit that the spacetime metric and the functions @xmath388 and @xmath383 satisfy ( [ nss : solution : eq ] ) . if we consider the case where @xmath3 1-branes are located at the origin of the z space , we have where @xmath115 is the total mass of 1-branes since the dependence on @xmath91 and @xmath435 in ( [ nh : h3:eq ] ) is negligible in the near - horizon limit @xmath273 , the six - dimensional metric is reduced to the following form : where @xmath438 has been used .
the line elements of a three - dimensional space ( m@xmath439 ) and a three - sphere are given by @xmath440 , @xmath441 , respectively .
thus we see that the near - horizon limit of the 1-brane system is a warped product of m@xmath439 with a certain internal 3-space with a circle .
we next study the behavior of the time - dependent 0-brane solution ( [ nss : solution3:eq ] ) . substituting ( [ nss : solution3:eq ] ) into the metric ( [ nss : metric : eq ] ) , the six - dimensional metric is expressed as where @xmath40 is given by ( [ nss : lapw : eq ] ) , and the function @xmath395 is defined by ( [ nss : bh : eq ] ) . since the time dependence appears through the function @xmath388 , the next task is to study the time evolution of the solutions carefully
. hereafter we will consider it by focusing upon collision of 0-branes .
we also discuss smearing out some of the directions in the transverse space to decrease the number of transverse dimensions to 0-brane effectively .
now we consider the case that the number of the smeared direction is given by @xmath396 .
then the function @xmath395 can be expressed as where @xmath396 is the number of smeared dimensions and should satisfy @xmath397 , and we assume that at least one direction of @xmath442 is not smeared in order to fix the location of our universe in the transverse space . in the following , we will use the function ( [ nss : sme : eq ] ) .
we will discuss the asymptotic behavior of the time - dependent solutions . in the limit of @xmath443 , the time dependence in the function @xmath388
can be ignored because the harmonic function @xmath395 dominates near a position of 0-brane . on the other hand ,
function @xmath395 vanishes in the limit of @xmath444 .
then the system becomes static near 0-brane while @xmath388 depends only on time @xmath91 in the far region from 0-branes .
thus the six - dimensional metric in the limit of @xmath444 , is rewritten by the metric has singularity at @xmath445 .
then the spacetime is regular if it is restricted inside the domain specified by the conditions , where the function @xmath395 is defined in ( [ nss : sme : eq ] ) .
since the spacetime evolves into a curvature singularity , the six - dimensional spacetime can not be extended beyond this region .
the regular spacetime with 0-branes ends up with the singularities .
the evolution of the spacetime highly depends on the signature of @xmath446 .
the system with @xmath447 has the time reversal one of @xmath448 .
now we will discuss the case with @xmath448 .
for @xmath264 , the spacetime is not singular because the function @xmath388 is positive everywhere . in the limit of @xmath449 ,
the solution is approximately given by a time - dependent uniform spacetime apart from a position of 0-branes .
in the vicinity of branes , the geometry takes a cylindrical form of infinite throat .
we study the time evolution for @xmath259 and @xmath450 . at @xmath262 , the spacetime is regular everywhere and has a cylindrical topology near each 0-brane .
as time slightly evolves , a curvature singularity appears as @xmath451 .
the singular hypersurface cuts off more and more of the space as time increases further .
when time continues to evolve , the singular hypersurface eventually splits and surrounds each of the 0-brane throats individually .
hence the spatial surface is composed of each isolated throats . for @xmath264 , the time evolution of the six - dimensional spacetime
is the time reversal of @xmath259 . since the metric ( [ nss : surface2:eq ] ) in the regular domain implies that the overall transverse space tends to expand asymptotically like @xmath452 , for any values of fixed @xmath42 , the solutions describe static 0-branes near the positions of the branes . in the far region as @xmath453 ,
the solutions approach frw universes with the power law expansion @xmath452 .
the emergence of frw universes is an important feature of the time - dependent 0-brane solutions .
we will discuss whether two 0-branes can collide or not .
we put the two 0-branes at @xmath454 and @xmath455 , where @xmath456 is a constant .
if we introduce the following quantity the proper distance at @xmath457 between the two 0-branes is given by where @xmath458 and @xmath459 are the charges of the 0-brane . for @xmath460 ,
this is a monotonically decreasing function of @xmath91 .
the behavior of the proper length is different depending on the number of the smeared directions @xmath396 .
we will discuss it for each of the values of @xmath396 below .
first we consider the case with @xmath461 .
the proper length is plotted in fig .
[ fig : nss0 ] for the cases with @xmath462 and @xmath463 .
since both cases show that a singularity appears before the proper distance becomes zero , the singularity between two 0-branes appears before collision .
the two 0-branes approach very slowly , and then the singular hypersurface suddenly appears at a finite proper distance .
the spacetime finally splits into two isolated 0-brane throats
. therefore one can not see collision of the 0-branes in these examples . for the other case with @xmath464 ,
the result is the same .
( a ) and @xmath463 ( b ) in the six - dimensional nishino - salam - sezgin gauged supergravity . for both cases , the two 0-brane charges are identical , @xmath465 and the parameters are taken as @xmath450 , @xmath466 , @xmath460 , and @xmath467 .
the result is also the same and a singularity develops before collision of 0-branes.,title="fig : " ] ( -175,120)@xmath468 ( 10,10)@xmath91 2.0 cm ( a ) and @xmath463 ( b ) in the six - dimensional nishino - salam - sezgin gauged supergravity . for both cases ,
the two 0-brane charges are identical , @xmath465 and the parameters are taken as @xmath450 , @xmath466 , @xmath460 , and @xmath467 .
the result is also the same and a singularity develops before collision of 0-branes.,title="fig : " ] ( -175,120)@xmath468 ( 10,10)@xmath91 + ( a ) ( b ) next we consider the case with @xmath469 , and assume that the @xmath42 directions apart from @xmath408 are smeared .
since the function @xmath470 is linear in @xmath471 , the behavior of the proper distance is different from the previous case .
the six - dimensional metric is now given by ( [ nss : surface : eq ] ) . by choosing @xmath472 , the harmonic function @xmath470 is written by we discuss the time - dependent solutions in the case that one 0-brane charge @xmath458 is located at @xmath473 and the other @xmath459 at @xmath474 .
the proper length between the two 0-branes is given by for @xmath460 , the proper distance decreases with time .
if we set @xmath475 , a singularity appears again at a certain finite time @xmath476 , while the proper distance is still finite , where @xmath477 is defined as this is the same result as the case with @xmath478 . on the other hand ,
two 0-branes have the same brane charge @xmath479 , the proper distance vanishes at a certain finite time @xmath480 , where @xmath481 is defined by hence two 0-branes can collide completely . in terms of @xmath481 ,
the proper length is expressed as if we choose the values as @xmath450 , @xmath466 , @xmath467 and @xmath460 , the proper distance @xmath468 is plotted in fig .
[ fig : nss ] for the two cases ( a ) the same 0-brane charges @xmath465 and ( b ) different charges @xmath482 , @xmath483 . in the case
( a ) the two 0-branes can collide completely .
however , in the case ( b ) a singularity appears before collision , as we have already discussed analytically .
( a ) and @xmath482 , @xmath483 ( b ) in the six - dimensional nishino - salam - sezgin gauged supergravity .
we fix @xmath484 , @xmath450 , @xmath466 , @xmath460 , and @xmath467 .
the proper distance rapidly vanishes near two 0-branes collide for the case of @xmath465 .
while for the case of @xmath482 , @xmath483 , it is still finite when a curvature singularity appears.,title="fig : " ] ( -175,115)@xmath468 ( 10,10)@xmath91 2.0 cm ( a ) and @xmath482 , @xmath483 ( b ) in the six - dimensional nishino - salam - sezgin gauged supergravity .
we fix @xmath484 , @xmath450 , @xmath466 , @xmath460 , and @xmath467 .
the proper distance rapidly vanishes near two 0-branes collide for the case of @xmath465 . while for the case of @xmath482 , @xmath483 , it is still finite when a curvature singularity appears.,title="fig : " ] ( -175,115)@xmath468 ( 10,10)@xmath91 + ( a ) ( b ) now we apply our time - dependent solutions to a collision of 1-brane systems . in the case of @xmath381 , the function @xmath383
is assumed to be where @xmath485 is a constant parameter , we choose @xmath486 , and the harmonic function @xmath487 is expressed as here @xmath385 are charges of 1-branes located at @xmath386 and because the harmonic function @xmath487 is defined on the @xmath488-dimensional euclidean subspace in @xmath80 .
the six - dimensional metric , scalar field , and gauge field of the solution are given by eqs .
( [ nss : metric : eq ] ) , and ( [ nss : ansatz : eq ] ) , respectively .
we see that @xmath463 case is critical . for @xmath489 , the function @xmath487 is written by the sum of linear functions of @xmath117 .
the possibility of 1-brane collisions depends on the difference in the transverse dimensions because the behaviors of the gravitational field in the transverse space depends on the number of the transverse dimensions .
although the six - dimensional metric ( [ nss : metric : eq ] ) is regular if and only if @xmath490 , the spacetime shows curvature singularities at @xmath491 .
hence the regular six - dimensional spacetime is restricted to the region of @xmath490 , which is bounded by curvature singularities .
let us study the time evolution for time - dependent 1-brane solution ( [ nss : solution : eq ] ) .
we perform the following coordinate transformation : @xmath492 if we choose @xmath486 , we find @xmath493^{-1/2}\tilde{t}^{-1 } \left[-d\tilde{t}^2+\tilde{t}^2\left\{d\tilde{y}^2 + \left(1+\tilde{t}^{-2}\tilde{h}(\tilde{z})\right ) \delta_{ab}d\tilde{z}^ad\tilde{z}^b\right\ } \right ] \nonumber \\ \hspace{-0.5cm}&=&\frac{2}{\lambda}\ , \left[1 + 16\bar{t}^{-4}\tilde{h}(\tilde{z})\right]^{-1/2 } \left[-d\bar{t}^2+\frac{1}{4}\,\bar{t}^2 \left\{d\tilde{y}^2 + \left(1 + 16\bar{t}^{-4}\tilde{h}(\tilde{z})\right ) \delta_{ab}d\tilde{z}^ad\tilde{z}^b\right\}\right ] , \label{nsc : sufrace : eq_m}\end{aligned}\ ] ] where @xmath494 , @xmath495 , and @xmath496 are defined , respectively , by @xmath497 here , @xmath498 obeys @xmath499 .
the six - dimensional metric ( [ nsc : sufrace : eq_m ] ) represents a homogeneous and isotropic spacetime whose scale factor evolves as the cosmic time @xmath495 , which is described as the milne universe .
hence , we can consider that the present solution with @xmath369 gives a system of 1-branes in the milne universe .
the existence of the expanding milne universe is guaranteed by the scalar field with the exponential potential in the six - dimensional action ( [ nss : action : eq ] ) .
now let us consider the collision of 1-branes .
the solution ( [ nss : metric : eq ] ) without 0-branes can be written in the form where we choose @xmath486 , @xmath43 denotes the four - dimensional metric , and the function @xmath500 is given by ( [ nsc : h2:eq ] ) .
the behavior of the harmonic function @xmath500 is divided into two classes depending on the dimensions of the 1-brane , that is , @xmath501 and @xmath463 , which we will study below separately . for @xmath463 ,
the harmonic function @xmath500 diverges both at infinity and near 1-branes .
in particular , there is no regular spacetime region near 1-branes , because @xmath502 .
hence , such a 1-brane solution is not physically relevant .
since the harmonic function @xmath500 becomes dominant in the limit of @xmath503 ( near 1-branes ) , we find a static structure of the 1-brane system . in the far region from 1-branes , that is , in the limit of @xmath504 , the function @xmath383 depends only on time @xmath91 , because @xmath500 vanishes .
the metric is thus written by in the following , we will analyze one concrete example , in which two 1-branes are located at @xmath505 and @xmath506 in order to study in more detail . since the metric function is singular at @xmath491 , the regular spacetime exists inside the domain restricted by where the function @xmath500 is given by ( [ nsc : h2:eq ] ) .
the six - dimensional spacetime can not be extended beyond this region , because not only does the dilaton @xmath12 diverge but also the spacetime evolves into a curvature singularity .
the regular spacetime with two 1-branes ends on these singularities .
the time dependence appears in the form of @xmath507 . for @xmath508 and @xmath509 ,
the function @xmath188 is positive everywhere and the six - dimensional spacetime is not singular .
it is asymptotically a time - dependent uniform spacetime except for near branes in the limit of @xmath503 , where the background geometry becomes the cylindrical forms of infinite throats when @xmath261 , the spatial metric is initially ( @xmath510 ) regular apart from @xmath511 , and the spacetime has a cylindrical topology near each 1-brane . as @xmath91 evolves slightly , a curvature singularity appears at @xmath511 and from a far region ( @xmath512 ) .
as @xmath91 evolves further , the singularity cuts off the space .
as the time continues to increase , the singular hypersurface eventually splits and surrounds each of the 1-brane throats individually .
then the spatial surface is composed of two isolated throats .
the six - dimensional metric ( [ nsc : surface2:eq ] ) implies that the transverse dimensions expand asymptotically as @xmath513 for fixed spatial coordinates @xmath435 and @xmath45 .
however , this is observer dependent , because it becomes static near branes , and the spacetime approaches a friedmann - robertson - walker universe in the far region ( @xmath514 ) , which expands in the background isotropically .
if we define the proper length at @xmath515 between two 1-branes is written by @xmath516^{\frac{1}{4 } } \ , .
\label{nsc : distance : eq}\end{aligned}\ ] ] this is a monotonically decreasing function of time for @xmath261 . in fig .
[ fig : ns ] , we show @xmath517 for the case of the 1-brane system .
we set @xmath518 , @xmath519 , @xmath509 , and @xmath520 .
all of the six - dimensional space is initially ( @xmath521 ) regular except at @xmath522 and @xmath523 .
although the spacetime becomes asymptotically time dependent and has the cylindrical form of an infinite throat near the 1-brane , the singularity appears from a far region ( @xmath523 ) and @xmath522 .
as time increases ( @xmath264 ) , the singularity erodes the region with the large @xmath524 region .
the region of transverse space is also invaded in time . as a result
, only the region of small @xmath524 and near 1-branes remains regular .
when we study the evolution on the @xmath435 and @xmath45 plane , the singularity appears at infinity @xmath512 , @xmath522 , and comes to the region of two 1-branes .
a singular hypersurface eventually surrounds each 1-brane individually , and then the regular regions near 1-branes split into two isolated throats . for the period of @xmath259
, we find the time - reversed behavior of the case of @xmath264 .
figures [ fig : ns ] and [ fig : ns1 ] show that this singularity appears before the distance @xmath282 vanishes . and ( b ) @xmath464 in the six - dimensional nishino - salam - sezgin gauged supergravity .
we fix @xmath509 , @xmath520 , @xmath519 , and @xmath518 .
the proper distance rapidly vanishes near where two 1-branes collide for the case of @xmath489 , while for the case of @xmath464 , it is still finite when a curvature singularity appears.,title="fig : " ] ( -195 , 80)@xmath517 ( -105 , 0)@xmath91 ( -5,40)@xmath435 2.cm and ( b ) @xmath464 in the six - dimensional nishino - salam - sezgin gauged supergravity .
we fix @xmath509 , @xmath520 , @xmath519 , and @xmath518 .
the proper distance rapidly vanishes near where two 1-branes collide for the case of @xmath489 , while for the case of @xmath464 , it is still finite when a curvature singularity appears.,title="fig : " ] ( -195 , 80)@xmath517 ( -105 , 0)@xmath91 ( -5,40)@xmath435 + ( a ) ( b ) then a singularity between two branes forms before their collision except for @xmath489 .
two 1-branes approach very slowly , a singularity suddenly appears at a finite distance , and the six - dimensional spacetime splits into two isolated 1-brane throats . )
is depicted for ( a ) @xmath337 , @xmath338 and ( b ) @xmath525 , @xmath338 in the six - dimensional nishino - salam - sezgin gauged supergravity .
we fix @xmath509 , @xmath462 , @xmath519 , and @xmath518 . in both cases
, a singularity appears at @xmath526 when the distance is still finite.,title="fig : " ] ( -195 , 80)@xmath517 ( -105 , 0)@xmath91 ( -5,40)@xmath435 2.cm ) is depicted for ( a ) @xmath337 , @xmath338 and ( b ) @xmath525 , @xmath338 in the six - dimensional nishino - salam - sezgin gauged supergravity .
we fix @xmath509 , @xmath462 , @xmath519 , and @xmath518 . in both cases
, a singularity appears at @xmath526 when the distance is still finite.,title="fig : " ] ( -195 , 80)@xmath517 ( -105 , 0)@xmath91 ( -5,40)@xmath435
+ ( a ) ( b ) on the other hand , we can discuss a brane collision for @xmath489 and @xmath264 .
if @xmath350 , a singularity appears at @xmath526 when the distance is still finite ( see fig . [ fig : ns2 ] ) .
this is just the same as the case in sec .
[ sec : ns - c ] .
however , if @xmath527 , the result completely changes ( fig .
[ fig : ns ] ) .
since the distance eventually vanishes at @xmath480 , two 1-branes collide with each other .
the proper length for fixed @xmath435 decreases as time increases from @xmath521 , and it eventually vanishes at @xmath480 .
hence , one 1-brane approaches the other as time evolves , causing the complete collision at @xmath480 .
if we fix the 1-brane charges such that @xmath527 , the branes first collide at larger @xmath524 , and as time progresses , the subsequent collisions occur at the smaller @xmath524 .
we show @xmath517 integrated numerically in figs .
[ fig : ns ] @xmath1 [ fig : ns2 ] . , @xmath338 and ( b ) @xmath525 , @xmath338 in
the six - dimensional nishino - salam - sezgin gauged supergravity .
we fix @xmath509 , @xmath489 , @xmath519 , and @xmath518 . for @xmath350 ,
a singularity appears at @xmath526 when the distance is still finite .
then , the solution does not describe the collision of two 0-branes.,title="fig : " ] ( -195 , 80)@xmath517 ( -105 , 0)@xmath91 ( -5,40)@xmath435 2.cm , @xmath338 and ( b ) @xmath525 , @xmath338 in the six - dimensional nishino - salam - sezgin gauged supergravity .
we fix @xmath509 , @xmath489 , @xmath519 , and @xmath518 . for @xmath350 ,
a singularity appears at @xmath526 when the distance is still finite .
then , the solution does not describe the collision of two 0-branes.,title="fig : " ] ( -195 , 80)@xmath517 ( -105 , 0)@xmath91 ( -5,40)@xmath435 + ( a ) ( b ) we also calculate the distance @xmath517 at @xmath528 and @xmath515 between two branes before the singularity appears except for the case of @xmath489 if @xmath529 .
the proper length is also given by eq .
( [ nsc : distance : eq ] ) . in the present case
, @xmath282 is a monotonically decreasing function of @xmath530 when @xmath264 .
we show the time evolution of the distance in fig .
[ fig : ns3 ] for the case of @xmath529 . on the other hand , for the case of @xmath531 ,
a singularity appears , when the proper distance is still finite . for the period of @xmath259 , the behavior of six - dimensional spacetime
is the time reversal of the period of @xmath264 .
we show the proper distance @xmath468 integrated numerically in fig .
[ fig : ns3 ] for the cases of @xmath489 and @xmath532 . and
@xmath515 for the case of @xmath462 in the six - dimensional nishino - salam - sezgin gauged supergravity is depicted .
we fix @xmath450 , @xmath519 , and @xmath518 . for @xmath533 the proper length decreases as time increases .
the bold line denotes the case of @xmath520 , while the solid one corresponds to the @xmath337 , @xmath338 case .
( b ) for the case of @xmath529 in the six - dimensional nishino - salam - sezgin gauged supergravity , the time evolution of the proper distance between two dynamical 1-branes at @xmath534 and @xmath535 given in ( [ nsc : distance : eq ] ) is depicted .
we fix @xmath450 , @xmath519 , and @xmath518 .
we show the lengths for @xmath462 ( bold line ) , @xmath464 ( solid line ) , and @xmath489 ( dashed line ) .
the proper distance rapidly vanishes near where two 1-branes collide in the case of @xmath489 , while in the case of @xmath532 , is still finite when a curvature singularity appears.,title="fig : " ] ( -95,120)@xmath468 ( 10,10)@xmath91 2.0 cm and @xmath515 for the case of @xmath462 in the six - dimensional nishino - salam - sezgin gauged supergravity is depicted .
we fix @xmath450 , @xmath519 , and @xmath518 . for @xmath533 the proper length decreases as time increases .
the bold line denotes the case of @xmath520 , while the solid one corresponds to the @xmath337 , @xmath338 case .
( b ) for the case of @xmath529 in the six - dimensional nishino - salam - sezgin gauged supergravity , the time evolution of the proper distance between two dynamical 1-branes at @xmath534 and @xmath535 given in ( [ nsc : distance : eq ] ) is depicted .
we fix @xmath450 , @xmath519 , and @xmath518 .
we show the lengths for @xmath462 ( bold line ) , @xmath464 ( solid line ) , and @xmath489 ( dashed line ) .
the proper distance rapidly vanishes near where two 1-branes collide in the case of @xmath489 , while in the case of @xmath532 , is still finite when a curvature singularity appears.,title="fig : " ] ( -95,120)@xmath468 ( 10,10)@xmath91 + ( a ) ( b ) similarly , for the six - dimensional romans theory @xcite , following the discussion in ref .
@xcite , the coupling of the 3-form and of the 2-form field strengths to the dilaton are given by @xmath536 and @xmath537 , respectively : where @xmath16 denotes the ricci scalar constructed from the six - dimensional metric @xmath11 , @xmath18 is the six - dimensional gravitational constant , @xmath19 is the hodge operator in the six - dimensional spacetime , @xmath12 denotes the scalar field , @xmath538 is cosmological constants , and @xmath371 and @xmath370 are 3-form and 2-form field strengths , respectively .
this has a negative scalar potential . in terms of eq .
( [ pl : c : eq ] ) , romans model is given by choosing @xmath539 , @xmath363 , @xmath373 , and @xmath98 .
from the six - dimensional action ( [ ro : action : eq ] ) , we find the field equations : where @xmath33 denotes the laplace operator with respect to the six - dimensional metric @xmath11 .
we assume the six - dimensional metric of the form ( [ nss : metric : eq ] ) .
the scalar field and the gauge field strengths are assumed to be the einstein equations ( [ ro : einstein : eq ] ) then reduce to where @xmath78 denotes the laplace operator on z space and @xmath84 is the ricci tensor constructed from the metric @xmath88 .
we next consider the gauge field . under the ansatz ( [ ro : ansatz : eq ] ) ,
the bianchi identity is automatically satisfied .
also the equation of motion for the gauge field becomes where @xmath374 denotes the hodge operator on z. although the roles of the bianchi identity and field equations are interchanged , the net result is the same .
finally , we consider the equation of motion for the scalar field . substituting the scalar field and the gauge field in ( [ ro : ansatz : eq ] ) into the equation of motion for the scalar field ( [ ro : scalar : eq ] ) , we have then , the functions @xmath388 and @xmath383 satisfy the equations where laplace operator @xmath382 is defined in eq .
( [ nss : lapw : eq ] ) .
if we set @xmath381 , the field equations give now we will focus upon a case by imposing the conditions where @xmath106 is the four - dimensional euclidean metric .
then , the solution for @xmath383 can be obtained explicitly as where @xmath384 are constants .
one can easily get the solution for @xmath387 , @xmath540 and @xmath377 if the roles of @xmath388 and @xmath383 are exchanged .
the solution of field equations is thus expressed as where @xmath541 , @xmath390 , and @xmath391 are constants , and the five - dimensional coordinate @xmath42 defined by .
hence , there is no cosmological 0-brane solution in terms of the ansatz of fields ( [ nss : metric : eq ] ) and ( [ ro : ansatz : eq ] ) if @xmath540 .
finally , we consider the five - dimensional romans theory @xcite . the five - dimensional action is given by where the expectation values of yang - mills potential is assumed to vanish , @xmath16 denotes the ricci scalar constructed from the five - dimensional metric @xmath11 , @xmath18 is the five - dimensional gravitational constant , @xmath19 is the hodge operator in the five - dimensional space - time , @xmath12 denotes the scalar field , @xmath542 is cosmological constant , @xmath370 and @xmath543 are two - form field strengths , and the couplings of the 2-form field strengths and cosmological constant to the dilaton are given by @xmath544 , @xmath545 , @xmath546 , in the action ( [ pl : action : eq ] ) , respectively .
this has also a negative scalar potential . in terms of eq .
( [ pl : c : eq ] ) , romans five - dimensional model is given by setting @xmath547 , @xmath363 , @xmath373 and @xmath98 .
the five - dimensional action ( [ ro5:action : eq ] ) , gives the field equations : where @xmath33 denotes the laplace operator with respect to the five - dimensional metric @xmath11 .
we assume the five - dimensional metric of the form where @xmath88 is the four - dimensional metric which depends only on the four - dimensional coordinates @xmath45 . the scalar field and the gauge field strengths
are assumed to be then , the field equations are reduced to where @xmath78 is the laplace operator on @xmath80 space , and @xmath84 is the ricci tensor with respect to the metric @xmath88 . setting @xmath548 , there is no cosmological solution because of eq .
( [ ro5:eq - h2:eq ] ) .
let us consider the case where @xmath106 the four - dimensional euclidean metric .
then we can construct the solution where @xmath384 , @xmath549 , and @xmath385 are the constants .
now we discuss the cosmological evolution for time - dependent solution ( [ ro5:solution : eq ] ) .
we define the cosmic time @xmath132 , which is given by the five - dimensional metric can be expressed as where the functions @xmath550 and @xmath551 are given by here @xmath396 is the number of smeared dimensions and should satisfy @xmath422 . here
we assume that one direction of @xmath423 is not smeared in order to fix the location of our universe in the transverse space .
our universe is given by the solutions with the five - dimensional coordinates @xmath552 .
the time direction is written by @xmath91 .
our choice is to take the three - dimensional from the overall transverse space with @xmath45 .
the four - dimensional universe is spanned by @xmath91 , @xmath425 , @xmath426 and @xmath427 , for instance .
the @xmath428 direction is preserved to measure the position of our universe in the overall transverse space of 0-branes .
since the metric depends on @xmath45 explicitly , we have to smear out @xmath429 , @xmath430 and @xmath431 so as to define our universe . then the number of the smeared directions @xmath396 should satisfy the condition @xmath432 .
unfortunately , the power exponent of four - dimensional universe becomes 1/4 .
hence , we have to conclude that in order to obtain a realistic expansion of the universe in this type of models , one has to include additional fields on the background .
we study the asymptotic behavior of the dynamical 0-brane background .
the time dependence in the function @xmath388 can be ignored in the limit of @xmath553 , because the harmonic function @xmath551 dominates near a position of 0-brane . in the limit of @xmath554 , as function @xmath551 vanishes , the system becomes static near 0-brane .
then , the function @xmath555 depends only on time in the far region from 0-branes .
the five - dimensional metric in the limit of @xmath554 , is thus given by the metric has singularity at @xmath556 .
then the five - dimensional spacetime does not have any singularity if it is restricted inside the domain satisfied by the conditions , where the function @xmath551 is defined in ( [ ro5:sme : eq ] ) .
the five - dimensional spacetime can not be extended beyond this region .
since the spacetime evolves into a curvature singularity , the regular spacetime with dynamical 0-branes ends up with the singularities .
although the evolution of dynamical 0-brane with @xmath557 has the time reversal one of @xmath558 , the behavior of the background spacetime strongly depends on the signature of @xmath559 . in the following
, we will focus on the case with @xmath558 . for @xmath264 ,
the function @xmath388 is positive everywhere .
then the spacetime is not singular . in the limit of @xmath449 ,
the solution becomes a time - dependent uniform spacetime apart from a position of 0-branes .
the five - dimensional background geometry can be described as a cylindrical form of infinite throat near the dynamical 0-branes , let us consider the time evolution the five - dimensional spacetime . at @xmath262
, the five - dimensional spacetime does not have any curvature singularity in the background .
the background geometry has a cylindrical topology near each 0-brane .
as time slightly increases , a curvature singularity appears far from 0-branes @xmath560 .
after that , the singular hypersurface cuts off more and more of the space as time increases further .
the singular hypersurface splits and surrounds each of the 0-brane throats individually after time continues to evolve . the spatial surface is finally composed of two isolated throats .
the time evolution of the five - dimensional spacetime for @xmath264 is the time reversal of @xmath259 .
we find that the overall transverse space tends to expand asymptotically like @xmath561 , for any values of fixed @xmath45 , in the regular domain of the five - dimensional metric ( [ ro5:surface : eq ] ) , while the solutions describe static 0-branes near the positions of the branes . in the far from 0-branes , where @xmath562 , the background geometry becomes frw universes with the power law expansion @xmath561 .
next we consider the case of the near - horizon limit that the spacetime metric and the functions @xmath388 , @xmath555 are given by ( [ ro5:solution : eq ] ) .
if we consider the case where all 0-branes are located at the origin of the z space , we have where @xmath414 , @xmath115 are the total mass of 0-branes since the dependence on @xmath91 in ( [ ro5:hk2:eq ] ) is negligible in the near - horizon limit @xmath273 , the five - dimensional metric is reduced to the following form where @xmath438 has been used .
the line elements of a two - dimensional ads space ( ads@xmath416 ) and a three - sphere with the unit radius ( @xmath563 ) are given by @xmath418 and @xmath441 , respectively .
thus we see that the near - horizon limit of the 0-brane system is a ads@xmath416 with a certain internal 3-space . before closing this subsection
, we discuss the collision of 0-brane .
there are two kinds of 0-brane in the five - dimensional spacetime .
one is static 0-brane coming from the function @xmath550 .
the other is dynamical 0-brane given by @xmath564 .
we set the two dynamical 0-branes at @xmath565 and @xmath566 , where @xmath567 is a constant . on the other hand
, we suppose that @xmath568 static 0-branes are sitting at a point , now we consider the following quantity then the proper length at @xmath535 between the two dynamical 0-branes is given by where @xmath343 and @xmath346 are the charges of the dynamical 0-brane and @xmath414 is defined by for @xmath569 , the length @xmath468 is a monotonically decreasing function of time .
since the time evolution of the proper length depends on the number of the smeared directions @xmath396 , we shall analyze it for each of the values of @xmath396 below .
first we consider the case with @xmath570 . for @xmath571 , the harmonic functions @xmath388 and @xmath555 diverges both at infinity and near 0-branes .
in particular , there is no regular spacetime region near 0-branes because of @xmath572 and @xmath573 .
then , these are not physically relevant .
hence , we show the proper length in fig . [
fig : ro0 ] for the cases with @xmath462 and @xmath464 . for both cases
the singularity between two dynamical 0-branes appears before collision because a singularity appears before the proper distance becomes zero .
although two dynamical 0-branes initially approach very slowly , the singular hypersurface suddenly appears at a finite distance , and the spacetime finally splits into two isolated 0-brane throats .
therefore we can not analyze the collision of the dynamical 0-branes in these examples .
( a ) and @xmath464 ( b ) in the five - dimensional romans theory . for both cases ,
the two dynamical 0-brane charges are identical , @xmath520 and the parameters are taken as @xmath574 , @xmath569 , @xmath509 , @xmath327 , @xmath575 , and @xmath576 .
the result is also the same and a singularity appears before collision of dynamical 0-branes.,title="fig : " ] ( -175,120)@xmath468 ( 10,10)@xmath91 2.0 cm ( a ) and @xmath464 ( b ) in the five - dimensional romans theory . for both cases ,
the two dynamical 0-brane charges are identical , @xmath520 and the parameters are taken as @xmath574 , @xmath569 , @xmath509 , @xmath327 , @xmath575 , and @xmath576 .
the result is also the same and a singularity appears before collision of dynamical 0-branes.,title="fig : " ] ( -175,120)@xmath468 ( 10,10)@xmath91 + ( a ) ( b ) however , for the case with @xmath432 , the function @xmath388 and @xmath577 are written by the linear function of @xmath45 .
if we assume that the @xmath45 directions apart from @xmath428 are smeared , the time evolution of the proper length is different from the previous case .
hence , the harmonic functions @xmath388 and @xmath577 are expressed as we study the dynamics of 0-branes , where one 0-brane charge @xmath343 is located at @xmath578 and the other @xmath346 at @xmath579 .
the proper distance between the two dynamical 0-branes is given by where we assume again that @xmath568 static 0-branes are sitting at a point @xmath580 , and @xmath414 is defined by ( [ ro5:l : eq ] ) .
for @xmath558 , the proper distance decreases with time . setting @xmath350 , a curvature singularity appears again at a certain finite time @xmath476 before the dynamical 0-branes collide .
then , @xmath477 is written by this is the same result as the case with @xmath581 . on the other hand , two 0-branes have the same brane charge @xmath527
, the proper distance vanishes at a certain finite time @xmath480 , where @xmath481 is defined by then two dynamical 0-branes collide completely .
if we set @xmath582 , for simplicity , the proper length between two dynamical 0-branes can be written by if we choose the physical parameters as @xmath574 , @xmath569 , @xmath509 , @xmath576 , @xmath582 , and @xmath327 , the proper distance @xmath468 is depicted in fig .
[ fig : ro ] for the two cases ( a ) the same 0-brane charges @xmath520 and ( b ) different charges @xmath525 , @xmath338 . for the case ( a )
, the two dynamical 0-branes can collide completely . on the other hand , in the case ( b ) a singularity appears before the collision of dynamical 0-branes , as we have already discussed in sec.[sec : ns - c ] .
( a ) and @xmath525 , @xmath338 ( b ) in the five - dimensional romans theory .
we fix @xmath489 , @xmath574 , @xmath569 , @xmath509 , @xmath582 , @xmath327 , and @xmath576 .
the proper length rapidly vanishes near two 0-branes collide for the case of @xmath520 . for the case of @xmath525 , @xmath338 , it is still finite when a curvature singularity appears.,title="fig : " ] ( -175,115)@xmath468 ( 10,10)@xmath91 2.0 cm ( a ) and @xmath525 , @xmath338 ( b ) in the five - dimensional romans theory .
we fix @xmath489 , @xmath574 , @xmath569 , @xmath509 , @xmath582 , @xmath327 , and @xmath576 .
the proper length rapidly vanishes near two 0-branes collide for the case of @xmath520 . for the case of @xmath525 , @xmath338 ,
it is still finite when a curvature singularity appears.,title="fig : " ] ( -175,115)@xmath468 ( 10,10)@xmath91 + ( a ) ( b )
in this section , we briefly discuss the nature of the singularities appearing in the time - dependent solutions and present the stability analysis for the dynamical brane background .
we follow the method used by @xcite ( see also @xcite ) and present the preliminary analysis performed , where the klein - gordon modes are analyzed . an analysis of such possibility will definitely make the property of singularity more clear , even if it is just a simple preliminary study to asses this issue .
let us first consider the stability for the 0-brane solution in nishino - salam - sezgin gauged supergravity .
the six - dimensional metric becomes static space near 0-brane while the background depends only on the times far from 0-brane .
we will study the stability in the 0-brane solution far from the branes .
for the limit @xmath583 in the solution ( [ nh : h2:eq ] ) , the six - dimensional metric is expressed as where @xmath584 denotes the metric of four - dimensional sphere .
the six - dimensional metric has curvature singularity at @xmath585 . in order to study the stability
, we consider the klein - gordon equation for a massive scalar field propagating in the background ( [ st : metric : eq ] ) : where @xmath586 denotes the determinant of the six - dimensional metric ( [ st : metric : eq ] ) . in terms of the metric ( [ st : metric : eq ] ) , the klein - gordon equation can be written by where @xmath587 denotes the laplace operator on the @xmath588 .
the six - dimensional metric involved permits separation of variables , so we take where the functions @xmath589 , @xmath590 obeys the eigenvalue equations here @xmath591 is the laplace operator on the w space , @xmath592 is eigenvalue and @xmath589 , @xmath590 satisfy @xcite where @xmath593 , @xmath594 are constants , and @xmath595 , @xmath596 denote the bessel functions and @xmath597 is related to the eigenvalue @xmath598 as the klein - gordon equation thus is rewritten by then the solution for @xmath599 is oscillatory , having the form where @xmath600 , @xmath601 are constants , @xmath602 , @xmath603 denote the bessel functions and @xmath62 is given by let us consider the energy of the klein - gordon modes to study whether the instability occurs or not . using the asymptotic solution ( [ st : solutions : eq ] ) , we will see that @xmath604 as the singularity is approached , where there is a curvature singularity at @xmath585 in the six - dimensional background ( [ st : metric : eq ] ) .
since the velocity is well - behaved besides the singularity , the energy of the klein - gordon modes can estimate as where @xmath376 is velocity . in terms of the normalization condition @xmath605 , the behavior of the @xmath606 and @xmath607
are determined in order to remain non - singular .
then , we find as @xmath608 , @xmath606 and @xmath607 have to behave in the limit @xmath583 for the dynamical 0-brane background . upon setting ( [ st : coe : eq ] )
, one then finds in terms of the asymptotic solution ( [ st : solutions : eq ] ) with @xmath62 ( [ st : gamma : eq ] ) , we find that @xmath604 as the singularity is approached if we set @xmath609 and @xmath369 .
hence the 0-brane solution implies that the energy momentum tensor of the scalar field mode diverges far from 0-brane .
however , it is necessary to study a full analysis of the metric perturbations whether the mode of klein gordon field is not likely to destabilize the metric modes near the singularity or not .
next we consider the stability for the 1-brane solution in nishino - salam - sezgin gauged supergravity .
for the metric ( [ nsc : h : eq ] ) , the harmonic function @xmath500 dominates in the limit of @xmath503 ( near a position of 1-branes ) and the time dependence can be ignored .
thus the background becomes static . on the other hand , in the limit of @xmath267
, @xmath500 vanishes .
then @xmath383 depends on @xmath91 and @xmath435 in the far region from 1-branes and the resulting metric is given by where @xmath610 is the metric of three - dimensional sphere .
for the six - dimensional metric , a curvature singularity may appear at in the following , we will again discuss the stability in the 1-brane solution far from the branes .
let us consider the klein - gordon equation for a massive scalar field in the six - dimensional background ( [ st1:metric : eq ] ) : where @xmath586 denotes the determinant of the six - dimensional metric ( [ st1:metric : eq ] ) . substituting the six - dimensional metric ( [ st1:metric : eq ] ) into the klein - gordon equation for a massive scalar field ( [ st1:kg : eq ] ) , we find where we set @xmath611 , and @xmath612 denotes the laplace operator on the @xmath613 .
the six - dimensional metric involved permits separation of variables , so we take where the functions @xmath614 , @xmath615 obeys the eigenvalue equations here @xmath616 is the laplace operator on the z space , @xmath617 is eigenvalue for the equation ( [ st1:eigen : eq ] ) , and @xmath614 , @xmath615 are satisfy where @xmath593 , @xmath594 are constants , @xmath595 , @xmath596 denote the bessel functions and @xmath597 is related to @xmath618 : hence , the klein - gordon equation is reduced to we shall discuss the massless cases in the following . in terms of @xmath574 and @xmath619 ,
the particular solutions of @xmath620 and @xmath621 are given , respectively , by where @xmath622 and @xmath623 are constants and @xmath624 is defined by we study the stability in terms of the energy of the klein - gordon modes . using the asymptotic solution given by eq .
( [ st1:solutions : eq ] ) , we can estimate the energy near the singularity , where there is a curvature singularity at @xmath625 in the six - dimensional background ( [ st1:metric : eq ] ) .
since the velocity is well defined besides the singularity , the energy of the klein - gordon modes can be written as where @xmath471 is velocity . by using the normalization condition @xmath626 , the behavior of the @xmath627 , @xmath628 , and @xmath629
are determined in order to remain nonsingular .
then , we find in the limit @xmath583 and @xmath608 , for the dynamical 1-brane background ( [ st1:metric : eq ] ) , @xmath627 , @xmath628 , and @xmath629 provided if we set the parameters ( [ st1:coe : eq ] ) , one then finds in terms of the asymptotic solution ( [ st1:solutions : eq ] ) , we find that @xmath604 as the singularity is approached at @xmath625 .
let us next consider the case @xmath619 and @xmath434 . in the limit @xmath583 for the solution ( [ nh : h3:eq ] ) , the six - dimensional metric becomes where we set @xmath611 and @xmath630 denotes the metric of the three - dimensional sphere .
there is a curvature singularity at @xmath631 .
now we study the behavior of the klein - gordon field .
the scalar field equation ( [ st1:kg : eq ] ) in the six - dimensional background ( [ st1:metric2:eq ] ) reads if we assume that the scalar field @xmath632 is given by ( [ st1:ansatz : eq ] ) , where @xmath614 and @xmath615 can be written by ( [ st1:eigen2:eq ] ) , the function @xmath633 is determined by the eigenvalue equation : here , @xmath634 is constant .
then , the equation for @xmath620 becomes for the massless case , the solution of @xmath620 is given by the oscillatory form where @xmath635 denotes the hypergeometric function , @xmath636 is the laguerre polynomial , @xmath637 and @xmath638 are constants , and @xmath639 is defined by we estimate the energy of the klein - gordon modes whether the instability exists or not . in terms of the asymptotic solution ( [ st1:solution : eq ] ) , we can present the behavior of the energy as the singularity is approached . since the velocity is well behaved except for the singularity , the energy of the klein - gordon modes is given by ( [ st1:e : eq ] ) . by using the normalization condition @xmath626 , @xmath627 , @xmath628 , and @xmath629 are determined by as @xmath608 and @xmath583 , the functions @xmath627 , @xmath628 , and @xmath629 are set to be if we use eq .
( [ st1:coe2:eq ] ) , the energy can be expressed as then , for the asymptotic solution ( [ st1:solution : eq ] ) , the energy becomes @xmath604 as the singularity is approached , that is , @xmath608 .
since the 1-brane solution gives that the energy - momentum tensor of the klein - gordon field mode diverges in this limit , the mode of the scalar field can not stabilize the metric modes near the singularity . in this subsection
, we analyze the stability of the dynamical 0-brane background in the five - dimensional romans gauged supergravity .
we will study the stability of the scalar field far from 0-branes . for @xmath583 in the dynamical 0-brane background ( [ ro5:hk2:eq ] )
, the five - dimensional metric becomes where we set @xmath640 and @xmath630 denotes the metric of the three - dimensional sphere .
there is a curvature singularity at @xmath556 .
let us consider the behavior of the klein - gordon field : where @xmath586 denotes the determinant of the five - dimensional metric ( [ r5:metric : eq ] ) .
the scalar field equation ( [ r5:kg : eq ] ) in the five - dimensional background ( [ r5:metric : eq ] ) reads here , @xmath612 denotes the laplace operator on the @xmath613 , and the scalar field @xmath641 is assumed to be where @xmath589 and @xmath590 are determined by the eigenvalue equation : here @xmath616 is the laplace operator on the z space , @xmath617 is the eigenvalue , and functions @xmath589 and @xmath590 obey @xcite where @xmath642 and @xmath643 are constants , @xmath595 and @xmath596 denote the bessel functions , and @xmath597 is related to the eigenvalue @xmath618 as by using eq .
( [ r5:e - z : eq ] ) , the equation for @xmath599 becomes for @xmath619 , the solution of @xmath599 is given by the oscillatory form where @xmath644 and @xmath645 are constants .
we calculate the energy of the klein - gordon modes to study the stability of the dynamical 0-brane background . by using the asymptotic solution ( [ r5:solution : eq ] )
, we can present the behavior of the energy as the singularity is approached . since it is possible to calculate the velocity except for the singularity , the energy of the klein - gordon modes is given by where @xmath376 denotes the velocity in the five - dimensional spacetime . in terms of the normalization condition @xmath605 , @xmath627 and @xmath629 are given by as @xmath646 and @xmath583 , the functions @xmath627 and @xmath629 are described as by using eq .
( [ r5:coe2:eq ] ) , the energy of the scalar field can be expressed as for the asymptotic solution ( [ r5:solution : eq ] ) , one can note that the energy becomes @xmath604 as the singularity is approached .
the dynamical 0-brane solution gives that the energy - momentum tensor of the scalar field mode diverges in the limit @xmath647 .
hence , the mode of the scalar field can not stabilize the metric modes near the singularity .
now we investigate the stability analysis for the dynamical @xmath303-brane background .
the geometry of the @xmath298-brane system becomes a static structure near branes , while the background geometry depends only on the time in the far region from branes . by setting @xmath648 in the @xmath6-dimensional background ( [ sa : exact : eq ] ) ,
the metric in the limit @xmath649 is thus given by where @xmath650 denotes the metric of the @xmath651-dimensional sphere and @xmath652 and @xmath653 are defined , respectively , by the @xmath6-dimensional spacetime has singularities at @xmath262 .
let us consider the klein - gordon equation to discuss the stability analysis where @xmath586 denotes the determinant of the @xmath6-dimensional metric ( [ sp : metric : eq ] ) .
equation ( [ sp : kg : eq ] ) on the @xmath6-dimensional background ( [ sp : metric : eq ] ) becomes where @xmath654 denotes the laplace operator on the @xmath655 .
we assume that the scalar field @xmath641 is given by where the functions @xmath589 and @xmath590 obey the eigenvalue equations here @xmath73 denotes the laplace operator on the @xmath80 space , and @xmath656 is the eigenvalue for the equation . the functions @xmath589 and @xmath590 also satisfy the equations @xcite where @xmath657 and @xmath658 are constants , @xmath659 and @xmath660 denote the bessel functions , and @xmath661 is related to the eigenvalue @xmath662 as by using eqs .
( [ sp : metric : eq ] ) , ( [ sp : ansatz : eq ] ) , and ( [ sp : eigen : eq ] ) , the field equation for @xmath599 becomes let us consider the case of @xmath619 .
the solution of @xmath599 is given by the oscillating form where @xmath663 and @xmath664 are constants and @xmath665 and @xmath666 are the bessel functions .
the energy of the klein - gordon modes can be calculated by where @xmath376 is velocity .
then , @xmath606 and @xmath607 are determined by where we used the normalization condition @xmath605 . in the case of @xmath608 and @xmath583 , @xmath606 and @xmath607
must behave as in order to remain nonsingular .
if we use the expression ( [ sp : coe : eq ] ) , the energy of the scalar field is given by for the asymptotic solution ( [ sp : solution : eq ] ) , one can note that the energy becomes @xmath604 as the singularity is approached .
hence , the energy - momentum tensor of the klein - gordon field mode diverges .
the classical solution gives the mode of the scalar field which can not stabilize the metric modes near the singularity .
finally , we discuss the stability analysis for the @xmath298-brane in the asymptotically de sitter universe . if we set @xmath667 and take @xmath649 in the background ( [ ds2:exact : eq ] ) , the @xmath6-dimensional metric becomes where @xmath650 denotes the metric of the @xmath651-dimensional sphere , @xmath108 is given by ( [ ds2:c0:eq ] ) , and the cosmic time @xmath132 is defined by ( [ ds2:cosmic : eq ] ) .
there is a curvature singularity at @xmath668 in the @xmath6-dimensional spacetime . in the following , we set @xmath146
otherwise , the scale factor of @xmath6-dimensional spacetime becomes complex or negative .
we consider the klein - gordon field to analyze the stability where @xmath586 is the determinant of the six - dimensional metric ( [ sd : metric : eq ] ) .
substituting the @xmath6-dimensional metric ( [ sd : metric : eq ] ) into eq .
( [ sd : kg : eq ] ) , we obtain where @xmath654 denotes the laplace operator on the @xmath655 .
we assume an ansatz for the scalar field @xmath641 : where the functions @xmath589 and @xmath590 satisfy the eigenvalue equation and obey the equations here @xmath616 is the laplace operator on the z space , and @xmath669 and @xmath670 denote constants , @xmath671 and @xmath672 are the bessel functions , and @xmath673 is written by the eigenvalue @xmath662 as in terms of eqs .
( [ sp : eigen : eq ] ) , ( [ sd : metric : eq ] ) , and ( [ sd : ansatz : eq ] ) , the field equation for @xmath599 becomes let us first consider the solution of @xmath599 for @xmath674 and @xmath364 . in the case of @xmath674 , the solution of @xmath599 can be expressed as where @xmath675 and @xmath676 are constants , @xmath677 and @xmath678 are the bessel functions , and @xmath679 is defined by on the other hand , setting @xmath364 , we can also find the solution of @xmath599 : here @xmath680 and @xmath681 denote constants , @xmath682 and @xmath683 are the bessel functions , and the constant @xmath684 is given by for @xmath685 , the solution of @xmath599 can be written in the following form : where @xmath686 and @xmath687 are constants , @xmath688 and @xmath689 denote the bessel functions , and @xmath113 is defined by since the energy of the scalar field can be written as ( [ sp : e : eq ] ) , the energy of the klein - gordon modes can be given by the expression ( [ sp : e2:eq ] ) .
then , we can find @xmath606 and @xmath607 in this way : where we used the normalization condition @xmath605 . in the case of @xmath690 and @xmath583 , @xmath606 and @xmath607 have to behave as for the solution ( [ sd : solution - g : eq ] ) , the energy becomes @xmath604 in the limit @xmath690 .
since the energy is not convergent with the asymptotic solution , the mode of the scalar field does not stabilize the metric modes near the singularity .
in this paper , we have discussed the time - dependent intersecting branes with cosmological constants for not only the delocalized case but also the partially localized one in @xmath6-dimensional gravitational theory .
we are everywhere brief , and on some points , we simply call attention to questions that might be investigated in the future .
the function @xmath691 depends on time as well as the coordinate of the relative and overall transverse spaces .
the coupling constants between the field strengths and the dilaton are given by the assumptions ( [ pl : ansatz : eq ] ) or ( [ cn : fields : eq ] ) and depend on the parameter @xmath3 .
in the case of the eleven- or ten - dimensional supergravity theory , the dilaton coupling requires @xmath4 .
the power of the time dependence depends on the number of the brane and total dimensions with the parameter @xmath3 of the dilaton coupling constant .
an exceptional case arises if the parameter @xmath3 in the dilaton coupling takes another value than 4 .
there are static solutions with @xmath5 in the lower - dimensional supergravity theories as well as einstein - maxwell theory @xcite . in this case one
gets asymptotically power - law expanding solutions if the dilaton is nontrivial . for the trivial dilaton ,
the einstein equations give an asymptotically de sitter solution for a single 2-form field strength .
since the cosmological constant is related to the field strength , the time derivative of the warp factor arises only from the ricci tensor and can be compensated by the cosmological constant in the einstein equations .
this is the same structure as in refs .
@xcite and the generalization of the solutions @xcite . in @xmath4 case ,
the equation of motion in the presence of the cosmological constant gives the static delocalized or partially localized intersecting brane solution because of the ansatz of the fields .
thus , one expects that the recipe for picking an accelerating expansion from the dynamical intersecting brane solutions depends on the dilaton coupling constant , and this is the case for the proposal in ref .
once the de sitter solution is obtained in the single @xmath0-brane solutions , it is possible to apply it to the intersecting brane systems .
an immediate point is that the time - dependent solutions make dynamical compactification more or less obvious , since cosmological evolution is a general property of the solution ( with constant parameters ) once the function @xmath190 is properly endowed with the time dependence .
the power of the scale factor in some solutions gives an accelerating expansion law even in the case that functions @xmath190 depend on both the time and coordinates of overall transverse space , while the extra dimension will shrink as cosmic time increases .
however , something is still missing , because the scale factor of our universe diverges at @xmath225 . at the moment , it is not clear how to do this . we have discussed the dynamics of the brane collisions . as the spacetime is contracting in the @xmath6-dimensional spacetime , each 0-brane approaches others as the time evolves for @xmath336 but separates for @xmath315 in the asymptotically power - law expanding solutions .
thus 0-branes never collide . in the case of asymptotically
de sitter solutions , all domains between branes are connected at @xmath331 ( @xmath349 ) .
the domain shrinks as the time decreases , while the proper distance becomes constant as @xmath132 increases . for the @xmath298-brane system ( @xmath10 ) ,
a singularity appears before 0-branes collide , and eventually the topology of the spacetime changes so that branes are separated by singular hypersurfaces surrounding each brane if branes are not smeared .
thus , we can not describe the collision of two 0-branes in terms of these solutions . on the other hand
, the @xmath334-brane system in ten dimensions or the smeared @xmath2-brane system in @xmath6-dimensional theory can provide examples of colliding branes if they have the same brane charges and only one overall transverse space .
we have analyzed the collision of the brane where the @xmath692-branes are localized at the same position along the overall transverse directions , in the case of equal charges .
the brane collision would not occur if the brane charges are different .
moreover , if these branes are localized at different positions , it raises the possibility that the curvature singularities appear .
we have also studied the dynamics of the five- or six - dimensional supergravity models with applications to cosmology and collision of branes .
first we have discussed the brane solutions to study the time evolution in the nss model . in the case of vanishing 3-form field strength in the five - dimensional effective theory ,
the scale factor of our four - dimensional spacetime is a linear function of the cosmic time which is the same evolution as the milne universe . on the other hand , for the dynamical
1-brane without 2-form field strength , the solution tells us that the function @xmath188 depends on all the world - volume coordinates of the 1-brane .
hence , the contribution of the field strength except for the 2-form leads to an inhomogeneous universe .
we have investigated the dynamics of 0-branes and found that , when the spacetime is contracting in six dimensions , each 0-brane approaches the others as the time evolves .
all domains between branes connected initially ( @xmath262 ) , but it shrinks as @xmath91 increases . however , for the 0-brane system without smearing branes , a singularity appears before 0-branes collide , and eventually the topology of the spacetime changes such that parts of the branes are separated by a singular region surrounding each brane .
thus , the solution can not describe the collision of two 0-branes .
in contrast , the smeared 0-brane system with @xmath484 can provide an example of colliding 0-branes and collision of the universes , if they have the same brane charges .
we have next constructed the time - dependent 1-brane solution in the nss supergravty model . in the asymptotic far 1-brane region ,
the 1-brane spacetime in the nss model approaches the six - dimensional milne universe . in regions close to the 1-branes , for concreteness
, we have studied the case of two 1-branes in detail .
the 1-brane is approaching the other as the time progress for @xmath264 .
we have found that , in the case of @xmath264 , all of the domains between the 1-branes are initially connected , but some region ( near small @xmath435 ) shrinks as the time increases , and eventually the topology of the spacetime changes such that parts of the branes are separated by a singular region surrounding each 1-brane .
thus , in the case of @xmath532 , 1-branes never collide . on the other hand , the case of @xmath489 , for @xmath264 , could provide an example of colliding 1-branes .
we found that the collision time depends on both brane charges and the place in the world volume of the 1-brane . since this case has the time - reversal symmetry , the evolution for @xmath259 is obtained by the time - reversal transformation .
we also investigated the time - dependent solution in the five - dimensional supergravity model .
the power of the scale factor is so small that the solutions can not give a realistic expansion law .
then , it is necessary to include additional matter on the background in order to obtain a realistic expanding universe .
we finally analyzed the classical instability of the dynamical brane background towards singularity . in order to present the instability of the dynamical brane background
, we have estimated whether an instability does exist by computing the energy of the klein - gordon modes .
one can find that the energy seen by an observer diverges as the curvature singularity is approached .
this implies that the mode of the scalar field is likely to destabilize the background metric modes near the singularity .
although this result has been given by preliminary analysis , it has made the property of singularity in the dynamical brane background more clear .
it is also necessary for us to perform a more rigorous analysis by considering in detail the metric perturbation whether the stability analysis arrives to the same conclusion or not .
a recent study of intersecting systems depending on the time coordinate and overall transverse space shows that all warp factors in the solutions can depend on time @xcite .
it is interesting to study if similar more general solutions can be obtained by relaxing some of our assumptions .
we hope to report on this subject in the near future elsewhere .
we thank m. minamitsuji for numerous valuable discussions and careful reading of the manuscript and also thank m. cvetic and g.w .
gibbons for discussions and valuable comments . | we discuss the dynamics of intersecting @xmath0-branes with cosmological constants in the higher - dimensional gravity theories . for the delocalized brane case , these solutions describe an asymptotically de sitter or power - law expanding universe , while for the partially localized intersecting branes , they describe homogeneous and isotropic universes at each position of the overall transverse space .
we then apply these time - dependent branes to the study on the collision of two 0-branes and show that the 0@xmath18-brane system or the smeared @xmath2-brane system can provide an example of colliding branes if they have the same brane charges and only one overall transverse space .
finally , we argue some applications of the solutions in supergravity models . |
three - dimensional fluid dynamic simulations of turbulent convection in an oxygen - burning shell of a presupernova star show bursty fluctuations which are not seen in one - dimensional stellar evolutionary calculations ( which use various versions of mixing - length theory , mlt , @xcite ) .
this paper explores the underlying physics of this new phenomena . since the formulation of mlt @xcite ,
there have been a number of significant developments in the theoretical understanding of turbulent convective flow .
first , @xcite and @xcite developed the modern version of the turbulent cascade , and published in journals easily accessible in the west ; the original theory @xcite was not used in mlt although it pre - dated it . this explicit expression for dissipation of turbulent velocities ,
@xmath1 where @xmath2 is the root - mean - square of the turbulent velocity and @xmath3 is the dissipation length .
it is found both experimentally and numerically that @xmath4 , where @xmath5 is the depth of the convective zone .
simulations for low - mach number flow show that the average of this dissipation over the convective zone closely compensates for the corresponding average of the buoyant power @xcite .
this additional constraint allows an alternative to present practice : fixing the free parameter ( e.g. , the mixing length factor @xmath6 ) directly by terrestrial experiments and numerical simulations which deal with the process of turbulence itself @xcite , instead of calibrating it from complex astronomical systems ( stellar atmospheres ) as is now done .
second , there has been a considerable development in understanding the nature of chaotic behavior in nonlinear systems ; see @xcite for a review and reprints of original papers , and @xcite .
@xcite presented a simplified solution to the rayleigh problem of thermal convection @xcite which captured the seed of chaos in the lorenz attractor , and contains a representation of the fluctuating aspect of turbulence not present in mlt .
this advance was allowed by the steady increase in computer power and accessibility , which lead to the exploration of solutions for simple systems of nonlinear differential equations ( see @xcite and references therein ) .
it became clear that the landau picture @xcite of the approach to turbulence was incorrect both theoretically @xcite and experimentally @xcite .
a striking feature of these advances has been the use of simple mathematical models , which capture the essence of chaos in a model with much reduced dimensionality compared to the physical system of interest .
third , it has become possible to simulate turbulence on computers .
this realizes the vision of john von neumann @xcite , in which numerical solutions of the navier - stokes equations by computer are used to inform mathematical analysis of turbulence . in this paper
we will follow this idea of von neumann , in the style which proved successful for chaos studies : building simple mathematical models of a more complex physical system ( in this case , the numerical simulations of turbulent convection ) .
this approach should lead to algorithms suitable for implementation into stellar evolution codes , which , unlike mlt , are ( 1 ) based upon solutions to fluid dynamics equations , ( 2 ) non - local , ( 3 ) time - dependent , and ( 4 ) falsifiable by terrestrial experiment and future improved simulations .
our particular example is a set of simulations of oxygen burning in a shell of a star of @xmath7 @xcite .
this is of astronomical interest in its own right as a model for a supernova progenitor , but also happens to represent a relatively simple and computationally efficient case , and has general implications for the convection process in all stars .
three - dimensional hydrodynamic simulations of shell oxygen burning exhibit bursty , recurrent fluctuations in turbulent kinetic energy ( @xcite and below ) .
the reason for this behavior has not been explained theoretically .
these simulations show a damping , and eventual cessation , of turbulent motion if we artificially turn off the nuclear burning @xcite .
further investigation @xcite shows that nearly identical pulsations are obtained with a volumetric energy generation rate which is constant in time , so that _ the cause of the pulsation is independent of any temperature or composition dependence in the oxygen burning rate .
_ localized heating is necessary to drive the convection ; even with this time - independent rate of heating , pulses in the turbulent kinetic energy still occur . such behavior is fundamentally different from traditional nuclear - energized pulsations dealt with in the literature ( e.g. ,
the @xmath8-mechanism , @xcite ) , and is a consequence of time - dependent turbulent convection ( it might be called a `` @xmath0-mechanism '' , with @xmath0 standing for turbulence ) .
it appears to be relevant to all stellar convection .
@xcite found , in a very different context , that non - linear interaction of the largest modes excited pulsations of a red - giant envelope - mechanism , which depends upon variations in opacity , is not required to drive such pulsations .
] , which is another example of the @xmath0-mechanism . in section 2
we examine the the physics context of the turbulence , including implications of subgrid and turbulent dissipation for the implicit large eddy simulations ( iles ) upon which our analysis is based , and the effect of the convective mach number on the nature of the flow . in section 3 we review the 3d numerical results of shell oxygen burning which are relevant to the theory . in section 4
we present the results of the classical lorenz model @xcite for conditions similar to those in section 3 . in section 5
we consider implications of turbulent intermittency on stellar variability , and provide a model light curve from this effect alone .
section 6 summarizes the results .
the appendix gives a short derivation of the lorenz model .
in this section we summarize concepts which are needed for the interpretation of later results .
approximation of partial differential equations by discrete methods inevitably leads to a loss of information at scales smaller than the grid size .
a single element in space is approximated as a homogeneous entity ; this is equivalent to complete mixing at this scale , at each time step , of mass , momentum , and energy .
the loss of information that occurs with this mixing corresponds to an increase in entropy @xcite , the mixing of momentum is equivalent to the action of viscosity , and the mixing of internal energy corresponds to the transport of heat ( @xcite , 15 and 49 ) . in 3d flow
, turbulent energy will cascade from large scales to small , at a rate set by the largest scales @xcite . at sufficiently small scales ,
microscopic processes homogenize the flow and dissipate the kinetic energy .
thus , there is a deep connection between the turbulent cascade and sub - grid scale mixing .
@xcite have demonstrated that the piece - wise parabolic method ( ppm , which we use ) , based on the euler equation ( which has no explicit viscosity ) , converges to the same limit as methods based on navier - stokes equation ( which do have explicit viscosity ) , as the grid is refined to smaller zones and smaller effective viscosity ( the relevant limit for astrophysics ) .
the subgrid scale dissipation for monotonicity preserving hydrodynamic algorithms @xcite , which is implicit in these methods , automatically gives a reasonable treatment of the turbulent cascade down to the grid scale .
we use this implicit sub - grid dissipation in our large eddy simulation ( iles ) ; this is the most computationally efficient way to deal with turbulent systems with a large range of scales .
the largest scales , which set the rate of cascade and contain most of the energy are explicitly calculated , while the sub - grid scales are dissipated in a way consistent with the kolmogorov cascade .
@xcite have presented a refinement of the ppm algorithm which has improved behavior at the smaller resolvable scales .
we have not yet implemented this modification .
the theoretical approach used here involves integrated properties of convective cells ; we find by direct resolution studies that these properties are well estimated even with surprisingly modest resolution , because they are determined primarily by the largest scales in the convective region .
it appears that the iles simulations are adequate for the present analysis .
@xcite have shown that the numerical damping at sub - grid scales in our iles simulations is quantitatively consistent with the introduction _ analytically _ of the kolmogorov cascade into the theoretical discussion .
the turbulent velocity field was found to be dominated by two components : 1 . a non - isotropic flow of the largest scale modes in the convection zone , which is coupled to the fluctuations .
this has aspects of a `` coherent structure '' @xcite .
the largest scales are unstable toward break - up , but are least affected by dissipation , and in this sense the most laminar .
2 . a more isotropic , homogeneous turbulent flow which carries the kinetic energy via the turbulent cascade to scales small enough for dissipation to occur @xcite . because of the vast size difference , the small scales are weakly coupled to the largest scales , which determine the rate at which energy flows through the cascade .
if we approximate the non - isotropic component of the flow ( the largest scale of convection ) with that described by the lorenz model , this interpretation captures the oxygen - burning fluctuations .
there are two limiting cases for convective flow , depending upon the convective mach number @xmath9 ( the ratio of the fluid speed to the local sound speed ) ; these are usually termed the incompressible " ( @xmath10 ) and compressible " ( @xmath11 ) regimes .
stars are stratified in density , so that the notion of incompressiblity " is misleading .
we will use the term low - mach number flow " in place of
incompressible " when we mean flows in which acoustic radiation is small , but may be compressed due to stratification . for turbulent motion , the pressure perturbation @xmath12 is related to the convective mach number by @xmath13 .
sound waves outstrip fluid motion , so that pressure differences quickly become small , except possibly for a static background stratification .
most of the historical research on convection ( e.g. , the bnard problem , @xcite ) is done in this limit . using the reynolds decomposition , @xmath14 , with horizontal averaging @xmath15 and @xmath16 , mass conservation for a steady state flow
can be written @xmath17 the navier - stokes equation , using mass conservation , is @xmath18 where @xmath19 is the reynolds stress tensor .
mass conservation implies a convenient identity involving total co - moving derivatives , @xmath20 taking the dot product of the velocity vector @xmath21 with the navier - stokes equation , gives a kinetic energy equation , @xmath22 if on average the system is in a steady state , the time derivative must integrate to zero over the convective region , and the mass conservation law implies that the total buoyancy power term is zero , @xmath23 ( assuming constant @xmath24 on horizontal averaging ) , and therefore does not contribute to the production of kinetic energy anywhere in the flow .
the only other term , which remains to balance the viscous dissipation of the kinetic energy , is the pressure term @xmath25 which may be rewritten as @xmath26 the divergence term vanishes upon integration over the volume . using hydrostatic equilibrium for the background state , @xmath27 , and mass conservation , @xmath28 @xmath29 when the mach number is small , the second term the right hand expression is nearly zero because @xmath30 and the turbulent pressure is negligible . in this limit the kinetic energy production is best understood as due to the remaining buoyancy power term @xmath31 , which is directly related to the enthalpy flux @xcite . when the mach number is no longer small , the second term on the rhs of eq .
[ pfluct ] increases in importance : both the divergence of the fluctuating velocity field and the pressure perturbation begin to play a role .
the velocity field changes character ; it is no longer dominated by rotational flow , but develops an irrotational component ( @xmath32 . the flow becomes diverging ( consider the extreme limit of a point explosion which is pure divergence ) . also , the ram
pressure " ( a tensor @xmath33 ) is not negligible and must be included in the momentum equation ( hydrostatic equilibrium " ) .
sound wave generation increases rapidly as @xmath34 ( @xcite , 75 ) .
the compressible limit is @xmath35 .
shock formation is the most startling change in the flow character . which of the @xmath9 limits is relevant for astrophysics ?
almost all the matter in stellar convection zones , during almost all evolution , is in the limit of low - mach number flow , as are our turbulence simulations .
the exceptions are important : ( 1 ) explosions , such as supernovae and novae , ( 2 ) vigorous thermonuclear flashes , ( 3 ) vigorous pulsations , especially radial ones , and ( 4 ) the sub - photospheric layers of stellar surface convection zones , which are strongly non - adiabatic , to name a few .
and its rayleigh number @xmath36 . turbulent velocity squared ( solid ) and buoyancy power to the 2/3 power ( dotted ) are plotted versus time . the axes where chosen so that the same number of peaks would be shown , and the average kinetic energy be the same .
the curves are surprisingly similar , even though ( 1 ) the lorenz model used only a single mode , ( 2 ) @xmath37 and @xmath36 were not adjusted to give a fit , ( 3 ) the lorenz model imposes thermal balance while thermal imbalance adds to the background slope in the simulation ( see fig . [ fig2 ] ) , and ( 4 ) the simulations have @xmath38 more degrees of freedom than the lorenz model . ] figure [ fig1 ] illustrates the behavior of two important integral quantities , the total turbulent kinetic energy and the total buoyancy power , in the oxygen - burning shell simulations for a 23 @xmath39 star @xcite .
the flow has a low - mach number ( @xmath40 ) , although the numerical simulations use the equations for fully compressible fluid flow , and would have correctly treated high - mach number flows .
also shown are the same quantities for the lorenz model , which is discussed in section [ lor ] . at any instant in time ,
the total convective kinetic energy is @xmath41 , where @xmath42 is the mass of the convective zone and @xmath43 is the rms velocity is the solenoidal velocity in the convective region , with overall translational velocity removed , so that @xmath43 is its root - mean - square value .
] , while the kinetic energy in the isotropic part of the turbulent flow field is @xmath44 , where @xmath45 is the turbulent velocity , which we define as the isotropic part of the turbulent flow , where @xmath46 and @xmath47 are the velocity components perpendicular to the direction of the gravity vector , i.e. , the tangential velocities .
the vertical velocity also contains a significant contribution ( @xmath48 ) from the non - isotropic flow of the largest eddies .
a more careful discussion of the velocities in terms of principal orthogonal decomposition is in preparation . ] .
the reader is warned that division of the flow into `` turbulent '' and large scale " flow is useful but an oversimplification , so that the exact relative values of these two kinetic energies depend upon the algorithm used in their definition , but is of order unity @xcite .
consequently the precise distinction does not change the qualitative picture .
the buoyancy power is the rate at which kinetic energy per unit mass is increased by buoyant acceleration .
if it is integrated over the space containing the convection zone , we have @xmath49 , where @xmath50 is the buoyant acceleration times the turbulent velocity ; @xmath51 has units of velocity cubed . in figure [ fig1 ]
the simulations show a phase lag of about 20 seconds between the peaks in buoyancy term ( @xmath52 ) and turbulent kinetic energy .
this is about half the time it takes the flow to transit a distance @xmath5 , the depth of the convection zone .
it also corresponds to an e - folding time for turbulent kinetic energy decay due to kolmogorov damping , where @xmath53 .
power spectra for both variables peak at 89 seconds ; an average transit time is @xmath54 .
figure [ fig1 ] shows multimodal behavior in the 3d simulations ; preliminary results from a quantitative analysis @xcite using principal orthogonal decomposition ( pod ) indicates that a single dominant mode has about 43% of the kinetic energy , the first five modes have 75% , and 90% is reached with the eleventh mode . there is a strong dominant mode but also significant energy in several other modes
; the modes interact in a nonlinear and dynamic way .
the buoyancy power is a large scale feature and is strongly anisotropic ( plumes move vertically ) .
the dissipation implied by the turbulence occurs at the kolmogorov scale ( which is tiny ) ; this dissipation is wide - spread in space @xcite , including the entire turbulent region on average , and bounded by the stably stratified layers . because buoyancy and dissipation occur at vastly different length scales , they are weakly coupled . in low mach number flow , on average over time ,
the buoyant driving must balance the turbulent damping for a quasi - steady state to exist ( see @xcite for a detailed discussion , especially their eq .
the time averaged viscous dissipation is @xmath55 where @xmath3 is the dissipation length , @xmath5 is the depth of the convection zone , and the over - lines indicate an average over time ( two turnover times or four transit times , about 200 seconds , in the analysis ) . from equation [ eturb ]
it is clear that the turbulent dissipation is of third order in the turbulent velocity , while the buoyancy power , @xmath50 , is second order ( because @xmath56 is first order ) .
this means that in the turbulent regime there is a unique turbulent velocity which satisfies the condition that buoyancy power balance turbulent damping .
this has the nature of an eigenvalue problem .
the assumption of a turbulent cascade implies finite turbulent velocities , with macroscopic structure .
equation [ eturb ] applies as an integral over a region much larger than the kolmogorov scale ; at the kolmogorov scale microscopic dissipation occurs by the navier - stokes term ( @xcite , 15 ) .
if we dot this navier - stokes term with the velocity to generate a rate of kinetic energy destruction by viscosity ( see equation [ ke - eq ] ) , we have @xmath57 which does scale as the velocity to second order . when integrated over the turbulent cascade , equation [ enavier ] becomes equation [ eturb ] .
the extra factor of velocity comes from the rate at which turbulent kinetic energy is delivered to the kolmogorov scale .
turbulence is an inherently nonlinear process , which sets up its own scales and structures .
care must be taken in analyzing it with techniques involving expansion in series .
the dissipation is seen to be driven by the isotropic part of the flow @xmath45 , which accounts for about three - fourths of the kinetic energy according to estimates by @xcite ( which are roughly similar to those from pod analysis by @xcite ) .
note that @xmath58 which gives the vertical offset between the solid and dotted curves in figure [ fig1 ] .
consider the average conditions from 200 to 800 seconds in the simulations .
the average level of buoyancy ( actually @xmath52 is plotted ) is higher than that of kinetic energy in isotropic turbulence . if we equate ( 1 ) the power generated by buoyancy , to ( 2 ) the dissipation due to turbulence , both averaged over a few transit times , we find @xmath59 thus , in the @xcite simulations , _ the dissipation length is found to be essentially the depth of the convective zone _ , consistent with kolmogorov theory .
these results are far more general than the particular stellar situation we have discussed . while the neutrino cooling may seem exotic to some stellar evolutionists , in fact it behaves somewhat like the more familiar cooling by radiative diffusion , and has no strange effect on the turbulence .
we note that the original simulations @xcite , which included core hydrogen burning cooled by radiative diffusion as well as oxygen shell burning cooled by neutrinos , explicitly showed this similarity . in the long term
, the thermal state of the convection zone is supposed to evolve toward a global thermal balance between total heating by nuclear burning and total cooling by neutrino emission @xcite .
this is illustrated in figure [ fig2 ] , which shows time average luminosities of heating and of cooling , as a function of the entropy of the convection zone . at lower entropy , as in the 3d simulation
, the heating is dominant , causing the entropy to increase .
this is accomplished primarily by expansion , with a small increase in temperature .
the decrease in density causes a bigger change in nuclear burning than in neutrino cooling , so that a thermal balance would be attained ( shown as a vertical line ) when the heating and cooling curves cross .
the gradual rise in turbulent kinetic energy in the simulation , shown in figure [ fig1 ] , occurs in a shell which is below the entropy of balanced heating and cooling , so that heating dominates .
the pulses are much faster than this secular evolution , which changes on a time scale of @xmath60 seconds .
the lorenz model is a convective roll , or cell , whose dynamics are described by three amplitude equations .
this is a simple example of an elegant method of reduction of turbulent flow to a low - order set of dynamical equations using amplitudes of a proper orthogonal decomposition ( pod ) of numerical simulations or extensive experimental data sets @xcite .
the lorenz model is a better mathematical representation of the dynamics of a convective cell than mlt in that the acceleration and deceleration over a convective cycle are integrated to determine the motion , rather than prescribed .
because it uses a single mode , the lorenz model does not have sensitivity to local variation ; it is a _
global _ model , unlike mlt , which is local .
lorenz devised the model as a point of principle test of meteorological convection , which like the stellar problem , is damped by a turbulent cascade . to make use of this extensive literature @xcite
, we use the original version of @xcite , with the same prandtl and rayleigh numbers , to explore the implications on dynamics of the strange attractor .
the original lorenz formulation is a 2d , low mach number , and single mode model ( equivalent to figure [ fig3 ] ) . a transition to 3d
may be made using the solutions of @xcite , but is not necessary for present purposes .
real convection @xcite is expected to be single mode only near the onset of convective instability .
( -12,-2)(4,12 ) ( -5,5)5 ( -5,5)3.6 ( 2,2)(2,5 ) ( 2,1.5)*@xmath61 * ( 1,5)(1,7 ) ( 1,7.5)*@xmath62 * ( -11,5)*@xmath63 * ( -8,5.5)*@xmath64 * ( -8.2,5.5)(-9.3,5.5)(-9.3,5.0 ) ( -4.5,5.5)*@xmath65 * ( -1.7,5.5)(-0.8,5.5)(-0.8,5.0 ) ( -5,5)3.6270315 ( -4.6,3.3)*@xmath66 * ( -5,5)0.8270315 ( -3.1,4.0)*@xmath67 * ( -6,10.8)*@xmath68 * ( -6,9.3)*@xmath69 * ( -6,-0.5)*@xmath70 * ( -6,0.8)*@xmath71 * ( -2,1.7)*@xmath72 * ( -1.6,2)(-1,2.8)(-0.5,4 ) ( -10,5)(2,5 ) the lorenz equations ( see appendix a ) are : @xmath73 in the classical lorenz model , @xmath74 is proportional to the speed of convective motion , @xmath75 is proportional to the temperature difference between ascending and descending flow , and @xmath76 is proportional to the distortion of the vertical temperature profile from adiabatic @xcite . here
@xmath0 is a time in thermal diffusion units , @xmath37 is the effective prandtl number , @xmath36 is the ratio of the rayleigh number to its critical value for onset of convection , and @xmath77 is the ratio of the depth to the width of the convective cell , so that @xmath78 ; see @xcite .
the prandtl number is the ratio of coefficients of the viscous dissipation term to the thermal mixing term .
lorenz took @xmath79 , and chose the most unstable mode so that @xmath80 , and @xmath81 .
for this mode , and a rayleigh number of @xmath82 times the critical value , steady flow becomes unstable .
figure [ fig3 ] illustrates the structure and notation in the lorenz model .
the prandtl number may be expressed as @xmath83 , the ratio of the radiative cooling time scale to the viscous time scale .
the viscous damping time is taken to be constant , @xmath84 .
the rayleigh number , in units of its value at the onset of laminar convection , is @xmath85 where @xmath86 , @xmath87 , and @xmath61 is the gravitational acceleration .
the lapse rate of the adiabatic background @xmath88 is @xmath89 , so @xmath90 . in astrophysical notation ,
@xmath91 where @xmath92 is the temperature variation in the vertical direction ( fig .
[ fig3 ] ) .
the depth of the roll , in pressure scale heights , is @xmath93.\ ] ] in the meteorological case the fluid motion ( wind ) is of primary interest , however the energy fluxes is strictly zero , and in this case only heat content can carry energy , so heat flux is relevant .
however , we may interpret the equations in terms of a stratified system with low - mach number flow , having a stratification in density as well as temperature .
then @xmath94 is not zero , @xmath95 work is done , and the relevant flux is the enthalpy flux . ]
provide a means to connect with stellar observations . the vertical enthalpy flux due to radiative diffusion of the background is @xmath96 which is constant in the lorenz model .
the thermal conduction coefficient is @xmath97 and @xmath98 .
here @xmath99 , the radius of the roll being @xmath100 . because the background temperature is taken to be linear in @xmath62 , the divergence of this flux is identically zero , and therefore gives no local heating or cooling .
the additional vertical enthalpy flux , due to the temperature perturbation , is @xmath101 if we define a potential temperature @xmath102 , the vertical flux is separated into two parts , so that the adiabatic background value is denoted by @xmath103 and the changes caused by motion are contained in @xmath104 .
the flux in the horizontal direction is @xmath105 which averages to zero by symmetry ; here @xmath106 .
both @xmath107 and @xmath108 are proportional to a potential temperature which varies in time
. the net vertical enthalpy flux will not be constant in height @xmath62 , so that its divergence is nonzero .
this implies local heating / cooling , which would have to be compensated for in a steady state , and is a consequence of considering only a single mode .
smaller scale modes would be needed to deal with the local heating / cooling ; see @xcite for efforts to include a full spectrum of modes . to get the average flux through the cell for a single mode , we take the double projection , of @xmath72 and of @xmath109 on the vertical direction , @xmath110 the enthalpy flux is zero at the top and bottom of the roll ( @xmath111 and @xmath112 ) , and a maximum at the midpoint ( @xmath113 ) .
the average value of @xmath114 over the roll , @xmath115 to @xmath116 , is @xmath117 .
this is to be compared with the @xmath118 of @xcite , which was the range @xmath119 to @xmath120 for the simulations then available .
because the simulations are multi - mode , this slightly higher value for the correlation seems natural .
integrating over the cell , @xmath121 , gives @xmath122 which is the vertically averaged enthalpy flux of the cell .
this is the convective flux " in stellar models .
the up - flow enthalpy flux equals the down - flow value , and both are positive ( hot up - flows and cold down - flows both transport energy upward ) .
the ratio of enthalpy flux to the total radiative flux ( @xmath123 ) is @xmath124 which is a function of time .
the kinetic flux exactly cancels in this formulation , as in mlt , with the up - flow being the negative of the down - flow .
this is _ not _ true in the simulations .
the kinetic flux does approach zero for convective regions which are so thin that they are almost unstratified .
however , stratified convection zones have an asymmetry in up and down flows , giving a modest net kinetic energy flux ( downward for pure top driving , up for pure bottom driving , and both for more general cases , @xcite ) . we may express fluxes in units of the radiative flux of the background , @xmath125 , and in terms of the variables in the lorenz equations .
the excess vertical radiative flux is @xmath126 where @xmath127 , @xmath128 and @xmath129 , and @xmath130 the horizontal radiative flux is @xmath131 the enthalpy flux of the cell , in the vertical direction , is @xmath132 in general , without need for any additional mechanism to cause variability , _ the net flux of energy through a turbulent cell varies with time . _
a steady state solution for the lorenz equations is @xmath133 and , for @xmath134 , a second solution , for a stable convective roll , appears : @xmath135^{1/2 } ; z = r-1.\ ] ] the second solution is unstable if @xmath136 and @xmath137 where @xmath138 and this instability gives rise to fluctuations in velocity @xcite .
the instability also gives fluctuations in energy flow through the cell .
the steady state solutions might be of practical value for stellar evolutionary codes , to the extent that they provide an estimate of the average behavior of the convective cell .
however , the fluctuations are large ( in no sense are they `` small perturbations '' ) , and may cause nonlinear complications .
for example , a thermonuclear runaway would be sensitive to the largest value of the temperature fluctuations because that would affect the burning rate @xcite .
the fluctuations drive entrainment episodes , and affect the mixing of composition @xcite .
the fluctuations may be able to modify the driving of pulsations in stars with vigorous convection zones .
in red supergiants , coupling of turbulent fluctuations in the surface convection zones to both radial and non - radial pulsations seems likely ( see below ) . in order to compare the lorenz model to 3d simulations , it is desirable to have comparable starting conditions .
arbitrary choices can give large initial transients before the attractor controls the behavior .
the 3d simulations start with zero velocity but finite temperature deviations from an adiabatic gradient .
we take @xmath139^{1\over 2 } , z= r-1,\ ] ] as initial conditions , which sets the velocity to zero and uses steady state values for the temperature fluctuations .
figure [ fig1 ] shows the behavior of the lorenz model of convection ( panel labeled lorenz ) , for a similar number of pulses as shown for the 3d simulation , and the same variables : buoyancy power @xmath140 to the two - thirds power , and kinetic energy per unit mass @xmath141 .
the factors @xmath142 are chosen so that the steady state values are normalized to unity , but are qualitatively correct for comparison with the 3d simulations .
the velocities in fig .
[ fig1 ] are not precisely the same , the 3d simulations giving turbulent velocity while the lorenz speeds are more appropriate to coherent structures " .
they are related ( see 3.1 ) , as the large scale velocities become unstable and turbulent , and their kinetic energies are the same order of magnitude @xcite .
the time is measured in the dimensionless units of lorenz ; a turnover time is two transit times , and roughly the time between peaks .
we see that the peaks in buoyancy power slightly precede those in kinetic energy , as in the 3d simulations but less dramatically .
this difference is related to the fact that the lorenz model has viscous dissipation acting directly on the large scale velocities , while the 3d simulations have dissipation at the kolmogorov scale , which is separated from the large scale by the turbulent cascade , involving many , many reductions in length scales ( see section [ cascade ] ) .
additional modes would fill in the valleys " in the lorenz model ( see [ scell ] ) . over the time shown ,
the average kinetic energy is 0.968 of the formal steady state solution , which is shown in figure [ fig1 ] as the dashed horizontal line in the lorenz panel .
the degrees of freedom in the three - dimensional simulations , and in the lorenz model , are dramatically different .
the floating - point operation count differs by a large factor : @xmath143 ( several times @xmath144 zones times 7 effective scalar variables for simulations , versus three amplitudes for the lorenz model ) .
with such an extreme compaction , it is striking that they give a similar picture for fluctuations in convection . unlike the full 3d simulations , it is trivial to extend the lorenz model to later times
. figure [ fig8 ] shows an extension in time by a factor of 10 .
there is a relatively steady growth in amplitude up to time @xmath1457 , at which point chaotic behavior begins to appear .
the fluctuations in kinetic energy ( and @xmath146 ) are large , sometimes exceeding the steady state value ( @xmath1451 ) by a factor of 4 .
the convective luminosity from a single lorenz model shows the same qualitative behavior .
this drastic behavior raises two interesting possibilities : 1 .
the numerical simulations will follow the solutions of the lorenz equation , and exhibit vigorous and chaotic fluctuations at later times , or , 2 .
the multimode behavior of the simulations will allow even stronger dissipation , preventing extreme behavior .
either way , the result is important for the evolution of supernova progenitors , especially regarding the effects of fluctuations on mixing ( yields ) and outbursts .
simulations of multiple burning shells ( c , ne , o and si ) in 2d have been continued further than the 3d simulations for the oxygen burning shell .
they appear to follow the first option , so that the prediction from 2d simulations is that core collapse progenitors will have violent eruptions prior to core collapse @xcite .
full 3d simulations need to be extended to later times at which the lorenz model predicts chaotic behavior .
the oxygen shell might not last for the @xmath14570 transit times shown in figure [ fig8 ] ; a linear estimate suggests that it consumes its fuel in about 100 traversal times ( the damkhler number is @xmath147 , @xcite ) , so the background evolution should not be neglected for such time intervals .
oxygen burning is likely to be a more dynamic event than previously supposed ( e.g. , by @xcite ) . averaging over multiple cells
may moderate the net fluctuation ( see below ) . for convection zones in other stars
, the number of traversal times available may be much larger .
the sun has a deep surface convection zone ( 20 pressure scale heights ) ; a plume would , if unimpeded , fall through the convection zone in about 2 hours ; it would take about 7 centuries to process all the mass of the convection zone through the surface layers .
these times bracket the effective mixing time , so that of order @xmath148 to @xmath149 mixing times would occur over the age of the present - day sun .
figures [ figxy ] and [ figyz ] show the familiar long term behavior of the lorenz model @xcite .
figure [ figxy ] shows the trajectory in x and y space for the first 70 dimensionless time intervals . x is the dimensionless speed and y the dimensionless temperature variation in the horizontal direction .
both x and y switch signs , but there is an overall correlation because there is a net enthalpy flux , which is proportional to xy .
the correlation is in the sense that hot regions rise and cool regions sink .
the radiative flux associated with the horizontal temperature variation is proportional to y , and while on average over long times is zero , it achieves this in segments of time in which first one then the other direction ( sign ) are favored .
this is the strange attractor at work .
figure [ figyz ] shows the same time interval , but plots y and z ( the dimensionless temperature in the vertical direction , parallel to the gravity vector ) .
notice that z has positive values about which it orbits .
the stratification breaks the symmetry in vertical versus horizontal directions .
the radiative flux in the vertical direction has a steady part and a fluctuating part .
the latter is proportional to z , and is positive ( outward energy flow ) . in mlt ,
the schwarzschild discriminant ( @xmath150 ) is the temperature excess above adiabatic ; this excess is assumed to be proportional to the enthalpy flux .
this is inconsistent with the lorenz model , because the velocity x can have both signs while z does not .
the error arises because in mlt the speed of convection is taken to be intrinsically positive ( to avoid this problem ) , and may be traced back to a lack of conservation of mass ( `` blobs dissolve '' into the environment rather than flow back to complete a cycle ) . what happens if we reduce the three equations of lorenz to two , forcing one variable to be at its steady state value
this is similar to the mlt approach ( assuming the mixing length parameter is chosen so that the kinetic energy scale is physically correct ; see @xcite ) .
enforcing the steady state value for the vertical temperature excess , @xmath151 , but allowing @xmath74 and @xmath75 to vary , corresponds to a model with a single temperature variable ( like mlt ) .
such integrations are shown in fig .
[ fig11 ] .
there are no pulses in kinetic energy or buoyancy ; the curves quickly approach a constant value .
convection proceeds by steady motion in a roll ; a finite xy is required to give torque to make the roll . this `` two equation model '' no longer has a strange attractor ; the pulses have been eliminated .
this explains why stellar evolutionary calculations which use mlt do not show these fluctuations .
( -9,-8)(9,8 ) ( 0,0)8160200 ( -4,0)(3,6 ) ( -9.5,0)9 - 2020 ( 9.5,0)9160200 ( 4,0)(3,6 ) ( 0,0)8 - 2020 ( -2,-7)*up - draft * ( -4,7)*toroidal flow * ( -12,0)(4,10 ) ( -10,8)1.8135225 ( -10,8)1.5 ( -10,8)1.8 - 4545 ( -6,8)1.8135225 ( -6,8)1.5 ( -6,8)1.8 - 4545 ( -2,8)1.8135225 ( -2,8)1.5 ( -2,8)1.8 - 4545 ( 2,8)1.8135225 ( 2,8)1.5 ( 2,8)1.8 - 4545 ( 0,1)*up - draft * ( -4,1)*down - draft * ( -8,1)*up - draft *
figure [ fig1 ] shows that the primary difference in the lorenz model and the 3d simulations is that the lorenz model has only a single mode , while the simulations are obviously multimodal
. this difference may be superficial .
the lorenz model in this application ( as in the original meteorological one ) has additional modes implied by the turbulent cascade which mediates the damping ( i.e. , they are implicitly in @xmath152 ; see appendices [ viscous ] and [ prandtl ] ) .
a simple richardson cascade was discussed in [ cascade ] , in which @xmath153 , the fractional change in length scale for each step in the cascade , is assumed to be a constant @xcite .
this is not very plausible for the largest scale modes because they are the most sensitive to boundary conditions ( they must fit into the convective region ) , but is simple and instructive .
the fractional time spent in the cascade for each mode may be shown to be the fractional kinetic energy in that mode . using [ cascade ] , this gives @xmath154 for @xmath155 , or roughly 0.37 , 0.23 , and 0.14 for the first three .
there is a dominant mode accompanied by several weaker but significant ones .
one way to proceed would be to introduce additional modes into the lorenz model ( chosen with guidance by the 3d simulations ) , and to generate an expanded set of amplitude equations which generalize the three of lorenz ( eq .
[ lor1 ] , [ lor2 ] , and [ lor3 ] ) .
this would make the system multimodal , and allow for modal interactions .
this approach is a simplified version of an elegant proposal by lumley and collaborators @xcite : empirical eigenfunctions are constructed by pod from simulations , introduced into the differential equations to derive a set of nonlinear ode s for the amplitudes of each mode ( a galerkin projection ) , and this set of ode s is solved to generate the evolution of the average properties of the turbulence .
this is being explored @xcite as a way to capture more fully the dynamics and multimodal behavior seen in the 3d simulations .
suppose we envisage the convective region to be populated by cells , each of which is a separate lorenz model representing the largest mode , a convective roll , in that cell .
a special case is the convective core : the geometry is not that of a layer , but a sphere , as is indicated in figure [ fig5 ] .
the flow has a toroidal structure . in this case
the gravitational acceleration goes to zero at the origin ( the center of mass of the star ) , but the velocity need not be zero there .
if there is no net rotation , then the direction of the up - flow is undetermined , and will be chosen chaotically by the turbulent flow . for a convective shell
, we imagine that the layer is filled by cells , as shown schematically in figure [ fig4 ] . for the oxygen burning shell ,
the inner and outer radii are about @xmath156 and @xmath157 cm , respectively .
the area of the spherical shell , evaluated at the midpoint in radius , is @xmath158 , and the cell is taken to be roughly square , so that its area is @xmath159 .
the ratio of shell area to cell area is about @xmath160 , so we assume there are roughly @xmath161 cells spread over the spherical shell .
in general , the number of cells in a shell will depend upon the geometry of the convection zone .
if the cells are not synchronous , but act independently , the effect of the pulses will be smoothed when summed over the whole shell .
however , the cells may interact constructively ; the solution to this more complex problem remains open in theta and in phi .
simulations with larger aspect ratio ( larger angle wedges ) do show a moderation of the total fluctuation , in qualitative agreement with the discussion above .
figure 2 in @xcite indicates the complexity of interaction between multiple burning regions even in 2d . ] . at issue
are both the interactions between cells in a single convective region , and interactions between multiple convective regions associated with different burning shells .
the individual cells exhibit fluctuations not only in time , but also in space .
each cell represents a mode which is unstable , and destroys itself and reforms , usually somewhere else . because the medium is fluid
, the pattern of cells is much more dynamic and less regular than that of a crystalline solid , so the fig .
[ fig4 ] should be interpreted as representing a snapshot of a system which fluctuates in both space and time .
hinode / sot observations of the solar surface @xcite reveal such a highly dynamic , complex structure .
figure [ fig12 ] explores an idea of martin schwarzschild @xcite , who estimated the number of convective cells in the sun and in red giants and supergiants .
he argued that only a modest number of cells ( @xmath14512 ) would exist at any given time in a red giant or supergiant . to illustrate the point
, we use the lorenz model to approximate the behavior of a convective cell .
we have estimated the convective flux for 12 cells at random phase , by adding 12 time sequences from an computation like figure [ fig8 ] ( but extended to t=800 ) , starting at 12 randomly chosen times in this interval .
this time sampling is intended to approximate a spatial ensemble average .
their flux , summed and normalized , is shown as a solid line , for a dimensionless time from 60 to 70 , which corresponds to roughly 20 pulses .
the signal is noisy and looks chaotic " .
these fluctuating cells make up a convective region , and will couple to the normal modes of the star to cause both radial and non - radial pulsations .
the amplitude of these pulsations will depend upon the overlap integrals between the normal modes and the cell motion , and the stellar damping .
this suggests that the noisy behavior will be combined with the relatively cleaner periodicity of the normal modes , giving a power spectrum with a base like that shown in figure [ fig13 ] , but with superimposed spikes corresponding to the excited normal modes .
while turbulent convection alone is sufficient to cause luminosity fluctuations , it occurs in regions of high opacity and partial ionization , which also drive pulsation , so that composite behavior and multiple periods may be expected .
joel stebbins ( pioneer of photoelectric astronomy ) monitored the brightness of betelgeuse ( @xmath6 orionis ) from 1917 to 1931 , and concluded that `` there is no law or order in the rapid changes of betelgeuse '' @xcite , which seems apt for fig .
[ fig12 ] ( the lorenz strange attractor ) as well .
more modern observations @xcite show a strong broad - band noise component in the photometric variability .
the irregular fluctuations of the light curve are aperiodic , and resemble a series of outbursts .
direct 3d simulations of betelgeuse @xcite show the same complex behavior ( and have the advantage that they predict the detailed spectral behavior as well ) .
this should be no surprise ; the 3d equations have embedded in them a strange attractor .
increasing the number of cells reduces the level of fluctuation about the mean .
averaging over the 2 million granules of the sun gives a very stable luminosity , which would plot as a straight line in fig .
[ fig12 ] ( even 2000 random cells do this ) . however , the size of the cells at the bottom of the solar convection zone will be larger ( hence fewer cells ) , and if chaotic might give a long term modulation to the solar luminosity . full star simulations of the whole solar convective zone , with sufficient numerical resolution to give well developed turbulence , should shed light on this issue . application to turbulent stellar atmospheres
requires surmounting two difficulties : ( 1 ) the flow is no longer low - mach number ( see 2.2 ) , and ( 2 ) the ionization zone causes dramatic changes in opacity ( assumed constant in the lorenz model ) . fortunately 3d atmospheres exist , and analysis such as we have done on stellar interior convection simulations is feasible . while this paper was in preparation , @xcite re - examined the idea that giant convective cell turnover is the explanation of the long secondary period observed in semiregular red variable stars @xcite , including betelgeuse and antares .
stothers used mlt to derive a velocity scale for the overturn , also relying on general features of simulations of @xcite ; see 3 in @xcite .
this theory appears to work directly as a complement to the discussion above .
the use of a lorenz model for the giant cells already implies real dynamics .
the strange attractor necessarily provides variability in luminosity , with its own quasi - period @xcite and velocity scale ( see 4 in @xcite ) .
several giant cells are at work , each with a quasi - period of order of the transit time , and therefore similar to the estimate of stothers , and the observations .
the larger convective velocities needed are simply a necessary consequence of the dynamics implied by the convective luminosity @xcite . the introduction of turbulence as an _ active _ agent in the discussion of stellar variability ( e.g. , @xcite ) seems timely .
an interesting improvement would be to use pod empirical eigenvalues from simulations ( e.g. , @xcite , [ cascade2 ] above ) , and develop a low order dynamical model to explore long duration behavior .
we have identified a major new feature of stellar physics : chaotic behavior due to turbulent fluctuations in stellar convection , and corresponding luminosity fluctuations .
while the simulations upon which the analysis was based were fully compressible , the theory uses the approximation of sub - sonic flow . both numerically and analytically , a strange attractor like that of @xcite seems to appear naturally in stellar convection . as a first approximation to more rigorous analysis
, we have applied the lorenz model to kinetic energy fluctuations in the oxygen burning shell , to the turbulent energy cascade , and to fluctuations in luminosity in irregular variables .
figure [ fig1 ] shows a comparison of the turbulent kinetic energy fluctuations in 3d simulations of turbulent flow and in the lorenz model .
no parameters were adjusted to give a fit .
additional modes , appropriate for turbulent flow , would improve the comparison further .
this suggests a new , _ inherently nonlinear _
mechanism for variability in stars , the @xmath0-mechanism , which is caused by luminosity fluctuations _
directly _ associated with turbulent convective cells . because the mechanism is nonlinear , it is not captured by linear stability analysis , which is a mainstay of variable star theory .
such luminosity fluctuations may have been observed already in the broad - band noise seen in betelgeuse ( alpha orionis @xcite , and in the long secondary periods in pulsating agb stars @xcite , and are expected to be observable in principle in all stars with extensive surface convection zones , including those with
solar - like " variability .
this mechanism is probably the cause of the strongly driven pulsations found by @xcite in their red giant " model ; the development of those large pulsations was a clue which may now be more fully understood .
such fluctuations provide a source of perturbations for instabilities , and may induce mixing not presently accounted for in stellar evolutionary calculations .
the fluctuations in convective velocity are comparable to average values .
if these fluctuations couple to nuclear burning , as for example in cases of degenerate ignition , shell flashes , or later stages of oxygen and silicon burning , outbursts may develop .
this work was supported in part by nsf grant 0708871 and nasa grant nnx08ah19 g at the university of arizona , and by the clear sub - contract from university of michigan .
we wish to thank fr .
j. fues ( specalo vaticano ) , prof .
r. ruffini ( icranet ) , and prof . j. lattanzio ( monash ) , p. wood ( australian national university ) , and the aspen center for physics for their hospitality , prof f. timmes and s. starrfield for discussions , r. stothers for helpful email , and the second anonymous referee for constructive comments .
* mass conservation . * conservation of mass is enforced by use of a stream function ( @xcite , 9 ) ; the simplest solution is a two dimensional , cylindrical `` roll '' ( @xcite , p. 44 ) . the flow is assumed to be subsonic .
* momentum conservation.[amomcons ] * the navier - stokes equation may be written as @xmath162 in the low mach number limit ; the last term is the buoyant acceleration .
the flow executes a circle of radius @xmath163 ; see figure [ fig3 ] .
the density fluctuation is related to the temperature fluctuation by @xmath164 , where @xmath165 .
we separate the variable @xmath21 by considering the flow to be a mass flux which is constant around the ring at any given moment , represented by an average speed @xmath72 which is a function of time only .
the hydrostatic background in the convective region has an entropy @xmath166 which is constant .
then , @xmath167 where @xmath168 is the enthalpy per unit mass , and @xmath169 , and @xmath170 .
now @xmath171 if @xmath61 is constant . lorenz assumed a linear temperature decrease with height , which corresponds to constant gravitational acceleration @xmath61 . using a height @xmath172 , @xmath173 or @xmath174 we denote @xmath90 , so @xmath175 which corresponds to the environmental temperature .
for @xmath176 the system is convectively unstable , while for @xmath177 , the background temperature can never be negative .
we represent the temperature by @xmath178 * viscous damping.*[viscous ] the viscous damping term may be approximated by @xmath179 . here
the constant @xmath152 is the inverse of the viscous dissipation time scale @xmath180 .
the buoyant acceleration in the vertical direction is @xmath181 only a temperature ( buoyancy ) difference in the horizontal direction ( @xmath113 ) gives a net torque to turn the convective roll . with damping due to a linear viscosity term , @xmath182 integrating eq .
[ eqb1b ] over a complete cycle in @xmath183 gives ) generate a factor of @xmath184 while the integral over @xmath185 gives a factor of @xmath186 . ]
t_{3 } \label{eq1c}\ ] ] the temperature terms in @xmath188 integrate to zero because of the @xmath189 factor in the buoyancy term .
thus we have just two terms in eq . [ eq1c ] , a sink from the viscous damping and a source from buoyancy . in the turbulent case
, we might identify this rate of dissipation as an integral over the turbulent region , so it is transformed into a global quantity , @xmath190 .
this is the deceleration times the velocity , giving a deceleration of @xmath191 , so that @xmath192 .
the absolute value @xmath193 is used here because the characteristic time is @xmath194 , and the deceleration is thus @xmath195 .
the length scale is the depth of the convective shell ( i.e. , @xmath196 ) .
the convective speed @xmath72 is constant in space , so we may use a nonlinear damping term , as implied by @xcite , @xmath197 t_{3}. \label{eq1ca}\ ] ] * energy conservation.[aecons ] * at constant pressure , the first law of thermodynamics simplifies to @xmath198 where @xmath199 is the net heating - cooling rate from nuclear burning and neutrino emission , @xmath200 the specific heat at constant pressure , @xmath109 the temperature , @xmath201 the mass ( nucleon ) density , @xmath202 the flux of radiative energy , and @xmath203 the volumetric heating by turbulence . ignoring @xmath199 and @xmath203 , @xmath204 the divergence of flux of radiative luminosity is @xmath205 , \end{aligned}\ ] ] so that @xmath206 where @xmath207 is the thermal conductivity for radiative diffusion , for constant @xmath208 and @xmath200 .
the background temperature @xmath88 is chosen such that @xmath209 .
the radiative diffusion term is @xmath210 where @xmath211 is the extra radiative diffusion due to the deviation of temperature from @xmath88 , and @xmath212 is the inverse of the radiation diffusion time scale @xmath213 . * the classical lorenz equations.[aclassicl ] * following @xcite , we correct for an aspect ratio @xmath77 by a factor @xmath214 which deals with the excess in vertical over horizontal heat conduction ( @xmath215 versus @xmath216 ) . substituting for @xmath109 and @xmath88 in equation [ eqb2 ]
, we have @xmath217 coefficients of orthogonal functions must separately sum to zero , so @xmath218 we introduce a potential temperature @xmath219 and eliminate @xmath92 . if @xmath220 is independent of time , @xmath221 we define dimensionless variables @xmath222 and use equations [ eq1c ] , [ eqt3 ] , and [ eqt4 ] to get the lorenz equations in their usual form : equations [ lor1 ] , [ lor2 ] , and [ lor3 ] .
@xmath223 where @xmath0 is a time in thermal diffusion units , @xmath224 is the effective prandtl number , and @xmath225 is the ratio of the rayleigh number to its critical value for onset of convection .
the prandtl number is the ratio of coefficients of the viscous dissipation term to the thermal mixing term , @xmath226 , where @xmath227 is the viscous damping time .
the original lorenz parameters appear to be a fair choice to represent the flow in 3d simulations ( see figure [ fig1 ] ) .
this is consistent with an effective prandtl number for the numerical simulation of @xmath228 ; however , the severe reduction in degrees of freedom from the simulations to the lorenz model warns against taking the numerical value too literally .
we note for comparison that water has @xmath229 and air has @xmath230 .
apparently lorenz felt he was lucky ( @xcite , p. 137 ) ; he took a suggested value ( @xmath231 ) which gave chaotic behavior instead of a lower value , actually appropriate for air , for which his equations give stable rolls .
as a mathematical example this quantitative difference in the prandtl number is not significant , but for physical applications , it is . using the ratio of microscopic thermal diffusion to viscous time scale , @xcite ( page 178 and 185 ) suggest that @xmath232 .
@xcite finds the critical value for @xmath36 for instability of steady convection to be @xmath233 . for @xmath234 ,
steady convection is always stable , so that the @xcite value would never give turbulence . if @xmath235 , steady convection is unstable for sufficiently large rayleigh numbers .
this precise value for instability is a characteristic of the canonical lorenz model , and is affected by the particular choice of dissipation function ( see appendix a ) . in the numerical simulations ,
the effective prandtl number is dominated by the turbulent cascade , which gives mixing of both momentum and heat at a rate determined by the largest eddy size .
thus , in numerical simulations , the turbulent flow defines its own effective value of this parameter @xmath236 , and whatever value @xmath237 has , it is clearly above the threshold for instability for the system we have numerically simulated .
numerical simulations and experiment suggest , for developed turbulence at high reynolds numbers , @xmath238 , a value typical of many common gases .
we note that in the reynolds analogy " ( @xcite , p. 341 ) , osborne reynolds argued that the mechanisms for transport of heat and momentum were essentially the same in a turbulent medium , so that the effective turbulent prandtl number should be of order unity ( @xmath239 ) .
suppose we think of the prandtl number as the relative strength of the process which makes the velocity field isotropic to that which converts the kinetic energy into heat .
these occur at different ends of the turbulent cascade .
to better understand what this might mean , consider @xmath240(time to change direction)/(time to heat ) .
we approximate the time to change direction by the time to halve the kinetic energy , ( @xmath241 ) . at any length scale @xmath242 the turbulent cascade
has a transfer rate for kinetic energy of @xmath243 .
this implies a time spent at each length scale of @xmath244 .
if each level in the cascade is smaller on average in length scale by a factor @xmath245 , the total time for the cascade is @xmath246 , where the geometric series has been used for the summation ( see also @xcite , 7.8 ) .
this gives an effective prandtl number @xmath247 .
some representative values : for @xmath248 , this gives @xmath249 , for @xmath250 this gives @xmath251 , and @xmath252 gives @xmath253 .
this argument may tend to overestimate the prandtl number , so that @xmath237 would be smaller than the historical value chosen by lorenz ( @xmath231 ) .
the separation in size of the large scale eddies with those at the dissipation scale suggests that this coupling might be approximated in some ingenious way .
the microscopic viscosity might be considered a bare " value , to be renormalized to a dressed " value , in analogy to the field - theoretic treatment of interacting particles . @xcite
proposed the idea of reducing the size of a system a step at a time by grouping neighboring entities ( in this case molecules ) and treating each group as a single interaction .
@xcite has successfully implemented the general idea of coarse - graining , or weeding out the small scales " .
these general ideas may be applied to the turbulent cascade .
@xcite use renormalization - group ( rg ) analysis of turbulence with some success , and in particular , estimate the effective prandtl number for turbulent flow to be @xmath254 ; see also @xcite .
however there is some debate : in their review , @xcite warn the rg method leads to suggestive results when applied to turbulence however , its application to turbulence can not yet be called a major success , owing to the uncontrolled approximations currently required to implement it . "
a similar sentiment is found in @xcite , p. 290 .
we have captured the behavior of the pulses in the simulations , by a lorenz model using the parameters that lorenz used . in this sense , these values are relevant to our problem , although it is unclear from first principles what the precise value of the effective turbulent prandtl number should be .
the form for dissipation that lorenz used is not the same as implied by kolmogorov , so that the threshold for instability changes @xcite .
we may interpret the prandtl number @xmath37 and the rayleigh number @xmath36 in terms of a renormalization in which the existence of turbulence implies effective prandtl and rayleigh numbers for the convective cell .
the actual values of the microscopic viscosity and thermal conductivity have little feedback on the behavior of the largest eddies ; see section [ iles ] .
magnetic fields in real stars may affect the effective prandtl number further , which in the classical lorenz model affects in turn the development of instability in the convective roll . for problems which are insensitive to the details of the small scale flow , the values of the microscopic prandtl do not matter ; the turbulent system bootstraps to an effective dissipation for which the rules of kolmogorov hold .
this approximation applies to hydrostatic ( and mildly dynamic ) stellar evolution , in which the burning times are long compared to sound travel times . for these problems ,
iles is appropriate .
there are notable exceptions , such as a flame front in a medium of unmixed fuel and ash ( snia progenitor models ) , for which the small scales are important , and direct numerical simulations ( dns ) of the small scales is necessary .
finally , we recall how the lorenz model was constructed : a low order mode was chosen , a rayleigh number was chosen just above the onset of instability for that mode , and a prandtl number was chosen which gave interesting behavior .
while different parameter choices are mathematically interesting @xcite , their physical relevance must be re - evaluated .
the lorenz equations are a spartan subset of the fluid equations which contain the germ of chaos ; it is probably better to use them as a guide rather than a gospel . | three - dimensional ( 3d ) hydrodynamic simulations of shell oxygen burning @xcite exhibit bursty , recurrent fluctuations in turbulent kinetic energy .
these are shown to be due to a general instability of the convective cell , requiring only a localized source of heating or cooling .
such fluctuations are shown to be suppressed in simulations of stellar evolution which use mixing - length theory ( mlt ) .
quantitatively similar behavior occurs in the model of a convective roll ( cell ) of @xcite , which is known to have a strange attractor that gives rise to chaotic fluctuations in time of velocity and , as we show , luminosity .
study of simulations suggests that the behavior of a lorenz convective roll may resemble that of a cell in convective flow .
we examine some implications of this simplest approximation , and suggest paths for improvement .
using the lorenz model as representative of a convective cell , a multiple - cell model of a convective layer gives total luminosity fluctuations which are suggestive of irregular variables ( red giants and supergiants @xcite ) , and of the long secondary period feature in semiregular agb variables @xcite .
this @xmath0-mechanism " is a new source for stellar variability , which is inherently non - linear ( unseen in linear stability analysis ) , and one closely related to intermittency in turbulence .
it was already implicit in the 3d global simulations of @xcite .
this fluctuating behavior is seen in extended 2d simulations of cneosi burning shells @xcite , and may cause instability which leads to eruptions in progenitors of core collapse supernovae _ prior _ to collapse . |
let us consider the following nonlinear oscillator described by the so called modified emden equation with linear forcing term @xcite , @xmath1 here @xmath2 is a parameter . equation ( 1 ) can be considered as the cubic anharmonic oscillator with additional position dependent damping type nonlinear force @xmath3 .
this type of equation has been well studied in the literature .
for example , eq .
( 1 ) with @xmath4 arises in a wide range of physical problems : it occurs in the study of equilibrium configurations of a spherical gas cloud acting under the mutual attraction of its molecules and subject to the laws of thermodynamics @xcite and in the modelling of the fusion of pellets @xcite .
it also governs spherically symmetric expansion or collapse of a relativistically gravitating mass @xcite .
this equation can also be thought of as a one - dimensional analog of the boson ` gauge - theory ' equations @xcite .
equation ( [ mod01a ] ) has been shown to posses an unusual property which is not a general characteristic of a nonlinear equation : the frequency of oscillation of the oscillator is independent of the amplitude similar to that of a linear harmonic oscillator @xcite .
an oscillator which possesses this property is also known as an isochronous oscillator @xcite . for
a detailed study about isochronous orbits and isochronous oscillators one may refer to refs .
@xcite . equation ( [ mod01a ] ) admits the following nonsingular , periodic solution : @xmath5 here @xmath6 and @xmath7 are arbitrary constants , expressible in terms of the two integrals of motion or integration constants obtained by solving ( [ mod01a ] ) ( for details see ref .
@xcite ) . note that the angular frequency of oscillation @xmath2 continues to be the same as that of the linear oscillation . from this solution
it is obvious that for @xmath8 , equation ( [ mod01a ] ) exhibits the property of amplitude independence of the frequency of oscillation .
one can starightforwardly write down the solution of the initial value problem from the general solution ( [ mod02a ] ) .
for example , for the initial condition @xmath9 , @xmath10 , from ( [ mod02a ] ) we have the solution as @xmath11}{\sqrt{b^2+\omega^2}-b\cos\left[\omega t+\cos^{-1}\left(\frac{b}{\sqrt{b^2+\omega^2}}\right)\right]}.\end{aligned}\ ] ] note that @xmath12 is the amplitude of oscillation .
figure [ fig1 ] shows the periodic oscillations admitted by eq .
( [ mod01a ] ) for three different sets of initial conditions @xmath13 and @xmath14 with @xmath15 in terms of three different colours .
we note here that the frequency of the oscillations is independent of the initial conditions as in the case of the linear harmonic oscillator . ) exhibiting periodic oscillation for three different initial conditions ( three different colours ) and @xmath15 ( b ) phase space portrait of eq .
( [ mod01a]),width=529 ] one can trace the origin of this property of equation ( [ mod01a ] ) to the fact that it can be transformed to the linear harmonic oscillator equation , @xmath16 through a nonlocal transformation , @xmath17 the solution ( [ mod02a ] ) can be obtained ( see below , equation ( [ nld05 ] ) ) from the solution of ( [ horm1 ] ) , @xmath18 , where @xmath6 and @xmath7 are arbitrary constants and the frequency , @xmath2 , is independent of the amplitude .
such a linearization property is one of the fascinating features associated with a class of nonlinear equations exhibiting large number of symmetries and extensive search for such linearizing transformations is being made in the recent literature @xcite .
in fact , there exists a class of nonlinear oscillators which are connected to the linear oscillator equation ( [ horm1 ] ) through the following nonlocal transformation @xcite @xmath19 where @xmath20 is an arbitrary function of @xmath21 . now substituting ( [ int02 ] ) into ( [ horm1 ] )
we get a nonlinear ordinary differential equation ( ode ) of the form @xmath22 where prime denotes differentiation with respect to @xmath23 .
equation ( [ int03 ] ) is a special case of the well known lienard equation ( le ) @xcite @xmath24 one can also consider a more general nonlocal transformation of the form @xmath25 and substituting this in ( [ horm1 ] ) we get @xmath26 we find the above equation reduces to a linard type equation only for the choice @xmath27 .
interestingly for @xmath28 , equation ( [ int03 ] ) becomes the well known isotonic oscillator @xcite equation , @xmath29 the solution of the nonlinear equation ( [ int03 ] ) is obtained by using the identity @xmath30 since @xmath31 , where @xmath6 and @xmath7 are integration constants , is the solution of the linear harmonic oscillator ( [ horm1 ] ) , equation ( [ nld05 ] ) can be rewritten as the first order nonlinear differential equation of form @xmath32 now one can get the solution of ( [ int03 ] ) by solving ( [ mod07aa ] ) . in particular , for the specific case @xmath33 equation ( [ mod07aa ] ) becomes a bernoulli equation of the form @xmath34 the corresponding ode ( [ int03 ] ) becomes @xmath35 and equation ( [ mod01a ] ) is the special case corresponding to @xmath36 . upon integrating ( [ mod07b ] )
we get the periodic solution of ( [ mod01 ] ) as @xmath37^{\frac{1}{(2m+1)}}},\end{aligned}\ ] ] where @xmath38 , @xmath39 , @xmath40 , @xmath41 and @xmath7 are arbitrary constants . here
@xmath42 is a non - negative integer and @xmath2 is the angular frequency .
one can note that solution ( [ mod02 ] ) is also isochronous .
this has indeed been reported recently by iacono and russo @xcite using a different procedure . in figure [ fig2 ]
we show the periodicity of the solution for the case @xmath43 and with the initial conditions @xmath44 and @xmath10 .
we additionally remark here that the case @xmath45 , @xmath46 of equation ( [ mod01 ] ) is also exactly solvable but the solutions are of damped oscillatory type as will be proved later in this article ( sec .
[ sec2 ] ) even for coupled systems . ) for @xmath43 exhibiting periodic oscillations with the initial condition @xmath44 and @xmath10 for @xmath47 ( b ) phase space portrait of eq .
( [ mod01]),width=529 ] in this paper we will show that this unusual ( amplitude independent frequency ) property possessed by the class of equations ( [ int03 ] ) is not limited to scalar les alone but is also possessed by a class of @xmath0 coupled nonlinear oscillator equations . in order to demonstrate the existence of periodic and quasiperiodic solutions for such systems , we extend the above mentioned procedure to @xmath0 coupled les .
we also derive the integrals of motion and the general solution to the system of @xmath0 coupled les .
we also point out that there exists another class of @xmath0 coupled nonlinear oscillators which is almost integrable in the sense that it admits @xmath48 independent integrals and reduces to a single first order nonautonomous nonlinear differential equation that can be solved numerically .
the frequency of oscillation of its periodic solutions is again independent of amplitude . in the following section
we show that a system of @xmath0 coupled les can be explicitly integrated to find periodic and quasiperiodic solutions whose frequencies are identical to a system of uncoupled linear oscillators . in sec .
[ sec3 ] we deduce the underlying @xmath49 integrals of motion for the system and analyze their structure . in sec . [ nearlyintegrable ] we numerically solve the system of @xmath0 coupled les for which the explicit general solution is not known , but @xmath48 integrals are known .
we finally summarize our results in sec .
generalizing the above results of the scalar system ( [ int03 ] ) to a system of coupled nonlinear oscillators of lienard type , we relate them to a system of uncoupled @xmath0-dimensional linear harmonic oscillators . for this purpose , we consider the @xmath0-dimensional anisotropic harmonic oscillator equation of the form @xmath50 where the freqencies @xmath51 are in general different .
let us now introduce the nonlocal transformation @xmath52 where @xmath53 are arbitrary functions of the variables .
subsituting ( [ nhorm2 ] ) into ( [ nhorm1 ] ) we get a system of @xmath0 coupled lienard type second order nonlinear oscillator equations of the form @xmath54 where @xmath55 , @xmath56 . in the following we shall demonstrate the existence of periodic and quasiperiodic solutions with amplitude independent frequencies , @xmath51 ( isochronous property ) for the above mentioned nonlinear oscillator equation with specific forms of @xmath57 s . to obtain the desired results , we make use of the following identities , @xmath58 derived from equation ( [ nhorm2 ] ) .
obviously the solution of the system of second order linear odes ( [ nhorm1 ] ) is @xmath59 where @xmath60 s and @xmath61 s , @xmath62 , are @xmath49 integration constants . now substituting the solution ( [ ham_sol ] ) in equation ( [ nhorm4 ] )
we find the following system of coupled first order odes to represent ( [ nhorm3 ] ) , @xmath63 it may be noted that @xmath0 integration constants , @xmath61 s , @xmath62 , of the @xmath0 coupled second order odes ( [ nhorm3 ] ) appear explicitly in ( [ mod07 ] ) . in order to find the general solution of ( [ nhorm3 ] ) we need @xmath0 more integration constants which are to be obtained by integrating ( [ mod07 ] ) . for general forms of @xmath64 in ( [ nhorm3 ] ) this can not be done .
however for the special choice , @xmath65 equation ( [ mod07 ] ) can be integrated . in order to perform this integration
we multiply each equation of the system ( [ mod07 ] ) by @xmath66 for @xmath67 and the last equation by @xmath68 and subtract them to get @xmath69 dividing eq .
( [ int1 ] ) throughout by @xmath70 , we get @xmath71 integrating the above equation , we get @xmath72 rewriting the above equation , we get @xmath73 where @xmath74 s , @xmath75 , are @xmath76 integration constants .
we are also left with a first order ode of the form @xmath77 where @xmath78 and @xmath79 s are given by equation ( [ mod08 ] ) .
now the problem of solving the set of @xmath0 coupled autonomous second order odes ( [ nhorm3 ] ) is reduced to the problem of solving a single non - autonomous first order ode ( [ mod07a ] )
. therefore one can get the general solution for equation ( [ nhorm3 ] ) for the case ( [ mod07e ] ) by solving equation ( [ mod07a ] ) .
equation ( [ mod07a ] ) again can not be in general solved explicitly for arbitrary form of the function @xmath80 @xcite .
hence , in order to solve ( [ mod07a ] ) we assume that the function @xmath80 has a symmetry @xmath81 , where @xmath82 , @xmath83 and @xmath84 are arbitrary parameters .
this implies that @xmath80 is a homogeneous polynomial and we assume the following form of @xmath80 , @xmath85 where @xmath86 s are arbitrary functions of @xmath87 .
however , in the present case we assume @xmath88 s to be constants only for simplicity . even when they are functions of @xmath87 the following integration procedure holds good .
with the above choice of @xmath80 , equation ( [ nhorm3 ] ) reduces to the system of coupled nonlinear oscillator equations , @xmath89 the solution of this system of coupled nonlinear oscillators can be obtained by solving the following first order nonlinear differential equation obtained by substituting ( [ mod10e ] ) into ( [ mod07a ] ) along with ( [ mod08 ] ) : @xmath90 where @xmath91 , @xmath92 , and @xmath93 .
the above equation is of the first order bernoulli equation type @xcite , namely @xmath94 where @xmath95 and @xmath96 are arbitrary functions of the independent variable @xmath97 . with the substitution @xmath98 in ( [ bern - eq1 ] )
we get the following first order linear inhomogenous ode @xmath99 the general solution of ( [ bern1 ] ) is obviously @xmath100},\end{aligned}\ ] ] where @xmath101 is the integration constant . rewriting the above in terms of @xmath66 , we get @xmath102^{1/q}},\end{aligned}\ ] ] where @xmath101 is the @xmath103 integration constant which we are looking for .
the integral appearing in the denominator of the above expression can be integrated explicitly for arbitrary values of @xmath104 .
however , we find that the system admits oscillatory solutions only when @xmath104 is a positive integer and for this choice we find the above expression reduces to either of the following forms depending on the value of @xmath104 @xcite . *
( i ) case 1 - @xmath104 odd :* we take @xmath105 , @xmath106 .
then equation ( [ mod11 ] ) reduces to the form [ sol_odd ] @xmath107^{\frac{1}{2m+1}}},\end{aligned}\ ] ] where @xmath108 , @xmath109 and @xmath110 . using ( [ mod08 ] ) and ( [ mod12 ] ) we get the remaining @xmath76 variables as @xmath111^{\frac{1}{2m+1}}},\end{aligned}\ ] ] where @xmath75 .
note that for the solution ( [ sol_odd ] ) to be nonsingular periodic , we require the condition @xmath112 .
we note here that for the special case @xmath113 the solutions ( [ mod12 ] ) and ( [ mod13 ] ) become the periodic / quasiperiodic solution @xmath114 which exactly matches with the solution given in @xcite for the so called coupled modified emden equation . *
( ii ) case 2 - @xmath104 even * , here we take @xmath45 , @xmath115
. then equation ( [ mod11 ] ) reduces to the form [ sol_even ] @xmath116^{\frac{1}{2m+1}}},\end{aligned}\ ] ] where @xmath117 , @xmath118 , @xmath119 . using ( [ mod08 ] ) and ( [ mod12b ] ) we get the remaining @xmath76 variables as @xmath120^{\frac{1}{2m+1}}}\,\,.\end{aligned}\ ] ] from ( [ mod12b ] ) and ( [ mod12c ] ) we find that equation ( [ nhorm3a ] ) admits only oscillatory dissipative type solution for the choice @xmath45 , @xmath115 .
note that for @xmath104 odd positive integer in ( [ nhorm3a ] ) , one can have either periodic or quasiperiodic solutions , depending on whether the uncoupled frequencies @xmath121 s are commensurate or not . in fig .
1 we have presented quasiperiodic and periodic solutions for suitable choices of the uncoupled frequencies @xmath122 with @xmath123 in the form of projected phase space plots in the @xmath124 plane . we find that for the choice @xmath125 and @xmath126 , the system ( [ nhorm3a ] ) exhibits @xmath127 period oscillations .
similarly for the choice @xmath128 and @xmath126 , the system ( [ nhorm3a ] ) is found to exhibit quasiperiodic oscillations .
one can note that equation ( [ nhorm3a ] ) can also be rewritten in the following first order form as was done by iacono and russo @xcite for the scalar case as @xmath129 [ mod05 ] multiplying eq .
( [ mod05aa ] ) by @xmath130 and eq .
( [ mod05ab ] ) by @xmath68 and subtracting the resulting equations we get @xmath131 dividing throughout by @xmath132 and rewriting we get @xmath133 upon introducing the angle variable @xmath134 , equation ( [ mod06 ] ) becomes @xmath135 from ( [ mod06c ] ) it is obvious that the angle and hence the frequency ( which is similar to that of the linear harmonic oscillator ) are independent of the amplitude of oscillation , irrespective of the form of @xmath136 .
however , one may note that this does not always imply isochronocity as the amplitude of oscillation may decay with time , as shown in equation ( [ sol_even ] ) . ) in the @xmath124 plane with @xmath123 , for two different values @xmath43 ( figures ( a , c ) ) and @xmath137 ( figures ( b , d ) ) , respectively . figures ( a ) and ( b ) describe the @xmath127 period oscillations for the choice @xmath125 and @xmath138 .
figures ( c ) and ( d ) describe the quasiperiodic oscillations for the choice @xmath128 and @xmath138 . ,
in this section we show the existence of @xmath49 independent integrals of motion for equation ( [ nhorm3 ] ) with @xmath139 , @xmath140 , being a homogeneous polynomial ( [ mod10e ] ) . in order to show that there exists @xmath0 time dependent integrals ,
let us consider the equivalent form of ( [ mod07 ] ) , that is , @xmath141 rewriting the above equation ( the first and the last expressions ) we get @xmath0 time dependent integrals as @xmath142-\omega_it , \ ; i=1,2 , \ldots , n,\end{aligned}\ ] ] where @xmath143 . now to find the remaining @xmath0 integrals for @xmath104 odd or even integer in ( [ mod10e ] ) we proceed as follows . from ( [ mod16 ] )
we get @xmath144 . using this expression in the well known trignometric identity @xmath145 and @xmath146
, we get @xmath147 [ mod17 ] rewriting now ( [ mod08 ] ) in the form @xmath148 we obtain @xmath149 now by using ( [ mod17a ] ) we can write @xmath150 substituting the above in ( [ rev2 ] ) we obtain the first @xmath76 time independent integrals as @xmath151 for @xmath104 odd positive integer , substituting ( [ mod17 ] ) and ( [ mod15 ] ) into ( [ mod12 ] ) and rearranging we arrive at the following form for the integral @xmath152 , @xmath153 which is the @xmath0th time independent integral . for @xmath113
the integral ( [ mod14 ] ) becomes @xmath154 one can note that for @xmath113 the integrals ( [ mod16a ] ) , ( [ mod15 ] ) and ( [ mod14a ] ) exactly match with those presented in @xcite .
for instance , equation ( [ mod14 ] ) exactly reduces to the corresponding form given in @xcite for the scalar case ( @xmath155 ) .
similarly , for the case where @xmath104 is even positive integer , we substitute ( [ mod17 ] ) and ( [ mod15 ] ) into ( [ mod12b ] ) and rearranging we arrive at the following form for the integral @xmath156 , @xmath157 we note here that the above first integral is a time dependent one .
in sec . [ sec3 ] we obtained the general solution of ( [ nhorm3 ] ) for the choice @xmath158 , @xmath140 and @xmath80 is a homogeneous polynomial as in ( [ mod10e ] ) .
however , we wish to point out that the system ( [ nhorm3 ] ) for any arbitrary choice of ( [ mod07e ] ) is almost integrable as there always exist @xmath0 time dependent integrals ( [ mod16a ] ) and @xmath159 time independent integrals ( [ mod15 ] ) . for complete integrability only the first order nonautonomous differential equation ( [ mod07a ] ) needs to be integrated . for those forms of @xmath80
for which this can not be done explicitly , one can always carry out a numerical integration of ( [ mod07a ] ) or apply a suitable approximation method to find @xmath160 .
the remaining @xmath161 s , @xmath162 , can be obtained readily using ( [ mod08 ] ) and @xmath160 .
equation ( [ mod06c ] ) ensures that if the solutions are periodic or quasiperiodic , the frequency is independent of amplitude . ) in the @xmath163 plane for the choices ( a ) @xmath164 exhibiting @xmath165 period oscillation , ( b ) @xmath166 exhibiting quasiperiodic oscillation ] in order to demonstrate the above , let us consider the special case @xmath167 in ( [ nhorm3a ] ) and @xmath168 .
this choice reduces equation ( [ nhorm3a ] ) to the following system of coupled second order nonlinear odes , [ nonintegrable ] @xmath169 the solution of the above system of equations can be deduced by solving the following first order ode obtained by the procedure discussed in the previous section , @xmath170 however , we find that the explicit general solution of the above first order ode is not known @xcite
. one can apply suitable numerical methods to solve this equation .
figure [ fig4 ] is plotted by solving eq .
( [ first ] ) using a variable step size runge - kutta fourth order method .
we find the system ( [ nonintegrable ] ) exhibits periodic and quasiperiodic oscillations which is shown in fig [ fig4 ] . here
the projection of the phase space of ( [ nonintegrable ] ) in the @xmath163 plane for the set @xmath171 and @xmath172 are shown for @xmath165 periodic behaviour and quasiperiodic behaviour , respectively .
.[table1]a comparison of the solution of eq .
( [ first ] ) obtained using the variable step size runge - kutta fourth order ( rk4 ) numerical procedure and the homotopy perturbation method ( hpm ) .
[ cols= " < , < , < " , ] one can also apply perturbation techniques such as the homotopy analysis @xcite to find approximate solution of high accuracy for eq .
( [ first ] ) and compare the results with the numerical analysis
. the homotopy perturbation method @xcite involves the introduction of an artificial parameter , say @xmath173 , into the original nonlinear equation @xmath174 as @xmath175 where @xmath176 is the linear operator corresponding to the linear part of the given equation and the @xmath6 is the operator corresponding to the given nonlinear equation . here @xmath177 is the lowest order approximate solution . expressing the solution @xmath97 as a power series in @xmath173 , i.e. @xmath178 where @xmath179 , @xmath180 are the higher order approximations , one can substitute this series solution in ( [ first ] ) to obtain a system of linear first order ordinary differential equations . solving this system with suitable initial conditions one can obtain the approximate solution to the given equation in the limit @xmath181 .
for further details on this procedure one can refer to refs .
@xcite . in table [ table1 ] we compare the solution of eq .
( [ first ] ) obtained through this procedure with the numerical solution for the parametric choice @xmath182 .
the values listed in table [ table1 ] for the homotopy perturbation method are calculated up to third order approximation .
we note here that the accuracy of the perturbation solution will improve if further higher order approximations are taken into the calculation @xcite , which we do not pursue here .
in this paper , we have shown that a system of @xmath0 coupled nonlinear lienard type oscillators admits periodic and quasiperiodic solutions or damped oscillatory periodic solutions with amplitude independent frequency of oscillations .
we have derived explicit general solution and @xmath49 integrals of motion for this system .
thus we prove this system to be completely integrable .
we have also shown that another system of @xmath0 coupled lienard type oscillators is almost integrable in the sense that it admits @xmath48 independent integrals and reduces to a single nonautonomous first order nonlinear differential equation .
we have also shown that this almost integrable system also exhibits periodic and quasiperiodic oscillations for suitable parametric choices .
for the general system of @xmath0 coupled nonlinear oscillators ( [ nhorm3 ] ) with arbitrary form of nonlinearity , the nonlocal transformations reduce it to a system of @xmath0 coupled first order odes ( [ mod07e ] )
. it will be interesting to investigate further whether other forms of nonlinearity ( different from ( [ mod10e ] ) ) are also amenable to analysis .
the work is supported by a department of science and technology ( dst)ramanna fellowship project and a dst
irhpa research project of m. l. , who is also supported by a dae raja ramanna fellowship .
jhs is supported by a dst fast track young scientist research project . v. j. erwin ,
w. f. ames and e. adams _ wave phenomena : modern theory and applications _ ed c rogers and j b moodie ( amsterdam : north - holland ) g. c. mcvittie , _ mon . not .
r. astron .
soc . _ * 93 * 325 ( 1933 ) ; _ ann .
inst . h poincar _ * 6 * 1 ( 1967 ) ; * 40 * 231 ( 1984 ) a. kimiaeifar , a. r. saidi , g. h. bagheri , m. rahimpour and d. g. domairry , _ chaos , solitons & fractals _ * 42 * 2660 ( 2009 ) ; m. bayat , m. shahidi , a. barari and g. domairry , _ int . j. phys .
sci . _ * 5 * 1074 ( 2010 ) | existence of amplitude independent frequencies of oscillation is an unusual property for a nonlinear oscillator .
we find that a class of n coupled nonlinear linard type oscillators exhibit this interesting property .
we show that a specific subset can be explicitly solved from which we demonstrate the existence of periodic and quasiperiodic solutions .
another set of @xmath0-coupled nonlinear oscillators , possessing the amplitude independent nature of frequencies , is almost integrable in the sense that the system can be reduced to a single nonautonomous first order scalar differential equation which can be easily integrated numerically .
nonlinear oscillators , coupled ordinary differential equations , complete integrability , isochronous systems |
a lot of effort has been focused on studying the spectrum and the properties of light quark mesons ( see @xcite-@xcite ) and references therein ) . in these systems
non - perturbative effects are dominant , therefore they are the best candidates for understanding the mechanisms underlying confinement and dynamical chiral symmetry breaking which are fundamental elements and of crucial importance for the theory . in some of these studies
the quark propagator equation has provide useful insights in the light quark sector of qcd @xcite , @xcite .
it is also interesting to explore , using that fundamental block of the theory and the quarks bound state equation , the transition to heavy quark physics .
we plan to canvass , by using an effective kernel for the gap equation , how the quark mass affects the infrared and ultraviolet dressing of the propagator and how that in turn will alter the dynamics of the bound state of quarks .
a more realistic case of a quark mass dependent version of the effective kernel is also explored in the same light .
the compton size of quarks and mesons is also used to qualitatively understand the relative infrared ( ir ) and ultraviolet ( uv ) qcd dynamics and inspired an approach that enabled us to reach a physical bound state for @xmath0 q = u / d , s , c mesons .
aspects related to the finite size of hadrons involving recent ideas by brodsky and shrock @xcite are also briefly discussed .
the dyson - schwinger equation for the quark propagator ( gap equation ) has the form : @xmath1 @xmath2 is the renormalized dressed gluon propagator , @xmath3 is the renormalized dressed quark - gluon vertex , @xmath4 is the regularization mass scale , with @xmath5 being the gluon and quark propagator renormalization constants . using the rainbow truncation for the gap equation and introducing @xmath6 : @xmath7 where @xmath8 is the free gluon propagator , we can disentangle the equation from the rest of the dyson - schwinger equations .
the unrenormalized quark self - energy term of the gap equation in the rainbow truncation is : @xmath9 where we have set @xmath10 , @xmath11 is the gluon momentum and the factor @xmath12 comes from the trace over the color indexes . by taking the dirac trace of the last equation we get : @xmath13 and
if we multiply by @xmath14 and then take the dirac trace , we get the second equation : @xmath15 where we have introduced the quark propagator amplitudes @xmath16 , @xmath17 : @xmath18 the quark propagator in terms of @xmath19 , @xmath20 is then : @xmath21 using the propagator renormalization condition , @xmath22 , we get @xmath23 @xmath24 where @xmath25 is the renormalized current quark mass at point @xmath26 and it is a parameter we fit to experimental data .
the amplitude ( bsa ) @xmath27 for a meson state of quarks of flavors a and b is given from the solution of the homogeneous bethe - salpeter equation ( bse ) : @xmath28_{tu}= \int^{\lambda}\frac { d^{4}\tilde{q}}{(4\pi)^{4 } } k_{tu}^{rs}(p,\tilde{q},p ) \nonumber\\ \times [ s^{a}(\tilde{q}+\eta p ) \gamma^{ab}(\tilde{q},p ) s^{b}(\tilde{q}-\bar{\eta}p)]_{sr } \label{eq : bse}\end{aligned}\ ] ] p is the total momentum , @xmath29 ( @xmath30 ) is the momentum partitioning parameter for the quark ( antiquark ) and @xmath31 , @xmath32 $ ] .
@xmath33 is the unknown renormalized amputated irreducible quark - antiquark scattering kernel .
physical observables are independent of the partitioning parameter .
the most general form of the bsa for psudoscalar mesons has four invariants while for the vector mesons has eight ( see @xcite,@xcite ) and we use a four chebychev polynomial expansion for each one of them .
these amplitudes are lorentz scalar functions of @xmath34 , @xmath35 , @xmath36 and the momentum partitioning parameter @xmath29 . for qq mesons that parameter
help us avoid having the propagator singularities inside their mass shell bse integration domain . since for the mass shell momentum @xmath37 , where @xmath38 is the meson mass , the quark momenta in ( eq . [ eq : bse ] ) are in general complex numbers .
this requires the solution of the gap equation in the appropriate parabolic region in the complex plane .
the ladder truncation for the bse is an approximation for the equation s kernel : @xmath39^{rs}_{tu } \to } \nonumber \\ & & -4\pi\,\alpha(q^2)\ , d_{\mu\nu}^{\rm free}(q ) \textstyle[{\frac{\lambda^i}{2}}\gamma_\mu]^{ru } \otimes \textstyle[{\frac{\lambda^i}{2}}\gamma_\nu]^{ts } \
, , \label{eq : ladder}\end{aligned}\ ] ] the electroweak decay constant @xmath40 of a charged pseudoscalar meson @xcite expressed in terms of the meson normalized bsa and quark propagators : @xmath41\bigg\ } \label{eq : decayconstps } \end{aligned}\ ] ] where @xmath42 and @xmath43 is the number of colors , from the trace over the color indexes .
similar expression exists for vector mesons .
for the unknown effective running coupling we are going to use a kernel - model that has been developed within the rainbow - ladder truncation of dyson - swchwinger equations .
the model respects some of the most important symmetries of qcd , like chiral symmetry and poincare covariance , while it provides quark dressing , dynamical chiral symmetry breaking , and most important , quark confinement .
it has been used to study the physics of dcsb and related phenomena , like the spectrum of light quark mesons ( @xcite @xcite , @xcite ) , decay constants ( @xcite , @xcite , @xcite , @xcite , @xcite ) and other physical observables ( @xcite , @xcite , @xcite ) , in good agreement with experimental data ( @xcite , @xcite , @xcite , @xcite ) .
the so called maris - tandy ( mt ) model @xcite has the form : @xmath44 } \label{eq : mtmodel}\end{aligned}\ ] ] for the parameters we have @xmath45 and @xmath46 and the u / d- and s - current quark masses at the renormalization scale @xmath26 , fitted to the experimental masses of pion and kaon , are @xmath47 and @xmath48 .
for the other two quarks , we use the masses in @xcite , where @xmath49 and @xmath50 .
the model essentially simulates both gluon and quark - gluon vertex dressing effects .
the phenomenological first term defines the behavior in the infrared region and provides the infrared enhancement necessary for the right value of the quark condensate in the chiral limit .
the second term is important in the ultraviolet region and is set up so that will reproduce the 1-loop perturbative qcd running coupling behavior .
recently an effort has started to extend the applications of the mt model in the heavy quarks region ( @xcite , @xcite @xcite ) . due to its behavior in the complex domain
the solution of the heavy quark gap equation is challenging . by changing the integration variable in the equation , from the quark internal momentum to the gluon momentum , solved that problem and the only limitation comes from the propagator s singularities .
the type and the exact location of these points is not known .
notice that the thranslationally invariant regularization of the integrals allows that change in the integration variable and variations of parameter @xmath51 and that our approach to numerically solve the equation is different than that in @xcite . in the case of the c quark for example with current mass @xmath52 and for the parabolic region determined by @xmath53 where @xmath54 is the bse integration variable , with @xmath55 and @xmath56
( the peak of the region will be then at @xmath57 ) amplitudes @xmath58 , @xmath59 and @xmath60 are plotted in figures ( [ fig : real_a_kcp_32_32_32_32_e-7_p2 - 14_n_0.5_0.88 ] , [ fig : real_m_kcp_32_32_32_32_e-7_p2 - 14_n_0.5_0.88 ] , [ fig : real_sigma_s_kcp_32_32_32_32_e-7_p2 - 14_n_0.5_0.88 ] ) correspondingly . for our present studies of c
- quark mesons we need to solve the gap equation for @xmath61 as small as @xmath62 , so the mass shell point ( peak ) is at @xmath63 and the singularities are far from the bse integration domain and they present no problem .
this is also tested by varying parameter @xmath64 . for c quark in a parabolic region in the complex plane ( @xmath65 , @xmath29=0.50 ) . in this plot and in the following ones
the straight lines connecting the end - points have no significance and should be ignored.,width=302,height=188 ] for c quark in a parabolic region in the complex plane ( @xmath65 , @xmath29=0.50).,width=302,height=188 ] for c quark in a parabolic region in the complex plane ( @xmath65 , @xmath29=0.50 ) with integration over the gluon momentum ( kcp ) .
the peak of the parabolic region is at @xmath66 .
we see indications of the existence of a pair of complex conjugate singularities near the peak of the region , approximately located at ( @xmath67 , @xmath68)@xmath69(-2.7 , @xmath703.5 ) @xmath71 .
the type of the singularities is unknown.,width=302,height=188 ] from these plots we can see that only amplitude @xmath72 ( and the same is true for @xmath73 ) has singularities , but not a and m. we can conclude then that the singularity is the point where the denominator of @xmath74 vanishes , i.e. @xmath75 . by keeping only the infrared first term of the model the solution for @xmath76 for the same quark
is plotted in fig .
( [ fig : re_sigma_s_1st_mt_term_kcp_32_32_32_32_e-7_p2 - 8.0_n_0.5_0.88 ] ) . for c quark in a parabolic region in the complex plane ( @xmath77 , @xmath29=0.50 ) keeping only the first ir term of the kernel .
the peak of the parabolic region is at @xmath78 .
the solution is different but we still have the general characteristics of the solution from the complete mt model .
the first pair of singularities approximately appear to be ( @xmath67 , @xmath68)@xmath69(-1.7,@xmath702 ) @xmath71 .
, width=302,height=188 ] in the case where we keep only the uv term in the mt model , the solution has the first singularity on the real axis ( see fig .
[ fig : re_sigma_s_2st_mt_term_kcp_32_32_32_32_e-7_p2 - 8.0_n_0.5_0.88 ] ) . for the c quark in a parabolic region in the complex plane ( @xmath77 , @xmath29=0.50 ) with integration over the gluon momentum ( kcp ) and keeping only the uv term in the model .
the peak of the parabolic region is at @xmath78 .
the first singularity is on the real axis around ( @xmath67 , @xmath68)@xmath69(-1.9 , 0 ) @xmath71 .
, width=302,height=188 ] comparing the c and b quark real part of the mass amplitude re(m ) , the last one appears to be almost flat near the peak of the parabolic region .
re(a ) also appears to vary little , just above one , in the complex plane .
one may consider that such behavior of re(m ) and re(a ) justifies for a constituent like approximation for the b quark propagator . from the b quark re(@xmath72 )
plot we found the singularities have not only moved , as was expected , deeper in the time - like region , along the real axis , but also further apart along the imaginary axis .
this is exactly the opposite from a constituent - like behavior of a propagator . that also signifies the importance of the imaginary part of the dressed quark mass amplitude in the behavior of the propagator amplitudes @xmath72 and @xmath73 . for the b quark in a parabolic region in the complex plane ( @xmath79 , @xmath29=0.50 ) with integration over the gluon momentum ( kcp ) with the full mt model .
the peak of the parabolic region is at @xmath80 .
traces of a first pair of singularities can be seen near the peak of the region and they are approximately located at ( @xmath67 , @xmath68)@xmath69(-24.0,@xmath7012 ) @xmath71 . ,
width=302,height=188 ] in table [ table : full_ir_uv_approximate_singularities_location ] we collect the data for the approximate location ( @xmath67,@xmath68 ) of the singularities for the c and b propagators for the full , ir and uv gap solution . in that table
we include a qualitative estimation of the quark mass dressing for each case .
we assume that the location of the singularities along the real axis can be used to extract that piece of information . from these data
we observe there is an increase in the quark mass dressing as we go from c to b quark using the full model .
for the c quark there is an almost equal dressing from the ir and uv term of the interaction while for b quark the uv term provides more than twice the mass dressing of the ir term .
[ cols= " < , < , < , < , < , < , < " , ] with this modification of the dressing of the b - quark we can find meson masses that are only about 7 - 12 % smaller than the experimental ones . for the only decay constant experimentally known there is a -24.4 % difference .
the present results and the earlier results for the equal quark systems appear to provide a partial confirmation of the recent suggestion by brodsky and shrock of a maximum wavelength for quarks and gluons in mesons @xcite , @xcite , @xcite .
the existence of such a wavelength for quark and gluons would be due to confinement in mesons with size of the order of 1 fm .
that in turn , through the compton relation , will introduce a minimum quark and gluon momentum inside hadrons .
same reasoning can be applied to baryons .
these scales though will depend on the bound state of quarks , requiring different scales to be artificially introduced for each case and mathematically accommodate the dyson - schwinger equations solution . if qcd is the correct theory for the interaction of quarks and gluons something like that is not desirable .
the present studies may indicate a possible scenario for the natural realization of these scales through the combined ir dynamics of the quark mass dependent kernel and the quark propagators in full qcd calculations .
some simple preliminary studies qualitatively support that idea but further and more detailed investigation is required .
the size of the qq mesons becomes smaller as quark mass increases and the dressed quark quasi - particles themselves become smaller in size .
the infrared component of the kernel relates to large distance gluons ; such components should be less physically relevant for internal dynamics of heavy , small compton size , particles .
this is supported by our finding that the heavier qq quarkonia receive diminishing contributions from the infrared component of our model kernel ; the uv component alone provides a very good description of the @xmath81 states .
we stress that this refers not only to the binding interaction but also to the quark self - energy dressing . in detail , as the quark mass increases , due to its smaller compton size , uv dressing will increase and ir dressing will be strongly suppressed .
model independent evidence indicate that the short distance dressing will be , after certain mass scale , more than the long distance dressing , and at a higher mass scale , it may be over and above the long distance dressing of the lightest u / d quarks .
the heavy quarks will be limited in a smaller area in space and the binding force will be provided by the exchange of sort distance gluons , signifying a smaller overall binding energy .
the heavy quark mass and the now relatively stronger uv dressing , _ triggered by _ that large mass , make feasible , with a very weak force , to bind the particles together . on the other hand ,
the qq mesons have a size that does not diminish significantly with increasing heavy quark mass .
so the infrared sector of the binding interaction should remain relevant .
however the self - energy dressing of the heavy quark should not receive strong contributions from large distance gluons .
this becomes evident by our finding that a suppression of the infrared component of the b - quark dressing kernel allows a physical @xmath0 , @xmath82 meson state even though the binding effective interaction and the dressing of the light quarks remains unchanged . in a similar way
as the quark momentum increases the particle will become more localized , receiving _ simultaneously _ smaller infrared _ and _ ultraviolet dressing and feeling less of the color field of the other quark(s ) in the bound state .
the wave - like nature of quarks can accommodate that interpretation .
aspects of this study may be related to brodsky and shrock s suggestion of a maximum wavelength of quarks and gluons in hadrons .
the present work may actually suggest a possible scenario for the natural realization of such scales in qcd .
further studies are necessary in that direction .
l. chang , i. c. cloet , b. el - bennich , t. klahn and c. d. roberts , chin .
c * 33 * , 1189 ( 2009 ) [ arxiv:0906.4304 [ nucl - th ] ] .
m. s. bhagwat , a. hoell , a. krassnigg , c. d. roberts and s. v. wright , few body syst . * 40 * , 209 ( 2007 ) [ arxiv : nucl - th/0701009 ] .
p. maris and c.d .
roberts , int .
j. mod .
e * 12 * , 297 ( 2003 ) [ arxiv : nucl - th/0301049 ] .
p. maris and c.d .
roberts , phys . rev .
c * 56 * , 3369 ( 1997 ) [ arxiv : nucl - th/9708029 ] .
p. maris , c.d .
roberts and p.c .
tandy , phys .
b * 420 * , 267 ( 1998 ) [ arxiv : nucl - th/9707003 ] .
l. chang , y. x. liu , m. s. bhagwat , c. d. roberts and s. v. wright , phys .
c * 75 * , 015201 ( 2007 ) [ arxiv : nucl - th/0605058 ] .
s. j. brodsky and r. shrock , phys .
b * 666 * , 95 ( 2008 ) [ arxiv:0806.1535 [ hep - th ] ] .
p. maris and p. c. tandy , phys .
c * 62 * , 055204 ( 2000 ) [ arxiv : nucl - th/0005015 ] .
bhagwat , m.a .
pichowsky , c.d .
roberts and p.c .
tandy , phys .
c * 68 * , 015203 ( 2003 ) [ arxiv : nucl - th/0304003 ] .
p. maris and p.c .
tandy , phys .
c * 61 * , 045202 ( 2000 ) [ arxiv : nucl - th/9910033 ] .
lei chang , yu - xin liu , mandar s. bhagwat , craig d. roberts , stewart v. wright .
phys . rev .
c * 75 * , 015201 ( 2007 ) .
r. alkofer , p. watson and h. weigel , phys .
d * 65 * , 094026 ( 2002 ) [ arxiv : hep - ph/0202053 ] .
p. maris , a. raya , c. d. roberts and s. m. schmidt , eur .
j. a * 18 * , 231 ( 2003 ) [ arxiv : nucl - th/0208071 ] .
amendolia _ et al . _
[ na7 collaboration ] , nucl .
b * 277 * , 168 ( 1986 ) .
g. eichmann , r. alkofer , i. c. cloet , a. krassnigg and c. d. roberts , arxiv:0802.1948 [ nucl - th ] .
s. j. brodsky and r. shrock , arxiv:0803.2554 [ hep - th ] .
s. j. brodsky and r. shrock , arxiv:0803.2541 [ hep - th ] .
although we use an approximate relation to calculate @xmath83 and there is some uncertainty , the parameters vary slowly in that mass area and we do nt have to worry much about the exact value of the quark mass corresponding to these parameter values . | by using a previously developed phenomenological kernel for the study of the light quark qcd sector and dynamical chiral symmetry breaking effects we will examine the relative infrared and ultraviolet qcd dynamics for j=0,1 meson properties . for the same reasons we extend and explore a quark mass depended generalization of the kernel in the heavy quark region and we also compare with the original model .
the relation between the dynamics of the quark propagator and the effective kernel with the j=0,1 qq and qq mesons and quarks compton size is also discussed . |
the standard model ( sm ) of particle physics has proven to be a robust theoretical model that very accurately describes the properties of elementary particles observed in nature and the forces of interaction between them . in this model , the electromagnetic and weak forces
are unified into a single electroweak theory .
the measured masses of the particles that mediate the electroweak force , however , are vastly different the photon has zero mass while the @xmath1 and @xmath2 bosons have masses almost 100 times heavier than the mass of a proton .
to explain this difference , the theory predicts the existence of a higgs field which interacts with the elecroweak field via electroweak symmetry breaking to produce masses for the @xmath1 and @xmath2 bosons while leaving the photon massless .
interaction with the higgs field would also explain how other fundamental particles acquire mass .
an additional spin-0 particle , the higgs boson , is also predicted to arise from the higgs field .
this particle is the only sm particle that has not been observed in nature and evidence of this boson would be a direct test of the theory .
the higgs mechanism is predicted to give mass to other particles , yet the mass of the higgs boson itself is a free parameter of the theory that must be determined experimentally .
direct searches at the large electron - positron collider ( lep ) at cern and indirect electroweak measurements result in a preferred sm higgs boson mass @xmath3 between 114.4 and 185 gev / c@xmath4 at 95% confidence level ( c.l ) . in this region ,
the range 156 @xmath5 177 gev / c@xmath4 has additionally been excluded at 95% c.l . by direct searches at the fermilab tevatron @xmath6 collider @xcite and the range above @xmath7 146 ( 145 )
gev / c@xmath4 has been excluded at 95% c.l . by direct searches at the atlas ( cms )
experiment from the @xmath8 large hadron collider ( lhc ) at cern @xcite . at the tevatron ,
the higgs boson would be produced most often through gluon fusion ( @xmath9 ) , followed by associated production with either a @xmath1 or @xmath2 vector boson ( @xmath10 , @xmath11 ) and vector boson fusion ( @xmath12 ) .
figure [ production ] shows diagrams of these processes and table [ tab : sigma ] gives the predicted sm cross sections for higgs boson masses between 100 and 150 gev / c@xmath4 .
the sm higgs boson search strategy at the tevatron is driven by the most dominant decay modes . at lower mass ( @xmath13 135 gev / c@xmath4 ) , @xmath14 @xmath15 provides the greatest sensitivity to higgs boson observation despite the exclusion of the gluon fusion process for this channel due to large multijet backgrounds . for @xmath7 135 gev / c@xmath4
, @xmath16 provides the greatest sensitivity .
further sensitivity to a sm higgs observation at the tevatron is obtained by the inclusion of more challenging channels such as @xmath17 .
the sm branching fraction for a higgs boson decaying to a photon pair b(@xmath18 ) is very small , reaching a maximal value of about 0.2% at @xmath3 = 120 gev / c@xmath4 ( see table [ tab : sigma ] ) .
the greatest sensitivity for this channel , however , is obtained for 110 @xmath5 140 gev / c@xmath4 , in the preferred region from electroweak constraints and in a region where combined tevatron higgs boson searches are least sensitive @xcite .
the diphoton final state is also appealing due to its cleaner signature compared to @xmath19 jets .
the better reconstruction efficiency for photons provides a larger relative acceptance of @xmath18 events and the photon s better energy resolution leads to a narrow mass peak for the higgs boson , which is a powerful discriminant against smoothly falling diphoton backgrounds .
these experimental signatures help make the diphoton final state one of the most promising search modes for higgs boson masses below 140 gev / c@xmath4 at atlas and cms experiments at the lhc , which have recently presented first results in this channel @xcite .
in addition to sm @xmath18 production , one can devise many possible beyond the standard model ( bsm ) scenarios where b(@xmath18 ) is enhanced . ) can be found in reference @xcite . ]
any resonance observed could also then be evidence for a bsm higgs . in the sm , the spontaneous symmetry breaking mechanism requires a single doublet of a complex scalar field .
however , it is likely that nature does not follow this minimal version and that a multi - higgs sector may be required . here
, we also consider a model which requires a doublet higgs field for which the symmetry breaking mechanism responsible for giving higgs masses to gauge bosons is separate from that which generates the fermion masses . in the benchmark model
considered , a `` fermiophobic '' higgs boson ( @xmath20 ) is predicted that assumes sm couplings to bosons and vanishing couplings to all fermions .
the gluon fusion process is then suppressed and only vh and vbf processes remain , which results in a reduction in the production cross section by a factor of four .
this reduction is compensated , however , by the branching fraction for this model , which can be larger than that predicted by the sm scenario by more than two orders of magnitude for low higgs boson masses ( see table [ tab : sigma ] ) .
the higher branching fraction causes a larger number of predicted fermiophobic higgs boson events compared to the sm higgs boson .
direct searches at lep set a lower limit on the fermiophobic higgs boson mass of 109.7 gev / c@xmath4 with 95% c.l . here , we present a search for both a sm and fermiophobic higgs boson in the diphoton final state from @xmath6 collisions at @xmath21 1.96 tev from the fermilab tevatron collider
. an inclusive sample of diphoton data are collected by the d0 and cdf experiments , corresponding to an integrated luminosity of 8.2 and 7.0 fb@xmath0 , respectively . by combining the results from each analysis , tevatron limits on the sm cross section multiplied by b(@xmath18 ) are also presented relative to the sm prediction .
the dominant backgrounds to prompt photons originating from the event vertex are electrons faking photons and jets faking photons .
the latter is more frequent and typically occurs when a jet fragments into a @xmath22 or @xmath23 meson which then decays to multiple photons .
these delayed photons are collinear and are often mis - reconstructed as a single photon . in order to identify high - energy prompt photons and reduce these backgrounds ,
both the d0 and cdf analyses start by searching for photon candidates with the following signature : ( i ) the majority of the energy should be deposited in the em calorimeter rather than the hadronic calorimeter , ( ii ) the em cluster should be isolated in the calorimeter , ( iii ) there should be no high - momentum tracks originating from the primary event vertex that are associated with the em cluster , and ( iv ) the em shower profile is consistent with that of a prompt photon .
more details on this identification can be obtained from the primary references , ref .
@xcite for d0 and ref .
@xcite for cdf .
at d0 , we select events with at least two photon candidates with @xmath24 1.1 and transverse momentum @xmath25 25 gev / c . in addition to the basic photon selection outlined above , a neural network ( nn ) is used to further discriminate jet backgrounds from prompt photons .
this nn discriminant is trained using photon and jet monte carlo ( mc ) samples and constructed from well - understood detector variables sensitive to differences between photons and jets .
the output for this discriminant ( o@xmath26 ) is shown in fig .
[ d0_kin_input ] for the photon and jet mc samples , where the signal peaks near one and the background near zero .
the photon efficiency @xmath27 for any cut on this output is determined from a true photon sample in the data obtained from @xmath28 events ( @xmath29 or @xmath30 ) .
the jet efficiency @xmath31 ( fraction of jets misidentified as a photon ) is determined from a sample of jets misidentified as photons in the data .
a cut of 0.1 is applied which retains more than 98% of true photons and rejects 40% of misidentified jets .
the two highest @xmath32 photons that pass this selection are used to form the diphoton system and the diphoton mass @xmath33 is then required to be greater than 60 gev / c@xmath4 .
the difference in azimuthal angle of the two photons @xmath34 is also required to be greater than 0.5 , which keeps 99% of the higgs boson signal but reduces prompt qcd photons originating from fragmentation , a process not well modeled in the simulation . for a higgs boson signal decaying to two photons , data with the above selection
is composed of both an irreducible and a reducible background .
the irreducible background is from two sm qcd photons from the hard interaction where the shape for different kinematic variables is modeled from sherpa mc and the normalization of this background is obtained from a fit made to the final discriminant distribution when setting limits on higgs boson production .
the reducible background is composed of fake events where at least one photon candidate is misidentified as a prompt photon .
both the shape and normalization of kinematic distributions for drell - yan @xmath35 events are obtained from pythia mc prediction . the shape for @xmath36jet and jet+jet background distributions
is obtained from an independent data sample selected from diphoton events that pass all other photon selection requirements but have a reversed requirement on the @xmath37 output for one or both photon candidates .
the normalization for this sample is determined from the data using a 4@xmath384 matrix method . for each data event that passes the full selection ,
a 4-component vector is constructed @xmath39 where the value of one element is 1 and other elements are 0 based on whether one or both photon candidates pass a stronger requirement of @xmath40 . the weight @xmath41 ( @xmath42 ) then represents events where both photon candidates pass ( fail ) and @xmath43 ( @xmath44 ) represents events where only the leading ( subleading ) photon candidate in the event and subleading refers to the second highest . ] photon candidate passes .
the efficiency of this cut for the photon and jet samples ( @xmath27 and @xmath31 ) are parametrized as a function of @xmath23 and used to construct a 4@xmath384 efficiency matrix @xmath45 .
a system of linear equations @xmath46 is then solved on an event - by - event basis in order to obtain a weight for events with two true photons @xmath47 , a weight for events with two jets faking a photon @xmath48 , and a weight for events where only the leading ( subleading ) candidate is a true photon @xmath49 ( @xmath50 ) .
then , for example , the number of dijet events is taken as the sum of the dijet weights obtained for each data event : @xmath51 similarly , the @xmath52jet background is determined from the sum of the @xmath49 and @xmath50 components .
the resulting background composition is then determined for each higgs boson mass hypothesis ( @xmath53 gev ) where the approximate percentage of @xmath54 events is 53% , of @xmath55 events is 44% , and of drell - yan events is 3% .
the simulated shape of the higgs boson signal is generated separately for each production process and normalized using cross section and branching fraction values shown in table [ tab : sigma ] .
see the primary d0 reference for more detail on signal expectation ( ref .
@xcite ) .
the most recent searches for a higgs boson in the diphoton final state exploit the narrow @xmath3 resolution to discriminate between the signal and background . with a good understanding of the background modeling for different kinematic distributions , however ,
further sensitivity is gained at d0 by using these variables with a multivariate technique .
in addition to the diphoton mass @xmath33 , four other variables with kinematic differences between the signal and background are also considered for this analysis : the transverse momentum of the diphoton system @xmath56 , @xmath57 , and the transverse momentum of the leading and subleading photon , @xmath58 and @xmath59 respectively .
these five well - modeled kinematic variables are used to construct a single discriminant from a boosted decision tree ( bdt ) algorithm , trained to distinguish a higgs boson signal from the backgrounds .
this process is repeated for each higgs mass hypothesis for 100 @xmath5 150 gev / c@xmath4 in 2.5 gev / c@xmath4 steps , and is trained separately for a sm and fermiophobic higgs boson signal . as an example , the @xmath33 and @xmath56 distributions ( @xmath60 115 gev / c@xmath4 ) for the data , background , and expected sm and fermiophobic higgs boson signal are shown in fig .
[ d0_kin_input ] , along with the resulting bdt response for both the sm and fermiophobic scenario ( figure [ d0_kin_bdt ] ) .
at cdf , the leading two photons are required to have @xmath61 gev / c .
plug photons ( @xmath62 ) are selected using a standard i d used at cdf @xcite .
central photons ( @xmath63 ) are identified in a similar manner as in the d0 analysis , using a nn output constructed from variables sensitive to distinguishing prompt photons from jet backgrounds .
this nn discriminant , used for the first time at cdf in this analysis , increases photon signal efficiency by 5% and background rejection by 12% relative to standard photon i d at cdf .
photon efficiencies for both central and plug photons are validated using z @xmath64 decays in both data and mc .
@xmath18 signal acceptance is further increased by reconstructing events in which a single central photon converts into an electron - positron pair , which is found to occur approximately 15% of the time for @xmath63 .
a base set of selection requirements is applied that searches for a central electron or @xmath65 . ] with a colinear , oppositely signed track nearby .
track associated with it is referred to as the primary ( secondary ) electron . ]
the proximity of the two electron tracks is determined from their @xmath66 separation at the radius of the conversion and from the difference in @xmath67 of the two tracks , where @xmath68 .
the tracks from both electrons are required to point to a fiducial electromagnetic energy cluster .
photons of a higher @xmath32 range are selected by requiring the secondary electron to have @xmath25 1.0 gev / c and the reconstructed conversion photon to have @xmath25 15 gev / c . in order to reject jet backgrounds , only a small fraction of hadronic @xmath69 associated with the primary electron s cluster
is allowed .
additionally , requirements are made on the conversion candidate s calorimeter isolation which is obtained from the primary electron s isolation energy @xcite with the secondary electron s @xmath32 subtracted if its track points to a different calorimeter phi tower .
the shape describing the ratio of transverse energy to transverse momentum ( @xmath70 ) is peaked at one for isolated photon conversions , but has a long tail for photon conversions from @xmath22 or @xmath23 @xmath71 decays due to the extra energy from the unconverted photon .
restrictions on this ratio then provide a further way to remove jet backgrounds .
the conversion @xmath69 is obtained from the primary electron s @xmath69 with the secondary electron s @xmath32 added if it is in a different calorimeter tower while the photon s reconstructed transverse momentum is obtained by adding the vector sum of the two track s momenta at the radius of the conversion .
a final requirement removes events with a small radius of conversion , primarily to reduce prompt electron - positron pairs from dalitz decays of neutral pions @xmath72 .
the direction of the conversion photon s momentum is obtained by taking the vector sum of the individual track momenta ; however , better @xmath18 mass resolution is obtained by setting the total momentum to be the conversion s energy obtained from em calorimeters , which additionally constrains the photon s mass to zero .
reconstruction of photon conversions in the cdf analysis provides an improvement of about 13% in sensitivity to a higgs boson signal .
data events in the cdf analysis are divided into four independent subsamples according the position and type of the photon candidate . in cc events ( the most sensitive category ) , there are two photons in the central region of the detector . in cp events ,
one photon is in the central region and one is the plug region . for each of these categories ,
the two highest @xmath32 photons in the sample are selected .
if a cc or cp event is not identified , then two additional categories are considered . in c@xmath73c events ,
both photons are central but one has converted and is reconstructed from its @xmath74 decay products . finally , in c@xmath73p events , one photon is in the plug region and the other is a central conversion photon .
region used for the sm higgs boson search and then divided into three @xmath56 regions ( b)(d ) used for the @xmath20 search .
each distribution shows a fit to the data for the higgs boson mass hypothesis of 120 gev / c@xmath4 .
the gap in the fit centered at 120 gev / c@xmath4represents the signal region for this mass point , which was excluded from the fit .
the expected shape of the signal from simulation is also shown in the inset in ( a).,title="fig : " ] region used for the sm higgs boson search and then divided into three @xmath56 regions ( b)(d ) used for the @xmath20 search .
each distribution shows a fit to the data for the higgs boson mass hypothesis of 120 gev / c@xmath4 .
the gap in the fit centered at 120 gev / c@xmath4represents the signal region for this mass point , which was excluded from the fit .
the expected shape of the signal from simulation is also shown in the inset in ( a).,title="fig : " ] for the fermiophobic model , the higgs boson is typically produced in association with either a @xmath1 or @xmath2 boson or two jets from the vbf process .
as a result , the fermiophobic higgs boson has a higher than average @xmath56 relative to the background processes as it is typically recoiling against another object .
therefore , in the fermiophobic scenario , the data are further divided into three regions of @xmath56 , where the highest @xmath56 region provides the greatest @xmath20 sensitivity , retaining about 30% of the signal and removing 99.5% of the background . by also including the two lower @xmath56 regions , a gain in @xmath20 sensitivity of about 15%
is obtained compared to using just the higher @xmath56 region alone . at cdf
, we use a data - driven background model which takes advantage of the higgs boson mass resolution ( 3 gev or less ) and smoothly falling background in the signal region of the diphoton mass spectrum .
fits are made to the data excluding a 12 gev window centered around each higgs mass hypothesis .
the fit is interpolated into the signal region to determine the background estimation in that region and the process is repeated for each subsample ( cc , cp , c@xmath73c , c@xmath73p and also each @xmath56 region for the fermiophobic higgs search ) .
fits are performed separately for each higgs boson mass hypothesis for 100 @xmath5 150 gev / c@xmath4 in 5 gev / c@xmath4 steps .
the statistical uncertainties on the total background in the signal region , taken from the fit , are considered when setting limits .
fits for the cc channel are shown in figure [ fig : cdf_fits ] for the whole @xmath56 region used in the sm search and for each of the three @xmath56 regions used in the fermiophobic search .
no obvious evidence of a signal is observed in the tevatron diphoton data and bdt discriminants ( invariant mass distributions ) are used by the d0 ( cdf ) analysis to set upper limits on the cross section multiplied by the branching ratio @xmath75 for both the sm and fermiophobic higgs boson searches at 95% confidence level ( c.l . ) .
systematic uncertainties on both the predicted number of signal and background events are considered , in addition to systematics on the shape of the bdt discriminant for the d0 analysis .
correlations between uncertainties are also taken into account .
the d0 analysis uses a modified frequentist approach to set upper limits on the higgs boson production rate and the cdf analysis uses a bayesian method . the cdf and d0 observed and expected ( median , background - only hypothesis ) limits are shown relative to the sm prediction in figure [ fig : sm_limits ] .
the bands indicate the 68% and 95% probability regions where the limits can fluctuate , in the absence of signal . for the cdf limit at @xmath76 120
gev / c@xmath4 , a deviation of greater than two sigma from the expectation is observed .
however , the statistical significance of this discrepancy is reduced below two sigma after the trial factor associated with performing multiple mass points is considered .
results from the individual searches from cdf and d0 are combined @xcite in order to extract limits on sm higgs production @xmath75 relative to the sm prediction for @xmath77 gev / c@xmath4 in 5 gev / c@xmath4 steps .
figure [ fig : tev_limits ] shows these results obtained from a bayesian method . in order to reduce model dependence from the sm predictions ,
limits are also calculated on the inclusive cross section times the branching ratio @xmath78 with theoretical uncertainties on the total production cross section removed ( fig .
[ fig : tev_limits ] ) . in the fermiophobic higgs boson search , sm cross sections and uncertainties are assumed with the @xmath79 process excluded , and the limits on @xmath80 are converted into limits on @xmath81 , seen in fig .
[ fig : bh_limits ] . based on an intersection between the observed limit and the model prediction ,
fermiophobic higgs boson masses are excluded below 112.9 ( 114 ) gev / c@xmath4 for the d0 ( cdf ) analysis .
analyses for the d0 and cdf experiments searched for a higgs boson in the diphoton final state using data corresponding to 8.2 and 7.0 fb@xmath0 , respectively . the results are interpreted in the context of both a sm and fermiophobic higgs boson and are found to significantly improve upon the most recent diphoton searches from the tevatron @xcite by more than doubling the amount of data included and by implementing enhanced search techniques .
the d0 analysis uses a neural network to identify photons with @xmath82 and gains considerable improvement by implementing a boosted decision tree technique to better separate the higgs boson signal from the background .
the cdf analysis now also uses a nn to identify central photons with @xmath82 and additionally gains signal acceptance by including forward ( @xmath62 ) and central photon conversions .
the combined results from these analyses are used to set 95% c.l .
upper limits on sm higgs boson production . for a higgs boson mass of @xmath60 115
gev / c@xmath4 , the observed ( expected ) limit is a factor of 10.5 ( 8.5 ) times the sm prediction .
these results significantly extend the sensitivity of the separate d0 and cdf results and are the most stringent limits on the sm @xmath18 process obtained from tevatron data . in the fermiophobic interpretation , the d0 ( cdf ) data set a lower limit on on the fermiophobic higgs boson mass of 112.9 ( 114 ) gev / c@xmath4 at 95% c.l .
each experiment alone , therefore , produces a more stringent lower limits than that of 109.7 gev / c@xmath4 obtained from combined searches at lep .
results obtained by combining the d0 and cdf data ( after the dpf conference ) .
this yields a lower limit of @xmath83 119 gev / c@xmath4 , the most stringent limit on the fermiophobic higgs model . ] | recent searches conducted at the fermilab tevatron for the higgs boson in the diphoton decay channel are reported using 7.0 fb@xmath0 and 8.2 fb@xmath0 of data collected at the cdf and d0 experiments , respectively .
although the standard model ( sm ) branching fraction is small , the diphoton final state is appealing due to better diphoton mass resolution compared with dijet final states .
in addition , other models such as fermiophobic models where the higgs does not couple to fermions predict much larger branching fractions for the diphoton decay . here , results are presented for both a sm and fermiophobic higgs boson as well as a sm search based on a combination of the cdf and d0 analyses . |
the models studied in this paper are composed of spinless charged particles , classical or quantum , which are non - relativistic , i.e. they behave according to schrdinger and not dirac . on the other hand ,
the interaction of charged particles via the radiated electromagnetic ( em ) field can be considered either non - relativistic ( nonretarded ) or relativistic ( retarded ) . in the nonretarded regime ,
magnetic forces are ignored by taking the speed of light @xmath0 , so that the particles interact only via instantaneous coulomb potentials . in the retarded regime , @xmath1 is assumed finite and the particles are fully coupled to both electric ( longitudinal ) and magnetic ( transverse ) parts of the radiated field .
one of the tasks in the equilibrium statistical mechanics of charged systems is to determine how fluctuations of microscopic quantities like charge and current densities , induced electric and magnetic fields , etc .
, around their mean values are correlated in time and space .
a special attention is devoted to the asymptotic large - distance behavior of the correlation functions and to the sum rules , which fix the values of certain moments of the correlation functions .
two complementary types of approaches exist in the theory of charged systems .
the microscopic approaches , based on the explicit solution of models defined by their microscopic hamiltonians , are usually restricted to the nonretarded regime .
a series of sum rules for the charge and current correlation functions has been obtained for infinite ( bulk ) , semi - infinite and fully finite geometries ( see review @xcite ) .
the quantum sum rules are available only for the jellium model of conductors ( sometimes called the one - component plasma ) , i.e. the system of identically charged pointlike particles immersed in a neutralizing homogeneous background , in which there is no viscous damping of the long - wavelength plasma oscillations .
the macroscopic approaches are based on the assumption of validity of macroscopic electrodynamics .
being essentially of mean - field type , they are expected to provide reliable results only for the leading terms in the asymptotic long - wavelength behavior of correlations . in general , these approaches are able to predict basic features of physical systems also in the retarded regime . a macroscopic theory of equilibrium thermal fluctuations of the em field in quantum media , conductors and dielectrics , was proposed by rytov @xcite . in a recent work @xcite , a sequence of static or time - dependent sum rules , known or new , was obtained for the bulk charge and current density correlation functions in quantum media fully coupled to the radiation by using rytov s fluctuational electrodynamics .
a technique was developed to extract the classical and purely quantum - mechanical parts of these sum rules .
the sum rules were critically tested on the jellium model .
a comparison was made with microscopic approaches to systems of particles interacting through coulomb forces only @xcite ; in contrast to microscopic results , the current - current density correlation function was found to be integrable in space , in both classical and quantum cases .
[ fig.1 ] this paper is a continuation of the previous study @xcite .
it aims at generalizing the previous sum rules to inhomogeneous situations , in particular to the interface geometry of two semi - infinite media with different dielectric functions pictured in fig .
it should be emphasized that this is not exactly the configuration considered in some previous studies .
the standard configuration was a conductor in contact with an `` inert '' ( not fluctuating ) wall of the static dielectric constant @xmath2 .
the presence of a dielectric wall is reflected itself only via the introduction of charge images ; the microscopic quantities inside the inert wall do not fluctuate , they are simply fixed to their mean values .
such a mathematical model can provide a deformed description of real materials and , as is shown in this paper , it really does . the only exception from the described inert - wall systems is represented by the specific ( two - dimensional ) two - densities jellium , i.e. the interface model of two joint semi - infinite jelliums with different mean particle densities , treated in @xcite .
it stands to reason that in the case of the vacuum @xmath3 plain hard wall , there is no charge which could fluctuate and the inert - wall model is therefore adequate . to our knowledge
, the sum rules for a ( fluctuating ) conductor medium in contact with a dielectric ( inert ) wall obtained up to now were restricted to the charge - charge density correlation functions .
the inhomogeneous charge - charge sum rules are either of dipole type or they are related to the long - ranged decay of the surface charge correlation function along the interface .
the classical dipole sum rule for the static charge - charge density correlations follows directly from the carnie and chan generalization to nonuniform fluids of the second - moment stillinger - lovett condition @xcite .
the time - dependent classical dipole sum rule was derived in @xcite .
a time - dependent generalization of the carnie - chan rule to the quantum ( nonretarded ) jellium and the consequent derivation of the quantum dipole sum rule for the time - dependent charge - charge density correlations were accomplished in ref .
@xcite .
the bulk charge correlation functions exhibit a short - ranged , usually exponential , decay in classical conductors due to the screening . on the other hand , for a semi - infinite conductor in contact with a vacuum or ( inert ) dielectric wall , the correlation functions of the surface charge density on the conductor decay as the inverse cube of the distance at asymptotically large distances @xcite . in the classical static case ,
this long - range phenomenon has been obtained microscopically @xcite as well as by using simple macroscopic argument based on the electrostatic method of images @xcite ; the prefactor to the asymptotic decay was found to be universal , i.e. independent of the composition of the coulomb fluid . in the quantum case of the specific jellium model , ignoring retardation effects , a nonuniversal prefactor to the asymptotic decay was obtained , for both static @xcite and time - dependent @xcite correlation functions .
recently @xcite , by using the inhomogeneous version of rytov s fluctuational theory , we have extended the quantum analysis of the jellium to the retarded case . we got a surprising result : for both static and time - dependent surface charge correlation functions , the inclusion of retardation effects causes the quantum prefactor to take its universal static classical form , for any temperature .
the restoration of the classical prefactor by retardation effects was observed subsequently for arbitrary ( conductor , dielectric , vacuum ) configurations of two semi - infinite quantum media @xcite . in this paper
, we apply the inhomogeneous version of the rytov fluctuational theory to extend the bulk sum rules , derived in @xcite , to the geometry of two joint semi - infinite media with distinct dielectric functions in fig .
the sum rules derived for the charge - charge density correlations represent a generalization of the previous ( dipole moment and surface charge ) ones , valid for a conductor system in contact with an inert dielectric wall , to two fluctuating semi - infinite media of any kind .
the fundamental differences between the results for the inert and fluctuating wall descriptions are pointed out .
the charge - current and current - current sum rules obtained here are , to our knowledge , new .
the current - current sum rules indicate a breaking of the directional invariance of the diagonal current - current correlations by the interface .
the sum rules are expressed explicitly in the classical high - temperature limit ( the static case ) and for the jellium model ( the time - dependent case ) .
the paper is organized as follows . in sec .
2 , we review the inhomogeneous rytov theory of em field fluctuations and write down basic expressions for the charge and current densities ; explicit results for the elements of the retarded green function tensor for the two semi - infinite media configuration in fig.1 are presented in appendix .
dipole sum rules for the charge - charge density correlation functions are derived in sect .
the sum rules related to the long - ranged tail of the surface charge density correlation function along the interface between two media are discussed in sect .
4 which is divided into three parts .
in the first part 4.1 , we generalize the classical static analysis of a medium in vacuum @xcite to arbitrary media configurations .
part 4.2 concerns the derivation of a classical static relation between the dipole moment and the large distance asymptotic of the surface charge density .
part 4.3 is a brief recapitulation of the quantum case , in both retarded and nonretarded regimes .
the sum rules for the charge - current and current - current density correlation functions are the subject of sects . 5 and 6 , respectively .
section 7 is the conclusion .
we consider the ( 3 + 1)-dimensional space of points , defined by euclidean vectors @xmath4 and time @xmath5 .
we shall deal with semi - infinite geometries , inhomogeneous say along the first coordinate @xmath6 ; it is useful to denote the remaining two coordinates normal to @xmath6 as @xmath7 .
the model consists in two distinct semi - infinite media ( conductors , dielectrics or vacuum ) with the frequency - dependent dielectric functions @xmath8 and @xmath9 which are localized in the complementary half spaces @xmath10 and @xmath11 , respectively , so that the interface between the media is localized at @xmath12 ( see fig .
we shall assume that the media have no magnetic structure , i.e. they are not magnetoactive , and the magnetic permeabilities @xmath13 .
the two - point functions studied in this paper will be translationally invariant in time and in the vector space @xmath14 perpendicular to the @xmath6 axis , and so we shall use the ( partial ) fourier representation @xmath15 where @xmath16 denotes the frequency and @xmath17 is the two - dimensional wave vector .
the induced em fields inside a material medium are random variables which fluctuate in time and space due to the random motion of charged particles . in the long - wavelength scale , the em fluctuations are described by the rytov theory @xcite .
this theory is usually formulated in the weyl gauge with the scalar potential @xmath18 , so that the ( classical ) vector potential @xmath19 with components @xmath20 @xmath21 determines the microscopic electric and magnetic fields as follows @xmath22 ( we use gaussian units ) , where @xmath1 is the speed of light . in the context of the quantized em field theory , the crucial role
is played by the retarded photon green function tensor @xmath23 defined by @xmath24 where @xmath20 denotes the vector - potential operator in the heisenberg picture and the angular brackets represent the equilibrium averaging at temperature @xmath25 , or the inverse temperature @xmath26 . for non - magnetoactive media ,
the green function tensor possesses the symmetry @xmath27 in the fourier space , the symmetry is expressible as @xmath28 the validity of macroscopic maxwell s equations for the mean values of the em fields implies , in the frequency fourier space , a set of differential equations of dyadic type fulfilled by the green function tensor : @xmath29 d_{lk}(\omega;{\bf r},{\bf r } ' ) = - 4 \pi \hbar \delta_{jk } \delta({\bf r}-{\bf r } ' ) .\ ] ] here , in order to simplify the notation , the vector @xmath4 is represented as @xmath30 .
the _ source _ point @xmath31 and the index @xmath32 only act as some fixed parameters , the boundary conditions are with respect to the _ field _ point @xmath33 .
there is an obvious boundary condition of regularity @xmath34 for asymptotically large distances @xmath35 . at an interface between two different media , the boundary conditions
correspond to the macroscopic requirement that the tangential components of the fields @xmath36 and @xmath37 , considered in the gauge ( [ 2.2 ] ) , be continuous .
the green function tensor for the geometry pictured in fig .
1 was obtained in a number of papers , see e.g. the method using vector wave functions @xcite or the weyl expansion method @xcite .
the results are usually written in a complicated way , by using the dyadic notation for the tensors . in order to enable the reader to reproduce easily calculations performed in this work , in the appendix we present explicitly the fourier transforms ( [ 2.1 ] ) of the tensor elements @xmath38 for two possible cases : the points @xmath33 and @xmath31 are in the same half space or they are in different half spaces . applying the fluctuation - dissipation theorem and assuming the symmetry ( [ 2.5 ] ) ,
the fluctuations of the vector potential are described by the formula @xmath39 where i m means the imaginary part and @xmath40 is the fourier transform of the symmetrized correlation function @xmath41 here , @xmath42 represents a truncated equilibrium average , @xmath43 .
the relation ( [ 2.7 ] ) enables us to calculate the symmetrized two - point correlation function of arbitrary statistical quantities .
let a scalar quantity @xmath44 be expressible in terms of the components of the vector potential , in the classical format and in the gauge ( [ 2.2 ] ) , as @xmath45 , where @xmath46 @xmath21 are operators acting on time and space variables . within the spectral representation with a single frequency @xmath16 and two - dimensional vector @xmath47 , @xmath48 , this relation takes a partially algebraic form @xmath49 ( @xmath50 can still act as operators on the @xmath6 coordinate ) .
it follows from the definition ( [ 2.1 ] ) that the fourier transform of the symmetrized two - point correlation function of statistical quantities @xmath44 and @xmath51 , @xmath52 , is then determined by @xmath53 where @xmath54 is given by ( [ 2.7 ] ) .
the statistical quantities of our interest are the volume charge density @xmath55 and the electric current density @xmath56 .
the charge density , defined by @xmath57 , is expressible in terms of the vector potential components as follows @xmath58 where the abbreviated notation @xmath59 is used .
the vector components of the electric current density , defined by @xmath60 , are expressible as @xmath61 , \label{2.11 } \\
j_y(\omega,{\bf q};x ) & = & \frac{c}{4\pi } \left [ - { \rm i } q_y \frac{\partial}{\partial x } a_x + \left ( q_z^2 - \frac{\omega^2}{c^2 } - \frac{\partial^2}{\partial x^2 } \right ) a_y - q_y q_z a_z \right ] , \phantom{eee } \label{2.12 } \\
j_z(\omega,{\bf q};x ) & = & \frac{c}{4\pi } \left [ - { \rm i } q_z \frac{\partial}{\partial x } a_x - q_y q_z a_y + \left ( q_y^2 - \frac{\omega^2}{c^2 } - \frac{\partial^2}{\partial x^2 } \right ) a_z \right ] .
\phantom{eee } \label{2.13}\end{aligned}\ ] ]
in the present and subsequent sections , we treat the symmetrized charge - charge density correlation function @xmath62 , where we set @xmath63 for simplicity . this function fulfills the obvious neutrality condition @xmath64 let us consider a ( partial ) dipole moment carried by the charge - charge density correlation function @xmath62 , with the point @xmath33 being constrained to the region @xmath65 : @xmath66 note that , due to the translational invariance of the correlation function in the space perpendicular to the @xmath6 axis , the integration @xmath67 can be equivalently rewritten as @xmath68 or @xmath69 .
interchanging naively the order of integrations over @xmath70 and @xmath6 in ( [ 3.2 ] ) , rewriting @xmath67 as @xmath68 and then applying the neutrality condition ( [ 3.1 ] ) , the quantity @xmath71 seems to vanish .
this is not true . as positive @xmath6 and
@xmath70 become large , @xmath62 tends to the bulk function @xmath72 corresponding to the medium 1 .
this correlation function is not small when the points @xmath33 and @xmath31 are close to one another .
consequently , the function in ( [ 3.2 ] ) is not absolutely integrable which prevents from permuting the integrations over @xmath70 and @xmath6 .
subtracting and adding the bulk correlation function in ( [ 3.2 ] ) leads to @xmath73 \nonumber \\ & & + \int_{-\infty}^{\infty}{\rm d}x ' \int_0^{\infty } { \rm d}x \int { \rm d}{\bf r}\ , x
s^{(1)}_b(t,{\bf r}-{\bf r } ' ) .
\label{3.3}\end{aligned}\ ] ] we assume that the convergence of the charge - charge density correlation function to the bulk function occurs on a microscopic scale , so that @xmath74 negative values of @xmath70 do not represent any complication for @xmath75 since the charge - charge density correlation function is expected to be short ranged along the normal to the interface . under condition (
[ 3.4 ] ) , we can permute the @xmath70 and @xmath6 integrals in the first term on the r.h.s . of ( [ 3.3 ] )
; regarding the neutrality sum rule ( [ 3.1 ] ) , this term becomes equal to 0 . using the translation plus rotation invariance of the bulk correlation function @xmath72 and the method of integration by parts , the second term on the r.h.s .
of ( [ 3.3 ] ) can be easily reexpressed as follows @xmath76 the calculations of this paragraph can be summarized by the equality @xmath77 note that this dipole sum rule depends on the bulk characteristics of the only one from the two media .
we can treat similarly the ( partial ) dipole moment carried by the charge - charge density correlation function @xmath62 , when the point @xmath33 is constrained to the region @xmath78 : @xmath79 the procedure analogous to the one outlined in the previous paragraph results in @xmath80 where @xmath81 is the bulk charge - charge density correlation function corresponding to the medium 2 . combining relations ( [ 3.6 ] ) and ( [ 3.8 ] )
, the total dipole moment reads @xmath82 .
\label{3.9}\end{aligned}\ ] ] we see that the dipole sum rules for an inhomogeneous configuration of two semi - infinite media are expressible in terms of the second moments of the symmetrized charge - charge density correlation function in an infinite medium with the frequency - dependent dielectric function @xmath8 or @xmath9 .
this subject was studied in sect .
3 of the previous paper @xcite .
the final result for the second - moment condition , derived by using the rytov fluctuational theory , reads @xmath83 where the index @xmath84 denotes the medium .
the introduced function @xmath85 fulfills @xmath86 , the equality @xmath87 takes place in the classical limit @xmath88 .
the integral over @xmath16 on the r.h.s . of (
[ 3.10 ] ) is expressible in terms of elementary functions perhaps only for the ( one - component ) jellium model of conductors , i.e. the system of identical particles with the number density @xmath89 , charge @xmath90 and mass @xmath91 , immersed in a neutralizing homogeneous background .
the dielectric function of the jellium is adequately described , in the long - wavelength limit @xmath92 , by the drude formula with the dissipation constant taken as positive infinitesimal @xcite , @xmath93 where the plasma frequency @xmath94 is defined by @xmath95 . inserting the representation ( [ 3.12 ] ) into ( [ 3.10 ] ) and using the weierstrass theorem @xmath96 ( p denotes the cauchy principal value )
, we arrive at @xmath97 in the static @xmath98 case , for all media , using path integration techniques and the general properties of dielectric functions in the complex frequency upper half - plane , the integral over @xmath16 on the r.h.s . of ( [ 3.10 ] ) can be formally expressed as @xcite @xmath99 + \frac{1}{\pi } \sum_{j=1}^{\infty } \left [ \frac{1}{\epsilon({\rm i}\xi_j ) } - 1 \right ] .\ ] ] here , @xmath100 are the ( real ) matsubara frequencies . for the general medium composed of species ( electrons and ions ) @xmath101 with the number density @xmath102 ,
charge @xmath103 and mass @xmath104 , the dielectric function fulfills the asymptotic relation @xcite @xmath105 in the high - temperature ( classical ) limit @xmath106 , each of the matsubara frequencies @xmath107 is much larger than @xmath94 , the corresponding terms in the sum on the r.h.s . of ( [ 3.15 ] ) vanish and so the formula ( [ 3.15 ] ) represents the split of the bulk second - moment condition onto its classical and purely quantum - mechanical parts . we conclude that the dipole sum rules ( [ 3.6 ] ) and ( [ 3.8 ] ) take in the classical limit the following forms @xmath108 in the above derivation of dipole sum rules , only the results of the bulk version of the rytov fluctuational theory were needed at the final stage of the analysis . in what follows
we shall show how rytov s theory can be adopted from the beginning in its inhomogeneous version ; the true value of this approach will be justified later .
let the point @xmath33 be in the region @xmath65 , i.e. @xmath75 , the position of the point @xmath31 is arbitrary . using the formalism of sect . 2 and the explicit results for the retarded green function tensor in the appendix , the fourier transform of the charge - charge density correlation function
is found to be @xmath109 \left ( q^2 + \frac{\partial^2}{\partial x \partial x ' } \right ) \delta(x - x ' ) .\ ] ] here , the delta function has to be understood in a macroscopic sense ( disregarding microscopic structure at small distances ) . taking in ( [ 3.20 ] ) @xmath110 and performing the inverse fourier transform in time , we obtain @xmath111 \nonumber \\ & & \qquad \times \left [ - \frac{\partial^2}{\partial x \partial x ' } \delta(x - x ' ) \right ] .
\label{3.21}\end{aligned}\ ] ] since the integration by parts implies @xmath112 = 1 , \ ] ] we recover the dipole sum rule ( [ 3.6 ] ) with the inserted bulk second - moment condition ( [ 3.10 ] ) . the dipole sum rule ( [ 3.8 ] )
can be verified analogously .
we extend the classical result of @xcite , sect .
ii , to the configuration in fig . 1 , i.e. , the two half - spaces @xmath65 ( @xmath75 ) and @xmath78 ( @xmath113 ) filled with media characterized by static dielectric constants @xmath114 and @xmath115 , respectively .
we recall that @xmath116 for conductors , @xmath117 for vacuum and @xmath118 ( finite ) for dielectrics .
we shall consider the static two - point correlation functions with zero time difference @xmath119 , so the time variables will be omitted in the notation .
for an arbitrary configuration of two points @xmath33 and @xmath31 in the media , we shall compute the correlation function @xmath120 , where @xmath121 is the microscopic electric potential created by the media at point @xmath33 .
it is related to the microscopic charge density @xmath122 by @xmath123 where the integral is over the whole space . in particular
, we shall first calculate microscopically the electric potential due to an infinitesimal charge @xmath124 placed in one of the two media and then complete this calculation with the phenomenological electrostatics result ( the method of images ) for that potential .
let us introduce a test infinitesimal pointlike charge @xmath124 at point @xmath33 .
the microscopic formula for the _ total _ potential @xmath125 induced at point @xmath31 is @xmath126 where @xmath127 means an equilibrium average in the presence of charge @xmath124 .
the additional hamiltonian is @xmath128 .
we now use the linear response theory for @xmath129 which says that @xmath130 where @xmath131 means the standard equilibrium average ( i.e. , in the absence of the test charge @xmath124 ) . combining ( [ 4.2 ] ) and ( [ 4.3 ] ) we arrive at @xmath132 .\ ] ] now , let the point @xmath33 be in region @xmath65 . according to phenomenological electrostatics @xcite , the shift of the potential average due to @xmath124 is in @xmath65 @xmath133 where @xmath134 is the position of the image charge @xmath135 .
we would like to emphasize that the relation ( [ 4.5 ] ) is valid for _ macroscopic _ distances @xmath136 which are much larger than the microscopic scale defined by the particle correlation length
. if the test charge is in region @xmath78 at @xmath33 , the average potential @xmath129 in @xmath78 is given by ( [ 4.5 ] ) with indices 1 and 2 interchanged .
finally , if the test charge is in region @xmath65 at @xmath33 , the average potential in region @xmath78 is given by @xmath137 a similar relation is also valid if the test charge is in @xmath78 for the average potential in region @xmath65 . using ( [ 4.4 ] ) and ( [ 4.5 ] ) , we obtain @xmath138 if @xmath139 , 1 and 2 should be interchanged .
similarly , ( [ 4.4 ] ) and ( [ 4.6 ] ) imply @xmath140 the surface charge @xmath141 on the plane @xmath12 at the point @xmath142 is related to the discontinuity of the normal @xmath6-component of the microscopic electric field @xmath36 on the interface : @xmath143 where the superscript @xmath144 @xmath145 means approaching the surface through the limit @xmath146 @xmath147 .
the surface charge correlation thus is @xmath148 the electric field is related to the potential by @xmath149 , so that @xmath150 using ( [ 4.7 ] ) , we obtain @xmath151 \bigg\vert_{x = x'=0 } .
\label{4.12}\end{aligned}\ ] ] since @xmath152 we find @xmath153 @xmath154 is obtained from ( [ 4.13 ] ) by interchanging 1 and 2 .
finally , @xmath155 using these relations in ( [ 4.10 ] ) gives the classical result @xmath156 where the argument of the prefactor @xmath157 equals to 0 for the considered static case @xmath119 .
we recall that this classical static result is valid for asymptotic distances @xmath158 much larger than any microscopic length scale .
if in @xmath78 there is vacuum ( @xmath159 ) , one retrieves equation ( 20 ) in @xcite . if furthermore in @xmath65 there is a conductor ( @xmath160 ) , one retrieves the old result of @xcite .
the surface charge density @xmath101 has to be understood as being the volume charge density @xmath55 integrated along the @xmath6 axis on some microscopic distance within the interface region . from this point of view
, the formula ( [ 4.15 ] ) implies a sum rule for the volume charge - charge density correlation function . in particular ,
if one assumes an asymptotic behavior of the form @xmath161 @xmath162 obeys , in the classical limit , the sum rule @xmath163 the above formalism can be extended straightforwardly to other geometries of the interface between media , e.g. cylindrically or spherically layered media , or to planarly multi - layered media .
the only modification consists in the application of the corresponding variant of the method of images .
it is instructive to compare the present classical result ( [ 4.15 ] ) , valid for two fluctuating media , with the previous result @xcite valid for a fluctuating medium in contact with the inert wall which `` produces '' the images , but does not fluctuate . for the special case of a coulomb conductor @xmath164 in contact with the fluctuating wall of the static dielectric constant @xmath165 , the prefactor @xmath166 in the formula ( [ 4.15 ] ) takes the form @xmath167 on the other hand , for a coulomb conductor in contact with the inert wall of the static dielectric constant @xmath2 , the prefactor @xmath166 was found to be @xcite @xmath168 we see that the two results ( [ 4.18 ] ) and ( [ 4.19 ] ) coincide with one anther only for the vacuum ( plain hard ) wall ; in vacuum , there are no charges , so that the description by fluctuating and inert walls should lead to the same result . increasing @xmath2 beyond 1 , our formula ( [ 4.18 ] ) predicts the suppression of the surface charge fluctuations while ( [ 4.19 ] ) predicts their enhancement . in the classical limit , there exists a direct relation between the dipole moments ( [ 3.18 ] ) , ( [ 3.19 ] ) and the asymptotic behavior of the surface charge density correlations ( [ 4.15 ] ) .
the aim of the present part is to derive this relation .
let us consider the potential - potential correlation function , given by ( [ 4.7 ] ) or ( [ 4.8 ] ) , when the point @xmath33 is localized at the interface , say @xmath169 , the position of the point @xmath31 is arbitrary : @xmath170 applying the laplacian to both sides of this equation and using the poisson equation @xmath171 , we get @xmath172 with regard to the definition of the microscopic potential ( [ 4.1 ] ) , using in ( [ 4.21 ] ) the ( partial , two - dimensional ) fourier transform of the coulomb potential @xmath173 and the convolution theorem , we get @xmath174 this equation is valid for large distances @xmath175 or , equivalently , small @xmath176 . performing the small-@xmath176 expansion in ( [ 4.23 ] ) and then integrating over all @xmath177 , we arrive at @xmath178 this is the wanted relation . inserting here the relations for the dipole moments ( [ 3.18 ] ) and ( [ 3.19 ] ) , we end up with @xmath179 since , in the sense of distributions , the two - dimensional fourier transform of @xmath180 is @xmath181 , the result ( [ 4.25 ] ) is equivalent to the previous one described by equations ( [ 4.15])([4.17 ] ) .
the long - range decay of the quantum surface charge density correlation functions , in both retarded and nonretarded regimes , was the subject of refs .
@xcite . by using the rytov formalism for a plane between two media ,
the two - point electric field correlations were derived for any point positions in medium 1 and 2 and the discontinuity of the electric field across the interface was related to the surface charge density .
the consequent integrals over the frequency were treated using path integration techniques and the general properties of dielectric functions in the complex frequency upper half - plane .
in the static @xmath119 case , the final result for the fourier transform of the quantum surface charge density correlation function reads @xmath182 where the explicit form of the ( static ) function @xmath183 depends on the considered , retarded or nonretarded , regime . in the retarded regime
, we have @xmath184 ^ 2 } { \epsilon_1({\rm i}\xi_j ) \epsilon_2({\rm i}\xi_j ) } , \ ] ] where the matsubara frequencies @xmath185 @xmath186 are defined in ( [ 3.16 ] ) and the inverse lengths @xmath187 for the regions @xmath188 in ( [ a.1 ] ) . for the purely imaginary values of the frequencies @xmath189 , the values of the dielectric functions @xmath190 , and consequently of the inverse lengths @xmath191 , are real positive . in the considered limit @xmath92 , @xmath192 .
since @xmath193 and , according to ( [ 3.17 ] ) , @xmath194 , the sum in ( [ 4.27 ] ) converges .
this means that the function @xmath195 , being of the order @xmath196 , becomes negligible in comparison with the first term in ( [ 4.26 ] ) in the limit @xmath92 .
the prefactor associated with the asymptotic decay ( [ 4.15 ] ) thus reads @xmath197 this expression , which does not depend on the temperature and the planck constant , coincides with the classical result ( [ 4.15 ] ) . in other words ,
the consideration of retardation effects makes the quantum mechanics equivalent to its classical limit .
the situation is fundamentally different in the nonretarded regime . in the limit @xmath0
, it holds @xmath198 .
we thus get from the retarded representation ( [ 4.27 ] ) that @xmath199 .\ ] ] with regard to the asymptotic behavior @xmath194 , the sum in ( [ 4.29 ] ) converges , the nonretarded function @xmath200 is of the order @xmath201 and therefore contributes to the surface charge density correlation ( [ 4.26 ] ) .
the prefactor @xmath202 is a complicated function of temperature .
it was shown in @xcite that for distances @xmath203 much smaller than @xmath204 the retardation effects are negligible and so the nonretarded result ( [ 4.29 ] ) takes place , while for @xmath205 the retardation results describe adequately the decay of the surface charge density correlations . in the retarded regime , the time difference between points has no effect on the form of the asymptotic behavior ( [ 4.15 ] ) , i. e. @xmath206
in this section , we shall deal with the charge - current density correlation functions @xmath207 , where the component index @xmath32 equals @xmath208 , @xmath209 or @xmath210 .
we recall that for the bulk medium with the dielectric function @xmath211 , these correlations were shown to satisfy the following sum rules @xcite @xmath212 the static @xmath98 version of the sum rule ( [ 5.2 ] ) is trivial for any medium : since @xmath213 is an odd function of @xmath16 , the r.h.s . of ( [ 5.2 ] ) vanishes . in the special case of the jellium model with the dielectric function ( [ 3.12 ] ) ,
the weierstrass theorem ( [ 3.13 ] ) permits us to express explicitly the time - dependent sum rule ( [ 5.2 ] ) as follows @xmath214 we now consider the inhomogeneous situation pictured in fig 1 .
let the point @xmath31 be in the region @xmath65 @xmath215 , the position of the point @xmath33 is arbitrary .
we start with the inhomogeneous rytov theory ( see sect . 2 and the appendix ) , whose results for the charge - current density correlation function in the fourier space , up to terms linear in @xmath216 and @xmath217 , can be summarized as follows @xmath218 \frac{\partial}{\partial x } \delta(x - x ' ) + o(q_y^2,q_z^2,q_yq_z ) , \label{5.4 } \\
\beta \langle \rho(x ) j_y(x ' ) \rangle^s_{\omega,{\bf q } } & = & - \frac{g(\omega)}{2\pi } { \rm i m } \left [ \frac{1}{\epsilon_1(\omega ) } \right ] q_y \delta(x - x ' ) + o(q_y^2,q_z^2,q_yq_z ) , \label{5.5 } \\
\beta \langle \rho(x ) j_z(x ' ) \rangle^s_{\omega,{\bf q } } & = & - \frac{g(\omega)}{2\pi } { \rm i m } \left [ \frac{1}{\epsilon_1(\omega ) } \right ] q_z \delta(x - x ' ) + o(q_y^2,q_z^2,q_yq_z ) .
\label{5.6}\end{aligned}\ ] ] taking @xmath110 in equations ( [ 5.4])([5.6 ] ) and regarding that @xmath219 we obtain in the leading order @xmath220 this is the analog of the bulk sum rule ( [ 5.1 ] ) which holds also for the point @xmath31 being situated inside the region @xmath78 . with respect to the equality ( obtained with the aid of the integration by parts ) @xmath221 for ( [ 5.4 ] ) and in the next order in @xmath176 of @xmath222 for the relations ( [ 5.5 ] ) , ( [ 5.6 ] ) , we find @xmath223 this is the analog of the bulk sum rule ( [ 5.2 ] ) for the point @xmath224 . when @xmath225 , an analogous sum rule is obtained by substituting in ( [ 5.10 ] ) @xmath8 by @xmath9 .
there exist another sum rules for the inhomogeneous situation which have no obvious counterpart in the bulk case .
these sum rules follow from the application of the equalities @xmath226 to the formula ( [ 5.4 ] ) with @xmath227 .
namely , we have @xmath228 and , similarly , @xmath229 these relations can be verified independently by using the method for the dipole sum rules developed in sect .
we first subtract from and add to the correlation function @xmath207 on the l.h.s . of equations ( [ 5.13 ] ) and
( [ 5.14 ] ) its bulk counterparts , corresponding to medium 1 if @xmath230 and to medium 2 if @xmath231 . as before , assuming that @xmath232 and , similarly , @xmath233 the permutation of the @xmath6 and @xmath70 integrations nullifies the contribution of the correlation function minus its bulk counterpart due to the sum rule ( [ 5.8 ] ) . using the translational plus rotational invariance of the bulk correlation function in the nonzero term ,
we obtain @xmath234 and , similarly , @xmath235 in view of these relations , the inhomogeneous sum rules ( [ 5.13 ] ) and ( [ 5.14 ] ) are in fact the consequences of the bulk sum rule ( [ 5.2 ] ) for media 1 and 2 , respectively .
as concerns the current - current density correlations @xmath236 @xmath237 , for the bulk medium with the dielectric function @xmath211 , they satisfy the sum rule @xcite @xmath238 in the case of the jellium model with the dielectric function ( [ 3.12 ] ) , the weierstrass theorem ( [ 3.13 ] ) implies @xmath239 in the static @xmath98 case , the formula ( [ 6.1 ] ) can be formally expressed as @xcite @xmath240 \right\ } , \ ] ] where @xmath185 are the matsubara frequencies defined by ( [ 3.16 ] ) . the first term on the r.h.s . of ( [ 6.3 ] )
represents the classical @xmath106 limit , the second term is the purely quantum - mechanical contribution to the sum rule . for the studied configuration in fig .
1 , the inhomogeneous version of the rytov method gives in the limit @xmath241 , for distinct current indices , @xmath242 for any positions of points @xmath33 and @xmath31 in media 1 and 2 .
the relation ( [ 6.4 ] ) is equivalent to @xmath243 where the position of the point @xmath31 in media 1 or 2 is irrelevant .
this is the generalization of the bulk sum rule ( [ 6.1 ] ) for @xmath244 .
let the point @xmath31 be localized in the region @xmath65 , i.e. @xmath230 , the position of point @xmath33 is arbitrary . for the @xmath92 limit of the diagonal correlation function of the @xmath245 current components , the rytov theory implies @xmath246 \delta(x - x ' ) .\ ] ] integrating over @xmath6 , this equation gives @xmath247 a similar expression can be derived when @xmath225 .
the inhomogeneous sum rules obtained up to now are quite trivial generalizations of the corresponding bulk sum rule ( [ 6.1 ] ) .
this is no longer true for the diagonal correlation functions of the yy and zz current components . for @xmath224
, the inhomogeneous rytov method implies @xmath248 , \label{6.8}\end{aligned}\ ] ] where the additional position - dependent function @xmath249 , which does not exist in the bulk case , reads @xmath250 \frac{n_1(\omega)}{n_1(\omega)\epsilon_2(\omega)+n_2(\omega)\epsilon_1(\omega ) } \frac{\epsilon_1(\omega)-\epsilon_2(\omega)}{\epsilon_1(\omega ) } \nonumber \\ & & \times \exp\left ( -\frac{\vert\omega\vert}{c } n_1(\omega ) x'\right ) \label{6.9}\end{aligned}\ ] ] with @xmath251 @xmath252 defined by @xmath253 the function @xmath249 is equal to zero in three cases : the homogeneous case @xmath254 , the medium 1 is the trivial vacuum @xmath255 and far away from the boundary @xmath256 . since @xmath257 , the function @xmath249 does not contribute to the classical limit of ( [ 6.8 ] ) , but it does contribute to quantum - mechanical corrections .
a similar expression can be derived when @xmath225 .
we conclude this section by noting that , according to the inhomogeneous rytov theory , the interface between two media breaks up the directional invariance of the diagonal current - current correlations in the bulk . while the sum rule for the xx correlations ( [ 6.7 ] ) has the form of the bulk one ( [ 6.1 ] ) , the yy and zz correlations ( [ 6.8 ] ) exhibit an additional dependence on the distance from the interface .
we believe that this is not an artificial anomaly of the applied method , but the true phenomenon occurring in the current - current correlations functions .
in this paper , we applied the rytov fluctuational electrodynamics to the inhomogeneous geometry in fig . 1 to derive a sequence of sum rules for the charge - charge , charge - current and current - current density correlation functions .
the validity of some of these sum rules was controlled independently by using methods developed previously in the context of the model of a fluctuating semi - infinite conductor in contact with an inert wall . in the realistic model considered here , both semi - infinite media in contact
fluctuate . comparing the classical static results ( [ 4.18 ] ) and
( [ 4.19 ] ) for the fluctuating and inert walls , respectively , we see that they coincide , as it should be , in the vacuum case @xmath258 , but for @xmath259 these results are fundamentally different .
some of the inhomogeneous sum rules represent a straightforward generalization of their bulk counterparts .
this is not the case of the current - current density correlation functions ; the sum rules ( [ 6.7 ] ) and ( [ 6.8 ] ) indicate a breaking of the directional invariance of the diagonal current - current density correlations by the interface .
is grateful to lpt for very kind hospitality .
the support received from the european science foundation ( `` methods of integrable systems , geometry , applied mathematics '' ) , grant vega no .
2/0113/2009 and ce - sas qute is acknowledged .
in this appendix , we present explicit forms of the retarded green function tensor elements @xmath38 for the two semi - infinite media geometry pictured in fig
the half spaces @xmath65 @xmath260 and @xmath78 @xmath261 are characterized , besides the dielectric functions @xmath262 @xmath252 , by the inverse lengths @xmath187 @xmath252 defined as follows @xmath263 here , from the two possible solutions for each @xmath264 we choose the one with the positive real part in order to ensure the regularity of tensor elements @xmath38 at asymptotically large distances from the interface @xmath265 . for simplification reasons , we shall omit in the notation the dependence of functions on the frequency @xmath16 and the wave number @xmath176 .
\i ) if the two points @xmath266 are localized in the same half - space , say @xmath267 ( i.e. @xmath268 ) , we introduce a pair of functions @xmath269 .
\label{a.2 } \\
v(x , x ' ) & = & \frac{2\pi\hbar c^2}{\omega^2\epsilon_1\kappa_1 } \left [ { \rm e}^{-\kappa_1\vert x - x'\vert } - \frac{\epsilon_1\kappa_2-\epsilon_2\kappa_1}{\epsilon_1\kappa_2 + \epsilon_2\kappa_1 } { \rm e}^{-\kappa_1(x+x ' ) } \right ] .
\label{a.3}\end{aligned}\ ] ] these functions satisfy the same type of the differential equation @xmath270 and are related by @xmath271 the third function we shall need is defined by @xmath272 in terms of the introduced functions , the elements of the retarded green function tensor are given by @xmath273 @xmath274 the remaining @xmath275 and @xmath276 components are given by the replacement rule @xmath277 , @xmath278 u(x , x ' ) + q_z^2 w(x , x ' ) , \ ] ] @xmath279 and @xmath280 .\ ] ] \ii ) if the two points @xmath266 are localized in the different half - spaces , say @xmath224 and @xmath281 ( i.e. @xmath230 and @xmath113 ) , we introduce the function @xmath282 in terms of this function , the elements of the retarded green function tensor are given by @xmath283 @xmath284 @xmath285 , @xmath286 s(x , x ' ) , \ ] ] @xmath279 and @xmath287 | this paper is a continuation of the previous study [ amaj , l. : j. stat .
phys . * 137 * , 1 - 17 ( 2009 ) ] , where a sequence of sum rules for the equilibrium charge and current density correlation functions in an infinite ( bulk ) quantum media coupled to the radiation was derived by using rytov s fluctuational electrodynamics . here
, we extend the previous results to inhomogeneous situations , in particular to the three - dimensional interface geometry of two joint semi - infinite media .
the sum rules derived for the charge - charge density correlations represent a generalization of the previous ones , related to the interface dipole moment and to the long - ranged tail of the surface charge density correlation function along the interface of a conductor in contact with an inert ( not fluctuating ) dielectric wall , to two fluctuating semi - infinite media of any kind .
the charge - current and current - current sum rules obtained here are , to our knowledge , new .
the current - current sum rules indicate a breaking of the directional invariance of the diagonal current - current correlations by the interface .
the sum rules are expressed explicitly in the classical high - temperature limit ( the static case ) and for the jellium model ( the time - dependent case ) . |
as is well known , event horizons , whether black holes or cosmologicals , mimic black bodies and possess a nonvanishing temperature and entropy , the latter obeying the bekenstein - hawking formula , @xcite @xmath0 this expression , in which @xmath1 stands for the boltzmann constant , @xmath2 the area of the horizon , and @xmath3 the planck s length , points to a deep connection between gravitation , thermodynamics , and quantum mechanics , still far from being fully unveiled though progress in that direction are being made -see , e.g. , @xcite and references therein .
recently it was demonstrated that cosmological apparent horizons are also endowed with thermodynamical properties , formally identical to those of event horizons @xcite .
the connection between gravity and thermodynamics was reinforced by jacobson , who associated einstein equations with clausius relation @xcite , and later on by padmanabhan who linked the macroscopic description of spacetime , given by einstein equations , to microscopic degrees of freedom , @xmath4 , through the principle of equipartition of energy , i.e. , @xmath5 in particular , padmanabhan , starting from the field equations arrived to the equipartition law @xcite and indicated how to obtain the field equations of any diffeormorphism invariant theory of gravity from an entropy extremising principle @xcite , the entropy of spacetime being proportional to @xmath4 .
+ on the other hand , quantum corrections to the semi - classical entropy - law ( [ entropy ] ) have been introduced in recent years , namely , logarithmic and power - law corrections .
logarithmic corrections , arises from loop quantum gravity due to thermal equilibrium fluctuations and quantum fluctuations @xcite , @xmath6 \ , .\ ] ] on its part , power - law corrections appear in dealing with the entanglement of quantum fields in and out the horizon @xcite , @xmath7 \ , .\ ] ] in the last two expressions , @xmath8 denotes a dimensionless parameter whose value ( in both cases ) is currently under debate .
+ for the connection between horizons and thermodynamics to hold , these quantum entropy corrections must translate into modifications of the field equations of gravity , see e.g. @xcite . in any sensible cosmological context
these modifications must fulfill the generalized second law ( gsl ) of thermodynamics .
the latter asserts that the entropy of the horizon plus the entropy of its surroundings must not decrease in time .
as demonstrated by bekenstein , this law is satisfied by black holes in contact with their radiation @xcite . the aim of this paper is to see whether the modified friedmann equations coming from logarithmic corrections and from power - law corrections , in conjunction with clausisus relation or the equipartition principle , are compatible with the generalized second law
this will set constraints on the parameter @xmath8 introduced above , whose value is theory dependent and rather uncertain .
we hope this may be of help in discriminating among quantum corrections , via a purely classical analysis .
+ the plan of this work is as follows : section [ modified friedmann equations ] derives friedmann equations from different entropy corrections , considering a friedmann - robertson - walker ( frw ) metric sourced by a perfect fluid .
section [ generalized second law of thermodynamics ] considers the entropy rate enclosed by the apparent horizon and in which the fluid is in thermal equilibrium with it .
section [ gsl with phantom fluid or particle production ] e xtends the analysis by allowing the fluid to be phantom , for a period , or by allowing the production of particles .
finally , section [ conclusions ] summarizes and discusses our findings .
in this section we recall the derivation of the modified friedmann equations in case of the classes of entropy corrections to the bekenstein - hawking entopy in eq.([entropy ] ) .
as said above , this semiclassical relation gets modified when quantum corrections are taken into account .
+ logarithmic corrections lead to the expansion @xcite @xmath9,\ ] ] while power - law corrections yield @xcite @xmath10,\ ] ] for the horizon entropy . in ( [ power ] ) @xmath11 is a parameter that depends on the power @xmath8 of the entropy correction .
+ we wish to examine thermodynamical behavior of the system consisting in the apparent horizon of a spatially flat frw universe and the fluid within it .
+ the frw metric can be written as @xmath12 where @xmath13 and @xmath14 .
the apparent horizon is defined by the condition @xmath15 so that its radius turns out to be @xmath16 where @xmath17 denotes the hubble function . from expressions ( [ log ] ) and ( [ power ] ) different cosmological scenarios can be considered , depending on whether use is made of clausius relation @xcite @xmath18 or the principle of equipartition of the energy , @xmath19 @xcite . in either case , the left hand side represents the amount of energy that crosses the apparent horizon within a time interval @xmath20 in which the apparent horizon evolves from @xmath21 to @xmath22 @xmath23 here @xmath24 is the energy - momentum tensor of the perfect fluid , and @xmath25 is the approximate generator of the horizon , @xmath26 .
it follows that @xmath27 the change of the area of the apparent horizon , @xmath28 , induces the entropy shift , @xmath29 the number of degrees of freedom is assumed proportional to the entropy whereby it also changes , and the same holds true for the temperature of the system , that we take as the temperature of the horizon @xcite @xmath30 by using eq.([energy ] ) with either clausius relation ( [ clausius ] ) or the equipartition principle ( [ equipartition ] ) , one gets the modified friedmann equations , @xmath31&=&\frac{8\pi g}{3 } \rho,\label{h}\\ \dot{h}\left[1+f(\alpha , h)\right]&=&-4\pi g\left(\rho+\frac{p}{c^2}\right)\label{hdot},\end{aligned}\ ] ] the explicit expressions of @xmath32 and @xmath33 depend on both the entropy corrections and the thermodynamical relation employed , as shown in table [ table1 ] .
it should be noted that , eq.([h ] ) can be recovered from eq.([hdot ] ) and the evolution equation for the perfect fluid : @xmath34 .expressions for @xmath35 and @xmath36 .
[ cols="^,^",options="header " , ] [ table1 ] in the case of the logarithmic correction the two approaches differ by a logarithmic term that comes from the @xmath37 contribution .
moreover , contrary to ref .
@xcite , we believe it should not be neglected because @xmath38 is larger than zero when the area of the apparent horizon is of order of the planck area . + on its part , power - law entropy correction gives the same correction to the friedmann equations , the only difference lying in the value of the coefficients .
the coefficient @xmath39 , there , is related to the dimensionless parameter @xmath8 by @xmath40 . as can be noted
, the @xmath8 parameter directly comes from quantum corrections to the entropy and it consequently affects cosmological scenarios .
its value depends on the details of the quantum calculations , and for the time being there is not agreement on it .
the following analysis determines in which intervals this parameter results compatible with the gsl .
equipped with the entropy expressions ( [ log ] ) and ( [ power ] ) , we set out to study whether the gsl is satisfied when the modified friedmann equations ( [ h ] ) and ( [ hdot ] ) are employed .
since the entropy depends on the area of the apparent horizon , @xmath41 , it varies as @xmath42 using eq.([hdot ] ) , it can be cast in terms of the hubble parameter and the energy density and pressure of the fluid that fills the universe @xmath43 where @xmath44 and @xmath45 depends on the entropy corrections and the thermodynamic relation used to derive friedmann equations , namely @xmath46 for logarithmic entropy corrections , and @xmath47 for power - law corrections .
+ for the sake of clarity in what follows we split the analysis for the two classes of entropy corrections , but in this section we will only consider perfect fluids assuming that the dominant energy condition ( dec ) holds true ( i.e. , @xmath48 ) all along the expansion .
then , in view of eq.([entropyrate ] ) the gsl is satisfied provided @xmath45 is non - negative which occurs only for some values of the parameter @xmath8 . in all the cases the results of general relativity @xcite
are recovered , as it should , by setting @xmath49 . for convenience ,
we introduce the dimensionless variable @xmath50 so that @xmath51 at the quantum regime and , provided @xmath52 , it decreases as time goes on . in terms of this new variable and using the equipartition theorem , eq.([equipartition ] ) ,
we have that @xmath53 we require @xmath54 to be non - negative , if the gsl is to hold , as well as @xmath55 , not to deal with ghosts . from friedmann s equation ( [ h ] ) with @xmath33 given for the logarithmic correction of entropy , it translates into the condition @xmath56 fig.[fig : inequality1 ] depicts the regions in the plane @xmath57 where @xmath58 is fulfilled and those in which @xmath59 .
we believe that the allowed values for @xmath8 are just those such that the gsl holds throughout the expansion of the universe .
the upper bound on @xmath8 is given by the local minimum of the dashed curve , that is @xmath60 , and the lower bound by the intersection of the dotted curve with the line @xmath61 ( i.e. , when horizon area equals planck s area ) , that is @xmath62 ( not shown ) .
depending on @xmath8 along the universe expansion , in the case of the logarithmic correction and the equipartition principle . here , @xmath63 .
the plot focuses on the range @xmath64 but in the remaining region the curves behave monotonically .
bear in mind that the smaller @xmath65 , the older the universe is.,width=302 ] positive values of @xmath8 seem to be largely favored , which is also consistent with some quantum calculations of the entropy corrections @xcite . + by using clausius relation ( [ clausius ] ) , instead , and adopting the above defined variable @xmath65 we get @xmath66 as seen in fig.[fig : inequality2 ] , the gsl is satisfied , @xmath67 , for @xmath68 . + but with the equipartition principle replaced by clausisus relation , width=302 ] .
[ fig : inequality2 ] before applying the gsl let us look at the friedmann equations for power - law entropy correction .
inspection of eq.([power ] ) shows that the values of @xmath69 and @xmath70 are special , in the sense that for @xmath49 there are no entropy corrections and the equations reduce to the corresponding general relativity expressions with a cosmological constant .
likewise @xmath70 represents just a renormalisation of newton constant , @xmath71 . bearing this in mind
, we start the analysis by introducing the dimensionless variable @xmath72 , and identifying the crossover scale @xmath39 with @xmath73 , as in @xcite .
thus @xmath65 tends to zero in the far past and its today value is @xmath74 ( provided , again , that @xmath52 ) .
+ by virtue of clausius relation ( [ clausius ] ) , the positive energy condition @xmath75 implies @xmath76 likewise , the function @xmath77 in eq.([entropyrate ] ) adopts the form @xmath78 and it is non - negative in the ranges @xmath79 as seen in fig.[fig : inequality3 ] .
the latter shows that the gsl remains valid throughout the expansion so long as @xmath80 .
depending on @xmath8 along the universe expansion in the case of power - law correction and clausius relation . here
@xmath81 . , width=264 ] .
[ fig : inequality3 ] when use is made of the equipartition relation one is led to the same functional modification of the friedmann equations , @xmath82 , but with different coefficients , which depend on @xmath8 . in this case
, @xmath55 implies @xmath83 and @xmath84 now , @xmath77 is non - negative in the following ranges @xmath85 as can be easily learnt from fig.[fig : inequality4 ] , this approach allows a wider range for the parameter @xmath8 which can now lie in the range @xmath86 , as well . but with clausius relation replaced by the equipartition principle.,width=264 ]
this section investigates whether by relaxing the dominant energy condition ( dec ) or by allowing particle production , the range of values of @xmath8 compatible with the gsl gets wider .
the above analysis shows that this may be the case but there are regions where the gsl is still violated so long as ghost fields are excluded . in figs .
[ fig : inequality1][fig : inequality4 ] these regions are explicitly marked .
we begin with the logarithmic corrections to the horizon entropy .
first we note that an equivalent rewriting of the entropy rate in eq.([entropyrate ] ) , for both eq.([flog1 ] ) and eq.([flog2 ] ) , is @xmath87 then , for non - superaccelerated expansion @xmath88 , the gsl will be fulfilled if @xmath89 .
this means that we the line @xmath62 should not be crossed if a meaningful description of the expansion is required .
+ nevertheless , we can focus on the positive range of the @xmath8 parameter in case of equipartition of energy that , as inspection of fig.[fig : inequality1 ] reveals , can be enlarged from the minimum of the dashed curve , @xmath90 , to the minimum of the dot - dashed curve , @xmath91 , by allowing the fluid to be phantom during the evolution .
these positive values of @xmath8 may be seem too big , but they are not inconsistent with quantum calculations @xcite .
+ let us note that , at present , @xmath92 .
this means that the phantom phase lies far in the past .
could it give rise to a `` reasonable '' inflation ?
the first thing to check is whether there are accelerated phases in this modified theory of gravity ( because of quantum corrections proportional to @xmath8 ) . introducing the parameter @xmath93 appearing in the equation of state , @xmath94 , in the phantom region , when @xmath95 and @xmath96 , positive values of @xmath8 are compatible with acceleration @xmath97 provided that the inequality @xmath98 is fulfilled .
then , an inflationary period can be obtained for @xmath99 ; the latter bound corresponds to the minimum of the dot - dashed curve in fig.[fig : inequality1 ] .
+ for this early inflation to be successful in solving the problems of the standard big - bang model it must yield a sufficient number of e - folds , @xmath100 . bearing in mind eqs.([h ] ) and ( [ hdot ] ) we get @xmath101 the question now reduces to finding an appropriate expression for @xmath102 such that the field can be phantom for a period that gives a suitable number of e - folds , with the initial and final values of the phantom period being the intersection of the line of a given @xmath8 with the appropriate curve in fig.[fig : inequality1 ] .
+ in the case of logarithmic entropy correction and equipartition of energy , eq.([e - folds ] ) can be analytically integrated to give a constant value of @xmath93 .
for example , for @xmath103 , one has , @xmath104 although this is just a rough estimate it makes clear that , given an evolution for the equation of state parameter , it suffices that it slightly crosses the phantom divide - line to get a convenient amount of inflation . in the case of power - law entropy corrections , allowing for a phantom field may enlarge the range of the @xmath8 parameter towards negative values , but a deeper analysis of the evolution equation shows that the requirement that the gsl is fulfilled forces the hubble parameter and the dimensionless variable we have defined , not to cross the curve @xmath105 of fig.[fig : inequality4 ] . in fact , the entropy rate can be written as @xmath106 \dot{x } x,\ ] ] hence in the region above that curve @xmath107 must be negative , and in the region below , positive ( @xmath52 and @xmath108 ) .
+ actually this depends on the choice of the crossover scale @xmath39 that we have identified as @xmath73 .
nevertheless , if this also depends on the parameter @xmath8 , through a factor , the whole negative region can correspond to a meaningful expansion , with @xmath52 and @xmath48 ; while all non accessible regions shifting to the positive sector , with a wider range in case the clausius relation to be used .
+ thus , the results strictly depend on the explicit choice of the crossover scale , once assumed its order of magnitude is about @xmath73 ; but it seems that the negative region is accessible , provided a redefinition of the scale is made , while the positive part remains largely determined by the power - law dependence .
as is well known , on a phenomenological level particle production can be described in terms of an effective bulk viscosity @xmath109 , with @xmath110 the coefficient of bulk viscosity @xcite .
thus the total pressure is @xmath111 where @xmath94 , denotes the equilibrium pressure . in the case of a isentropic particle production
there exists a general relation between the viscous coefficient and the particle production rate @xcite . in this effective picture ,
in which particles are accessible to a perfect fluid description soon after their creation , eqs .
( [ hdot ] ) and ( [ conservation ] ) stay as they are only that @xmath112 is now given by eq.([viscousp ] ) .
likewise , the entropy rate acquires a new term entirely due to the increase in the number of particles : @xmath113 where @xmath114 is the @xmath115-spatial volume enclosed by the horizon , @xmath116 . as before ,
the fluid is assumed in thermal equilibrium with the horizon ( see eq.([temperature ] ) ) .
+ by using eqs.([entropyrate ] ) and ( [ fluidrate ] ) , the total entropy rate is @xmath117.\ ] ] the first term , with particle creation , can allow some region to be reached , in the sense that gsl can hold with a non - phantom fluid .
inspection of figs.[fig : inequality1 ] and [ fig : inequality2 ] shows that in case of logarithmic entropy corrections , the @xmath8 negative range from @xmath118 to @xmath119 could be accessible .
in fact , in this region the field is non - ghost and it was discarded not to have a superaccelerating evolution so long as @xmath120 .
moreover , @xmath121 often appears in the literature @xcite now , because of effective viscosity , a positive term has been included in the entropy rate .
+ so , gsl will hold if @xmath122 in arriving to this expression we have made use of @xmath123 in terms of the dimensionless variable @xmath65 , eq.([particleineq ] ) can be written as @xmath124 where @xmath125 .
+ it follows that particle production allows to enlarge the @xmath8 range from @xmath126 in the case of clausius relation , and from @xmath127 in case of equipartition of energy .
this is depicted in the @xmath115-dimensional plot of fig.[fig : inequalityviscosity ] that refers to the latter , and in which a constant value of @xmath8 can be maintained all along the expansion , provided a certain amount of particle production is permitted ( @xmath128 ) . ) in terms of the pair of parameters ( i.e. @xmath129 ) and the dimensionless @xmath65 variable.,width=302 ]
in this paper we investigated the constraints imposed by the gsl on modified friedmann equations that arise from quantum corrections to the entropy - area relation , eq .
( [ entropy ] ) .
as is well known , the gsl is a powerful tool to set bounds on astrophysical and cosmological models -see
e.g. @xcite . cosmological equations follow either from jacobson s approach , that connects gravity to thermodynamics by associating einstein equations to clausius relation ( [ clausius ] ) , or padmanabhan s suggestion that relates gravity ( i.e. , einstein equations ) to microscopic degrees of freedom through the principle of equipartition of energy ( [ equipartition ] ) .
we analyzed two entropy - area terms , logarithmic ( [ log ] ) , and power - law corrections ( [ power ] ) , the former coming from loop quantum gravity , the latter from the entanglement of quantum fields .
both quantum corrections have been widely investigated but , since they come from very different techniques , one should not be surprised that total agreement on these corrections is still missing . in particular , there is a lack of consensus on the value of the constant parameter @xmath8 .
our work aimed to discriminate among quantum corrections by requiring , via a classical analysis , the gsl to be fulfilled throughout the evolution of the universe .
this sets constraints on the value of the parameter .
we first investigated the intervals of values of @xmath8 compatible with the gsl by assuming that the dec holds true for the perfect fluid that sources the gravitational field of the frw universe . in the case of logarithmic corrections to the horizon
entropy this gives a wide range , in which positive values are largely favored , with no upper bound in the case of the modified friedmann equation derived from clausius relation .
negative values of @xmath8 are consistent with the gsl only up to @xmath62 , hence discarding two negative values that have been suggested in the literature , namely , @xmath121 @xcite , and @xmath130 @xcite . + in the case of power - law entropy corrections ,
our analysis favors positive values of @xmath8 , though they are expected to be negative .
only a small concordance range appears when modified friedmann equations from the equipartition of energy are employed .
specifically , the interval @xmath131 corresponds to a power - law correction with an index between @xmath119 and 0 , that has been analytically or numerically obtained @xcite . the second part of our analysis considered the possibility of enlarging the allowed interval of @xmath8 in the case of logarithmic entropy corrections by considering either a phantom phase for the fluid or particle production modelled as an effective bulk viscosity .
phantom behavior affects positive values of @xmath8 while particle production could enlarge the negative range and incorporate the value @xmath121 that seems widely accepted in literature @xcite . | we apply the generalized second law of thermodynamics to discriminate among quantum corrections ( whether logarithmic or power - law ) to the entropy of the apparent horizon in spatially friedmann - robertson - walker universes .
we use the corresponding modified friedmann equations along with either clausius relation or the principle of equipartition of the energy to set limits on the value of a characteristic parameter entering the said corrections . |
the analysis of the epic data from a 30 ks observation of the seyfert 1 mrk 335 ( longinotti et al . 2006 ) shows that the 2 - 10 kev spectrum can be phenomenologically fitted with a power law , a broad gaussian line with [email protected] kev , @[email protected] kev , ew 490 ev , and a narrow ( @xmath2 = 1ev ) absorption line with [email protected] kev and ew=50 ev .
these features are visible in the residuals of the spectrum plotted in fig.1 .
the significance of the absorption line estimated through monte carlo simulations is 99.7% .
the most obvious identification for the absorption feature is with redshifted iron k@xmath0 resonance absorption .
the identification with iron is favoured since the observed energy of the line is too high to be readily explained by k@xmath0 absorption in any of the other astrophysically abundant elements .
we present a simple model for the inflow ( centre and right panels of fig.1 ) , accounting approximately for relativistic and radiation pressure effects , and use monte carlo methods to compute synthetic spectra for qualitative comparison with the data .
this modeling was developed following sim ( 2005 ) and assuming spherically symmetric radially infalling gas .
it does show that the absorption feature can plausibly be reproduced by infalling gas providing that the feature is identified with fe xxvi .
a smooth continuous flow is ruled out by the poor agreement with the data : although the presence of a broad inverse p cygni fe xxvi k@xmath0 line profile is predicted , the absorption line is insufficiently redshifted and too broad .
an inflow over a limited range of radii ( as discrete blobs or section of infalling gas ) is more consistent with the data ( fig.1 ) .
the narrowness of the absorption line tends to argue against a purely gravitational origin for the redshift of the line , but given the current data quality we stress that such an interpretation can not be ruled out .
a longer ( 100 ks ) xmm - newton observation of mrk 335 has been performed this year .
the analysis of the integrated spectrum reveals a double - peaked fe line .
a broad accretion disc line is likely to be present but the peaks are due to narrower components at 6.4 and 7 kev ( oneill et al . in prep . ) .
a preliminary time - resolved analysis shows that the narrow peaks are variable .
2 shows the spectra from three portions of the light curve .
the narrow peak at 6.4 kev , clearly present in the first 30 ks , disappears in the central portion .
the 7 kev peak instead shows up only in the last 60 ks .
the variable narrow lines could be tentatively associated to the presence of flares in the light curve , but a much more detailed and careful analysis is necessary before speculating on their origin . as regard to the absorption line observed in the archival 30 ks observation ,
it is marginally detected only in the first portion of the longer exposure .
further investigation on these data will hopefully clarify all the issues reported here . | the analysis of hard x - ray features in _ xmm - newton _ data of the bright sy 1 galaxy mrk 335 is reported here .
the presence of a broad , ionised iron k@xmath0 emission line in the spectrum , first found by gondoin et al.(2002 ) , is confirmed .
the broad line can be modeled successfully by relativistic accretion disc reflection models .
regardless of the underlying continuum we report , for the first time in this source , the detection of a narrow absorption feature at the rest frame energy of 5.9 kev .
if the feature is identified with a resonance absorption line of iron in a highly ionised medium , the redshift of the line corresponds to an inflow velocity of 0.11 - 0.15 c. preliminary results from a longer ( 100ks ) exposure are also presented . |
flares with duration of a few to tens of minutes and energies of @xmath2 ergs have been observed on low mass dwarf stars for many decades ( eg bopp & moffett 1973 , gershberg & shakhovskaia 1983 ) .
the origin of these flares is thought to be similar to solar flares in that they are produced during magnetic reconnection events ( eg haisch , strong & rodono 1991 ) .
studying stellar flares from a wide range of stars can give important insight to how magnetic activity varies as a function of stellar mass and age .
in more recent years , the affects of flares on the atmosphere of exo - planets around dwarf stars has been the subject of much interest ( eg segura et al 2010 ) .
historically the study of stellar flares was performed on known m dwarf stars . however , with the advent of large scale surveys such as sdss it has become possible to identify events from many previously unknown flare stars ( eg davenport et al 2012 ) . whilst this will no doubt prove a goldmine for stellar flare researchers , the issue of separating extra - galactic transient events and flares from m dwarfs will become increasingly difficult in future surveys such as that made using lsst .
one survey which allows the virtually uninterrupted observation of sources is nasa s _ kepler _ mission which covers an area of 116 square degrees .
the light curves extend over many months ( or years ) and have a precision of parts per million and allows models of stellar structure to be tested in a way not previously possible .
a key point is that the actual targets which are observed using _
kepler _ can be updated every month .
walkowicz et al ( 2011 ) presented _ kepler _ observations of flares seen in cool stars , while balona ( 2012 ) reported observations of stars with a / f spectral type and maehara et al ( 2012 ) presented some examples of ` super ' flares on solar type stars .
in june 2011 we started the _ rats - kepler _ project whose aim was to identify sources which showed flux variations on short ( @xmath330 min ) timescales ( ramsay et al 2013 ) .
we do this by taking a series of short exposures ( 20 sec ) using wide field cameras on telescopes such as the isaac newton telescope on la palma on specific fields for one hour .
light curves of each object are derived and variable sources identified .
one variable source which we identified was kic 5474065 which showed a short duration ( @xmath320 min ) flare with an amplitude of 0.6 mag in the @xmath4 band .
we were successful in placing kic 5474065 on the _ kepler _ 1 min sampling target list .
this paper presents the results of these _ kepler _ observations and a comparison of the energy distribution of the flares with other low mass flare stars .
although kic 5474065 ( @xmath5=19h 53 m 02.3s , @xmath6=+40@xmath7 40@xmath8 34.6@xmath9 ) is included in the kepler input catalog ( brown et al 2011 ) , it does not have a measured temperature or surface gravity .
it is , however , included in the kepler - int survey ( u=20.61 , g=19.00 , r=17.33 , i=15.60 ; greiss et al 2012a , b ) ; the ubv survey of the kepler field ( b=18.79 , v=18.07 ; everett , howell & kinemuchi 2012 ) and also the 2mass survey ( j=14.015 , h=13.397 , k=13.215 ; skrutskie et al 2006 ) .
the optical colours indicate a relatively late - type star . since kic 5474065 is variable ( due to its rotational modulation and flare activity ) some degree of caution
is required when determining its colours unless it is known that multi - band observations are made simultaneously .
however , lpine & gaidos ( 2011 ) show the relationship between the colour @xmath10 and spectral type for late - type stars . for kic 5474065
, @xmath11 = 4.06 implies a spectral type of m3 to m4 .
we carried out low - resolution spectroscopy with the optical system for imaging and low resolution integrated spectroscopy ( osiris ) tunable imager and spectrograph ( cepa et al .
2003 ) at the 10.4 m gran telescopio canarias ( gtc ) , located at the observatorio roque de los muchachos in la palma , canary islands , spain .
the heart of osiris is a mosaic of two 4k@xmath122k e2v ccd4482 detectors that gives an unvignetted field of view of [email protected]@xmath13 with a plate scale of 0.127arcsecpix@xmath14 .
however , to increase the signal - to - noise ratio of our observations , we chose the standard operation mode of the instrument , which is a 2@xmath122-binning mode with a readout speed of 100khz .
two spectra each with an exposure of 300 sec were obtained using the osiris r1000r grism in service mode on 11 may 2013 .
they were made as part of a gtc filler programme which utilies poor weather conditions .
we used the 1.0@xmath9-width slit , oriented at the parallactic angle to minimise losses due to atmospheric dispersion .
the resulting resolution , measured on arc lines , was r @xmath15 700 in the approximate 52509200 spectral range .
the star ross 640 was used to remove the instrumental response .
the data were reduced using standard figaro routines .
we show the optical spectrum of kic 5474065 in figure [ spectrum ] : it is clearly a late - type dwarf star . examining figure 1 of bochanski et al ( 2007 ) ,
kic 5474065 is later than an m0v spectral type .
judging by the depth of the na i ( 8190 ) feature and the ca ii triplet around 8500 it is most likely that kic 5474065 has a m4v spectral type although m3v and m5v are also possible . in order to determine the energies of the flares
, we must first estimate the intrinsic luminosity of kic 5474065 .
lpine & gaidos ( 2011 ) include @xmath16 and parallax data for late type stars .
we were able to extract data as a function of spectral type and estimate the mean @xmath17 using relationship between @xmath10 and @xmath17 outlined in lpine & gaidos ( 2011 ) .
we show in table [ mv ] the mean absolute @xmath18 magnitude for spectral types m3v m5v and we assume the sun has @xmath17 = 4.83 and @xmath19 erg s@xmath14 .
( 8,5.5 ) ( 0,0 ) .the absolute magnitude and luminosity of stars with spectral types m3 v
m5 v based on the data in lpine & gaidos ( 2011 ) . [ cols="<,>,>",options="header " , ] [ flare - rate ] ( 16,10 ) ( 0,11 )
we present _ kepler _ short cadence observations of the m4 v star kic 5474065 which has a rotation period of 2.47 days .
it shows two high amplitude short duration flares ( @xmath200.4 which have integrated energies of @xmath21 ergs .
additional flares energies as low as @xmath22 ergs are also seen .
we compare the flare rate of a second m4 v star kic 9726699 which is more than 5 mag brighter than kic 5474065 and has a more rapid rotation period of 0.60 days .
compared to kic 5474065 , kic 9726699 does not show such high amplitude flares but since the _ kepler _ data has a higher signal to noise , it allows us to detect many short duration , low energy flares reaching energies as low as @xmath23 ergs .
although the effect of flares with energies of @xmath24 ergs on the atmosphere s of exoplanets in the habitable zone have been investigated , it is of great interest to determine what effect the presence of many numbers of lower energy events will have on exo - planet atmospheres .
this paper includes data collected by the kepler mission . funding for the kepler mission
is provided by the nasa science mission directorate .
some of the data presented in this paper were obtained from the mikulski archive for space telescopes ( mast ) .
stsci is operated by the association of universities for research in astronomy , inc .
, under nasa contract nas5 - 26555 .
support for mast for non - hst data is provided by the nasa office of space science via grant nnx09af08 g and by other grants and contracts .
observations were also made with the gran telescopio canarias ( gtc ) , installed in the spanish observatorio del roque de los muchachos of the instituto de astrofsica de canarias , in the island of la palma . armagh observatory is supported by the northern ireland government through the dept culture , arts and leisure .
we thank the referee for a constructive and helpful report .
baliunas , s. l. et al .
1995 , apj , 438 , 269 balona , l. , 2012 , mnras , 423 , 3420 batyrshinova , v. m. , ibragimov , m. a. , 2001 , astl , 27 , 29 bochanski , j. j. , munn , j. a. , hawley , s. l. , west , a. a. , covey , k. r. , schneider , d. p. , 2007 ,
apj , 134 , 2418 bopp , b. w. , moffett , t. j. , 1974 , apj , 185 , 239 brown , t. m. , latham d. w. , everett m. e. , esquerdo g. a. , 2011 , aj , 142 , 112 byrne , p.b . , doyle , j.g . butler , c.j . ,
andrews , a.d . , 1984 , mnras 211 , 607 byrne , p.b . ,
doyle , j.g .
menzies , j.w . , 1985 , mnras 214 , 119 cepa , j. , et al . , 2003 , rmxac , 16 , 13 dal , h.a . , evren , s. , 2010 , aj 140 , 483 dal , h.a . ,
evren , s. , 2011 , pasp 123 , 659 dal , h.a .
, 2011 , pasa 28 , 365 dal , h.a .
, 2012 , pasj 64 , 82 davenport , j. r. a. , becker , a. c. , kowalski , a. f. , hawley , s. l. , schmidt , s. j. , hilton , e. j. , sesar , b. , cutri , r. , 2012 , apj , 748 , 58 donati , j. -f . , et al .
2008 , mnras , 385 , 1179 doyle , j.g . , butler , c.j .
, 1985 , nat 313 , 378 doyle , j.g . , byrne , p.b . , 1986 ,
a&a 154 , 370 doyle , j.g . , byrne , p.b . , butler , c.j .
, 1986 , a&a 156 , 283 doyle , j.g . ,
van oord , g.h.j . ,
butler , c.j .
, 1989 , a&a 208 , 208 doyle , j.g . , butler , c.j . , van oord , g.h.j . , kiang , t. , 1990 , a&a 232 , 83 everett m. e. , howell s. b. , kinemuchi k. , 2012 , pasp , 124 , 316 gershberg , r. e. , shakhovskaia , n. l. , 1983 , ap&ss , 95 , 235 greiss s. , et al .
, 2012a , aj , 144 , 24 greiss s. , et al . , 2012b , arxiv:1212.3613 haisch , b. , strong , k. t. , rodono , m. , 1991 , ara&a , 29 , 275 hawley , s. l. , pettersen , b. r. , 1991 , apj , 378 , 825 hawley , s.l . , panov , k.p . , pettersen b.r . , sundland , s.r
, 1989 , a&a 220 , 218 hunt - walker n. m. , hilton e. j. , kowalski a. f. , hawley s. l. , matthews j. m. , 2012 , pasp , 124 , 545 kirkpatrick , j. d. , henry , t. j. , mccarthy , d. w. , 1991 , apjs , 77 , 417 kopparapu , r. k. , et al , 2013 , apj , 765 , 131 lacy , c.h . , moffett , t.j . ,
evans , d.s . , 1976 , apj s 30 , 85 lpine , s. , gaidos , e. , 2011 , aj , 142 , 138 leto , g. , pagano , i. , buemi , c.s . & rodono , m. , 1997 , a&a 327 , 1114 maehara , h. , shibayama , t. , notsu , s. , notsu , y. , nagao , t. , kusaba , s. , honda , s. , nogami , d. , shibata , k. , 2012 , nature , 485 , 478 moffett , t.j .
, 1974 , apjs 29 , 1 morin , j. , donati , j .- f . ,
petit , p. , delfosse , x. , forveille , t. , jardine , m. m. , 2010 , mnras , 407 , 2269 notsu , y. , shibayama , t. , maehara , h. , notsu , s. , nagao , t. , honda , s. , ishii , t. t. , nogami , d. , shibata , k. , 2013 , accepted apj , ( arxiv1304.7361 ) noyes , r. w. , 1985 , soph , 100 , 385 pettersen , b.r . , 1975a , a&a 41 , 87 pettersen , b.r .
, 1975b , a&a 41 , 113 pettersen , b.r . , 1981a , a&a 95 , 135 pettersen , b.r . , 1981a , a&a 97 , 199 pettersen , b.r . , 1983 , a&a 120 , 192 pettersen , b.r
. , 1985a , a&a 148 , 147 pettersen , b.r . , 1985b , a&a 148 , 151 pettersen , b.r . ,
2006a , obs 124 , 397 pettersen , b.r .
, 2006b , mnras 368 , 1392 pettersen b.r . ,
kern , g.a .
, evans , d.s . , 1983 , a&a 123 , 184 pettersen , b.r . , coleman , l.a . ,
evans , d.s .
, 1984 , apjs 54 , 375 pettersen , b.r .
, panov , k.p . , sandmann , w.h . ,
ivanova , m.s .
, 1986 , a&a s 66 , 235 pettersen , b.r . , sundland , s.r . , 1991 , a&a s 87 , 303 ramsay , g. , brooks , a. , barclay , t. , hakala , p. , still , m. , greiss , s. , gnsicke , b. , steeghs , d. , reynolds , m. , 2013 , submitted , mnras reid , i. n. , et al , 2004 , aj , 128 , 463 savanov , i. s. , dmitrienko , e. s. , 2011 , arep , 55 , 890 segura , a. , walkowicz , l. m. , meadows , v. , kasting , j. , hawley , s. , 2010 , asbio , 10 , 751 skrutskie , m. f. , et al .
, 2006 , aj , 131 , 1163 van den oord , g. h.j . , et al . , 1996 , a&a , 310 , 908 ventura , r. , peres , g. , pagano , i. , rodono , m. , 1995 , a&a , 303 , 509 walkowicz , l. m. , et al 2011 , aj , 141 , 50 | using data obtained during the _ rats - kepler _ project we identified one short duration flare in a 1 hour sequence of ground based photometry of the dwarf star kic 5474065 .
observations made using gtc show it is a star with a m4 v spectral type . _
kepler _ observations made using 1 min sampling show that kic 5474065 exhibits large amplitude ( @xmath0f / [email protected] ) optical flares which have a duration as short as 10 mins .
we compare the energy distribution of flares from kic 5474065 with that of kic 9726699 , which has also been observed using 1 min sampling , and ground based observations of other m dwarf stars in the literature .
we discuss the possible implications of these short duration , relatively low energy flares would have on the atmosphere of exo - planets orbiting in the habitable zone of these flare stars . physical data and processes : magnetic reconnection astrobiology stars : activity stars : flares stars : late - type stars : individual : kic 5474065 , kic 9726699 |
in recent work @xcite ( see also @xcite ) we have analysed some aspects of the cauchy problem for the einstein equations with data on a characteristic cone in all dimensions @xmath0 , see @xcite and references therein for previous work on the subject . in this note
we apply the results derived in @xcite to present an existence theorem for this problem , with initial data which approach rapidly the flat metric near the tip of the light cone , see theorem [ t23vii.1 ] below .
the reader s attention is drawn to @xcite , where sets of unconstrained data on a light - cone centered at past timelike infinity are given in dimension @xmath1 .
it is well known that by using normal coordinates centred at @xmath2 the characteristic cone @xmath3 of a given lorentzian metric can be written , at least in a neighbourhood of @xmath2 , as a cone in minkowski spacetime whose generators represent the null rays .
it is therefore no geometric restriction to assume that the characteristic cone of the spacetime we are looking for is represented in some coordinates @xmath4 , @xmath5 , @xmath6 of @xmath7 by the equation of a minkowskian cone with vertex @xmath2 , @xmath8 the parameter @xmath9 is an affine parameter when normal coordinates are used , and it is also going to be an affine parameter in the solutions that we are going to construct .
coordinates @xmath10 as above , which moreover satisfy the wave - equation , @xmath11 , will be called _ normal - wave _ coordinates . given a smooth metric ,
such coordinates can always be constructed ( see @xcite or @xcite ) by solving the wave equation in the domain of dependence of @xmath3 , with initial data the normal coordinates on @xmath3 . given a smooth metric , one obtains a coordinate system near the vertex , which suffices for our purposes . the coordinates
@xmath12 will be normal - wave coordinates for the solution which we aim to construct . given the coordinates @xmath13 we can define coordinates @xmath14 on @xmath7 by setting @xmath15 with @xmath16 local coordinates on @xmath17 .
the null geodesics issued from @xmath2 have equation @xmath18 , @xmath19constant , so that @xmath20 is tangent to those geodesics . on @xmath21 ( but not outside of it in general )
the spacetime metric @xmath22 that we attempt to construct takes the form ( we put an overbar to denote restriction to @xmath3 of spacetime quantities ) @xmath23 where @xmath24 are respectively an @xmath25-dependent scalar , one - form , and riemannian metric on @xmath17 .
the symbol @xmath9 will be used interchangeably with @xmath26 . to avoid ambiguities
, we will write @xmath27 for the components of a metric tensor in the @xmath28 coordinates , and @xmath29 for the components in the coordinate system @xmath10 .
this convention will be used regardless of whether @xmath22 is defined on the space - time , or only on the light - cone .
the analysis in @xcite uses a wave - map gauge , with minkowski target @xmath30 , with the light - cone of @xmath30 being the image by the wave - map @xmath31 of the light - cone of the metric @xmath22 that one seeks to construct .
quite generally , a metric @xmath22 on a manifold @xmath32 will be said to be _ in @xmath33-wave - map gauge _ if the identity map @xmath34 is a harmonic diffeomorphism from the spacetime @xmath35 onto the pseudo - riemannian manifold ( @xmath36 . recall that a mapping @xmath37 is a harmonic map if it satisfies the equation , in abstract index notation , @xmath38 in a subset in which @xmath31 is the identity map defined by @xmath39 , the above equation reduces to @xmath40 , where the _ wave - gauge vector _
@xmath41 is given in arbitrary coordinates by the formula @xmath42 where @xmath43 are the christoffel symbols of the _ target _ metric @xmath33 .
see @xcite for a more complete discussion .
there are various ways of choosing free initial data for the cauchy problem for the vacuum einstein equations on the light - cone @xmath3 . in this work
we choose as initial data a one - parameter family , parameterized by @xmath9 , of conformal classes of metrics @xmath44 $ ] on @xmath17 , thus @xmath45 is assumed to be conformal to @xmath46 , and where @xmath9 will be an affine parameter in the resulting vacuum space - time .
the initial data needed for the evolution equations are the values of the metric tensor on the light - cone , which will be obtained from @xmath46 by solving a set of wave - map - gauge constraint equations derived in @xcite , namely equations ( [ 3v.1 ] ) , ( [ 5vi.13 ] ) , ( [ xiabis])-([cafinal2 ] ) and ( [ 19viii.1 ] ) below . the main issue is then to understand the behaviour of the fields near the vertex of the light - cone , making sure that the @xmath47-coordinates components of the metric @xmath48 , obtained by solving the wave - map - gauge constraints , can be written as restrictions to the light - cone of sufficiently smooth functions on space - time , so that the pde existence theorem of @xcite can be invoked to obtain the vacuum space - time metric .
it is convenient to start with some notation .
for @xmath49 and @xmath50 we shall say that a tensor field @xmath51 , of valence @xmath52 , defined for @xmath53 , is @xmath54 if there exists a constant @xmath55 such that @xmath56 where @xmath57 are coordinate components of @xmath58 , in a coordinate system which will be made clear as needed .
[ t23vii.1 ] let @xmath60 , @xmath61 , @xmath62 , @xmath63 , @xmath64 , @xmath65 $ ] .
suppose that there exist smooth tensor fields @xmath66 on @xmath17 , @xmath67 , so that , in local charts on @xmath17 , the coordinate components @xmath68 satisfy @xmath69 then : 1 .
there exist functions @xmath70 such that @xmath71 if there exists @xmath72 so that @xmath73 for @xmath74 , then @xmath75 for @xmath76 .
if moreover @xmath77 then there exists a @xmath78 lorentzian metric defined in a neighbourhood of the vertex of @xmath3 , with @xmath79 $ ] , which is vacuum to the future of @xmath2 . roughly speaking , the index @xmath80 is the final sobolev differentiability of the solution .
the ranges of indices above are only needed for the second part of the theorem , and arise from the fact that dossa s existence theorem @xcite requires initial data which are of @xmath81 differentiability class in coordinates regular near the vertex .
there is a loss of three derivatives when going from the free data @xmath68 to the full initial data @xmath82 , which brings the threshold up to @xmath83 .
finally , the existence argument invokes the bianchi identity , which in its natural version requires a @xmath84 metric ; together with the sobolev embedding , this leads to the restriction @xmath62 .
straightforward taylor expansions at @xmath2 show that a smooth metric on a space - time @xmath85 will lead to the form ( [ 30vi.5x ] ) of @xmath44 $ ] , with @xmath80 , @xmath86 and @xmath52 which can be chosen at will , and with @xmath87 .
so ( [ 30vi.5x ] ) is necessary in this sense . in view of our theorem above , to obtain a complete solution of the problem at hand it remains to provide an exhaustive description of those @xmath44 $ ] s that lead to ( [ 23vii.5 ] ) .
we conjecture that ( [ 23vii.5 ] ) will hold for all @xmath46 s arising from the restriction of a smooth metric to a light - cone , where @xmath9 is an affine parameter . a similar problem arising on the null cone at past infinity in dimensions
@xmath1 has been solved by friedrich in @xcite , but no details have been presented .
proof of theorem [ t23vii.1 ] : we need to analyze the behaviour of the solutions of the wave - map - gauge constraint equations near the vertex of the cone .
we start by noting that , for some smooth functions @xmath95 on @xmath17 , @xmath96 where @xmath97 is the matrix inverse of @xmath68 .
further , for some smooth tensors @xmath98 on @xmath17 , @xmath99 note that @xmath100 is the @xmath59trace - free part of @xmath101 .
this leads to , for some functions @xmath102 on @xmath17 , @xmath103 where @xmath104 has been defined in the left - hand side of .
let @xmath105 where @xmath106 is the divergence of @xmath3 ( sometimes denoted by @xmath107 in the literature ; cf .
, e.g. , @xcite ) : @xmath108 in terms of @xmath12 , the vacuum raychadhuri equation @xmath109 , where @xmath110 is a null tangent to the generators of @xmath3 , reads @xmath111 using known arguments ( compare @xcite and ( * ? ? ?
* lemma 8.2 ) ) , there exist functions @xmath112 such that @xmath113 we rewrite this as @xmath114 so that @xmath115 using this last formula , it is easy to establish existence of functions @xmath116 such that @xmath117 let us write @xmath118 from @xmath119 there exist functions @xmath120 such that : @xmath121 equation ( [ 30vi.8 ] ) implies that there exist smooth tensor fields @xmath122 on @xmath17 such that @xmath123 subsequently , @xmath124 for some tensor fields fields @xmath125 . it is a consequence of the affine - parameterization condition and of the wave - gauge conditions that the function @xmath126 solves the equation ( see @xcite for details ) @xmath127 set @xmath128 then ( [ 5vi.13 ] ) integrates to @xmath129 where @xmath130 and @xmath131 we find successively @xmath132 for some @xmath133 .
closer inspection shows that @xmath134 . integrating ( [ 30vi.12 ] ) ,
we conclude that there exist functions @xmath135 such that @xmath136 if @xmath137 the sum from four to @xmath138 is understood as zero .
one has a similar formula for @xmath139 .
we pass now to a vector @xmath140 , defined as @xmath141 here the @xmath142 s are the christoffel symbols of the canonical metric @xmath59 on @xmath17 , @xmath143 is as in ( [ wavegauge0 ] ) , the @xmath144 s are the christoffel symbols of the @xmath145-dimensional metric @xmath146 , with associated covariant derivative operator @xmath147 , and @xmath148 , while @xmath149 in terms of @xmath150 the equation @xmath151 in wave - map gauge reads @xmath152 where @xmath153 . from what has been proved so far
this last equation can be written as @xmath154 for some @xmath155 .
the unique solution of this equation with the relevant behaviour at the origin satisfies @xmath156 for some @xmath157 . viewing ( [ xiabis ] ) as an equation for @xmath158 , we obtain @xmath159 for some @xmath160 .
we finally need to integrate the third wave - map - gauge constraint , @xmath161 , where @xmath162 is the einstein tensor : @xmath163 we need the formula , for some functions @xmath164 , @xmath165 and the result , for a minkowski target , @xmath166 with @xmath167 .
hence @xmath168 for some functions @xmath169 .
a straightforward analysis of the remaining terms in ( [ 19viii.1 ] ) shows that solutions with the required asymptotics satisfy , for some functions @xmath170 , @xmath171 in order to apply dossa s existence theory from @xcite to the wave - map gauge reduced einstein equations , we transform the metric to coordinates @xmath172 , regular near the tip of the light - cone , defined as @xmath173 where @xmath174 . in these
coordinates our wave - map - reduced einstein equations become the usual harmonically - reduced equations , which have the right structure for the existence results in @xcite .
one finds that there exist functions @xmath175 such that @xmath176 this proves the first part of theorem [ t23vii.1 ] .
now , for general @xmath66 s the sum @xmath177 will _ not _ be the restriction of a polynomial to the light - cone .
assume , however , that this is the case .
then the remainder term in the difference @xmath178 will be of the differentiability class @xmath179 with @xmath180 , as required by dossa for existence @xcite , provided that @xmath181 the solution @xmath182 is then in dossa s space @xmath183 for some @xmath184 , which embeds into @xmath185 leading to , for small @xmath186 , @xmath187 in the coordinates @xmath172 the harmonicity vector can be calculated using the usual formula , @xmath188 which has bounded components in the @xmath172 coordinate system near the tip of the cone . the bianchi identity ( which , in its simplest version , requires a @xmath84 metric ;
this raises the differentiability threshold to @xmath62 ) together with the arguments in @xcite show then that @xmath189 and thus the solution of the wave - map reduced einstein equations , obtained from dossa s theorem , is also a solution of the vacuum einstein equations .
@xmath190 acknowledgements : ptc and ycb are grateful to the mittag - leffler institute , djursholm , sweden , for hospitality and financial support during part of work on this paper .
they acknowledge useful discussions with vincent moncrief , as well as comments from roger tagn wafo .
ycb wishes to thank thibault damour for making available his detailed manuscript calculations in the case @xmath191 leading to equations ( 22 ) of @xcite .
jmm thanks oxpde for hospitality .
rendall , _ the characteristic initial value problem for the einstein equations _ , nonlinear hyperbolic equations and field theory ( lake como , 1990 ) , pitman res .
notes math .
ser . , vol .
253 , longman sci .
harlow , 1992 , pp . | we prove an existence theorem for the cauchy problem on a characteristic cone for the vacuum einstein equations . |
methods and techniques inspired by statistical physics have been shown to be useful in the search for hidden features in social systems @xcite . particularly , it has been remarked that social systems can exhibit universal properties similarly to thermodynamical systems at the criticality .
for example , scaling laws have been reported in scientific research @xcite , biological systems @xcite , economics @xcite and religious @xcite activities .
in addition , there is some evidence about the relation between urban metrics and the population size , where non - linearities are explicitly manifested by power - laws @xcite .
more recently , phase transition was also found in a model for criminality @xcite .
crime is one of the major concerns of contemporary society and , therefore , there is a great interest in understanding features of its organization and dynamics .
we are living in a period when most people live in cities @xcite .
the increasing concentration of people in urban areas entails both opportunities and challenges @xcite .
in particular , cities have become one of the principal sources of problems such as pollution , spread of disease and crime @xcite .
studies on this last subject involve many areas of knowledge going from human sciences @xcite to exact sciences @xcite .
in particular , the economic relevance of social issues of crime has been discussed @xcite .
it has also been pointed out that social interactions may explain the large variance in crime in cities with different concentration of people @xcite . however , despite the broad investigations and the social relevance of criminality in connection with urbanization , our understanding of universality and pervasiveness of patterns in this subject remains limited . in this work ,
we apply a statistical approach to investigate patterns and growth characteristics of homicide . in a quantitative manner , our study provides insights on how the growth of crime and the size of cities are related .
more precisely , we study homicide growth rates in brazilian cities based on data from 1980 to 2009 ( section 2 ) , focusing on scaling laws related to probability distributions and standard deviations .
we investigate homicide growth rates of cities aiming to understand the mechanisms that govern criminality ( section 3 ) .
we show that our results have a similar formal structure to those ones found in complex systems such as scientific research , biological systems , economic and religious activities , fact that put the universality of our findings in evidence ( section 4 ) .
motivated by this universality , we indicate a scenario to our results in terms of stochastic models proposed in the context of complex organizations ( section 4 ) .
our conclusions are in section 5 .
for the study of statistical properties of crime , we consider a database containing the annual number of homicides in all brazilian cities spanning the years from 1980 to 2009 , obtained from the database of vital statistics of datasus @xcite .
the annual population of cities was obtained from the demographic and socio - economic database of datasus @xcite . in this last database ,
the years 1980 , 1991 , 1996 and 2000 report the population number obtained by the population census conducted by the ibge @xcite , while all the other years are actually approximated values of the population number estimated by ibge agency . due to this fact ,
our analysis will be mainly focused on the homicides database . .
the error bars are 95% confidence intervals obtained via bootstrapping @xcite . ] we selected 528 cities from this set , which present a significant number of homicides ( at least one per year ) in the period from 1980 to 2009 .
they are about @xmath0 of brazilian cities but represent approximately @xmath1 of the total number of homicides in brazil in the period considered
. moreover , the average percentage of the population of the country living in these cities during this period is about @xmath2 .
an illustration of our database is given in fig .
[ fig : crimepop ] . in this figure ,
a typical scaling law can be observed if we consider only the big cities ( population greater than 53.465 ) .
we find an exponent very close to those found in other studies on urban metrics and crime @xcite .
however , if we take the 528 cities into account the exponent is approximately one . for years subsequent to 1980 ,
the behavior of the exponents is similar
. given by equation ( [ eq:1 ] ) for some cities in the period from 1980 to 2009 .
( a ) so paulo ( @xmath3 ) and maring ( @xmath4 ) .
( b ) braslia ( @xmath5 ) and pereira barreto ( @xmath6 ) .
observe that the fluctuations in @xmath7 are generally larger in cities with small number of homicides in 1980 , @xmath8.,title="fig : " ] given by equation ( [ eq:1 ] ) for some cities in the period from 1980 to 2009 .
( a ) so paulo ( @xmath3 ) and maring ( @xmath4 ) .
( b ) braslia ( @xmath5 ) and pereira barreto ( @xmath6 ) .
observe that the fluctuations in @xmath7 are generally larger in cities with small number of homicides in 1980 , @xmath8.,title="fig : " ] in terms of the total number of homicides @xmath9 in the year @xmath10 in a given city , the annual , biannual and triannual ( logarithmic ) growth rates are defined as @xmath11 , \label{eq:1}\ ] ] with @xmath12 , @xmath13 and @xmath14 , respectively . to simplify the notation , we omit the sub - index @xmath15 when referring to annual growth rates , this is , we employ @xmath7 to represent @xmath16 .
examples of the temporal evolution of @xmath7 for some cities are shown in fig .
[ fig : evolution ] .
they illustrate the presence of fluctuations in the homicide growth rate @xmath7 .
this figure also exemplifies that the fluctuations of @xmath7 are generally larger in small towns than in bigger urban centers .
this is an expected result that motivates our analysis in the standard deviation in function of the city size , since fluctuations are larger in small systems than in bigger systems , as found in different contexts @xcite .
even without a detailed description of the system , studies on complex systems have shown that scaling properties of fluctuations in macroscopic quantities can lead to a better understanding of some features of these variables . in this direction ,
a variety of complex organizations has been investigated and it has been shown that , despite the differences , they all show non - gaussian tent - shaped distributions in their annual growth rates with a mean power - law behavior in its standard deviation @xcite .
these studies have suggested that there are some universal mechanisms , independent of the particular details , governing the fluctuations on growth rates of complex organizations .
these facts motivate us to investigate aspects of the crime growth focusing on these statistical properties .
we first consider the conditional distribution of @xmath7 by taking some ranges of initial values of @xmath9 into account . in an attempt to reduce effects of statistical fluctuations ,
we employ cumulative distribution functions ( cdfs ) . to use this method ,
we separate the growth rates @xmath7 in positive and negative values . for instance
, for positive @xmath7 we have the cdf @xmath17 where @xmath18 is the conditional probability density function .
figure [ fig : distribution ] shows examples of empirical cdfs by considering three ranges for the values of @xmath8 among the set of possible values , where we find a good agreement between the empirical data and the family of laplace distributions @xmath19 with @xmath20 being the mean value and @xmath21 being the standard deviation of @xmath22 within each selected interval of @xmath8 .
for some intervals of initial number of homicides @xmath8 : red circles ( @xmath23 ) , green squares ( @xmath24 ) and blue triangles ( @xmath25 ) on a mono - log scale . the dashed lines are laplace distribution fits .
( a ) positive values of @xmath22 and ( b ) module of the negative values of @xmath22.,title="fig : " ] for some intervals of initial number of homicides @xmath8 : red circles ( @xmath23 ) , green squares ( @xmath24 ) and blue triangles ( @xmath25 ) on a mono - log scale .
the dashed lines are laplace distribution fits . ( a ) positive values of @xmath22 and ( b ) module of the negative values of @xmath22.,title="fig : " ] scaling behavior
can be graphically represented considering an appropriate change in the scale of the independent and dependent variables . if scaling holds , the data for a variety of initial values should collapse upon a single curve .
for instance , we note from equation ( [ eq:3 ] ) that @xmath26 is the scaling variable , where @xmath20 is the mean value and @xmath21 is the standard deviation of @xmath22 within each interval of @xmath8 .
this changing of variable leads us to @xmath27 where @xmath28 is the probability distribution of the scaled variable @xmath29 , which has no parameters .
we divided the data in log - spaced intervals of @xmath8 .
then , by using the scaled variable @xmath29 , we have verified that all curves with different ranges of @xmath8 collapse approximately in one single cdf ( see fig . [
fig : cumulative ] ) , indicating that the distribution of growth rates exhibits the same functional form for distinct size scales . from this figure
, we also note that the laplace distribution presents small , but systematic , deviations from the empirical data .
, for each log - spaced range of initial values of @xmath8 .
the squares are the average values of the cdfs by windows in the scaled growth rate , @xmath29 .
the dashed lines show the laplace distribution with unitary standard deviation and zero mean ( [ eq:5 ] ) .
the continuous lines refer to a stretched gaussian , equation ( [ eq:6 ] ) with @xmath30 .
( a ) positive values of @xmath29 and ( b ) module of the negative values of @xmath29 .
the error bars are @xmath31 confidence intervals obtained via bootstrapping @xcite.,title="fig : " ] , for each log - spaced range of initial values of @xmath8 .
the squares are the average values of the cdfs by windows in the scaled growth rate , @xmath29 .
the dashed lines show the laplace distribution with unitary standard deviation and zero mean ( [ eq:5 ] ) .
the continuous lines refer to a stretched gaussian , equation ( [ eq:6 ] ) with @xmath30 .
( a ) positive values of @xmath29 and ( b ) module of the negative values of @xmath29 .
the error bars are @xmath31 confidence intervals obtained via bootstrapping @xcite.,title="fig : " ] in an attempt to overcome these deviations , we have fitted the data by using the one - parameter distribution @xmath32\ , .
\label{eq:6}\ ] ] this distribution is a `` stretched gaussian '' @xcite and it can recover the standard gaussian for @xmath33 and the laplace distribution for @xmath34 . in our case , we have found that @xmath35 gives a more accurate description for the empirical cdfs ( see fig .
[ fig : cumulative ] ) .
note that , if we consider directly @xmath28 instead of its cdf , we also obtain the tent - shaped distribution as shown in fig .
[ fig : tentshaped ] . of the normalized annual growth rates
@xmath29 on mono - log scale .
the different symbols represent each interval of @xmath8 employed in the analysis of the pdf . here , the intervals of @xmath8 are the same used in fig .
[ fig : cumulative ] .
the dashed black line is the laplace distribution ( [ eq:5 ] ) with unitary standard deviation and zero mean .
the continuous blue line is the stretched gaussian ( [ eq:6 ] ) with parameter @xmath30 . ]
returning to fig .
[ fig : distribution ] , we note that , for small values of @xmath36 the distributions have longer tails and that the tails become shorter as @xmath8 increases . this result and fig .
[ fig : evolution ] indicate that large fluctuations in the homicide growth rates are observed in small towns and that the standard deviation @xmath21 tends to decrease with the increasing of @xmath8 . at this point
, we are considering the number of homicides as a measure of the city size since , in principle , the average number of crimes increases with the number of inhabitants ( see fig .
[ fig : crimepop ] and @xcite ) . in our analysis
, we found that the main tendency of the relationship between @xmath21 and @xmath8 is well adjusted by a power - law : @xmath37 with @xmath38 for @xmath39 ( see fig .
[ fig : law]a ) .
however , the relation between the number of crimes and population is non - linear @xcite and not completely understood when we take all city sizes into account , i.e. , all brazilian cities . for this reason ,
we directly investigate @xmath21 as a function of the total population @xmath40 of each city in 1980 . as a result
, we verify that @xmath41 with @xmath42 as shown in fig .
[ fig : law]b .
this result is in agreement with fig .
[ fig : crimepop ] , since when taken the 528 cities into account , the relationship between the number of homicides and the population of the cities is approximately linear , as discussed in the second paragraph of section 2 . of homicide time - series for the 528 examined cities .
the black squares are window averaged values of the standard deviation @xmath21 of the growth rates @xmath22 as a function of the average value of @xmath8 .
( b ) here , we replace the initial value of @xmath8 by the population @xmath40 for the city .
the black squares are window averaged values of the standard deviation @xmath21 of the growth rates @xmath22 as a function of the average value of @xmath40 . for both figures ,
the continuous lines are linear fits to the main tendencies of data on log - log scale ( equation ( [ eq:7 ] ) ) , yielding an @xmath43 .
the error bars are @xmath31 confidence intervals obtained via bootstrapping @xcite.,title="fig : " ] of homicide time - series for the 528 examined cities .
the black squares are window averaged values of the standard deviation @xmath21 of the growth rates @xmath22 as a function of the average value of @xmath8 .
( b ) here , we replace the initial value of @xmath8 by the population @xmath40 for the city .
the black squares are window averaged values of the standard deviation @xmath21 of the growth rates @xmath22 as a function of the average value of @xmath40 . for both figures ,
the continuous lines are linear fits to the main tendencies of data on log - log scale ( equation ( [ eq:7 ] ) ) , yielding an @xmath43 .
the error bars are @xmath31 confidence intervals obtained via bootstrapping @xcite.,title="fig : " ] in addition to @xmath7 , we also investigate biannual , @xmath44 , and triannual , @xmath45 , homicide growth rates . in these last two cases , we have the same scenario of the annual growth rates . however , the numerical values of the parameters related to the annual , biannual and triannual growth rates , in general , are not equal , as shown in table 1 .
note that each @xmath46 value and the corresponding @xmath47 are very close to each other and that they decrease as the time interval increases .
if we consider @xmath21 as a function of @xmath9 or @xmath48 in years other than @xmath49 , @xmath21 also presents an asymptotic power - law behavior .
however , the @xmath50 exponents decrease as @xmath10 increases .
this fact is more pronounced when we take @xmath21 as a function of @xmath9 than @xmath48 . in the case of @xmath48 ,
the exponents are almost constant . as a last concern ,
we consider annual growth rates @xmath7 of cities with at least one homicide per year in smaller periods , i.e. , starting in years subsequent to 1980 and ending in 2009 .
naturally , this kind of analysis contains different initial conditions .
nevertheless , the @xmath46 and @xmath47 values fluctuate around @xmath51 and @xmath52 respectively .
all these results as well as the previous ones are robust to window size change .
[ tab : summary ] .*summary of the results * [ cols="^,^,^,^,^",options="header " , ] a summary of our empirical results about annual , biannual and triannual logarithmic homicide growth rates for brazilian cities is given ( see equation ( [ eq:1 ] ) ) considering data from 1980 to 2009 .
the second and third columns inform us the shape of the collapsed distributions , where the parameter @xmath53 refers to the stretched gaussian ( sg ) and indicates the deviation from the laplace distribution ( @xmath34 ) ( see equation ( [ eq:6 ] ) ) .
the parameters @xmath46 and @xmath47 are asymptotic exponents of the main tendencies of the standard deviations @xmath21 of the logarithmic growth rates ( see equation ( [ eq:7 ] ) ) .
a large part of the literature about criminality tries to relate crime rates to possible explicative variables as average income , unemployment , inequality , gender , age , education level , race and others related to delinquency , even though statistical data of crime rates can not give much information on this @xcite .
certainly , cities with the same size can present different numbers of crimes ( as exemplified in fig .
[ fig : crimepop ] ) usually due to socio - economic aspects . however , in our analysis we have not specified such dependences since our results arise from a more general statistical context .
this fact can be viewed as reminiscent of the concept of universality found in statistical physics , where different systems can be characterized by the same fundamental laws , independent of the microscopic details @xcite .
systems with the same values of critical - point exponents and scaling functions are said to belong to the same universality class @xcite . to find an exponent for the relationship between standard deviation and city size is , somehow , to indicate the class to which the system belongs .
in fact , in other contexts @xcite the analog of this exponent generally assumes different values .
the presence of non - systematic errors on two commonly employed databases , fbi ( federal bureau of investigation ) and ucr ( uniform crime report u.s.a . ) ones , has been pointed out due to a lack of reports from some local agencies @xcite . for our analysis as well as in another recent one @xcite
, we choose the homicide crime to minimize this unwanted aspect of data since , in principle , any kind of demise is recorded by datasus @xcite .
these facts indicate that our results are robust and can be a useful guide for future works on criminality modeling .
we found that the distributions of homicide growth rates exhibit a scale - invariant structure in distinct city sizes , as shown in fig .
[ fig : cumulative ] .
our results show that the distributions of annual , biannual and triannual homicide growth rates are close to a laplace distribution .
we also found that the standard deviations of growth rates have a tendency to decrease as the initial number of crimes increases ( fig .
[ fig : evolution ] and fig .
[ fig : distribution ] ) .
moreover , the mean value of standard deviations are well approximated by power - laws ( fig .
[ fig : law ] ) .
similar results on annual growth rates and power - laws have also been reported for other complex organizations but usually with different values of @xmath50 exponents @xcite .
this parallel between the criminality with complex organizations suggest that there are similarities in their growth dynamics . as pointed out ,
there is a universality in the way the fluctuations in growth rates of homicides in cities are distributed , even though they differ in many aspects , including the details of people education , public health , cultural development and ethnicity , for example .
our results also indicate that the power - law dependence in this social system could be due to some emerging scenario as well as it occurs in the growth rate dynamics of physical and biological systems , scientific publications , economic and religion activities . here , trying to encompass the above universal features and in a different perspective from several works that have applied a variety of techniques and models in the search for a characterization of crime behavior @xcite , we present motivations to stochastic models proposed in the context of complex organizations in the search for some insight on the study of crime .
although modeling is not the main objective of our empirical investigations , we discuss some aspects of our results that indicate a connection between criminality and the framework of complex organizations
. we can consider that complex organizations have an internal structure involving subunits @xcite
. also , qualitatively consistent with criminality , for something to occur other things must take place .
these facts go towards a multiplicative process for crime dynamics .
gibrat s model , initially proposed in the context of firm growth @xcite , plays the role of a first attempt to describe the growth of complex organizations . in our crime context , it can be written as @xmath54 , where @xmath9 is the number of homicides in a city in the year @xmath10 , @xmath55 is a random number following a gaussian distribution with zero mean and unitary standard deviation , and @xmath56 is a positive constant .
as we can see , this model assumes that homicide growth is independent of the number of homicides and uncorrelated in time .
unfortunately , this simple approach is not enough to describe our empirical findings such as the tent - shaped distributions in the growth rates and the asymptotic power - laws in the standard deviations .
however , its simplicity could be viewed as a first insight for the comprehension of mechanisms that govern the homicide growth rate .
generalizations of this model can take interaction among the subunits ( e.g. , non - linearity ) and relevance of previous number of homicides ( memory ) into account .
as a final remark of this brief discussion about modeling , we illustrate that , if non - linearity and memory are incorporated to gibrat s model , a substantial improvement is obtained from it . for this illustration , we follow ref .
@xcite , i.e. , we replace @xmath57 by @xmath58 , where @xmath59\,\epsilon(t)$ ] and @xmath56 , @xmath60 and @xmath61 are positive constants ( @xmath62 and @xmath63 recover gibrat s model ) .
the value of memory parameter @xmath60 basically dictates the degree that the probability distributions of homicide growth rates approach to tent - shapes . on the other hand ,
the parameter @xmath61 essentially fixes the @xmath46 value .
thus , this model is enough to reproduce two basic qualitative aspects of our annual homicide growth rate data : the tent - shaped form for the scaled distribution and the power - law behavior for the standard deviation .
our study on homicide growth rates gives a new quantitative characterization of criminality . using a framework of complex organizations
, we have investigated the relation between the number of homicides and the corresponding annual , biannual and triannual logarithmic growth rates for brazilian cities with at least one murder per year from 1980 to 2009 .
we have obtained two main empirical findings based on these three growth rates .
one is that the standard deviations of the growth rates as a function of the number of initial murders follows asymptotic power - law decays .
in addition , if the number of inhabitants is considered instead of the initial value of murders , the power - law behaviors of the standard deviations remains essentially unchanged .
the other main empirical result is that the distributions of growth rates are non - gaussian for distinct size scales but with the same shape , indicating the presence of scaling laws in the dynamics of crime growth rate .
this non - gaussian distribution is well described by the laplace distribution , where the small corrections can be accomplished by stretched gaussians . in a quantitative way ,
all these aspects are summarized in table 1 .
we report that similar empirical behaviors were also observed in a variety of other complex organizations , suggesting a universality and pervasiveness of patterns in their growth dynamics .
since this universal behavior is not dependent of details from the system , models from complex organizations are good candidates to be employed when investigating the crime dynamics . from the literature of these organizations , we consider a model that reproduces two basic qualitative aspects of our data : the non - gaussian distributions and the power - law exponent of the standard deviations .
the quantitative investigations of the emerging behaviors go in the direction that there are some interacting structures where random multiplicative processes take place . furthermore , since many works are still controversial due to difficulties inherent in data collection and also to technical issues in statistical treatment , our results can be a useful guide for further works on criminality modeling .
we are grateful to cnpq and capes ( brazilian agencies ) for financial support .
99 c. castellano , s. fortunato , v. loreto , statistical physics of social dynamics , reviews of modern physics 81 ( 2009 ) 591 - 646 .
n. boccara , modeling complex systems , springer , new york , 2010 .
f. havemann , m. heinz , r. wagner - dobler , firm - like behaviour of journals ?
scaling properties of their output and impact growth dynamics , journal of the american society for information science and technology 56 ( 2005 ) 3 - 12 .
s. picoli jr . , r.s .
mendes , l.c .
malacarne , e.k .
lenzi , scaling behaviour in the dynamics of citations to scientific journals , europhysics letters 75 ( 2006 ) 673 - 679 .
keitt , l.a.n .
amaral , s.v .
buldyrev , h.e .
stanley , scaling in the growth of geographically subdivided populations : invariant patterns from a continent - wide biological survey , philosophical transactions of the royal society of london b 357 ( 2002 ) 627 - 633 .
labra , p.a .
marquet , f. bozinovic , scaling metabolic rate fluctuations , proceedings of the national academy of sciences u. s. a. 104 ( 2007 ) 10900 - 10903 .
stanley , l.a.n .
amaral , s.v .
buldyrev , s. havlin , h. leschhorn , p. maass , m.a .
salinger , h.e .
stanley , scaling behaviour in the growth of companies , nature 379 ( 1996 ) 804 - 806 . l.a.n .
amaral , s.v .
buldyrev , s. havlin , m.a .
salinger , h.e .
stanley , power law scaling for a system of interacting units with complex internal structure , physical review letters 80 ( 1998 ) 1385 - 1388 .
d. fu , f. pammolli , s.v .
buldyrev , m. riccaboni , k. matia , k. yamasaki , h.e .
stanley , the growth of business firms : theoretical framework and empirical evidence , proceedings of the national academy of sciences u. s. a. 102 ( 2005 ) 18801 - 18806 .
m. ausloos , f. petroni , statistical dynamics of religions and adherents , europhysics letters 77 ( 2007 ) 38002 .
s. picoli jr . , r.s .
mendes , universal features in the growth dynamics of religious activities , physical review e 77 ( 2008 ) 036105 .
bettencourt , j. lobo , d. helbing , c. kuhnert , g.b .
west , growth , innovation , scaling , and the pace of life in cities , proceedings of the national academy of sciences u. s. a. 104 ( 2007 ) 7301 - 7306 .
bettencourt , j. lobo , g.b .
west , why are large cities faster ?
universal scaling and self - similarity in urban organizations and dynamics , the european physical journal b 63 ( 2008 ) 285 - 293 .
bettencourt , g.b .
west , a unified theory of urban living , nature 467 ( 2010 ) 912 - 913 .
bettencourt , j. lobo , d. strumsky , g.b .
west , urban scaling and its deviations : revealing the structure of wealth , innovation and crime across cities , plos one 5 ( 2010 ) e13541 .
a. gomez - lievano , h. youn , l.m.a .
bettencourt , the statistics of urban scaling and their connections to zipf s law , plos one 7 ( 2012 ) e40393 .
gordon , j.r .
iglesias , v. semeshenko , j.p .
nadal , crime and punishment : the economic burden of impunity , the european physical journal b 68 ( 2009 ) 133 - 144 .
iglesias , v. semeshenko , e.m .
scheneider em , m.b .
gordon , crime and punishment : does it pay to punish ?
physica a 391 ( 2012 ) 3942 - 3950 . p. crane , a. kinzig , nature in the metropolis , science 308 ( 2005 ) 1225 . r.w .
kates , t.m .
parris , long - term trends and a sustainability transition , proceedings of the national academy of sciences u. s. a. 100 ( 2003 ) 8062 - 8067 .
glaeser , b. sacerdote , why is there more crime in cities ?
journal of political economy 107 ( 1999 ) s225-s258 .
becker , crime and punishment : an economic approach , journal of political economy 76 ( 1968 ) 169 - 217 .
j. crane , the epidemic theory of ghettos and neighbourhood on dropping out and teenage childbearing , the american journal of sociology 96 ( 1991 ) 1226 - 1259 .
glaeser , b. sacerdote , j.a .
scheinkman , crime and social interactions , quarterly journal of economics 111 ( 1996 ) 507 - 548 .
i. ehrlich , crime , punishment , and the market for offenses , journal of economic perspectives 10 ( 1996 ) 43 - 67 .
s. d. levitt , using electoral cycles in police hiring to estimate the effect of police on crime , the american economic review 87 ( 1997 ) 270 - 290 .
kennedy , i. kawachi , d. prothrow - stith , k. lochner , v. gupta , social capital , income inequality , and firearm violent crime , social science & medicine 47 ( 1998 ) 7 - 17 .
messner , l. anselin , r.d .
baller , d.f .
hawkins , g. deane , s.e .
tolnay , the spatial patterning of county homicide rates : an application of exploratory spatial data analysis , journal of quantitative criminology 15 ( 1999 ) 423 - 450 .
a. blumstein , f.p .
rivara , r. rosenfeld , the rise and decline of homicide - and why , annual review of public health 21 ( 2000 ) 505 - 541 .
l. tedesco , la ata contra el vidrio : urban violence and democratic governability in argentina , bulletin of latin american research 19 ( 2000 ) 527 - 545 .
levitt , alternative strategies for identifying the link between unemployment and crime , journal of quantitative criminology 17 ( 2001 ) 377 - 390 .
k. keizer , s. lindenberg , l. steg , study shows how degraded surrounding can degrade behaviour , science 322 ( 2008 ) 1681 .
a. sachsida , m.j.c .
mendona , p.r.a .
loureiro , m.b.s gutierrez , inequality and criminality revisited : further evidence from brazil , empirical economics 39 ( 2010 ) 93 - 109 . j. crane , n. boccara , k. higdon , the dynamics of street gang growth and policy response , journal of policy modeling 22 ( 2000 ) 1 - 25 .
w. cook , p. ormerod , e. cooper , scaling behaviour in the number of criminal acts committed by individuals , journal of statistical mechanics : theory and experiment ( 2004 ) p07003 .
nuno , m.a .
herrero , m. primicerio , a triangle model of criminality , physica a 387 ( 2008 ) 2926 - 2936 .
short , m.r .
dorsogna , v.b .
pasour , g.e .
tita , p.j .
brantingham , a.l .
bertozzi , l.b .
chayes , a statistical model of criminal behaviour , mathematical models and methods in applied sciences 18 ( 2008 ) 1249 - 1267 .
gordon , a random walk in the literature on criminality : a partial and critical view on some statistical analyses and modelling approaches , european journal of applied mathematics 21 ( 2010 ) 283 - 306 .
short , a.l .
bertozzi , p.j .
brantingham , nonlinear patterns in urban crime : hotspots , bifurcations , and suppression , the siam journal on applied dynamical systems 9 ( 2010 ) 462 - 483 .
short , p.j .
brantingham , a.l .
bertozzi , g.e .
tita , dissipation and displacement of hotspots in reaction - diffusion models of crime , proceedings of the national academy of sciences u. s. a. 107 ( 2010 ) 3961 - 3965 .
http://www.datasus.gov.br/. accessed on april 2012 .
http://www.ibge.gov.br/. accessed on april 2012 .
b. efron , r. tibshirani , an introduction to the bootstrap , chapman & hall , new york , 1993 .
richardson , atmospheric diffusion shown on a distance - neighbour graph , the proceedings of the royal society a 110 ( 1926 ) 709 - 737 .
mendes , h.v .
ribeiro , f.c.m .
freire , a.a .
tateishi , e.k .
lenzi , universal patterns in sound amplitudes of songs and music genres , physical review e 83 ( 2011 ) 017101 .
r. gibrat , les ingalits economiques , sirey , paris , 1931 .
h. e. stanley , scaling , universality , and renormalization : three pillars of modern critical phenomena , reviews of modern physics 71 ( 1999 ) s358-s366 . | the increasing number of crimes in areas with large concentrations of people have made cities one of the main sources of violence .
understanding characteristics of how crime rate expands and its relations with the cities size goes beyond an academic question , being a central issue for contemporary society . here , we characterize and analyze quantitative aspects of murders in the period from 1980 to 2009 in brazilian cities .
we find that the distribution of the annual , biannual and triannual logarithmic homicide growth rates exhibit the same functional form for distinct scales , that is , a scale invariant behavior .
we also identify asymptotic power - law decay relations between the standard deviations of these three growth rates and the initial size .
further , we discuss similarities with complex organizations . |
the experimental observation of bose - einstein condensation ( bec ) in dilute trapped and supercooled atomic vapors @xcite has stimulated enormous experimental and theoretical interest in the physics of the weakly interacting bose gas ( see @xcite for a review ) .
especially intriguing is the role of dimensionality and two - body interactions .
lower dimensional atomic gases have been realized with strong quantum confinement in one or more directions .
these systems exhibit very peculiar properties , such as the exact mapping between interacting bosons and noninteracting fermions ( tonks - girardeau gas ) @xcite , because the role of quantum fluctuations and strong correlations is enhanced .
although these gases are very dilute , bec of trapped atomic gases is strongly influenced by two - body interactions . within a mean - field approximation ,
the strength of the interatomic interactions can be typically characterized by a single parameter , the @xmath0-wave scattering length @xmath1 , at very low temperatures .
this van der waals type interaction is isotropic and short - range in comparison to the average interparticle separation .
accordingly , bose - einstein condensates created thus far have been very accurately described by the gross - pitaevskii theory ( see @xcite for a review ) , a mean - field approach .
furthermore , it has been possible to tune the s - wave scattering length @xmath1 to essentially any value , positive or negative , by utilizing a magnetic atom - atom feshbach resonance @xcite .
studies of the influence of the interparticle interactions on the properties of bec have focused mainly on the short - range interactions @xcite . however , understanding new kinds of systems will require the investigation of the effects of additional interactions , since bose - einstein condensation of molecules has now been observed using new techniques for creating ultracold molecules @xcite .
in addition , ling _
et al . _
@xcite proposed an efficient method to convert atomic condensates into stable molecules using a generalized raman adiabatic passage scheme .
very recently , the generation of a bec in a gas of @xmath2cr atoms was reported @xcite .
as molecules can potentially have large dipole moments and a chromium atom has a very high magnetic moment of 6@xmath3 ( bohr magneton ) in its ground state , bec with dipole - dipole interactions in ultracold gases has attracted considerable theoretical attention over the past few years @xcite .
compared to the short - range interaction , the dipole - dipole interactions are long - range , anisotropic , and partially attractive . as a result ,
nonlinear effects due to long - range dipole - dipole interactions lead to new interesting properties ; different quantum phases such as superfluid , supersolid , mott insulator , checkerboard , and collapse phases @xcite , a self - bound bose condensate in the field of a traveling wave @xcite , a roton minimum in the excitation spectrum @xcite , and promising candidates for the implementation of fast and robust quantum - computing schemes @xcite .
first of all , yi and you @xcite suggested an effective pseudopotential for anisotropic dipolar interactions in the born approximation and recently derevianko @xcite derived a more rigorous velocity - dependent pseudopotential for dipole - dipole interactions .
et al . _
@xcite showed a strong dependence of the stability of trapped dipolar bec on the trapping geometry .
odell _ et al .
_ @xcite derived exact results in the thomas - fermi regime for the statics and dynamics of a trapped dipolar bec .
however , in all the theoretical studies so far the t=0 k situation has been investigated . in this paper
we address the properties of trapped condensates with dipolar interactions at finite temperatures .
several sources for dipolar bec are atoms polarized by an electric field or atoms with large magnetic moments in addition to polar molecules .
the magnitude and sign of dipole - dipole interactions can be changed by rapidly rotating an external field @xcite .
thus , by varying the strength of the dipolar interactions relative to the s - wave scattering ( tuning the dipolar interaction or changing the s - wave scattering length @xmath1 ) it is possible to study the effects of the dipolar interaction on a condensate in detail . in this paper we attempt to understand bose - einstein condensation of trapped dipolar gases , where the interparticle interaction consists of a long - range dipole - dipole interaction in addition to the usual short - range s - wave interaction ( a purely repulsive hard - sphere potential ) .
in particular , we demonstrate the influence of dipole - dipole forces on the properties of ultracold gases by comparing the energetics and structural properties of harmonically trapped atomic condensates with dipolar interactions with those for which only short - range interactions are present .
in addition , we study the dependence of the sign and the value of the dipole - dipole interaction energy on the trapping geometry by using two different types of cylindrical traps ( a cigar - shaped trap and a disk - shaped trap ) . for these purposes
, we use a finite - temperature path - integral monte carlo ( pimc ) method @xcite .
our paper is structured as follow .
section [ sec2 ] introduces the hamiltonian of trapped dipolar hard - sphere systems used in this studies and an effective pseudopotential to describe anisotropic dipolar interactions in the born approximation .
the path integral monte carlo technique briefly is also reviewed in sec .
section [ sec3 ] presents our simulation results .
finally , we conclude in sec .
we wish to study the problem of a quantum @xmath4-particle system to consider a harmonically trapped bose - einstein condensation in three dimensions .
the quantum particles have both a long - range dipole - dipole interaction and the usual short - range hard - sphere interaction .
we assume that all atomic dipole moments are equal and oriented along the @xmath5-axis ( trap axis ) . within the born approximation , the dipole - dipole interaction potential between two dipoles separated by @xmath6 is given by @xmath7 where @xmath8 characterizes the dipole moment , @xmath9 is the vector between the dipoles @xmath10 and @xmath11 , and @xmath12 the angle between the vector @xmath6 and the direction of the dipoles ( @xmath13 ) . to date , this pseudopotential has worked reasonably well for the long - range dipole - dipole interaction .
the corresponding hamiltonian is then @xmath14 where @xmath15 is the hamiltonian for trapped ideal boson gases and @xmath16 is the hard - sphere potential defined by @xmath17 in order to investigate the dependence of the sign and the value of the dipole - dipole interaction energy on the trapping geometry , we consider bose gases under a cylindrically harmonic confinement ( @xmath18 and @xmath19 ) , where @xmath20 is the aspect ratio . consider a system of @xmath4 dipolar hard spheres .
the statistical mechanics of a quantum system at an inverse temperature @xmath21 , where @xmath22 and @xmath23 denotes the boltzmann constant , is governed by the density matrix .
the bose - symmetrized density matrix is given by @xmath24 where @xmath25 and @xmath26 are two configurations of @xmath4 dipolar hard spheres , which denote @xmath27 dimensional vectors , @xmath25 = @xmath28 .
@xmath29 denotes a permutation of particle labels among hard spheres and @xmath30 is one such permutation .
the inclusion of particle permutations is crucial and superfluidity is an immediate consequence of bose symmetry .
however , evaluating the density matrix for interacting systems at very low temperatures is complicated by the fact that the kinetic and potential terms in the exponent of the density matrix do not commute . in order to avoid this problem , we insert @xmath31 intermediate configurations into eq .
( [ eq - dm ] ) : @xmath32 where @xmath33 = @xmath34 is the imaginary time step and @xmath35 is referred to as the number of time slices .
[ eq - dm2 ] is the path - integral formulation of the density matrix , which contains the discrete sum over permutations and multiple integrations over density matrices at a higher temperature @xmath36 .
. [ eq - dm2 ] can be evaluated in the path integral monte carlo technique by a stochastic sampling of the discrete paths @xmath37 using multilevel monte carlo sampling @xcite , which accounts for permutations . in order to use monte carlo sampling
, we must first provide a pair - product form of the exact two - body density matrices .
for the short - range hard - sphere potential , we used the high - temperature approximation for the hard - sphere propagator derived by cao and berne @xcite .
unfortunately , it is impractical to develop an analytic pair - product form for the anisotropic dipole - dipole interaction potential . thus the primitive approximation , which separates the density matrix into kinetic and potential energy terms , was used . in order to choose the value of the number of time slices @xmath35 , we performed consistency checks by varying @xmath35 to see that the results had converged .
we used up to 160 time slices ( @xmath35 = 160 ) depending on the temperature and the number of particles .
we employed the canonical ensemble , @xmath38 in each simulation we fixed the temperature @xmath39 , the number of particles @xmath4 , the diameter of a hard sphere @xmath1 , and the strength of the dipolar interactions @xmath8 .
the thermal expectation value of an operator @xmath40 is given by @xmath41 .
expectation values are calculated at the same time the discrete paths are sampled . typically 20,000 - 30,000 mc steps are required for equilibration and statistical averages are collected from 150,000 - 180,000 mc steps after this . in each mc step , we attempted 60 - 1,250 trial moves at each time slice .
multiple runs were performed for each set of parameters so that statistical error bars could be determined .
we performed these calculations partly on ibm power4 and partly on athlon processors at the university of georgia . for @xmath4 = 125 dipolar particles and @xmath33 = 0.00446
, the simulation ( thermalization + computing averages ) at a single temperature took about 2100 ibm power4 cpu hours . the density profile @xmath42 , which indicates the likelihood of finding particles at a given distance @xmath43 from the center of the trap , is defined as @xmath44 where @xmath45 denotes the distance between the position of the @xmath10th particle and the center of the trap at time slice @xmath46 and the symbol @xmath47 denotes an ensemble average .
the superfluid fraction along the axis of rotation @xmath48 is given by @xcite @xmath49 where @xmath50 denotes the projected area , @xmath51 and @xmath52 the classical moment of inertia , @xmath53 ( 0,380)(0,0 ) ( -130,200 ) ( -130,390 ) ( 0,190)(0,0 ) ( -130,200 )
the path integral monte carlo ( pimc ) method allows one to calculate accurate quantum mechanical expectation values of many - body systems , the only input being the many - body potential . here
, we use pimc to simulate a system composed of dipolar hard spheres for various values of the number of hard spheres @xmath4 and the trap aspect ratio @xmath20 at two fixed dipolar moments @xmath8 as a function of temperature @xmath39 and to obtain simulation results for a bose - einstein condensation with magnetic dipole - dipole forces .
this section discusses the energetics and structural properties of trapped dipolar hard spheres obtained from our pimc simulations . for simplicity , throughout this paper , we use a cylindrically symmetric harmonic trap with an aspect ratio @xmath20 , @xmath54 and @xmath55 .
the approximate critical temperature @xmath56 for an ideal gas of @xmath4 atoms in a trap is given by @xcite @xmath57^{2/3 } } \left(\frac{t^{3d}_{c}}{t^{3d}_{0}}\right)^{2}n^{-1/3 } , \\ & & t^{3d}_{0 } = \frac{\hbar\omega_{ho}}{k_{b}}\left(\frac{n}{g(3)}\right)^{1/3},\nonumber\end{aligned}\ ] ] where @xmath58 is the critical temperature for an ideal bose gas in a trap in the thermodynamic limit , @xmath59 is the riemann zeta function , and @xmath60 and @xmath61 are the arithmetic average and the geometric average of the trapping frequencies , respectively .
a measure of the strength of the dipolar interaction relative to the @xmath0-wave scattering in general is given by @xmath62 for chromium , the more common bosonic isotope , @xmath2cr , has @xmath63 @xmath64 0.089 and the less common bosonic isotope , @xmath65cr , has @xmath63 @xmath64 0.36 @xcite . in this study , in order that the effects of dipolar interactions can be made visible , we used @xmath66 = @xmath67rb ) , @xmath19 ( @xmath68 ) = 2@xmath69 77.78 hz , @xmath68 = 10 @xmath19 ( @xmath19 = 10 @xmath68 ) , the @xmath0-wave scattering length @xmath1 = 10 @xmath70 = 0.0433 @xmath71 ( @xmath72 ) , and the dipole moment @xmath8 = 0.2 @xmath73 ( @xmath74 ) for a cigar - shaped trap ( a disk - shaped trap ) as an example for numerical calculations , giving @xmath63 @xmath64 0.31 . here ,
@xmath70 = 100 @xmath75 ( bohr radii ) is the @xmath0-wave scattering length of @xmath76rb , @xmath71 = @xmath77 , and @xmath72 = @xmath78 . for these parameters , the short - range interactions and
the dipole - dipole ones have a comparable strength , and we can see the interplay between short - range and dipole - dipole interactions and the deviation between the case with and without dipole - dipole interactions .
the results are scalable .
moreover , the magnitude and sign of the values of @xmath1 and @xmath8 can be easily tuned by feshbach resonance @xcite and by an external field @xcite , respectively .
( 0,190)(0,0 ) ( -130,200 ) ( 0,200)(0,0 ) ( -130,200 ) ( 0,400)(0,0 ) ( -130,225 ) ( -130,410 ) ( 0,190)(0,0 ) ( -130,200 ) we first consider a 3d dipolar bose gas in a cigar - shaped trap with @xmath20 = @xmath79 = 10 and @xmath80 by increasing the confinement in the transverse direction .
the length unit is @xmath71 = @xmath77 and energies are measured in units of @xmath81 .
figure 1(a ) shows the calculated density profiles @xmath82 as a function of @xmath83 , normalized such that @xmath84 and @xmath85 , at @xmath39 = 0.4 @xmath56 for three different numbers of particles , @xmath4 = 27 ( circles ) , 64 ( diamonds ) , and 125 ( squares ) . for comparison ,
we used two values @xmath8 = 0 and @xmath8 = 0.2 .
filled symbols indicate the density profiles @xmath82 for @xmath8 = 0.2 and open symbols the density profiles @xmath82 for @xmath8 = 0 . in all figures ,
when statistical errors can not be seen on the scale of the figure , the error bars are smaller than the symbol sizes .
qualitatively , the density profiles for @xmath8 = 0 and for @xmath8 = 0.2 are similar in shape , as predicted . however , the density profiles @xmath82 for @xmath8 = 0 are broader than those for @xmath8 = 0.2 . the difference with respect to the density profile @xmath82 without dipolar interactions
is clearly seen as the number of particles increases .
the density profiles @xmath86 along the axial direction as a function of @xmath5 , normalized such that @xmath87 , at @xmath39 = 0.4 @xmath56 for three different numbers of particles are also shown in fig .
in contrast to the density profile @xmath82 , the difference in the density profiles @xmath86 between for @xmath8 = 0 ( open symbols ) and for @xmath8 = 0.2 ( filled symbols ) is small as the number of particles grows .
in addition , a thick solid line in fig .
1(a ) shows the density profile @xmath82 for an ideal bose gas for comparison .
the density profile @xmath82 for the interacting gas approaches that for an ideal gas as the number of particles @xmath4 decreases .
correspondingly the excitations in the longitudinal direction are largely frozen out and the system behaves quasi - one dimensionally .
our density profile studies support the facts that i ) as we add more particles into the gas , the density profiles expand ; the half - width of the density profile increases with the number of particles @xmath4 , as shown in the inset of fig 1(a ) and ii ) the net effect of dipole - dipole interactions in the cigar - shaped trap is attractive , _
i.e. _ shrinking of the density profiles is observed . to further characterize the behavior of particles with dipolar interactions at @xmath39 = 0.4 @xmath56 , fig .
2 shows the total energy per particle @xmath88 of the system as a function of the number of particles @xmath4 .
the total energy per particle is strongly affected by the particle - particle interaction and the overall dipole - dipole interaction is attractive in the cigar - shaped trap , because the total energy per particle for @xmath8 = 0.2 is smaller than that for @xmath8 = 0 . from fig .
2 we can clearly see the dependence on the number of particles @xmath4 , and numerical calculations have previously shown that the finite - size effects are significant for rather small values of @xmath4 @xcite .
the difference in the total energy per particle @xmath88 between for an ideal bose gas and for an interacting gas increases with increasing @xmath4 , which is a consequence of the expansion of the density profiles with interaction , as shown in fig .
1 , although the ideal bosonic gas energy per particle @xmath88 changes considerably less as the number of particles @xmath4 increases .
accordingly , the dependence of the total energy per particle @xmath88 on the number of particles @xmath4 reflects essentially the interaction energy alone , as the kinetic energy depends weakly on @xmath4 .
figure 3 shows the dependence of the density profiles @xmath82 on the temperature @xmath39 for @xmath4 = 64 . as the temperature increases , the effect of interactions is vanishingly small . at @xmath39
= 2.0 @xmath56 , the density profiles @xmath82 for @xmath8 = 0 and for @xmath8 = 0.2 are indistinguishable and the density profiles have the gaussian profiles of a pure thermal distribution .
finally , we calculate the superfluid fraction @xmath89 along the axis of rotation @xmath13 and in fig .
4 , we plot the superfluid fraction @xmath89 as a function of the number of particles @xmath4 at @xmath39 = 0.4 @xmath56 . in addition ,
4 shows the superfluid fraction @xmath89 for an ideal bose gas . in general , the superfluid fraction is essentially unity at very low temperatures and decreases gradually with increasing temperature . for the ideal gas ,
the superfluid fractions show at most a very weak dependence on the number of particles considered . however , there are significant finite - size effects for the interacting gas .
as the presence of repulsive interactions has the effect of lowering the value of the critical temperature , the superfluid fraction for the interacting gas is smaller than that for the ideal gas at the same temperature @xmath90 , as shown in fig .
the dependence of the superfluid fraction @xmath89 on the presence of interactions increases with the number of particles @xmath4 .
it is not easy , however , to see the difference between the superfluid fractions @xmath89 for @xmath8 = 0.2 ( filled circles ) and those for @xmath8 = 0 ( open circles ) because of the size of the errorbars but the difference must clearly be small , if any .
( 0,175)(0,0 ) ( -130,200 ) for the disk - shaped trap with the asymmetry factor @xmath20 = @xmath79 = 0.1 , _
i.e. _ increasing the confinement in the axial direction , the length unit is @xmath72 = @xmath78 and the energy unit @xmath91 . in fig .
5 , as we did for the cigar geometry , we plot the density profiles @xmath82 and @xmath86 as a function of @xmath83 and @xmath5 , respectively , at @xmath39 = 0.4 @xmath56 for three different numbers of particles , @xmath4 = 27 ( circles ) , 64 ( diamonds ) , and 125 ( squares ) using @xmath8 = 0 ( open symbols ) and @xmath8 = 0.2 ( filled symbols ) .
it is compared to corresponding density profiles @xmath82 and @xmath86 for a bose gas with short - range interactions only and for an ideal bosonic gas .
compared to the density profiles for @xmath20 = 10 , they are similar in shape . as @xmath19 is larger than @xmath68 , the density profile @xmath86 for an interacting gas approaches that for an ideal gas ( see fig 5(b ) ) ,
indicating the excitations in the axial direction are largely frozen out and the system behaves quasi - two dimensionally . in a disk - shaped trap ,
the density profiles expand along both the @xmath5-axis and the @xmath83-axis as we increase @xmath8 , in agreement with those found by other authors . correspondingly , the net contribution of long - range dipolar forces is repulsive , in contrast to the cigar - shaped trap .
simulations for other values of @xmath20 show that bec in a disk - shaped trap is stable , even in very asymmetric traps , because the dipole - dipole interaction energy always remains positive .
figure 6 shows the total energy per particle @xmath88 as a function of the number of particles @xmath4 both for interacting atoms with @xmath8 = 0 ( open circles ) and @xmath8 = 0.2 ( filled circles ) and for noninteracting particles ( open squares ) .
the total energy per particle @xmath88 for @xmath8 = 0.2 is larger than that for @xmath8 = 0 due to the increasing repulsive interactions .
the total energy per particle @xmath88 for an interacting bose gas increases more rapidly than that for an ideal bosonic gas due to the finite - size effect , as @xmath4 increases .
therefore , in the thomas - fermi limit the kinetic energy of the particles in the trap is negligible in comparison to the interparticle interaction energy and the trapping potential . in addition ,
7 shows the superfluid fraction @xmath89 along the axis of rotation @xmath13 as a function of the number of particles @xmath4 at @xmath39 = 0.4 @xmath56 .
the superfluid fraction @xmath89 for an noninteracting gas is also shown in the figure .
the superfluid fraction @xmath89 for @xmath8 = 0.2 ( filled circles ) is clearly smaller than that for @xmath8 = 0 ( open circles ) .
the superfluid fraction @xmath89 shows no dependence on the number of particles @xmath4 , whereas the interparticle interaction plays a significant role in decreasing of the superfluid fractions @xmath89 .
interestingly , when we compare the superfluid fraction for @xmath20 = 0.1 to that for @xmath20 = 10 , we find that the superfluid fraction in the disk - shaped trap is significantly smaller than that in the cigar - shaped trap because geometric arguments imply that more excited longitudinal modes are occupied for @xmath20 = 0.1 than for @xmath20 = 10 .
in this paper , bose - einstein condensation of trapped dipolar bose gases has been studied using a finite - temperature path integral monte carlo technique .
the quantum particles have a dipole - dipole interaction and a @xmath0-wave scattering interaction .
the dipolar potential energy is long - range , anisotropic , and partially attractive , compared to the short - range and purely repulsive @xmath0-wave scattering interaction . using pseudopotential forms to describe the interatomic interactions ,
we have calculated the equilibrium properties , such as the energetics and structural properties , of a system composed of @xmath4 dipolar hard spheres .
we find that for a cigar - shaped trap , the net effect of dipolar forces is attractive and shrinking of the density profiles is observed ; and for a disk - shaped trap , the net contribution of dipole - dipole interactions is repulsive and the density profiles expand .
accordingly , the net contribution of dipolar interactions depends on the trapping aspect ratio @xmath20 .
lbl m. h. anderson _
et al . _ ,
science * 269 * , 198 ( 1995 ) ; k. b. davis _ et al .
lett . * 75 * , 3969 ( 1995 ) ; c. c. bradley _ et al .
lett . * 78 * , 985 ( 1997 ) .
f. dalfovo , s. giorgini , l. p. pitaevskii , and s. stringari , rev .
71 * , 463 ( 1999 ) .
l. tonks , phys . rev . *
50 * , 955 ( 1936 ) ; m. girardeau , j. math , phys .
* 1 * , 516 ( 1960 ) ; e. h. lieb and w. liniger , phys . rev . * 130 * , 1605 ( 1963 ) ; b. paredes , a. widera , v. murg , o. mandel , s. folling , i. cirac , g. v. shlyapnikov , t. w. hansch , and i. bloch , nature ( london ) * 429 * , 277 ( 2004 ) .
e. tiesinga _
et al . _ ,
a * 46 * , r1167 ( 1992 ) ; s. inouye _
et al . _ , nature ( london ) * 392 * , 151 ( 1998 ) ; s. l. cornish _ et al .
* 85 * , 1795 ( 2000 ) ; t. loftus _
et al . _ ,
lett . * 88 * , 173201 ( 2002 ) ; k. m. ohara _ et al .
_ , science * 298 * , 2179 ( 2002 ) ; t. bourdel _
et al . _ ,
lett . * 91 * , 020402 ( 2003 ) .
m. greiner , c. regal , and d. s. jin , nature ( london ) , * 426 * , 537 ( 2003 ) ; s. jochim _
_ , science * 302 * , 2101 ( 2003 ) ; m. w. zwierlein _
et al . _ ,
lett . * 91 * , 250401 ( 2003 ) ; t. bourdel _
et al . _ ,
lett . * 93 * , 050401 ( 2004 ) .
h. y. ling , h. pu , and b. seaman , phys . rev .
lett . * 93 * , 250403 ( 2004 ) a. griesmaier , j. werner , s. hensler , j. stuhler , and t. pfau , cond - mat/0503044 ( 2005 ) . k. goral _
et al . _ ,
a * 61 * , 051601(r ) ( 2000 ) ; j .-
a * 64 * , 037601 ( 2001 ) ; s. yi and l. you , phys .
a * 63 * , 053607 ( 2001 ) ; k. goral and l. santos , phys
a * 66 * , 023613 ( 2002 ) ; p. m. lushnikov , phys . rev . a * 66 * , 051601(r ) ( 2002 ) ; u. r. fischer , phys . rev . lett . * 89 * , 280402 ( 2002 ) ; b. damski _
et al . _ ,
lett . * 90 * , 110401 ( 2003 ) ; s. yi and l. you , phys .
a * 67 * , 045601 ( 2003 ) ; a. v. avdeenkov _
et al _ , phys .
a * 69 * , 012710 ( 2004 ) ; s. yi and l. you , phys . rev . lett .
* 92 * , 193201 ( 2004 ) ; z. w. xie and w. m. liu , phys .
a * 70 * , 045602 ( 2004 ) . s. yi and l. you , phys .
a * 61 * , 041604(r ) ( 2000 ) .
a. derevianko , phys .
a * 67 * , 033607 ( 2003 ) .
l. santos , g. v. shlyapnikov , p. zoller , and m. lewenstein , phys .
lett . * 85 * , 1791 ( 2000 ) .
k. goral , l. santos , and m. lewenstein , phys .
lett . * 88 * , 170406 ( 2002 ) .
s. giovanazzi , d. odell , and g. kurizki , phys .
lett . * 88 * , 130402 ( 2002 ) .
l. santos , g. v. shlyapnikov , and m. lewenstein , phys .
lett . * 90 * , 250403 ( 2003 ) .
d. h. j. odell , s. giovanazzi , and g. kurizki , phys .
lett . * 90 * , 110402 ( 2003 ) ; s. giovanazzi and d. h. j. odell , eur .
j. d * 31 * 439 ( 2004 ) .
d. jaksch _ et al .
lett . * 85 * , 2208 ( 2000 ) ; d. demille , phys . rev .
lett . * 88 * , 067901 ( 2002 ) .
d. h. j. odell , s. giovanazzi , and c. eberlein , phys .
lett . * 92 * , 250401 ( 2004 ) ; c. eberlein , s. giovanazzi , and d. h. j. odell , cond - mat/0311100 .
s. giovanazzi , a. gorlitz , and t. pfau , phys .
lett . * 89 * , 130401 ( 2002 ) .
d. m. ceperley , rev .
67 * , 279 ( 1995 ) .
d. p. landau and k. binder , _ monte carlo simulations in statistical physics _
( cambridge university press , cambridge , england , 2000 ) .
j. cao and b. j. berne , j. chem .
phys . * 97 * , 2382 ( 1992 ) k. nho and d. p. landau , phys .
a * 70 * , 053614 ( 2004 ) .
e. l. pollock and d. m. ceperley , phys .
b * 36 * , 8343 ( 1987 ) . p. o. schmidt , s. hensler , j. werner , a. griesmaier , a. gorlitz , t. pfau , and a. simoni , phys .
lett . * 91 * , 193201 ( 2003 ) .
w. ketterle and n. j. van druten , phys .
a * 54 * , 656 ( 1996 ) . | the path integral monte carlo method is used to simulate dilute trapped bose gases and to investigate the equilibrium properties at finite temperatures .
the quantum particles have a long - range dipole - dipole interaction and a short - range @xmath0-wave interaction . using an anisotropic pseudopotential for the long - range dipolar interaction and a hard - sphere potential for the short - range @xmath0-wave interaction , we calculate the energetics and structural properties as a function of temperature and the number of particles . also ,
in order to determine the effects of dipole - dipole forces and the influence of the trapping field on the dipolar condensate , we use two cylindrically symmetric harmonic confinements ( a cigar - shaped trap and a disk - shaped trap ) . we find that the net effect of dipole - dipole interactions is governed by the trapping geometry . for a cigar - shaped trap ,
the net contribution of dipolar interactions is attractive and the shrinking of the density profiles is observed . for a disk - shaped trap ,
the net effect of long - range dipolar forces is repulsive and the density profiles expand . |
a persistent problem in understanding the absorbing material in ngc 4151 has been reconciling the vastly different gas columns inferred for the x - ray absorption and for the uv absorption .
the x - ray absorbing column varies between @xmath1 and @xmath2 .
bromage et al .
( 1985 ) estimated a total column for the uv - absorbing material of no more than @xmath3 .
the neutral hydrogen column is variable ( kriss et al .
the bulk of the absorption is in low column density gas with @xmath4 and doppler parameter @xmath5 .
any low - b component has a neutral column no greater than @xmath6 .
one possibility for reconciling these differences has been the recent success of warm absorber models for characterizing the x - ray absorption and the associated uv absorption lines in 3c 351 and ngc 5548 ( mathur et al . 1994 ; mathur et al . 1995 ) . in such models
the absorption arises in gas photoionized by the central engine ( e.g. , netzer 1993 ; krolik & kriss 1995 ) . the x - ray absorption is dominated by highly ionized species of heavy ions ( e.g. , o vii and o viii ) .
the total gas columns can be quite high ( @xmath1@xmath2 ) , with relatively low columns in the lower ionization species responsible for the uv absorption .
warm absorber models with a reflection component can fit the x - ray spectrum of ngc 4151 ( weaver et al .
1994a , b ) .
kriss et al .
( 1995 ) find that similar models can also account for the high ionization lines in ngc 4151 ( e.g. , o vi , n v , and c iv ) , but they can not simultaneously match the particularly strong absorption in lower ionization species such as h i , c iii , and si iv .
they conclude that a single - zone warm absorber is insufficient . to search for absorption components that might possibly be identified with the x - ray absorbing gas
, i examined archival high resolution ghrs spectra of the c iv and mg ii line profiles in ngc 4151 .
fig.1 shows the spectrum of ngc 4151 in the c iv region with 14 @xmath7 resolution obtained in 8486 s using grating g160 m of the ghrs on 28 october 1994 . a model consisting of an underlying power law continuum , three broad gaussian emission lines , and 8 c iv absorption line doublets fits the data well and gives @xmath8 for 1800 points and 50 free parameters .
although the deepest and broadest c iv doublet is saturated , the bottom of the line profile is not black .
either this gas only partially covers the source ( at the 90% level , both continuum and broad line ) , or 10% of the continuum flux is scattered around the absorbing region back into our line of sight .
narrow - line emission is visible on the red side of the c iv absorption trough .
this emission is apparently unabsorbed by the broad absorbing gas ; a final layer of absorbing gas , however , lying at the systemic velocity of ngc 4151 , absorbs the core of the narrow - line profile .
this is presumably the ism or halo of ngc 4151 .
the spectrum of the mg ii region at 10 @xmath7 resolution obtained in a 1414 s integration with grating g270 m of the ghrs on 29 october 1994 is shown in fig.2 .
the best fit to the modeled emission and absorption profile gives @xmath9 for 1438 points and 22 free parameters .
as with c iv , the mg ii emission was modeled with 3 gaussians .
seven mg ii absorption doublets are required .
table 1 gives the velocities , equivalent widths , doppler parameters , and column densities of each of the absorption components fit in the c iv and the mg ii spectra .
& & & & & & & & + & & + & & & & & & & & + # & @xmath10 & ew & _ b _ & @xmath11 & @xmath10 & ew & _ b _ & @xmath12 + & @xmath13 & ( ) & @xmath13 & @xmath14 & @xmath13 & ( ) & @xmath13 & @xmath14 + & & & & & & & & + 1 & @xmath15 & 0.514 & 294 & @xmath16 & & & & + 2 & @xmath17@xmath18 & 0.120 & @xmath19 & @xmath20 & @xmath21 & 0.143 & @xmath22 & @xmath23 + 3 & @xmath17@xmath24 & 0.642 & 203 & @xmath25 & & & & + 4 & @xmath26@xmath27 & 0.310 & @xmath28 & @xmath29 & @xmath17@xmath30 & 1.259 & @xmath31 & @xmath32 + 5 & @xmath33 & 0.083 & @xmath34 & @xmath35 & @xmath36 & 0.052 & @xmath37 & @xmath38 + 6 & @xmath39 & 1.026 & 163 & @xmath40 & @xmath41 & 1.116 & 235 & @xmath42 + 7 & @xmath43 & 4.018 & 234 & @xmath44 & @xmath45 & 0.852 & 176 & @xmath46 + 8 & & & & & @xmath47 & 0.329 & @xmath48 & @xmath49 + 9 & @xmath50 & 0.407 & @xmath28 & @xmath51 & @xmath52 & 0.134 & @xmath53 & @xmath54 +
for the absorption components intrinsic to ngc 4151 , i assume that the gas is photoionized by the active nucleus .
computing photoionization models similar to those discussed by krolik & kriss ( 1995 ) and kriss et al .
( 1996 ) , i search for ionization parameters and total column densities that match the mg ii and c iv columns seen in the data .
table 2 summarizes the column density ratios of each of the absorption components and the matching ionization parameters and total column densities .
the velocities are now relative to the systemic velocity of ngc 4151 ( @xmath55 , mundell et al .
1995 ) . & & & & + # & @xmath56 & @xmath57 & log _ u _ & log @xmath58 + & @xmath13 & & & @xmath59 + & & & & + 1 & @xmath60 & @xmath61 & @xmath62 & + 2 & @xmath63 & 0.12 & @xmath64 & 18.3 + 3 & @xmath65 & @xmath61 & @xmath62 & + 4 & @xmath17@xmath66992 & 3.73 & galactic & 20.3 + 5 & @xmath17@xmath66830 & 0.060 & @xmath67 & 18.1 + 6 & @xmath17@xmath66805 & 0.085 & @xmath68 & 18.2 + 7 & @xmath17@xmath66321 & 0.004 & @xmath69 & 19.9 + 8 & @xmath17@xmath66193 & @xmath70 & @xmath71 & 17.018.0 + 9 & @xmath26@xmath661 & 0.026 & @xmath72 & 18.6 + note that all the absorbing systems have fairly low ionization parameters .
none of the systems in which mg ii absorption is visible is a good candidate for association with the warm x - ray absorbing gas , which typically has high ionization parameters @xmath73 and high total column densities log @xmath74 ( weaver et al .
1994a , b ) . while components 1 and 3 might be possible candidates , note that component 1 is visible only at this single epoch .
it is absent from all other ghrs observations at both higher and lower flux levels ( weymann et al .
observations of higher ionization species such as si iv or n v are required to set more stringent constraints on the ionization parameters and the total column densities of components 1 and 3 .
bromage , g. , et al .
1985 , mnras , 215 , 1 kriss , g. a. , et al .
1992 , apj , 392 , 485 kriss , g. a. , et al .
1995 , apj , 454 , l7 kriss , g. a. , et al .
1996 , apj , 467 , 622 krolik , j. h. , & kriss , g. a. 1995 , apj , 447 , 512 mathur , s. , et al .
1994 , apj , 434 , 493 mathur , s. , et al .
1995 , apj , 452 , 230 mundell , c. g. , et al .
1995 , mnras , 272 , 355 netzer , h. 1993 , apj , 411 , 594 weaver , k. , et al . 1994a , apj , 423 , 621 weaver , k. , et al .
1994b , apj , 436 , l27 weymann , r. j. , et al .
1997 , this volume | observations of the c iv and mg ii absorption lines in the seyfert 1 galaxy ngc 4151 obtained with the ghrs in october 1994 are presented .
the data from the stsci archive show multiple broad and narrow components in both species .
in addition to galactic absorption , four narrow and four broad systems associated with ngc 4151 are identified .
two broad systems dominate the total equivalent width , and their mean blueshift and width are comparable to the broad lyman line and continuum absorption seen in far - uv spectra from the hopkins ultraviolet telescope . narrow - line c iv emission is present on the red side of the broadest absorption trough , and narrow absorption at the systemic velocity of ngc 4151 , presumably in its own ism , absorbs the core of the narrow emission line .
strong mg ii absorption is present in all but two velocity systems .
ratios relative to the corresponding c iv components suggest a low ionization parameter for the absorbing gas : @xmath0 .
this makes none of the identified uv absorption systems a good candidate for association with the warm x - ray absorbing gas . |
an advantage of studying the cluster galaxy population to test theories of galaxy formation is that , once a galaxy has entered a cluster , it will likely remain within a high density environment .
we can therefore trace a more direct evolutionary connection between high redshift and local systems without needing to appeal to abundance matching to link populations at different epochs ( e.g. , @xcite _ et seq . _ ) , where the large intrinsic scatter in galaxy growth rates in simulations implies that one can not unambiguously identify galaxy progenitors and descendants between different observational epochs in the field @xcite .
the homogeneity of cluster samples offers a snapshot of galaxy properties at each epoch , providing a more direct link between galaxy progenitors and descendants . for this reason
, cluster galaxies may serve as an important benchmark for models of galaxy formation and evolution , as well as being useful cosmological probes .
the most distinctive and characteristic population of cluster galaxies at @xmath8 are the cluster red sequence galaxies ( crsgs ) .
they form such a distinctive population that their colours can be used to search for high redshift clusters in wide - field images ( e.g. , @xcite ) .
these predominantly early type galaxies have little or no ongoing star formation subsequent to forming the bulk of their stellar populations and assembling their stellar masses at high redshift ( e.g. , @xcite and references herein ) .
however , while their stellar populations and luminosities evolve passively , their morphologies may not , and may also evolve differently from those of similarly passive field galaxies .
for example , a population of dense and compact quiescent galaxies has been identified in the field at high redshift and appears to be much less common nearer the present epoch ( e.g. , @xcite ) .
cluster early type galaxies instead appear to have larger sizes than a similarly massive sample of quiescent field galaxies @xcite and to evolve more slowly @xcite .
galaxy size evolution may also have significant environmental dependencies @xcite . in our previous study ( @xcite , hereafter paper i )
, we explored the morphologies of a sample of crsgs drawn from four rich clusters at @xmath10 ( mean redshift @xmath11 ) .
we used deep archival _ hubble space telescope _ imaging in the rest - frame @xmath12 and @xmath13bands ( observed frame @xmath14 and @xmath15 ) . the latter passband probes the evolution of structural parameters in a rest - frame band more representative of the stellar mass distribution ( as opposed to earlier studies where the equivalent bandpasses are closer to @xmath16 or @xmath17 ) .
our data implied even weaker size evolution than previously measured ( e.g. , @xcite ) with no significant size growth compared to the virgo cluster .
while we confirmed the essentially passive evolution of crsgs in colour and luminosity , we also observed that a typical crsg ( with a luminosity around @xmath18 or fainter ) in these @xmath19 clusters has a considerably more disk - like morphology ( lower srsic index , more flattened light distribution ) than its present - day counterparts , structurally closer to that of field spirals . in agreement with this , @xcite also measured more oblate shapes for crsgs in a @xmath20 cluster partly included within the candels dataset , and argued for the presence of increasingly important disk - like components in high redshift cluster galaxies . in the field , luminous quiescent galaxies also appear to contain significant disk - like components @xcite , with at least some of these systems showing evidence of being rotationally supported @xcite .
one possible clue to the mechanisms that convert these high redshift ` disks ' to the more spheroid - dominated population of more local clusters is offered by the existence of significant radial colour gradients .
local early - type galaxies tend to show slightly bluer colours outwards , consistent with the existence of a metal abundance gradient ( e.g. , @xcite ) .
based on the measured colour gradients of @xcite it appears that cluster galaxies are already similar to local early - type galaxies in this respect by @xmath21 .
our sample in paper i showed significantly larger gradients ( also bluer outwards ) at @xmath22 , more consistent with the existence of age gradients and more similar to the evolution seen by @xcite in their sample of disks . together with the evidence for disk - like components and more flattened shapes ( e.g. , @xcite ; see figure 7 in paper i ) , we argued in paper i that a population of recently quenched thin disks transforms into the local population of crsgs ( dominated by e / s0 galaxies ) largely by mechanisms that do not change stellar mass ( given the observed passive evolution of the luminosity function ) or overall sizes ( as those of our sample were broadly consistent with local values as measured in the virgo cluster ) .
possible mechanisms that give rise to these morphological changes include minor mergers , harassment , fly - bys and tidal effects , among many others .
these objects may also be related to the red spirals found locally in the outskirts of the a901/2 supercluster by @xcite and in sdss clusters by @xcite and @xcite .
such ` anemic ' spirals with weak spiral structure and low star formation rates were originally identified by @xcite and @xcite in the virgo cluster .
these objects may represent the low redshift counterparts of our high redshify crsgs with significant disk components , and would argue that a process of transformation from field - like spirals to cluster e / s0 may have been taking place in a similar fashion across the past 2/3 of the hubble time .
this also suggests that most crsgs may contain residual disks , a hypothesis supported by earlier photometric work of @xcite and the recent atlas3d dynamical work @xcite .
here we extend this analysis by bridging the intervening redshift range with the full sample of clusters in the clash hst dataset @xcite .
thanks to the multi - band nature of clash , we are able to select our targets in the same way as the targets in paper i , i.e. by rest - frame @xmath23 band luminosity and @xmath24 colour . as before , we also measure structural parameters in ( restframe ) @xmath23 and colour gradients in @xmath24
. we can therefore choose objects over similar magnitude ranges and with very similar stellar populations at all redshifts ( tied by the observed passive evolution in both quantities ) and measure them in the same fashion as our higher redshift galaxies in paper i. this yields a comprehensive survey of crsg structural evolution for objects brighter than @xmath25 ( where m@xmath26 is the magnitude corresponding to @xmath18 in the luminosity function ) , over the last @xmath27 gyr , drawn from a broad sample of clusters , treated in a uniform manner across redshift ( as regards to object selection , use of the same passbands and style of analysis ) .
we discuss the dataset and our analysis in the next section .
results including photometry ( in ab system , unless otherwise stated ) , measurements of sizes , srsic indices , ellipticities and colour gradients are presented in section 3 .
the results are discussed and placed in context _ via _ a simple evolutionary model in section 4 .
conclusions are presented in section 5 . throughout this paper
we assume a cosmology based on the latest release of parameters from wmap9 @xcite .
the sample of clusters from which all our crsgs are drawn consists of all available systems targeted in the multi - wavelength clash treasury project @xcite .
we supplement this with four high redshift clusters ( xmm1229 @xcite , rdcs1252 @xcite , iscs1434 @xcite and xmm2235
@xcite ) with suitable archival data at @xmath28 as already discussed in paper i. see table [ tab ] for basic information .
the small number of higher redshift objects in our sample is due to the lack ( at present ) of suitable infrared imaging datasets for @xmath29 clusters in the hst archive .
.clusters studied in this paper split into the four redshift dependent groups discussed in the text . [ cols="<,<,<",options="header " , ] clash clusters were originally selected as representative of the most massive and dynamically relaxed systems at @xmath30 ( as discussed in @xcite ) , with bright and diffuse x - ray emission , and have typical halo masses in excess of 10@xmath31 m@xmath32 .
the objects studied in paper i are also massive clusters , already possessing hot x - ray gas atmospheres , and have richnesses in excess of abell class 1 at @xmath33 ; they will evolve into very rare ( @xmath34 mpc@xmath35 ) systems of richness class above 2 in the present - day universe . unlike previous studies that select for overdensities in wide - field images ( e.g. , @xcite ) , which are too small to identify rich systems ,
our targets are chosen as massive clusters at their respective redshifts , likely among the most massive systems in the universe at their epoch , tracing the highest peaks of the dark matter distribution .
for comparison , the largest structure studied by @xcite , corresponding to the system identified by @xcite , only has a velocity dispersion of @xmath36 km s@xmath37 @xcite and does not possess extended x - ray emission @xcite , indicating a far lower environmental density than any cluster in our sample .
photometry is carried out exactly as in paper i , running sextractor @xcite on the individual images with a set of parameters that were determined from previous experience ( e.g. , @xcite , paper i ) to be optimal for galaxy detection and photometry .
for each cluster we select the clash observations that best match the rest - frame @xmath17 and @xmath23 at the respective redshift , as in paper i , i.e. the choice of bands depends on redshift .
this strategy allows us to directly compare sizes , srsic indices , axial ratios and colour gradients , while minimizing uncertainties due to @xmath38- and @xmath39- corrections which often affect field studies where structural parameters are determined in several different rest - frame bands ( cf .
the discussion in @xcite and @xcite on the dependence of structural parameters on bandpass ) .
all data were retrieved as fully drizzled frames from the hubble legacy archive .
all calibrations are on the ab system using the appropriate zeropoints from the stsci server .
all detections were visually inspected to check for defects , arcs , spurious objects ( e.g. , fragmented bright galaxies ) and other contaminants ( such as diffraction spikes ) .
stars were removed using their constant light concentration as a function of luminosity . images in clash are already registered to a common frame , but galaxies in the higher redshift sample ( paper i ) were position matched to within @xmath40 in the acs and wfc3 images .
( assuming a passively evolving model as described in the text ) , and our @xmath33 sample ( of four clusters ) in paper i ( symbols as identified in the legend ) compared to the local cluster sdss @xmath41 band red sequence luminosity function as derived by rudnick et al .
( 2009 ) , with arbitrary normalization to best show the match in shape.,scaledwidth=55.0% ] within each dataset we observe a prominent red sequence , due to the passively evolving population of massive galaxies in the clusters targeted , reaching to the photometric limit of the data and showing no apparent deficit of galaxies at low luminosities . as with all clusters out
to even high redshifts , we observe the expected thin and well - defined red sequences ( e.g. , the @xmath42 cluster by @xcite).these are well separated from their surroundings in colour - magnitude space and therefore serve to isolate a sample of quiescent cluster members efficiently , especially to the more conservative magnitude limits we adopt for analysis . fig .
[ examplecmr ] shows the derived colour magnitude diagram in ( restframe ) @xmath24 vs. @xmath23 for galaxies in the fields of four clash clusters at @xmath43 . for these clusters ,
the observed bands are @xmath44 and @xmath45 ( at @xmath46 ) , equivalent to central wavelengths of 4310 and 7250 respectively , compared with 4400 and 6900 for the rest - frame @xmath17 and @xmath23 bands , respectively .
data for the highest redshift clusters are shown in paper i ( e.g. , rdcs12522927 where the red sequence is very prominent ) , while we show the corresponding colour - magnitude diagrams for all other clash clusters as supplementary information ( on - line only ) in the same fashion as fig .
[ examplecmr ] .
we select galaxies within @xmath47 mag . of the mean colour - magnitude relation for the red sequence , as determined by fitting a straight line to the data by linear minimum absolute deviation @xcite . in the absence of extensive spectroscopic information
, the red sequence selects a sample of cluster members with comparatively minimal contamination by galaxies in the foreground and background ( cf .
while this means that some contamination may be present , to the limits of our analysis its effect is not overly significant .
for example , fig .
[ examplecmr ] shows that to @xmath48 ( more than one magnitude fainter than our typical analysis limit ) the red sequence of a subset of @xmath49 clusters appears to be well separated from other objects : similar conclusions can be drawn for other clusters ( e.g. , see the equivalent figure in paper i ) .
in this and equivalent figures , the galaxies we have eventually studied ( for structural analysis ) are marked as red dots ; a few objects in each cluster ( even if they were on the red sequence ) could not be modelled or the residuals seemed to indicate a poor fit .
as these represent only a small fraction of our sample , we have not considered them further .
the red sequence has been commonly used as an indicator of cluster membership , at least for the dominant quiescent population of ( typically ) e / s0 galaxies . as a test of this method , two of our clusters have extensive public redshifts from the clash - vlt survey . in macs1206
, 62 of our 100 galaxies on the red sequence have a redshift measured from @xcite and rosati et al .
( in preparation ) . of these 61
are cluster members . in macs0416 , 122 of the 158 galaxies we study on the red sequence
also have a redshift from @xcite and of these 118 are cluster members .
therefore , our method selects cluster members with a fidelity exceeding 95% .
of course higher redshift clusters , and fainter galaxies in any clusters , may suffer from somewhat higher contamination , as we have discussed in paper i. since the redshift completeness , even for the clusters in clash - vlt , is no greater than 60% , and not all clash clusters are observable from vlt , we still need to use red sequences to select a sample of cluster members in all our clusters . by construction
, our red bands are chosen to match the rest - frame @xmath23 band , with a small colour term ( e.g. , as in the example above in figure 2 , the central wavelength of the adopted filter is only 5% different from the actual @xmath23 band ) .
this colour term is however difficult to calculate accurately , as it depends sensitively on the spectral energy distributions of galaxies and their evolution . in order to be able to compare the luminosity distributions at different redshifts
, we corrected all magnitudes to the rest frame @xmath23 at @xmath50 ( the mean redshift of the five closest clash clusters ) to provide a reference magnitude , denoted as @xmath51 . we assume a passively evolving model from @xcite with @xmath52 , solar metallicity and an e - folding time of 1 gyr .
previous work has shown that this model yields a good fit to the observed colours of crsgs across a wide range of redshifts @xcite .
we group the clash clusters into three subsamples : ( i ) fourteen clusters at @xmath53 , the ` low ' subsample ; ( ii ) nine at @xmath54 , the ` mid ' subsample ; ( iii ) two at @xmath55 and @xmath56 , the ` high ' sample and combine these with the ` very high ' @xmath10 sample from paper i ( see table [ tab ] ) .
the relatively small size of the ` high ' sample is due to the paucity of systems observed in the appropriate bands with hst ( acs f814w and wfc3 f125w or f160w ) .
the number of crsgs in each subsample ( identified via the red sequence ) are 1037 for ` low ' , 743 for ` mid ' , 136 for ` high ' and 87 galaxies for ` very high ' ( paper i ) to @xmath57 .
we select galaxies on the red sequence of each cluster and calculate the average number of objects in each of the ` low ' , ` medium ' , ` high ' and ` very high ' redshift subsamples , within 0.5 mag .
bins in @xmath51 down to a magnitude where we believe contamination to be minimal and red sequences are well isolated in the surrounding colour space ( see fig . 1 ) .
we plot the average counts ( i.e. , divided by the number of clusters in each subsample ) in fig .
[ rsld ] vs. @xmath51 , omitting the error bars to avoid confusion ( but the errors are poissonian , since these are simple counts and do not include the clustering contributions from the field ) .
this may be considered as a luminosity _ distribution _ rather than a luminosity _ function _ as we have not removed possible contamination from fore / back - ground objects ( all clusters are studied in different observed bands , not all of which have sufficiently deep , or even any , observations of comparison fields ) . for comparison
, we plot the @xmath58 @xmath59 band luminosity function of cluster red sequence galaxies ( corrected to our bandpass and redshift ) from the sdss @xcite .
the normalisation is chosen arbitrarily to best show the differences in the shapes of the red sequence luminosity distribution across the redshift ranges we study and to provide a local comparison ( this normalisation is usually left to float as it does not have a clear physical meaning for clusters ) .
the average composite luminosity distributions for the crsgs drawn from the ` low ' , ` mid ' and ` high ' subsamples are consistent with the passively evolved schechter functions for local and very high redshift clusters having @xmath60 ( the characteristic magnitude @xmath61 measured in the @xmath23 band shifted to @xmath50 as described above ) and @xmath62 .
the mean number of galaxies in the clash sample in each magnitude bin are also comparable , suggesting that we are typically observing objects of comparable environmental density .
the number counts of galaxies in the high redshift sample of paper i are also broadly consistent with those measured for clash clusters , suggesting that we are not observing objects with very different properties across the redshift range of interest . given this selection , brighter than an evolved @xmath51 of about 21 ( approximately @xmath63 , or down to about 1/5th the luminosity of the milky way ) we are studying objects that have broadly the same stellar mass and passively evolving stellar populations at all times out to @xmath19 .
the good match to the local luminosity function of @xcite and to the results of paper i confirms the essentially passive evolution of red sequence galaxies in clash clusters . while these results are already well known , their recovery confirms that our selection of the photometrically - defined red sequence galaxies as cluster members is appropriate for the purposes of this study .
therefore , our sample consists of similar galaxies at all redshifts , linked by the observed passive evolution on the red sequence .
to @xmath64 we have a complete sample where we study galaxies of similar luminosity and colour at all redshifts with relatively little contamination from non - cluster sources .
we determine structural parameters for the crsgs in the rest - frame @xmath17 and @xmath23 using galfit @xcite as in our previous analysis of @xmath19 clusters in paper i. for each galaxy we carry out an interactive and iterative process , where we adopt reasonable starting values and proceed to inspect the model and residuals to ascertain the quality of the fit ( i.e. , we do not run galfit in batch mode , but individually for each object ) .
where necessary , multiple models are run together , e.g. , in the case of close pairs or systems of galaxies .
we adopt a single srsic profile .
this is warranted by the relatively shallow clash data and can also be more easily compared to previous work in cluster and field galaxies . for each image and in each band
we use the median of several bright , unsaturated and isolated ( as possible ) stars to derive a psf ( point spread function ) reference , with min / max clipping of eventual interlopers in the chosen field .
typical psf image sizes are 100 times the fwhm of the psf ( as recommended by the galfit manuals ) .
this procedures ensures that the psf incorporates all processing steps to which the images of the galaxies have been subjected and is therefore preferable to generating artificial psfs ( e.g. , with tinytim ) . for each galaxy being considered the sky level is determined locally within the large aperture needed for analysis by galfit ( including assessment of eventual gradients by fitting a surface to the sky see below ) .
the results of this procedure are the effective radii ( in restframe @xmath17 and @xmath23 ) , the srsic indices ( also in both bands ) and the axial ratio b / a ( also in @xmath17 and @xmath23 restframe ) . from these
we derive the colour gradient @xmath65 , following the formula given in @xcite using the measured structural parameters in each band . particularly for the higher redshift samples ,
the pixel scale of the images ( @xmath66 to @xmath67 ) is a significant fraction of the effective radius ( at @xmath68 , @xmath69 corresponds to 8 kpc , with typical galaxies having @xmath70 of 1 kpc , e.g. , @xcite ) and the psf full width at half maximum ( fwhm ) has a size comparable to the effective radius ( @xmath71 to @xmath67 depending on the bandpass and instrument used ) .
any 2d profile , e.g. produced with the iraf ellipse package @xcite , will be heavily affected by sampling and by the psf size .
furthermore , mismatches in the galaxy centroids and the different psfs in the red and blue bands ( often using two different instruments : acs in the optical and wfc3 in the infrared ) will result in spurious colour gradients ( e.g. , @xcite ) .
using fits to structural parameters from galfit modelling of the full image , automatically removes these sources of error .
we therefore use our derived parameters in the subsequent analysis rather than attempting more detailed profiling .
it is known that the errors output by galfit can sometimes be significant underestimates in the case of small images .
however from the work of @xcite who simulated hst images very similar to those used here , in terms of magnitude , surface brightness and size , it can be seen that galfit will provide values of @xmath70 accurate to within 5% down to magnitude 22.5 , even for steep ( @xmath72 ) profiles .
errors in @xmath73 are @xmath74 at @xmath72 and much less at @xmath75 , while ellipticity errors are typically 0.04 . errors of this size , for the faintest objects we analyse , are sufficiently small that they do not influence any of our results . in addition , we have run a series of simple simulations of our own with iraf artdata to test our procedures .
these introduce idealised galaxies ( pure exponentials or pure spheroids obeying the devaucouleurs law ) in ` clean ' regions of the images . using galfit on the simulated images
confirmed that parameter errors were of the level noted above or smaller ( the dispersion in values was consistent with the errors returned by galfit , i.e. , within 1% ) and that galfit parameters could be used safely down to the faint limit of our analysed data .
we have further introduced artificial sky gradients to simulate the effects of intracluster light and galfit is found to be insensitive to any such smooth gradients , unless they vary rapidly across the galaxies over scales comparable to the effective radius ; even then , the result is not incorrect parameters but catastrophic failures of the iteration .
when this does not happen , the errors returned are consistent with the galfit output , i.e. , @xmath76 .
since we actually inspect each object ( model and residuals ) individually , we can rule out the presence of such pathological sky features at any levels that would interfere with the fitting procedure .
fig . [ sizes ] shows the size distribution , vs. @xmath23 band magnitude ( shifted to @xmath50 assuming the concordance cosmology and a passively evolving model as described above ) , for all galaxies in the clash cluster sample and the sample from paper i. the appearance of the distribution is similar to that found in local systems ( e.g. , coma in @xcite ) , with a cloud of systems at intermediate sizes of @xmath77@xmath78 kpc and a brighter sequence of large , luminous galaxies ; these latter objects tend to be high @xmath73 early - type spheroid - dominated galaxies ( classical e / s0 ) .
we compare the size distributions of galaxies brighter than @xmath57 in the right - hand panel of fig .
[ sizes ] ( including the @xmath33 sample in paper i ) . in order to take into account both the shape and absolute numbers of objects contributing to each distribution in fig .
[ sizes ] , each was compared to the others using a standard kolmogorov - smirnov ( ks ) test .
the ` very high ' distribution differed from the ` low ' and ` middle ' at the 3@xmath79 and 2@xmath79 levels ( @xmath80 and 0.06 ) respectively ( assuming a gaussian distribution ) i.e. unlikely to be drawn from the same distribution .
similarly , the high " distribution differed from the ` low ' and ` mid ' distributions at the @xmath81 ( p=0.02 and 0.04 ) levels respectively .
below @xmath82 there appears to be no significant size evolution , while for higher redshifts , we can characterise the change by the typical @xmath70 decreasing by _ at most _ a factor @xmath83 over @xmath84 given that the peak of the size histogram shifts by at most one bin ( 0.2 dex ) in fig .
[ sizes ] .
the large sizes of some of our brightest cluster galaxies are also consistent with the findings of @xcite and @xcite .
@xcite use both a method very similar to ours ( with galfit ) and also the iraf ellipse package from sdsdas , finding very similar results , suggesting that the measured sizes are insensitive to the precise modelling technique ( j. stott , private communication ) the results shown in fig .
[ sizes ] are broadly consistent with the form of size evolution identified by @xcite from ground - based near - ir imaging , through a comparison between systems at @xmath85 and @xmath86 ( @xmath87 ) with most , if not all of the evolution , taking place well above @xmath82 and probably beyond @xmath88 .
any size evolution in crsgs in the past @xmath89 gyr appears to be much smaller than that found in field samples of quiescent galaxies of comparable luminosity over the same redshift range ( @xcite , paper i ) .
@xcite observes that at @xmath90 early - type galaxies have roughly the same size as local ones , in agreement with our findings for the ` low ' , ` mid ' and ` high ' subsamples .
@xcite detects no more than a 30% change in the size of red sequence galaxies with stellar masses higher than @xmath91 @xmath92 between @xmath86 and @xmath42 ( also see a forthcoming paper by s. andreon , private communication ) .
@xcite and @xcite , among others , however find that the average field early type galaxy increases its @xmath70 by a factor of 3 or more , mainly since @xmath9 , with growth continuing all the way to @xmath86 , unlike this cluster sample .
the degree of size evolution appears to depend on environment : @xcite and @xcite also find evidence that size evolution is weaker in high density environments ( by as much as 50% in @xcite when looking at their densest systems ) .
however , note that our clusters are considerably more massive than any of the objects studied by @xcite or @xcite . given the already known passive evolution in colour and luminosity , the small change in size observed for these galaxies appears to suggest early formation and assembly , with little subsequent evolution , especially by mergers , and most other changes being due to internal processes or more gentle external inputs , a conclusion also reached by @xcite at least for the more massive spheroidals in the field . as in the previous subsection
, we now plot the evolution of rest - frame @xmath13band srsic indices ( @xmath73 ) as a function of redshift for crsgs in fig .
[ sersic ] .
the corresponding histograms for the distribution of @xmath73 values is shown in the right - hand panel of fig .
[ sersic ] , including the @xmath10 data in paper i. at low redshift ( @xmath93 ) , clusters are dominated by a population of spheroidal galaxies , with @xmath94 corresponding to classical e / s0 galaxies as expected , with very few objects with @xmath95 .
however , at higher redshifts an increasing fraction of the population have low srsic indices consistent with the presence of disk - like components ( fig .
[ sersic ] ) .
whereas for @xmath96 about 30% of galaxies are observed to have @xmath97 and the median @xmath73 is @xmath98 , this fraction rises slightly ( to about 35% ) in the @xmath99 sample and in the highest redshift subsample more than 50% of galaxies have @xmath97 and the median @xmath73 is 2.2 , with only brighter cluster members tending to be @xmath100 spheroids . again , ks - tests were carried out between each of the subsamples .
the srsic index distribution of the ` very high ' subsample differed from each of those of the other three subsamples at more than the 5@xmath79 level ( [email protected] ) .
the ` high ' and ` mid ' distributions showed no convincing differences from each other , but both differed from that of the ` low ' subsample at the approximately 2@xmath79 ( p=0.04 ) level , implying ongoing evolution towards the low redshift distribution to about @xmath6 @xmath102 .
looking at fig [ sersic ] any evolution below @xmath88 is likely to be more subtle than that at higher redshifts and may only require evolution in a subset of the galaxies to achieve the observed differences . inspection of the @xmath17 band images confirms this impression . in paper
i ( see fig .
7 ) we saw that most @xmath3 galaxies have significant disk - like components , even though they lie on the red sequence . in this paper , we plot similar postage stamps of all such galaxies in cl1226 + 3332 ( @xmath56 ) , the highest redshift cluster in the clash sample , in fig .
[ cl1226 ] , showing that disk - like objects start to become significant at this redshift , including some slightly clumpy systems .
both @xcite and @xcite also observe numerous faint disk - like galaxies in two @xmath68 clusters .
this reinforces our previous finding : red sequence galaxies in high redshift clusters appear to contain an increasingly important disk - like component at higher lookback times . while local s0 galaxies certainly have disks , their overall srsic indices are still typically @xmath103 .
consequently this disk - like component at high redshift must eventually either disappear or evolve by @xmath82 where the population closely resembles that of present - day clusters .
the end - state of these high redshift crsgs ( with significant disks ) must be the s0 and elliptical galaxies that dominate at higher @xmath73 and lower redshifts .
in some respects , this finding resembles the observations by @xcite in the general field , where apparently quiescent galaxies exhibited an increasing contribution from disk - like systems at @xmath104 .
similarly , @xcite observe bright @xmath105 spheroids formed at high redshift , and an increasing proportion of irregular and thin disks at lower luminosities and higher redshifts . in order to ascertain the nature of the disk - like components we next look at the shape of the objects via their axial ratios .
below @xmath82 our clusters contain mainly , if not solely , galaxies with effective radii , srsic indices and axial ratios ( see below ) consistent with @xmath86 systems .
such objects likely resemble conventional e / s0 galaxies ( to the extent that such classification is objective and appropriate , considering the nature of the data ) .
we do not carry out a study of such classifications here , as we are interested in the change in structural parameters and quantifiable measures of evolution . in fig .
[ axial ] we plot the distribution of b / a ( axial ratios ) vs. @xmath51 magnitude for all our clusters in the same fashion as for sizes ( fig .
[ sizes ] ) and srsic indices ( fig [ sersic ] ) .
we also plot the corresponding axial ratio distributions in fig .
[ axial ] . at @xmath93
the axial ratio distributions indicate a population dominated by objects with low apparent ellipticity , either true ellipticals or systems with comparatively thick disks such as s0s .
the distribution peaks close to the local ( @xmath86 ) values in @xcite , i.e. @xmath106@xmath107 . the axial ratio distributions of the two higher redshift populations instead appear by eye to be flatter , with a significantly higher proportion of systems with projected axial ratios reaching down to @xmath108 .
such distributions are unlike those achievable through a mixture of ellipticals and typical s0 galaxies @xcite and again indicate an additional population with significant disk - like components at @xmath109 with an axial ratio distribution closer to that of field spirals @xcite : the mean axial ratio in the highest redshift bin is 0.58 , closer to that observed in spiral disks @xcite .
ks - tests indicate significant differences between the ` very high ' and both the ` mid ' and ` low ' subsamples at about the 2@xmath79 levels ( p=0.05 and 0.09 respectively ) , but no other significant difference .
@xcite also finds some evidence for increasing flattening with redshift , implying a more prominent role for disk - like components in galaxies , while @xcite find an increase in the fraction of oblate objects at @xmath33 .
however , unlike their results we have an increasing fraction of flatter galaxies at the lower luminosity end in the higher redshift bins . @xcite already noted the presence of elongated galaxies on the red sequence for a @xmath20 cluster , while @xcite has found no change in the shape of cluster galaxies _ selected as s0 _ with redshift , which implies that our flattened ` disks ' might eventually thicken and become rounder , in order to join the population of present - day s0s . a further clue to the evolution of galaxies may be provided by radial colour gradients .
these reflect the 2d distribution of stellar populations as well as their kinematics .
the existence of colour gradients in nearby ellipticals ( e.g. , @xcite ) is best explained as being due to a mild metal abundance gradient with radius ( richer inward ) and its apparent dependence on galaxy luminosity ( e.g. , in coma : @xcite ) suggests a model where the majority of the stellar populations were assembled early in rapid mergers or dissipational collapses , as a more extended hierarchical merger scenario would tend to erase such gradients @xcite . in paper i we showed through the ratio of rest - frame @xmath12 and @xmath13band @xmath70 that the @xmath10 crsgs had significant radial colour gradients , becoming bluer with radius .
the strength of these gradients was such that they could not be explained by abundance variation , implying that they were a symptom of radial age gradients in the stellar population of each galaxy .
following paper i , we have derived colour gradients for our galaxies , extending the treatment by converting the observed @xmath70 and @xmath73 in the two @xmath17 and @xmath23 bands to a more conventional @xmath110 @xcite . to do this we determine the colours at 1 and 2@xmath70 , sufficiently large radii to not have the measurement compromised by either the psf or pixel size , even for the most distant objects . in order to display the results of this procedure and
gain insight into how the gradients depend on stellar mass , for each redshift range we show the derived colour gradients and best fitting straight line from a biweight estimator ( e.g. , @xcite ) in fig .
[ gradients ] , plotted vs. @xmath51 as in previous figures .
the units of the colour gradients are magnitudes per decade in radius .
the statistical uncertainty on each derived value is low ( typically @xmath111 mag per decade in radius ) given the errors on @xmath73 and @xmath70 from galfit ( typically less than 1% ) .
consequently the scatter in and breadth of the distributions is a real reflection of the scatter in the properties of the objects at each redshift . at lower redshifts ,
the mean values of the colour gradient at each @xmath13band magnitude are consistent with the usual mild colour gradients ( about 0.1 0.2 mag .
blueward per decade in radius ) seen in nearby clusters . at these redshifts we also confirm the existence of a mild trend with luminosity ( and presumably stellar mass ) extending over 34 magnitudes , in the sense that fainter galaxies have milder colour gradients , as seen in the coma cluster @xcite .
this relationship appears to persist at least out to @xmath82 in agreement with earlier studies by @xcite .
linear fits ( with minimum absolute deviation ) to the data are shown for the ` low ' and ` mid ' redshift bins in fig .
[ gradients ] . if we interpret the observed gradients as metal abundance gradients , then our data are consistent with a trend of metal abundance with stellar mass , in the sense of galaxies with steeper potential wells having more significant gradients , as in coma @xcite .
this is more consistent with an inner region dominated by old stellar populations formed dissipationally at high redshift , either by gas - rich early mergers or large - scale gaseous collapses ( e.g. , @xcite ) .
the picture changes at higher redshifts . at @xmath112
the gradient is typically -0.1 mag at @xmath113 while at @xmath109 this becomes 0.5 mag , but with a larger scatter . at @xmath109
the outskirts of galaxies seem to become increasingly blue relative to their centres and show an increasing object - to - object scatter in the strength of the gradient ( cf .
paper i,@xcite for a similar result ) . for this reason we do not carry out fits to the data in the ` high ' and ` very high ' bins in fig .
[ gradients ] .
the appearance of comparatively strong colour gradients at @xmath109 is consistent with the results at @xmath11 in paper i where the outskirts of the crsgs were clearly bluer than their centres ( even though they were globally red in colour by selection ) . at these redshifts
a significant minority of sources with @xmath114 appear to have gradients with strength more easily achievable through stellar age gradients , rather than variations in metallicity as is common at lower redshifts . in order to explore how the variation in the strengths of colour gradients maps onto potential differences in both age and metallicity of stellar populations in bulge and disk components
, we generated the luminosity and colour evolution with redshift for two possible components of the galaxies using @xcite ssp models .
we assumed a component that formed in a single burst at @xmath115 with solar metallicity and determined its variation in rest - frame @xmath24 colour and mass to light ratio at @xmath116 .
this can be taken to be representative of a maximally old bulge component ( or indeed a similarly old disk component ) . for purposes of this exercise
we assume that this represents a maximally red model for the central regions of one of our galaxies .
we then did the same for a component undergoing a similar burst at @xmath117 with either solar or 0.25 solar metallicity .
the lateness of this burst was chosen simply to give the maximally blue colour and most rapid colour evolution below @xmath118 rather than attempting to mimic any individual galaxy . in effect
, it places an upper bound on the variation in colour that can be attributed to variations in population age and metallicity across a galaxy .
the range of colour differences between the old and younger stellar populations , and its evolution with redshift demonstrates that the variation and evolution in colour gradients observed in the population of crsgs can be potentially explained through a combination of metallicity and age - related gradients .
the age - induced gradients decrease with time as both populations converge towards similar colours @xmath119 gyr after the most recent burst , revealing the underlying gradient due to the ( known ) radial metallicity gradients ( see also @xcite for a similar conclusion ) . while similar colours may in principle be achievable if the younger component included some ongoing star formation below @xmath118 , this behaviour is constrained by the lack of evidence for such activity in cluster red sequences below this redshift .
the differential colours shown in fig .
[ models ] are for two components of equal stellar mass . given that specific star formation rates in these galaxies are usually well below @xmath120
yr@xmath37 , it is difficult to see how the colour gradients observed at @xmath109 are achievable through ongoing star formation alone .
a similar argument is presented by @xcite who analyse one of the clusters we studied in paper i and reach consistent conclusions to ours : the presence of an age gradient superposed over a more ancient weak metallicity gradient , although they do not make the explicit connection to the presence of disk - like components .
studying red disks in the sdss @xcite also argue for a quenching timescale of around 23 gyr ( in agreement with our work and @xcite ) and the gradual morphological evolution of the red disks as they move towards cluster centres .
paper i explored the morphological differences between a sample of @xmath0 crsgs and their nearby counterparts .
a significant fraction of the reasonably bright ( @xmath121 ) high redshift systems had structural parameters that were consistent with more disk - like morphologies than those of their low redshift counterparts .
the brightest galaxies in the red sequence appeared to be different , in that their morphological parameters were already closer to local early type ellipticals , indicating any structural evolution was completed at earlier epochs for this brighter subsample .
what was not established in paper i was the timescale over which any morphological transformation occurred in the more typical red sequence galaxies .
the results within this paper help to determine this timescale by exploring the same structural parameters for crsgs drawn from clusters observed at several intervening epochs . a key result of this work appears to be that for most of the structural parameters in question , any strong evolution has largely been completed by @xmath122 , a look - back time of 56 gyr , and 34 gyr after the latest redshift ( @xmath5 ) that any star formation in the red sequence galaxies is likely to have been quenched ( e.g. * ? ? ?
* ) . at higher redshifts
there is good evidence ( albeit limited by the number of systems with useful hst data at @xmath123 ) that the relevant parameters are transitioning between the local distributions and those found in paper i. as the velocity dispersions are already very high in these massive clusters and the luminosity function evolution disfavours significant major merging , this process must take place secularly or by more gentle external inputs ( such as minor mergers or fly - bys ) .
all of this data is consistent with a simple picture outlined in paper i where , after quenching , @xmath121 red sequence galaxies do not evolve in colour and luminosity other than the passive evolution of their stellar populations , but a proportion undergo considerable morphological change over the past 2/3 of the hubble time .
the brightest and more massive cluster spheroids in this sample seem to have already arrived at their final configuration by the beginning of this period ( cf . , @xcite ) but fainter galaxies are likely to show significant evolution , analogous to the observations by @xcite in the candels field
. however , most of our discussion as to the former must be qualitative for now , as there are too few bright galaxies in the small number of high redshift clusters that we observe to draw statistically valid conclusions . within this picture , the crsg population as a whole appears to become slightly less compact with time ,
although the scale of this growth is significantly less than that seen for passively evolving field galaxies of similar mass over the same time period .
the evolution is reminiscent of that observed in previous studies @xcite , although it may be even weaker , compared to the field , because of the much higher environmental density ( if the trends observed by @xcite can be extended to clusters ) . at the same time crsgs
also become less disk - dominated with decreasing redshift .
the decreasing importance of the disk component is reflected in changes to all of the structural parameters , including the sersic index @xmath73 and the axial ratios .
similarly to @xcite and @xcite we observe an increasing contribution from disk - like components at @xmath104 and somewhat more oblate galaxies , with axial ratios closer to those of spiral disks ( cf . , @xcite ) .
the brighter and more massive galaxies may constitute a class apart .
brightest cluster galaxies , for instance , are often believed to follow a more distinct evolutionary history than the less massive cluster population .
we show the actual data in fig . 3 , 5 and 7 for all objects . however , if we consider only the 10% brighter ( and likely more massive ) systems , the values for @xmath70 and @xmath73 are consistent with no significant evolution ( i.e. , these bright ellipticals appear to have reached their final status early on in the history of the universe ) but the statistical power afforded by the small number of objects in the highest redshift samples does not suffice to draw strong conclusions on the evolution of these objects . here
we summarise a strawman model for this evolution which can explain the results described above , and poses further useful questions on the details of this evolution .
this model contains elements that will be familiar to anyone with a knowledge of the early studies of galaxy evolution ( e.g. * ? ? ?
the model need not apply to each individual galaxy on the red sequence , rather it describes the evolution necessary within a subset of crsgs in order to produce the changes in the distributions of structural parameters over time . at high redshift ( @xmath0 )
we begin with a significant fraction of crsgs having disk - like components that are both thin and have younger and more metal poor ( bluer ) stellar populations than their central spheroids .
we associate the disk component with recently quenched star formation seen at @xmath124 ( e.g. * ? ? ?
* ) , so while it is bluer than the spheroid component , it is in the first stage of passive evolution and its photometric properties evolve rapidly .
the model assumes a metallicity gradient ( e.g. , as for local samples of early - type galaxies
@xcite ) between the two morphological components , consistent with the many observations of radial metallicity gradients in disk galaxies ( e.g. * ? ? ?
* and references therein ) .
in local disks , one observes metal abundance gradients of 0.2 to 0.6 dex per decade in radius ( from the direct method of line ratios in planetary nebulae and hii regions within resolved disks , as in @xcite ) , which would yield colour gradients ( from fig .
[ models ] ) of the magnitude observed in local samples ( such as the brighter members in coma from @xcite ) .
the effect of the metallicity difference on radial colour gradients is only evident at lower redshift as the difference in age of the bulge and disk components dominates at higher redshift .
given the constraint imposed on the overall colours of the galaxies by red sequence selection , the fraction of the mass provided by any disk must be severely constrained , even when there is a colour gradient .
[ models ] shows the colour difference between two equal mass ssp bursts with different formation times .
given the width of the rest - frame @xmath24 red sequence is only a few tenths of a magnitude and includes objects with a range of srsic indices at all redshifts studied here , it is likely that any _ young _ disk component carries no more than @xmath125 per cent of the total stellar mass in a typical csrg .
obviously any extended period of star formation in such a disk has less effect on the overall colour and therefore can contain more of the mass .
@xcite shows how spiral galaxies evolve almost vertically in colour space and rapidly on to the red sequence once their star formation is terminated , in agreement with the scenario presented here , where such objects are present on the red sequence but still preserve their ` spiral ' morphologies , with thin disks .
with increasing time since quenching the disk fades and reddens faster than the older spheroidal population .
after 3 - 4 gyr the difference in colours between the two components is largely dominated by the different metallicities and both fade and redden at approximately the same rate ( as in figure [ models ] ) .
this leads to the decreasing magnitude of and scatter in colour gradient with time . over time
the disk should also thicken to explain the changing axial ratio distribution for the population .
this could simply be caused by the quenched thin disk fading to reveal an underlying thick disk , or other internal and external processes could continually act to thicken the disk .
the precise mechanisms are beyond the scope of this paper and there may be several processes contributing to this , such as disk instabilities or the addition of structure in the outskirts .
however , simulations by @xcite demonstrated that the continual tidal heating suffered by disk galaxies subsequent to their infall into clusters was sufficient to thicken their disks by a factor of @xmath126 over their lifetime in the cluster , effectively transforming the shapes of late type spirals into those of s0s @xcite .
over the same time , and for the same reason , the srsic indices of those systems with the ( initially ) most prominent disks change from @xmath95 to values more consistent with s0 galaxies , @xmath127 , straightforwardly explaining the evolution seen in fig .
[ sersic ] .
a clear implication of this model is that we can link the high redshift crsgs with relatively blue and thin disks to the s0 galaxies with small colour gradients and thicker disks seen to populate cluster red sequences at intermediate and low redshifts at @xmath121 and below .
as the shapes of galaxies selected morphologically as e / s0 by @xcite do not change over time , our disk - like red sequence galaxies must naturally evolve to resemble present - day e / s0s .
a key recent work on the low redshift cluster s0 population is that of @xcite , exploring the parameters of this population in the coma cluster .
the simple model outlined here produces galaxies with photometric properties consistent with those determined by @xcite .
note that because s0 galaxies exist on the red sequences of clusters at all redshifts , the model does not imply that all low redshift cluster s0s formed in this manner between @xmath128 , merely that the s0 population was significantly augmented over this time .
and therefore little evidence for bulge growth , scaledwidth=50.0% ] there are several significant issues that are not constrained by this simple model and the existing data .
the first is the reason for the quenching of star formation in disk components .
this could simply be attributed to a loss of fuel after infall into the cluster . as ever , suspects for such cluster - centric mechanisms include ram pressure stripping ( e.g. * ? ? ?
* ; * ? ? ?
* ) , virial shock heating @xcite and strangulation ( e.g. * ? ? ?
* ; * ? ? ?
* ) amongst others .
alternatively , it could be due to a secular process , such as morphological quenching @xcite , where the spheroidal component of a galaxy grows to a size sufficient to stabilise the disk against further star formation .
so long as star formation in any disk is ended by @xmath9 , any of these mechanisms may result in the systems at @xmath10 identified in paper i having disk - like components .
a second issue is that of bulge growth . although the model above was framed in terms of disk fading , because we are measuring the evolution in relative strength of bulge and disk components with time there could equally be a rle for bulge growth in this evolution
work has recently been presented on the maturity of bulges and the quenching of star formation in @xmath129 star forming galaxies drawn from a range of environments @xcite .
this indicated that these systems quench star formation from the inside out and that their central ( bulge ) surface densities are already comparable to those of similar stellar mass @xmath86 early type galaxies with stellar masses above @xmath130 , and within a factor of two of the @xmath86 values even for lower mass systems .
the implication is that barring any subsequent mass growth through mergers , the bulges of galaxies with similar masses to those explored here are already largely in place by @xmath131 .
the crsgs explored in paper i and here either originated in the field and subsequently fell in to their clusters , or formed in an accelerated manner in existing high density regions as part of the original growth of the hosting clusters ; it might be expected that their central spheroidal components are already in place before any disk component starts to quench .
given the already - noted limit to how much mass can be in any relatively young stellar disk given the narrow colour range in cluster red sequences , coupled with the lack of growth of total stellar mass in the same galaxies since @xmath5 ( paper i ) , it would seem that the bulges would have to be in place by @xmath5 .
however , that does not account for any older disk component which could evolve morphologically , growing the bulge through inflow of stars and other material over time . in order to test this we determine the luminosity in the @xmath23 band within the central 1 kpc ( @xmath132 ) for each galaxy by integrating its fitted sersic profile out to this point .
this may be thought of as a proxy for the central density ( as used by , e.g. , @xcite in their study of low redshift transitioning ` green valley ' galaxies ) .
results are plotted as histograms in fig [ concentration ] .
the distribution of this density ; measure is essentially the same for the ` low ' and ` mid ' redshift subsamples , implying no evolution in central density at @xmath93 ( see also @xcite .
the corresponding histogram for the ` high ' subsample shows if anything a higher typical central concentration .
however , given only two clusters contribute , this could easily be attributable to shot noise from cluster to cluster .
only a small fraction of objects at this redshift have concentrations below the mean , suggesting that the inner regions of the galaxy population have been built by this epoch .
the ` very high ' sample may be different from the others : there is a relative lack of bright ` bulges ' but the small number statistics and the increased uncertainty in each measurement caused by the relatively large size of the wfc3/@xmath133 psf in comparison to the effective radii of galaxies make interpretation of this plot difficult without more data .
consequently , it appears that the high central densities of the spheroidal components of the crsgs are in place by @xmath134 .
the origin of the outer , younger , more metal poor material is more uncertain : as an example , it may itself have been accreted through minor mergers at an earlier epoch ( before @xmath117 ) , perhaps even during pre - processing of galaxies before they join the main cluster . from both the limited results on concentration obtained here and the arguments based on the work of @xcite , it is likely that the spheroidal components of most of the @xmath10 crsgs are already in place
this would imply that the decrease in the contribution from disk - like components in the population over the 3 - 4 gyr between @xmath5 and @xmath135 was mainly or solely due to the fading or rearrangement of the disks and not from the buildup of a bulge component ( or at least the central regions of any bulge ) .
this evolution in the disks may be triggered by disk instabilities , tides , harassment , fly - bys and a series of other external and internal processes we are not in a position to explore here .
a further indication that the morphological transformation may be due to changes in the disk , is the concurrent evolution observed in the axial ratios and srsic indices , which tend to reinforce our impression that the spheroidal portion of these galaxies has been established at early times and only the younger , and perhaps more fragile , disks appear to be involved in the morphological changes we observe .
one result that does not directly follow from the strawman model is the ( albeit small ) increase in the typical size ( @xmath70 ) of the crsg population between @xmath136 and @xmath137 .
if a disks fades and reddens relative to its bulge , with no large - scale change to the bulge structure it would be expected that the overall @xmath70 would probably decrease ( albeit by a modest amount ) rather than increase , in the absence of any other morphological evolution .
given the small size of any evolution , it could be explained by other evolutionary effects in only a subset of the comparatively small ` high ' and ` very high ' subsamples .
we note here that the strawman model does not account for this potential behaviour , but is certainly not currently ruled out by it .
for instance , it is possible that any size evolution in the overall crsg population could be driven by galaxies other than those which are evolving morphologically in the manner discussed here .
alternatively , one may envisage the disk being disturbed , for instance by minor mergers or fly - bys , and then added on to the outer regions of the bulge , thus driving a modest increase in size and srsic index as structure is increased in the outer regions , and reducing the oblateness of galaxies .
such fly - bys would be frequent in the dense environment of clusters and may result in significant structural changes in halos @xcite .
disk instability would be another possibility ( e.g. , @xcite ) . in the model by @xcite , for instance , galaxies are born as highly flattened objects and their disks are later on destroyed or rearranged by interactions and mergers with minor satellites ( e.g. , @xcite ) to form a more spheroid - dominated population .
several other mechanisms are possible , as discussed in the vast literature on size growth of spheroids , but we are not in a position here to speculate
. a clear improvement to the analysis presented here would be to carry out detailed two - component surface brightness fitting to crsgs over the entire redshift range and perhaps beyond @xmath5 where the analysis may need to be extended to slightly bluer galaxies due to the increased influence of star formation in red sequence progenitors .
this is not reliably achievable with the current hst data , especially at high redshift where the most changes are seen ( cf .
deeper exposures could be obtained with hst . however , an equally important issue is to be able to resolve the galaxies on scales smaller than @xmath70 in the reddest bands .
higher spatial resolution will allow the analysis to separate out any disk and bulge component and therefore to securely determine how much of the observed evolution is caused by bulge growth and how much to disk fading .
this will be straightforwardly achievable with jwst .
we have explored the morphological evolution of cluster red sequence galaxies using a large sample of such galaxies drawn from multiple clusters over the redshift range @xmath1 split into four subsamples by redshift .
our previous work showed there were significant differences between the populations at @xmath10 and @xmath138 . here
we fill in details of their evolution over the intervening 9 - 10 gyr .
our major findings are : * we confirm that the distribution in structural parameters of high redshift cluster galaxies differs from those at lower redshift .
at least a proportion of the population evolves morphologically having arrived on the red sequence subsequent to quenching of star formation . *
the measurements indicate that this evolution largely ends by @xmath122 , a lookback time of 5 - 6 gyr and a likely 3 - 4 gyr after quenching of star formation .
this evolution is seen in the distribution of srsic parameters , axial ratios , sizes and colour gradients . *
the evolution in each of these parameters indicates typical objects become less disk - dominated with time , with the strength of any colour gradient ( bluer on the outside ) decreasing similarly .
the data broadly fit a simple model where ( thin ) disk components fade and redden more quickly than their central bulges , straightforwardly explained by more recent quenching of star formation in the disk than in the bulge .
the timescale for this matches the @xmath139 gyr over which we can detect changes in the structural parameters . during this time
the disks are required to thicken in order to explain the changing axial ratio distribution , agreeing with the timescale for the expected tidal thickening of disks in clusters .
the obvious implication of the model is that these initially disk - dominated galaxies are the progenitors of the passive s0 systems that populate ( indeed dominate ) the red sequence of intermediate and low redshift clusters at luminosities around @xmath18 and below ( also see @xcite ) .
* we confirm that the size growth in the population of red sequence galaxies in clusters is significantly less over the past @xmath140 gyr than that of similar mass passively - evolving field galaxies .
the data indicate that any growth is likely to occur at @xmath109 as the size distributions of our two lower redshift subsamples appear indistinguishable . *
while the decreasing prominence of any disk component could have a contribution from bulge growth over time , consideration of the fraction of the luminosity ( mass ) within the central 1kpc of each galaxy indicates that at least to @xmath134 this is not a significant factor in the evolution of the measured structural parameters . *
although further similar hst data sets ( with both rest - frame @xmath12 and @xmath13band imaging ) for @xmath109 clusters would improve the statistical significance of some of these results , improvements in the interpretation of these results will only come from the ability to securely fit multi - component surface brightness profiles to high redshift samples .
this will only be achievable through jwst imaging which will give sufficient signal - to - noise and spatial resolution at the highest redshifts to separate the co - evolution of disk and bulge components in cluster red sequence galaxies .
this work is based on observations made with the nasa / esa hubble space telescope , and obtained from the hubble legacy archive , which is a collaboration between the space telescope science institute ( stsci / nasa ) , the space telescope european coordinating facility ( st - ecf / esa ) and the canadian astronomy data centre ( cadc / nrc / csa ) .
some of the data presented in this paper were obtained from the mikulski archive for space telescopes ( mast ) .
stsci is operated by the association of universities for research in astronomy , inc .
, under nasa contract nas5 - 26555 . balestra i. , mercurio a. , sartoris b. et al .
2015 , arxiv 1511.02522 beers t. c. , flynn k. , gebhardt k. 1990 , aj , 100 , 32 bertin e. , arnouts s. 1996 , a&as , 117 , 393 birnboim y. , dekel a. , 2003 , mnras , 345 , 349 biviano a. , rosati p. , balestra i. et al .
2013 , a&a , 558 , a1 bruce v. a. , dunlop j. s. , cirasuolo m. et al .
2012 , mnras , 427 , 1666 bruce v. a. , dunlop j. s. , mclure r. j. et al .
2014 , mnras , 444 , 1001 bruzual g. , charlot s. 2003 , mnras , 344 , 1000 buitrago f. , trujillo i. , conselice c. , hauler b. 2013 , mnras , 428 , 1460 buitrago f. , conselice c. j. , epinat b. , bedregal a. j. , grtzbauch r. , weiner b. j. 2014 , mnras , 439 , 1494 bundy k. , scarlata c. , carollo c. m. et al .
2010 , apj , 719 , 1969 cameron e. , driver s. p. 2007 ,
mnras , 377 , 523 cassata p. et al .
2013 , apj , 775 , 106 cen r. 2014 , apj , 781 , a38 cerulo p. , couch w. j. , lidman c. , delaye l. , demarco r. , huertas - company m. , mei , s. , snchez - janssen r. 2014 , mnras , 439 , 2790 chan j. c. c. , beifiori a. , mendel j. t. et al .
2016 , arxiv1603.00468 chang y .- y . , van der wel a. , rix h .- w .
2013 , apj , 773 , a149 damjanov i. , hwang h. s. , geller m. , chilingarian i. 2014 , apj , 793 , 39 damjanov i. , zahid h. j. , geller m. j. , hwang h. s. 2015 , arxiv 1508.03346 delaye l. , huertas - company m. , mei s. et al .
2014 , mnras , 441 , 203 den brok m. , peletier r. f. , valentijn e. a. et al .
2011 , mnras , 414 , 3052 de propris r. , stanford s. a. , eisenhardt p. r. , dickinson m. , elston r. 1999 , aj , 118 , 719 de propris r. , phillipps s. , bremer m. n. , 2013 , mnras , 434 , 3469 de propris r. , bremer m. n. , phillipps s. 2015 , mnras , 450 , 1268 ( paper i ) eisenhardt p. r. , brodwin m. , gonzalez a. h. et al .
2008 , apj , 684 , 905 emsellem e. , cappellari m. , krajinovic d. et al .
2007 , mnras , 379 , 401 fang j. j. , faber s. m. , koo d. c. , dekel a. 2013 , apj , 776 , 63 ferrarese l. , cote p. , jordan a. et al . 2006 ,
apjs , 164 , 334 gladders m. , yee h. k. c. 2000 , aj , 120 , 2148 gnedin o. y. , 2003a , apj , 589 , 752 gnedin o. y. , 2003b , apj , 582 , 141 guo q. , white s. , li c. , boylan - kolchin m. , 2010 , mnras , 404 , 1111 graham a. w. , worley c. 2008 , mnras , 388 , 1708 gutirrez c. m.,trujillo i. , aguerri j. a. l. , graham a. w. , caon , n. 2004 , apj , 602 , 664 haussler b. , mcintosh d.h .
, barden m. et al .
2007 , apjs , 172 , 615 head j. t. c. g. , lucey j. r. , hudson m. j. , smith r. j. , 2014 , mnras , 440 , 1690 hinshaw g. , larson d. , komatsu e. et al .
2011 , apjs , 208 , a19 holden b. p. , franx m. , illingworth g. d. et al .
2009 , apj , 693 , 617 hoyos c. , den brok m. , verdoes kleijn g. et al .
2011 , mnras , 411 , 2439 huertas - company m. , perz - gonzalz p. g. , mei s. et al .
2015 apj , 809 , a95 jedrzejewski r. i. 1987 , mnras , 226 , 747 jrgensen i. , chiboucas k. , toft s. , bergmann m. , zirm a. , schiavon r. p. , grtzbauch r. 2014 , aj , 148 , a117 kelkar k. , aragon - salamanca a. , gray m. e. et al .
2015 , mnras , 450 , 1246 kim d. , i m m. 2013 , apj , 766 , a109 kobayashi c. 2004 , mnras , 347 , 740 kodama t. , arimoto n. 1997 , a&a , 320 , 41 koopmann r. a. , kenney j. d. p. 1998 , apj , 497 , l75 krajnovic d. , bacon r. , cappellari m. et al .
2008 , mnras , 390 , 93 la barbera f. , busarello g. , merluzzi p. , massarotti m. , capaccioli m. 2002 , apj , 571 , 790 la barbera f. , busarello g. , massarotti m , merluzzi p. , capaccioli m. 2003 , a&a , 409 , 21 lidman c. , suherli j. , muzzin a. et al .
2012 , mnras , 427 , 550 lopes p. a. a. , rembold s. b. , ribeiro a. l. b. , nascimento r. s. , vajgel b. 2016 , arxiv 1606.06280 mancone c. l. , gonzalez a. h. , brodwin m. , stanford s. a. , eisenhardt p. r. m. , stern d. , jones c. 2010 , apj , 720 , 284 martig m. , bournaud f. , teyssier r. , dekel a. , 2009 , apj , 707 , 250 masters k. l. , mosleh m. , romer a. k. et al .
2010 , mnras , 405 , 783 mclure r. j. , pearce h. j. , dunlop j. s. et al .
2013 , mnras , 428 , 1088 mei s. , stanford s. a. , holden b. p. et al .
2012 , apj , 754 , a141 michard r. 1994 , a&a , 288 , 401 moustakas j. , kennicutt r. c. , jr .
, tremonti c. a. , dale d. a. , smith j .- d .
t. , calzetti d. , 2010 , apjs , 190 , 233 naab t. , johansson p. , ostriker j. p. 2009
, apj , 699 , l178 newman a. b. , ellis r. s. , bundy k. , treu t. 2012 , apj , 746 , 162 newman a. b. , ellis r. s. , andreon s. , treu t. , raichoor a. , trinchieri g. 2014 , apj , 788 , a51 oser l. , ostriker j. p. , naab t. , johansson p. h. , burkert a. 2010 , apj , 725 , 2312 papovich c. , bassett r. , lotz j. m. et al .
2012 , apj , 750 , 93 peng c. y. , ho l. c. , impey c. d. , rix h .- w .
2002 , aj , 124 , 266 peng c. y. , ho l. c. , impey c. d. , rix h .- w .
2010 , aj , 139 , 2097 peng y. , maiolino r. , cochrane r. , 2015 , natur , 521 , 192 pipino a. , matteucci f. 2004 , mnras , 347 , 968 postman m. , coe d. , benitez n. et al .
2012 , apjs , 199 , 25 quilis v. , trujillo i. 2012 , apj , 752 , l19 rawle t. d. , smith r. j. , lucey j. r. 2010 , mnras , 401 , 852 rembold s. b. , pastoriza m. g. 2012 , mnras , 422 , 719 rosati p. , tozzi p. , ettori s. et al .
2004 , aj , 127 , 230 rosati p. , tozzi p. , gobat r. et al .
2009 , a&a , 508 , 538 rozo e. , rykoff e. s. , becker m. , reddick r. m. , wechsler r. h. 2015 , mnras , 453 , 38 rudnick g. , von der linden a. , pello r. et al .
2009 , apj , 700 , 1559 ryden b. s. , forbes d. a. , terlevich a. i. 2001 , mnras , 326 , 1141 ryden b. s. 2004 , apj , 601 , 214 skibba r. , bamford s. p. , nichol r. c. et al . 2009 , mnras , 399 , 966 stanford s. a. , eisenhardt p. r. , dickinson m. 1998 , apj , 492 , 461 stott j. p. , collins c. a. , burke c. , hamilton - morris v. , smith g. p. 2011
, mnras , 414 , 445 tamura n. , ohta k. 2000 , aj , 120 , 533 tanaka m. , finoguenov a. , mirkazemi m. et al .
2013 , pasj , 65 , 17 taylor e. n. , franx m. , glazebrook k. , brinchmann j. , van der wel a. , van dokkum p. g. 2010 , apj , 720 , 723 tran k .- v .
h. , et al .
, 2010 , apj , 719 , l126 tran k .- v .
h. , nanayakkara t. , yuan t. et al .
2015 , apj , 811 , a28 vader j. p. , vigroux l. , lachieze - rey m. , souviron j. 1988 , a&a , 203 , 217 van den bergh s. 1976 , apj , 206 , 883 van der wel a. , franx m. , van dokkum p. g. et al .
2014 , apj , 788 , a28 van dokkum p. g. , bezanson r. , van der wel a. et al .
2014 , apj , 791 , a45 vika m. , bamford s. j. , hauler b. , rojas a. l. 2014 , mnras , 444 , 3603 wolf c. , aragon - salamanca a. , balogh m. et al .
2009 , mnras , 393 , 1302 | galaxies arrive on the red sequences of clusters at high redshift ( @xmath0 ) once their star formation is quenched and evolve passively thereafter .
however , we have previously found that cluster red sequence galaxies ( crsgs ) undergo significant morphological evolution subsequent to the cessation of star formation , at some point in the past 9 - 10 gyr . through a detailed study of a large sample of cluster red sequence galaxies spanning @xmath1
we elucidate the details of this evolution . below @xmath2 ( in the last 5 - 6 gyr )
there is little or no morphological evolution in the population as a whole , unlike in the previous 4 - 5 gyrs . over this earlier time
( i ) disk - like systems with srsic @xmath3 progressively disappear , as ( ii ) the range of their axial ratios similarly decreases , removing the most elongated systems ( those consistent with thin disks seen at an appreciable inclination angle ) , and ( iii ) radial colour gradients ( bluer outwards ) decrease in an absolute sense from significant age - related gradients to a residual level consistent with the metallicity - induced gradients seen in low redshift cluster members .
the distribution of their effective radii shows some evidence of evolution , consistent with growth of _ at most _
a factor @xmath4 between @xmath5 and @xmath6 , significantly less than for comparable field galaxies , while the distribution of their central ( @xmath7kpc ) bulge surface densities shows no evolution at least at @xmath8 .
a simple model involving the fading and thickening of a disk component after comparatively recent quenching ( after @xmath9 ) around an otherwise passively evolving older spheroid component is consistent with all of these findings .
[ firstpage ] galaxies : formation and evolution galaxies : interactions |
in information theoretic key agreement problem @xcite , legitimate parties need to distill a secret key from a random variable in the situation such that an eavesdropper can access to a random variable that is correlated to the legitimate parties random variable .
the privacy amplification is a technique to distill a secret key under the situation by using a ( possibly random ) function @xcite .
the security of distilled key is evaluated by various kinds of measures . in this paper , we focus on the leaked information , which is the mutual information between the distilled key and eavesdropper s random variable ( the so - called strong security @xcite ) , because it is the strongest notion among security criterion @xcite ( see also ( * ? ? ?
* appendix 3 ) ) .
the privacy amplification is usually conducted by using a family of universal 2 hash functions @xcite . in @xcite ,
bennett _ et .
al . _ evaluated ensemble averages of the leaked information for universal 2 families , and derived an upper bound on the leaked information by using the rnyi entropy of order @xmath0 . in @xcite ,
renner and wolf evaluated ensemble averages of the leaked information for universal 2 families , and derived an upper bound on the leaked information by using the smooth minimum entropy . in @xcite , hayashi evaluated ensemble averages of the leaked information for universal 2 families , and derived a parametric upper bound on the leaked information by using the rnyi entropy of order @xmath1 . concerning the exponential decreasing rate of the leaked information , the exponent derived by hayashi s bound is state - of - the - art . in noisy channel coding problem ,
the exponential decreasing rate of the error probability is also regarded as an important performance criterion of codes , and has been studied for a long time .
the best exponent at high rates is the one derived by gallager s random coding bound @xcite .
however , gallager s exponent is not tight in general , and can be improved at low rates because the random code ensemble involves some bad codes and those bad codes become dominant at low rates . the improved exponent by expurgating those bad codes is usually called the expurgation exponent @xcite .
similar improved exponents are also known in the context of the slepian - wolf coding @xcite or the quantum error correction @xcite .
the purpose of this paper is to show a security analog of above results , i.e. , to derive an improved exponent of the leaked information in the privacy amplification at low rates . for this purpose , we concentrate our attention on the case such that the random variable possessed by the legitimate parties is the binary uniform source and the function used in the privacy amplification is a linear matrix .
we first consider the case such that the eavesdropper s random variable is generated via a binary erasure channel ( bec ) .
for this case , we first relate the leaked information to the maximum likelihood ( ml ) decoding error probability of the linear code whose generator matrix is the one used in the privacy amplification .
then an improved exponent is derived by using the result of the expurgation exponent of linear codes .
it should be noted that a similar approach to relate the leaked information to the erasure error correction has been appeared in @xcite .
however in this paper , we directly relate the leaked information to the ml decoding error probability , which enables us to derive the improved exponent .
it should be also noted that the approach in this paper is completely different from the error correction approach conventionally used to prove the so - called weak security and the problem pointed out in @xcite does not apply to our approach .
next , we consider the case such that the eavesdropper s random variable is generated via a binary symmetric channel ( bsc ) . for this case , the technique used in the bec case can not be directly applied .
thus , we first reduce the bsc case to the bec case by using the partial order between bscs and becs .
the reduction turns out to be quite tight .
indeed , the exponent derived via this reduction is as good as hayashi s exponent below the critical rate , and strictly better than hayashi s exponent below the expurgation rate , which resemble the relation between the expurgation exponent and the random coding exponent of the noisy channel coding .
our results suggest that the privacy amplification with a universal 2 family is not necessarily optimal .
the rest of the paper is organized as follows .
we first explain the problem formulation of the privacy amplification in section [ section : problem ] .
then , we consider the bec case and the bsc case in sections [ section : erasure ] and [ section : bsc ] respectively .
conclusions are discussed in section [ section : conclusion ] .
let @xmath2 be a correlated i.i.d .
source with distribution @xmath3 .
the alphabet is denoted by @xmath4 . in the privacy amplification problem , we are interested in generating the uniform random number on @xmath5 by using a function @xmath6 . the joint distribution of the generated random number and the side - information is given by @xmath7 where @xmath8 .
the security is evaluated by the leaked information @xmath9 where @xmath10 is the mutual information and we take the base of the logarithm to be @xmath11 . for given rate @xmath12
, we are interested in the exponential decreasing rate of @xmath13 , i.e. , @xmath14 in the privacy amplification problem , we typically use the universal @xmath0 hash family .
a family @xmath15 of functions @xmath16 is called universal @xmath0 if @xmath17 for every @xmath18 , where @xmath19 is the uniform random variable on @xmath15 . for parameter @xmath20 ,
let @xmath21 . \end{aligned}\ ] ] hayashi derived the following lower bound on @xmath22 . [
proposition : hayashi - exponent ] for any universal @xmath0 hash family @xmath15 , we have @xmath23 \\ & \ge & e_r(r;x|z ) \\ & : = & \max_{0 \le \theta \le 1 } \left [ \psi(\theta ; x|z ) - \theta r \right],\end{aligned}\ ] ] where @xmath24 means the average over randomly chosen function from @xmath15 .
in this section , we assume that @xmath25 is the uniform binary source and @xmath26 is the output of the binary erasure channel ( bec ) with erasure probability @xmath27 , i.e. , @xmath28 and @xmath29 , where @xmath30 represent the erasure symbol ( see fig .
[ fig : bec - e ] ) . for
given sequence @xmath31 , let @xmath32 be the set of those indices such that @xmath33 .
when the sequence @xmath31 is obvious from the context , we abbreviate @xmath34 as @xmath35 .
[ cols= " < , > " , ] in this section , we assume that @xmath25 is the uniform binary source and @xmath26 is the output of the binary symmetric channel ( bsc ) with crossover probability @xmath27 , i.e. , @xmath28 and @xmath36 ( see fig . [
fig : bsc - e ] ) .
let @xmath37 be the output of bec(@xmath38 ) with input @xmath25 .
since bec(@xmath38 ) ( see fig . [
fig : bec - convolution ] ) is less noisy than bsc(@xmath27 ) @xcite , we have @xmath39 thus , corollary [ corollary : expurgation - exponent ] can be applied to the case considered in this section .
[ theorem : expurgation - bsc - less - noisy - reduction ] let @xmath37 be the output of bec(@xmath40 ) with input @xmath25 .
then , we have @xmath41 hayashi s exponent for bsc(@xmath27 ) is @xmath42.\end{aligned}\ ] ] the exponents are compared in fig .
[ fig : ex - ed - less - noisy - reduction-011 ] and fig .
[ fig : ex - ed - less - noisy - reduction-025 ] for @xmath43 and @xmath44 respectively .
( dashed line ) and @xmath45 ( solid line ) for bsc(@xmath46 ) . ]
( dashed line ) and @xmath45 ( solid line ) for bsc(@xmath44 ) . ]
let @xmath47 be the critical rate , i.e. , the largest rate such that the optimization in @xmath48 is achieved by @xmath49 .
then , for @xmath50 , we have @xmath51 on the other hand , let @xmath52 be the expurgation rate , i.e. , the smallest rate such that the optimization in @xmath53 is achieved by @xmath49 .
then , for @xmath54 , we have @xmath55 thus , for @xmath56 , @xmath57 , which can be also observed in fig . [
fig : ex - ed - less - noisy - reduction-011 ] and fig .
[ fig : ex - ed - less - noisy - reduction-025 ] .
we also find that @xmath58 is strictly larger than @xmath59 at low rates .
for the bec case and the bsc case , we derived the expurgation exponent of the leaked information in the privacy amplification . the technique to relate the leaked information to the ml decoding error probability heavily relies on the specific structure of the bec .
thus , to derive the expurgation exponent for general cases , a method to expurgate bad functions directly might be needed .
hayashi derived a quantum counter part of proposition [ proposition : hayashi - exponent ] in @xcite .
it is also interesting to derive the expurgation exponent in the privacy amplification for quantum adversary . for the case
such that the eavesdropper s information is generated via the complementary channel of a pauli channel , the technique to relate the leaked information to the ml decoding error probability is already known @xcite , and it is not difficult to derive the expurgation exponent . in general , more refined technique is needed .
these topics will be investigated in elsewhere .
the author would like to thank prof .
yasutada oohama for valuable discussion .
the author also would like to thank prof .
prakash narayan for inviting the author to the workshop .
this research is partly supported by grand - in - aid for young scientist(b):2376033700 and grand - in - aid for scientific research(a):2324607101 .
u. maurer , `` the strong secret key rate of discrete random triples , '' in _ communication and cryptography two sides of one tapestry_.1em plus 0.5em minus 0.4emkluwer academic publishers , 1994 , pp .
271285 . r. renner and s. wolf , `` simple and tight bound for information reconciliation and privacy amplification , '' in _ advances in cryptology
asiacrypt 2005 _ , ser .
lecture notes in computer science , vol .
3788.1em plus 0.5em minus 0.4emspringer - verlag , 2005 , pp . 199216 .
a. thangaraj , s. dihidar , a. r. calderbank , s. w. mclaughlin , and j .-
merolla , `` application of ldpc codes to the wiretap channel , '' _ ieee trans .
inform . theory _
53 , no . 8 , pp . 29332945 , august 2007 .
s. watanabe , r. matsumoto , and t. uyematsu , `` strongly secure privacy amplification can not be obtained by encoder of slepian - wolf code , '' _ ieice trans .
fundamentals _ , vol .
e93a , no . 9 , pp .
16501659 , september 2010 , arxiv:0906.2582 . | we investigate the privacy amplification problem in which eve can observe the uniform binary source through a binary erasure channel ( bec ) or a binary symmetric channel ( bsc ) . for this problem
, we derive the so - called expurgation exponent of the information leaked to eve .
the exponent is derived by relating the leaked information to the error probability of the linear code that is generated by the linear hash function used in the privacy amplification , which is also interesting in its own right .
the derived exponent is larger than state - of - the - art exponent recently derived by hayashi at low rate . |
lovsz and schrijver ( 1991 ) have proposed a very intriguing successive convex relaxation procedure for 0 - 1 integer programming problems .
the procedure called @xmath0 , to be defined shortly , when applied to a classical linear programming ( lp ) relaxation of the stable set problem ( with only the edge and nonnegativity constraints ) produces a relaxation for which many well - known inequalities are valid , including the odd hole , odd antihole , odd wheel , clique , and even the orthonormal representation inequalities of grtschel , lovsz and schrijver ( 1981 ) .
this implies that for many classes of graphs , including perfect ( for which clique inequalities are sufficient ) or t - perfect graphs ( for which odd hole inequalities are sufficient ) , one can find the maximum stable set by using the @xmath0 procedure .
the @xmath0 procedure is a strengthening of another procedure , called @xmath1 , also introduced by lovsz and schrijver .
the main difference between the two procedures is that @xmath0 involves a positive semidefinite constraint .
when applied to a linear programming relaxation , @xmath1 will produce another ( stronger ) lp relaxation while @xmath0 will produce a semidefinite relaxation .
for the stable set problem , lovsz and schrijver have shown that the relaxation produced by @xmath1 is much weaker than the one derived from @xmath0 . in general
, it is however not clear in which situations the procedure @xmath0 is better or significantly better than @xmath1 ; especially , when @xmath1 and @xmath0 are applied iteratively . in this paper
, we try to shed some light on this question .
we generalize certain properties derived by lovsz and schrijver .
we also identify certain situations in which @xmath1 produces the same relaxation as @xmath0 .
several examples are discussed throughout the paper , including one in which the number of iterations of the @xmath0 procedure needed to derive the convex hull of 0 - 1 points is equal to the dimension of the space , hence resolving a question left open by lovsz and schrijver . in the next section ,
we review the lift - and - project procedures and their basic properties .
section [ sec : up ] includes upper bounds on the number of major iterations required by such procedures .
section [ sec : low ] discusses techniques to prove lower bounds on the number of major iterations required .
sections [ sec : lowercomp ] and [ sec : gen ] include geometric properties and characterizations of the convex relaxations produced by the procedures .
first , we describe two lift - and - project procedures proposed by lov ' asz and schrijver ( 1991 ) which produce tighter and tighter relaxations of the convex hull of @xmath2-@xmath3 points in a convex set . in what follows , @xmath4 is the @xmath5th unit vector and @xmath6 is the vector of all ones .
the sizes of @xmath6 and @xmath4 will be clear from the context .
the cone generated by all 0 - 1 vectors @xmath7 with @xmath8 is called @xmath9 .
let @xmath10 denote a convex cone ; for example , @xmath11 could be a polyhedral cone obtained from a polytope @xmath12 in @xmath13^d$ ] via homogenization using a new variable @xmath14 . that is , if @xmath15 then @xmath16 we are interested in determining ( or approximating ) @xmath17 , the cone generated by all 0 - 1 vectors of @xmath11 .
let @xmath18 , @xmath19 denote the dual cones of @xmath11 and @xmath9 under the standard euclidean inner - product , e.g. , @xmath20 @xmath21 denotes the space of @xmath22 symmetric matrices and @xmath23 denotes the cone of @xmath24 symmetric , positive semidefinite matrices .
for a matrix @xmath25 , we denote its positive semidefiniteness by @xmath26 .
when we deal with the duals of convex cones in the space of @xmath27 matrices ( or in the subspace of the symmetric matrices ) , we always take the underlying inner - product to be the trace inner - product ( or frobenius inner - product ) : @xmath28 .
let @xmath29 denote the linear operator which maps a symmetric matrix to its diagonal .
then its adjoint @xmath30 is the linear operator @xmath31 which maps a vector from @xmath32 to the diagonal atrix in @xmath21 whose @xmath33th component is the @xmath34th component of the original vector .
[ defi : m ] a @xmath27 symmetric matrix , @xmath35 , with real entries is in @xmath36 if * @xmath37 , and * @xmath38 lov ' asz and schrijver note that condition ( ii ) of the above definition is equivalent to @xmath39 ( where @xmath40 ) , or : @xmath41 @xmath42 for all @xmath43 and @xmath44 for all @xmath45 since the extreme rays ( after normalization ) of the cone @xmath19 are given by @xmath46 [ defi : mplus ] @xmath47 if @xmath48 and @xmath35 is positive semidefinite . observe that if we take any @xmath49 ( not necessarily integral ) and consider @xmath50 , @xmath35 satisfies @xmath51 and also ( ii) , but this specific @xmath35 satisfies ( i ) if and only if @xmath52 is such that @xmath53 for all @xmath34 , i.e. @xmath52 corresponds to a @xmath2-@xmath3 vector .
now , we define the projections of these liftings @xmath54 and @xmath55 : @xmath56 @xmath57 the above argument regarding @xmath58 shows that @xmath59 , the last inclusion following from the fact that @xmath60 and @xmath42 imply that @xmath61 . if @xmath12 is a polytope ( or any convex set ) in @xmath13^d$ ] then we simply write @xmath62 to represent @xmath63 where @xmath11 is the cone obtained via homogenization using the variable @xmath14 , and similarly for @xmath64 .
we also let @xmath65 and @xmath66 . we should point out that the definition of @xmath54 ( or @xmath55 ) is such that @xmath36 depends only on the sets @xmath67 and @xmath68 for all @xmath34 .
in particular , we have : let @xmath11 and @xmath69 be such that @xmath70 and @xmath71 for all @xmath72
. then @xmath73 ( and @xmath74 ) and @xmath75 ( and @xmath76 ) .
for example , @xmath77 and @xmath78 ( see figure [ fig1 ] ) have the same @xmath79 .
the definitions of @xmath54 , @xmath1 , @xmath55 and @xmath0 are invariant under various operations including flipping coordinates @xmath80 for any subset of the indices @xmath81 .
more formally , [ prop:2.1 ] let @xmath82 be a linear transformation mapping @xmath9 onto itself .
then @xmath83 one crucial feature of the operators @xmath1 and @xmath0 is that they can be iterated .
the iterated operators @xmath84 and @xmath85 are defined as follows .
@xmath86 , @xmath87 , @xmath88 and @xmath89 for all integers @xmath90 .
lovsz and schrijver ( 1991 ) show that , even without the positive semidefiniteness constraints , @xmath91 iterations are sufficient to get @xmath17 : [ thm : ls ] @xmath92 and @xmath93 let @xmath94 be a valid inequality for @xmath17 .
then the smallest nonnegative integer @xmath95 such that @xmath94 is valid for @xmath84 is called _ the @xmath1-rank of @xmath96
relative to @xmath11 . _
the @xmath0-rank of @xmath94 relative to @xmath11 is defined similarly . the above theorem states that these ranks are at most @xmath91 for any valid inequality .
the @xmath1-rank ( resp.@xmath0-rank ) of a cone @xmath11 is the smallest nonnegative integer @xmath95 such that @xmath97 ( resp.@xmath98 ) .
theorem [ thm : ls ] can also be proved using the results of balas ( 1974 ) , see balas , ceria and cornujols ( 1993 ) .
our interest , in this paper , mostly lies in understanding the strength of @xmath0 in comparison to @xmath1 .
consider the stable set polytope on a graph @xmath99 defined as the convex hull of incidence vectors of sets of non - adjacent vertices ( known as stable sets ) .
let @xmath100 be the relaxation defined by the edge constraints ( @xmath101 for all edges @xmath102 ) and the nonnegativity constraints ( @xmath103 for all @xmath104 ) .
then @xmath105 is exactly equal to the relaxation obtained by adding all odd hole inequalities , saying that @xmath106 for any odd cycle @xmath107 with no chords .
however , many more complicated inequalities have small @xmath0-rank .
lovsz and schrijver ( 1991 ) prove that odd hole , odd antihole , odd wheel , clique and orthogonal inequalities all have @xmath0-rank at most 1 , relative to @xmath100 .
these results are proved using lemma [ lem:2.2 ] of next section , except for the orthogonality constraints .
in contrast , the @xmath1-rank of a clique inequality for example is equal to @xmath108 where @xmath109 is the size of the clique .
note that the separation problem for the class of clique inequalities is np - hard ( and so is the problem of optimizing over the clique inequalities , see grtschel , lovsz and schrijver ( 1981 ) ) .
@xmath0 , however , leads to a polynomial - time separation algorithm for a broader class of inequalities .
this , and more generally the importance of @xmath1 and @xmath0 , stems from the following result .
[ thm : complexity ] if we have a weak separation oracle for @xmath11 then we have a weak separation oracle for @xmath84 and @xmath85 for any fixed constant @xmath95 . together with the equivalence between ( weak ) optimization and ( weak ) separation ( grtschel et al .
( 1981 ) ) , this implies for example that the stable set problem can be solved in polynomial time for any graph with bounded @xmath0-rank ( lovsz and schrijver ( 1991 ) ) .
next we study the upper bounds on @xmath1- and @xmath0-ranks of inequalities and convex sets .
lovsz and schrijver give some ways to upper bound the @xmath1-rank of an inequality .
they show the following .
[ lem:2.1 ] @xmath110 lovsz and schrijver ( 1991 ) define an operator @xmath111 by : @xmath112 thus , @xmath113 . the iterated operator @xmath114 , @xmath111-rank of inequalities , polytopes and convex cones are defined analogously to the corresponding definitions of @xmath1-and @xmath0-ranks .
lemma [ lem:2.1 ] shows that an inequality will be valid for @xmath115 if it is valid for @xmath116 and @xmath117 for some @xmath34 . in order to iterate lemma [ lem:2.1 ]
, we first need the following lemma .
it is stated in terms of the faces of @xmath9 , which can be obtained by intersecting @xmath9 with hyperplanes of the form @xmath118 or @xmath119 .
similar insights for a procedure related to the @xmath1- procedure were discussed by balas ( 1974 ) .
[ lem : face ] let @xmath120 be any face of @xmath9
. then @xmath121 similarly for @xmath0 and @xmath111 . *
* `` @xmath122 '' is clear from the definitions . for the converse , let @xmath123
this means that there exists a matrix @xmath124 with @xmath125 .
since @xmath126 and @xmath127 and their sum @xmath128 belongs to the face @xmath120 of @xmath9 , we have that @xmath129 and @xmath130 must belong to @xmath120 , by definition of a face .
thus , @xmath131 and @xmath132 for all @xmath34 implying that @xmath133 and @xmath134 .
the proof for @xmath0 is identical . ' '' '' + iterating lemma [ lem : face ] , we get : [ cor : face ] let @xmath120 be any face of @xmath9 .
then , for any @xmath95 , @xmath135 similarly for @xmath0 and @xmath111 .
repeatedly using lemma [ lem:2.1 ] and lemma [ lem : face ] ( or corollary [ cor : face ] ) , we can derive a condition that an inequality be valid for @xmath84 .
this , in particular , proves theorem [ thm : ls ] .
[ rankn ] @xmath136 where @xmath137 we should point out that even though @xmath138 and @xmath139 , @xmath140 is not necessarily equal to @xmath141 , if @xmath142 .
for example , for @xmath143 , one can show that @xmath144 . for @xmath145 , lovsz and schrijver ( 1991 ) give a different condition for the validity of an inequality . in the statement of the next lemma , the assumption that @xmath146 is without loss of generality ( by flipping coordinates if necessary , as shown in proposition [ prop:2.1 ] ) .
[ lem:2.2 ] let @xmath147 .
then @xmath94 is valid for + @xmath148 for all @xmath34 such that @xmath149 , implies @xmath94 is valid for @xmath145 . as mentioned previously
, the result that clique , odd hole , odd antihole , odd wheel inequalities for the stable set problem have @xmath0-rank 1 follows from the above lemma . for the stable set problem ( as for many combinatorial optimization problems )
, there exists several important constructions to derive facet - defining valid inequalities from other facet - defining inequalities .
the simplest is _ cloning _ a clique at a vertex @xmath150 , which consists of replacing the vertex by a clique , replacing all the edges incident to @xmath150 by corresponding edges incident to all clique vertices and substituting in the inequality the variable for @xmath150 by the sum of the variables of the clique vertices .
it can easily be shown that the resulting inequality is valid and facet - defining if the original inequality was a non - trivial ( i.e. different from the nonnegativity constraints ) facet - defining inequality .
in general , it is not clear how cloning influences the @xmath0-rank of an inequality .
however , if we perform cloning at the center vertex of an odd wheel inequality , lemma [ lem:2.2 ] implies that the @xmath0-rank still remains equal to 1 .
if we perform cloning at one or several vertices of an odd wheel , odd hole or odd antihole inequality , lemma [ lem:2.2 ] implies that the @xmath0-rank is at most 2 .
indeed , if we fix any variable ( of the corresponding subgraph ) to 1 , the resulting inequality can be seen to be a linear combination of clique inequalities and hence valid for @xmath151 .
lemma [ lem:2.2 ] can be extended to derive conditions under which the @xmath0-rank of an inequality is at most @xmath95 .
[ thm:3.5 ] let @xmath147 and let @xmath152 . if @xmath94 is valid for + @xmath153 for all sets @xmath154 satisfying either of the following two conditions 1 .
[ it:1 ] @xmath155 , 2 .
[ it:2 ] @xmath156 and @xmath157 , then @xmath94 is valid for @xmath158 .
observe , however , that the result mentioned previously regarding cloning does not follow from theorem [ thm:3.5 ] .
* we proceed by induction on @xmath95 . for @xmath159 ,
the result is lemma [ lem:2.2 ] .
assume now that @xmath160 , that the theorem was proved for @xmath161 ( and for any inequality and for any convex set @xmath11 ) , and that the hypothesis is satisfied for the inequality @xmath162 and @xmath95 . from corollary [ cor : face ] and lemma [ lem:2.2 ] , we know that @xmath162 is valid for @xmath163 if it is valid for @xmath164 for all @xmath165 .
this is equivalent to showing that @xmath166 is valid for @xmath167 .
now there are two cases .
if @xmath168 then condition [ it:2 ] implies that @xmath169 and thus any inequality is valid for @xmath170 . on the other hand , if @xmath171 , we can use induction to prove the result .
indeed , conditions [ it:1 ] and [ it:2 ] for inequality @xmath172 and @xmath95 imply that conditions [ it:1 ] and [ it:2 ] are satisfied for the inequality @xmath173 for @xmath174 .
thus , by the inductive hypothesis , @xmath166 is valid for @xmath167 , proving the inductive statement . ' '' '' + for the stable set problem , the above theorem implies that the @xmath0-rank of a graph is at most its stability number @xmath175 , the cardinality of the largest stable set in @xmath176 ; this was proved in corollary 2.19 of lovsz and schrijver ( 1991 ) . more generally , if we consider a polytope @xmath12 for which @xmath177 is only described by inequalities of the form @xmath162 with @xmath146 ( i.e. it is lower comprehensive , see section [ sec : lowercomp ] ) then its @xmath0-rank is upper bounded by the maximum number of variables that can be set to 1 in @xmath12 to obtain a unique integral point of @xmath177 ( in which the other variables are thus set to 0 ) .
similar , more complex , statements can be made if the polytope is not lower comprehensive .
consider the complete undirected graph on the vertex set @xmath178 ; let @xmath179 denote its edge set .
let @xmath180 in the above , @xmath181 is the set of edges in @xmath179 that are incident on @xmath150 ; for @xmath182 , @xmath183 represents @xmath184 . for @xmath185 ,
let @xmath186 refer to the set of edges with both endpoints in @xmath187 .
then the _ matching polytope _ for the complete graph is @xmath188 edmonds ( 1965 ) proved that @xmath189 the above inequalities are known as the blossom inequalities .
[ thm:3.1 ] the @xmath0-rank of the inequality @xmath190 with respect to @xmath12 is @xmath191 .
the fact that the @xmath0-rank is at most @xmath191 also follows directly from theorem [ thm:3.5 ] .
observe that since @xmath91 is @xmath192 , we derive that the @xmath0-rank of @xmath12 is equal to @xmath193 if @xmath194 odd and , @xmath195 if @xmath194 even . from theorem [ thm:3.1 ]
, the @xmath1-rank of the blossom inequality on @xmath187 is at least @xmath191 .
furthermore , using theorem [ rankn ] with @xmath196 being the complement of a complete bipartite graph on @xmath191 and @xmath197 vertices on each side , we derive that the @xmath111-rank of a blossom inequality is equal to @xmath198 .
this uses the fact that @xmath12 is an integral polytope if and only if the underlying graph is bipartite .
thus , the @xmath1-rank is at most @xmath198 .
these bounds are to be compared with those derived from corollary 2.8 of lovsz and schrijver ( 1991 ) ( since a matching in a graph can be viewed as a stable set in its line graph ) .
their results imply a lower bound of @xmath199 and an upper bound of @xmath200 .
consider @xmath201 then @xmath202 theorem [ rankn ] implies that the @xmath1-rank of @xmath203 is at most @xmath204 , while theorem [ thm:3.5 ] implies that the @xmath0-rank is at most @xmath205 .
these bounds are actually attained and this is discussed in section [ details ] .
we also show in that section that the positive semidefiniteness constraint does not help for many iterations .
in this section , we provide lower bounds on the @xmath1- and @xmath0-rank .
we also show a situation in which the positive semidefiniteness constraints do not help at all and both the @xmath1-rank and the @xmath0-rank of a polytope is @xmath91 .
we first provide a way to derive points in @xmath62 in certain cases . for @xmath206
define @xmath207 so , @xmath208 .
throughout this section , let @xmath209 .
[ thm:3.2 ] let @xmath210 such that @xmath211 then @xmath212
. simply stated , this result says that if we can replace any coordinate of @xmath52 ( strictly between 0 and 1 ) by 0 and 1 and remain in @xmath12 then @xmath212
. * proof .
* we define @xmath213 by definition , @xmath214 , @xmath215 .
moreover , @xmath216 therefore , @xmath217 for all @xmath218 .
similarly , @xmath219 therefore , @xmath220 for all @xmath218 .
finally , since @xmath221 for all @xmath222 , we have @xmath223 therefore , @xmath224 and @xmath225 as desired . ' '' '' + as a corollary , we derive the following ( see figure [ fig2 ] for an illustration ) . [ corequal ] let @xmath12 be such that @xmath226 for all @xmath227 . then @xmath228 * proof .
* let @xmath229 . by lemma [ lem:2.1 ]
, we know that @xmath230 . on the other hand , theorem [ thm:3.2 ] shows that @xmath231 . ' '' ''
+ in the proof of theorem [ thm:3.2 ] , we constructed a @xmath232 such that a certain @xmath233 would also be in @xmath62 .
the idea of the proof suggests a stronger technique to achieve such a goal .
we define @xmath234 where @xmath235 , @xmath236 then clearly we have @xmath237 @xmath238 moreover , using the schur complement of @xmath239 in @xmath240 , we have @xmath241 the latter can be assured in many simple ways , for example by diagonal dominance : it suffices to choose @xmath242 such that @xmath243 the entries of such a @xmath244 will be further restricted by the condition @xmath245 for every @xmath246 and @xmath247 for every @xmath248 .
if this condition is verified for some @xmath244 then the above argument would imply @xmath225 . in the case of theorem [ thm:3.2 ]
, we utilized diagonal dominance ; because of the special structure of @xmath12 , we could choose @xmath249 and satisfy all the conditions for @xmath225 .
we now give an example where both @xmath1 and @xmath0 require @xmath91 iterations , showing that theorem [ thm : ls ] can not be improved .
this result was independently obtained by cook and dash ( 1999 ) who also show additional results regarding the rank of inequalities .
previously , the worse example known in terms the number of repeated @xmath0 iterations needed to obtain @xmath17 was the matching polytope results of stephen and tunel ( 1999 ) where the @xmath0-rank was of the order of @xmath250 .
let @xmath251 [ thm:3.6 ] for @xmath252 , @xmath253 .
furthermore , @xmath254 while @xmath255 .
thus , the @xmath0 procedure requires @xmath91 iterations to prove @xmath256 * proof .
* follows from corollary [ corequal ] .
( in fact this corollary characterizes precisely @xmath257 . ) ' '' '' + one interesting feature of the example above is that @xmath258 can be described by @xmath259 inequalities , contains no integral point , but no inequality can be removed without creating an integral point .
this is actually an extreme situation in this regard as shown by the following result of doignon ( 1973 ) .
suppose we are given a set of @xmath260 linear inequalities @xmath261 where @xmath206 and @xmath262 .
a theorem of doignon ( 1973 ) implies that if this system does not contain any integer points then there is a subsystem ( of this system ) with at most @xmath259 inequalities which does not have an integer solution .
doignon s theorem is an integer analog of helly s theorem . in section [ ex2 ]
, we have shown that the @xmath1-rank and the @xmath0-rank of @xmath263 are at most @xmath204 and @xmath264 , respectively . here
we claim that these bounds are attained .
[ thm:3.8 ] the @xmath1-rank of @xmath265 relative to @xmath11 is @xmath204 .
the @xmath0-rank of the same inequality relative to @xmath11 is @xmath266 oreover , for @xmath267 , the optimum values of @xmath268 are the same . our proof of the first statement of the theorem , saying that the @xmath1-rank is @xmath204 is lengthy and is not included here .
the proof of the remainder of the theorem appears partly in this section and partly in the appendix .
the theorem indicates that the positive semidefiniteness constraint does not help for @xmath269 iterations .
unfortunately , neither theorem [ thm:3.2 ] nor corollary [ corequal ] is useful here . instead , exploiting the symmetry ( and convexity of @xmath115 and @xmath145 )
, we will only consider points in @xmath84 or @xmath85 such that @xmath270 takes only three possible values , @xmath2 , @xmath3 and a constant @xmath271 .
letting @xmath272 denote the number of @xmath270 set to 0 and letting @xmath273 denote the number of @xmath270 set to 1 , we define @xmath274 to be the largest common value @xmath271 of the remaining @xmath275 coordinates of @xmath52 such that @xmath276 .
we define @xmath277 similarly with respect to @xmath85 . by symmetry
, such a point @xmath52 belongs to @xmath84 ( resp . to @xmath158 ) if there exists a symmetric matrix @xmath124 ( resp .
@xmath278 ) of the form @xmath279 for some value @xmath280 ; here the columns of @xmath35 are partitioned in the way that the first column corresponds to the homogenizing variable @xmath14 , the next @xmath273 columns correspond to those @xmath281 that are set to one , the next @xmath272 columns correspond to those @xmath281 set to zero and the remaining @xmath282 columns correspond to the remaining @xmath281 s ( which are set to @xmath271 ) .
for @xmath283 and @xmath284 , we see by plugging @xmath52 into the description of @xmath11 that @xmath285 for @xmath286 , the condition that @xmath287 is equivalent to @xmath288 ( corresponding to @xmath289 ) and @xmath290 ( corresponding to @xmath291 ) . eliminating @xmath280 , we derive : @xmath292 the condition that @xmath51 reduces to ( by taking a schur complement ) @xmath293 ( where the matrices have size @xmath294 , or @xmath295 and @xmath296 .
this can be seen to imply that @xmath297 observe that the @xmath1-rank ( resp .
the @xmath0-rank ) of @xmath11 is the smallest integer @xmath95 such that @xmath298 ( resp.@xmath299 ) .
theorem [ thm:3.8 ] hence follows from the following proposition .
[ propapp ] 1 .
[ firstitem ] @xmath300 , 2 .
@xmath301 , 3 . for any @xmath302 such that @xmath303
, we have @xmath304 . the proof of [ firstitem ] is obtained by solving explicitly the recurrence for @xmath305 ; the details however , are omitted .
the proof of the rest of the proposition is given in the appendix .
theorem [ lower ] in the appendix actually illustrates a peculiar behavior of the @xmath0 operator ( as well as the @xmath1 operator ) on this example . in cutting plane procedures , it is usual that the improvement due to the addition of a cutting plane ( or a batch of them ) decreases as the algorithm progresses .
however , theorem [ lower ] shows that @xmath306 hence , as illustrated on figure [ fig : nplus ] for @xmath307 , the improvement in objective function value is negligible for many iterations and only towards the end increases considerably .
we should point out , however , that the procedures @xmath1 and @xmath0 are such that the number of `` important '' inequalities generated in each iteration could potentially increase tremendously in later iterations .
a nonempty convex set @xmath308 is called _ lower comprehensive _ if for every @xmath233 , every @xmath309 such that @xmath310 is also in @xmath12 .
[ defi:3.1 ] let @xmath311 .
a convex set @xmath312^d$ ] is said to be _ a convex corner with respect to @xmath150 _ if there exists a linear transformation @xmath82 of @xmath313 onto itself such that @xmath314 and @xmath315 is lower comprehensive .
[ thm:3.3 ] if @xmath12 is a convex corner with respect to @xmath311 then so are @xmath64 and @xmath62 .
* proof . * by proposition [ prop:2.1 ] and the definitions , it suffices to prove that if @xmath12 is lower comprehensive then so are @xmath64 and @xmath62 .
let @xmath12 be lower comprehensive and @xmath316 .
it suffices to show that @xmath317 for every @xmath5 such that @xmath318 . without loss of generality
suppose @xmath319 and @xmath318 .
then there exists @xmath320 such that @xmath321 .
let @xmath322 then using the fact that @xmath12 is lower comprehensive , it is easy to see that @xmath323 . since the above argument applies to every @xmath5 such that @xmath318 , we proved that @xmath64 is lower comprehensive .
we can prove that @xmath62 is lower comprehensive by a very similar argument .
we only have to note that if @xmath232 then the corresponding @xmath324 constructed as above will be positive semidefinite ( in addition to satisfying @xmath325 for every @xmath326 and @xmath327 for every @xmath328 ) since every principal minor of @xmath324 is a principal minor of @xmath35 and @xmath35 is positive semidefinite . ' '' ''
+ a similar fact , in a less general form , was observed independently by cook and dash ( 1999 ) .
in this section , we derive general conditions under which the positive semidefiniteness constraint is not useful .
this can be expressed in several ways as * @xmath329 , or as * @xmath330 or even as * @xmath331 for some given @xmath305 .
first , we rewrite condition ( ii ) of definition [ defi : m ] . since @xmath35 is symmetric , @xmath332 using the fact that @xmath333
, we see that condition @xmath334 is also equivalent to ( ii)^ & & y ^ * , where @xmath335 let s define @xmath336 note that the cone ( more specifically , the subspace in this case ) dual to @xmath337 in the space @xmath21 is the orthogonal complement of @xmath337 .
@xmath338 where @xmath339 .
we have [ thm : m = m+ ] @xmath340 * proof .
* by definition of the sets @xmath36 , @xmath341 , we have @xmath342^ * \cap d = \left[t(k)\right]^ * \cap d \cap { \cal s}_+^{d+1}.\ ] ] since the inclusion @xmath343^ * \cap d \supseteq \left[t(k)\right]^
* \cap d \cap { \cal s}_+^{d+1}$ ] is clear , we have @xmath342^ * \cap d \subseteq { \cal s}_+^{d+1}.\ ] ] noting that @xmath344^ * \cap d \subseteq { \cal s}_+^{d+1 } \,\ , \iff \,\ , ( \left[t(k)\right]^ * \cap d)^ * \supseteq { \cal s}_+^{d+1},\ ] ] ( we used the fact that @xmath345 is self dual under the trace inner - product , in the space @xmath21 ) and that @xmath346^ * \cap d)^ * = t(k ) + d^*,\ ] ] we conclude @xmath347 ' '' '' + this theorem completely characterizes when @xmath54 and @xmath55 differ or are equal . to make the condition more easily tractable , we can give a more explicit description of @xmath348 .
define @xmath349 to be set of all @xmath350 such that @xmath351 is a facet of @xmath12 ( or , more generally , for non - polyhedral convex sets , @xmath349 describes a set of valid inequalities exactly characterizing @xmath12 ) .
note that @xmath349 can be taken as the set of extreme rays of @xmath18 .
we arrive at the identity t(k)+d^ & = & \{(e_i v^t + v e_i^t ) , i \{1,2 ,
, d } , v f(k);. + & & , i \{1 , 2 , , d } , v f(k ) ; + & & .
( e_ii - e_0i ) , i \{1,2 , , d } , } where we have used the fact that @xmath352 since @xmath353 .
so , @xmath354 iff for every @xmath7 , we can express @xmath58 as an element of the above cone @xmath355 .
consider the clique on four vertices and the corresponding lp relaxation @xmath100 of the stable set problem ( with the edge and nonnegativity constraints only ) .
for this example , @xmath356^ * \cap d;\ ] ] but clearly @xmath357 . a proof of this is provided by the incidence vector of the clique inequality on the four vertices : @xmath358 this means , for @xmath359 , @xmath58 is not in the convex cone @xmath360 .
now , we relate these findings to @xmath115 and @xmath145 .
[ cor:6.1a ] if @xmath361 , then @xmath362 for every @xmath363
. * proof . * trivial for @xmath283 . by theorem [ thm : m = m+ ]
, the assumption of the corollary implies @xmath364 . by theorem [ thm : ls ] ,
thus , @xmath366 now , applying theorem [ thm : m = m+ ] recursively , we obtain the desired result . ' '' ''
+ now we look at the weaker condition that @xmath330 .
[ thm : n = n+ ] @xmath364 if and only if for every @xmath367 , @xmath368 before proceeding with the proof , observe that , for any convex cone @xmath369 , we have [ eqn : adjoint - dual ] ^ * = \{s ^d+1 : diag(s ) ^ * } . * proof . * as in the proof of theorem [ thm : m = m+ ] , we obtain @xmath370^ * \cap d \right ) \subseteq \mbox{diag}\left(\left[t(k)\right]^ * \cap d \cap { \cal s}_+^{d+1}\right).\ ] ] using equation ( [ eqn : adjoint - dual ] ) and the proof technique of theorem [ thm : m = m+ ] , we find @xmath371 @xmath372 ' '' '' + we should compare this result to lemma 1.2 of lovsz and schrijver ( 1991 ) .
note that our result is also based on cone duality , we also characterize the dual cones of @xmath115 and @xmath145 ; but , we only work in the space of symmetric matrices instead of the larger space of all matrices . as a result , the dependence of the characterization on the skew symmetric matrices is eliminated and our description is more explicit .
our ideas in the geometric characterizations above are also applicable in comparing the weaker procedure @xmath111 to @xmath1 .
recall @xmath373 we define @xmath374 the main difference with @xmath54 is that @xmath35 is not necessarily symmetric .
as is mentioned by lovsz and schrijver ( 1991 ) , we have @xmath375 we further define @xmath376 then @xmath377^ * \cap d_0 \right),\ ] ] where @xmath378^*$ ] is the dual of @xmath379 in @xmath380 under the trace inner - product .
[ thm:6.3 ] @xmath381 iff @xmath382 * proof . *
as we showed , @xmath383^ * \cap d_0 $ ] and it is clear from the definitions that @xmath384^ * \cap d_0 \cap { \mathcal s}^{d+1}.$ ] note that @xmath385 thus , @xmath386 let @xmath387 denote the subspace of @xmath388 skew - symmetric matrices with real entries .
therefore , @xmath389 now , using elementary cone geometry on closed convex cones and the definitions , we have the following string of equivalences : ( t_0(k ) + d_0^ ) \{(e_i e_j^t - e_j e_i^t):i , j \{1,2 , , d } } & & ( t_0(k ) + d_0^ ) ^d+1 + & & ^ * d_0 ^d+1 + & & m_0(k ) = m(k ) . ' '' '' + [ cor:6.1 ] if @xmath390 then @xmath391 for every @xmath363 .
* proof . *
trivial for @xmath283 . by theorem [ thm:6.3 ]
, the assumption of the corollary implies @xmath392 . by theorem [ thm : ls ] ,
thus , @xmath393 now , applying theorem [ thm:6.3 ] recursively , we obtain the desired result . ' '' '' + let @xmath176 denote the complete graph on @xmath91 vertices , and consider the lp relaxation @xmath100 of the stable set problem on @xmath176 .
for every @xmath394 such that @xmath395 , we have @xmath396 thus , for every @xmath394 such that @xmath395 , we have @xmath397 this implies , @xmath398 therefore , the condition of theorem [ thm:6.3 ] is satisfied and we have @xmath399 for every @xmath363 . as in theorem [ thm : n = n+ ] , we obtain [ cor:6.2 ] @xmath392 if and only if for every @xmath367 , @xmath400 instead of comparing @xmath36 and @xmath341 , or @xmath115 and @xmath145 , we might ask when are the set of optimal solutions of both relaxations the same . this is precisely when @xmath401^ * + \pmatrix{-z^ * \cr c } \,\ , \supseteq \,\ , \left[n_+(k)\right]^ * + \pmatrix{-z^ * \cr c},\ ] ] where @xmath402 is the optimal value of @xmath403 . sometimes we are only interested in the bound provided by the relaxation .
this is equivalent to finding the smallest @xmath404 for which @xmath405^*$ ] and the smallest @xmath406 for which @xmath407^*$ ] .
* references * balas e. ( 1974 ) .
disjunctive programming : properties of the convex hull of feasible points .
_ management science research report _ 348 gsia , carnegie mellon university , pittsburgh , pa , usa . balas e. , s. ceria and g. cornujols ( 1993 ) . a lift - and - project cutting plane algorithm for mixed 0 - 1 programs .
* 58 * 295323
. cook , w. , and s. dash , private communication , 1999 .
doignon , j .-
convexity in cristallographical lattices . _ journal of geometry _ * 3 * 71 - 85 .
edmonds , j. ( 1965 ) . maximum matching and a polyhedron with 0,1-vertices .
_ journal of research of the national bureau of standards - b _ * 69b * 125130
. grtschel , m. , l. lovsz and a. schrijver ( 1981 ) . the ellipsoid method and its consequences in combinatorial optimization .
_ combinatorica _ * 1(2 ) * 169197 .
lovsz , l. and a. schrijver ( 1991 ) .
cones of matrices and set - functions and 0 - 1 optimization .
_ siam j. optim .
_ * 1 * 166190 .
stephen , t. and l. tunel ( 1999 ) . on a representation of the matching polytope via semidefinite liftings , _ math .
* 24 * 17 .
in this appendix , we prove proposition [ propapp ] and derive additional properties of @xmath305 and @xmath408 .
we first start with a few preliminary lemmas . on the other hand
, we have that @xmath417 iff @xmath418 iff @xmath419 , which follows by assumption .
furthermore , @xmath420 iff @xmath421 iff @xmath422 . as a result ,
both terms in the minimum are greater than @xmath423 , and the second part of each inequality follows . * proof . * for @xmath159 , @xmath451 and @xmath452 ,
let @xmath453 , @xmath454 and @xmath455 .
observe that @xmath456 , @xmath457 and @xmath458 , implying that @xmath459 and @xmath460 .
thus , lemma [ lemint2 ] implies the result for @xmath159 .
we now proceed by induction and assume the result true for @xmath461 .
defining @xmath462 , @xmath423 and @xmath305 as above , we know from corollary [ cor1 ] that @xmath459 and from the inductive hypothesis that @xmath463 .
lemma [ lemint2 ] then implies the result for @xmath95 .
* proof . *
the proof is by induction on @xmath95 .
the base case is obvious .
assume the result is true for @xmath174 .
this implies that @xmath470 and @xmath471 ; we denote respectively by @xmath462 and @xmath423 these two quantities .
the result would then follow if we can show that @xmath472 where @xmath473 .
this inequality is equivalent to @xmath474 , or to @xmath475 . since @xmath476
, we need to prove that @xmath477 this follows from corollary [ cordiff ] since we have that @xmath478 . | we study the lift - and - project procedures of lovsz and schrijver for 0 - 1 integer programming problems .
we prove that the procedure using the positive semidefiniteness constraint is not better than the one without it , in the worst case .
various examples are considered .
we also provide geometric conditions characterizing when the positive semidefiniteness constraint does not help .
* keywords : * semidefinite lifting , semidefinite programming , lift - and - project , integer programming * ams subject classification : * 90c10 , 90c27 , 47d20 |
ferromagnets have the tremendous potential for the development of spintronics and subsequently will lay the foundation to realize quantum computing .
the field of spintronics require the incorporation of the spin - property of the electrons into the existing charge transport devices @xcite .
parallel to this , the technological potential of dms ( ref .
@xcite ) is associated to spintronics - device development , whereas manganites that show a large drop of resistance below @xmath2 lead to the colossal magnetoresistance effect ( cmr ) , which is also important in the new technologies such as read / write heads for high - capacity magnetic storage and spintronics @xcite . as such
, applications involving both dms and manganites very much depend on our understanding of their transport properties at various doping levels and temperatures ( @xmath3 ) .
in addition , dms also has several interesting physical properties namely , anomalous hall - effect @xcite , large magnetoresistance in low dimensional geometries @xcite , the changes of electron - phase - coherence time in the presence of magnetic impurities @xcite and negative bend resistance @xcite . as for the transport properties ,
there are several models developed to characterize the resistivity of dms .
in particular , the impurity band model coupled with the multiple exchange interactions for @xmath4 for ga@xmath5mn@xmath6as was proposed @xcite .
the electronic states of the impurity band can be either localized or delocalized , depending on doping concentration or the fermi - level ( @xmath7 ) . if @xmath7 is in the localized - state , then the conduction is due to carrier hopping .
if @xmath7 is in the extended - state , then the conduction is metallic and finite even for @xmath8 ( ref .
@xcite ) . on the other hand ,
the spin disorder scattering resistivity as a function of magnetic susceptibility can be used to estimate the magnitude of @xmath9 ( the ferromagnetic ( fm ) exchange interaction energy ) @xcite .
moreover , there are also theories that qualitatively explain the conductivity for @xmath10 , namely , the kohn - luttinger kinetic exchange model @xcite and the semiclassical boltzmann model @xcite . apart from that , for manganites , the one- and two - orbital models @xcite and the phase separated resistivity model @xcite have been used to qualitatively describe the resistivity curves for @xmath10 . however , in all these approaches , we are faced with two crucial problems , the need ( i ) to explain how the resistivity evolve with different doping elements , without any _ a priori _ assumption on carrier density and ( ii ) to understand how defects and clustering affect the evolution of @xmath0 curves . here
, we show unequivocally , a new method to analyse the evolution of @xmath0 curves for different doping elements using the concept of the @xmath1 invoked in the hamiltonian and fermi - dirac statistics . in doing so
, we can also understand the evolution of @xmath0 curves in the presence of defects and clustering , which is important for characterization of spintronics devices .
the @xmath1 concept has broad applications , where it has been applied successfully for the normal state ( above critical temperature ) of high temperature superconductors @xcite and ferroelectrics @xcite .
the @xmath1 model is for compounds obtained via substitutional doping , not necessarily homogeneous or defect - free .
a typical solid contains 10@xmath11 strongly interacting particles . therefore , their universal collective behavior is of paramount interest as compared to the microscopic details of each particular particle and the potential that surrounds it .
this universal collective behavior , being the focal point in this work , arises out of anderson s arguments in _ more is different_. @xcite that is , we intend to justify a universal physical parameter that could be used to describe the association between the transport - measurement data and the fundamental properties of an atom . in view of this , we report here the existence of such a parameter through the hamiltonian as given below ( eq . ( [ eq:100 ] ) ) .
the parameter is the ionization energy , a macroscopic , many - electron atomic parameter . @xmath12 where @xmath13 is the usual hamilton operator and @xmath14 is the total energy at @xmath3 = 0 .
the + sign of @xmath15 is for the electron ( @xmath16 ) while the @xmath17 sign is for the hole ( @xmath18 ) . here
, we define the ionization energy in a crystal , @xmath19 is approximately proportional to @xmath1 of an isolated atom or ion .
now , to prove the validity of eq .
( [ eq:100 ] ) is quite easy because @xmath20 is also an eigenvalue and we did not touch the hamilton operator .
hence , we are not required to solve eq .
( [ eq:100 ] ) in order to prove its validity
. we can prove by means of constructive ( existence ) and/or direct proofs , by choosing a particular form of wavefunction with known solution ( harmonic oscillator , dirac - delta and coulomb potentials ) and then calculate the total energy by comparison .
in doing so , we will find that the total energy is always given by @xmath21 , as it should be ( see appendix and ref .
@xcite ) . for an isolated atom
, the concept of ionization energy implies that ( from eq .
( [ eq:100 ] ) ) @xmath22 where @xmath1 is the ionization energy of an isolated atom .
the corresponding total energy is @xmath23 .
whereas for an atom in a crystal , the same concept of ionization energy implies that @xmath24 . here , @xmath25 is the many body potential averaged from the periodic potential of the lattice .
the corresponding total energy is @xmath26 . here , @xmath27 is the ionization energy of an atom in a crystal .
the exact values of @xmath1 are known for an isolated atom .
that is , one can still use @xmath1 obtained from isolated atoms for @xmath28 in order to predict the evolution of resistivity versus temperature curves for different doping elements .
therefore , eq .
( [ eq:100 ] ) can be approximately rewritten as @xmath29 .
it is obvious from the hamiltonian given in eq .
( [ eq:100 ] ) that we can not use it to determine interactions responsible for fm and the magnitude of @xmath2 , simply because we have suppressed the @xmath25 . on the other hand ,
if we have a free - electron system , then the total energy equation is given by @xmath30 , where , @xmath14 is the total energy of the compound at @xmath3 = 0 and @xmath31 is the sum over the constituent elements in a particular compound .
we also define here , @xmath32 , where @xmath33 is the averaged many - body potential value . here , @xmath27 can be noted as one of the many - body response functions with respect to transport properties . the total energy can be rewritten as @xmath34 \nn \\ & & = e_{\rm{kin } } + v_{\rm{pot } } + v_{\rm{many - body } } \leftrightarrow ~\rm{for~ electrons}~ \pm \rightarrow + \nn \\ & & = e_{\rm{kin } } + v_{\rm{total } } \nn \\ & & = e_{\rm{kin } } \leftrightarrow \rm{implies~free~electrons}. \nn \end{aligned}\ ] ] note that we have substituted for @xmath27 since the concept of ionization energy is irrelevant here simply because the electrons in these metals are free and do not require excitations from its parent atom to conduct electricity . as such , the carrier density is constant and independent of temperature . whereas , the scattering rate is the one that determines the resistivity with respect to temperature , impurities , defects , electron - electron and electron - phonon interactions .
therefore , the total energy from eq .
( [ eq:100 ] ) carries the _ fingerprint _ of each constituent atom in a compound and it refers to the difference in the energy levels of each atom rather than the absolute values of each energy level in each atom .
hence , the kinetic energy of each electron from each atom will be captured by the total energy and preserves the atomic level _ electronic - fingerprint _ in the compound .
parallel to eq .
( [ eq:100 ] ) , the electron and hole distribution functions can be derived as @xcite ( see fig .
[ fig:1 ] ) @xmath35/k_bt } + 1 } , \nn \\ & & f_h(e_0,e_i ) = \frac{1}{e^{[e_f^{(0 ) } - \left(e_0 - e_i \right)]/k_bt}+1}. \label{eq:360}\end{aligned}\ ] ] where @xmath36 is the fermi level at @xmath3 = 0 , independent of doping concentration and @xmath37 is the boltzmann constant . from eq .
( [ eq:360 ] ) , one can surmise that large @xmath1 corresponds to the difficulty of the electron to conduct due to a large coulomb attraction between the electron and the ionic core
. this effect will give rise to low conductivity .
the variation of @xmath1 with doping in our model somewhat resembles the variation produced from the impurity band model with doping @xcite .
the difference is that our model fixes the fermi level as a constant and we let @xmath1 capture all the changes due to doping , consistent with ionization energy based fermi - dirac statistics ( @xmath38fds ) . whereas
, in the impurity - band approach , both the impurity band and the fermi level evolve simultaneously with doping , consistent with fermi - dirac statistics .
the carrier density can be calculated from @xmath39 where @xmath40 is the density of states and @xmath7 is defined from now on as the fermi level at @xmath3 = 0 so as to comply with eq .
( [ eq:100 ] ) . here , we will show the relationship of the ionization energy to the energy level splitting and the spin - orbit coupling , as well as their association with resistivity .
the latter association is crucial because spin - orbit coupling is found to be an an important phenomenon that influences the electronic properties of the magnetic semiconductors @xcite .
it is well established that the energy associated with both the relativistic correction and the spin - orbit coupling for hydrogen - like atoms is given by the fine structure formula , @xmath41 ( ref .
@xcite ) @xmath42 , \label{eq:999}\end{aligned}\ ] ] where @xmath43 is the fine structure constant , @xmath44 = @xmath31 + 1/2 , @xmath31 = @xmath45 @xmath46 1/2 , where @xmath31 is the total angular momentum , @xmath45 is the orbital angular momentum and @xmath47 is the non - relativistic energy level . @xmath48 and @xmath49 are the atomic and the principal quantum numbers , respectively . using eqs .
( [ eq:100 ] ) and ( [ eq:999 ] ) , we can show the extreme magnitude of the energy level splitting ( i.e. , between @xmath44 = 1 and @xmath44 = @xmath49 ) is @xmath50 thus , the magnitude of the energy level splitting ( @xmath51 ) is proportional to the ionization energy ( @xmath20 ) and @xmath52 , while it is inversely proportional to @xmath49 .
that is , @xmath51 also decreases with decreasing ionization energy ( see fig . [ fig:2 ] ) .
therefore , we can surmise that a system with large ionization energy gives rise to large spin - orbit coupling , which is important for spin - injection @xcite in @xmath53-type dms . as discussed earlier however
, a large ionization energy also leads to a small carrier density ( from eq .
( [ eq:3600 ] ) ) , which in turn implies that spin - orbit coupling competes with the electrical conductivity .
in addition , fig .
[ fig:2 ] points out that @xmath54 and @xmath55 are larger than @xmath56 and @xmath57 , respectively , in accordance with eqs .
( [ eq:999 ] ) and ( [ eq:9999 ] ) .
therefore , we can surmise that @xmath54 @xmath58 @xmath55 , @xmath56 @xmath58 @xmath57 , @xmath59
@xmath58 @xmath60 and so on . in order to derive the resistivity model as a function of ionization energy , we need an expression that connects the carrier density with the total current . as a consequence ,
we propose that the total current in ferromagnets consists of contributions from both non - fm and fm phases , which is @xmath61 = @xmath62 , with @xmath63 = @xmath64 , @xmath65 , @xmath66 , @xmath67 , @xmath68 , @xmath69 . here ,
@xmath65 and @xmath68 are the spin - assisted electron and hole , respectively below @xmath2 , influenced by the @xmath3-dependent magnetization function ( @xmath70 ) through the spin disorder scattering rate ( @xmath71 ) . @xmath64 and @xmath67 are the electron and hole with electron - electron scattering rate ( @xmath72 ) . @xmath66 and @xmath69 are the electron and hole with spin disorder scattering and @xmath70 = constant ( const . ) for @xmath73 . for convenience ,
the spin - up , @xmath74 denotes the direction of the magnetic field or a particular direction below @xmath2 , while the spin - down , @xmath75 represents any other direction .
that is , a non - fm phase gives rise to @xmath76 and @xmath77 .
whereas , @xmath78 originates from a fm phase where ( @xmath77,@xmath76 ) @xmath79 @xmath78 for @xmath80 . hence
, this resistivity model is fundamentally different from the phase separation model @xcite .
that is , @xmath81 ) @xmath82 @xmath83 $ ] ) for @xmath84 , while @xmath85 ) and @xmath86[const .
] ) contribute for @xmath73 due to the _ non - existence _ of the fm phase .
in other words , for @xmath87 , the whole system is a non - fm phase and its conductivity is determined by @xmath85 ) and @xmath86[const . ] ) . for @xmath4 ,
some of the non - fm phase becomes fm due to spin and its conductivity is determined by @xmath88 and @xmath83 $ ] ) .
simply put , we assume that @xmath72 and @xmath71[const . ] contribute in the non - fm phase above @xmath2 .
below @xmath2 , we have both fm and non - fm phases contributing to the resistivity via @xmath72 and @xmath89 $ ] . as such
, the total current can be simplified as @xmath61 = @xmath90 + [ @xmath91,@xmath92 = @xmath93 + [ @xmath94,@xmath95 if the considered system is @xmath96-type , while @xmath61 = @xmath97 + [ @xmath98,@xmath99 if it is @xmath53-type . @xmath93 and @xmath97 are the spin independent charge current due to electron and hole , respectively , and are influenced by @xmath72 .
@xmath100 and @xmath101 are also the spin independent charge current due to electron and hole , respectively , but are influenced by @xmath71[const .
] in the non - fm phase , which is valid only above @xmath2 .
@xmath94 and @xmath98 are the spin - assisted charge current in the fm phase only , influenced by the @xmath89 $ ] .
thus , the total resistivity ( for @xmath96 or @xmath53-type ) can be written as ( after making use of the elementary resistivity equation , @xmath102 ) @xmath103 = \bigg[\frac { m_{e , h}^*}{(n , p)e^2 \tau_{e - e}}\bigg]^{-1 } \nn \\ & & + \bigg[\frac { m_{e , h}^*}{(n , p)e^2 \tau_{sd}\big[{\rm{const.}},m(t , m_0)\big]}\bigg]^{-1 } \nn \\ & & = \bigg[\frac { m_{e , h}^*}{(n , p)e^2 \tau_{e - e}}\bigg]^{-1 } + \bigg[\frac { m_{e , h}^*}{(n , p)e^2 \tau_{sd}}\bigg]^{-1 } , \label{eq:4 } \end{aligned}\ ] ] where @xmath104 denotes the effective mass of the electron or hole .
@xmath105 is the charge of an electron .
the carrier density for the electron and hole ( @xmath106 ) based on @xmath38fds are given by ( after substituting eq .
( [ eq:360 ] ) into eq .
( [ eq:3600 ] ) ) @xmath107^{3/2}(m^*_e)^{3/2 } \exp\left[\frac{e_f - e_i}{k_bt}\right ] , \label{eq:5}\end{aligned}\ ] ] @xmath108^{3/2}(m^*_h)^{3/2 } \exp\left[\frac{-e_f - e_i}{k_bt}\right ] .
\label{eq:6}\end{aligned}\ ] ] the spin disorder scattering resistivity as derived by tinbergen - dekker is given by @xcite @xmath109 , \label{eq:3}\end{aligned}\ ] ] where @xmath110 is the tinbergen - dekker magnetization function @xcite .
@xmath111 is the magnetization at zero temperature .
@xmath112 is the concentration of nearest neighbor ions ( for example , the concentration of mn ) .
@xmath113 = @xmath114 , @xmath115 denotes plancks constant and @xmath116 is the spin quantum number .
equation ( [ eq:3 ] ) is equal to the theory developed by kasuya @xcite if one replaces the term , @xmath117 $ ] with 1 . substituting @xmath118 = @xmath119 ( due to
the electron - electron interaction and @xmath120 is the @xmath3 independent electron - electron scattering rate constant ) , together with eqs .
( [ eq:3 ] ) and ( [ eq:5])(or ( [ eq:6 ] ) ) into eq .
( [ eq:4 ] ) , then one can arrive at @xmath121}{at^{3/2}[m_{\alpha}(t , m_0)]^{-1}+ bt^{-1/2 } } , \label{eq:7}\end{aligned}\ ] ] @xmath122^{3/2},\nn \\ & & b = \frac{2m^*_{e , h}n(\pi e_f)^{1/2}j_{ex}^2}{e^2\hbar k_b^{3/2 } } , \nn \\ & & \tau_{sd}^{-1 } = \left[\frac{n(2e_f)^{1/2}(m^*_{e , h})^{3/2}}{\pi \hbar^4}\right]j_{ex}^2m_{\alpha}(t , m_0 ) .
\nn\end{aligned}\ ] ] note here that we have invoked the strong correlation through the carrier density , which is a function of the real ionization energy .
secondly , to account for different spin polarization below and above @xmath2 , we have @xmath96 = @xmath123 , @xmath124 and @xmath125 .
notice that the spin - disorder scattering rate , @xmath126 is a function of the exchange energy , @xmath9 .
therefore , by redefining the variables in the elementary resistivity equation in accordance with strongly correlated effects , we have now derived a resistivity equation ( eq .
( [ eq:7 ] ) ) suitable for strongly correlated ferromagnets . here , @xmath127 is the @xmath3-dependent magnetization function for @xmath4 . for @xmath87 however ,
@xmath127 = constant .
the negative and positive signs in @xmath128 are for electrons and holes , respectively .
equation ( [ eq:7 ] ) points out that , @xmath129 is semiconducting and it only becomes metallic if @xmath130 and the fm phase sets in .
below @xmath2 , the fm metallic phase increases with lowering @xmath3 and once it achieves the maximum value ( saturation ) at a much lower @xmath3 , insulating character may set in as a result of @xmath131 .
we define this temperature @xmath132 which corresponds to the transition from fm metallic to insulating below @xmath2 .
hence , it is clear that our model is fundamentally different to the phase separation model as pointed out earlier .
the emergence of insulating ( @xmath132 ) character below @xmath2 is an intrinsic property based on the ionization energy model , regardless of defect densities and it was first predicted for both high-@xmath133 superconductors and ferromagnets @xcite . in a recent analysis using the one- and two - orbital models , a similar insulating character below @xmath2 was observed from numerical calculations @xcite , in support of our prediction .
however , the analysis stops short of explaining why the insulating behavior persists even in the clean limit " with no defects ( i.e. pure host material with zero interstitial and vacancy defects ) . here
we propose that @xmath132 or the insulating behavior below @xmath2 is intrinsic and is associated with the energy levels through the @xmath134 parameter .
if @xmath135 , then @xmath136 $ ] . in this limit , a large cmr effect could be observed if the metallic - ferromagnetism sets in .
the empirical function of the normalized magnetization is defined here as @xmath137 equation ( [ eq:8 ] ) is an empirical function that will be used to extract the magnetization curve from the resistivity curve via eq .
( [ eq:7 ] ) . in other words , eq .
( [ eq:8 ] ) is used to calculate the magnetization curve , after coupling it with eq .
( [ eq:7 ] ) .
for example , the magnetization curves associated with tinbergen - dekker ( @xmath138 ) @xcite , kasuya ( @xmath139 ) @xcite and resistivity curves ( @xmath140 ) are calculated using @xmath141,\label{eq : b8}\end{aligned}\ ] ] @xmath142 and eq . ( [ eq:8 ] ) , respectively .
that is , eqs .
( [ eq : b8 ] ) , ( [ eq : b9 ] ) and ( [ eq:8 ] ) are separately coupled with eq .
( [ eq:7 ] ) to fit the resistivity curves and subsequently obtain @xmath143 , @xmath144 and @xmath145 , respectively .
consequently , we can compare and analyze @xmath146 with the experimentally measured , @xmath147 data .
recall that , @xmath148 = @xmath139 ( calculated from eq .
( [ eq : b9 ] ) ) , @xmath138 ( from eq .
( [ eq : b8 ] ) ) , @xmath140 ( from eq .
( [ eq:8 ] ) ) , exp ( experimentally determined magnetization curves ) . for simplicity
, we will drop the term @xmath111 from the equations that contain @xmath149 . for the electron- or hole - doped
strongly correlated non - fm semiconductors , one needs eq .
( [ eq:9 ] ) given below , which is again based on @xmath38fds , @xcite @xmath150 , \label{eq:9}\end{aligned}\ ] ] where the @xmath17 sign is for the electron and + sign is for the hole .
equation ( [ eq:9 ] ) will be used to justify the importance of the term @xmath94 even if the resistivity is semiconductor - like in the fm phase .
the values of @xmath1 can be averaged using the approximation given below @xmath151 where @xmath152 is a function of the many - body potential , @xmath25 and varies with different background " atoms ( host lattice ) .
the @xmath38 @xmath153 1 , 2 , ... , @xmath154 represent the first , second, ... etc ionization energies and , @xmath154 is the oxidation number of a particular ion .
@xmath155 is actually equal to the energy needed to ionize an atom or ion in a crystal such that the electron is excited to a distance @xmath156 . on the other hand
, @xmath1 corresponds to taking that particular electron to @xmath157 with @xmath158 .
prior to averaging , the ionization energies for all the elements mentioned above were taken from ref .
@xcite , also given in fig .
[ fig:3 ] .
we can also use eq .
( [ eq : z12 ] ) that originates from eq .
( [ eq:25 ] ) to predict the change in the valence state of a particular ion .
@xmath159 here , @xmath160 has @xmath38 @xmath153 @xmath154 + 1 , ... ,
@xmath154 + @xmath31 and @xmath31 @xmath153 1 , 2 , .... it is solely due to the multivalence ion .
that is , the first term is due to mn@xmath161 ion s contribution ( mn@xmath162 @xmath17 electron @xmath163 mn@xmath161 = 51.200 ev atom@xmath164 ) , hence @xmath31 equals 1 ( @xmath165 ) in this case and @xmath166 represents the additional contribution from mn@xmath161 . the second ( @xmath38 @xmath153 1 , 2 , ... , @xmath154 ) and last ( @xmath38 @xmath153 1 , 2 , ... , @xmath167 ) terms respectively are due to mn @xmath17 3(electrons ) @xmath163 mn@xmath162 and ga @xmath17 3(electrons ) @xmath163 ga@xmath162 .
recall that @xmath167 = @xmath154 = 3 + and @xmath38 = 1 , 2 , ... represent the first , second , ... ionization energies while @xmath31 = 1 , 2 , ... represent the fourth , fifth , ... ionization energies .
we will first apply the @xmath1-based carrier density to the resistivity ( @xmath0 ) curves of dms above @xmath2 by means of @xmath169 , in order to avoid the influence of fitting parameters in our analysis .
subsequently , we will discuss the evolution of resistivity curves with appropriate fittings .
the ga@xmath5mn@xmath6as system is the most studied material among the dms with @xmath2 around 170 k ( ref .
the origin of ferromagnetism and the transport properties in this class of materials are still unclear quantitatively @xcite .
the evolution of the resistivity curve ( @xmath170 ) for sample _ x1 _ , ga@xmath5mn@xmath6as ( see fig . 3 of ref .
@xcite ) is such that @xmath171 @xmath172 @xmath173 @xmath172 @xmath174 @xmath172 @xmath175 @xmath172 @xmath176 .
that is , the @xmath170 curve shifts downwards with increasing dopant concentration , @xmath177 .
this is expected from the @xmath38fds model since @xmath1 for mn@xmath162 ( 18.910 ev atom@xmath164 ) is less than @xmath1 for ga@xmath162 ( 19.070 ev atom@xmath164 ) , assuming mn@xmath162 substitutes ga@xmath162 . however , the resistivity curve , @xmath178 is above @xmath174 , unexpectedly . for sample
_ y2 _ ( see fig . 1 of ref .
@xcite ) , @xmath171 @xmath172 @xmath173 @xmath172 @xmath174 , also complies with the @xmath38fds model . again
, unexpectedly we have , @xmath175 @xmath82 @xmath176 @xmath82 @xmath178 , where @xmath170 shifts upwards with increasing @xmath177 .
that is , @xmath170 switched over from decreasing ( expected ) to increasing with @xmath177 at critical concentrations , @xmath179 = 0.071 and @xmath180 = 0.043 for samples _
x1 _ and _ y2 _ , respectively .
these switch - overs that seem to violate @xmath38fds can be explained if we can understand what causes @xmath1 to deviate from its averaged value .
the only reason that could give rise to such deviation is the change in the ions valence states .
using eq .
( [ eq : z12 ] ) , we obtain : @xmath181 + \frac{1}{3}[e_i { \rm{(mn^{3 + } ) } } + e_i { \rm{(mn^{2 + } ) } } + e_i { \rm{(mn^{1 + } ) } } ] = \frac{1}{3}[e_i ( { \rm{ga^{3 + } } } ) + e_i ( { \rm{ga^{2 + } } } ) + e_i ( { \rm{ga^{1+}}})]$ ] and @xmath182 , therefore @xmath183 . since the average @xmath1 ( 18.910 ev atom@xmath164 ) for mn@xmath162 is less than the average @xmath1 ( 19.070 ev atom@xmath164 ) for ga@xmath162 , and if the resistivity curve shifts upward with mn substitution , then @xmath154 + @xmath166 gives the minimum valence number for mn ( mn@xmath184 ) which can be calculated from eq .
( [ eq : z12 ] ) . if however , the resistivity curve shifts downward with mn substitution , then @xmath154 + @xmath166 gives the maximum valence number for mn ( mn@xmath185 ) .
consequently , the maximum valence state for mn in ga@xmath5mn@xmath6as is 3.003 + ( mn@xmath186 ) for the case where the @xmath170 curve shifted downwards with @xmath177 .
if the valence state of mn is larger than 3.003 + , then @xmath170 is expected to shift upwards with @xmath177 .
therefore , the switch - over from decreasing to increasing @xmath170 with doping , @xmath177 is due to larger average mn valence state , and it is calculated to be larger than 3.003 + .
next , we need to understand what can cause this change in the average mn valence state .
the most likely reason comes from mn occupation at non - substitutional sites , i.e. , the mn@xmath162 ions do not substitute ga@xmath162 ions .
this has been predicted experimentally where the formation of mn interstitials ( mn@xmath187 ) is found to be substantial above a critical concentration , @xmath188 ( ref .
interestingly , occupations at interstitial sites is found to change the charge states of mn ions , as predicted from the first - principles calculations , where mn@xmath189 and mn@xmath161 are more stable at interstitial sites , while mn@xmath190 are stable at substitutional sites @xcite .
in addition , interstitial in the presence of clustering gives rise to larger positive charge states as compared to clustering due to substitutional alone in which , larger positive charge implies larger valence state for mn ions @xcite .
therefore , we can understand that at certain doping , the average mn valence states increased due to mn@xmath187 and clustering and eventually reduces the carrier density and shifts the @xmath170 curve upwards with doping .
now we will apply the ionization energy incorporated resistivity model to the resistivity measurements @xcite and the fits based on eqs .
( [ eq:7 ] ) and ( [ eq:9 ] ) are shown in figs .
[ fig:4](a ) and [ fig:4](b ) respectively for ga@xmath5mn@xmath6as .
one needs two fitting parameters ( @xmath192 and @xmath134 ) for @xmath193 and another two ( @xmath194 and @xmath195 ) for @xmath196 .
all the fitting parameters are listed in table [ table : i ] .
.calculated values of @xmath192 , @xmath194 and the ionization energy ( @xmath1 ) .
@xmath3 " denotes the annealing temperature and @xmath197 = @xmath198 , is the temperature that corresponds to the transition from fm metallic to insulating character .
" denotes calculated values from the fits . [ cols="<,^,^,^,^,^ " , ] [ table : i ] note that @xmath116 = 1 and 5/2 are used for the fits of @xmath199 while @xmath2 and @xmath198 ( @xmath197 ) are determined from the experimental resistivity curves .
the @xmath197 s observed in ga@xmath200mn@xmath201as ( annealed : 370@xmath202c ) and ga@xmath203mn@xmath204as ( as grown ) are 10 k and 12 k , respectively , which are very close to the calculated values of 8 k and 12 k , respectively from eq .
( [ eq:7 ] ) .
the calculated carrier density is 2 @xmath205 10@xmath206 @xmath207 , using @xmath134 @xmath208 10 k , and eq . ( [ eq:5 ] ) . in this calculation , we take the effective mass as @xmath209 = 10@xmath210 , in accordance with the optical spectroscopy measurements @xcite and compensation due to annealing , where @xmath210 is the rest mass of an electron . figures [ fig:4](c ) and [ fig:4](d ) show the normalized magnetization , @xmath211 .
@xmath212 are compared with the experimentally determined magnetization @xcite ( @xmath213 ) as depicted in fig .
[ fig:4](d ) .
one can easily notice the inequality , @xmath214 @xmath172 @xmath215 @xmath172 @xmath216 @xmath172 @xmath213 from figs . [ fig:4](c ) and [ fig:4](d ) . as such , @xmath216 is the best fit as compared with @xmath213 , better than the models developed by tinbergen - dekker @xcite and kasuya @xcite . another dms material that we will consider here is the mn@xmath6ge@xmath5 system .
it is a @xmath53-type dms , with carrier density of the order of @xmath217 @xmath207 for 0.006 @xmath218 @xmath177 @xmath218 0.035 at room temperature @xcite . from the @xmath1 model ,
mn@xmath162 ( 18.910 ev atom@xmath164 ) substitution into ge@xmath161 ( 25.941 ev atom@xmath164 ) sites will shift the @xmath0 curve downwards since ( @xmath1)@xmath219 @xmath172 ( @xmath1)@xmath220 . this has been observed experimentally ( see fig . 2b of ref .
@xcite ) where , @xmath221 @xmath172 @xmath222 @xmath172 @xmath223 .
subsequently , we can estimate the maximum valence state for the mn ion in mn@xmath6ge@xmath5 using eq .
( [ eq : z12 ] ) and we find , @xmath224 or mn@xmath225 : @xmath226 + \frac{1}{3}[e_i ( { \rm{mn}}^{3 + } ) + e_i ( { \rm{mn}}^{2 + } ) + e_i ( { \rm{mn}}^{1 + } ) ] = \frac{1}{4}[e_i ( { \rm{ge}}^{4 + } ) + e_i ( { \rm{ge}}^{3 + } ) + e_i ( { \rm{ge}}^{2 + } ) + e_i ( { \rm{ge}}^{1+})]$ ] .
therefore , ( @xmath227 and @xmath224 . if the valence state of mn is larger than 3.137 + , then the @xmath170 curve is expected to shift upwards with @xmath177 .
the result , @xmath228 indicates that mn@xmath162 in this doping range may not have substituted ge , instead it could have occupied non - substitutional sites that eventually causes the change in the valence state of mn . since the resistivity , @xmath228 , we can calculate the maximum increment of mn@xmath161 content from @xmath177 = 0.02 to @xmath177 = 0.033 .
using @xmath166 = 0.137 , we estimate that there are 13.7% more mn@xmath161 in mn@xmath229ge@xmath230 as compared to mn@xmath231ge@xmath232 samples .
resistivity measurements @xcite and the fit using eq .
( [ eq:7 ] ) are shown in fig .
[ fig:5](a ) . from the fit , we find that @xmath134 = 15 k for mn@xmath231ge@xmath232 and the hole density is 2.38 @xmath205 10@xmath233 @xmath207 using eq .
( [ eq:5 ] ) , where we take @xmath234 = @xmath210 , as the lower limit .
this lower limit value is still comparable with the experimental value given above .
we also find that the semiconductor - like behavior of @xmath223 below @xmath2 is not exponentially driven as indicated by eq .
( [ eq:7 ] ) in fig .
[ fig:5](a ) . the pronounced effect of eq .
( [ eq:8 ] ) can be noticed by comparing the calculated plots between eq .
( [ eq:7 ] ) , and eq . ( [ eq:7 ] ) with the additional constraint @xmath235 = 0 , as depicted in fig .
[ fig:5](a ) .
the normalized magnetization , @xmath236 for mn@xmath231ge@xmath232 is given in fig .
[ fig:5](b ) .
again , we find that @xmath216 gives the best fit for the experimental data , which is better than @xmath215 and @xmath214 .
however , @xmath216 is significantly larger than @xmath213 , which makes the fit poor .
the reason is that the resistivity measures only the lowest @xmath1 path , regardless of the temperature and with easily - aligned spin path that gives rise to high conductivity and complies with the principle of least action .
that is , the ability of both @xmath93 and @xmath94 to follow the easiest path .
on the contrary , magnetization measurements quantify the average spin of an ensemble of electrons , which will be smaller in magnitude as compared with transport measurements . for manganites ,
the la@xmath5ca@xmath6mno@xmath237 system has a maximum @xmath2 of 260 k ( ref .
@xcite ) . unlike dms
, the resistivity of this class of materials above @xmath2 is exponential and thus gives rise to the large drop in resistance below @xmath2 that leads to the cmr effect .
we will evaluate this scenario with the ionization energy concept and show the validity of @xmath38fds in both dms and manganites .
based on @xmath1 model , ca@xmath238 ( @xmath1 = 9.000 ev atom@xmath164 ) @xmath82 la@xmath162 ( @xmath1 = 11.940 ev atom@xmath164 ) , therefore the resistivity @xmath170 curve is expected to shift downward with ca@xmath238 doping . on the contrary , the overall @xmath0 curve between @xmath177 = 0.1 and 0.2 , above @xmath2 are almost identical ( see fig . 3 of ref .
again , this could be due to the change in the valence state of mn , as a result of mn , ca and/or la occupying the non - substitutional sites , provided that the valence states of ca@xmath238 and la@xmath162 are invariant to doping and defects .
indeed , the content of mn@xmath161 is found to increase with ca doping @xcite , from 19% ( for @xmath177 = 0.1 ) to 25% ( for @xmath177 = 0.2 ) .
we can calculate the maximum increment of mn@xmath161 from @xmath177 = 0.1 to 0.2 , and we obtain , 5.75% using eq .
( [ eq : z12 ] ) : @xmath239 + \frac{1}{2}[e_i ( { \rm{ca}}^{2 + } ) + e_i ( { \rm{ca}}^{1 + } ) ] = \frac{1}{3}[e_i ( { \rm{la}}^{3 + } ) + e_i ( { \rm{la}}^{2 + } ) + e_i ( { \rm{la}}^{1+})]$ ] . therefore , @xmath240 and @xmath241 .
this value is remarkably close to the experimental value of 6%(=25%@xmath1719% ) , determined via redox titrations @xcite .
this implies that , @xmath242 , which in turn implies that the averaged many - body or lattice potential ( @xmath152 ) can be approximated as a constant .
this comes as no surprise because the averaged crystal or lattice potential is indeed a constant due to the periodicity .
using eq .
( [ eq:7 ] ) for la@xmath5ca@xmath6mno@xmath237 @xcite , @xmath134 is calculated for @xmath177 = 0.1 and 0.2 samples and we obtain 0.121 ev ( 1400 k ) and 0.112 ev ( 1300 k ) respectively .
the calculated carrier density , using @xmath243 = @xmath210 and eq .
( [ eq:5 ] ) gives 10@xmath244 @xmath207 . in the presence of the magnetic field , * h * = 6 tesla
, we obtain @xmath134 = 0.0776 ev for @xmath177 = 0.2 and its hole concentration @xmath53 = 10@xmath245 @xmath207 .
the fits are shown in figs .
[ fig:6](a ) and [ fig:6](b ) , while the fitting parameters are listed in table [ table : i ] .
the @xmath166 value determined previously ( 5.75 % ) needs to be corrected because in the previous calculation , we used @xmath246 that gives maximum @xmath166 . therefore , the actual increment after the fitting is @xmath247 .
figures [ fig:6](c ) and [ fig:6](d ) depict the calculated @xmath211 with @xmath116 = 1 and @xmath213 for @xmath177 = 0.2 , respectively .
it can be seen that the magnetization curve calculated from eq .
( [ eq:8 ] ) shows better agreement with the experimental data than @xmath199 .
hence , the ionization energy model is suitable for both types of ferromagnets , be it diluted or concentrated .
however , this does not imply that the ferromagnetic interactions are identical between diluted and concentrated ferromagnets .
in conclusion , we have developed a theoretical model based on the ionization energy concept that can be used to analyze the evolution of the resistivity versus temperature curves of both diluted and concentrated ferromagnets , for different doping elements . by identifying the cause that deviate the ionization energy from its averaged value ,
we come to understand how defects and clustering contribute to the changes in valence states of ions and eventually how they affect the resistivity of ferromagnets .
a.d.a . is grateful to the school of physics , university of sydney for the usirs award , and kithriammah soosay for the partial financial support .
special thanks to a. stroppa for his explanation on the half - metallic character of mnge and mnsi . x.y.c . and c.s .
gratefully acknowledge support from the australian research council ( arc ) .
k.r . acknowledges partial funding from the malaysian grant no .
saga 66 - 02 - 03 - 0077 .
author - contributions ; a.d.a . designed the overall structure of the theory , developed and explained all the ideas related to the theory with proofs , carried out all the analysis and wrote both the manuscript and the appendix ; x.y.c and c.s . contributed to the idea that the valence states in the ionization energy theory can be related to the first - principles charge states ( in section iii - a-1 ) , and edited the manuscript ; k.r . edited the manuscript .
the relation between the electron affinity and ionization energy is through @xmath156 , where the electron affinity is connected to @xmath28 , and for any other finite @xmath156 , electron affinity is undefined microscopically and it is not a good quantum mechanical variable , as opposed to our @xmath20 .
for example , if we were to work with the electron affinity then we will face two problems , firstly , considering the anions , electrons from different cations will be treated with equal affinity toward the anion , which is incorrect . secondly
, considering the cations , there is no mathematical analogy as the first electron affinity , second , etc . , which we have used for first , second , etc .
ionization energies .
hence , the electron affinity in our case , is not a good many - body variable , except for qualitative descriptions . as for the reason behind the name ,
ionization energy is that in the early stages of the ionization energy theory , we used the atomic ionization energy ( @xmath28 ) as the input parameter to compute carrier concentrations @xcite , and for this reason , @xmath20 was labeled as the ionization energy .
the technically correct label is the excitation energy .
we can also explain the resistivity above the curie temperature with any exponential function that either contains the activation energy or the energy related to the electrons hopping rates . however , the whole point of our theory is to associate our exponential function ( as given in eq .
( [ eq:7 ] ) ) with each constituent atoms that exist in a given compound .
this enable us to estimate and predict the evolution of the carrier density and the resistivity , before the experiments are carried out .
on the contrary , the theories with activation energy and with hopping rates do not give us that freedom , hence these other approaches can not predict the evolution of the resistivity for different doping _ a priori_. in other words , these other approaches need to estimate the carrier density by other means to feed into their theory so as to predict the doping dependent resistivity . apart from that
, our resistivity theory captures the resistivity curve completely , from @xmath10 to @xmath87 with a single equation , as presented by eq .
( [ eq:7 ] ) .
we do not have one equation for @xmath87 , another for @xmath4 and another one for @xmath248 .
our theory captures all three mechanisms with a single equation , which will be very useful for experimental evaluations .
however , this approach has not been developed to the extend where we can apply it to magnetic heterostructures to estimate the spin transmission probability as discussed by egues @xcite , dai et al .
@xcite , and papp and peeters @xcite .
the proof that is given in the subsequent section is a form of proof - of - existence .
this appendix is to prove that eq .
( [ eq:100 ] ) is mathematically and physically correct , regardless of the type of potential used . in order to prove this
, we used the simplest case of 1d harmonic oscillator for convenience .
even if we were to use the hydrogenic wavefunction , we will still arrive at eq .
( [ eq:100 ] ) as derived in ref .
@xcite . in mathematics
, there are many techniques to prove the existence of a solution for a differential equation .
of course , one of the techniques is to _ solve _ the equation .
there are also the so - called existence and direct proofs .
now , to prove the validity of eq .
( [ eq:100 ] ) is quite easy because @xmath20 is also an eigenvalue and we did not touch the hamilton operator .
hence , we are not required to solve eq .
( [ eq:100 ] ) in order to prove its validity . in other words , we can prove by means of constructive ( existence ) and/or direct proofs , by choosing a particular form of wavefunction with known solution and then calculate the total energy by comparison .
in doing so , we will find that the total energy is always given by @xmath21 , as it should be .
for example , for 1d harmonic oscillator , the known solution is a gaussian function , for 1d dirac - delta potential , the known solution is an exponential function , for a 1d square well potential , the known solution is a sinusoidal function , for a 3d coulomb potential in hydrogen atom , the known radial solution is an exponential function .
therefore , it depends on which system we are interested in and if we are interested in 1d harmonic oscillator , then we write the solution for eq .
( [ eq:100 ] ) in its gaussian form and then we derive the schrodinger equation and its wavefunction in terms of @xmath21 . after that , we compare this new exponential wavefunction with the known solution for 1d harmonic oscillator .
when we compare the constants , we will be able to show that the total energy is always given by @xmath21 .
such that @xmath15 is the energy needed for a particle to overcome the bound state and the potential that surrounds it .
the + sign for @xmath15 is for the electron ( @xmath16 ) while the @xmath17 sign is for the hole ( @xmath18 ) . @xmath254 and @xmath14 denote the total energy at @xmath255 = 0 and the energy at @xmath3 = 0 , respectively , i.e. , @xmath254 = kinetic energy . in physical terms , @xmath20 is defined as the ionization energy . that is , @xmath20 is the energy needed to excite a particular electron to a finite distance , @xmath156 , not necessarily @xmath28 . where @xmath113 @xmath153 @xmath114 , @xmath115 is planck s constant and @xmath264 is the frequency
therefore , we obtain @xmath265 , @xmath266 and @xmath267 $ ] . on the other hand
, we can rewrite eq .
( [ eq : a7 ] ) to obtain where @xmath253 and @xmath14 for a given system range from @xmath269 to 0 for electrons and 0 to @xmath270 for holes which explains the @xmath46 sign in @xmath20 .
using eqs .
( [ eq : a1 ] ) , ( [ eq : a7 ] ) and , ( [ eq : a12 ] ) , we obtain @xmath271 .
normalizing @xmath261 gives in this section , we will show that the total energy , @xmath21 is a function of the potential energy . that is , using eqs .
( [ eq : a1 ] ) , ( [ eq : a2 ] ) and ( [ eq : a7 ] ) , we will show the potential energy can be written in terms of @xmath21 .
for example , from eq .
( [ eq : a17 ] ) the harmonic oscillator schrodinger eq . can be shown as | the evolution of resistivity versus temperature ( @xmath0 ) curve for different doping elements , and in the presence of various defects and clustering are explained for both diluted magnetic semiconductors ( dms ) and manganites .
here , we provide unambiguous evidence that the concept of ionization energy ( @xmath1 ) , which is explicitly associated with the atomic energy levels , can be related quantitatively to transport measurements .
the proposed ionization energy model is used to understand how the valence states of ions affect the evolution of @xmath0 curves for different doping elements .
we also explain how the @xmath0 curves evolve in the presence of , and in the absence of defects and clustering .
the model also complements the results obtained from first - principles calculations . |
the goods - n field @xcite covers @xmath5arcmin@xmath6 centered on the hubble deep field north @xcite and is unrivaled in terms of its ancillary data .
these include extremely deep _ chandra _ , _ hubble space telescope _ and _ spitzer _ observations , deep @xmath7 ground - based imaging and @xmath03,500 spectroscopic redshifts from 810-m telescopes .
previous radio observations of this region , however , fell short of complementing this unique dataset .
cccc nov - dec 1996 & 42 & ar368 & a + feb 2005 & 28 & am825 & b + aug 2005 & 7 & am825 & c + dec 2005 & 2 & am825 & d + feb - apr 2006 & 86 & am857 & a + radio emission is a relatively unbiased tracer of star formation and can probe heavily obscured active galactic nuclei ( agn ) objects that are missed by even the deepest x - ray surveys .
radio observations thus allow us to fully exploit the wealth of data taken at x - ray through millimeter wavelengths , providing a unique extinction - free probe of galaxy growth and evolution through the detection of starbursts and agn .
the recent imaging of @xcite ( om08 ) ( @xmath8jybeam@xmath2 at 1.4ghz ) have shown that the techniques exist to make radio images that approach the theoretical noise limit . to this end , we have obtained new , deep radio imaging of the goods
- n field .
while goods - n was selected to be free from bright sources at optical wavelengths , the field contains several very bright radio sources which place severe limitations on the dynamic range that can be obtained and hence the ultimate sensitivity of the radio map . before new techniques were developed to deal with this issue , only moderately - deep radio imaging was possible @xcite .
the earliest vla data in the goods - n field were reprocessed using new techniques by @xcite and @xcite , achieving a noise level of 5.75.8@xmath1jybeam@xmath2 a 23 - 25% improvement on the original map by @xcite , and close to the theoretical noise limit when one considers the increase in system temperature at the low elevations permitted during the original observations .
while the reduction of @xcite provided improved access to the @xmath1jy radio population , even deeper radio imaging is required to properly complement the extremely deep goods _
mid - infrared data , as well as forthcoming deep observations at far - infrared and submillimeter wavelengths from _ herschel _ , scuba-2 , the large millimeter telescope , and other facilities . to this end
, we have added 123hr to the existing data .
the reduced full resolution image ( beam=1.7 ) and its rms map are available onlinemorrison / goodsn and at the nasa / ipac infrared science archive ( irsa ) http://irsa.ipac.caltech.edu/ as an ancillary data product associated with the goods spitzer legacy survey ] .
the paper is laid out as follows : in 2 we describe the observations and the reduction of the data . in
3 we discuss the cataloging of radio emitters .
4 contains the results and some discussion of the catalog .
we present our conclusions in 5 .
in 1996 november , @xcite observed a region centered at 12:36:49.4 , + 62:12:58 ( j2000 ) for a total of 50hr at 1.4ghz using the national radio astronomy observatory s ( nrao s ) vla in its a configuration . of this , only 42hr was considered usable by @xcite . adopting the same position and frequency
, we obtained 28hr of data in the vla s b configuration in february
april 2005 , 7hr in c configuration in august 2005 , and 2hr in d configuration in december 2005 , and 86hr in a configuration in 2006 february april ( see table [ table1 ] ) for a useful combined total of 165hr .
observations were done at night to avoid solar interference .
we followed the 1:4 scaling of visibility data between the arrays described by om08 .
this empirically derived scaling relation provides for more uniform weighting of @xmath9 data
. in most regards the new observations were taken using the same parameters as those used in 1996 .
however , the integration time was changed from 3.33s to 5s because of difficulties experienced by the correlator with the shorter integration time @xcite .
the data were all obtained using spectral - line mode 4 , which yields [email protected] channels in each of two intermediate frequencies ( ifs ) , centered at 1,365 and 1,435mhz , in each of two circular polarizations . the channel width and integration time were compromises chosen to maximize reliability and sensitivity while minimizing bandwidth ( radial ) and time ( tangential ) smearing effects , respectively .
the upcoming evla correlator , widar , will offer much shorter integration times , narrower channels , and greater overall bandwidth .
@xmath10 was used to reduce and analyze all the radio data .
the first step was the calculation and corrections of the spectral bandpass shape .
this was done in the following manner using the task bpass .
the bright point - source phase calibrator was split from the raw database and phase self - calibration was applied .
the self - calibrated data were then used to calculate a bandpass correction and flatten the spectral response across the band for the uncalibrated multi - source database .
standard flux density calibration was applied next , using the baars flux - density scale with 3c286 @xcite as the calibrator .
the antenna - based weights for each 5-s integration were also calibrated .
the @xmath9 data for the target field were split from the database and clipped using the @xmath10 task clip at a level well above the total flux density found in the field to remove any interference prior to the self - calibration process .
only minor interference was encountered during the observations .
in verifying the astrometry of the a - array data from 1996 and then again in 2006 , we found a rotation between the two astrometric frames .
the rotational offset was about 1 at a radial distance of 20 from the phase center , and is likely the cause of the optical
radio offsets reported regularly over the last decade ( e.g. * ? ? ?
the offset problem was traced back to the vla on - line system ( the so - called modcomps ) and its calculated @xmath11 and @xmath12 terms .
this problem was corrected by using the aips task uvfix .
this task computes values of @xmath11 and @xmath12 using the time index and baseline information recorded during the observations .
the @xmath10 task uvfix was used on all the @xmath9 datasets after the initial bandpass and flux calibration .
bright sources and their sidelobe patterns affect the noise structure and the dynamic range of the final reduced maps if they are not cleaned .
such sources were located using d - configuration data .
we imaged to a radius of 8degrees using a pixel size of 30 .
the goods - n field was chosen to be free from bright objects at most wavelengths , but not the radio waveband . as a result ,
bright radio sources limit the dynamic range .
we imaged sources as far as 1.5degrees from the phase center a total of 56 facets , selected to cover distant , bright sources , plus another five facets within the primary beam to help with the peelr process which is fully discussed in om08 .
each facet is constructed from the fourier transform of the data , phase shifted to its tangent point on the celestial sphere . in all , 98 facets were used .
this technique is known as polyhedral imaging ( also known as faceting ) and is discussed at length by @xcite .
one night of a - configuration data was used to lay out 37 facets to cover the primary beam at 1.4ghz .
each of the 37 facets comprised 1024@xmath13 pixels , each 0.5 pixels .
the other 56 facets were 512@xmath13-pixels in size , again with 0.5@xmath13pixels , centered on the bright , outer sources .
initial cleaning of the fields with imagr revealed faint sources within the 98 facets and tight clean boxes were placed around each one , thus limiting cleaning to real features .
maps were cleaned down to the 1-@xmath4 noise level to create clean components for use in the self - calibration process .
clean boxes are a vital aspect of this process .
they avoid the possibility of cleaning into the noise , which can change the noise structure of a map and result in so - called ` clean bias ' .
failure to use clean boxes can artificially reduce the noise .
this one night of @xmath9 data were first self - calibrated in phase only , via the @xmath10 task calib , and a new map was made .
self - calibration in amplitude and phase followed .
the clean components from this one night were used to construct the fiducial model for the remaining 13 nights of new a - configuration data ( am857 nrao project number ) as well as the seven nights of 1996 ( ar368 ) a - configuration data . for the four nights of new b - configuration data ( am825 ) ,
one of the b - configuration datasets were used to provide the clean model for self - calibration .
the c- and d - configuration data ( am825 ) were bootstrapped to the clean model in a similar fashion .
editing of the @xmath9 data was done before and after the self - calibration process using the @xmath10 task tvflg .
after each episode of self - calibration , the clean component model was removed from the @xmath9 database using uvsub .
uvplt was then used to determine the residual amplitude distribution .
any high residual points were deleted with clip , except those on the shortest baselines .
the 165hr of data resulted in a very large database so we reduced the data volume by exploiting the fact that the a- and b - configuration ( programs am825 and am857 ) observations were taken at the same hour angles ( @xmath143.5hr of transit ) over many nights .
the smaller hour angle interval helped mitigate the higher system temperature that occurred at lower dish elevations .
this is due to the fact that the antenna feeds see the ground at low elevations .
the ar368 observing tracks covered @xmath147.5hr of transit yielding extended periods of low - elevation observations .
the antenna weights were calibrated as part of the standard calibration process minimizing the impact of the more noisy low - elevation data .
finally , stuffr is used to average data within a small range of u , v ( so there is no effect on the image ) , which was weighted properly . in this way the number of data points used for the imaging were reduced by an order - of - magnitude .
this is fully described in om08 .
the five separate datasets were combined into one large @xmath9 dataset using dbcon .
this was then divided up by hour angle , by stokes parameter ( ll and rr , separately ) and by if , thus yielding eight separate datasets .
this last step minimizes the problem of ` beam squint ' wherein the two circular polarizations have different pointing centers due to the slight off - axis location of the feeds .
this , combined with the alt - az telescope mount , causes the effective gain for an off - axis source to vary as a function of time . moreover
, the two ifs are at different frequencies and so yield a different primary beam shapes hence correspondingly different gains for off - axis sources .
crrrccccccc 5 & 12 34 3.51 0.061 & 62 14 20.7 0.03 & 27.6 & 191.6@xmath14 6.9 & 795.9@xmath14 44.9 & 2.6 & 0.7 & 95 & 0.0 & 1.7 + 9 & 12 34 8.79 0.164 & 62 9 20.5 0.10 & 8.0 & 51.6@xmath14 6.5 & 142.4@xmath14 19.5 & 0.0 & 0.0 & 0 & 3.8 & 1.7 + 12 & 12 34 9.23 0.098 & 61 56 45.5 0.09 & 44.3 & 397.0@xmath14 8.9 & 2767.1@xmath14 171.5 & 7.2 & 0.0 & 129 & 0.0 & 6.0 + 13 & 12 34 9.87 0.238 & 62 3 57.6 0.12 & 7.3 & 57.3@xmath14 7.9 & 208.6@xmath14 30.4 & 2.8 & 0.0 & 89 & 0.0 & 1.7 + 14 & 12 34 10.61 0.085 & 62 6 15.8 0.05 & 18.6 & 132.1@xmath14 7.1 & 564.4@xmath14 43.5 & 2.4 & 0.8 & 67 & 0.0 & 1.7 + 18 & 12 34 10.78 0.083 & 62 2 54.8 0.05 & 19.2 & 173.4@xmath14 9.0 & 848.0@xmath14 63.7 & 2.6 & 0.5 & 63 & 0.0 & 1.7 + 22 & 12 34 11.16 0.029 & 62 16 17.3 0.02 & 239.1 & 1674.0@xmath14 7.0 & 7741.0@xmath14 236.7 & 8.5 & 1.6 & 126 & 0.0 & 6.0 + 23 & 12 34 11.74 0.003 & 61 58 32.5 0.00 & 438.5 & 4385.0@xmath14 10.0 & 27684.0@xmath14 834.7 & 1.7 & 0.6 & 52 & 0.0 & 1.7 + 35 & 12 34 16.09 0.071 & 62 7 22.4 0.05 & 22.8 & 127.5@xmath14 5.6 & 346.2@xmath14 18.2 & 0.0 & 0.0 & 0 & 3.7 & 3.0 + 36 & 12 34 16.17 0.157 & 62 25 58.8 0.13 & 8.6 & 66.1@xmath14 7.7 & 284.2@xmath14 34.6 & 0.0 & 0.0 & 0 & 3.6 & 1.7 + after self - calibration , some artifacts associated with two bright sources ( within the 37 facets that map out the primary beam ) remained . a local self - calibration technique in @xmath10 called , peelr was used to remove these sidelobes .
the method is described as follows . in brief
, peelr subtracts the clean component ( cc ) model for all facets from the @xmath9 data , except for the facet containing the source to be ` peeled ' . then peelr self - calibrates using this information ( flux ) in this facet , writes a new calibration table .
next , peelr removes the ccs for the chosen field , thereby removing the bright source . the special calibration for this one field aids in the sidelobe correction process .
peelr then goes back to the original calibration tables and adds the ccs back to the @xmath9 dataset .
having the @xmath9 data in eight separate files allows for a more accurate cc model , enabling more accurate removal of these bright sources .
full details on this important process can be found in om08 .
the final maps were constructed by combining the 37 central facets made with each of the eight @xmath9 datasets , weighting them by @xmath15 , using flatn .
the r.m.s .
, before correcting for the shape of the primary beam , is @xmath03.9@xmath1jy over a region of 100@xmath13 pixels .
the synthesized beam size is [email protected] with a position angle of @xmath165degrees .
the final step was to run @xmath10 task mwflt which low - pass filters the image with a 101@xmath13-pixel kernel to yield a more uniform background .
the @xmath10 task , rmsd , was used to construct a noise image , calculating a histogram based r.m.s . for each pixel using the surrounding pixels within a radius of 100 pixels .
a multi - iteration process rejects pixels outside of the @xmath17 range , thereby offering a more robust estimate of the noise .
[ fig1 ] shows contours of constant noise over the central region , after correcting for the primary beam response .
the angular size of discrete sources in the image are broadened by three effects : ( 1 ) the finite bandwidth of each channel where @xmath18 where @xmath19 is the fractional bandwidth ( i.e. 3.125mhz ) and @xmath20 is the distance from the phase center ; ( 2 ) is the finite data sampling rate ( estimated at a few percent ) ; ( 3 ) and the true angular size of the source .
thus , in order to detect sources above 4.5-sigma , images at 3and 6are also needed to detect extended sources whether truly extended or experimentally extended .
om08 showed that 3 and 6 maps are a useful complement to the full - resolution images to recover the full flux of extended sources .
the 3 and 6 images were made using the @xmath10 task , convl .
the @xmath10 task , sad , ( ` search and destroy ' ) was used to generate an initial source catalog . for each resolution , sad was used in ` signal - to - noise ratio ( snr ) mode ' to search for peaks more than 4.5@xmath3 the local noise and to correct for radial smearing and primary beam attenuation .
this mode uses the noise maps created by rmsd .
the final catalog was clipped at @xmath215@xmath4 .
residual maps created by sad were searched by eye to find sources missed by the automatic procedure .
we made snr maps from the residual maps , searching for peaks above four .
this lower snr limit was picked because bws reduces the peak flux of a source .
next , the properties of these sources were measured using @xmath10 task , jmfit .
sources with snr greater than 4.5 were retained .
any such sources were added to the catalog for the appropriate resolution .
sad sources with snr @xmath22 5.5 were re - measured by hand using jmfit the same task used by sad to determine the source properties . at this low flux limit sad s automatic detection routine
does not always choose the optimal area for analysis .
those pixels within a defined radius ( 100 pixels ) were used to calculate the local noise for each source .
following om08 , resolved sources were classified on the basis of the best - gaussian - fit major axis .
if the lower limit for this parameter was greater than zero , the source is classed as resolved and the integrated flux was used for the total flux .
if lower limit for the major axis was equal to zero ( i.e. , unresolved ) then the peak flux was adopted as the total flux for the source .
there are several extended sources within the central field , as shown in fig .
[ kntr2 ] .
tvstat was used to measure the total flux density of extended sources .
the snrs of extended sources were determined as follows : imean was used to determine the brightest peak and its position was then examined in the noise map ; the ratio of these values was adopted as snr .
results were then collated following the prescription given in om08 .
an additional noise term of 3% was included for each source to account for calibration errors and uncertainties in the primary beam correction .
a sample of the radio catalog is given in table [ table2 ] .
the complete catalog is available electronically .
column 1 is the source number : source numbers below 10,000 relate to those sources found by sad while those above 10,000 were those found and investigated by hand . columns 2 and 3 contain the right ascension ( r.a . ) and the declination ( dec . ) in j2000 , with the positional error .
column 4 represents the snr using the ratio of the observed peak flux density and the local noise . in columns 5 we list the peak flux density and its error , in @xmath1jybeam@xmath2 .
columns 6 contain the total , primary - beam - corrected flux density and uncertainty , in @xmath1jy .
the best - fit deconvolved size ( in arcsec ) is given in column 7 - 9 .
using our criteria for resolved sources , if a resolved 2-dimensional gaussian size was the best fit then we give the resulting major and minor fwhm size ( in arcsec ) along with the position angle ( p.a . ) .
if this was not the case then in column 10 we report the major axis upper limit , as estimated by jmfit or sad .
if the source is resolved then we report 0.0 for the upper limit size .
as noted above , very extended sources , as shown in fig .
[ kntr2]-[kntr4 ] , had their positions , sizes , fluxes and errors measured interactively using imval and tvstat .
the final column reports the resolution of the map that was used for the measurements for that particular source .
cccccccccl 5 & 20 & 30 & 32 & 37 & 1.2 & 0.1 & 1.16 & 46 + 5 & 30 & 100 & 49 & 24 & 1.8 & 0.2 & 1.15 & 67 + 10 & 100 & 300 & 34 & 26 & 2.3 & 0.4 & 1.09 & 57 + 20 & 300 & 1000 & 48 & 11 & 3.4 & 0.6 & 1.57 & 81 + as a check on the radio angular size distribution in the goods - n field , we have repeated the source size analysis of om08 .
the focus of this analysis is to estimate the angular size of the faint radio population . as noted in column 1 of table [ table3 ] , we have analyzed the more sensitive inner region ( @xmath235 ) of the radio map for fainter sources while for the brighter sources we expanded our search outward ( to 20 ) to improve the statistics .
columns 2 and 3 give the flux range and column 4 lists the sources within that size range .
column 5 shows the number of upper limits used with the measured source size distribution in the kaplan - meier analysis to derive a robust mean size ( column 6 ) and its associated error ( column 7 ) .
the median size is reported in column 8 and the percentage of resolved sources used in this analysis is listed in column 9 .
the results in table [ table3 ] are consistent with both om08 and @xcite , adding to the evidence that the size distribution of the radio population does not continue to decrease below 100@xmath1jy and that the median size of such faint emitters is around 1 .
rrccc 28.0 & 40.0 & 34.0 & 185 & [email protected] + 40.0 & 55.0 & 46.7 & 187 & [email protected] + 55.0 & 90.0 & 71.1 & 274 & [email protected] + 90.0 & 150.0 & 113.1 & 150 & [email protected] + 150.0 & 300.0 & 199.8 & 107 & [email protected] + 300.0 & 1,500.0 & 557.3 & 65 & [email protected] + 1,500.0 & 60,000.0 & 5,260.5 & 17 & [email protected] + we have determined the differential normalized sources counts from our radio catalog using the code and analysis techniques described in detail by om08 .
the results are tabulated in table [ table4 ] .
the relative point - source sensitivity as a function of the distance from the field center for the three resolutions can be seen in figure 4 of om08 .
their figure 5 also shows the relative sensitivity as a function of source size . to summarize , multiple effects can make the source list incomplete , even at rather large snr .
the sensitivity for resolved sources is thus a function of both beam size and distance from the phase center .
these effects include ( i ) primary beam attenuation , ( ii ) reduction of peak flux due to bws , ( iii ) finite time - average smearing and ( iv ) resolution bias .
sources with integrated fluxes near the survey limit , if extended , will have peak sources below the limit . as seen in table [ table3 ] , the median source size is similar to that of the synthesized beam ( @xmath01.7 ) . to correct for the effects of this ,
we adopt the om08 size distribution , which is based on the properties of sources at higher snr .
as described by @xcite , and seen in their figure 2 , the brightest goods - n radio source within the primary beam of the vla is near 12:34:52 , @xmath2662:02:35 . with a primary - beam - corrected total flux of 263mjy and a peak flux of @xmath086mjybeam@xmath2 ,
it is bright sources like this one ( see fig . [ kntr3 ] upper right ) that can limit the dynamic range in an image .
this is because there is a limit to how well we can clean these bright sources a limit determined by how well we can characterize the bandpass , the pointing , the antenna - based gains , and many other parameters . as a result of the imperfect cleaning , residual flux spills into the image from the sidelobes of bright sources .
peak flux density , in jybeam@xmath2 , is therefore a key quantity when determining the suitability of a field for deep radio imaging , where the ` beam@xmath2 ' indicates a dependence on resolution .
the reduction by @xcite yielded an image with a dynamic range of @xmath27 , with @xmath28jy .
this was due in part to the fact that self - calibration was not applied to the @xmath9 data .
self - calibration and newly developed techniques allowed @xcite to create a more sensitive map , with @xmath29jy and a dynamic range of @xmath30 . our reduction , which includes a large quantity of new data ( in all vla configurations ) , reaches @xmath31jybeam@xmath2 and a dynamic range of @xmath32 .
the @xmath0140-hr vla observation by @xcite in 1046 + 59 , a field deliberately chosen to be free of strong radio sources has @xmath33jy , which is close to the theoretical noise limit , yet the dynamic range is only @xmath34 because of the paucity of bright radio emitters .
[ fig2 ] shows our differential source counts below 1mjy , compared to other deep , 1.4-ghz continuum surveys .
our source counts agree with those for the 1046 + 59 field where the same methodology was employed for cataloging and for determining log @xmath24log @xmath25 : use of multi - resolution maps to build a master source catalog , and the adoption of the same source size distribution . above 100@xmath1jy
our new survey agrees with the counts presented by @xcite .
below 100@xmath1jy , the counts in the new survey flatten out while those of @xcite began to turn down .
this is a well - known issue in deep radio surveys .
the difference can be related to a number of complications : 1 .
we believe that one of the main problems arises from the gaussian fit not always being the best approach to estimate source properties .
indeed , the intrinsic shape of the sources gets badly affected due to bandwidth and time delay smearing .
we have found evidence that flux densities are underestimated systematically after sources have been convolved to lower resolutions
flux densities are larger in the convolved maps .
also , a particular uncertainty comes from the assumption that in unresolved sources we consider the peak flux density as an estimate of the total flux density .
these problems clearly suggest the necessity for a more robust source extraction method in radio observations .
another , smaller uncertainties coming from the ` clean bias ' and ` flux boosting ' are not expected to be important due to the tight clean boxes we used , and the high 5-@xmath4 threshold in the source extraction , respectively .
a knowledge for the intrinsic radio source size distribution is essential when estimating the number of sources missed by our imaging approach .
@xcite use high - resolution merlin and a - configuration vla data to tackle this problem , but they miss a non negligible number of low - surface - brightness sources , as is indicated clearly by cross - matching the catalog with lower resolution observations using the wsrt .
we believe the combination of vla data in its a , b , c , and d configurations followed by a source extraction performed in convolved images is essential for detecting these low surface brightness sources ( @xmath354 ) .
a simple source extraction using sad in the original non - convolved image misses a considerable number of sources as we show in this work .
we find that our survey is in good agreement with om08 , so given by the similar treatment , we have adopted their source size distribution for number count estimations .
3 . finally , the source extraction based on a snr criterion highly depends on reliable estimates of noise as a function of position in the map and the efficiency for detection .
monte - carlo simulations for the source extraction method
( e.g. inserting randomly placed sources in the map and then doing a random search to extract them ) are typically used for these purposes .
nevertheless , given our detailed source extraction and our constant examination of residual maps , we believe that our number counts do not suffer significantly from incompleteness down to a threshold of 5@xmath4 .
this is why we prefer to perform theoretical corrections , as explained in om08 , rather than a simple monte - carlo approach as was in @xcite .
it is expected that the cosmic microwave background ( cmb ) temperature would increase above 2.7k at 1.4 ghz if the counts do not turn over at some flux density below where we have reached .
( e.g. * ? ? ?
the log @xmath24log @xmath25 results from both the om08 and the current work suggest that if the euclidean nature of the counts continue down to the natural confusion as an extrapolation of our log n - log s , this suggests that the cutoff must be fairly sharp below 1 @xmath1jy .
we have presented a deep , new , high - resolution radio image of the goods - n field and a catalog containing 1,230 secure radio sources in the central 40@xmath340 region .
the image reaches an r.m.s .
of @xmath03.9@xmath1jybeam@xmath2 in the best central region .
it is thus one of the deepest 1.4-ghz radio surveys ever undertaken , well matched to the excellent _ spitzer _ , _ hubble space telescope _ and _ chandra _ data in goods - n , and to the upcoming _ herschel _ , scuba-2 , and the large millimeter telescope imaging at 100850@xmath1 m and @xmath01.1 - 4 mm , respectively , providing an unbiased probe of star formation out to @xmath36 and a means to test the calibration of other star - formation rate indicators .
our analysis of the source size distribution and the counts provide further evidence that the faint 1.4-ghz emitters have a median angular size of @xmath01.2 , that log @xmath24log @xmath25 remains flat below 100@xmath1jy , and that the natural confusion limit at 1.4ghz must be near 1@xmath1jy a prediction that can be tested in the coming years using the expanded vla ( evla ) and e - merlin .
g.e.m . acknowledges financial support from nrao and travel support from the uk s science and technology facilities council via the uk astronomy technology centre at the royal observatory , edinburgh .
the national radio astronomy observatory is a facility of the national science foundation operated under cooperative agreement by associated universities , inc g.e.m .
also acknowledges financial support for this work , part of the space infrared telescope facility legacy science program , was provided by nasa through contract 1224666 issued by the jet propulsion laboratory , california institute of technology , under nasa contract 1407 . we would also like to thank the anonymous referee for useful suggestions to enhance the quality of the paper . ,
m. , dickinson , m. , ferguson , h. c. , ravindranath , s. , kretchmer , c. , moustakas , l. a. , madau , p. , fall , s. m. , gardner , j. p. , livio , m. , papovich , c. , renzini , a. , spinrad , h. , stern , d. , & riess , a. 2004 , , 600 , l103 , r. j. , greve , t. r. , smail , i. , dunlop , j. s. , roche , n. d. , scott , s. e. , page , m. j. , stevens , j. a. , almaini , o. , blain , a. w. , willott , c. j. , fox , m. j. , gilbank , d. g. , serjeant , s. , & hughes , d. h. 2002 , , 337 , 1 , g. , dickinson , m. , owen , f. , daddi , e. , chary , r. , bauer , f. , mobasher , b. , macdonald , e. , koekemoer , a. , & pope , a. 2008 , in astronomical society of the pacific conference series , vol .
381 , infrared diagnostics of galaxy evolution , ed . r .- r .
chary , h. i. teplitz , & k. sheth , 376 , t. w. b. , richards , a. m. s. , garrington , s. t. , wilkinson , p. n. , anderson , b. , richards , e. a. , axon , d. j. , fomalont , e. b. , kellermann , k. i. , partridge , r. b. , & windhorst , r. a. 2005 , , 358 , 1159 , r. e. , blacker , b. , dickinson , m. , dixon , w. v. d. , ferguson , h. c. , fruchter , a. s. , giavalisco , m. , gilliland , r. l. , heyer , i. , katsanis , r. , levay , z. , lucas , r. a. , mcelroy , d. b. , petro , l. , postman , m. , adorf , h .-
, & hook , r. 1996 , , 112 , 1335 | we describe deep , new , wide - field radio continuum observations of the great observatories origins deep survey north ( goods - n ) field .
the resulting map has a synthesized beamsize of @xmath01.7 and an r.m.s .
noise level of @xmath03.9@xmath1jybeam@xmath2 near its center and @xmath08@xmath1jybeam@xmath2 at 15 from phase center .
we have cataloged 1,230 discrete radio emitters , within a 40@xmath340region , above a 5-@xmath4 detection threshold of @xmath020@xmath1jy at the field center .
new techniques , pioneered by @xcite , have enabled us to achieve a dynamic range of 6800:1 in a field that has significantly strong confusing sources .
we compare the 1.4-ghz ( 20-cm ) source counts with those from other published radio surveys .
our differential counts are nearly euclidean below 100@xmath1jy with a median source diameter of @xmath01.2 .
this adds to the evidence presented by @xcite that the natural confusion limit may lie near 1@xmath1jy .
if the euclidean slope of the counts continues down to the natural confusion limit as an extrapolation of our log n - log s , this indicates that the cutoff must be fairly sharp below 1 @xmath1jy else the cosmic microwave background temperature would increase above 2.7k at 1.4 ghz . |
the problem of collective classification is a widely studied one in the context of graph mining and social networking applications . in this problem
, we have a network containing nodes and edges , which can be represented as a graph .
nodes in this network may be labeled , but it is not necessary that all nodes have a label . typically , such labels may represent some properties of interest in the underlying network .
this is a setting that appears in several situations in practice .
some examples of such labeled networks in real scenarios are listed below : * in a bibliographic network , nodes correspond to authors , and the edges between them correspond to co - authorship links .
the labels in the bibliographic network may correspond to subject areas that experts are interested in .
it is desirable to use this information in order to classify other nodes in the network . * in a biological network , the nodes correspond to the proteins .
the edges may represent the possibility that the proteins may interact .
the labels may correspond to properties of proteins @xcite . * in a movie - actor network , the nodes correspond to the actors .
the edges correspond to the co - actor relationship between the different actors .
the labels correspond to the pre - dominant genre of the movie of the actor . * in a patent network , the nodes correspond to patent assignees .
the edges model the citations between the respective patents .
the labels correspond to the class categories . in such networks ,
only a small fraction of the nodes may be labeled , and these labels may be used in order to determine the labels of other nodes in the network .
this problem is popularly referred to as _ collective classification _ or _
label propagation _
@xcite , and a wide variety of methods have been proposed for this problem .
the problem of data uncertainty has been widely studied in the database literature @xcite , and also presents numerous challenges in the context of network data @xcite .
in many real networks , the links are uncertain in nature , and are derived with the use of a probabilistic process . in such cases , a probability value may be associated with each edge .
some examples are as follows : * in biological networks , the links are derived from probabilistic processes . in such cases , the edges have uncertainty associated with them
. nevertheless , such probabilistic networks are valuable , since the probability information on the links provides important information for the mining process . *
the links in many military networks are constantly changing and may be uncertain in nature .
in such cases , the analysis needs to be performed with imperfect knowledge about the network .
* networks in which some links have large failure probabilities are uncertain in nature .
* many human interaction networks can be created from real interaction processes , and such links are often uncertain in networks . thus , such networks can be represented as probabilistic networks , in which we have probabilities associated with the existence of links .
such probabilities can be very useful for improving the effectiveness of problems such as collective classification .
furthermore , these networks may also have properties associated with nodes , that are denoted by labels .
recent years have seen the emergence of numerous methods for uncertain graph management @xcite and mining @xcite , in which uncertainty is used directly as a first - class citizen . however , none of these methods address the problem of collective graph classification .
one possibility is to use _ sampling _ of possible worlds on the edges in order to generate different instantiations of the underlying network .
the collective classification problem can be solved on these different instantiations , and voting can be used in order to report the final class label .
the major disadvantage with this approach is that the sampling process could result in a sparse or disconnected network which is not suited to the collective classification problem . in such cases ,
good class labels can not be easily produced with a modest number of samples . in this paper , we investigate the problem of collective classification in uncertain networks with a more direct use of the uncertainty information in the network .
we design two algorithms for collective classification .
the first algorithm uses a probabilistic approach , which explicitly accounts for the uncertainty in the links in the classification .
the second algorithm works with the assumption that most of the information in the network is encoded in high - probability links , and low - probability links sometimes even degrade the quality .
therefore , the algorithm uses the links with high probability in earlier iterations , and successively relaxes the constraints on the quality of the underlying links .
the idea is that a greater caution in early phases of the algorithm ensures convergence to a better optimum .
the contributions we make in this paper can be summarized as follows .
* we introduce the problem of collective classification in uncertain graphs , where uncertainty is associated with the edges of the graph , and provide a formal definition for this problem . *
we introduce two algorithms based on iterative probabilistic labeling that incorporate the uncertainty of edges in their operation .
these algorithms are based on a bayes formulation , which enables them to capture correlations across different classes , leading to improved accuracy . *
we perform an extensive experimental evaluation , using two real datasets from diverse domains .
we evaluate our techniques using a multitude of different conditions , and input data characteristics .
the results demonstrate the effectiveness of the proposed techniques and serve as guidelines for the practitioners in the field .
this paper is organized as follows . in section
[ sec : relwork ] we survey prior studies on collective classification and on mining uncertain networks . in section
[ sec : cc ] , we formally define the problem of collective classification in uncertain networks . in section [ sec : proposal ] , we present our model and two algorithms for collective classification .
we discuss the space and time complexity of our proposal in section [ sec : complexity ] , and we present the results of our experimental evaluation in section [ sec : results ] .
finally , we discuss the conclusions in section [ sec : conclusions ] .
the problem of node classification has been studied in the graph mining literature , and especially relational data in the context of _ label or belief propagation _ @xcite .
such propagation techniques are also used as a tool for semi - supervised learning with both labeled and unlabeled examples @xcite .
collective classification @xcite refers to semi - supervised learning methods that exploit the network structure and node class labels to improve the classification accuracy .
these techniques are mostly based on the assumption of homophily in social networks @xcite : neighboring nodes tend to belong to the same class .
a technique has been proposed in @xcite , which uses link - based similarity for node - classification in directed graphs .
recently , collective classification methods have also been used in the context of blogs @xcite . in @xcite ,
bilgic et al .
discuss the problem of overcoming the propagation of erroneous labels by asking the user for more labels . a method for performing collective classification of email speech acts has been proposed by carvalho et al . in @xcite , exploiting the sequential correlation of emails . in @xcite ,
integrate the classification of nodes in heterogeneous networks with ranking .
methods for leveraging label consistency for collective classification have been proposed in @xcite .
recently , the database and data mining community has investigated the problem of uncertain data mining widely @xcite .
a comprehensive review of the proposed models and algorithms can be found in @xcite .
several database systems supporting uncertain data have been proposed , such as conquer @xcite , trio @xcite , mistiq @xcite , maymbs @xcite and orion @xcite .
the `` possible worlds '' model , introduced by abiteboul et al .
@xcite , formalizes uncertainty by defining the space of the possible instantiations of the database .
instantiations must be consistent with the semantics of the data .
for example , in a graph database representing moving object trajectories there may be be different configurations of the edges where each node represents a region in the space .
however , an edge can not connect a pair of nodes that represent a pair of non - neighboring regions .
the main advantage of the `` possible worlds '' model is that the formulations of the queries originally designed to cope with certain data can be directly applied on each possible instantiation .
many different alternatives have then been propose to aggregate the results across the different instantiations . despite its attractiveness , the number of possible worlds explodes very quickly and even their enumeration becomes intractable problem . to overcome these issues , simplifying assumptions
have been introduced to leverage its simplicity : the tuple- and the attribute - uncertainty models @xcite . in the attribute - uncertainty model , the uncertain tuple
is represented by means of multiple samples drawn from its probability density function ( pdf ) .
in contrast , in the tuple - uncertainty model the value of the tuple is fixed but the tuple itself may not exist .
similar simplifications have been considered for graph databases where nodes may or may not exist ( node - uncertainty ) and edges are associated with an existence probability ( edge - uncertainty ) .
the underlying uncertainty model can then be used to generate graph instances , eventually considering additional generation rules to consider correlations across different nodes and edges . in this study
we combine a bayes approach and the edge - uncertainty model .
the problem of uncertain graph mining has also been investigated extensively .
the most common problems studied in uncertain graph management are those of nearest neighbor query processing @xcite , reachability computation @xcite and subgraph search @xcite . in the context of uncertain graph _ mining
_ , the problems commonly studied are frequent subgraph mining @xcite , reliable subgraph mining @xcite , and clustering @xcite .
recently , the problem of graph classification has also been studied for the uncertain scenario @xcite , though these methods are designed for classification of many small graphs , in which labels are attached to the entire graph rather than a node in the graph .
typical social and web - based scenarios use a different model of collective classification , in which the labels are attached to nodes in a single large graph . in this work ,
we study the problem of _ collective classification _ in the context of uncertain networks , where the underlying links are uncertain .
uncertainty impacts negatively on the classification accuracy .
first , links may connect sub - networks of very different density , causing the propagation of erroneous labels .
second , the farthest distance between two nodes tends to be smaller in very noisy networks , because of the presence of a larger number of uncertain edges , which include both true and spurious edges .
this reduces the effectiveness of iterative models because of the faster propagation of errors .
some of our techniques , which drop uncertain links at earlier stages of the algorithm , are designed to ameliorate these effects .
in this section , we formalize the problem of collective classification after introducing some definitions .
an uncertain network is composed of nodes whose connections may exist with some probability .
[ def : unet ] an uncertain network is denoted by @xmath0 , with node set @xmath1 , edge set @xmath2 and probability set @xmath3 . each edge @xmath4 is associated with a probability value @xmath5 .
this is the probability that edge @xmath6 exists in the network .
we assume that the network is undirected , though the method can easily be extended to the directed scenario .
we can assume that the @xmath7 matrix @xmath3 has entries which are denoted by @xmath8 and @xmath9 .
a node @xmath10 can be associated with a label , representing its membership in a class .
for ease in notation , we assume that node labels are integers .
[ def : label ] given a set of labels @xmath11 drawn from a set of integers @xmath12 , we denote the label of node @xmath13 by @xmath14 .
if a node @xmath13 is unlabeled , the special label @xmath15 is used .
we can now introduce the definition of the collective classification problem on uncertain graphs .
[ problem : cc ] given an uncertain network @xmath0 and the subset of labeled nodes @xmath16 , predict the labels of nodes in @xmath17 .
figure [ fig : net1 ] shows an example of an uncertain network .
nodes @xmath18 , @xmath19 , and @xmath20 are labeled _ white _ , and nodes @xmath21 , @xmath22 , and @xmath23 are labeled _ black_. the label of nodes @xmath24 and @xmath25 is unknown .
the aim of collective classification is to assign labels to nodes @xmath24 and @xmath25 .
in this section , we first present the algorithm for iterative probabilistic labeling .
a bayes approach is used in order to perform the iterative probabilistic labeling .
this method models the probabilities of the nodes belonging to different classes on the basis of the adjacency behavior of the nodes .
the bayes approach can directly incorporate the edge uncertainty probabilities into the estimation process .
we continue with a second algorithm that builds upon the first one , and is based on iterative edge augmentation .
finally , we describe a variation of the second algorithm that is a linear combination of two classifiers .
the overall approach for the labeling process uses a bayesian model for the labeling . in the rest of the paper
, we refer to this algorithm as _
ubayes_. given that we have an unlabeled node @xmath26 , which is adjacent to @xmath27 other nodes denoted by @xmath28 , how do we determine the label of the node @xmath26 ?
it should be noted that the concept of adjacency is also uncertain , because the edges are associated with probabilities of existence .
this is particularly true , when the edge probabilities are relatively small , since the individual network instantiations are likely to be much sparser and different than the probabilistic descriptions .
furthermore , for each edge @xmath6 we need to estimate the probability of the node @xmath29 having a particular label value , given the current value of the label at node @xmath13 .
this is done with the use of training data containing the labels and edges in the network .
these labels and edges can be used to construct a bayesian model of how the labels on the nodes and edges relate to one another .
the algorithm uses an iterative approach , which successively labels more nodes in different iterations .
this is the set @xmath30 of nodes whose labels will not be changed any further by the algorithm .
initially , the algorithm starts off by setting @xmath30 to the initial set of ( already ) labeled nodes @xmath31 .
the set in @xmath30 is expanded to @xmath32 in each iteration , where @xmath33 is the set of nodes not yet labeled that are adjacent to the labeled nodes in @xmath30 . if @xmath33 is empty , either all nodes have been labeled or there is a disconnected component of the network whose nodes are not in @xmath31 .
the expanded set of labeled nodes are added to the set of training nodes in order to compute the propagation probabilities on other edges .
thus , the overall algorithm iteratively performs the following steps : * estimating the bayesian probabilities of propagation from the current set of edges . *
computing the probabilities of the labels of the nodes in @xmath34 . *
expanding the set of the nodes in @xmath30 , by adding the set of nodes from @xmath33 , whose labels have the highest probability for a particular class . these steps are repeated until no more nodes reachable from the set @xmath30 remain to be labeled .
we then label all the remaining nodes in a single step , and terminate .
the overall procedure for performing the analysis is illustrated in algorithm [ iterative ] .
it now remains to discuss how the individual steps in algorithm [ iterative ] are performed .
+ * algorithm * _ ubayes_(graph : @xmath35 + uncertainty prob .
: @xmath3 , initial labeling : @xmath31 ) ; + + * begin * + = @xmath36 ; + ( not termination ) * do * + + = compute edge propagation probabilities ; + compute node label probabilities in @xmath34 ; + expand @xmath30 with @xmath33 nodes ; + + * end * the two most important steps are the computation of the edge - propagation probabilities and the expansion of the node labels with the use of the bayes approach . for a given edge @xmath6 we estimate @xmath37 .
this is estimated from the data in _ each iteration _ by examining the labels of nodes which have already been decided .
therefore , the training process is successively refined in each iteration .
therefore , the value of @xmath38 can be estimated by examining those edges for which one end point contains a label of @xmath39 . among these edges , we compute the fraction for which the other end point contains a label of @xmath40 .
for example , in the network shown in figure [ fig : net1 ] the probability @xmath41 is estimated as @xmath42 .
the label of node @xmath25 is unknown , and it is not considered in the calculation . note that this is simply equal to the probability that both end points of an edge are _ black _ , if one of them is _
black_. therefore , one can compute the uncertainty weighted conditional probabilities for this in the training process of each iteration .
this provides an estimate for the conditional probability .
we note that in some cases , the number of nodes with a label of either @xmath40 or @xmath39 may be too small for a robust estimation .
the following smoothing techniques are useful in reducing the effect of ill - conditioned probabilities : * we always add a small value @xmath43 to each probability .
this is similar to laplacian smoothing and prevents any probability value from being zero , which would cause problems in a multiplicative bayes model . * in some cases
, the estimation may not be possible when labels do not exist for either nodes @xmath40 or @xmath39 . in those cases
, we set the probabilities to their prior values .
the prior is defined as the value of @xmath44 , and is equal to the fraction of currently labeled nodes with label of @xmath40 .
the prior therefore defines the default behavior in cases where the adjacency information can not be reasonably used in order to obtain a better posteriori estimation . for an unlabeled node @xmath26 , whose neighbors @xmath45 have labels @xmath28 , we estimate its ( unnormalized ) probability by using the naive bayes rule over all the adjacent labeled neighbors .
this is therefore computed as follows : @xmath46 note that the above model incorporates the uncertainty probabilities directly within the product term of the equation .
we can perform the estimation for each of the different classes separately .
if desired , one can normalize the probability values to sum to one .
however , such a normalization is not necessary in our case , since the only purpose of the computation is to determine the highest probability value in order to assign labels .
the approach mentioned above is not very effective when a large fraction of the edges are noisy .
in particular , if many edges have a low probability , this can have a significant impact on the classification process .
figure [ fig : net2 ] shows an example .
nodes @xmath18 , @xmath19 , are labeled _ white _ , and nodes @xmath20 , @xmath24 , @xmath25 , @xmath22 , @xmath23 and @xmath47 are labeled _
black_. the label of node @xmath21 is unknown and must be assigned by the algorithm .
we observe that ignoring the edges whose existence probability is lower than @xmath48 is beneficial for the correct classification of node @xmath21 .
therefore , we use an iterative augmentation process in order to reduce the impact of such edges , by instead favoring the positive impact of high quality edges in the collective classification process .
the idea is to _ activate _ only a subset of the edges for use on the modeling process . in other words , edges which are not activated
are not used in the modeling .
we call this algorithm _ ubayes+_. we adopt a model inspired by automatic parameter selection in machine learning . note that , analogous to parameter selection , the choice of a particular subset of high quality links , corresponds to a configuration of the network , and we would like to determine an optimal configuration for our approach . in order to do this , we split the set of labeled nodes @xmath31 into two subsets : a _ training set _ denoted by @xmath49 and a _ hold out _ set denoted by @xmath50 .
the ratio of the @xmath31 nodes that are assigned to the training set @xmath49 is denoted by @xmath51 , a user - defined parameter .
the purpose of the hold out set is to aid optimal configuration selection by checking the precise value of the parameters at which the training model provides optimal accuracy over the set of nodes in @xmath50 .
we use labels of nodes in @xmath49 for the learning process , while using labels of nodes in @xmath50 as for the evaluation of accuracy at a particular configuration of the network .
( note that a label is never used for both the training and the hold out set , in order to avoid overfitting . )
the idea is to pick the ratio of active edges in such a way so as to optimize the accuracy on the hold out set .
this ensures that an optimal fraction of the high quality edges are used for the labeling process .
we start off considering a small fraction of the high probability edges , iteratively expanding the subset of active edges by enabling some of the inactive edges with the highest probabilities .
the ratio of active edges is denoted by the parameter @xmath52 .
ideally , we want to activate only the edges that contribute positively to the classification of unlabeled nodes . given a configuration of active edges , we measure their goodness as the estimated accuracy on labels of nodes in @xmath50 .
the value of @xmath52 that leads to the highest accuracy , denoted by @xmath53 , is used as the ratio of edges with the highest probability to activate on the uncertain network @xmath35 .
the resulting network is then used as input for the iterative probabilistic labeling algorithm ( _ ubayes _ ) . despite optimizing accuracy by selecting the best ratio of edges to be considered ,
the basic model described above is not very efficient , because it requires multiple evaluations of the iterative probabilistic labeling algorithm . in particular
, it requires us to vary the parameter @xmath52 and evaluate accuracy , in order to determine @xmath53 .
a more efficient technique for identifying @xmath53 can be obtained by evaluating the accuracy for different values of @xmath52 on a sample of the uncertain network @xmath35 ( rather than the full network ) as follows .
we generate a new uncertain network @xmath54 by sampling @xmath55 nodes from @xmath35 uniformly at random , and retaining the edges from @xmath2 and probabilities from @xmath3 referring to these sampled nodes .
@xmath56 is a user - defined parameter that controls the ratio of nodes sampled from @xmath35 and it implies the size of the sampled uncertain network @xmath57 .
the initial set of labeled nodes in the sampled uncertain network @xmath57 is @xmath58 .
we split the set of nodes in @xmath59 into two random subsets , @xmath60 and @xmath61 , respectively .
the number of nodes in @xmath60 is @xmath62 .
we start off considering @xmath63 edges with the highest probabilities , expanding iteratively the subset of active edges at each iteration by increasing @xmath52 .
the goodness of parameter @xmath52 is estimated as the accuracy of node labels in @xmath61 .
let @xmath53 be the value of @xmath52 leading to the highest accuracy .
we activate @xmath64 edges with highest probability in @xmath35 .
the resulting network is then used as input for the iterative probabilistic labeling ( algorithm [ iterative ] ) .
the overall algorithm is illustrated in algorithm [ iterative2 ] .
* algorithm * _ ubayes+_(graph : @xmath35 + uncertainty prob . : @xmath3 , initial labeling : @xmath31 , + sampled nodes ratio : @xmath56 , train nodes ratio : @xmath51 ) ; + + * begin * + = @xmath65 = random sample of @xmath55 nodes from @xmath1 ; + = @xmath66 = edges @xmath67 in @xmath2 with @xmath68 ; + = @xmath69 ; + = @xmath70 edges in @xmath66 with + greatest existence probability ; + ( @xmath71 ) * do * + + = construct graph @xmath72 ; + = @xmath73 ; + = test accuracy using nodes in @xmath50 ; + = expand edges in @xmath74 with top edges in @xmath66 ; + + = construct graph @xmath75 with best + = configuration ( corresponding to @xmath53 ) ; + = @xmath76 ; + * end * we note that the frequencies used to estimate conditional and prior probabilities across the different configurations in algorithm [ iterative2 ] can be efficiently maintained in an incremental fashion . in this section
we propose a third algorithm , _
ubayes+rn_. it uses an ensemble methodology in order to further improve robustness in scenarios , where some deterministic classifiers can provide good results over _ some _ subsets of nodes , but not over all the nodes .
_ ubayes+rn _ is the linear combination of two classifiers : the _ ubayes+ _ algorithm and the relational neighbor ( rn ) classifier @xcite .
the rn classifier is defined as follows : @xmath77 where @xmath78 is the probability and @xmath79 .
ubayes+ _ and _ rn _ algorithms are combined as follows : @xmath80 where @xmath43 controls the influence of the rn classifier during the collective classification process .
when @xmath81 then _ ubayes+rn _ degenerates to _ ubayes+ _ , while when @xmath82 the two classifiers are weighted equally .
note that this is a simple linear combination .
we used this combination , since it sometimes provides greater robustness in the classification process .
in this section , we discuss the complexity of the proposed algorithms . we start with _ ubayes _ , which for the computation of the initial statistics requires @xmath83 ( label priors and conditional label probabilities ) . assuming that the cardinality of the set of immediate unlabeled neighbors of nodes in @xmath30 ( remember that @xmath30 represents the set of currently labeled nodes ) is at most @xmath84 , and that the number of neighbors for a particular node is at most @xmath85 , each iteration can be decomposed as follows .
the computation of new unlabeled nodes requires @xmath86 .
the computation of edge propagation probabilities requires @xmath86 .
the computation of node label probabilities requires @xmath86 . summing up
, each iteration requires @xmath86 . assuming that all unlabeled nodes will be labeled in @xmath87 iterations , the algorithm cost is @xmath88 , where @xmath89 .
space complexity is @xmath90 . for algorithm _
ubayes+ _ , the computation of the uncertain network sample @xmath91 requires @xmath92 .
active edges are maintained using a priority list , whose initialization requires @xmath93 .
each iteration of the iterative automatic parameter selection procedure can be decomposed as follows .
algorithm [ iterative ] ( used by _
ubayes+ _ ) requires @xmath94 .
testing the classification accuracy requires @xmath95 .
expanding the set of active edges requires @xmath96 . summing up
, each iteration requires : @xmath97 finally , the last call to algorithm [ iterative ] requires @xmath88 .
assuming that the parameter selection procedure terminates after @xmath98 iterations , the algorithm cost is @xmath99 .
simplifying , the cost is @xmath100 .
the space complexity is @xmath90 .
note that algorithms _
ubayes+rn _ and _ ubayes+ _ have the same space and time complexity .
in this section , we evaluate the proposed techniques under different settings , in terms of both accuracy and performance .
we implemented all techniques in c++ using the standard template library ( stl ) and boost libraries , and ran the experiments on a linux machine equipped with an intel xeon 2.40ghz processor and 16 gb of ram .
the reported times do not include the initial loading time , which was constant over all methods .
the results were obtained from @xmath21 independent runs .
for all experiments we report the averages and @xmath101 confidence intervals . in our experiments
, we used two data sets for which edge probabilities can be estimated , as described below .
_ dblp : _ the _ dblp data set _ @xcite is the most comprehensive citation network of curated records of scientific publications in computer science . in our experiments
, we consider the subset of publications from @xmath102 to @xmath103 .
the data set consists of @xmath104 nodes and @xmath105 edges .
nodes represent authors and edges represent co - authorship relations .
the edge probability is an estimate of the probability that two authors co - authored a paper in a year selected randomly during their period of activity .
for example , if a pair of authors published papers in ten different years and they both published papers for twenty years , then their edge probability is @xmath48 .
( we consider the union of their periods of activity . )
we used 14 class labels , that represent different research fields in computer science .
the corresponding labels and their frequencies are illustrated in table [ table : labels_dblp ] .
the labels were generated by using a set of top conferences and journals in these areas , and the most frequent label in the author s publications is used as the author s label . in our data set , @xmath106 of the nodes are labeled .
the rest were not labeled , because the corresponding authors did not have publications in the relevant conferences and journals .
+ .node labels and label priors of the _ dblp _ dataset . [ cols="^,^,^",options="header " , ] finally , we report the accuracy of the _ ubayes+rn _ algorithm when varying the parameter @xmath43 between @xmath15 and @xmath18 . recall that @xmath43 controls the influence of the _ rn _ classifier on the overall classification process . in our experiments with both data sets , the accuracy of _
ubayes+rn _ was always slightly better than _ ubayes+ _ , but never more than @xmath107 .
we observed nearly no difference among the different @xmath43 configurations . in the interest of space , we omit the detailed results .
we next provide some real examples of labeling results obtained with _
ubayes+ _ on the patent dataset .
the _ atari inc . _ " and _ sega enterprises , ltd _ " companies , which belong to the hall of fame of the video game industry , were not assigned to any category .
our algorithm correctly classified them as computers & communications " .
similarly , the companies _ north american biologicals , inc _ " and _ bio - chem valve , inc _ " were correctly labeled as drugs & medical " , since they are both involved in drug development and pharmaceutical research .
interestingly , _ starbucks corporation _ " was labeled as chemical " . taking
a close look at their patents , it turns out that a large fraction of them describe techniques for enhancing flavors and aromas that involve chemical procedures .
evidently , having labels for all the nodes in the graph allows for improved query answering and data analysis in general . in this section ,
we assess running time efficiency on a variety of settings using both real and perturbed data sets .
figures [ fig : time_gamma](a ) and [ fig : time_gamma](b ) show the cpu time required by the algorithms when varying the ratio of noisy edges , for the _ dblp _ and _ patent _ data sets , respectively .
note that _ sampling _ has the same time performance as _
ubayes_. the _
ubayes+ _ algorithm is nearly three times slower than _
ubayes_. this is due to the automatic parameter tuning approach employed by the _
ubayes+ _ algorithm we observe that the performance of _ wvrn-20 _ almost always _ considerably _ worse than both _
ubayes _ and _ ubayes+_. the same observation is true when we vary the standard deviation of the probability of the noisy edges ( see figures [ fig : time_theta](a ) and [ fig : time_theta](b ) ) .
note that the inference in the _ wvrn _ algorithm is based on labeling relaxation , whose complexity is proportional to the size of the network and remains constant across iterations . on the contrary ,
the iterative labeling that _ ubayes _ and _ ubayes+ _ use for their inference model becomes faster with each successive iteration , since it needs to visit a smaller part of the network . as the results show
, the standard deviation does not affect the time performance of the algorithms .
these experiments demonstrate that the two proposed algorithms effectively combine low running times with high accuracy and robustness levels . in the final set of experiments
, we evaluated the accuracy of all algorithms as a function of the time required for algorithmic execution by the baselines . since the baselines tradeoff between running time and accuracy , it is natural to include the running time in the comparison process . in this case , we removed the constraint that _ wvrn _ and _ sampling _ end their processing after a fixed amount of time or a specific number of iterations , and examined how their accuracy changes when the number of iterations ( and consequently , processing time ) increases . for reference , we also include the _ ubayes _ and _ ubayes+ _ algorithms , which execute in a fixed amount of time .
the results for the _ dblp _ and _ patent _ data sets are depicted in figures [ fig : dblp_timeacc ] and [ fig : patents_timeacc ] respectively .
the graphs show that the accuracy of _ wvrn _ and _ sampling _ is slightly increasing with time , but reaches an almost stable state after the first @xmath108 iterations .
( in our experiments , we stopped _ wvrn _
after @xmath109 iterations in the _ dblp _ data set and @xmath110 iterations in the _ patent _ data set , and the _ sampling _ algorithm after @xmath111 iterations in the _ dblp _ data set and @xmath109 iterations in the _ patent _ dataset ) .
nevertheless , the
_ and _ ubayes+ _ algorithms achieve significantly better results in a much lower running time .
uncertain graphs are becoming increasingly popular in a wide variety of data domains .
this is due to the statistical methods used to infer many networks , such as protein interaction networks and other link - prediction based methods .
consequently , the problem of collective classification has become particularly relevant for determining node properties in such networks . in this paper
, we formulate the collective classification problem for uncertain graphs , and describe effective and efficient solutions for this problem . to this effect
, we describe an iterative probabilistic labeling method , based on the bayes model , that treats uncertainty on the edges of the graph as first class citizens . in the proposed approach ,
the uncertainty probabilities of the links are used directly in the labeling process .
furthermore , the methodology we describe allows for automatic parameter selection .
we have performed an experimental evaluation of the proposed approach using diverse , real - world datasets .
the results show significant advantages of using such an approach for the classification process over more conventional methods , which do not directly use uncertainty probabilities . *
acknowledgments * + part of this work was supported by the fp7 eu ip project kap ( grant agreement no .
work of the second author was sponsored by the army research laboratory under cooperative agreement number w911nf-09 - 2 - 0053 .
1000010000 = 1000 a. fuxman , e. fazli , and r. j. miller .
conquer : efficient management of inconsistent databases . in _ proceedings of the 2005 acm
sigmod international conference on management of data _ , pages 155166 .
acm , 2005 .
p. agrawal , o. benjelloun , a. d. sarma , c. hayworth , s. u. nabar , t. sugihara , and j. widom .
trio : a system for data , uncertainty , and lineage . in
_ international conference on very large data bases ( vldb ) _ , pages 11511154 , 2006 .
s. singh , c. mayfield , s. mittal , s. prabhakar , s. e. hambrusch , and r. shah .
orion 2.0 : native support for uncertain data . in _
acm sigmod international conference on management of data _ , pages 12391242 , 2008 .
x. zhu , z. ghahramani , j. lafferty , semi - supervised learning using gaussian fields and harmonic functions , , 2003 . z. zou , h. gao , j. li .
discovering frequent subgraphs over uncertain graph databases under probabilistic semantics . _
acm sigkdd conference on knowledge discovery and data mining ( kdd ) _ , 2010 .
j. ren , s. d. lee , x. chen , b. kao , r. cheng , and d. cheung .
naive bayes classification of uncertain data . in _ data mining , 2009 .
ninth ieee international conference on _ , pages 944949 .
ieee , 2009 . | in many real applications that use and analyze networked data , the links in the network graph may be erroneous , or derived from probabilistic techniques . in such cases , the node classification problem can be challenging , since the unreliability of the links may affect the final results of the classification process .
if the information about link reliability is not used explicitly , the classification accuracy in the underlying network may be affected adversely . in this paper , we focus on situations that require the analysis of the uncertainty that is present in the graph structure .
we study the novel problem of node classification in uncertain graphs , by treating uncertainty as a first - class citizen .
we propose two techniques based on a bayes model and automatic parameter selection , and show that the incorporation of uncertainty in the classification process as a first - class citizen is beneficial .
we experimentally evaluate the proposed approach using different real data sets , and study the behavior of the algorithms under different conditions .
the results demonstrate the effectiveness and efficiency of our approach .
= 1 |
many neural network models@xcite@xcite , in addition to their potential applications to computation , robotics and artificial intelligence , constitute intriguing dynamical systems in their own right , showing unusual manifestations of the statistical mechanics phenomena of order , disorder and frustration .
the connection between neural networks and statistical mechanics became especially clear with the introduction of the hopfield @xcite@xcite model , which furnishes a model of associative memory , or the recall of a memorized pattern from an incomplete stimulus .
this model has a well - defined energy function and is closely related to the sherrington - kirkpatrick spin glass model @xcite@xcite . in this paper
we consider a hopfield - like network of @xmath0 bistable elements , the bistable gradient network or bgn , previously introduced in @xcite .
a closely related model was also discussed in @xcite and suggested as a model for the so - called `` bistability of perception '' in the interpretation of ambiguous visual stimuli @xcite .
the network s dynamics consists of a continuous gradient descent described by the coupled differential equations @xmath1 where @xmath2 are continuous - valued real state variables associated with the @xmath3 nodes of the network and the hamiltonian or energy function is given by @xmath4 the quantities @xmath5 are a symmetric matrix of coupling strengths , and the quantities @xmath6 are bias terms or external inputs to the network . for the remainder of this paper
we will set all @xmath7 unless otherwise stated ; we include them here only for the sake of generality .
@xmath8 is a control parameter determining the strength of the internode couplings relative to the local terms .
the variables @xmath9 can be viewed as components of an @xmath3-dimensional state vector @xmath10 .
we define a normalized inner product between two state vectors @xmath11 and @xmath12 by @xmath13 .
the first term in the hamiltonian represents a local double - well potential for each node , making each node individually bistable .
this local potential constitutes the main difference between the bgn and the hopfield model .
the classical hopfield network ( hn ) which we consider by way of comparison is described by the hamiltonian @xmath14 where the @xmath9 are now _ discrete _ state variables restricted to the values @xmath15 .
although continuous versions of the hn have also been studied , these generally lack the bistability property , and their behaviour is essentially similar to that of the discrete version @xcite .
the variables @xmath9 can be thought of as the outputs of proccessing units or neurons .
their dynamical equations can be written as @xmath16 where @xmath17 is the input to the neuron from internal and external connections .
by analogy with ising spin systems , we also refer to @xmath18 as a magnetic field .
the steady - state output for a given input is a solution of the fixed - point equation @xmath19 when @xmath20 , there are stable fixed points at @xmath21 and an unstable fixed point at @xmath22 an applied field shifts the positions of the fixed points .
a saddle - node bifurcation occurs when @xmath23 so that for larger values of the field there is only one equilibrium , aligned with the field ( @xmath24 and @xmath25 have the same sign ) .
@xmath24 is in principle unbounded ; the output does not truly saturate when the input is large .
the double - valuedness and the lack of saturation are the principal differences between the input - output relation for the bgn and that of the hopfield model , including its continuous versions .
numerous experimental studies have been made on intriguing chemical analogs of the bgn ( see , e.g. @xcite , @xcite ) .
these studies involved networks of bistable chemical reactors coupled either electrically or through mass exchange .
previous computational work on small bgn s @xcite suggested that under some conditions the network might permit the storage of a larger number of patterns than in a hn of the same size , without any modification of the basic hebb learning rule .
it was noted , however , that the stability of a particular attractor can depend on the control parameter @xmath8 .
some dependence of pattern stability on the coupling strength had also been noted in the experiments on the mass - coupled chemical networks @xcite . in this paper
we focus on the behaviour of the network in the case where the number of nodes is large and the number of memorized patterns is small . using both analytical techniques and numerical simulations fell below a convergence criterion which for most examples was taken as 0.001 .
our hn simulations , which were used for comparisons , were performed using asynchronous updating in random order .
unless otherwise indicated , all simulations were done on an @xmath26 network with @xmath27 stored memory patterns . ] , we examine the retrieval of stored patterns and classify the attractors that occur .
we find that there are three types of attractors .
in addition to _ memory _ or _ retrieval states _ , there are spurious attractors in which no pattern is fully recognized .
these include the _ mixture _ or _ spin glass _ _ states _ familiar from hn studies , as well as an additional category specific to the bgn which we refer to as _
uncondensed states_. we examine how the attractors and their basins of attraction change as the control parameter @xmath8 is changed . throughout the paper ,
we compare our model to the zero - temperature or deterministic discrete hopfield model .
it is hoped that these results can illuminate some of the novel behavior of the bgn and clarify its relation to the hn .
the behavior of the bgn under higher memory loading and the question of its maximum storage capacity will be addressed elsewhere .
as in previous work on hopfield networks@xcite@xcite@xcite , we define the task of associative memory as follows .
we are given a set of @xmath28 distinct @xmath3-dimensional vectors or _ memory patterns _ @xmath29 @xmath30 which are to be recognized by the network .
the patterns should correspond to attractors of the network s dynamics .
we will refer to these attractors as _ retrieval states .
_ input is given by imposing a particular initial condition on the network . if that initial condition is sufficiently close to one of the memorized patterns , then the network s state should converge to the correct nearby attractor , and we say that the network has recognized or retrieved the pattern .
in this paper we follow the hn literature in considering the case where the patterns are random and uncorrelated strings of @xmath31 s and @xmath32 s .
we read the output of the network according to the _ signs _ of the @xmath9 .
thus we say that the network has recalled pattern 1 , for example , if @xmath33 for all @xmath34 .
although variations in the magnitude of @xmath9 can be important to the dynamics , we will for the moment ignore them for the purpose of reading the output . as we will see below , the retrieval states in general do not have @xmath35 even though the patterns have @xmath36 we focus here on the limiting case @xmath37 , @xmath38 , or large networks with low memory loading .
( strictly speaking , we take @xmath3 to infinity while @xmath28 remains finite . ) in this case the inner product of a pair of patterns * * * * @xmath39 behaves as a gaussian random variable with zero mean and variance @xmath40 , so that in the @xmath37 limit the pattern vectors are nearly orthogonal to each other and form a basis for a @xmath28-dimensional subspace of the @xmath3-dimensional configuration space . as in the hn
, we construct the coupling matrix from the stored patterns according to the hebb @xcitelearning rule : @xmath41 the term @xmath42 is included to make all diagonal elements of the coupling matrix zero .
non - zero diagonal entries would have the effect of adding an additional quadratic self - interaction term .
following the usual practice we omit them here so that the quadratic term is contained only in the local potential . for the case @xmath43 , however , the effect of the diagonal elements is negligible and we can substitute the simpler learning rule @xmath44 a useful set of order parameters are the overlaps @xmath45 , which are inner products of the network s state with each of the stored patterns : @xmath46 for the discrete hn , these variables take values @xmath47 , while for the bgn any real values are possible
. it will be useful to define another set of variables , which we will call `` bit overlaps , '' by should not be confused with the biases or external fields @xmath6 in ( [ hamiltonian ] ) , which are set to zero for the remainder of this paper . ]
@xmath48 the bit overlap is simply related to the hamming distance by @xmath49 where the hamming distance @xmath50 between two vectors is defined as the number of elements for which their signs differ , or the number of positions @xmath34 such that @xmath51 .
unlike @xmath52 the bit overlaps always obey @xmath53 .
they encode information about sign agreements but not about magnitudes of the outputs @xmath9 .
the definitions of the overlap variables allow us to rewrite the hamiltonian and the dynamical equations in useful forms .
@xcite in particular , if the synaptic matrix @xmath54 is given by the simplified hebb rule ( [ simplehebb ] ) , then the interaction term of the hamiltonian can be rewritten in terms of @xmath45 as follows : @xmath55 and the net input to a given node from the other nodes is given by @xmath56 to show that the network functions properly as an associative memory , we exhibit the attractor states corresponding to the stored patterns , demonstrate their stability , and show that a pattern can be retrieved from an initial condition which lies close to the pattern but differs from it by one or more incorrect signs .
consider the state @xmath57 where @xmath58 is a scalar and @xmath59 is a particular one of the stored patterns .
we will show that for a suitable value of @xmath58 this state represents a stable fixed point of the dynamics and is therefore the retrieval state we seek .
in this state , @xmath60 , @xmath61 , and all other overlaps are small .
the field acting on the @xmath34-th node can be written as follows:@xmath62 the sum over patterns @xmath63 is called the crosstalk term . for the overlaps with these other patterns we have@xmath64 since the patterns are random , each of these overlaps is of order @xmath65 .
the number of patterns remains finite as @xmath37 , so the sum in the crosstalk term vanishes in this limit and @xmath66 . a stationary state must satisfy the fixed point condition ( [ fixedpt ] ) for each node , which leads to a self - consistency condition on @xmath58 : @xmath67 in the last step , we have used the fact that @xmath68 when @xmath69 and then divided out the common factor of @xmath70 . solutions to this condition are an unstable equilibrium @xmath71 and two stable equilibria@xmath72 the two stable solutions represent perfect retrieval of pattern @xmath73 and its mirror state , respectively .
the doubled state is a consequence of the @xmath74 symmetry in the hamiltonian . since the overlap @xmath75 is equal to @xmath58 and all other overlaps vanish in the thermodynamic limit , the energy of this retrieval state is easily calculated using expression ( [ mham ] ) for the energy in terms of the overlap variables , giving:@xmath76 note that this energy expression is extensive ( proportional to @xmath3 ) and a monotonically decreasing function of @xmath8 .
having identified the state @xmath77 as an equilibrium state , we now demonstrate its stability using a linear stability analysis of the dynamical equations@xmath78 evaluating the jacobian @xmath79 at the fixed point ( [ retrievalstate ] ) we obtain : @xmath80 where @xmath81 is the kronecker delta .
the fixed point is linearly stable if and only if the jacobian has no positive eigenvalues .
this depends in turn on the eigenvalues of the synaptic connection matrix @xmath54 .
but in the limit where all of the stored patterns @xmath29 are mutually orthogonal , the stored patterns are themselves eigenvectors spanning a degenerate subspace with eigenvalue 1 , while the complement of this subspace has eigenvalue 0 .
( the hebb rule ( [ simplehebb ] ) itself gives a spectral decomposition of @xmath54 . )
since the maximum eigenvalue of @xmath54 is 1 , we see that the jacobian at the retrieval fixed point has no positive eigenvalues and so the retrieval state is linearly stable for any value of @xmath8 .
in fact , all eigenvalues become more negative as @xmath8 increases .
@xmath82 we reiterate that this result is valid in the ideal limit of large @xmath3 and low loading where the stored patterns are orthogonal . for finite - sized networks with finite overlaps among the patterns , it is possible for the memory states to be destabilized by the crosstalk terms .
this issue will be examined elsewhere .
numerical results for a network with @xmath26 nodes and @xmath27 random patterns agree excellently with the above description . to study a retrieval state numerically , we initialized the network to the state @xmath83 ( arbitrarily choosing the first pattern ) . starting at @xmath84 , we increased @xmath8 by small steps to @xmath85 .
at each step , we integrated the dynamical equations until they converged .
this procedure allows us to examine the evolution of a state under quasistatic changes in the control parameter @xmath86 we verified that @xmath87 remained equal to 1 over the whole range @xmath88 indicating that the retrieval state is stable .
the measured values of @xmath89 and @xmath90 were within 1% of the theoretical expressions ( [ retrievalm ] ) and ( [ retrievale ] ) respectively .
linear stability analysis has shown that the retrieval states are stable against _ infinitesimal _ perturbations for any value of @xmath8 , but this does not guarantee their stability against the flipping of signs of one or more nodes . in order to function as an associative memory ,
a network must be capable of dynamically correcting sign errors . when presented with an input at a small , nonzero hamming distance from one of the stored patterns ( i.e. , differing from it by a few reversed signs ) it must be able to flip the reversed signs and restore the correct pattern .
we will now show that there is a critical value @xmath91 above which the correction of sign - flip errors can occur . for smaller values of @xmath8
the bgn does not correct sign errors and thus does not truly function as an associative memory , but as @xmath8 increases above @xmath92 the retrieval states develop increasingly large basins of attraction . consider a state of the network which is a slightly corrupted retrieval state : all node variables have the values @xmath93 with the exception of one or possibly some number @xmath94 of nodes which may be misaligned .
in such a state the few misaligned nodes make only a small contribution to the overlap sums , so we have @xmath95 and @xmath96 .
the field acting on each node is therefore @xmath97 the misaligned bits experience a field opposite to their signs .
if the field becomes larger than the critical value @xmath98 then there is only one stable equilibrium for each node , and the misaligned nodes will flip to conform with the stored pattern .
error correction therefore occurs if @xmath99 the critical value occurs when the equality @xmath100 holds , or at @xmath91 .
if the pattern is more strongly corrupted ( a significant number of bits misaligned ) then the situation is more complicated , because the presence of a larger number of misaligned bits may reduce the value of @xmath75 and thus the magnitude of the field .
the misaligned bits have a significant back - reaction on the ones with the correct sign .
the correction of larger numbers of sign errors requires higher values of @xmath8 .
we will return to this point later ; the basic result is that when @xmath8 is only slightly above the threshold of @xmath101 the memory states have rather small basins of attraction , but these basins grow as @xmath8 increases .
in the case of the hn , the hebb learning rule results in a large number of `` spurious '' attractors in addition to the retrieval states . the energy function defines a rugged landscape , and a trajectory which does not start sufficiently close to one of the stored patterns may become trapped in one of the spurious local minima instead of one corresponding to a recalled pattern .
it is possible to suppress the spurious minima by introducing thermal noise which allows trajectories to jump out of the shallower basins of attraction into deeper ones . at low levels of loading ,
the hn possesses spurious attractors which are nonlinear combinations of the stored patterns .
there is a hierarchy of symmetric _ mixture states _ of the form@xcite@xcite@xmath102 these states overlap equally with @xmath103 different patterns : @xmath104 . for the hn ,
only the mixtures with odd @xmath103 are stable .
the @xmath105 mixtures have the lowest energy in this category , and the energies increase with @xmath106 asymptotically approaching @xmath107 .
as the number of stored patterns @xmath28 increases , these spurious states proliferate exponentially ; their number is of order @xmath108 .
there are also non - symmetric mixtures .
the proliferation of spurious states is associated with spin glass type behaviour in the hn .
accordingly we also refer to these mixed states somewhat loosely as _ spin glass states_. we will show here that the bgn possesses mixture states analogous to those of the hn , but their structure is slightly more complex .
let us focus on the @xmath105 symmetric mixture state with positive signs .
for the hn , this state is given by @xmath109 this state is stable against individual sign flips because each node is subject to a non - zero magnetic field which maintains its alignment . to see this , note that there are two possibilities for each bit
either all three patterns agree at that particular site ( @xmath110 or one of the patterns has the opposite sign from the other two , for example @xmath111 when all three agree , we say that the @xmath34-th bit is a unamimous bit .
if the patterns are random , then each @xmath112 is @xmath15 with equal probability , giving a probability of @xmath113 that a given bit is unanimous .
the mixture state has equal overlaps with all three of the patterns . since for a given @xmath34 there is a probability of @xmath114 that @xmath115 , we have @xmath116 , and likewise @xmath117 the field acting at the @xmath34-th site is @xmath118 ( consistent with the low - loading , large-@xmath3 limit , we ignore all other overlaps which are of order @xmath40 . ) this gives @xmath119 for unanimous bits and @xmath120 for the others . in any case ,
each node of the hn experiences a field which stabilizes its alignment .
we will now show that the bgn , like the hn , possesses a mixture state in which the sign of @xmath9 is given by the majority vote of three of the stored patterns : @xmath121 this state has a more complicated structure , however , because the magnitude of @xmath9 at a given node depends significantly on the local field at that node .
since the field at a unanimous bit is stronger than the field at a non - unanimous bit , we expect the magnitude of @xmath9 to be larger for a unanimous bit
. therefore , we make the _ _
ansatz__@xmath122{ll}% a\operatorname{sign}(\xi_{i}^{\mu_{1}}+\xi_{i}^{\mu_{2}}+\xi_{i}^{\mu_{3 } } ) & \text{if \ } \xi_{i}^{\mu_{1}}=\xi_{i}^{\mu_{2}}=\xi_{i}^{\mu_{3}}\\ d\operatorname{sign}(\xi_{i}^{\mu_{1}}+\xi_{i}^{\mu_{2}}+\xi_{i}^{\mu_{3 } } ) & \text{otherwise}% \end{array } \right .
\label{bgnthreemix}%\ ] ] where @xmath123 and @xmath124 are real numbers . the dynamical equations for the network give a pair of self - consistency equations which can be solved numerically for @xmath123 and @xmath125 first
, we need an expression for the overlap of the mixture state with one of the three patterns , say , pattern @xmath126 .
if the @xmath34-th bit is a unanimous bit , then @xmath127 has magnitude @xmath123 and agrees in sign with @xmath128 . on the other hand , if it is a non - unanimous bit , then @xmath127 has magnitude @xmath124 and has a @xmath129 probablity of agreeing in sign with @xmath128 .
the result is that @xmath130 note that in the special case @xmath131 , the above expression reduces to the hopfield value @xmath132 , as it should .
again , all three overlaps have the same size : @xmath133 . the total energy of the network in this state is given by@xmath134 we can think of this as an an energy function on the restricted family of states parametrized by ( [ bgnthreemix ] ) .
a necessary condition for the mixture state ( [ bgnthreemix ] ) to be a fixed point is that @xmath135 .
this gives us two self - consistency conditions for the parameters @xmath123 and @xmath124 : @xmath136 these are the equations of the nullclines of the energy function ( [ bgnthreee ] ) .
alternatively , the above equations could be derived directly from the dynamical equations for each node and the expressions for @xmath18 instead of using the energy function . for a _
stable _ fixed point , @xmath137 must be a local minimum of the energy function ( [ bgnthreee ] ) . the self - consistency equations ( [ selfconsist ] ) can be rewritten as follows:@xmath138 \nonumber\\ a & = \frac{4}{\gamma}\left [ d^{3}-(1+\frac{\gamma}{4})d\right ] .\end{aligned}\ ] ] the graphs of these two cubic equations are plotted for @xmath139 in figure [ nullclinesboth]a .
solutions are points where the two nullclines intersect .
note that the slope , @xmath140 , of the first curve at the origin is @xmath141 and for the second curve @xmath142 .
these two inequalities satisfied by the slopes ensure that the curves always intersect in at least 5 points .
the five solutions can be classified by looking at the energy function and its gradient .
we see that the solution @xmath143 is an unstable node ( maximum of the energy ) , the two solutions in the 2nd and 4th quadrants are saddle points , and the two solutions in the first and third quadrants are the stable solutions we seek .
( there are two because of the @xmath74 sign reversal symmetry one is a mirror state of the other . )
[ ptb ] fig1.eps the self - consistency equations for @xmath123 and @xmath124 were solved numerically for a range of @xmath8 values using a gradient descent algorithm .
these values were in turn substituted into ( [ bgnthreee ] ) to find the energy as a function of @xmath8 .
the results are plotted in figure [ mixcompare ] where they are also compared with numerical results from dynamical simulations of a bgn with @xmath26 .
we studied the mixture state numerically by initializing a network to the state @xmath144 and incrementing @xmath8 beginning at 0 , much as was done for the retrieval states .
figure [ mixcompare]a shows the magnitudes of unanimous and non - unanimous bits in the mixture state as functions of @xmath86 the solid lines show the solutions of the self consistency equations for @xmath123 and @xmath124 .
the symbols show the observed magnitudes @xmath145 for @xmath146 in the simulated network state .
two of these first six bits are unanimous while the other four are not .
there is good agreement between the observed values of @xmath147and the values obtained from the self - consistency equations . for comparison ,
@xmath148 is plotted as a dotted line on the same axes .
( recall that this is the value of all @xmath149 in a pure retrieval state . )
figure [ mixcompare]b shows a corresponding comparison of the observed and theoretical energies .
finally , figure [ mixcompare]c compares the @xmath105 mixture state with the retrieval state by plotting the ratios @xmath150 and @xmath151 as well as the ratio of the mixture state energy @xmath152to that of a retrieval state @xmath153 .
all three of these ratios appear to approach constant asymptotic values as @xmath8 increases .
asymptotically , @xmath154 , while for the hn the corresponding ratio is 0.75 .
the strength of the field acting on each unanimous bit , @xmath155 , and that acting on the non - unanimous bits , @xmath156 , both increase as @xmath8 increases .
the mixture state is stable against single sign flips of the unanimous bits when @xmath157 and stable against any single sign flip when @xmath158 .
thus as @xmath8 increases , the mixture state begins to develop a non - trivial basin of attraction of its own .
[ ptb ] fig2.eps it is interesting to note that at @xmath159 a saddle - node bifurcation occurs and for @xmath160 two aditional stable solutions to the self - consistency equations ( [ selfconsist ] ) appear in the second and fourth quadrants , at @xmath161 and @xmath162 .
( see figure [ nullclinesboth]b . )
these are states with @xmath163 in which @xmath164 and the net field acting on each node exactly cancels .
they are not stable against sign flips , and become completely destabilized when @xmath165 .
as we will see in the following section , they do not properly belong to the category of mixture or spin glass states , but rather to another class of spurious attractors present only in the bgn at low values of @xmath8 .
here we examined only the @xmath105 mixture state , but similar methods may be used to characterize higher - order mixtures .
in general , they are more complex as there are more possibilities for the size of the majority by which the sign is determined .
the magnitudes of @xmath9 then take a greater number of distinct values .
in section [ errorcorrect ] , we noted that for values of @xmath8 not far above @xmath92 , it is possible that a state may have a significant overlap with one of the stored patterns but that the field acting on the nodes may nonetheless not be strong enough to overcome the potential barrier and correct the sign errors .
indeed if @xmath8 is below @xmath101 then even a single sign error may go uncorrected .
this consequence of the bistability of the bgn units contrasts with the behavior of the hn .
consider first the case of the hn .
a typical random initial state has small but nonzero overlaps with the memorized patterns , @xmath166 , resulting in fields @xmath167 which are random with zero mean and variance of order @xmath168 . typically , for approximately half of the nodes @xmath9 and @xmath18 will initially have opposite signs .
since there is no potential barrier against sign flips , those nodes will change their signs , and the sign flips will continue until the field experienced by every node is aligned with @xmath9
. every sign flip will increase the magnitude of one or more of the overlap variables .
if , for example , one overlap @xmath75 is larger than all of the others , then most nodes will experience fields which tend to align them with pattern @xmath73 .
every sign flip further increases the value of @xmath75 , and eventually @xmath75 will be fully retrieved even though the initial overlap may have been quite small . however
, if one overlap does not clearly dominate the others , then the trajectory may arrive at a spurious attractor which has roughly equal overlaps with several patterns , instead of at a single one of the memory states .
even a state which is initially orthogonal to all of the memorized patterns can be rendered unstable by changing a single sign : even a single sign flip will create a small but nonzero field affecting the other nodes , resulting in further sign flips , and so on . in summary , for the hn
essentially any initial condition converges under the dynamics to an attractor lying in or close to the @xmath28-dimensional subspace of the patterns . for the bgn ,
on the other hand , the situation is different due to the presence of potential barriers . just as with the hn , given any state @xmath10
every node @xmath34 experiences a field @xmath18 , which may be aligned with or opposed to @xmath9
. however , the antiparallel local fields may not be strong enough to flip their nodes into the parallel direction .
if most @xmath18 are well below the threshold @xmath169 then the flipping of one or a few nodes will not change the field enough to cause any further flips .
thus there might be a large number of initial conditions which remain stuck with low overlaps , far away from any of the patterns .
we refer to such states with sub - threshold fields as uncondensed states , because in those states none of the order paramenters @xmath45 are condensed
. however , we will show below that states with low overlaps can not remain stable for @xmath165 , and thus for higher values of @xmath8 the behavior is hopfield - like , with all trajectories collapsing toward the pattern subspace .
consider a hypothetical state which is strictly orthogonal to all memory patterns , so that @xmath170 for all @xmath171 .
( the extra solutions appearing in the self - consistency equations for the mixture state when @xmath160 are examples of such states . ) if @xmath170 for all @xmath171 , then @xmath172 for all @xmath34 . in this case
, the steady state of each node is @xmath173 .
proceeding with linear stability analysis as above , we find the relevant jacobian@xmath174 the equilibrium is unstable if @xmath175 , where @xmath176 is the largest eigenvalue of the coupling matrix .
@xmath177 is at least 1 .
therefore , if @xmath165 , states orthogonal to the stored patterns are all unstable . if all of the @xmath28 stored patterns are mutually orthogonal ( which is approximately true in the limit @xmath37 ) then for @xmath165 there are @xmath28 unstable eigenvalues
. we can examine one of those unstable directions , say , the one associated with the @xmath73-th pattern , more closely by making an explicit _ ansatz .
_ let us denote the orthogonal , zero - overlap state by @xmath178 ; by assumption @xmath179 for all @xmath171 including @xmath180 . for half of the nodes @xmath181 ; for the other half @xmath182 .
let us then consider a family of states described by the _ _
ansatz__@xmath183{ll}% a\xi_{i}^{\bot}=a\xi_{i}^{\nu } & \text{if } \xi_{i}^{\bot}=\xi_{i}^{\nu}\\ b\xi_{i}^{\bot}=-b\xi_{i}^{\nu } & \text{if } \xi_{i}^{\bot}=-\xi_{i}^{\nu}% \end{array } \right .
\label{unstable}%\ ] ] this parametrizes a 2-dimensional subspace of the state space which contains @xmath184 and one of its unstable eigenvectors . using the methods of section [ mixture ] ,
one obtains a pair of cubic self - consistency equations for @xmath123 and @xmath185:@xmath186 for all @xmath8 there are two solutions @xmath187 which correspond to the retrieval states @xmath188 . for @xmath160
there are additional stable solutions @xmath189 corresponding to @xmath190 @xmath191 .
a bifurcation occurs at @xmath159 and these solutions become saddle points .
thus , we see that there is an absolute upper limit for the existence of stable uncondensed states .
in fact , @xmath159 turns out to be a high upper bound .
the example of a state with all @xmath45 equal to zero is a sort of worst - case scenario . for a finite - sized network
the typical random initial condition has small but nonzero overlaps .
in addition , if the patterns are truly random then they will not be exactly orthogonal but have small overlaps and so the largest eigenvalue of the synaptic matrix will be slightly larger than unity . because of these factors the typical uncondensed state becomes unstable at values of @xmath8 lower than 2 ; in numerical simulations we found that for the case @xmath192 most become unstable between @xmath193 and @xmath194 .
figures [ dyna05]-[dyn12b ] , we show numerical results for the fate of a typical random initial condition of the bgn with @xmath26 , @xmath27
. figures [ dyna05]-[dyna12 ] show the dynamical evolution of the same initial condition , at different values of @xmath8 .
the initial condition was a random string of @xmath15 values .
we plot the energy per node , all five overlap variables @xmath45 and all five bit overlap variables @xmath195 as functions of time .
recall that @xmath195 contain information about sign agreements only . for @xmath196 ( figure [ dyna05 ] ) the state changes very little before convergence occurs .
the energy per node remains very close to @xmath197 .
the overlaps @xmath45 increase slightly in magnitude , but the bit overlaps do not change at all , indicating that no sign flips occur . for @xmath193 ( figure [ dyna10 ] ) the trajectory is similar , except that the small initial overlaps are amplified to a greater extent ( we will explain this effect below . ) the bit overlaps still do not change . when @xmath198 , however , the trajectory changes qualitatively ( figure [ dyna12 ] ) .
the magnitudes of the overlaps @xmath45 grow slowly until at @xmath199 the resulting field becomes strong enough to begin flipping some signs . at this point
the bit overlaps begin to change , the energy drops significantly and the trajectory moves close to the pattern subspace . after some further evolution
, the system converges to a mixture state which overlaps with several patterns .
a different random initial condition , followed again at @xmath198 , leads instead to a memory state ( figure [ dyn12b ] ) .
here one of the five overlaps becomes dominant and the others shrink away .
in this case the mirror state of one of the five patterns is retrieved .
these trajectories are typical examples representing descent on a rugged energy landscape .
different initial conditions lead to different attractors , of which some are memory states and some are mixtures .
frequently the trajectory lingers at one or several states before settling at its asymptotic attractor .
[ ptb ] fig3.eps [ ptb ] fig4.eps [ ptbptb ] fig5.eps [ ptbptbptb ] fig6.eps figures [ dyna05]-[dyna10 ] illustrated that for sufficiently small values of @xmath8 , the dynamics amplifies small initial overlaps without flipping the sign of any node . for more insight into this phenomenon , consider an initial state in which all @xmath9 are @xmath15 but somewhat more nodes are aligned parallel with one particular pattern @xmath200 than are antiparallel .
in other words , @xmath201 is nonzero but less than unity . for simplicity
let us neglect all other overlaps .
initially , each node experiences a small field given by @xmath202 this field will push @xmath9 to larger magnitudes @xmath203 for those nodes which are aligned with pattern @xmath200 , and it will push the others to smaller magnitudes @xmath204 . this adjustment in turn increases the value of @xmath75 , until an equilibrium is reached with @xmath205 .
we might think of this as a kind of subliminal recognition of the pattern . the effect becomes stronger as @xmath8 increases .
clearly it has a nonlinear dependence on both @xmath8 and @xmath201 .
when the field becomes large enough it will exceed the threshold for sign flips and the state will be attracted toward the pattern retrieval state .
the larger @xmath8 is , the smaller the initial @xmath201 that is necessary to fully retrieve the pattern @xmath200 . in other words ,
the basins of attraction of the patterns expand as @xmath8 increases .
in this section , we provide numerical support for the three - way classification of attractors into retrieval , spin glass and uncondensed states and we show how the respective attractor basins change with the control parameter @xmath206 .
we observe an interpolation between two different regimes .
as we showed above , for @xmath207 there are no stable uncondensed states . for lower values of @xmath8 , on the other hand ,
uncondensed states are numerous .
recall that uncondensed states are characterized by local fields too weak to overcome the potential barriers against sign flips , and so their dynamics is dominated by the local potential . in the extreme case @xmath84 ,
there are of course no magnetic fields at all and only the local potential is present .
the classification of attractors is very clearly reflected in the energy spectrum . to explore attractors and their basins , we seeded the network with 500 random initial conditions ( taken with several different realizations of the five random patterns ) , integrated the dynamical equations until they converged , and constructed a histogram of the final energies ( figure [ seedhist ] ) . for the case @xmath193 ( figure [ seedhist]a ) , there are three clearly separated clusters of attractors
. those with the lowest energies are retrieval states , while the states clustered at @xmath208 are the uncondensed states , and those in the intermediate range are the glassy states .
the picture is qualitatively similar at the slightly larger value @xmath209 ( figure [ seedhist]b ) , but the peak at @xmath208 has shrunk relative to the other two .
note that the energies of the retrieval and spin glass states change with @xmath8 , while the uncondensed states remain at nearly the same energy because their dynamics is dominated by the local potential .
for @xmath159 , ( figure [ seedhist]c ) on the other hand , the cluster at @xmath208 is absent as there are no stable uncondensed states .
the histogram for a hn ( figure [ seedhist]d ) resembles that for the bgn with @xmath159 .
one quantitative difference is that the retrieval state peak is slightly higher for the bgn with @xmath159 , while the glassy states are comparatively suppressed .
we performed this experiment at a range of values of @xmath8 . in all cases
the classification of states was clear from the energy spectrum and was verified by examining the final values of @xmath195 .
figure [ attractorprobs ] shows the probabilities of convergence to each of the three types of attractors from a random initial condition as functions of @xmath8 . at @xmath196
the landscape is dominated by the uncondensed states .
even though @xmath196 lies above the threshold of @xmath92 and the retrieval states have non - trivial basins of attraction , these basins still occupy a very small fraction of the total configuration space volume .
the patterns can be retrieved only if the initial overlaps are relatively high , and the probability of a _
random _ initial condition being sufficiently close is very low .
the retrieval probability becomes significant only as @xmath8 approaches 1 . as @xmath8 increases from 1 to 1.5 , basins for the memory and spin glass states grow at the expense of of the uncondensed states until the latter disappear .
the retrieval state basins grow faster than those of the spin glass states . beyond @xmath210 ,
the probability of retrieving a memory state saturates at approximately a @xmath211 higher value than in the hopfield case , and the probability of falling into a spin glass state is correspondingly lower .
[ ptb ] fig7.eps [ ptbptb ] fig8.eps in an attempt to map the attractor basins in more detail , we generated configurations at specified initial hamming distances from particular memory patterns .
this was done by starting with a pattern @xmath59 and flipping the signs of a specified number of randomly chosen bits . using an ensemble of such initial conditions
, we measured the probability of retrieval of the target pattern @xmath59 as a function of the initial distance from it . as a rule ,
the probability of recognizing the pattern is high if only a few signs are flipped , but drops sharply if a certain threshold hamming distance is exceeded .
we are interested in learning where this threshold lies , and thus answering the question of how close an initial condition must be to a pattern in order to be attracted to it .
we are also interested in the fate of states lying just outside the boundaries of a basin of attraction .
in other words , does the basin share a boundary with the basins of other patterns , or only with spurious attractors ?
the results are presented in figure [ boundaries ] for a bgn with @xmath26 and @xmath27 , for the three values @xmath212 and @xmath213 , and also for the hn . in each of these cases
, we generated an ensemble of initial conditions at a particular initial value of @xmath201 for some pattern .
each initial condition was allowed to evolve under the dynamics and the resulting attractor was classified as : a ) the target pattern @xmath214 b ) one of the other patterns @xmath215 c ) a spin glass spurious state or d ) an uncondensed state . in this case
, we classified a state as uncondensed if no sign flips occured during the dynamical evolution .
the probabilities of each of these four outcomes were averaged over several realizations of the random patterns and plotted as functions of the initial bit overlap @xmath201 .
[ ptb ] fig9.eps consider first the hn data from figure [ boundaries]d .
a pattern can evidently be retrieved even if the initial overlap is fairly small ; the probability is close to unity if @xmath216 .
if the target pattern is not retrieved , then either a spurious attractor or one of the other patterns may be retrieved .
there is a range of @xmath217 over which all three probabilities are significant , indicating that the basins for the memory states border on each other as well as those of spurious states . for an @xmath26 network ,
the expected magnitude of the overlap of a random state with any given one of the stored patterns is @xmath218 , which is not much smaller than the apparent threshold of @xmath219 this is consistent with the view that for the hn , a pattern is likely to be retrieved as long as the initial overlap with that pattern is significantly larger than all of the other overlaps .
the @xmath159 bgn ( figure [ boundaries]c ) shares the qualitative features of the hn .
note however that the probability of becoming trapped in a spurious state is smaller for the bgn , consistent with the results in figure [ attractorprobs ] .
a contrasting case is the @xmath196 bgn ( figure [ boundaries]a ) . in this case retrieval of the target pattern
requires an initial overlap of more than 0.5 .
although this represents a significant basin , it is highly unlikely that a _ random _ initial condition will have such a large overlap , thus explaining why random initial conditions almost never flow to a memory state . the basins of the memory states are bordered only by spurious states , not by other memory states .
interestingly , the states which lie adjacent to the basin of a memory pattern are not all uncondensed some are spurious states of the mixture or spin glass type .
examination of the states retrieved near the boundary shows that these are typically _ asymmetric _ mixture states with one large overlap and two or more smaller ( but greater than random ) overlaps .
finally , for @xmath193 ( figure [ boundaries]b ) , the basins of the memory state are almost as large as in the hn case , and near the boundaries there is a small but nonzero probability of retrieving one of the other memory patterns , indicating that the basins of different memory states almost touch each other . taken together ,
the above results suggest a qualitative , schematic picture of the energy landscape illustrated in figures [ landscape05]-[landscape2 ] .
the representation of the configuration space by two dimensions is not to be taken literally , since it is of course @xmath3-dimensional . at low values of @xmath8 , such as 0.5 ,
the energy landscape is dominated by uncondensed states , which form a series of shallow basins , each limited to roughly a single orthant of @xmath3-dimensional space .
these are represented in the diagram by a series of shallow pits .
the basins of attraction for the retrieval and mixture states form isolated depressions in this pitted plateau .
they occupy nontrivial volumes but do not lie adjacent to each other ( with the exception of certain spin glass states which lie near the retrieval states ) . at intermediate values @xmath220 , the basins of attraction of the retrieval states are much larger and in some places almost touch each other , but significant islands of uncondensed states remain . by @xmath221 ,
however , the uncondensed states have disappeared and the basins of attraction for the other two types of states occupy the entire energy landscape and share boundaries with each other .
[ ptb ] fig10.eps [ ptbptb ] fig11.eps [ ptbptbptb ] fig12.eps
we have studied the behaviour of the bistable gradient network in the thermodynamic low - loading limits @xmath37 , @xmath38 .
we described and classified the attractors of the dynamics and also observed the effectiveness of pattern retrieval as a function of the coupling parameter @xmath8 .
we found that states corresponding to perfect retrieval of the stored patterns are linearly stable at all values of @xmath8 , and have an energy that decreases monotonically with @xmath8 . above the threshold
@xmath222 the retrieval states become stable against sign flips of one or more nodes , and the network begins to function as an associative memory . if @xmath8 is not far above this threshold , then the basins of attraction of the retrieval states are small and input must be very close in hamming distance to a pattern for recognition to occur .
the basins of attraction of the retrieval states grow as @xmath8 increases .
there are two regimes of behaviour , distinguished by the types of attractors that occur . at low @xmath8
the configuration space is dominated by the uncondensed states , or states in which no node experiences a field strong enough to overcome its potential barrier . in these states
, @xmath145 remains close to 1 for all nodes , and the energy remains close to @xmath197 .
each of these states occupies a basin of attraction confined to approximately a single orthant . in the limit @xmath84
there are @xmath223 such states , all degenerate in energy . as @xmath8 increases above the threshold @xmath91
, the retrieval states and the mixture or spin glass states at first occupy small isolated basins among the many uncondensed states
. however , as @xmath8 increases further , these basins grow until they lie adjacent to each other . at some value of @xmath8 (
observed to lie between 1 and 1.5 ) , the uncondensed states disappear and there is a transition to a hopfield - like regime where the basins of attraction for retrieval and spin glass states cover the whole configuration space .
as @xmath8 increases still further , the retrieval basins grow at the expense of the spin glass states , so that the latter can be noticeably suppressed compared to the deterministic hopfield case .
this suppression of the spurious states occurs without thermal noise or a modification of the hebb learning rule .
the uncondensed states represent a phase which is neither `` ferromagnetic '' ( i , e , strongly ordered and correlated with one pattern ) nor `` glassy '' in the sense that frustration is an important effect , yet they can not properly be described as `` paramagnetic , '' as paramagnetism is characterized by spins which are able to flip freely from one orientation to the other .
a few words on the application of such networks to practical problems of associative memory are in order .
the goal of associative memory is to reconstruct a pattern from a more or less corrupted version or from a fragment of the pattern , without becoming trapped in a spurious local minimum . from this point of view
, it appears that increasing @xmath8 improves the performance of the network expanding the basins of attraction for the retrieval states and suppressing the spurious states .
the low-@xmath8 regime , on the other hand , may be suited to applications where the goal is a _ selective _ associative memory , one which only recognizes a pattern from a fairly close approximation and thus avoids false recognition . in the low-@xmath8 regime , if the input is not close to one of the stored patterns , then the network is likely to remain in an uncondensed state .
these can in general be distinguished clearly from other states ( especially retrieval states ) by their relatively high energy ( @xmath224 ) or by the fact that the magnitudes of all @xmath149 remain close to 1 .
the magnitudes of the outputs can therefore be read as a signal of whether recognition has occurred .
persistence in an uncondensed state corresponds to an i do nt know or nonrecognition response . in a subsequent publication
, we will examine the behavior of the bgn when the loading level @xmath225 is of order unity , and we will demonstrate another performance trade - off . specifically , we will show that the maximum storage capacity of the network decreases as @xmath8 increases . for a low-@xmath8 regime ,
it is possible to stabilize more memorized patterns than in the hopfield case , while at higher @xmath8 , even though the low - loading fault tolerance is increased , the storage capacity decreases .
w. hohmann , m. krauss and f.w .
schneider , j. phys .
a * 103 * , 7606 ( 1999 ) ; j phys .
chem a 102 , 3103 ( 1998 ) ; j. phys .
chem a 101 , 7364 ( 1997 ) .
g. dechert , k .-
zeyer , d. lebender and f.w .
schneider , j. phys chem a 100 , 19043 ( 1996 ) .
laplante , m. pemberton , a. hjelmfelt and j. ross , j. phys .
chem 99 , 10063 ( 1995 ) .
v. booth , t. erneux and j .- p .
laplante , j. phys .
chem 98 , 6537 ( 1994 ) .
a. hjelmfelt and j. ross , j. phys .
97 , 7998 ( 1993 ) .
amit , h. gutfreund and h. sompolinsky , phys.rev . *
a32 * , 1007 ( 1985 ) ; phys .
55 * , 1530 ( 1985 ) ; ann . phys .
* 173 * , 30 ( 1987 ) .
a. chrisanti , d.j .
amit and h. gutfreund , europhys .
* 2 * , 337 ( 1986 ) . | we examine the large - network , low - loading behaviour of an attractor neural network , the so - called bistable gradient network ( bgn ) .
we use analytical and numerical methods to characterize the attractor states of the network and their basins of attraction .
the energy landscape is more complex than that of the hopfield network and depends on the strength of the coupling among units . at weak coupling ,
the bgn acts as a highly selective associative memory ; the input must be close to the one of the stored patterns in order to be recognized .
a category of spurious attractors occurs which is not present in the hopfield network .
stronger coupling results in a transition to a more hopfield - like regime with large basins of attraction .
the basins of attraction for spurious attractors are noticeably suppressed compared to the hopfield case , even though the hebbian synaptic structure is the same and there is no stochastic noise . |
proposed by @xcite , cca is a technique for finding pairs of vectors that maximises the correlation between a set of paired variables .
the set of paired variables can be considered as two views of the same object , a perspective we adopt throughout the paper . since the debut of cca , a multitude of analyses , adaptations and applications have been proposed @xcite . + +
the potential disadvantage of cca and similar statistical methods , such as principle component analysis ( pca ) and partial least squares ( pls ) , is that the learned projections are a linear combination of all the features in the primal and dual representations respectively . this makes the interpretation of the solutions difficult .
studies by @xcite and the more recent @xcite have addressed this issue for pca and pls by learning only the relevant features that maximise the variance for pca and covariance for pls . a previous application of sparse cca has been proposed in @xcite where the authors imposed sparsity on the semantic space by penalising the cardinality of the solution vector @xcite . the scca presented in this paper is novel to the extent that instead of working with covariance matrices @xcite , which may be computationally intensive to compute when the dimensionality of the data is large , it deals directly with the training data .
+ + in the machine learning ( ml ) community it is common practice to refer to the input space as the primal - representation and the kernel space as the dual - representation . in order to avoid confusion with the meanings of the terms primal and dual commonly used in the optimisation literature , we will use ml - primal to refer to the input space and ml - dual to refer to the kernel space for the remainder of the paper , though note that the references to primal and dual in the abstract refer to ml - primal and ml - dual .
+ + we introduce a new convex least squares variant of cca which seeks a semantic projection that uses as few relevant features as possible to explain as much correlation as possible . in previous studies ,
cca had either been formulated in the ml - primal ( input ) or ml - dual ( kernel ) representation for both views .
these formulations , coupled with the need for sparsity , could prove insufficient when one desires or is limited to a ml primal - dual representation , i.e. one wishes to learn the correlation of words in one language that map to documents in another .
we address these possible scenarios by formulating scca in a ml primal - dual framework in which one view is represented in the ml - primal and the other in the ml - dual ( kernel defined ) representation .
we compare scca with kcca on a bilingual english - french and english - spanish data - set for a mate retrieval task .
we show that in the mate retrieval task scca performs as well as kcca when the number of original features is small and scca outperforms kcca when the number of original features is large .
this emphasises scca s ability to learn the semantic space from a small number of relevant features .
+ + in section [ sec1 ] we give a brief review of cca , and section [ sec2 ] formulates and defines scca . in section [ sec4 ]
we derive our optimisation problem and show how all the pieces are assembled to give the complete algorithm .
the experiments on the paired bilingual data - sets are given in section [ sec5 ] .
section [ sec6 ] concludes this paper .
we briefly review canonical correlation analysis and its ml - dual ( kernel ) variant to provide a smooth understanding of the transition to the sparse formulation .
first , basic notation representation used in the paper is defined @xmath0 + the correlation between @xmath1 and @xmath2 can be computed as @xmath3 where @xmath4 and @xmath5 are the within - set covariance matrices and @xmath6 is the between - sets covariance matrix , @xmath7 is the matrix whose columns are the vectors @xmath8 , @xmath9 from the first representation while @xmath10 is the matrix with columns @xmath8 from the second representation .
we are able to observe that scaling @xmath11 does not effect the quotient in equation ( [ math : sim ] ) , which is therefore equivalent to maximising @xmath12 subject to @xmath13 . + + the kernelising of cca @xcite offers an alternative by first projecting the data into a higher dimensional feature space @xmath14 before performing cca in the new feature spaces .
the kernel variant of cca is useful when the correlation is believed to exist in some non linear relationship .
given the kernel functions @xmath15 and @xmath16 let @xmath17 and @xmath18 be the linear kernel matrices corresponding to the two representations of the data , where @xmath7 is now the matrix whose columns are the vectors @xmath19 , @xmath9 from the first representation while @xmath10 is the matrix with columns @xmath20 from the second representation .
the weights @xmath21 and @xmath22 can be expressed as a linear combination of the training examples @xmath23 and @xmath24 .
substitution into the ml - primal cca equation ( [ math : sim ] ) gives the optimisation @xmath25 which is equivalent to maximising @xmath26 subject to @xmath27 .
this is the ml - dual form of the cca optimisation problem given in equation ( [ math : sim ] ) which can be cast as a generalised eigenvalue problem and for which the first @xmath28 generalised eigenvectors can be found efficiently .
both cca and kcca can be formulated as symmetric eigenproblems .
+ + a variety of theoretical analyses have been presented for cca @xcite .
a common conclusion of some of these analyses is the need to regularise kcca .
for example the quality of the generalisation of the associated pattern function is shown to be controlled by the sum of the squares of the weight vector norms in @xcite .
although there are advantages in using kcca , which have been demonstrated in various experiments across the literature , we clarify that when using a linear kernel in both views , regularised kcca is the same as regularised cca ( since the former and latter are linear ) .
nonetheless using kcca with a linear kernel can have advantages over cca , the most important being speed when the number of features is larger than the number of samples .
the motivation for formulating a ml primal - dual scca is largely intuitive when faced with real - world problems combined with the need to understand or interpret the found solutions .
consider the following examples as potential case studies which would require ml primal - dual sparse multivariate analysis methods , such as the one proposed .
* enzyme prediction ; in this problem one would like to uncover the relationship between the enzyme sequence , or more accurately the sub - sequences within each enzyme sequence that are highly correlated with the possible combination of the enzyme reactants .
we would like to find a sparse ml - primal weight representation on the enzyme sequence which correlates highly to sparse ml - dual feature vector of the reactants .
this will allow a better understanding of the enzyme structure relationship to reactions . *
bilingual analysis ; when learning the semantic relationship between two languages , we may want to understand how one language maps from the word space ( ml - primal ) to the contextual document ( ml - dual ) space of another language . in both cases
we do not want a complete mapping from all the words to all possible contexts but to be able to extract an interpretable relationship from a sparse word representation from one language to a particular and specific context ( or sparse combination of ) in the other language .
* brain analysis ; here , one would be interested in finding a ( ml - primal ) sparse voxelmm@xmath29 mm . ] activation map to some ( ml - dual ) non - linear stimulus activation ( such as musical sequences , images and various other multidimensional input ) .
the potential ability to find only the relevant voxels in the stimuli would remove the particularly problematic issue of thresholding the full voxel activation maps that are conventionally generated .
for the scope of this paper we limit ourselves to experiments with the bilingual texts problems .
+ + throughout the paper we only consider the setting when one is interested in a ml - primal representation for the first view and a ml - dual representation for the second view , although it is easily shown that the given derivations hold for the inverted case ( i.e. a ml - dual representation for the first view and a ml - primal representation for the second view ) which is therefore omitted .
+ + consider a sample from a pair of random vectors ( i.i.d assumptions hold ) of the form @xmath30 each with zero mean ( i.e. centred ) where @xmath9 .
let @xmath7 and @xmath10 be matrices whose columns are the corresponding training samples and let @xmath31 be the kernel matrix of the second view and @xmath22 be expressed as a linear combination of the training examples @xmath32 ( note that @xmath33 is a general vector and should not be confused with notation sometimes used for unit coordinate vectors ) .
the primal - dual cca problem can be expressed as a primal - dual rayleigh quotient @xmath34 where we choose the primal weights @xmath21 of the first representation and dual features @xmath33 of the second representation such that the correlation @xmath35 between the two vectors is maximised . as we are able to scale @xmath21 and @xmath33 without changing the quotient , the maximisation in equation ( [ eig : origin ] ) is equal to maximising @xmath36 subject to @xmath37 .
for simplicity let @xmath38 , @xmath39 and @xmath40 . + + having provided the initial primal - dual framework we proceed to reformulate the problem as a convex sparse least squares optimisation problem .
we are able to show that maximising the correlation between the two vectors @xmath41 and @xmath42 can be viewed as minimising the angle between them .
since the angle is invariant to rescaling , we can fix the scaling of one vector and then minimise the norm to be the @xmath43norm . ] between the two vectors @xmath44 subject to @xmath45 .
this intuition is formulated in the following theorem , [ theo1 ] vectors @xmath46 are an optimal solution of equation ( [ eig : origin ] ) if and only if there exist @xmath47 such that @xmath48 are an optimal solution of equation ( [ eigsec ] ) .
theorem [ theo1 ] is well known in the statistics community and corresponds to the equivalence between one form of alternating conditional expectation ( ace ) and cca @xcite . for an exact proof
see theorem @xmath49 on page @xmath50 in @xcite .
+ + constraining the @xmath43norm of @xmath41 ( or @xmath42 ) will result in a non convex problem , i.e we will not obtain a positive / negative - definite hessian matrix
. motivated by the rayleigh quotient solution for optimising cca , whose resulting symmetric eigenproblem does _ not _ enforce the @xmath45 constraint , i.e. the optimal solution is invariant to rescaling of the solutions .
therefore we replace the scaling of @xmath45 with the scaling of @xmath33 to be @xmath51 .
we will address the resulting convexity when we achieve the final formulation .
+ + after finding an optimal cca solution , we are able to re - normalise @xmath33 so that @xmath45 holds .
we emphasis that even though @xmath52 has been removed from the constraint the link to kernels ( kernel tricks and rkhs ) is represented in the choice of kernel @xmath52 used for the dual - view , otherwise the presented method is a sparse linear cca .
we can now focus on obtaining an optimal sparse solution for @xmath46 .
+ + it is obvious that when starting with @xmath53 further minimising is impossible . to avoid this trivial solution and to ensure that the constraints hold in our starting condition ,
therefore there must be at least one @xmath54 for some @xmath55 that is equal to @xmath56 .
] we set @xmath51 by fixing @xmath57 for some fixed index @xmath58 so that @xmath59 $ ] . to further obtain a sparse solution on @xmath33
we constrain the @xmath60norm of the remaining coefficients @xmath61 , where we define @xmath62 $ ] .
the motivation behind isolating a specific @xmath28 and constraining the @xmath60norm of the remaining coefficients , other than ensuring a non - trivial solution , follows the intuition of wanting to find similarities between the samples given some basis for comparison . in the case of documents ,
this places the chosen document ( indexed by @xmath28 ) in a semantic context defined by an additional ( sparse ) set of documents .
this captures our previously stated goal of wanting to be able to extract an interpretable relationship from a sparse word representation from one language to a particular and specific context in the other language .
the @xmath63 choices of @xmath28 correspond to the @xmath64 projection vectors .
we discuss the selections of @xmath28 and the ensuring of orthogonality of the sparse projections in section [ sccaalgsec ] . + + we are also now able to constrain the @xmath60norm of @xmath65 without effecting the convexity of the problem .
this gives the final optimisation as @xmath66 subject to @xmath51 .
the expression @xmath67 is quadratic in the variables @xmath65 and @xmath33 and is bounded from below @xmath68 and hence is convex since it can be expressed as @xmath69 h [ { \mathbf{w } } ' { \mathbf{e}}']'$ ] . if @xmath70 were not positive definite taking multiple @xmath71 of the eigenvector @xmath72 $ ] with negative eigenvalue @xmath73 would give @xmath74 creating arbitrarily large negative values . when minimising subject to linear constraints ( 1-norms are linear ) this makes the whole optimisation convex .
+ + while equation ( [ eqn : stacca1 ] ) is similar to least absolute shrinkage and selection operator ( lasso ) @xcite , it is not a standard lasso problem unless @xmath33 is fixed .
the problem in equation ( [ eqn : stacca1 ] ) could be considered as a double - barreled lasso where we are trying to find sparse solutions for _ both _ @xmath46 .
we propose a novel method for solving the optimisation problem represented in equation ( [ eqn : stacca1 ] ) , where the suggested algorithm minimises the gap between the primal and dual lagrangian solutions using a greedy search on @xmath75 .
the proposed algorithm finds a sparse @xmath76 vectors , by iteratively solving between the ml primal and dual formulation in turn .
we give the proposed algorithm as the following high - level pseudo - code .
a more complete description will follow later ; * repeat 1 .
use the dual lagrangian variables to solve the ml - primal variables 2 .
check whether all constraints on ml - primal variables hold 3 .
use ml - primal variables to solve the dual lagrangian variables 4 .
check whether all dual lagrangian variable constraints hold 5 .
check whether 2 .
holds , if not go to 1 .
* end we have yet to address how to determine which elements in @xmath46 are to be non - zero .
we will show that from the derivation given in section [ primald ] a lower and upper bound is computed .
combining the bound with the constraints provides us with a criterion for selecting the non - zero elements for both @xmath65 and @xmath33 .
the criteria being that only the respective indices which violate the bound and the various constraints need to be updated .
+ + we proceed to give the derivation of our problem .
the minimisation @xmath77 subject to @xmath51 can be written as @xmath78 subject to @xmath51 , where @xmath71 , @xmath79 are fixed positive parameters .
+ + to simplify our mathematical notation we revert to uniformly using @xmath33 in place of @xmath80 , as @xmath28 will be fixed in an outer loop so that the only requirement is that no update will be made for @xmath81 , which can be enforced in the actual algorithm .
we further emphasis that we are only interested in the positive spectrum of @xmath33 , which again can be easily enforced by updating any @xmath82 to be @xmath83 constraint by updating any @xmath84 to be @xmath85 . ] .
therefore we could rewrite the constraint @xmath86 as @xmath87 + + we are able to obtain the corresponding lagrangian @xmath88 subject to @xmath89 where @xmath90 is the dual lagrangian variable on @xmath33 and @xmath47 are positive scale factors as discussed in theorem [ theo1 ] and @xmath91 is the all ones vector .
we note that as we algorithmically ensure that @xmath92 we are able to write @xmath93 as @xmath94 .
+ + the constants @xmath47 can also be considered as the hyper - parameters ( or regularisation parameters ) common in the lasso literature , controlling the trade - off between the function objective and the level of sparsity .
we show that the scale parameters can be treated as a type of dual lagrangian parameters to provide an underlying automatic determination of sparsity .
this potentially sub - optimal setting still obtains very good results and is discussed in section [ sec : hyper ] .
+ + to simplify the @xmath60norm derivation we express @xmath65 by its positive and negative components will only have the positive / negative values of @xmath65 and zero elsewhere . ]
such that @xmath95 subject to @xmath96 .
we limit ourselves to positive entries in @xmath33 as we expect to align with a positive subset of articles .
+ + this allows us to rewrite the lagrangian as @xmath97 the corresponding lagrangian in equation ( [ eqnlarg ] ) is subject to @xmath98 the two new dual lagrangian variables @xmath99 are to uphold the positivity constraints on @xmath100 . in this section
we will show that the constraints on the dual lagrangian variables will form the criterion for selecting the non - zero elements from @xmath65 and @xmath33 .
first we define further notations used . given the data matrix @xmath101 and kernel matrix @xmath102 as defined in section [ sec2 ] , we define the following vectors @xmath103\\ { \mathbf{w}}^- & = & \left[w^-_1,\ldots , w^-_m\right]\\ \bm{\alpha}^+ & = & \left [ \alpha^+_1,\ldots , \alpha^+_m \right]\\ \bm{\alpha}^- & = & \left [ \alpha^-_1,\ldots , \alpha^-_m \right]\\ { \mathbf{e } } & = & \left[e_1,\ldots , e_\ell \right]\\ \bm{\beta } & = & \left[\beta_1,\ldots,\beta_\ell \right].\end{aligned}\ ] ] throughout this section let @xmath55 be the index of either @xmath46 that needs to be updated .
we use the notation @xmath104 or @xmath105_i$ ] to refer to the @xmath55th index within a vector and @xmath106 to refer to the @xmath55th element on the diagonal of a matrix .
+ + + taking derivatives of equation ( [ eqnlarg ] ) in respect to @xmath107 , @xmath108 , @xmath109 and equating to zero gives @xmath110 adding the first two equations gives @xmath111 implying a lower and upper component - wise bound on @xmath112 of @xmath113 we use the bound on @xmath114 to indicate which indices of the vector @xmath65 need to be updated by only updating the @xmath115 s whose corresponding @xmath116 violates the bound .
similarly , we only update @xmath54 that has a corresponding @xmath117 value smaller than @xmath118 .
+ + we are able to rewrite the derivative with respect to @xmath107 in terms of @xmath119 @xmath120 we wish to compute the update rule for the selected indices of @xmath65 .
taking the second derivatives of equation ( [ eqnlarg ] ) in respect to @xmath107 and @xmath108 , gives @xmath121 so for the @xmath122 , the unit vector with entry @xmath56 , we have an exact taylor series expansion @xmath123 and @xmath124 respectively for @xmath125 and @xmath126 as @xmath127 giving us the exact update for @xmath125 by setting @xmath128_i.\end{aligned}\ ] ] therefore the update for @xmath125 is @xmath129 .
we also compute the exact update for @xmath126 as @xmath130_i,\end{aligned}\ ] ] so that the update for @xmath126 is @xmath131 . recall that @xmath132 , hence the update rule for @xmath115 is @xmath133
therefore we find that the new value of @xmath115 should be @xmath134_i.\end{aligned}\ ] ] we must also consider the update of @xmath115 when @xmath116 is within the constraints and @xmath135 , i.e. previously @xmath116 had violated the constraints triggering the updated of @xmath115 to be non zero .
notice from equation ( [ sccdev1 ] ) that @xmath136 it is easy to observe that the only component which can change is @xmath137 , therefore as we need to update @xmath115 towards zero .
hence when @xmath138 the absolute value of the update is @xmath139 else when @xmath140 then the update is the negation of @xmath141 so that the update rule is @xmath142 . in the updating of @xmath115
we ensure that @xmath143 do not have opposite signs ,
i.e. we will always stop at zero before updating in any new direction .
+ + we continue by taking second derivatives of the lagrangian in equation ( [ eqnlarg ] ) with respect to @xmath109 , which gives @xmath144 so for @xmath122 , the unit vector with entry @xmath56 , we have an exact taylor series expansion @xmath145 giving us the following update rule for @xmath146 @xmath147_i,\end{aligned}\ ] ] the update for @xmath33 is @xmath148 .
the new value of @xmath54 will be @xmath149_i,\ ] ] again ensuring that @xmath150 .
input : data matrix @xmath151 , kernel matrix @xmath152 and the value @xmath28 .
@xmath153 initialisation : @xmath154 , @xmath155 , @xmath156 , @xmath57 @xmath157 , @xmath158 @xmath159 @xmath160 @xmath153 update the found weight values : converge over @xmath161 using algorithm [ alg : sccaw ] @xmath153 find the dual values that are to be updated @xmath162 @xmath163 @xmath153 update the found dual projection values converge over @xmath109 using algorithn [ alg : sccae ] @xmath153 find the weight values that are to be updated @xmath164 @xmath160 @xmath165 , @xmath166 * output * : feature directions @xmath167 @xmath168 @xmath169 $ ] @xmath170 @xmath171 $ ] @xmath172 @xmath173 @xmath174 @xmath175 @xmath176 $ ] @xmath177 @xmath178 observe that in the initial condition when @xmath179 from equations ( [ sccdev1 ] ) we are able to treat the scale parameters @xmath180 as dual lagrangian variables and set them to @xmath181 we emphasise that this is to provide an underlying automatic determination of sparsity and may not be the optimal setting although we show in section [ sec : hyper ] that this method works well in practice . combining all the pieces we give the scca algorithm as pseudo - code in algorithm [ alg : scca ] , which takes @xmath28 as a parameter . in order to choose the optimal value of @xmath28 we would need to run the algorithm with all values of @xmath28 and select the one giving the best objective value .
this would be chosen as the first feature .
+ + to ensure orthogonality of the extracted features @xcite for each @xmath182 and corresponding @xmath183 , we compute the residual matrices @xmath184 , @xmath185 by projecting the columns of the data onto the orthogonal complement of @xmath186 , a procedure known as deflation , @xmath187 where @xmath188 is a matrix with columns @xmath189 and @xmath190 is a matrix with columns @xmath191
. the extracted projection directions can be computed ( following @xcite ) as @xmath192 .
similarly we deflate for the dual view @xmath193 where @xmath194 and compute the projection directions as @xmath195 where @xmath196 is a matrix with columns @xmath197 and @xmath198 has columns @xmath199 . the deflation procedure is illustrated in pseudocode in algorithm [ completealgo ] , for a detailed review on deflation we refer the reader to @xcite . + + checking each value of @xmath28 at each iteration is computationally impractical . in our experiments
we adopt the very simplistic strategy of picking the values of @xmath28 in numerical order @xmath200 .
clearly , there exists intermediate options of selecting a small subset of values at each stage , running the algorithm for each and selecting the best of this subset .
this and other extension of our work will be focused on in future studies .
input : data matrix @xmath151 , kernel matrix @xmath152 .
@xmath201 , @xmath202 @xmath203 @xmath204 = $ ] scca_algorithm@xmath205 @xmath194 @xmath189 @xmath191 @xmath206 @xmath207
in the following experiments we use two paired english - french and english - spanish corpora .
the english - french corpus consists of @xmath208 samples with @xmath209 english features and @xmath210 french features while the english - spanish corpus consists of @xmath211 samples with @xmath212 english features and @xmath213 spanish features .
the features represent the number of words in each language .
both corpora are pre - processed into a term frequency inverse document frequency ( tfidf ) representation followed by zero - meaning ( centring ) and normalisation .
the linear kernel was used for the dual view . the best test performance for the kcca regularisation parameter for the paired corpora
was found to be @xmath214 .
we used this value to ensure that kcca was not at a disadvantage since scca had no parameters to tune . in the following section
we demonstrate that the proposed approach for automatically determining the regularisation parameter ( hyper - parameter ) @xmath71 ( or alternatively @xmath79 ) is sufficient for our purpose .
the scca problem @xmath215 subject to @xmath51 can be simplified to a general lasso solver by removing the optimisation over @xmath33 , resulting in @xmath216 where , given our paired data , @xmath217 is the inner product between the query and the training samples and @xmath218 is the second paired data samples
. this simplified formulation is trivially solved by algorithm [ alg : scca ] by ignoring the loops that adapt @xmath33 .
the simplification of equation ( [ eqn : stacca ] ) allows us to focus on showing that @xmath71 is close to optimal , which is also true for @xmath79 , and therefore omitted .
+ + the hyper - parameters control the level of sparsity .
therefore , we test the level of sparsity as a function of the hyper - parameter value .
we proceed by creating a new document @xmath219 from a paired language that best matches our query and observe how the change in @xmath71 affects the total number of words being selected .
an ideal " @xmath71 would generate a new document , in the paired language , and select an equal number of words in the query s actual paired document . recall that the data has been mean corrected ( centred ) and therefore no longer sparse .
+ + we set @xmath71 to be in the range of @xmath220 $ ] with an increment of @xmath221 and use a leave - paired document - out routine for the english - french corpus , which is repeated for all @xmath208 documents . figure [ gif : larschange ] illustrates , for a single query , the effective change in @xmath71 on the level of sparsity .
we plot the ratio of the total number of selected words to the total number of words in the original document .
an ideal choice of @xmath71 would choose a ratio of @xmath56 ( the horizontal lines ) i.e. create a document with exactly the same number of words as the original document or in other words select a @xmath71 such that the cross would lie on the plot .
we are able to observe that the method for automatically choosing @xmath71 ( the vertical line ) is able to create a new document with a close approximation to the total number of words in the original document . in table
[ tab : rat ] we are able to show that the average ratio of total number of selected words for each document generated in the paired language is very close to the ideal " level of sparsity , while a non - sparse method ( as expected ) generates a document with an average of @xmath222 times the number of words from the original document . now that we have established the automatic setting of the hyper - parameters , we proceed in testing how ` good ' the selected words are in the form of a mate - retreiveal experiment .
.french - english corpus : the ratio of the total number of selected words to the actual total number of words in the paired test document , averaged over all queries . the optimal average ratio if we always generate an ` ideal ' document is @xmath56 . [ cols="^,^",options="header " , ] [ tab : rat ] our experiment is of mate - retrieval , in which a document from the test corpus of one language is considered as the query and only the mate document from the other language is considered relevant . in the following experiments the results are an average of retrieving the mate for both english and french ( english and spanish ) and have been repeated @xmath223 times with a random train - test split .
+ + we compute the mate - retrieval by projecting the query document as well as the paired ( other language ) test documents into the learnt semantic space where the inner product between the projected data is computed .
let @xmath224 be the query in one language and @xmath225 the kernel matrix of the inner product between the second language s testing and training documents @xmath226 the resulting inner products @xmath227 are then sorted by value .
we measure the success of the mate - retrieval task using average precision , this assesses where the correct mate within the sorted inner products @xmath227 is located .
let @xmath228 be the index location of the retrieved mate from query @xmath229 , the average precision @xmath230 is computed as @xmath231 where @xmath232 is the number of query documents .
+ + we start by giving the results for the english - french mate - retrieval as shown in figure [ fig : enfr ] .
the left plot depicts the average precision ( @xmath233 standard deviation ) when @xmath234 documents are used for training and the remaining @xmath235 are used as test queries .
the right plot in figure [ fig : enfr ] gives the average precision ( @xmath233 standard deviation ) when @xmath236 documents are used for training and the remaining @xmath237 for testing .
it is interesting to observe that even though scca does not learn the common semantic space using all the features ( average plotted in figure [ fig : enfrp ] ) for either ml primal or dual views ( although scca will use full dual features when using the full number of projections ) its error is extremely similar to that of kcca and in fact converges with it when a sufficient number of projections are used .
it is important to emphasise that kcca uses the full number of documents ( @xmath234 and @xmath236 ) and the full number of words ( an average of @xmath238 for both languages ) to learn the common semantic space .
for example , following the left plot in figure [ fig : enfr ] and the additional plots in figure [ fig : enfrp ] we are able to observe that when @xmath239 projections are used kcca and scca show a similar error . however , scca uses approximately @xmath240 words and @xmath241 documents to learn the semantic space , while kcca uses @xmath238 words and @xmath234 documents . + +
the second mate - retrieval experiment uses the english - spanish paired corpus . in each run
we randomly split the @xmath242 samples into @xmath236 training and @xmath243 testing paired documents .
the results are plotted in figure [ fig : enes ] where we are clearly able to observe scca outperforming kcca throughout .
we believe this to be a good example of when too many features hinder the learnt semantic space , also explaining the difference in the results obtained from the english - french corpus as the number of features are significantly smaller in that case .
the average level of scca sparsity is plotted in figure [ fig : enesp ] . in comparison
to kcca which uses all words ( @xmath244 ) scca uses a maximum of @xmath245 words .
+ + the performance of scca , especially in the latter english - spanish experiment , shows that we are indeed able to extract meaningful semantics between the two languages , using only the relevant features . despite these already impressive results
our intuition is that even better results are attainable if the hyper - parameters would be tuned to give optimal results .
the question of hyper - parameter optimality is left for future research .
although , it seems that the main gain of scca is sparsity and interpretability of the features .
despite being introduced in @xmath246 , cca has proven to be an inspiration for new and continuing research . in this paper
we analyse the formulation of cca and address the issues of sparsity as well as convexity by presenting a novel scca method formulated as a convex least squares approach .
we also provide a different perspective of solving cca by using a ml primal - dual formulation which focuses on the scenario when one is interested in ( or limited to ) a ml - primal representation for the first view while having a ml - dual representation for the second view . a greedy optimisation algorithm is derived . + +
the method is demonstrated on a bi - lingual english - french and english - spanish paired corpora for mate retrieval .
the true capacity of scca becomes visible when the number of features becomes extremely large as scca is able to learn the common semantic space using a very sparse representation of the ml primal - dual views .
+ + the papers raison dtre was to propose a new efficient algorithm for solving the sparse cca problem .
we believe that while addressing this problem new and interesting questions which need to be addressed have surfaced * how to automatically compute the hyperparameters @xmath47 values so to achieve optimal results ? * how do we set @xmath28 for each @xmath182 when we wish to compute less than @xmath247 projections ? * extending scca to a ml primal - primal ( ml dual - dual ) framework .
we believe this work to be an initial stage for a new sparse framework to be explored and extended .
david r. hardoon is supported by the epsrc project le strum , ep - d063612 - 1 .
we would like to thank zakria hussain and nic schraudolph for insightful discussions .
this publication only reflects the authors views .
dhanjal , c. , gunn , s. r. & shawe - taylor , j. ( 2006 ) .
sparse feature extraction using generalised partial least squares . in _ proceedings of the ieee international workshop on machine learning for signal processing _ ( pp . 2732 ) .
friman , o. , borga , m. , lundberg , p. & knutsson , h. ( 2001a ) . a correlation framework for functional mri data analysis . in _ proceedings of the 12th scandinavian conference on image analysis_. bergen , norway .
hardoon , d. r. & shawe - taylor , j. ( 2003 ) . for different level precision in content - based image retrieval . in _ proceedings of third international workshop on content - based multimedia indexing_. irisa , rennes , france .
sriperumbudur , b. k. , torres , d. & lanckriet , g. ( 2007 ) .
sparse eigen methods by d.c . programming . in _ in proceedings of 2nd international conference on machine learning _
( pp . 831838 ) .
morgan kaufmann , san francisco , ca .
szedmak , s. , bie , t. d. & hardoon , d. r. ( 2007 ) .
a metamorphosis of canonical correlation analysis into multivariate maximum margin learning . in _
15th european symposium on artificial neural networks ( esann)_. | we present a novel method for solving canonical correlation analysis ( cca ) in a sparse convex framework using a least squares approach .
the presented method focuses on the scenario when one is interested in ( or limited to ) a primal representation for the first view while having a dual representation for the second view .
sparse cca ( scca ) minimises the number of features used in both the primal and dual projections while maximising the correlation between the two views .
the method is demonstrated on two paired corpuses of english - french and english - spanish for mate - retrieval .
we are able to observe , in the mate - retreival , that when the number of the original features is large scca outperforms kernel cca ( kcca ) , learning the common semantic space from a sparse set of features . |
quantum interference between different excitation path ways is an intriguing phenomenon in quantum optics , laser and atomic physics .
notable examples are electromagnetically induced transparency ( eit ) @xcite , multi - wave mixing @xcite etc . among these ,
an interesting category is the phase - sensitive process formed by closed - loop interactions , where the phases of the optical fields can dramatically change the steady state of the atoms and the optical susceptibility .
there has been a large amount of work in this field since the 1980s @xcite , and the interests boosted after the experimental demonstration of phase - sensitive eit @xcite .
applications of phase sensitive processes include large nonlinearity @xcite , slow and fast light @xcite , and optical switch @xcite etc . to the best of our knowledge ,
it is a tradition to use oscillating electromagnetic fields , i.e. , either all optical fields @xcite , or a combination of microwave fields and optical fields @xcite , to form a closed loop . in this paper
, we propose to use a static magnetic field to close the interaction loop .
this approach is suitable as long as the ground states can be coupled by the magnetic field .
for example , ground states formed by superpositions of zeeman sublevels can be mixed if their spin orientations are not parallel to the external magnetic field . in this case
, the magnetic field can in effect coherently couple the two ground states through larmor precession .
for instance , as we will show below , in a lambda system , without the magnetic field the system is a normal eit configuration , whose steady state susceptibility is not influenced by the optical phases ; with the magnetic field , a closed loop is formed and the field absorption becomes dependent on the relative phase .
similar idea also applies to a double - lambda system .
although its level configuration naturally allows a closed - cycle four - wave - mixing ( fwm ) which is already phase sensitive , applying a magnetic field can still alter the degree of phase sensitivity , as also shown below both theoretically and experimentally . adding a magnetic field as a new knob to the double - lambda system with zeeman sublevel ground states
might be particularly useful , since such a system has attracted considerable amount of interests recently for use in light squeezing @xcite , light entanglement @xcite and amplified slow and stored light @xcite .
the following advantages of this double - lambda system are probably responsible for its popularity : ( 1 ) it corresponds to the energy configuration of the d1 line of the alkali atoms , and can be easily implemented by experiments .
( 2 ) its relevance to magnetic fields allows various magnetometer schemes to be realized @xcite .
( 3 ) laser fields with only one frequency are sufficient to address this system due to the zeeman ground states , which greatly simplifies experiments . ( 4 ) the long lifetime of the ground states enables coherence - enhanced nonlinear optics @xcite . this paper is organized as follows . in section 2
, we introduce the system theoretically , and lay out key steps in our numerical simulation . then in section 3
, we describe our experimental setup and measurement methods . in section 4 , we present the experiment results along with corresponding numerical results , which mainly include : ( i ) the presence of a magnetic field increases the dependence of the optical absorption on the relative phase between laser fields ; ( ii ) when the magnetic field reverses and the relative phase changes by @xmath3 , the optical responses remain the same ; ( iii ) in the dynamical regime , the probe pulse can experience either slow light or fast light depending on the magnetic field and the relative phase . finally in section 5 , we conclude .
the double - lambda system under consideration is shown in fig .
[ fwmpic ] , where ( a ) and ( b ) use different ground state basis . @xmath4 and @xmath5 are degenerate zeeman sublevels ( with opposite magnetic quantum numbers ) and are eigenstates of the atomic angular momentum component along the direction of the static magnetic field . @xmath6 and @xmath7 are the two hyperfine excited states separated by @xmath8 .
the optical fields are nearly resonant with the lower excited state .
we assume that the selection rules and c - g coefficients determine that a sigma plus ( minus ) field @xmath9 ( @xmath10 ) can couple @xmath4 ( @xmath5 ) to both @xmath6 and @xmath7 , with rabi frequency @xmath11 ( @xmath12 ) and @xmath11 ( @xmath13 ) respectively . since in the experiment we use two orthogonally linearly polarized laser fields , a weak probe @xmath14 and a strong control @xmath15 , it is more convenient to use the linear basis in fig .
[ fwmpic ] ( b ) , where @xmath16 , and @xmath17 . in this basis , the probe field couples @xmath18 and @xmath19 transitions with the same rabi frequency @xmath20 , and the control field couples @xmath21 and @xmath22 transitions with the same rabi frequency @xmath23 .
when a magnetic field along the light propagation direction is applied , @xmath4 and @xmath5 have zeeman shift @xmath24 , @xmath25 respectively ; and in the linear basis , @xmath26 and @xmath27 are coherently coupled by the magnetic field with an effective rabi frequency equal to @xmath28 . here , @xmath29 with @xmath30 the bohr magneton , @xmath31 the land @xmath31 factor , and @xmath32 the magnetic quantum number . and @xmath33 , and ( b ) in the linear basis where the two linearly polarized fields are @xmath34 and @xmath35 . here , @xmath16 , @xmath17
due to the selection rules and c - g coefficients of rb , @xmath36 couples @xmath37 both with rabi frequency @xmath38 , and @xmath33 couples @xmath39 with rabi frequency @xmath40 , @xmath12 respectively . similarly , @xmath35 couples @xmath41 both with rabi frequency @xmath42 , and @xmath34 couples @xmath43 both with rabi frequency @xmath44.,scaledwidth=80.0% ] the hamiltonian of the system takes the following form in the circular basis ( fig . [ fwmpic ] ( a ) ) : @xmath45 here , @xmath46 and @xmath47 , with @xmath48 the transition dipole moment between level @xmath49 and @xmath50 .
the two circularly polarized laser fields @xmath36 and @xmath33 are connected to the linearly polarized fields @xmath35 and @xmath34 by @xmath51 various decay channels can be taken into account by the master equation @xmath52+\hat{\gamma}_{exc}-\hat{\gamma}_{rel } \label{masterequation}\ ] ] where @xmath53 and @xmath54 are excitation matrix and relaxation matrix .
we assume that both excited states have a decay rate @xmath55 , and in the circular basis , the ground states population difference decay rate is @xmath56 and the coherence decay rate is @xmath57 . in our model ,
when cold atoms are considered , we set @xmath58 mhz , and when a warm atomic vapor is considered , we set @xmath59 mhz to take into account of doppler broadening .
we consider the steady state of the atoms by setting @xmath60 , and then the propagation of the two circular polarized fields are described by the maxwell equation in the slowly varying amplitude approximation : @xmath61 where @xmath62 is the coupling constant , with @xmath63 the laser frequency , @xmath64 the dipole moment of the transitions @xmath65 and @xmath66 ( assumed to be equal ) , and @xmath67 the atomic density .
numerically , we can solve for the fields complex amplitudes @xmath68 and @xmath69 at the vapor cell output ( @xmath70 the cell length ) , and then have the probe field s output power as : @xmath71 where @xmath72 is the relative phase between @xmath68 and @xmath69 .
it can be seen that the probe field s transmission is related to both the amplitudes and relative phase of @xmath11 and @xmath13 at the output .
we performed experiments ( schematically shown in fig . [ setup ] ) using a cylindrical paraffin coated vapor cell containing isotropically enriched @xmath2rb .
the cell ( 2.5 cm diameter , 7.5 cm length ) was heated by a blown - air oven residing in a solenoid for magnetic field control , all within a four - layer magnetic shield . tuned close to the @xmath2rb d1 line @xmath73 @xmath74 @xmath75 transition , the laser passed through a polarized beam splitter ( pbs ) to generate the linearly polarized pump and the probe field .
then they each went through an acoustic - optical modulator ( aom ) for independent power control and pulse shaping .
the two identical aoms shifted the laser frequency by 80 mhz , and the two 80 mhz rf sources were phase locked to ensure the nearly perfect phase coherence between the probe and the control .
the first order beams from the aoms were then combined by another pbs and directed into the vapor cell .
at the cell output , a high quality pbs separated the probe and the pump , and they were independently detected by two photo - detectors ( pd ) with tunable gains .
one of the key elements of this experiment was to have good control on the relative phase between the control and the probe , which was tuned by varying the voltage on the piezoelectric transducer ( pzt ) attached to the mount of a mirror right before the combining pbs . to reduce fluctuations of the relative phase ,
the splitting and combining optics were covered by a box to minimize air flow .
the maximal phase change the pzt could provide was about @xmath76 . to measure the relative phase at the cell input
, we used a flip mirror to direct the beam to a half wave plate followed by a pbs to interfere the control and the probe .
also , it was ensured that this calibration optics at the side gave the same phase with a temporary calibration optics at the cell output ( added a half wave plate before the detection pbs ) while the entire shield was pushed aside . to check the phase stability ,
the interference signal was taken before and after each absorption measurement for a particular phase .
in all the experiments reported below , the control field and probe field power were about 225 @xmath77w and 15 @xmath77w respectively ( unless otherwise stated ) , and both of them were 3.5 mm in diameter .
we used a single mode fiber at the laser output to ensure good beam profile .
the overlap of the two beams after combination were checked by a beam profiler at several locations along the light stream .
the temperature of the cell was about 55 @xmath78 , corresponding to an atomic density of about @xmath79 @xmath80 and an optical depth about 15 @xcite .
in this section , we discuss the effect of the magnetic field on the phase sensitivity of the double - lambda system as shown in fig
. 1(b ) .
let s start from a simpler case which we call the cold atom regime , where the spacing ( @xmath81 mhz ) between the two excited states is much larger than the linewidth ( @xmath82 mhz ) of the excited states .
[ phasesweepcoldatom ] shows the calculated probe transmission ( normalized to the input value ) vs the relative phase between the probe and control , for three magnetic field values ( represented by @xmath83 ) .
since these curves have a period of @xmath84 , only one period is plotted .
it can be seen that , when the magnetic field is off , the probe transmission is insensitive to the phase .
this is because the negligible effect of the upper excited state effectively reduces the system to a normal lambda system , which is not a closed loop without the magnetic field .
when the magnetic field is on , the loop is formed .
indeed , when @xmath83 increases to 10 hz and then to 80 hz , the transmission becomes increasingly sensitive to the phase . in the cold atom regime ,
where the excited state decay rate @xmath58 mhz .
( a ) the black solid , red dashdotted and blue dash curves are for @xmath85 hz , 10 hz and 80 hz respectively .
( b ) the black solid , red dashdotted and blue dash curves are for @xmath85 hz , @xmath86 hz and @xmath87 hz respectively .
simulation parameters : @xmath88 khz , @xmath89 khz , @xmath90 hz , and the optical depth was 0.15 .
changes in the optical depth do not affect the main features of the curves as described in the text.,scaledwidth=80.0% ] .
the black dashed , black solid and grey solid curves are for @xmath85 hz , 40 hz and @xmath91 hz respectively .
experimental conditions are in section 3 .
( b ) corresponding theoretical results where the excited state decay rate @xmath59 mhz .
other simulation parameters : @xmath92 hz , @xmath93 hz , @xmath94 khz , @xmath95 khz , and the optical depth was 15.,scaledwidth=80.0% ] fig .
[ phasesweepcoldatom](b ) shows the calculated probe transmission for magnetic fields opposite to that in fig .
[ phasesweepcoldatom](a ) .
it was found that two curves with opposite @xmath83 values coincide if we translate one of them by @xmath3 on the x - axis .
this phenomenon can be explained using the linear basis ( fig .
[ fwmpic](b ) ) . in each lambda loop , for example , in the cycle of @xmath96 , the coherence term @xmath97 is determined by the product of @xmath98 and @xmath28 which remains unchanged if the sign of @xmath28 and @xmath98 reverses simultaneously .
indeed , we found that the master equation remains valid when the following substitution is made : @xmath99 , @xmath100 , @xmath101 , @xmath102 , @xmath103 , @xmath104 , @xmath105 , @xmath106 , @xmath107 ( @xmath108 ) , which indicates these are two sets of identical solutions to the system .
in other words , the probe field susceptibility should be the same if we change the sign of the magnetic field and change the relative phase from @xmath109 to @xmath110 simultaneously . to verify above predictions , we measured the probe transmission vs the relative phase for @xmath85 , @xmath111 hz ( fig . [ phasesweepwarmatom ] ) .
the transmission was normalized to the off - resonant transmission of the probe when the control was absent .
we can see that , in contrast to the cold atom regime , the phase sensitivity was already very pronounced without the magnetic field .
this is because that doppler broadening makes the effects of the upper excited state not negligible , and thus the fields form a closed loop four - wave - mixing ( even without the magnetic field ) which is phase sensitive . when @xmath28 was nonzero , the phase sensitivity increased , as in the simulation for cold atoms .
also , the two curves for @xmath112 hz can almost overlap when one curve is shifted by @xmath3 along the x - axis , consistent with above theoretical analysis . to account for the shape difference of the measured curves compared to the cold atom case , we performed simulations with @xmath59 mhz ( fig .
[ phasesweepwarmatom](b ) ) and found good agreement with our experiment .
the experiment results show that a double - lambda scheme has distinct phase dependence than a single lambda scheme . for several relative phases .
( a ) measured results , with experimental conditions stated in section 3 .
( b ) calculated results , with identical simulation parameters as used in fig .
[ phasesweepwarmatom](b).,scaledwidth=80.0% ] in this section , we investigate the combined effects of the magnetic field and the relative phase on the probe s transmission and its group velocity .
we first measured the probe transmission vs @xmath28 for various relative phase @xmath109 .
as shown in fig . [ 55](a ) , the curves combine features of absorption and dispersion plots , which can be interpreted using the circular basis .
( 6 ) shows that the probe transmission carries the information of both the absorption and refractive index ( phase ) of the circular laser beams .
in particular , the first two terms of eq .
( 6 ) represent the total transmission of the two circular fields , and should give a normal lorentzian eit spectra when @xmath83 is swept .
as for the interference term of eq .
( 6 ) , @xmath113 can be written as @xmath114 if @xmath115 is small . here
@xmath116 is the input relative phase between the two circular beams , and @xmath115 is the accumulated phase difference which has a dispersive feature . also , it is clearly seen from fig . [ 55 ] that , the two curves with relative phase @xmath109 and @xmath110 are mirror images of each other . this again verifies that the probe transmission remains unchanged if we change the sign of @xmath28 and add @xmath3 to the relative phase simultaneously .
[ 55](b ) is the numerically calculated results , which agree well with the experiment .
remaining discrepancy between the experiment and the simulation results is mainly due to that in our simplified model , we did not include the effects of atomic motion in the coated vapor cell @xcite , and doppler broadening was not strictly taken into account by integrating over all velocities .
next , we studied the dynamic behavior of the probe field , and found that the refractive index and the group velocity can be also controlled by the magnetic field and the relative phase .
we generated the probe pulse field via programming the aom s driving field by labview .
for optimal slow and fast light effects , the full width half maximum ( fwhm ) of the probe pulse was chosen to be about 5 ms .
the continuous control field power was 225 @xmath77w and the peak power of the probe pulse was 15 @xmath77w .
when @xmath28 was set at @xmath117 hz , we could turn the slow light into fast light by changing the relative phase from about 1.75 @xmath3 to 0 ( fig .
[ fastslowlight](a ) ) .
the base of the fast light or slow light curve was the result of the optical rotation of the control field due to the magnetic field .
on the other hand , if the relative phase was set to be 0 , we could also turn the slow light into fast light by changing @xmath28 from 0 hz to @xmath117 hz ( fig .
[ fastslowlight](b ) ) . the fractional delay and advance were about 25@xmath118 and 10@xmath118 respectively .
( b ) relative phase @xmath119.,scaledwidth=80.0% ] vs the probe frequency , with the pump frequency fixed . ( a ) relative phase manipulation for @xmath120 hz . ( b )
magnetic field manipulation for the relative phase @xmath119 .
simulation parameters are the same as in fig .
[ phasesweepwarmatom](b).,scaledwidth=80.0% ] to explain such phenomena , we calculated the refractive index experienced by the probe field while fixing the control field frequency and sweeping the probe s one - photon detuning @xmath121 . for this calculation
, we rewrote the hamiltonian of the system in the linear basis as depicted in fig .
[ fwmpic](b ) , and computed the real part of @xmath122 which is proportional to the refractive index for the probe . according to the group velocity formula @xmath123 the slope of the refractive index spectra of the probe near @xmath124 hz in fig .
[ refractiveindex ] determines the group velocity of the probe pulse .
[ refractiveindex](a ) shows that when @xmath120 hz , the slope near @xmath124 changes from positive to negative when the phase changes from 1.75 @xmath3 to 0 .
[ refractiveindex](b ) shows that when the relative phase @xmath119 , the slope near @xmath124 changes from positive to negative when @xmath28 varies from 0 hz to @xmath117 hz .
this verifies that switching between slow and fast light can be realized by either adjusting the magnetic field or the relative phase .
in conclusion , we demonstrated a new method to manipulate the phase sensitivity of a double - lambda system . a magnetic field coupling the two ground states can function as an effective oscillating electromagnetic field , and thus can either close an otherwise open interaction loop , rendering the system phase - sensitive , or alter the phase sensitivity of an initially phase - dependent system .
we have theoretically and experimentally verified that a static magnetic field can drastically influence the absorption and refractive index of an optical field and its dependence on the relative phase .
this work should be useful for group velocity manipulation , nonlinear optics and magnetometery .
we are grateful to xiangming hu and lei feng for useful discussions .
this work was supported by the nbrpc ( 973 program grants no .
2011cb921604 and no .
2012cb921604 ) , the nnsfc ( grant no .
61078013 ) , and fudan university . | we study the effect of a dc magnetic field on the phase sensitivity of a double - lambda system coupled by two laser fields , a probe and a pump .
it is demonstrated that the gain and the refractive index of the probe can be controlled by either the magnetic field or the relative phase between the two laser fields .
more interestingly , when the system reduces to a single - lambda system , turning on the magnetic field transforms the system from a phase - insensitive process to a phase - sensitive one . in the pulsed - probe regime , we observed switching between slow and fast light when the magnetic field or the relative phase was adjusted .
experiments using a coated @xmath0rb vapor cell produced results in good agreement with our numerical simulation .
this work provides a novel and simple means to manipulate phase sensitive electromagnetically - induced - transparency or four - wave mixing , and could be useful for applications in quantum optics , nonlinear optics and magnetometery based on such systems .
99 k. j. boiler , a. imamoglu , and s. e. harris , `` observation of electromagnetically induced transparency , '' 25932596 ( 1991 ) .
m. d. lukin , `` _ colloquium _ : trapping and manipulating photon states in atomic ensembles , '' 457472 ( 2003 ) .
m. fleischhauer , a. imamoglu , and j. p. marangos , `` electromagnetically induced transparency : optics in coherent media , '' 633673 ( 2005 ) .
y. zhang and m. xiao , _ multi - wave mixing processes _
( springer , berlin , 2009 ) .
s. j. buckle , s. m. barnett , p. l. knigt , m. a. lauder , and d. t. pegg , `` atomic interferometers : phase - dependence in multilevel atomic transitions , '' opt .
acta * 33 , * 11291140 ( 1986 ) .
s. p. krinitzky and d. t. pegg , `` coherent irradiation of multilevel atoms in branched and cyclic configurations , '' 403406 ( 1986 ) .
e. a. korsunsky , n. leinfellner , a. huss , s. baluschev , and l. windholz , `` phase - dependent electromagnetically induced transparency , '' 23022305 ( 1999 ) .
m. sahrai , a. maleki , r. hemmati , and m. mahmoudi , `` transient dispersion and absorption in a v - shaped atomic system , '' eur .
phys .
j. d * 56 , * 105112 ( 2010 ) .
t. m. preethi , m. manukumara , k. asha , j. vijay , d. a. roshi , and a. narayanan , `` phase - sensitive microwave optical double resonance in an n system , '' euro .
. lett .
* 95 , * 34005 ( 2011 ) .
j. dimitrijevi and d. arsenovi , `` simple analytical expressions for the analysis of the phase - dependent electromagnetically induced transparency in a double-@xmath1 atomic scheme , '' phys .
scr .
* t149 , * 014007 ( 2012 ) .
d. bortman - arbiv , a. d. wilson - gordon , and h. friedmann , `` phase control of group velocity : from subluminal to superluminal light propagation , '' 043818 ( 2001 ) .
h. sun , h. guo , y. bai , d. han , s. fan , and x. chen , `` light propagation from subluminal to superluminal in a three - level lambda - type system , '' phys .
lett . a * 335 , * 6875 ( 2005 ) .
d. jafari , m. sahrai , h. motavalli , and m. mahmoudi , `` phase control of group velocity in a dielectric slab doped with three - level ladder - type atoms , '' 063811 ( 2011 ) .
j. kou , r. wan , z. kang , l. jiang , l. wang , y. jiang , and j. gao , `` phase - dependent coherent population trapping and optical switching , '' 063807 ( 2011 ) .
w. maichen , f. renzoni , i. mazets , e. korsunsky , and l. windholz , `` transient coherent population trapping in a closed loop interaction scheme , '' 34443448 ( 1996 ) .
e. a. korsunsky and d. v. kosachiov , `` phase - dependent nonlinear optics with double-@xmath1 atoms , '' 49965009 ( 1999 ) .
x. fan , z. liu , y. liang , k. jia , and d. tong , `` phase control of probe response in a doppler - broadened n - type four - level system , '' 043805 ( 2011 ) .
a. f. huss , e. a. korsunsky , and l. windholz , `` phase control of electromagnetically induced transparency in a double-@xmath1 system , '' 141155 ( 2002 ) .
h. li , v. a. sautenkov , y. v. rostovtsev , g. r. welch , p. r. hemmer , and m. o. scully , `` electromagnetically induced transparency controlled by a microwave field , '' 023820 ( 2009 ) .
d. kosachiov , b. matisov , and yu .
rozhdestvensky , `` coherent population trapping : sensitivity of an atomic system to the relative phase of exciting fields , '' 209212 ( 1991 ) .
m. s. shahriar and p. r. hemmer , `` direct excitation of microwave - spin dressed states using a laser - excited resonance raman interaction , '' 18651868 ( 1990 ) .
e. e. mikhailov and i. novikova , `` low - frequency vacuum squeezing via polarization self - rotation in rb vapor , '' 12131215 ( 2008 ) .
s. barreiro , p. valente , h. failache , and a. lezama , `` polarization squeezing of light by single passage through an atomic vapor , '' 033851 ( 2011 ) .
w. wasilewski , t. fernholz , k. jensen , l. s. madsen , h. krauter , c. muschik , and e. s. polzik , `` generation of two - mode squeezed and entangled light in a single temporal and spatial mode , '' 1444414457 ( 2009 ) .
i. novikova , d. f. phillips , and r. l. walsworth , `` slow light with integrated gain and large pulse delay , '' 173604 ( 2007 ) .
w. wasilewski , k. jensen , h. krauter , j. j. renema , m. v. balabas , and e. s. polzik , `` quantum noise limited and entanglement - assisted magnetometry , '' 133601 ( 2010 ) .
b. patton , o. o. versolato , d. c. hovde , e. corsini , j. m. higbie , and d. budker , `` a remotely interrogated all - optical @xmath2rb magnetometer , '' 083502 ( 2012 ) .
n. b. phillips , a. v. gorshkov , and i. novikova , `` light storage in an optically thick atomic ensemble under conditions of electromagnetically induced transparency and four - wave mixing , '' 063823 ( 2011 ) .
m. klein , m. hohensee , d. f. phillips , and r. l. walsworth , `` electromagnetically induced transparency in paraffin - coated vapor cells , '' 013826 ( 2011 ) .
i. novikova , r. walsworth and y. xiao , `` electromagnetically induced transparency - based slow and stored light in warm atoms , '' laser photonics rev .
* 6 , * 333353 ( 2012 ) .
z. xu , w. qu , r. gao , x. hu , and y. xiao , `` linewidth of electromagnetically induced transparency under motional averaging in a coated vapor cell , '' chin .
phys .
b * 22 , * 033202 ( 2013 ) . |
in this paper we are interested in relative fractal drums @xmath0 in which the set @xmath1 has degenerated to the point at infinity . in short ,
a relative fractal drum @xmath0 generalizes the notion of a bounded subset of @xmath2 and is defined as an ordered pair of subsets of @xmath2 where @xmath1 is nonempty and @xmath3 is of finite @xmath4-dimensional lebesgue measure satisfying a mild technical condition .
the lapidus ( or distance ) zeta function of @xmath0 is then defined as the lebesgue integral @xmath5 for all @xmath6 such that @xmath7 is sufficiently large , where @xmath8 denotes the euclidean distance from @xmath9 to @xmath1 .
its main property is that the _ abscissa of convergence _ @xmath10 of @xmath11 coincides with the upper box dimension of @xmath0 , i.e. , @xmath12 . in other words ,
the integral converges absolutely and defines a holomorphic function in the open half - plane @xmath13 . for the study of relative fractal drums , their corresponding fractal zeta functions and the general higher - dimensional theory of complex dimensions
see @xcite along with the survey articles @xcite .
this higher - dimensional theory generalizes the well known theory of geometric zeta functions for fractal strings and their complex dimensions developed by michel l. lapidus and his collaborators in the last two decades ( see @xcite and the relevant references therein ) . in the case
when the set @xmath1 degenerates to the point at infinity , we will denote this new kind of relative fractal drum with @xmath14 . in this case
it is clear that the fractal properties of such a relative fractal drum will depend only on the set @xmath3 .
we will extend the notions of minkowski content and box dimension for such relative fractal drums and define a new class of lapidus zeta functions associated to them .
furthermore , it will be shown that this new class of lapidus zeta functions has analogous properties as in the case of ordinary relative fractal drums and hence , provides an analytic approach to the study of fractality of unbounded sets .
the motivation to study the fractal properties of unbounded sets comes from a variety of sources .
in particular , the notion of `` unbounded '' or `` divergent '' oscillations appears in problems in oscillation theory ( see , e.g. @xcite ) , automotive industry ( see , e.g. , @xcite ) , civil engineering ( see , e.g , @xcite ) and mathematical applications in biology ( see , e.g. , @xcite ) .
unbounded ( divergent ) oscillations are oscillations the amplitude of which increases with time .
for instance , the oscillations of an airplane that has positive static stability but negative dynamic stability is an example of divergent oscillations that appears in aerodynamics ( see , e.g. @xcite ) . furthermore , unbounded domains themselves are also interesting in the theory of elliptic partial differential equations .
more precisely , the question of solvability of the dirichlet problem for quasilinear equations in unbounded domains is addressed in @xcite and ( * ? ? ?
* section 15.8.1 ) . also , unbounded domains can be found in other aspects of the theory of partial differential equations ;
see , for instance @xcite and @xcite .
let @xmath3 be a lebesgue measurable subset of the @xmath4-dimensional euclidean space @xmath2 of finite lebesgue measure , i.e. , @xmath15 .
firstly , we will introduce a new notation for the sake of brevity , namely , @xmath16 where @xmath17 and @xmath18 denotes the complement of the open ball of radius @xmath19 centered at @xmath20 .
we introduce the _ tube function of @xmath3 at infinity _ by @xmath21 for @xmath17 where @xmath22 denotes the @xmath4-dimensional lebesgue measure and we will be interested in the asymptotic properties of this function when @xmath23 .
furthermore , for any real number @xmath24 we define the _
upper @xmath24-dimensional minkowski content _ of @xmath3 _ at infinity _ @xmath25 and , analogously , by taking the lower limit in as @xmath26 , we define the _
lower @xmath24-dimensional minkowski content _ of @xmath3 _ at infinity _ denoted by @xmath27 .
it is easy to see that the above definition implies the existence of a unique @xmath28 such that @xmath29 for @xmath30 and @xmath31 for @xmath32 and similarly for the lower minkowski content .
the value @xmath33 is called the _ upper box dimension of @xmath3 at infinity _ and we denote it with @xmath34 .
similarly as in the case of ordinary relative fractal drums , we have @xmath35 analogously , by using the lower minkowski content of @xmath3 at infinity , we define the _ lower box dimension of @xmath3 at infinity _ and denote it by @xmath36 and the analog of is also valid in this case .
of course , if the upper and lower box dimensions coincide , we define the _ box dimension _ of @xmath3 _ at infinity _ and denote it with @xmath37 . in the case
when the upper and lower minkowski content at infinity coincide we define the _
@xmath24-dimensional minkowski content _ of @xmath3 _ at infinity _ and denote it with @xmath38 . furthermore , in the case when @xmath39 for some @xmath28 ( which implies that @xmath40 ) , we say that @xmath3 is _ minkowski nondegenerate at infinity_. we say that @xmath3 is _ minkowski measurable at infinity _ if it is minkowski nondegenerate at infinity and its lower and upper minkowski content at infinity coincide .
[ < = -n ] let @xmath3 be a lebesgue measurable subset of @xmath2 of finite lebesgue measure .
then @xmath41 , i.e. , the upper and lower box dimensions of @xmath3 at infinity are always negative , that is , less than or equal to @xmath42 . from the definitions and and the fact that @xmath15 we have that @xmath43 as @xmath44 which implies that if @xmath45 , then @xmath31 .
[ = -n ] intuitively the conclusion of proposition [ < = -n ] is expected , since @xmath3 having finite lebesgue measure implies that it must have a certain flatness property relative to infinity .
( compare with the notion of flatness introduced in @xcite . ) furthermore , if @xmath46 , then it follows from the definition that @xmath47 and , consequently , @xmath3 must be minkowski degenerate at infinity .
the next two results about the monotonicity are simple consequences of the definitions involved .
[ subset_inf ] let @xmath48 be two lebesgue measurable sets and @xmath49 .
then for any real number @xmath24 we have that @xmath50 and @xmath51 [ subset_inf_dim ] let @xmath48 be two lebesgue measurable sets with @xmath49 . then , @xmath52 and @xmath53 let us now take a look at a few examples .
[ omega(a , b ) ] let @xmath54 and @xmath55 be fixed and define @xmath56 , @xmath57 and @xmath58 we define @xmath59 that is , as a union of countably many intervals @xmath60
. [ 1dimexample ] for the set @xmath61 defined by we have that @xmath62 firstly , we observe that for @xmath63 large enough the intervals @xmath64 become disjoint , i.e. , @xmath65 . as we see , @xmath61 is a union of intervals that `` escape '' to infinity and @xmath66 .
let us compute the box dimension and minkowski content of @xmath61 at infinity .
for @xmath17 let @xmath67 be such that for every @xmath68 it holds that @xmath69 , that is , @xmath70 .
now we fix @xmath19 large enough so that the intervals @xmath71 are disjoint for @xmath72 . from this
, we have @xmath73 with @xmath74 being the characteristic function of @xmath3 .
this implies the following estimate @xmath75 furthermore , using the integral criterion @xmath76 for estimating the sum , we have @xmath77 finally , by using the fact that @xmath78 , we conclude that @xmath79 which implies that @xmath80 is different from @xmath20 and @xmath81 if and only if @xmath82 , i.e. , if holds .
as we can see , the minkowski content in the above case depends only on the parameter @xmath83 , i.e. , the rate at which @xmath61 `` escapes '' to infinity is not relevant for it .
furthermore , by changing the values of parameters @xmath84 and @xmath83 , we can obtain any prescribed value in @xmath85 for @xmath86 .
moreover , we have that @xmath87 and @xmath88 as @xmath89 .
[ standardniprim ] for @xmath90 let @xmath91 .
then we have that @xmath92 let @xmath93 and let @xmath94 be such that @xmath95 then we have @xmath96 which implies that @xmath97 furthermore , from we have that @xmath98 as @xmath99 , and we conclude that holds .
note that @xmath100 and @xmath101 as @xmath102 .
next we will prove a useful lemma which states that the box dimension and minkowski measurability at infinity are independent on the choice of the norm on @xmath2 in a sense that we can replace the ball @xmath103 in the definition of the minkowski content at infinity with a ball in any other norm on @xmath2 .
more precisely , let @xmath104 be another norm on @xmath2 .
we denote by @xmath105 the open ball of radius @xmath19 around @xmath20 in the new norm ; define the associated upper minkowski content at infinity @xmath106 and analogously , @xmath107 and @xmath108 .
[ eqnorms ] let @xmath109 with @xmath15 and assume that two norms , @xmath22 and @xmath104 , are given on @xmath2 , i.e. , there are @xmath110 such that @xmath111
. then , for any @xmath112 we have @xmath113 and analogously for the corresponding lower minkowski contents .
from @xmath114 we have that @xmath115 for any @xmath17 and , consequently , @xmath116 taking the upper limit as @xmath44 , we obtain the first statement of the lemma . the second one is obtained by taking the lower limit instead of the upper .
let @xmath3 be an arbitrary lebesgue measurable subset of @xmath2 with finite @xmath4-dimensional lebesgue measure .
then @xmath117 the upper and lower box dimensions of @xmath3 at infinity do not depend on the choice of the norm on @xmath2 in which we measure the neighborhood of infinity .
@xmath118 the minkowski nondegeneracy of @xmath3 is independent of the choice of the norm on @xmath2 in which we measure the neighborhood of infinity .
there are special cases when we even get the same values for the minkowski contents for different norms on @xmath2 .
one of these cases is addressed in the next lemma which will prove to be useful in some of the future calculations .
it can easily be generalized to the @xmath4-dimensional case but we will need it only in the case of @xmath119 .
[ infty_norm ] let @xmath120 with @xmath15 such that @xmath3 is a subset of a horizontal @xmath121vertical@xmath122 strip of finite width .
let @xmath105 be an open ball in the @xmath22-norm of radius @xmath17 with center at the origin and @xmath24 a real number .
then , we have that @xmath123 and @xmath124 without loss of generality we will assume that the set @xmath3 is contained in the horizontal half - strip @xmath125 . then , for @xmath126 we have that @xmath127 and consequently for @xmath112 @xmath128 taking the upper and lower limits as @xmath23 completes the proof . in the next example we will show that the value @xmath129 can be achieved
. let @xmath130 and let us calculate the box dimension of @xmath3 at infinity using the @xmath131-ball in @xmath119 : @xmath132 consequently , we have that @xmath133 when @xmath44 for every @xmath112 and therefore @xmath134 . from now on , we will always implicitly assume that @xmath135 when dealing with relative fractal drums of the type @xmath136 ( unless stated otherwise ) .
as we have shown in proposition [ < = -n ] , the upper box dimension of any subset of the plane of finite lebesgue measure does not exceed @xmath137 .
the next proposition will show that the value @xmath137 can be achieved and it can be easily adapted for constructing a subset @xmath3 of @xmath2 with finite lebesgue measure such that @xmath46 .
[ prim_d=-2 ] there exists a lebesgue measurable subset @xmath120 with @xmath138 such that @xmath139 let @xmath140 for @xmath141 and we define @xmath142 we will `` stack '' the sets @xmath143 on top of each other . in order to do so , we define @xmath144 to be an @xmath145-translated image of @xmath143 along the @xmath146-axis where @xmath147 and define @xmath148 .
we observe that @xmath3 is contained in the horizontal strip of finite height @xmath149 where @xmath150 .
furthermore , we have that @xmath151 so that @xmath152 . using the same calculation as in proposition [ standardniprim ]
yields @xmath153 finally , by using corollary [ subset_inf_dim ] we have that @xmath154 for every @xmath141 which implies .
let @xmath109 be a measurable set with @xmath138 .
we define the _ lapidus zeta function _ of @xmath3 _ at infinity _ by the lebesgue integral @xmath155 for a fixed @xmath156 and @xmath157 in @xmath158 with @xmath7 sufficiently large
. we will also call this zeta function the _ distance zeta function of @xmath3 at infinity _ and use the two notions interchangeably . from now on ,
our main goal will be to show that this new zeta function has analogous properties as the distance zeta function for relative fractal drums studied in @xcite .
first of all , the dependence of the distance zeta function at infinity on @xmath156 is inessential in the sense that for @xmath159 the difference @xmath160 with @xmath161 is an entire function of @xmath157 . indeed , since @xmath162 for @xmath163 this will follow from theorem [ an2]@xmath164 with @xmath165 , @xmath166 and @xmath167 in the notation of that theorem . therefore , from now on , we will emphasize the dependence of the lapidus zeta function of @xmath3 at infinity on @xmath168 and write @xmath169 only when it is explicitly needed . also note that if @xmath3 is bounded , then for @xmath168 sufficiently large ,
we have that @xmath170 .
the definition of the lapidus zeta function of @xmath3 at infinity is , as we will demonstrate immediately , closely related to the distance zeta function of a certain relative fractal drum .
this relative fractal drum is actually the image of @xmath136 under the geometric inversion in @xmath2 , i.e. , it is equal to @xmath171 here , but we will always abuse notation in this way for a relative fractal drum @xmath0 when the set @xmath1 consists of a single point . ] , where @xmath172 and @xmath173 is the origin . to derive the mentioned relation we will need to compute the jacobian of the geometric inversion and use the change of variables formula for the lebesgue integral . to compute the jacobian
we will use the well - known matrix determinant lemma ( see , e.g. , @xcite ) which we state here for the sake of exposition .
[ mdl ] let @xmath174 be an invertible matrix and @xmath175 , @xmath176 column vectors .
then we have that @xmath177 where @xmath178 and @xmath179 denotes the transpose operator . [ jacobian ]
let @xmath180 be the geometric inversion on @xmath2 .
then for the jacobian of @xmath181 we have@xmath182 @xmath183 with @xmath184 and @xmath185 the kronecker delta we have that @xmath186 and consequently @xmath187 where @xmath188^{\tau}$ ] and @xmath189 is the identity matrix .
now we can apply the matrix determinant lemma with @xmath190 , @xmath191 and @xmath192 from which we obtain @xmath193 the next theorem will show that , from the point of view of the distance zeta functions , there is no difference between the unbounded relative fractal drum @xmath136 and the relative fractal drum @xmath194 obtained from it by geometric inversion .
[ zeta_inversion ] let @xmath3 be a lebesgue measurable subset of @xmath2 of finite measure , @xmath173 the origin and fix @xmath156 .
then we have @xmath195 defining @xmath196 and using lemma [ jacobian ] this is a consequence of the change of variables formula once we observe the fact that @xmath197 : @xmath198 this result suggests that we can analyze fractal properties of @xmath109 at infinity by analyzing the fractal properties of the ` inverted ' relative fractal drum @xmath194 . a similar approach ( in the context of unbounded subsets of @xmath2 )
was made in @xcite .
of course , in that approach , we can use results of @xcite about relative fractal drums and relative distance ( and tube ) zeta functions . on the other hand ,
we stress that in that case we are dealing with the usual relative box dimension of the inverted relative fractal drum , i.e. , with @xmath199 which is defined via the @xmath24-dimensional relative minkowski content , namely , @xmath200 .
however , it is not evident what are the relations between the `` classical '' relative box dimension ( and minkowski content ) of the inverted relative fractal drum with the notions of box dimension and minkowski content at infinity introduced in section [ inf_box_def ] .
we will give an answer to this question in a future work as well as to the natural question about the effect of the one - point compactification on the fractal properties of unbounded sets at infinity as well as how to analyze fractal properties of unbounded sets of infinite measure at infinity .
( see also @xcite . ) to prove the holomorphicity theorem , we will need the following proposition which complements ( * ? ? ?
* lemma 3 ) .
[ integralna_veza ] let @xmath109 be a lebesgue measurable set with @xmath138 , @xmath156 and let @xmath201 be a strictly monotone @xmath202 function . then the following equality holds @xmath203 we will use a well - known fact ( see , e.g. , ( * ? ? ?
* theorem 1.15 ) ) that for a non - negative borel function @xmath204 on a separable metric space @xmath205 the following identity holds @xmath206 we let @xmath207 , @xmath208 and consider separately the cases of strictly decreasing and strictly increasing function @xmath209 .
@xmath117 let @xmath209 be strictly decreasing and @xmath210 .
for the set appearing on the right - hand side of we have @xmath211 for @xmath212 it is true that @xmath213 for any @xmath214 because @xmath215 and we have @xmath216 . furthermore ,
if @xmath217 , it is clear that @xmath218 finally , for @xmath219 we have that @xmath220 because @xmath221 and using we get @xmath222 where we have introduced the new variable @xmath223 in the last equality .
@xmath118 let now @xmath209 be a strictly increasing function and @xmath224 $ ] .
in this case we have @xmath225 for @xmath226 we have that @xmath213 for any @xmath214 because @xmath227 and we have @xmath216 . furthermore , if @xmath228 it is clear that @xmath229 , and for @xmath230 the set @xmath231 is an empty set .
altogether , we have @xmath232 where , again , we have introduced the new variable @xmath223 in the last equality .
this concludes the proof of the proposition .
[ cor_infint ] let @xmath109 be a measurable set with @xmath138 , @xmath156 .
then for every @xmath233 , the following identity holds@xmath182 @xmath234 furthermore , the above integrals are finite for such @xmath235 .
the proposition is a direct consequence of proposition [ integralna_veza ] with @xmath236 when @xmath237 and for @xmath238 the equation is trivially fulfilled .
namely , let us fix @xmath239 .
then for @xmath168 large enough we have that for a constant @xmath240 we have @xmath132 for every @xmath241 . from this
we get that @xmath242 and the last integral above is finite because @xmath243 . in order to prove the holomorphicity theorem we will need the following theorem which we cite from @xcite along with its proof for the sake of exposition .
[ an2 ] let @xmath244 be a measure space , where @xmath245 is a locally compact metrizable space , @xmath246 is the borel @xmath247-algebra of @xmath245 , and @xmath248 is a positive or complex @xmath121local@xmath122 measure , with total variation @xmath121local@xmath122 measure denoted by @xmath249 .
furthermore , let @xmath250 be a measurable function .
then@xmath182 @xmath117 if @xmath251 is essentially bounded @xmath121that is , if there exists @xmath252 such that @xmath253 for @xmath249-a.e .
@xmath254 , and if there exists @xmath255 such that @xmath256 @xmath257 , then @xmath258 is holomorphic on the right half - plane @xmath259 , and @xmath260 in that region .
@xmath118 if there exists @xmath252 such that @xmath261 for @xmath249-a.e .
@xmath262 , and if there exists @xmath263 such that @xmath264 , then @xmath265 is holomorphic on @xmath259 , and @xmath266 in that region .
@xmath164 finally , if there exist positive constants @xmath267 and @xmath268 such that @xmath269 for @xmath249-a.e .
@xmath262 , and there exists @xmath255 such that @xmath270 , then the dirichlet - type integrals @xmath271 and @xmath272 in @xmath117 and @xmath118 , respectively , are entire functions
. we use ( * ? ? ?
* theorem b.4 , page 295 ) ( see also @xcite ) . in our case , @xmath273 , @xmath274 .
note that for any @xmath275 , we have @xmath276 , so that @xmath277 implies that @xmath278 .
in particular , since @xmath279 , it follows that @xmath280 for all @xmath6 such that @xmath281 .
let @xmath282 be a compact subset of @xmath283 .
since @xmath284 for all @xmath285 and @xmath249-a.e .
@xmath262 , where @xmath286 .
this proves part @xmath117 of the theorem .
part @xmath118 follows from part @xmath117 applied to @xmath287 .
finally , part @xmath164 follows similarly as in @xmath117 , by noting that @xmath288 for every complex number @xmath157 .
now we can state and prove the holomorphicity theorem for the lapidus zeta function at infinity , but firstly we will introduce a new notation for the sake of brevity , namely , @xmath289 [ analiticinf ] let @xmath3 be any lebesgue measurable subset of @xmath2 of finite @xmath4-dimensional lebesgue measure .
assume that @xmath168 is a fixed positive number .
then the following conclusions hold .
@xmath117 the abscissa of convergence of the lapidus zeta function at infinity @xmath290 is equal to the upper box dimension of @xmath3 at infinity , i.e. , @xmath291 consequently , @xmath292 is holomorphic on the half - plane @xmath293 and for every complex number @xmath157 in that half - plane we have that @xmath294 @xmath118 if @xmath40 exists and @xmath295 , then @xmath296 for @xmath297 as @xmath298 .
@xmath117 if we let @xmath299 , then from the definitions of the upper minkowski content and of the upper box dimension at infinity we deduce that @xmath300 for every @xmath301 .
now , let us fix @xmath302 such that @xmath303 and take @xmath304 large enough , such that for a constant @xmath240 it holds that @xmath132 furthermore , we estimate @xmath305 in the following way @xmath306 the last inequality follows from the fact that @xmath304 and @xmath307 .
we let now @xmath308 , @xmath166 and @xmath167 and note that @xmath309 for @xmath163 .
part @xmath117 follows now directly from theorem [ an2]@xmath118 . to conclude the proof that @xmath33 is the abscissa of convergence of @xmath292 we take @xmath310 and use proposition [ cor_infint ] : @xmath311 now , we fix @xmath235 such that @xmath312 . from @xmath313
we conclude that there exists a sequence @xmath314 such that @xmath315 when @xmath316 it is clear that the function @xmath317 is nonincreasing and we have @xmath318 therefore , @xmath319 for every @xmath320 which proves that @xmath321 .
@xmath118 let us assume now that @xmath40 exists , and @xmath295 . from proposition
[ < = -n ] we have that @xmath322 . on the other hand , the condition @xmath295 and remark [ = -n ]
imply that @xmath323 .
consequently , we may assume that @xmath324 .
furthermore , @xmath295 implies that there exists a constant @xmath252 such that for a sufficiently large @xmath168 we have that @xmath325 for every @xmath241 .
hence , for @xmath326 we have the following : @xmath327 when @xmath298 , and this proves part @xmath118 . in the special case
when @xmath46 we have from the definition of the upper minkowski content at infinity that @xmath328 and @xmath329 this shows that the condition @xmath295 from part @xmath118 of theorem [ analiticinf ] can not be omitted in the general case .
similarly as in the case of standard relative fractal drums ( see @xcite ) , it is easy to see that theorem [ analiticinf ] is still true if we replace the norm appearing in the definition of the distance zeta function at infinity with any other norm on @xmath2 .
let us now revisit propositions [ 1dimexample ] and [ standardniprim ] from the previous section and compute the corresponding distance zeta functions at infinity .
[ 1dimexamplezeta ] let @xmath330 be the set from definition [ omega(a , b ) ] .
then , for @xmath331 large enough so that @xmath332 is a countable union of disjoint intervals we have that @xmath333 furthermore , we have that @xmath334 and @xmath335 is a removable singularity of @xmath336 . for the distance zeta function of @xmath3 at infinity
we have : @xmath337 from which follows after integrating . by setting @xmath338 and using the mean value theorem for integrals
, we estimate @xmath339 for some @xmath340 so that @xmath341 as @xmath342 which , in turn , implies that @xmath343 the right - hand side is convergent if and only if @xmath344 from which we conclude by using that @xmath345 , which is in accord with theorem [ analiticinf ] .
let @xmath91 for @xmath90 .
then for the distance zeta function of @xmath3 at infinity calculated using the @xmath131 norm on @xmath119 we have @xmath346 it is meromorphic on @xmath158 with a single simple pole at @xmath347 .
in particular , @xmath348 .
let us compute the distance zeta function of @xmath3 at infinity : @xmath349 the last equation holds if and only if @xmath350 . from this and
, we conclude that @xmath351 which is , of course , in accord with theorem [ analiticinf ] .
moreover , the distance zeta function @xmath352 of @xmath3 at infinity can be meromorphically extended to the whole complex plane with a single simple pole at @xmath353 .
revisiting proposition [ prim_d=-2 ] will show that the conditions of theorem [ analiticinf ] can not be relaxed .
[ primd=-2_zeta ] let @xmath3 be as in proposition [ prim_d=-2 ] .
then for the corresponding lapidus zeta function at infinity calculated via the @xmath131-norm on @xmath119 we have @xmath354 furthermore , we also have that @xmath355 and @xmath356 moreover , @xmath352 is holomorphic on the set @xmath357 and @xmath358 is an accumulation point of its simple poles . finally , for the residues of @xmath352 we have that @xmath359 for every @xmath141 .
let us calculate the distance zeta function at infinity using the @xmath131 norm on @xmath2 .
for @xmath360 we have that @xmath361 for @xmath362 and consequently @xmath363 the last equation above is valid if and only if @xmath364 for every @xmath141 .
furthermore , by using the weierstrass @xmath365-test we have that the last sum appearing above defines a holomorphic function on @xmath366 , which implies that @xmath367 . on the other hand , by direct computation we have that @xmath368 , but the zeta function can not be even meromorphically extended to a neighborhood of @xmath358 .
this follows from the fact that for @xmath369 we have that @xmath370 where the functions @xmath371 are meromorphic on @xmath158 with simple poles at @xmath372 .
furthermore , the above sum converges uniformly on compact subsets of @xmath373 , i.e. , it defines a holomorphic function on that set , but it has an accumulation of simple poles at @xmath358 , and by the principle of analytic continuation , the same is true for @xmath352 . in other words ,
@xmath367 and this , in turn , is equal to @xmath37 according to .
although proposition [ primd=-2_zeta ] is stated in terms of the distance zeta function calculated via the @xmath131-norm , proposition [ euc_ty ] below will guarantee that the difference @xmath374 is holomorphic at least on the half - plane @xmath375 . from this
we conclude that is also true for @xmath292 , @xmath376 , and @xmath292 is holomorphic ( at least ) on the set @xmath377 with @xmath358 being an accumulation point of its simple poles .
in this section we will derive results which relate the the upper and lower minkowski content of @xmath136 with the residue of the distance zeta function at infinity at @xmath378 .
[ resdistinf ] let @xmath109 be such that @xmath138 and @xmath379 , @xmath380 .
if @xmath292 has a meromorphic continuation to a neighborhood of @xmath353 , then @xmath381 is a simple pole and it holds that @xmath382 moreover , if @xmath3 is minkowski measurable at infinity , then we have @xmath383 firstly , using the fact that @xmath295 we can apply part @xmath164 of theorem [ analiticinf ] to get that @xmath296 as @xmath384 .
in fact , by looking at the proof of part @xmath164 of theorem [ analiticinf ] we can see that @xmath353 is a singularity of @xmath292 that is at least a simple pole .
it remains to show that the order of this pole is not greater than one .
let us define @xmath385 .
from @xmath386 we have that @xmath387 for @xmath168 large enough .
now , for @xmath297 such that @xmath326 by using proposition [ cor_infint ] we have @xmath388 this implies that @xmath389 where @xmath390 is a constant independent of @xmath157 and @xmath168 and from this we conclude that @xmath353 is a pole of at most order one , i.e. , it is a simple pole . to compute the residue at @xmath353
we observe that its value is independent of @xmath168 because the difference @xmath391 is an entire function .
furthermore , from we have @xmath392 and taking limits on both sides as @xmath298 yields @xmath393 finally , by taking the limit as @xmath394 we get @xmath395 .
the proof of the inequality involving the lower minkowski content is completely analogous and this completes the proof .
the next technical proposition is needed in order to establish a finer connection between the zeta function at infinity defined via the euclidean norm and the one defined via the @xmath131-norm .
it is very useful since the later zeta function can be calculated explicitly in the examples we are interested in .
the proof follows from a more general theorem ( see ( * ? ? ?
* theorem 4.55 ) ) which is proved by using the complex mean value theorem ( * ? ? ?
* theorem 2.2 ) and the theorem about complex differentiation under the integral sign ( see , e.g. , @xcite ) . due to the technical nature
we omit the proof here and refer the reader to ( * ? ? ?
* theorem 4.55 and proposition 4.58 ) for the detailed proof .
[ euc_ty ] let @xmath109 with @xmath138 be such that it is contained in a cylinder @xmath396 for some constant @xmath252 where @xmath184 .
furthermore , let @xmath397 and @xmath156 .
then @xmath398 is holomorphic on @xmath121at least@xmath122 the half - plane @xmath399 .
furthermore , if any of the two distance zeta functions possesses a meromorphic extension to some open connected neighborhood @xmath400 of the critical line @xmath401 , then the other one possesses a meromorphic extension to @xmath121at least@xmath122 @xmath402 .
moreover , their multisets of poles in @xmath403 coincide .
we now introduce the notion of complex dimensions of @xmath136 analogously as in the case of ordinary relative fractal drums .
[ pcw ] let @xmath109 be of finite @xmath4-dimensional lebesgue measure and such that its lapidus zeta function at infinity can be meromorphically extended to some open connected neighborhood @xmath404 of the half - plane @xmath405 .
we define the _ set of visible complex dimensions of @xmath136 through @xmath404 _ as the set of poles of the distance zeta function @xmath292 that are contained in @xmath404 and denote it by @xmath406 which we will abbreviate to @xmath407 when there is no ambiguity concerning the choice of @xmath404 ( or when @xmath408 ) .
furthermore , if @xmath292 possesses a meromorphic continuation to the whole of @xmath158 , we will call the set @xmath409 the _ set of @xmath121all@xmath122 complex dimensions of @xmath136_. the subset of @xmath410 consisting of poles with real part equal to @xmath411 is called the set of _ principal complex dimensions of @xmath136 _ and is denoted by @xmath412 .
in this section we will construct a subset of @xmath119 with prescribed box dimension @xmath413 at infinity that will have a cantor - like structure in a sense that will be described below .
this set depends on two parameters and is denoted by @xmath414 in definition [ omega(a , b , ty ) ] .
furthermore , these sets can be used as building blocks for the construction of ( algebraically and transcendentally ) quasiperiodic sets at infinity by using some classical results from transcendental number theory ( see @xcite ) . from definition [ omega(a ,
b , ty ) ] . here , @xmath415 and @xmath416 .
note that the axes are not in the same scale and only the first four steps in the construction of the set @xmath417 are shown ; that is , for @xmath418.,width=453,height=132 ] [ omega(a , b , ty ) ] for @xmath419 and @xmath420 we define a _ two parameter unbounded set _ denoted by @xmath414 .
we start with the countable family of sets @xmath421 now , we will construct the set @xmath414 by `` stacking '' the translated images of the sets @xmath422 along the @xmath146-axis on top of each other .
more precisely , for each @xmath423 we take @xmath424 copies of @xmath422 and arrange all of these sets by vertical translations so that they are pairwise disjoint and lie in the strip @xmath425 . here
, @xmath426 is the sum of widths of all of these sets , i.e. , @xmath427 moreover , without loss of generality , we can arrange them in an `` increasing fashion '' , i.e. , stacking them from bottom to top as @xmath428 increases @xmath121see figure [ cantor_u_besk]@xmath122 . finally , we define @xmath414 as the disjoint union of all of these sets .
[ twop ] the distance zeta function of the two parameter unbounded set @xmath414 calculated via the @xmath131-norm on @xmath119 is given by @xmath432 and is meromorphic on @xmath158 .
furthermore , the set of complex dimensions of @xmath414 at infinity visible through @xmath433 is given by @xmath434 finally , we also have that @xmath435 let us choose @xmath436 and calculate : @xmath437 where the last two equalities follow since @xmath438 . from this
we see that @xmath439 and the zeta function has a ( unique ) meromorphic extension to all of @xmath158 defined by .
furthermore , we have that @xmath440 since @xmath441 is contained in a strip of finite width , we can apply proposition [ euc_ty ] to conclude that the difference @xmath442 is holomorphic on the half - plane @xmath443 from which we conclude that the complex dimensions of @xmath441 visible through @xmath404 are given by .
the two parameter set @xmath414 is cantor - like in the sense that its construction parallels , in a way , the construction of the ( generalized ) cantor set .
for instance , if we choose @xmath444 then the construction of the sets @xmath445 for @xmath423 parallels the deletion of the middle - third interval in the standard middle - third cantor set .
this cantor - like structure can also be seen in the structure of the complex dimensions of the two sets .
namely , the set of principal complex dimensions of the middle - third cantor set is given by @xmath446 while the set of principal complex dimensions of @xmath447 is equal to @xmath448 .
as we can see , the _ oscillatory period _
@xmath449 of these two sets coincides . in the definition of fractality proposed in @xcite ,
we have defined a set or a relative fractal drum to be fractal if it possesses a nonreal complex dimension .
the motivation for this definition is justified , under mild hypotheses , in the case of relative fractal drums since it is shown in @xcite that nonreal complex dimensions generate oscillations in the inner geometry of the relative fractal drum .
we expect that analogous results can also be derived in the case of fractal sets at infinity .
it is not difficult to compute the box dimension of @xmath414 at infinity directly and obtain that @xmath450 .
furthermore , one also obtains that @xmath414 is not minkowski measurable at infinity which is expected due to the presence of nonreal complex dimensions . for the detailed calculation see ( *
* example 4.63 ) . , _ unbounded oscillation of higher - order nonlinear delay dynamic equations of neutral type with oscillating coefficients _ , electronic journal of qualitative theory of differential equations * 34 * ( 2009 ) , 114 . , _ on the behaviour of solutions of higher - order elliptic equations in unbounded domains _ , tr
mosk . mat .
ob^ s. * 31 * ( 1974 ) , 3558 ( in russian ) .
english translation : trans .
* 31 * ( 1974 ) , 3054 . , _ fractal zeta functions and complex dimensions of relative fractal drums _ , journal of fixed point theory and applications no . 2 , * 15 * ( 2014 ) , 321378 .
festschrift issue in honor of haim brezis 70th birthday .
( doi : 10.1007/s11784 - 014 - 0207-y . )
( also : e - print , arxiv:1407.8094v3 [ math - ph ] , 2014 ; ihes preprint , ihes / m/15/14 , 2015 . ) , _ fractal zeta functions and complex dimensions : a general higher - dimensional theory _ , survey article , in : _ fractal geometry and stochastics v _
( c. bandt , k. falconer and m. zhle , eds . ) , proc . fifth internat . conf .
( tabarz , germany , march 2014 ) , progress in probability , vol . * 70 * , birkhuser / springer internat . ,
basel , boston and berlin , 2015 , pp .
229257 ; doi:10.1007/978 - 3 - 319 - 18660 - 3@xmath451 - 1pt13 .
( based on a plenary lecture given by the first author at that conference . )
( also : e - print , arxiv:1502.00878v3 [ math.cv ] , 2015 ; ihes preprint , ihes / m/15/16 , 2015 . ) , _ fractal geometry , complex dimensions and zeta functions@xmath182 geometry and spectra of fractal strings _ , second revised and enlarged edition ( of the 2006 edn . ) , springer monographs in mathematics , springer , new york , 2013 . , _ the dirichlet problem for elliptic equations of arbitrary order in unbounded regions _ , dokl .
nauk sssr * 150 * ( 1963 ) , 12211224 ( in russian ) .
english translation : sov .
dokl . * 4 * ( 1963 ) , 860863 . | we study fractality of unbounded sets of finite lebesgue measure at infinity by introducing the notions of minkowski dimension and content at infinity .
we also introduce the lapidus zeta function at infinity , study its properties and demonstrate its use in analysis of fractal properties of unbounded sets at infinity . |
in the first part of this paper ( sections [ sec : cluster - complexes][sec : recip ] ) , we introduce and study a simplicial complex @xmath3 associated to a finite root system @xmath4 and a nonnegative integer parameter @xmath0 . for @xmath5 ,
our construction specializes to the ( simplicial ) generalized associahedra @xmath6 introduced in @xcite and identified in @xcite as the underlying complexes for the cluster algebras of finite type .
we enumerate the faces of the complexes @xmath3 and determine their euler characteristics .
for the classical types in the cartan - killing classification , we provide explicit combinatorial descriptions of these complexes in terms of dissections of a convex polygon into smaller polygons . in types
@xmath7 and @xmath8 , we rediscover the constructions given by e. tzanaki @xcite .
enumerative combinatorial invariants of the complexes @xmath3 provide natural generalizations of the fuss - catalan , kirkman - cayley , and przytycki - sikora numbers to arbitrary types , suggesting connections with a host of well studied problems in the algebraic combinatorics of finite coxeter groups , root systems , and hyperplane arrangements .
the second part of the paper ( sections [ sec : coxeter - inv][sec : fake ] ) is devoted to combinatorial algorithms for determining coxeter - theoretic invariants . starting with a coxeter diagram of a finite coxeter group ( or with the corresponding dynkin diagram or cartan matrix ) , we compute the coxeter number , the exponents , and other related invariants by a procedure ( in fact , by several alternative procedures ) which is entirely combinatorial in nature .
that is , these procedures only use the most elementary graph - theoretic concepts applied to the input diagram , and do not involve , in any way , the coxeter group itself , the associated root system , the root lattice , or any other group - theoretic , lattice - theoretic , or lie - theoretic notions and constructions .
the crucial ingredients in all of these calculations are some identities for the face numbers of the generalized cluster complexes established in the first part of the paper . formally extending these procedures
, one can calculate `` fake '' coxeter invariants ( such as a `` fake coxeter number '' ) of various infinite coxeter groups .
the only prerequisites for this paper are the fundamentals of root systems and finite coxeter groups ; see , e.g. , ( * ? ? ?
* lectures 12 ) , or the standard references @xcite .
the paper is organized as follows .
section [ sec : cluster - complexes ] reviews the main notions related to the generalized associahedra @xmath6 .
readers familiar with the original source ( * ? ? ?
* section 3 ) or with the survey in ( * ? ? ?
* section 4.3 ) may proceed directly to section [ sec : gcc ] .
there , we define the generalized cluster complexes @xmath3 and describe their basic structural properties .
the proofs of two of these results ( theorems [ th : alt def ] and [ th : restrict ] ) are given in sections [ sec : alt def ] and [ sec : restrict ] , respectively , and can be skipped at first reading .
section [ sec : examples ] presents some examples of the complexes @xmath3 .
combinatorial models for @xmath3 when @xmath4 is of classical type ( i.e. , @xmath9 , @xmath8 , @xmath10 , or @xmath11 ) are given in section [ sec : classical ] .
recursions and explicit multiplicative formulas for the face numbers of @xmath3 are given in section [ sec : face - numbers ] ; these formulas are proved in section [ sec : proof - product - f ] . in section [ sec
: h - vectors ] , we compute the @xmath2-vector of @xmath3 , recovering the @xmath0-narayana numbers introduced by c. a. athanasiadis @xcite .
the reduced euler characteristic of @xmath3 is determined in section [ sec : euler ] . in section [ sec : recip ] , we enumerate the @xmath0-analogues of `` positive clusters . '' section [ sec : coxeter - inv ] presents several alternative algorithms for computing classical invariants of a root system @xmath4 ( or the associated coxeter group @xmath12 ) . in section [ sec : fake ] , these algorithms are applied to calculate `` fake invariants '' of some infinite coxeter groups .
this section reviews the basic facts from ( * ? ? ?
* section 3 ) concerning ( simplicial ) generalized associahedra ; alternatively , see ( * ? ? ?
* section 4.3 ) .
the only difference is that we do _ not _ assume that the underlying root system @xmath4 is crystallographic .
this does not in fact create any additional complications : as noted in @xcite , the constructions in @xcite extend verbatim to the non - crystallographic case .
let @xmath4 be a finite root system of rank @xmath13 .
we denote by @xmath14 the set of positive roots in @xmath4 .
the set of simple roots in @xmath4 is denoted by @xmath15 , where @xmath16 is an @xmath13-element indexing set .
( the standard choice is @xmath17 . )
accordingly , @xmath18 is the set of negative simple roots .
the set @xmath19 of simple reflections corresponding to the simple roots @xmath20 generates a finite reflection group @xmath12 .
the pair @xmath21 is a coxeter system ; the coxeter group @xmath12 naturally acts on the set of roots @xmath4 .
let us temporarily assume that the root system @xmath4 is _
irreducible_. let @xmath22 be a decomposition of @xmath16 such that the sets @xmath23 and @xmath24 are disjoint , and each of them labels a totally disconnected set of vertices in the coxeter diagram of @xmath4 .
the ground set for the cluster complex @xmath6 is the set @xmath25 of _ almost positive roots_. we define the involutions @xmath26 by @xmath27 \displaystyle\bigl(\prod_{i \in i_\varepsilon } s_i\bigr)\,(\alpha ) & \text{otherwise , } \end{cases}\ ] ] for @xmath28 .
the product @xmath29 can be viewed as a deformation of the coxeter element in @xmath12 .
we denote by @xmath30 the cyclic group generated by @xmath31 .
let @xmath2 denote the _ coxeter number _ of @xmath12 and let @xmath32 be the longest element of @xmath12 .
[ lem : dihedral ] the order of @xmath31 is @xmath33 if @xmath34 , and is @xmath35 otherwise .
every @xmath30-orbit in @xmath36 has a nonempty intersection with @xmath37 .
these intersections are precisely the @xmath38-orbits in @xmath37 .
the following theorem is a reformulation of results in ( * ? ? ?
* section 3.1 ) .
[ th : compat ] there is a unique symmetric binary relation on @xmath39 ( called `` compatibility '' ) such that : * @xmath40 and @xmath41 are compatible if and only if @xmath42 and @xmath43 are compatible ; * a negative simple root @xmath44 is compatible with a positive root @xmath41 if and only if the simple root expansion of @xmath41 does not involve @xmath20 .
following @xcite , we define the _ cluster complex _
@xmath6 ( of type @xmath4 ) as the clique complex for the compatibility relation .
that is , a subset of roots in @xmath39 forms a simplex in @xmath6 if and only if every pair of roots in this subset is compatible . if @xmath4 is _ reducible _ , with irreducible components @xmath45 , then @xmath46 ( disjoint union ) .
we declare two roots in @xmath39 compatible if and only if they either belong to different components , or belong to the same component and are compatible within it .
thus , the simplicial complex @xmath6 is the _ join _ of the complexes @xmath47 .
the simplicial complex @xmath6 is homeomorphic to a sphere @xcite .
moreover , it can be explicitly realized as a boundary of a convex polytope @xcite , a polar dual to the ( simple ) _ generalized associahedron _ of type @xmath4 .
this is why @xmath6 is sometimes referred to as the _ simplicial _ generalized associahedron ( of type @xmath4 ) . by ( * ?
? ? * theorem 1.13 ) , the definition of @xmath6 given above is equivalent to the algebraic definition @xcite of a cluster complex for a cluster algebra of finite type .
although inspired by cluster algebra theory , this paper does not rely on any of its results ; cf.remark [ rem : m - cluster - algebras ] .
let @xmath0 be a nonnegative integer . in this section , we define and begin to study the main object of this paper , the generalized cluster complex @xmath3 .
the ground set of @xmath3 is the set @xmath48 of _ colored almost positive roots_. it consists of @xmath0 copies of the set @xmath14 of positive roots in @xmath4 together with one copy of the negative simple roots
. it will be convenient to use the following notation for the elements of @xmath48 . for each @xmath49 ,
let @xmath50 denote the @xmath0 `` colored '' copies of @xmath40 occurring in @xmath51 .
each negative simple root @xmath40 occurs in @xmath48 as @xmath52 .
thus , @xmath53 the simplicial complex @xmath3 is defined using the binary _ compatibility relation _ on @xmath48 .
this relation can be defined in ( at least ) two different ways , which we will now describe . as in the non - colored ( @xmath1 ) case , we assume that @xmath4 is irreducible , since the reducible case can be obtained by taking joins . for a root @xmath54 , let @xmath55 denote the smallest @xmath56 such that @xmath57 is a negative root .
in particular , @xmath58 if @xmath41 is negative simple .
[ def : m - compat ] two colored roots @xmath59 are called _ compatible _ if and only if one of the following conditions is satisfied : * @xmath60 , @xmath61 , and the roots @xmath42 and @xmath41 are compatible ( in the original `` non - colored '' sense of theorem [ th : compat ] ) ; * @xmath62 , @xmath63 , and the roots @xmath40 and @xmath43 are compatible ; * @xmath60 , @xmath64 , and the roots @xmath40 and @xmath41 are compatible ; * @xmath62 , @xmath65 , and the roots @xmath40 and @xmath41 are compatible ; * @xmath66 , and the roots @xmath40 and @xmath41 are compatible .
the compatibility relation on @xmath48 is symmetric .
immediate from definition [ def : m - compat ] and the symmetry of the compatibility relation on @xmath67 .
we next define @xmath68 , the @xmath0-analogue of @xmath31 , and prove an @xmath0-analogue of theorem [ th : compat ] .
[ def : rm ] for @xmath69 , we set @xmath70 ( r(\alpha))^1 & \text{otherwise . }
\end{cases}\ ] ] [ th : alt def ] the compatibility relation on @xmath48 has the following properties : * @xmath71 is compatible with @xmath72 if and only if @xmath73 is compatible with @xmath74 ; * @xmath75 is compatible with @xmath72 if and only if the simple root expansion of @xmath41 does not involve @xmath20 .
furthermore , conditions ( i)(ii ) uniquely determine this relation .
the proof of theorem [ th : alt def ] is given in section [ sec : alt def ] .
it does not rely on any statements proved in between .
[ cor : tau - switch ] simultaneously replacing @xmath29 by @xmath76 and changing the colors by @xmath77 does not change the compatibility relation on @xmath78 .
immediate from theorem [ th : alt def ] , since all we are really doing is replacing @xmath68 by @xmath79 .
[ cor : subcomplex ] if @xmath80 , then @xmath81 is a ( vertex - induced ) subcomplex of @xmath82
. follows directly from theorem [ th : alt def ] . for a reducible root system @xmath83
, compatibility is defined by analogy with the non - colored case .
two colored roots @xmath84 and @xmath85 are compatible if either @xmath86 or else @xmath87 and @xmath71 and @xmath72 are compatible as elements of @xmath88 . for a non - empty subset @xmath89 ,
let @xmath90 denote the parabolic root subsystem of @xmath4 spanned by the simple roots @xmath91 for @xmath92 . for @xmath93
, we denote @xmath94 .
thus @xmath95 consists of all roots in @xmath4 whose simple root expansion does not involve @xmath96 .
the compatibility relation is preserved under restriction to a parabolic subsystem : [ th : restrict ] if @xmath97 and @xmath98 , then @xmath40 and @xmath41 are compatible in @xmath99 if and only if they are compatible in @xmath100 .
the proof of theorem [ th : restrict ] is given in section [ sec : restrict ] using the classification of finite coxeter groups together with the combinatorial models for generalized cluster complexes of types @xmath101 presented in section [ sec : classical ] .
[ def : gcc ] the _ generalized cluster complex _ @xmath3 is the clique complex for the compatibility relation on @xmath48 .
that is , the ground set for @xmath3 is the set of colored almost positive roots , and a subset of such roots forms a simplex if and only if any two of them are compatible .
if @xmath4 is irreducible , corollary [ cor : tau - switch ] implies that the complex @xmath3 is invariant , up to an isomorphism , under interchanging @xmath102 and @xmath103 .
if @xmath4 is reducible , @xmath104 , then @xmath3 is , by design , the join of the complexes @xmath105 .
as in the non - colored case , all maximal simplices of @xmath3 have cardinality @xmath13 : [ th : pure ] the simplicial complex @xmath3 is pure of dimension @xmath106 .
let @xmath107 be a maximal face of @xmath3 and let @xmath71 be a colored root in @xmath107 .
choose @xmath108 so that @xmath109 is a negative simple root @xmath110 .
then @xmath111 is a maximal face of @xmath3 containing @xmath110 . by theorem
[ th : alt def ] , all other roots in @xmath112 are of the form @xmath72 for @xmath113 , and these roots form a maximal face @xmath114 in @xmath115 .
( here we use theorem [ th : restrict ] . ) by induction , this simplex is of dimension @xmath116 , so @xmath107 is of dimension @xmath106 .
the base of the induction ( @xmath117 ) is trivial ; cf .
example [ example : gcc : n=1 ] below .
[ prop : m - fold ] each codimension @xmath118 face of @xmath3 is contained in exactly @xmath119 faces of maximal dimension .
let @xmath107 be a codimension @xmath118 face of @xmath3 and let @xmath71 , @xmath108 , @xmath120 , @xmath112 and @xmath114 be defined as in the proof of theorem [ th : pure ] . then @xmath114 is a codimension @xmath118 face of @xmath115 , which by induction is contained in exactly @xmath119 faces of maximal dimension .
adjoining @xmath110 to each of these faces and applying @xmath121 , one obtains all maximal faces containing @xmath107 . [
prop : link ] the link of any face in @xmath3 is isomorphic to a join of generalized cluster complexes of the form @xmath122 , for some irreducible parabolic root subsystems @xmath123 . it is enough to verify the statement for links of vertices .
applying @xmath68 to a vertex until it becomes a negative simple root , the statement follows .
[ rem : dual - root - system ] assume that @xmath4 is a crystallographic root system , and let @xmath124 be the root system dual to @xmath4 . as shown in ( * ? ? ?
* proposition 3.3 ) , two roots @xmath125 are compatible if and only if the corresponding coroots @xmath126 and @xmath127 are compatible in @xmath128 .
thus the cluster complex @xmath6 is canonically isomorphic to @xmath129 .
it follows directly from definition [ def : m - compat ] that the same is true about @xmath3 and @xmath130 .
in particular , the generalized cluster complexes of types @xmath131 and @xmath132 are canonically isomorphic . in view of this
, we do not consider type @xmath132 in what follows .
[ rem : m - cluster - algebras ] it would be very interesting to find a generalization of the notion of a cluster complex of a cluster algebra of finite type ( as defined in @xcite ) that would yield the complexes @xmath3 .
this might also lead to an extension of the concept to infinite types .
here we illustrate the definition of a generalized cluster complex ( see definition [ def : gcc ] ) by considering the special cases where @xmath133 or @xmath134 .
[ example : m=0 ] the ground set of @xmath135 consists of the @xmath13 negative simple roots in @xmath4 .
any two such roots are compatible .
hence @xmath135 is an @xmath136-dimensional simplex .
[ example : m=1 ] by design , @xmath137 is the cluster complex of ( finite ) type @xmath4 . in particular , in type @xmath7 we obtain stasheff s associahedron , while in type @xmath131/@xmath132 we get the bott - taubes cyclohedron .
see @xcite or section [ sec : classical ] below for further details .
[ example : gcc : n=1 ] for a root system @xmath4 of type @xmath138 , the simplicial complex @xmath3 consists of @xmath119 disconnected points @xmath139 .
[ example : gcc , n=2 ] let @xmath4 be a root system of type @xmath140
. the ( @xmath118-dimensional ) simplicial complex @xmath3 is an @xmath141-regular graph on @xmath142 vertices .
figure [ fig : b2.3 ] shows this graph for a root system of type @xmath143 and @xmath144 .
the figure is color - coded to highlight the subgraphs corresponding to @xmath145 and @xmath1 and thus to illustrate how @xmath146 changes with increasing @xmath0 ( cf .
corollary [ cor : subcomplex ] ) .
the simple roots are @xmath147 and @xmath148 , with @xmath149 and @xmath150 .
the other two roots are denoted by @xmath151 and @xmath152 .
the black edges are the edges of @xmath153 , the black and dark - gray edges form @xmath154 , and the edges of all colors form @xmath155 .
figure [ fig : b2.3star ] shows the same graph , drawn so as to make its symmetry apparent .
the map @xmath68 acts by a @xmath156 counterclockwise rotation of this picture . in general , the graph @xmath157 can be constructed in the plane as follows .
take the vertex set to be the integers modulo @xmath158 , identified with the vertices of a regular @xmath158-gon so that @xmath159 lists them in clockwise order .
for @xmath160 odd , the edge set has @xmath158-fold rotational symmetry and connects each vertex @xmath161 to the @xmath119 vertices of the form @xmath162 , for @xmath163 .
figure [ fig : i25.4 ] shows this graph for @xmath164 and @xmath165 . for @xmath160
even , fix an odd integer @xmath108 .
the edge set of @xmath157 has @xmath166-fold rotational symmetry , and connects @xmath167 to the vertices @xmath168 .
figure [ fig : b2.3star ] is @xmath155 drawn in this style with @xmath169 .
[ rem : rp2 ] in the special case @xmath170 , the complex @xmath171 can be obtained from the boundary of a regular @xmath172-gon by adding an edge connecting every pair of antipodal vertices , as illustrated in figure [ fig : b2.2rp ] .
thus , @xmath171 is the of a polygonal subdivision of @xmath173 .
it would be interesting to determine which complexes @xmath3 ( perhaps all of them ? ) can be realized as skeleta of polyhedral @xmath174-dimensional manifolds .
in this section , we describe combinatorial models realizing the generalized cluster complexes @xmath3 for the classical types @xmath9 , @xmath175 , and @xmath176 .
each of these models is constructed in terms of collections of non - intersecting diagonals in a certain convex polygon . in every case
, the isomorphism between the suggested model and the root - theoretic description of @xmath3 can be verified by straightforward induction on @xmath0 .
the base of the induction is the non - colored ( @xmath1 ) case , described in ( * ? ? ?
* section 3.5 ) . for the types
@xmath7 and @xmath8 , we rediscover the constructions given earlier by e. tzanaki @xcite .
the type-@xmath7 construction was inspired by the paper @xcite by j. h. przytycki and a. s. sikora .
it is easy to see that a convex polygon @xmath177 with @xmath178 vertices can be dissected into ( convex ) @xmath179-gons by pairwise non - crossing diagonals if and only if @xmath180 .
( here and in what follows , `` non - crossing '' means that diagonals have no common points in the _ interior _ of @xmath177 . )
let @xmath177 be a convex polygon with @xmath181 vertices . to simplify descriptions , we take @xmath177 to be regular . a diagonal of @xmath177
is called _ @xmath0-allowable _ if it cuts @xmath177 into two polygons each of which can be dissected into @xmath179-gons .
that is , an @xmath0-allowable diagonal connects two vertices such that each of the two paths connecting them along the perimeter of @xmath177 passes through a number of vertices that is divisible by @xmath0 .
let us now define a simplicial complex @xmath182 on the set of all @xmath0-allowable diagonals in @xmath177 .
such diagonals form a simplex in @xmath182 if and only if they are pairwise non - crossing .
the complex @xmath182 is easily seen to be pure , with maximal simplices formed by @xmath13-tuples of diagonals that cut @xmath177 into @xmath179-gons .
we now construct an isomorphism between @xmath182 and the generalized cluster complex @xmath3 for a root system @xmath4 of type @xmath9 . under the isomorphism
, @xmath68 will correspond to a clockwise rotation of @xmath177 taking @xmath183 to @xmath184 , etc . to avoid additional notation
, we refer to this clockwise rotation as @xmath68 while defining the isomorphism .
we use the standard labeling of the simple roots in @xmath4 ( see , e.g. , @xcite ) and take @xmath23 to be the odd indices . let @xmath185 be the vertices of @xmath177 , labeled counterclockwise . for @xmath186 , identify the negative simple root @xmath187 with the diagonal of @xmath177 connecting @xmath188 to @xmath189 . for @xmath190 , identify @xmath191 with the diagonal connecting @xmath192 to @xmath189 .
collectively , these @xmath13 diagonals ( each of them @xmath0-allowable ) form what we call the _ @xmath0-snake _
* figure 3 ) ) .
the positive roots of @xmath4 are @xmath193 for each @xmath108 and @xmath194 with @xmath195 . for each such @xmath108 and @xmath194 , there are exactly @xmath0 diagonals which are @xmath0-allowable and intersect the diagonals @xmath196 and no other diagonals in the @xmath0-snake .
this collection of diagonals is of the form @xmath197 for some diagonal @xmath198 . for @xmath199
, we identify @xmath200 with @xmath201 .
figure [ fig : a4.5snake ] shows the @xmath0-snake for @xmath202 and @xmath203 , along with the diagonals identified with the colored roots @xmath204 . for @xmath1 , the complex @xmath205 is , by definition ( see , e.g.
, @xcite and references therein ) , the dual complex for ( the boundary of ) the ordinary @xmath13-dimensional _ associahedron _ , also known as the _ stasheff polytope_. the first non - trivial example with @xmath206 is @xmath207 , the complex of quadrangulations of an octagon .
example [ example : gcc , n=2 ] with @xmath145 and @xmath208 , and also remark [ rem : rp2 ] with @xmath208 . )
this @xmath118-dimensional complex is a non - planar 3-regular graph with @xmath209 vertices and @xmath210 edges .
thus @xmath211 is homotopy equivalent to a wedge of 5 circles .
when @xmath4 is of type @xmath131 ( or @xmath132cf.remark [ rem : dual - root - system ] ) , the complex @xmath3 can be realized as follows .
let @xmath177 be a centrally symmetric regular polygon with @xmath213 vertices . the map @xmath68 will correspond to a @xmath214 clockwise rotation about the center of @xmath177 .
the vertices of the complex are of two kinds : * the _ diameters _ of @xmath177 , i.e. , diagonals connecting antipodal vertices ( all such diagonals are @xmath0-allowable in the sense of section [ subsec : gcc - an ] ) ; * the pairs @xmath215 of distinct @xmath0-allowable diagonals of @xmath177 such that @xmath198 is related to @xmath216 by a half - turn about the center of @xmath177 .
two vertices are called `` non - crossing '' if no diagonal representing one vertex crosses a diagonal representing the second vertex .
the simplices in the complex are the sets of pairwise non - crossing vertices .
it is easily verified that this is a pure complex of dimension @xmath13 .
its maximal simplices correspond to centrally symmetric dissections of @xmath177 into @xmath179-gons .
figure [ fig : b2.2 ] illustrates this complex for @xmath170 .
the edges of the graph correspond to centrally symmetric quadrangulations of a @xmath217-gon . for @xmath1
, this construction specializes to one of the common definitions of the @xmath13-dimensional _ cyclohedron _ , or _ bott - taubes polytope _ ( see , e.g. , @xcite and references therein ) . in order to describe an isomorphism between this complex and @xmath218
, we relate the former to @xmath182 .
label the simple roots @xmath219 so that @xmath220 is the only simple root in its @xmath12-orbit .
the negative simple roots correspond to the orbits , under central symmetry , of the diagonals on the @xmath0-snake of type @xmath221 ( see section [ subsec : gcc - an ] ) .
if @xmath222 are the simple roots of @xmath223 , then @xmath44 is encoded by the pair of diagonals corresponding to @xmath224 and @xmath225 , for @xmath226 , while @xmath227 is encoded by the diameter corresponding to @xmath228 .
the positive roots of @xmath131 can be divided into three categories : @xmath229 a colored positive root @xmath71 for @xmath40 in category ( i ) is encoded by the pair consisting of the diagonal for @xmath230 and the diagonal for @xmath231 . for @xmath40 in category ( ii )
, @xmath71 is encoded by the diameter corresponding to @xmath232 . for @xmath40 in category ( iii )
, @xmath71 is encoded by the diagonal corresponding to @xmath233 and the diagonal corresponding to @xmath234 .
let @xmath177 be a regular polygon with @xmath235 vertices .
the vertices in the combinatorial realization of @xmath236 fall into two groups .
the vertices in the first group correspond one - to - one to pairs of distinct non - diameter diagonals in @xmath177 related by a half - turn . in the second group , each vertex
is indexed by a diameter of @xmath177 , together with one of two _ flavors _ , which we will call `` dashed '' and `` gray , '' and picture accordingly .
thus , each diameter occurs twice , in each of the two flavors .
we label the vertices of @xmath177 counterclockwise : @xmath237 we call @xmath238 $ ] the _ primary diameter_. by construction , the map @xmath68 acts by rotating @xmath183 clockwise to @xmath184 and switching the flavor of certain diameters .
specifically , @xmath68 preserves flavor when applied to a diameter of the form @xmath239 $ ] ( for @xmath240 ) unless @xmath241 or @xmath242 . if @xmath241 or @xmath242 , the flavor is switched . by analogy with types @xmath7 and @xmath8 , the complex @xmath236
is realized as a clique complex for a certain `` compatibility '' relation defined as follows .
two vertices at least one of which is not a diameter are compatible or not according to precisely the same rules as in the type-@xmath131 case .
a more complicated condition determines whether two diameters are compatible .
two diameters with the same location and different flavors are compatible .
two diameters at different locations are compatible if and only if applying @xmath68 repeatedly until either of them is in position @xmath238 $ ] results in diameters of the same flavor .
explicitly , let @xmath198 be a flavored diameter and let @xmath243 denote @xmath198 with its flavor reversed .
then for @xmath244 , the flavored diameter @xmath245 is compatible with @xmath198 ( and incompatible with @xmath243 ) if and only if @xmath246 is even .
figure [ fig : d3.2diam ] shows the @xmath210 maximal simplices of @xmath247 containing the gray diameter @xmath238^\text{gr}$ ] .
the points @xmath248 are marked with a white dot .
some edges of @xmath177 are broken , indicating the locations at which diameters change flavors under the action of @xmath68 .
specifically , @xmath249 and @xmath250 are connected by a broken edge if and only if @xmath68 maps a flavored diameter @xmath251 $ ] to the opposite - flavored diameter @xmath252 $ ] . to illustrate the rules of compatibility for pairs of diameters ,
figure [ fig : d3.2orb ] shows the orbit of a maximal simplex under the action of @xmath68 .
we next describe an isomorphism between the complex defined above and the original construction of @xmath236 .
label the simple roots @xmath219 so that @xmath253 and @xmath220 correspond to the ends of the fork in the dynkin diagram of type @xmath11 .
the negative simple roots correspond to the orbits , under central symmetry , of the diagonals in the @xmath0-snake for @xmath254 , except that @xmath220 and @xmath253 both correspond to the same diameter ( in two different flavors ) .
specifically , if @xmath255 are the simple roots of @xmath256 , then @xmath44 is encoded by the pair of diagonals corresponding to @xmath224 and @xmath257 , for @xmath258 . the dashed copy of the diameter corresponding to @xmath228 encodes @xmath259 , while @xmath227 is encoded by the gray diameter corresponding to @xmath228 .
we label the vertices of @xmath177 so that this is the primary diameter @xmath238 $ ] .
the positive roots of @xmath11 can be divided into two categories : @xmath260 a colored positive root @xmath71 for @xmath40 in category ( i ) is encoded by the pair consisting of the diagonal for @xmath230 and the diagonal for @xmath261 if @xmath262 . if @xmath263 then @xmath71 is encoded by the gray diameter corresponding to @xmath264 .
for @xmath40 in category ( ii ) , @xmath71 is encoded by the diagonal corresponding to @xmath265 and the diagonal corresponding to @xmath266 if @xmath267 . if @xmath268 then @xmath71 is encoded by the dashed diameter corresponding to @xmath264 . in types
@xmath7 and @xmath8 , the combinatorial models for the complex @xmath3 presented in sections [ subsec : gcc - an ] and [ subsec : gcc - bn ] describe this complex as a subcomplex of the ordinary cluster complex @xmath269 for a larger root system @xmath270 of type @xmath271 or @xmath272 , respectively .
it is not clear whether such an embedding exists for the type @xmath176 .
the proof makes use of the following results , which can be extracted from @xcite . 1 .
each @xmath31-orbit in @xmath39 either has size @xmath35 and contains two negative simple roots , or has size @xmath33 and contains one negative simple root .
if @xmath2 is even , then for every orbit of size @xmath35 , the two negative simple roots are placed symmetrically in the orbit .
that is , applying @xmath31 to one of the negative simple roots @xmath33 times yields the other negative simple root .
3 . if @xmath2 is odd , then necessarily all @xmath31-orbits have size @xmath35 .
the two negative simple roots in each orbit are placed so that applying @xmath31 to one of the negative simple roots @xmath273 times yields the other negative simple root . then applying @xmath31 an additional @xmath274 times yields the original negative simple root .
4 . in every orbit ,
positive simple roots are adjacent to their negatives .
when @xmath2 is odd , both positive simple roots in an orbit are located on the longer of the two paths between the two negative simple roots .
if @xmath1 , then both characterizations of compatibility specialize to the non - colored version , so from now on we assume that @xmath206 .
let @xmath97 .
without loss of generality , let @xmath275 .
proving condition ( i ) requires us to consider several cases .
first , if neither @xmath40 nor @xmath41 is negative simple and neither @xmath276 nor @xmath277 equals @xmath0 , then condition ( i ) is immediate .
let us consider the cases where neither @xmath40 nor @xmath41 is negative simple but @xmath278 or @xmath279 or both .
if @xmath280 , then @xmath71 and @xmath72 are compatible if and only if @xmath40 and @xmath41 are compatible , and @xmath281 is compatible with @xmath282 if and only if @xmath43 and @xmath42 are compatible , so ( i ) follows by the @xmath31-invariance of non - colored compatibility .
if @xmath283 , then @xmath281 and since @xmath284 , @xmath285 is compatible with @xmath286 if and only if @xmath42 is compatible with @xmath41 , and this is exactly the requirement for @xmath71 to be compatible with @xmath72 . for @xmath287
there are two cases : @xmath65 and @xmath288 . if @xmath65 , then @xmath71 is compatible with @xmath72 if and only if @xmath40 is compatible with @xmath41 .
also , @xmath289 , so @xmath290 is compatible with @xmath282 if and only if @xmath42 and @xmath43 are compatible , and ( i ) follows by @xmath31-invariance of non - colored compatibility . if @xmath288 , then @xmath71 is compatible with @xmath72 if and only if @xmath40 is compatible with @xmath43 .
now @xmath291 , so @xmath290 is compatible with @xmath282 if and only if @xmath40 and @xmath43 are compatible . if both @xmath40 and @xmath41 are negative simple , then @xmath292 , @xmath293 , and @xmath294
. now ( i ) is immediate by @xmath31-invariance .
if only one of the two roots is negative simple , then @xmath295 implies that the negative simple root must be @xmath40 , and in particular @xmath241 and @xmath65 .
the colored roots @xmath52 and @xmath72 are compatible if and only if @xmath40 and @xmath41 are compatible . if @xmath279 , then @xmath293 is compatible with @xmath296 if and only if @xmath42 and @xmath43 are compatible , and ( i ) follows by @xmath31-invariance .
we now assume @xmath297 . if @xmath298 , then @xmath293 is compatible with @xmath286 if and only if @xmath42 and @xmath43 are compatible , so ( i ) follows by @xmath31-invariance .
the condition @xmath298 will hold for most negative simple roots @xmath40 and positive roots @xmath41 .
the only way for it to fail is if @xmath2 is odd , @xmath299 and @xmath300 . in this case , @xmath293 is compatible with @xmath286 if and only if @xmath42 is compatible with @xmath41 .
thus condition ( i ) in this case amounts to verifying that @xmath40 and @xmath41 are compatible if and only if @xmath42 and @xmath41 are compatible .
because @xmath299 , the root @xmath41 lies on the long path between two negative simple roots in its @xmath31-orbit and is adjacent to a negative simple root .
therefore @xmath41 must be a ( positive ) simple root . if @xmath301 , then @xmath40 and @xmath41 are not compatible , and @xmath302 so @xmath42 and @xmath41 are not compatible . if @xmath303 , then @xmath40 and @xmath41 are compatible and @xmath304 is a negative simple root distinct from @xmath40 , so @xmath304 and @xmath40 are compatible .
therefore @xmath42 and @xmath41 are compatible .
if @xmath40 is negative simple , then the third condition in definition [ def : m - compat ] always applies , and condition ( ii ) follows .
it remains to prove the last statement of theorem [ th : alt def ] .
the following is a generalization of lemma [ lem : dihedral ] .
[ lem : m - dihedral ] the order of @xmath68 is @xmath305 if @xmath34 , and is @xmath306 otherwise .
every @xmath307-orbit in @xmath308 has a nonempty intersection with the set @xmath309 .
these intersections are precisely the @xmath38-orbits in @xmath310 .
immediate from lemma [ lem : dihedral ] and definition [ def : rm ] .
it follows from lemma [ lem : m - dihedral ] that conditions ( i ) and ( ii ) uniquely define the compatibility relation on @xmath311 .
it is sufficient to treat the case @xmath312 .
let @xmath313 . by theorem
[ th : alt def ] , @xmath314 if and only if the @xmath0-allowable diagonal in @xmath177 corresponding to @xmath71 does not intersect the diagonal @xmath198 corresponding to @xmath75 .
the diagonal @xmath198 dissects @xmath177 into two polygons @xmath315 and @xmath316 .
the remaining negative simple roots ( besides @xmath44 ) form two @xmath0-snakes , one in each of the two polygons . the diagonal corresponding to a colored positive root @xmath200 is uniquely determined as the @xmath276th ( in clockwise order ) of the exactly @xmath0 diagonals which intersect the diagonals for @xmath317 and no other diagonals in the @xmath0-snake .
thus the correspondence between @xmath318 and the diagonals in the polygons @xmath315 and @xmath316 is identical to the correspondence between colored roots @xmath319 with @xmath314 and diagonals in @xmath177 .
theorem [ th : restrict ] for @xmath4 of type @xmath131 follows easily from the type-@xmath221 case by identifying parabolic root subsystems of @xmath131 with parabolic root subsystems of @xmath221 which are fixed by the involution @xmath320 .
again , it suffices to consider the case @xmath312 .
first , assume that @xmath321 . by theorem
[ th : alt def ] , @xmath319 has @xmath314 if and only if the corresponding diameter or diagonal pair in @xmath177 does not intersect the diagonal pair @xmath322 corresponding to @xmath75 .
these two diagonals dissect @xmath177 into three polygons , one of which is centrally symmetric and two of which are related to each other by central symmetry .
the proposition follows in this case by an argument analogous to the type @xmath7 case . by symmetry ,
we need only consider one additional case : @xmath323 .
the colored almost positive roots compatible with @xmath227 correspond to the dashed primary diameter , all gray non - primary diameters , and all pairs of diagonals which do not intersect the primary diameter .
the parabolic root subsystem @xmath324 is of type @xmath325 .
we will realize @xmath115 as the complex of non - crossing @xmath0-allowable diagonals in a convex polygon @xmath326 with the vertices @xmath327 in counterclockwise order . the vertices @xmath249 are vertices of @xmath177 and the vertices @xmath328 are new . for a diagonal pair @xmath322 in @xmath177 which does not intersect the primary diameter ,
exactly one of the diagonals @xmath198 and @xmath216 connects vertices in @xmath329 , and is therefore a diagonal in @xmath326 .
the dashed primary diameter of @xmath177 corresponds to the diagonal @xmath238 $ ] of @xmath326 .
the correspondence between gray diameters in @xmath177 and the remaining diagonals of @xmath326 is as follows .
the gray diameter @xmath251 $ ] in @xmath177 is mapped to the diagonal @xmath330 $ ] in @xmath326 for @xmath331 .
figure [ fig : diag map ] illustrates this correspondence ( for @xmath144 and @xmath4 of type @xmath332 ) by showing a maximal simplex in @xmath115 represented as a collection of gray diameters in @xmath177 and as the corresponding collection of diagonals in @xmath326 .
@xmath333 it is immediate that if two diagonal pairs in @xmath177 do not intersect the primary diameter , then they intersect each other if and only if the corresponding diagonals in @xmath326 intersect each other . it is also immediate that a diagonal pair in @xmath177 that does not intersect the primary diameter intersects a gray diameter in @xmath177 if and only if the corresponding diagonals in @xmath326 intersect each other .
intersections with the dashed primary diameter of @xmath177 also correspond to intersections with the diagonal @xmath238 $ ] in @xmath326 .
one can check that for @xmath334 , a gray diameter @xmath251 $ ] is compatible with a gray diameter @xmath335 $ ] if and only if the fractional part of @xmath336 is greater than or equal to the fractional part of @xmath337 .
this is exactly the requirement that @xmath330 $ ] and @xmath338 $ ] do not intersect .
thus our correspondence also maps the compatibility relation among diameters and @xmath0-allowable diagonal pairs in @xmath177 compatible with the gray primary diameter to the compatibility relation among @xmath0-allowable diagonals of @xmath326 .
it remains to verify that the process of mapping from @xmath318 to diagonal pairs and flavored diameters in @xmath177 , then mapping to diagonals in @xmath326 produces the same result as the process of mapping directly from @xmath318 to diagonals in @xmath326 by the type @xmath325 construction .
this is a straightforward check . for all exceptional types ( including the non - crystallographic ones ) ,
theorem [ th : restrict ] was checked with the help of a computer . in definition
[ def : m - compat ] , the compatibility of @xmath71 and @xmath72 depends on the comparison between @xmath276 and @xmath277 but not on the actual values of @xmath276 and @xmath277 .
thus it is sufficient to check the theorem in the case @xmath145 , so that in particular , for each type the theorem reduces to a finite computation . in this case , the statement is trivial ( cf .
examples [ example : gcc : n=1][example : gcc , n=2 ] ) .
in this section , we enumerate the faces of various dimensions in the generalized cluster complexes @xmath3 .
let @xmath339 denote the @xmath276th _ face number _ @xmath3 or , more precisely , the number of @xmath276-element simplices ( i.e. , @xmath340-dimensional faces ) in @xmath3 .
we denote by @xmath341 the number of maximal simplices in @xmath3 .
we note that @xmath342 for any @xmath4 .
[ example : f : n=1 ] in view of example [ example : gcc : n=1 ] , we have @xmath343 [ example : f : n=2 ] by example [ example : gcc , n=2 ] , the complex @xmath157 is an @xmath141-regular graph on @xmath142 vertices .
hence @xmath344 the face numbers @xmath339 satisfy the following recurrences which generalize ( * ? ? ?
* proposition 3.7 ) .
[ prop : fk - recur ] let @xmath276 be a positive integer . for a reducible root system @xmath345 , @xmath346 in particular , @xmath347
if @xmath4 is irreducible , then @xmath348 in particular , @xmath349 formula follows from the fact that @xmath350 is a join of @xmath351 and @xmath352 .
the proof of is analogous to the proof of ( * ? ? ?
* proposition 3.7 ) .
let us count in two different ways the number of pairs @xmath353 where @xmath354 is a @xmath276-element simplex in @xmath3 and @xmath355 .
on the one hand , the number of such pairs is @xmath356 . on the other hand , combining lemma [ lem : m - dihedral ] , theorem [ th : alt def ] , and theorem [ th : restrict ] , we conclude that the colored roots @xmath357 belonging to each @xmath358-orbit @xmath359 contribute @xmath360 to the count ( per negative simple root @xmath361 ) , implying .
it follows from by induction on @xmath0 that @xmath339 is a polynomial in @xmath0 of degree @xmath276 . in this section
, we calculate these polynomials for all irreducible root systems @xmath4 .
( the formulas for reducible cases follow at once . )
it turns out that , with a single exception of @xmath362 , the face numbers @xmath339 factor over the rationals into linear factors .
we present the special case @xmath363 first . for the rest of this section , we assume that the root system @xmath4 is irreducible . let @xmath364 be the _ exponents _ of @xmath4 .
[ pr : product - n ] the number of maximal simplices in @xmath3 is given by @xmath365 proposition [ pr : product - n ] is a special case of the general formula for the face numbers @xmath339 given in theorem [ th : product - f ] below .
the product appearing in the right - hand side of comes up in a variety of contexts related to the combinatorics of the root system @xmath4 , the corresponding coxeter group @xmath12 , and associated lie algebras and lie groups .
we refer the reader to @xcite and references therein ; see also the discussion in section [ sec : h - vectors ] below . one would hope for a uniform formula for the face numbers in the spirit of .
this does not quite happen .
the factorization of @xmath339 into irreducible polynomials in @xmath0 yields a subset of the factors appearing on the right - hand side of together with , in types @xmath11 ( @xmath366 ) , @xmath367 , @xmath368 , @xmath369 , @xmath370 , and @xmath371 , one additional `` mysterious '' factor , which we denote by @xmath372 .
we emphasize that @xmath372 is not a `` fudge factor '' needed to force our formulas to look like .
rather , it is forced upon us by the unique factorizations of the polynomials @xmath339 into irreducibles .
[ th : product - f ] the face numbers of the generalized cluster complex @xmath3 for an irreducible root system @xmath4 are given by @xmath373 here , the `` levels '' of exponents @xmath374 are defined in table [ table : levels ] , with each column showing the exponents of appropriate type , each of them placed in the row corresponding to its level .
the factor @xmath372 is a polynomial in @xmath0 given as follows : * in types @xmath7 , @xmath131 , @xmath375 , and @xmath376 , we have @xmath377 ; * in type @xmath11 , we have : @xmath378 * in types @xmath379 , @xmath380 , @xmath381 , @xmath382 , and @xmath383 , the factors @xmath372 are shown in table [ table : correction_f ] . in particular , for the classical types @xmath7 ,
@xmath131 , and @xmath11 , the face numbers are given by @xmath384 .levels of exponents [ cols="^,^,^,^,^,^,^,^,^,^,^",options="header " , ] the values of @xmath385 for all finite irreducible coxeter groups are listed in table [ table : m(g ) ] .
all of these values are integers .
an explanation of this integrality is provided by the following interpretation of the invariants @xmath385 , due to f. chapoton .
[ prop : plein ] let @xmath12 be a finite irreducible coxeter group with coxeter diagram @xmath386 . then the number of reflections in @xmath12 which do not belong to any proper parabolic subgroup @xmath387 is equal to @xmath385 .
accordingly , the number of all reflections in @xmath12 is equal to @xmath388 proposition [ prop : plein ] can be restated as saying that in a finite root system @xmath4 , the number of positive roots whose simple root expansion has full support ( i.e. , involves every simple root ) is equal to @xmath385 , where @xmath386 is the corresponding coxeter diagram .
the inclusion - exclusion relation uniquely defines the numbers @xmath385 , and is satisfied by the number of reflections not lying in a proper parabolic subgroup .
consequently , it suffices to prove in order to prove the proposition . substituting into both sides of ,
we obtain : @xmath389 implying @xmath390 dividing by @xmath0 and substituting @xmath391 , we get ( the only surviving term on the left - hand side is the one for @xmath392 ) : @xmath393 and follows . it would be interesting to find a direct proof of proposition [ prop : plein ] based on definition [ def : m(g ) ] and one of the combinatorial interpretations of @xmath394 , rather than on calculations involving product formulas and/or recurrence relations .
it follows from that the numbers @xmath385 satisfy the recurrence @xmath395 there is also a more complicated recurrence for these numbers that does not require knowing the coxeter number @xmath2 .
[ lem : m(g)-recur ] let @xmath386 be the coxeter diagram of a finite irreducible coxeter group of rank @xmath396 .
set @xmath397 \scriptstyle h\subset g\\[-.05 in ] \scriptstyle 2\leq |h|\leq n-1\end{array } } m(h)\,.\ ] ] then @xmath398 identities and can be written as @xmath399 substituting into yields .
formulas can be used to recursively compute the values @xmath385 . in view of remark [ rem : m(g)-disconn ]
, the sums in can be restricted to _ connected _ subgraphs @xmath400 and @xmath401 , simplifying the calculations .
the coxeter numbers are then recovered using or , equivalently , .
[ example : k3-m(g ) ] for @xmath386 of type @xmath140 , one has @xmath402 . for a @xmath403-vertex diagram with edge labels @xmath404 ( cf .
examples [ example : euler - n=3 ] , [ example : symmetry - n=3 ] and [ example : recip - n=3 ] ) , with @xmath405 , we get @xmath406 . substituting this into , we obtain @xmath407 matching the results of our earlier calculations .
[ example:4-chain ] now let @xmath408 -\!\!\!-\!\!\!-\\[-.1 in ] \scriptstyle { } \end{array } \ ! } \bullet { \!\begin{array}{c } \scriptstyle{b_2}\\[-.1 in ] -\!\!\!-\!\!\!-\\[-.1 in ] \scriptstyle { } \end{array } \ ! } \bullet { \!\begin{array}{c } \scriptstyle{b_3}\\[-.1 in ] -\!\!\!-\!\!\!-\\[-.1 in ] \scriptstyle { } \end{array } \ ! }
\bullet)\ ] ] be a linear coxeter diagram of rank @xmath409 .
( here @xmath410 are integers @xmath411 . ) using , we compute : @xmath412 \sigma_2&=b_1+b_2+b_3 - 6+\sigma_1\ , .
\end{aligned}\ ] ] the values of @xmath385 and @xmath413 are then obtained using .
the resulting formula for the coxeter number @xmath413 is a cumbersome rational expression in @xmath410 .
a diligent reader might want to verify that , in the special cases where @xmath386 is of types @xmath414 , @xmath415 , @xmath382 , or @xmath383 , the outcomes coincide with the known values of coxeter numbers .
each of the methods described in section [ sec : coxeter - inv ] can , in principle , be applied to a very general coxeter diagram @xmath386 .
the resulting `` fake coxeter invariants '' may or may not be meaningful .
be that as it may , it is tempting to calculate these invariants in the simplest non - finite cases ; to catalogue the infinite types where such invariants are `` nice ; '' and possibly find intrinsic interpretations for them .
one also wonders what the limits of applicability are for each of the methods of section [ sec : coxeter - inv ] , and how their outputs compare if more than one of them works .
examples [ example : euler - n=3 ] , [ example : symmetry - n=3 ] , [ example : recip - n=3 ] , and [ example : k3-m(g ) ] illustrate the methods in the @xmath416 case , and provide the first example in which all of them break down : when @xmath417 , the denominator in the formula for @xmath2 vanishes .
( for @xmath418 , we get @xmath419 , a negative `` fake coxeter number . '' ) here is another such example .
[ example : k4 ] let @xmath420 be the complete graph on @xmath409 vertices , with all @xmath421 edges labeled by @xmath403 .
then each @xmath403-vertex subgraph @xmath422 is of affine type @xmath423 , so gives . then yields @xmath424 , and the denominator in vanishes .
so the method based on lemma [ lem : m(g)-recur ] does not work for @xmath425 .
in fact , none of the methods described above will work in this case .
consequently , the methods will fail for any coxeter diagram that has @xmath426 as an induced subgraph .
another similar example is the @xmath409-cycle with edge labels @xmath427 ( in this order ) .
actually , each of our three main methods ( see sections [ sec : euler - char - method][sec : reciprocity - based ] ) will typically fail in less drastic ways . in the euler characteristic method and the ( general ) reciprocity - based method , solving for @xmath2 can yield a non - constant rational function of @xmath0 ; see and , respectively . in the symmetry - based method
, the polynomial @xmath428 can fail to be symmetric with respect to the involution . in these cases ,
the methods yield an answer which should be viewed with some skepticism . in what follows , references to the failure of one of the three methods mean that @xmath2 is non - constant or that symmetry fails .
( see remark [ rem : recip - special ] for a discussion of the special versions of the reciprocity - based method , described in sections [ sec : reciprocity - based][sec : m(g ) ] . )
we now summarize the results of applying our methods to various coxeter diagrams of infinite type , including the affine diagrams . the only infinite family for which the methods succeed wonderfully is the affine series of type @xmath429 .
[ prop : affinea - fake ] for a coxeter diagram of type @xmath430 with @xmath431 , the fake invariants arising from all methods described in section [ sec : coxeter - inv ] agree , and coincide with the corresponding invariants of type @xmath432 .
that is , @xmath433 and the exponents are @xmath434 .
let the coxeter diagram @xmath386 be a `` simply - laced @xmath13-cycle , '' i.e. , an unoriented graph with the vertices @xmath435 such that the labels of the edges @xmath436 are equal to @xmath403 , and the labels of all other edges are @xmath437 . to prove proposition [ prop : affinea - fake ] , one needs to verify the following statements for this particular diagram @xmath386 and the invariants given by the type-@xmath131 formulas : the recurrence ; the formula for the `` fake euler characteristic ; '' and the inclusion - exclusion formula .
this is a straightforward verification , omitted here .
( note that each connected induced subgraph of @xmath386 is of type @xmath438 . )
the recurrences that determine the fake invariants of type @xmath439 are different from the type-@xmath131 recurrences , so the coincidence of the final answers is somewhat mysterious . on the other hand ,
the fact that the @xmath13-cycle is combinatorially related to the type-@xmath131 cluster complex ( the @xmath13-dimensional cyclohedron ) is well known : there is a _ different _ generalization of associahedra @xcite which produces the cyclohedron in case of an @xmath13-cycle .
it would be interesting to clarify the relationship between the two constructions , and provide a conceptual explanation for proposition [ prop : affinea - fake ] .
our computations suggest that the complete list of connected coxeter diagrams for which all methods described in sections [ sec : euler - char - method][sec : reciprocity - based ] produce _ positive integer _
( fake ) exponents is as follows : * coxeter diagrams of finite irreducible coxeter groups ; * coxeter diagrams of affine type @xmath429 ; * diagrams of rank @xmath403 with the sum of edge labels equal to @xmath209 , @xmath440 , @xmath217 , or @xmath441 .
[ rem : recip - special ] the discussion below does not refer to the methods based on and on lemma [ lem : m(g)-recur ] , respectively .
this is because these methods produce results which are , by construction , the specializations ( at @xmath1 and @xmath391 , respectively ) of the results delivered by the general reciprocity method of section [ sec : reciprocity - based ] based on the recurrences and .
so whenever the general reciprocity method produces a fake coxeter number @xmath2 that is a non - constant rational function in @xmath0 , the specializations of this rational function at @xmath391 and @xmath1 should be viewed with suspicion .
still , it is tempting to mention some instances of coxeter diagrams of affine type where the general reciprocity method fails whereas the @xmath385-based method of section [ sec : m(g ) ] gives a positive integer value of @xmath2 .
they are : * type @xmath442 , with @xmath443 and @xmath444 ( in type @xmath445 , @xmath2 is a fraction ) ; * type @xmath446 , with @xmath447 and @xmath448 ( but irrational exponents ) ; * type @xmath449 , with @xmath450 and @xmath451 .
the fake coxeter number , if defined , is always a rational number ( by design ) , and sometimes even an integer . on the other hand , in most cases , some of the fake exponents are irrational , as the generalized fuss - catalan invariant @xmath452 would not factor into linear polynomials in @xmath0 ( over @xmath453 ) .
still , @xmath452 would sometimes be an integer - valued polynomial , potentially allowing for a meaningful interpretation .
for the affine types other than @xmath429 , calculations give the following results . the types @xmath454 and @xmath455 fall into the @xmath416 case mentioned above , with @xmath456 and @xmath457 , respectively . in type @xmath458 , we get @xmath459 , with exponents @xmath460the same invariants as in type @xmath375 . in type
@xmath461 , we get @xmath443 , with exponents @xmath462 . in type @xmath463 , all methods yield @xmath464 , with exponents @xmath465 .
in type @xmath466 , all methods yield @xmath467 , with exponents @xmath468 .
example [ example:4-chain ] with @xmath469 , @xmath470 , @xmath471 .
the type @xmath472 is the only ambiguous case that we have found .
the reciprocity method gives @xmath473 with the exponents @xmath474 .
this sequence of exponents is not symmetric , and accordingly , the symmetry - based method fails .
however , the answer produced by blindly applying the formulas from the symmetry - based method still agrees with the result of the reciprocity method . the euler characteristic method fails .
for all other affine types , computations up to rank 12 suggest that all methods fail . beyond rank
@xmath403 , our tests succeeded in a few isolated examples of non - affine infinite type , although they did not ordinarily produce integer values of @xmath2 .
here are a few typical examples . for a @xmath409-cycle with one edge labeled @xmath409 and the others labeled @xmath403
, we get @xmath475 with exponents @xmath476 . for a @xmath477-cycle with similar labels ,
all tests fail . for a @xmath409-cycle with one edge labeled @xmath477 and the others labeled @xmath403 ,
all methods yield @xmath478 with exponents @xmath479 .
we also found two more examples with positive integer fake coxeter numbers : @xmath480 -\!\!\!-\!\!\!-\\[-.1 in ] \scriptstyle { } \end{array } \ ! } \bullet { \!\begin{array}{c } \scriptstyle{4}\\[-.1 in ] -\!\!\!-\!\!\!-\\[-.1 in ] \scriptstyle { } \end{array } \ ! } \bullet { \!\begin{array}{c } \scriptstyle{4}\\[-.1 in ] -\!\!\!-\!\!\!-\\[-.1 in ] \scriptstyle { } \end{array } \ ! } \bullet & \qquad & h=98\\ \bullet { \!\begin{array}{c } \scriptstyle{4}\\[-.1 in ] -\!\!\!-\!\!\!-\\[-.1 in ] \scriptstyle { } \end{array } \ ! } \bullet { \!\begin{array}{c } \scriptstyle{3}\\[-.1 in ] -\!\!\!-\!\!\!-\\[-.1 in ] \scriptstyle { } \end{array } \ ! } \bullet { \!\begin{array}{c } \scriptstyle{5}\\[-.1 in ] -\!\!\!-\!\!\!-\\[-.1 in ] \scriptstyle { } \end{array } \ ! } \bullet & \qquad & h=104 \end{array}\ ] ] we note that in all cases we have tried , if all methods work , then they all agree .
we are grateful to christos athanasiadis for valuable comments on the preliminary versions of this paper , for explaining some of his work , and for making available his paper @xcite ( joint with e. tzanaki ) which we cite in section [ sec : recip ] .
christian krattenthaler , david speyer and hugh thomas contributed several suggestions which improved the quality of the final version .
we also thank egon schulte , richard stanley , john stembridge , and andrei zelevinsky for helpful conversations . the first version of this paper was circulated at the workshop `` braid groups , clusters and free probability '' held at the american institute of mathematics on january 1014 , 2005 .
we thank the organizers ( jon mccammond , alexandru nica , and victor reiner ) as well as the rest of participants for stimulating discussions .
our computations of ( fake ) coxeter invariants were done with ` maple ` using john stembridge s package ` coxeter ` .
the ` maple ` source code is currently available at ` http://www.math.lsa.umich.edu/~nreading/papers/gcccc/ ` . c. a. athanasiadis , deformations of coxeter hyperplane arrangements and their characteristic polynomials , _ in : _ arrangements
tokyo 1998 ( ed . m. falk and h. terao , kinokuniya ,
tokyo ) , _ adv .
stud . pure math .
_ * 27 * ( 2000 ) , 126 .
s. fomin and n. reading , root systems and generalized associahedra , lecture notes for the _ ias / park city graduate summer school in geometric combinatorics _
, 2004 , available at ` http://www.math.lsa.umich.edu/~fomin/papers.html ` .
v. reiner and v. welker , on the charney - davis and neggers - stanley conjectures , _
j. combin .
theory ser .
a _ * 109 * ( 2005 ) , 247280 . n. j. a. sloane , the on - line encyclopedia of integer sequences , published electronically at ` http://www.research.att.com/~njas/sequences/ ` , sequence ` a001764 ` . | we introduce and study a family of simplicial complexes associated to an arbitrary finite root system and a nonnegative integer parameter @xmath0 . for @xmath1 ,
our construction specializes to the ( simplicial ) generalized associahedra or , equivalently , to the cluster complexes for the cluster algebras of finite type .
our computation of the face numbers and @xmath2-vectors of these complexes produces the enumerative invariants defined in other contexts by c. a. athanasiadis , suggesting links to a host of well studied problems in algebraic combinatorics of finite coxeter groups , root systems , and hyperplane arrangements .
recurrences satisfied by the face numbers of our complexes lead to combinatorial algorithms for determining coxeter - theoretic invariants .
that is , starting with a coxeter diagram of a finite coxeter group , one can compute the coxeter number , the exponents , and other classical invariants by a recursive procedure that only uses most basic graph - theoretic concepts applied to the input diagram . |
let us state the definition of the minkowski sum of two sets @xmath0 .
the _ minkowski sum _ is @xmath1 in this paper we consider minkowski sums in the plane .
it is well - known , that the minkowski sum of two convex sets is again convex . in the case of convex polygons
it is computed by a simple `` edge merging and slope sorting '' algorithm .
if we consider non - convex polygons , the computation of the minkowski sum may require more complicated algorithms , see @xcite for example . in the cited papers the problem of finding the minkowski sum arised from packing or motion planning problems .
indeed , the set of possible shifts of a region @xmath2 , that intersect another region @xmath3 is the minkowski sum @xmath4 , where minus denotes the reflection w.r.t . the origin .
the most straightforward way to find the minkowski sum of non - convex regions is to partition every non - convex region into convex polygons , calculate minkowski sums of parts , and then take the union @xcite . in some practical applications , where the regions are essentially non - convex , this approach can be too complicated , in such cases it is convenient to use the intuitive `` orbital '' ( or `` sliding '' ) methods , see @xcite .
the latter methods deal with non - convex regions quite well , but the essential ( and hard ) part of these methods is finding `` holes '' in the minkowski sum .
therefore , it is important to give a computable criterion for the minkowski sum to have no `` holes '' . in section [ definitions ]
we define a measure of non - convexity @xmath5 for a simply - connected polygonal region ( simple polygon ) @xmath6 in the plane so that @xmath7 in general and @xmath8 iff @xmath6 is convex .
this measure of non - convexity uses some essential properties of @xmath9 , some other definitions of non - convexity measures valid for spaces of arbitrary ( even infinite ) dimensions are reviewed in @xcite .
another general non - convexity measure based on the path metric in @xmath6 can be found in @xcite .
the main result is stated as follows .
[ aconv - add ] suppose @xmath6 and @xmath10 are simple polygons , and @xmath11 .
then @xmath12 is a simple polygon and @xmath13 this theorem shows that the property @xmath14 is stable under the minkowski sum , and in this case the sum of an arbitrary number of simple polygons is simple . in the proof of theorem
[ aconv - add ] we use the following fact , which has its own value .
it generalizes the separation theorem for convex sets .
[ aconv - sep ] suppose @xmath6 is a simple polygon with @xmath14 .
then for any point @xmath15 there exist an angular region @xmath2 with apex @xmath16 , such that @xmath17 and @xmath18
now we make some definitions and fix some notation .
the sequence of points @xmath19 ( vertices ) and the corresponding segments @xmath20 ( edges ) is called a _
polyline_. we require that the consecutive vertices do not coincide @xmath21 .
for a polyline @xmath22 we call the sequence of vectors @xmath23 the _ shift sequence _ and denote it @xmath24 .
we call a polyline @xmath25 _ closed _ if @xmath26 . in this case
we often index its vertices modulo @xmath27 .
we call a polyline _ simple _ if it has no self - intersections , i.e. its edges may intersect in one point if they are consecutive , otherwise they do not intersect . in a closed polyline
we , of course , allow the first vertex to equal the last one .
we call a compact set @xmath28 a _ simple polygon _ if its boundary is a closed simple polyline .
it is obvious that generally the minkowski sum of two simple polygons can be not simple , speaking informally it can have `` holes '' .
the objective of this paper is to find some sufficient conditions on the polygons that guarantee that the minkowski sum is simple .
we need some more definitions . for two plane vectors
define the _ skew product _
@xmath29 = v_xw_y - v_yw_x.\ ] ] two nonzero vectors @xmath30 are called _ parallel _ if @xmath31 . in this case , if @xmath32 we call the vectors _
opposite_. for two non - opposite vectors
@xmath30 denote @xmath33 the angle between vectors with sign , positive if @xmath34 > 0 $ ] , negative if @xmath34 < 0 $ ] .
if the shift sequence of a polyline @xmath35 does not contain a consecutive pair of opposite vectors , we call @xmath35 _ non - reverse_. let a polyline @xmath35 with shift sequence @xmath36 be non - reverse .
the _ rotation _ of @xmath35 ( the same as rotation of @xmath24 ) is @xmath37 if the polyline has @xmath38 vertices we put @xmath39 .
if the polyline @xmath35 is closed , then ( the indices are modulo @xmath27 ) @xmath40 let @xmath35 be a non - reverse polyline . denote the _ angular convexity _ of @xmath35 @xmath41 where the minimum is taken over subpolylines @xmath42 ( obtained from @xmath35 by removing some vertices from its front and/or its back ) .
let @xmath6 be a simple polygon and @xmath43 be oriented so that @xmath44 .
denote the _ angular convexity _ of @xmath6 @xmath45 where the minimum is taken over simple polylines @xmath42 , oriented along @xmath35 .
the angular convexity of a polygon @xmath6 is illustrated on its slope diagram ( dependance of the slope angle on the anticlockwise parameter ) in figure [ slope - diag ] is the full length . ]
here we formulate and prove some lemmas that generalize the algorithm of finding minkowski sum for convex polygons by edge sorting .
we start from the definition .
denote the concatenation of sequences @xmath46 and @xmath47 by @xmath48 .
denote the first element of nonempty @xmath46 as @xmath49 , denote the sequence @xmath46 without its first element by @xmath50 .
[ sorted - sum ] suppose two polylines @xmath51 and @xmath52 are non - reverse , let their shift sequences be @xmath53 and @xmath54 .
suppose also that @xmath55 and @xmath56 .
define the _ sorted sum _ @xmath57 as follows .
let the first vertex be @xmath58 , let the shift sequence @xmath59 be formed by the following rule : \1 ) start from @xmath60 ; \2 ) if @xmath61 and @xmath62 then put @xmath63 and quit ; \3 ) if @xmath61 then put @xmath64 and quit ; \4 ) if @xmath62 then put @xmath65 and quit ; \5 ) if @xmath66 , then put @xmath67 , @xmath68 and go to step @xmath38 ; \6 ) if @xmath69 , then put @xmath70 , @xmath71 and go to step @xmath38 .
informally , we merge the shifts from two sequences in such a way , that from the two possibilities to choose the next shift we always choose the `` rightmost '' one , preferring the first sequence if the directions coincide .
the condition that the starts and the ends of the shift sequences have the same direction , and the rotations are the same , can be relaxed in general , but it is crucial in lemmas [ sorted - sum - correct ] , [ sorted - sum - rot ] , and [ sorted - sum - aco ] .
an example of a sorted sum is shown in figure [ sorted - sum - pic ] .
first , we show how the sorted sum is related to the minkowski sum .
[ sorted - sum - param ] let the polylines @xmath35 and @xmath72 be as in definition [ sorted - sum ] and let them have piece - wise linear parameterizations @xmath73\},\quad q = \{q(t ) : t \in [ 0 , 1]\}.\ ] ] then there are non - decreasing piece - wise linear functions ( onto ) @xmath74 \to [ 0,1]$ ] such that @xmath75 is a parameterization of @xmath76 , in particular @xmath77 ( the minkowski sum of @xmath35 and @xmath72 as subsets of the plane ) .
let @xmath78 , @xmath79 , then @xmath80 .
without loss of generality we may assume that the parameterization @xmath81 maps points @xmath82 to the vertices of @xmath35 , and @xmath83 maps points @xmath84 to the vertices of @xmath72 .
the functions @xmath85 and @xmath86 will be continuous , piece - wise linear .
we will construct them so that @xmath85 on a segment @xmath87 $ ] ( @xmath88 ) is either constant or linearly increases by @xmath89 , and @xmath86 on the same segment is either constant or linearly increases by @xmath90 . by definition
put : if the @xmath91-th shift in @xmath92 is taken from @xmath24 , then @xmath85 increases and @xmath86 remains constant , if it is taken from @xmath93 , @xmath85 remains constant and @xmath86 increases .
in the first case denote @xmath94 , where @xmath95 , then for @xmath96 and @xmath97 @xmath98 and @xmath99 the above equations mean that the points @xmath100 and @xmath101 are two consecutive vertices of the polyline @xmath102 , and for @xmath103 $ ] the point @xmath104 is on the corresponding segment of @xmath102 .
the second case is considered similarly . [
sorted - sum - correct ] suppose @xmath105 are as in definition [ sorted - sum ] . if @xmath106 , then @xmath57 is non - reverse .
assume the contrary : its consecutive shifts @xmath107 and @xmath108 ( indexed as they were in @xmath35 and @xmath72 ) are opposite .
take the sequences @xmath109 and @xmath110 from the definition in the state before adding @xmath107 and @xmath108 to @xmath111 , so @xmath112 , @xmath113 . since @xmath107 and @xmath108 are opposite @xmath114 , for some @xmath115 . since @xmath107 comes before @xmath108 in @xmath92 , then @xmath66
thus @xmath116 if @xmath117 , then by the construction @xmath118 , so @xmath119 , which can not be the case . if @xmath120 , then @xmath121 , which contradicts with the angular convexity condition .
[ sorted - sum - rot ] suppose @xmath105 are as in definition [ sorted - sum ] . if @xmath106 , then @xmath122 .
denote @xmath123 . to prove @xmath124 , it is sufficient to prove the following statement : if on some cycle of the construction the last element of @xmath111 is from @xmath24 , then @xmath125 , if the last element of @xmath111 is from @xmath93 , then @xmath126 .
let us prove it by induction .
if @xmath127 the statement is true . let @xmath128 , if @xmath16 and @xmath129 are shifts from @xmath24 , then @xmath130 by the inductive assumption .
the same is true if @xmath131 .
it is left to consider the case @xmath132 .
consider two cases depending on whether @xmath109 is empty or not .
case 1 : @xmath109 is non - empty . by the construction of the sorted sum
we have : @xmath133 and since @xmath134 , then we conclude that @xmath135 by the inductive assumption @xmath136 .
the difference @xmath137 should be a multiple of @xmath138 .
but its absolute value is less than @xmath138 , since @xmath139 and @xmath140 .
. case 2 : @xmath109 is empty .
then , similar to the previous case : @xmath142 and the difference @xmath143 should be a multiple of @xmath138 .
since @xmath139 and @xmath144 , this difference should be zero .
[ sorted - sum - aco ] suppose @xmath105 are as in definition [ sorted - sum ] . if @xmath106 , then @xmath145
. denote @xmath146 , @xmath147 .
assume the contrary : for some subsequence @xmath46 of the sequence @xmath92 we have @xmath148 .
it is equivalent to the following statement : the sequence @xmath111 in the construction of sorted sum takes two values @xmath149 and @xmath150 ( in this order ) so that @xmath151 put @xmath152 , @xmath153 , and denote the corresponding values of @xmath109 and @xmath110 by @xmath154 .
if @xmath16 and @xmath129 are from @xmath24 then ( see the proof of lemma [ sorted - sum - rot ] ) @xmath155 and by the assumption @xmath156 if @xmath157 then @xmath158 and therefore , @xmath159 , which is a contradiction with @xmath160 . if @xmath16 and @xmath129 are both from @xmath93 , the same contradiction is obtained
. now assume that @xmath16 is from @xmath93 , @xmath129 is from @xmath24 , then @xmath161 and by the assumption @xmath162 by the construction @xmath163 ( when @xmath16 was added to @xmath111 ) , and @xmath164 if @xmath157 then @xmath165 and therefore , @xmath166 , which is a contradiction with @xmath160 .
the case @xmath16 is from @xmath24 , @xmath129 is from @xmath93 is considered the same way .
in this section we consider a polyline @xmath35 that has self - intersection . then we convert it to a polyline without self - intersections and try to describe how the rotation of @xmath35 changes .
first we need a definition .
a polyline @xmath35 is said to be in _ general position _ , if \1 ) all its vertices are distinct ; \2 ) any two edges intersect in at most one point , the common point being either the common vertex , or in the relative interiors of both edges ; \3 ) any three edges do not have a common point .
it is clear that , by arbitrarily small movements of the vertices , a polyline can be put to a general position .
let a polyline @xmath35 be in general position .
suppose it is given by a piece - wise linear parameterization @xmath167\}$ ] and has a self - intersection @xmath168 for some @xmath169 .
call the _ loop removal _ the transform that replaces @xmath35 by the concatenation of polylines @xmath170\}$ ] and @xmath171\}$ ] .
it is clear that generally @xmath172 changes by a multiple of @xmath138 after the loop removal .
the following lemma tells more .
[ loop - removal ] let @xmath35 be a polyline in general position . suppose that the subpolyline of @xmath35 between @xmath173 and @xmath174 has rotation @xmath175 .
then the rotation @xmath172 can not increase after loop removal .
let @xmath173 lie on the edge with direction @xmath176 , @xmath174 line on another edge with direction @xmath177 .
denote the subpolyline between @xmath173 and @xmath174 by @xmath10 .
if we identify @xmath173 and @xmath174 in @xmath10 we obtain a closed polyline , so @xmath178 since @xmath179 and @xmath180 , we obtain @xmath181 .
after removal of @xmath10 the rotation of the new polyline differs from @xmath172 by @xmath182 , so it does not increase .
[ loop - elimination ] let @xmath35 be a polyline in general position with @xmath183 . then we can remove all loops in @xmath35 so that its rotation does not increase .
consider all the self - intersections in @xmath35 , choose the first ( w.r.t .
the parameterization ) such point @xmath184 , and remove the loop between @xmath173 and @xmath174 . then continue such steps until all the self - intersections are removed .
it is clear that the removed segments of parameterization do not intersect , each removed subpolyline had angular convexity @xmath175 and by lemma [ loop - removal ] the angular convexity never increased .
an example of removing all loops is shown in figure [ loop - removal - pic ] .
let us give some definitions . a convex hull of two rays @xmath185 and @xmath186 with common apex is called an _ angular region_. for an angular region @xmath2 we denote its _ angular measure _ @xmath187 .
we generalize the notion of rotation to piece - wise smooth curves .
let @xmath188\}$ ] be a smooth curve with non - zero derivative .
then the unit tangent @xmath189 is well defined and can be considered as a continuous map @xmath190\to s^1 $ ] to the unit circle .
consider the universal cover @xmath191 given by the anti - clockwise angle .
the map @xmath192 is lifted continuously to a map @xmath193 \to \mathbb r$ ] so that @xmath194
. then the _ rotation number _ of @xmath195 is @xmath196 .
let @xmath188\}$ ] be a piece - wise smooth curve , composed as a concatenation of smooth curves @xmath197 .
denote the values of the parameter , where @xmath198 is changed to @xmath199 by @xmath200 .
we require the vectors @xmath201 and @xmath202 ( left and right derivative ) to be non - opposite . then the _ rotation number _ of @xmath195 is @xmath203 it is clear that if the curve @xmath195 is approximated by a polyline @xmath35 with small enough step , then @xmath204 .
the angular convexity of a curve is defined similarly .
now we are going to prove theorem [ aconv - sep ] .
the polyline @xmath43 can be approximated by a smooth curve as follows : let @xmath205 be its vertex with nonzero angle @xmath206 , in a small neighborhood of @xmath205 we replace the union of two small equal segments of @xmath35 ( the first segment is from @xmath205 back along @xmath207 , the second is from @xmath205 forth along @xmath207 ) by a circular arc @xmath208 , so that the tangent becomes continuous along the resulting curves .
if the segments are taken small enough , then the new boundary @xmath209 has no self - intersections .
if we denote by @xmath210 the region , bounded by @xmath209 , then the ( tested for the separation property ) point @xmath16 still lies outside @xmath209 for close enough approximation .
note that @xmath211 by the construction , so it is clear that @xmath212 .
for any two points @xmath213 let us denote @xmath214_{p'}$ ] the subcurve of @xmath209 , oriented along @xmath209 , that starts at @xmath215 and ends at @xmath216 .
consider the following functions of a point @xmath217 @xmath218_{p'},\quad \gamma_-(p ) = \min_{q\in p ' } \operatorname{rot}[q , p]_{p'}.\ ] ] it is clear that they are continuous and @xmath219 .
take the rays @xmath220 and @xmath221 with apex @xmath81 so that @xmath222 note that the rays @xmath220 and @xmath221 point outside or tangentially to @xmath210 , and the angle between them is at least @xmath223 .
denote the angular region @xmath224 .
we are going to show that @xmath225 .
assume the contrary , in this case @xmath209 has to pass through the interior of @xmath226 .
the situation is outlined in figure [ separation - pic ] . and the angular region @xmath226 .
] let us move the point @xmath83 from @xmath81 along @xmath209 .
note the first time @xmath83 gets inside @xmath227 .
suppose @xmath83 intersects the ray @xmath220 , then the rotation of the closed ( piece - wise smooth ) non - self - intersecting curve @xmath228_{p ' } { \circ}[q , p]$ ] ( the last segment is a straight line segment ) is @xmath229 , and it is clear that @xmath230_{p'}$ ] should be less than @xmath231 , which is a contradiction .
the only possibility left is that @xmath83 ( when moving from @xmath81 along @xmath209 ) gets into @xmath226 across @xmath221 .
let @xmath83 be the first such point on @xmath221 , in this case we consider the closed non - self - intersecting curve @xmath232_{p ' } { \circ}[q , p]$ ] that bounds a region @xmath10 .
if we move a point @xmath233 backwards from @xmath81 along @xmath209 , then it should go out of @xmath10 , in fact it has to intersect @xmath234 $ ] and get into @xmath227 .
let @xmath233 be the first such point on @xmath234 $ ] , in this case the closed simple curve @xmath235_{p ' } { \circ}[p , s]$ ] has rotation @xmath229 , and @xmath236_{p ' } < \gamma_-(p)$ ] , which is a contradiction again .
thus the theorem is proved for points @xmath16 , close enough to @xmath209 , now we prove it for any
@xmath16 outside @xmath209 . for @xmath217 denote @xmath237 the direction of the bisector of @xmath226 , and @xmath238 the unit direction of @xmath239 .
they give two maps @xmath240 and @xmath241 .
the map @xmath242 is homotopic to the gauss map of @xmath209 , so @xmath243 .
the map @xmath244 is homotopic to a constant map ( the homotopy is obtained by moving @xmath16 to infinity without crossing @xmath209 ) , so @xmath245 .
thus @xmath242 and @xmath244 should coincide on some @xmath81 . the other ( more elementary ) way to prove
the coincidence is to note that the rotation of the vector @xmath237 when @xmath81 moves along @xmath246 is @xmath138 , and the rotation of @xmath238 is zero .
it is easy to see that for such @xmath81 the point @xmath16 lies inside @xmath226 , and the angular region @xmath247 gives what we need .
denote @xmath248 . consider the set @xmath249 .
later we prove that it is simply connected ( has no _ holes _ , i.e. bounded connected components of the complement @xmath250 ) , but now we assume that the holes may exist and denote its outer piece of boundary by @xmath3 ( the boundary of the component of @xmath251 of @xmath250 ) .
we are going to show that @xmath252 .
suppose that for some two points @xmath253 we have @xmath230_b = \delta < \gamma$ ] ( the notation is the same as in the proof of theorem [ aconv - sep ] ) .
we may assume that @xmath81 and @xmath83 lie inside the edges of @xmath3 .
in this case from the definition of the minkowski sum we deduce that @xmath254 one of the points @xmath255 and @xmath256 ( let it be @xmath255 ) lies on an edge ( since @xmath81 lies on an edge ) of @xmath257 with outer normal @xmath242 .
the other point @xmath256 either lies on an edge of @xmath10 with the same normal , or is a vertex of @xmath10 . in the latter case the inner product @xmath258 on some small enough neighborhood of @xmath256 in @xmath10
attains its maximum exactly at @xmath256 , in this case we can insert a `` virtual edge '' to @xmath10 of length zero at point @xmath256 , so that it has normal @xmath242 .
this virtual edge does not affect the rotation of any subpolyline of @xmath259 . in the sequel
we assume that @xmath260 lie on ( possibly virtual ) edges , the edges of @xmath261 have the same direction , and the edges of @xmath262 have the same direction .
note that if @xmath263 , then for any subpolyline @xmath264_{\partial k}$ ] its rotation is less than @xmath265 , otherwise the subpolyline @xmath266_{\partial k}$ ] would have the rotation @xmath267 .
now we see that @xmath268_{\partial k } = \delta + 2\pi n$ ] , @xmath115 . from the angular convexity condition and the above note
we deduce that @xmath268_{\partial k } = \delta + 2\pi$ ] .
so we have @xmath269_{\partial k } = \operatorname{rot}[d , b]_{\partial l } = -\delta.\ ] ] now consider the sorted sum @xmath270_{\partial k}+_s [ d , b]_{\partial l}$ ] . by lemma [ sorted - sum - rot ] and lemma [ sorted - sum - aco ] we have @xmath271 , @xmath272 .
lemma [ sorted - sum - param ] shows that @xmath273 .
the situation is illustrated in figure [ sum - aco - pic ] .
note that @xmath276 , @xmath277 is connected , so we may put @xmath46 to general position so that it remains in @xmath278 and the only points of @xmath279 are @xmath81 and @xmath83 , we assume that the rotation of @xmath46 under this perturbation changes by arbitrarily small value , and @xmath280 remains @xmath175 .
then we eliminate loops on @xmath46 by lemma [ loop - elimination ] and obtain a simple polyline @xmath281 with @xmath282 . from lemma [ sorted - sum - param ] we know that @xmath283 goes from @xmath83 to @xmath81
. moreover , by the construction of the sorted sum its edge directions at @xmath81 and @xmath83 coincide ( up to some small change on passing to general position ) with the edge directions of @xmath232_b$ ] ( edges may be virtual ) .
so the closed polyline @xmath284_b$ ] has rotation number @xmath285 and does not have self - intersection . hence @xmath286 , and the region that it bounds is to the right of @xmath195 .
but @xmath287 , and along the polyline @xmath232_b$ ] , the set @xmath278 is to the left of @xmath232_b$ ] ( @xmath3 is oriented anti - clockwise ) , which is a contradiction .
thus we have proved that @xmath288 .
now we are going to prove that @xmath278 actually has no holes .
assume the contrary : a point @xmath289 is not connected to @xmath251 in the complement of @xmath278 .
let us perturb the vertices of @xmath6 and @xmath10 by values @xmath290 .
if @xmath291 is chosen small enough then @xmath292 condition is kept , and @xmath16 remains outside @xmath278 . moreover ,
if @xmath278 does not have a hole containing @xmath16 after perturbation , then the complement @xmath250 contains by theorem [ aconv - sep ] an angular region @xmath2 with apex @xmath16 and bounded from below measure @xmath293 .
hence for small enough @xmath291 the region @xmath2 can not be `` blocked '' by returning to the non - perturbed @xmath6 and @xmath10 , we have a contradiction .
thus we have proved that for small enough ( @xmath290 ) perturbations of @xmath6 and @xmath10 their sum @xmath278 still has a hole .
i.e. the property of having a hole is open .
we are going to consider the parameterized family of polygons @xmath294 note that the edges of @xmath295 are contained in the lines ( we define lines by two points ) @xmath296 where @xmath297 and @xmath298 are vertices . by perturbing @xmath6 and @xmath10
we may require that all such lines ( there is a finite number of them ) have different directions .
we also require that when @xmath299 changes in the range @xmath300 , no @xmath301 of the lines meet at a single point , this can be achieved by perturbation .
if the parameter @xmath299 is close enough to zero , then the polygon @xmath295 has the following structure : near an edge of @xmath6 it has a `` long '' edge , and near a vertex of @xmath6 it may be complicated .
but since @xmath6 is an angular region @xmath302 ( possibly concave ) in some neighborhood of a vertex @xmath303 , then @xmath295 in @xmath304 ( and for small enough
@xmath299 ) is a union of a family of translates of @xmath302 , that is starshaped .
hence , for small enough @xmath299 , locally @xmath295 has the same topology as @xmath6 , and can not have holes in global .
now consider @xmath305 it is already proved that @xmath306 , assume @xmath307 . since the `` has hole '' property is open , @xmath308 has no holes .
consider what could happen if we increase @xmath299 by a small value @xmath290 . from the general position assumption @xmath295
may be changed by adding some extra edge instead of a vertex , but this can not give a hole .
hence @xmath309 and the proof is complete .
burkea , r.s.r .
helliera , g. kendalla , g. whitwell .
complete and robust no - fit polygon generation for the irregular stock cutting problem . //
european journal of operational research , 179(1 ) , 2007 , 2749 . | in this paper a measure of non - convexity for a simple polygonal region in the plane is introduced .
it is proved that for `` not far from convex '' regions this measure does not decrease under the minkowski sum operation , and guarantees that the minkowski sum has no `` holes '' . |
in classical general relativity a causal particle inside the event horizon is inevitably pulled towards the center of the black hole , at least until it reaches the cauchy horizon representing a barrier to predictability @xcite . in particular , no classically relativistic particle , be it of positive or vanishing rest - mass , can escape from the black hole interior .
however , hawking @xcite has shown that photons in effect can leave the black hole if a quantized photon field in the curved spacetime is supposed , cf .
@xcite . in general , quantization of spin - weighted waves in a curved spacetime
yields the notion of spin particles in a gravitational field . as such
, spin - weighted waves are the basis for semiclassical quantum gravity .
their properties and behavior in the outer space of kerr black holes have been extensively studied @xcite , and could in part even be extended to charged kerr - newman black holes @xcite . by the symmetries of these spacetimes the equations turn out to be separable in special coordinate frames , thus being mathematically tractable to a certain extent . in the framework of classical general relativity
the examination of waves in the outer space of a black hole seemed sufficient , since the event horizon acts like a perfect semi - permeable membrane , letting in any form of energy and matter but allowing none to get out . in the present paper
we extend these considerations to the region beyond the event horizon .
topologically , this means nothing particular , since the spacetime curvature remains finite in this domain and the event horizon is nothing more than a coordinate singularity .
as expected , the event horizon causes a singular repulsive potential in the respective wave equations .
however , more detailed examination shows the remarkable , and to our knowledge yet unmentioned , property of the potential barrier to be singular enough to prevent nearly all kinds of spin - weighted waves from tunneling through it unless electromagnetic waves .
our analysis below shows that in fact a photon inside an uncharged rotating black hole may tunnel through the event horizon to the outside region , but a graviton , a neutrino or an electron may not . if the black hole is non - rotating , no particle at all can tunnel out of it . in case of a rotating and electrically charged black hole
it remains still unknown what kind of particle may tunnel out , it is only sure that electrons and neutrinos _ may not_. the present paper is organized as follows . in section [ sec - kn ]
we introduce the basic notation and properties of kerr - newman spacetimes . in section [ sec - spin - waves ]
we analyze the equations for massless waves in kerr geometry and present a proof that only photons of certain discrete frequencies may tunnel out of a rotating but non - extremal kerr black hole .
the case of extremal rotation is considered in section [ sec - rotation ] , the mathematically not completely tractable case of massless waves in a charged reissner - nordstrm spacetime in section [ sec - rn ] , and the dirac equation in a general kerr - newman spacetime in section [ sec - dirac ] .
finally , in section [ sec - discussion ] we sum up the results and discuss them .
suppose a kerr - newman black hole with the three real parameters @xmath0 , @xmath1 , and @xmath2 .
they are related to the mass @xmath3 ( in kg ) , the angular momentum @xmath4 ( in kg m@xmath5 s@xmath6 ) and the electrical charge @xmath7 ( in kg@xmath8 m@xmath9 s@xmath6 ) by the relations @xmath10 here @xmath11 is the gravitational constant , and @xmath12 the speed of light .
the non - vanishing contravariant components @xmath13 of the metric tensor in boyer - lindquist coordinates @xmath14 , , @xmath15 , @xmath16 , @xmath17 , @xmath18 are then @xmath19 \displaystyle g^{t\varphi } = { ( 2mr - q^2 ) a \over \rho \bar\rho \delta},\quad g^{\varphi\varphi } = - { \delta - a^2 \sin^2\theta \over \rho \bar\rho \delta \sin^2\theta } , \end{array}\ ] ] ( @xmath20 ) , where @xmath21 with the _ event horizon _ @xmath22 and the _ cauchy horizon _ @xmath23 given by @xmath24 only the points of the set @xmath25 have infinite riemann curvature and thus locate the curvature singularity of the spacetime . for @xmath26
, it in fact forms a ring of radius @xmath27 in the equatorial plane @xmath28 , cf . @xcite .
because the _ cosmic censorship hypothesis _ forbids naked singularities , the event horizon @xmath22 must necessarily exist .
thus the square root of ( [ horizonte ] ) must be real , i.e. @xmath29
let be @xmath30 , and define @xmath31 @xmath32 \{0 , @xmath33 , @xmath34 , @xmath35 } as the _ spin - weight_. then the sourcefree perturbation equations for scalar ( @xmath31 @xmath36 0 ) , two - component neutrino ( @xmath31 @xmath36 @xmath33 ) , electromagnetic ( @xmath31 @xmath36 @xmath34 ) , and gravitational fields ( @xmath31 @xmath36 @xmath35 ) are given by wave equations which by the the symmetries of the kerr spacetime , viz .
stationarity and axialsymmetry , admit the separable solutions @xmath37 with the constants @xmath38 and @xmath39 @xmath32 @xmath40 . here
@xmath41 obeys the radial equation @xmath42 \displaystyle + 4 { \mathrm{i } } s\omega r - \lambda \bigg ) r_s = 0 , \label{radial equation}\end{aligned}\ ] ] and @xmath43 solves the angular equation @xmath44 \displaystyle - { ( m + s \cos\theta)^2 \over \sin^2\theta } - s(s-1 ) + a \bigg ) s_s = 0 , \label{angular equation}\end{aligned}\ ] ] with @xmath45 and @xmath46 , cf .
@xcite the constants @xmath47 , @xmath48 @xmath32 @xmath49 are separation constants obtaining some discrete values depending on the boundary conditions of @xmath43 .
( for details see @xcite , 2.1 . )
to be more explicit , @xmath47 @xmath36 @xmath50 @xmath36 @xmath51 @xmath52 @xmath53 @xmath54 @xmath55 , where @xmath51 @xmath36 @xmath56 @xmath54 @xmath57 @xmath52 @xmath58 , with @xmath59 @xmath60 @xmath61 and @xmath59 @xmath32 @xmath62 for integer spin @xmath31 , @xmath59 half - odd integer for @xmath63 @xmath36 1/2 , cf .
@xcite eq .
( 3.10 ) .
[ satz - potential ] let be given the operator @xmath64 @xmath65 @xmath66 @xmath65 , @xmath67 with @xmath68 with the functional transformation @xmath69 thes radial equation ( [ radial equation ] ) is equivalent to @xmath70 @xmath36 @xmath71 . _
_ first we see that with the functions @xmath72 given by the transformation @xmath73 ( i.e. , @xmath74 @xmath36 @xmath75 and @xmath76 @xmath36 @xmath77 ) the radial differential equation ( [ radial equation ] ) is equivalent to @xmath78 , with the differential operator @xmath79 where @xmath80 and @xmath81 ( cf .
@xcite , eqs .
( 3 ) , ( 96 ) , ( 97 ) in chapter 8) .
with @xmath82 we see that @xmath83 since d@xmath84/d@xmath16 @xmath36 @xmath85 , we have @xmath86 by elementary means ( e.g. @xcite , p. 323 ) we deduce from this result that , by the transformation @xmath87 the equation @xmath78 is equivalent to @xmath88 .
[ lemma - tunneling ] let @xmath89 be a solution of the differential equation @xmath70 @xmath36 0 .
then the asymptotic behavior of @xmath89 is given by @xmath90 \phantom{\scriptstyle r \to \infty}\\[-1.5ex ] \longrightarrow\\[-1.5ex ] { \scriptstyle r \to \infty } \\[-1.5ex ] \end{array } a_{\mathrm{in}}\ , r^s\ , { \mathrm{e}}^{-{\mathrm{i } } \omega r } + a_{\mathrm{out}}\ , r^{-s}\ , { \mathrm{e}}^{{\mathrm{i } } \omega r},\ ] ] @xmath90 \phantom{\scriptstyle r \to m}\\[-1.5ex ] \longrightarrow\\[-1.5ex ] { \scriptstyle r \to m } \\[-1.5ex ] \end{array } b_{\mathrm{in}}\ , { \mathrm{e}}^{-{\mathrm{i } } k r } + b_{\mathrm{out}}\ , { \mathrm{e}}^{{\mathrm{i } } k r } \label{u - asymptotic}\ ] ] with the complex amplitudes @xmath91 , @xmath92 , @xmath93 , @xmath94 @xmath32 @xmath95 and the complex exponent @xmath96 given by ( cf .
figure [ fig - sqrt ] ) @xmath97 ^ 2 \over ( m^2-a^2)^2 } + { |s| - 1 - \lambda - 4 { \mathrm{i } } s \omega m \over m^2 - a^2 } } \!.
\label{k - def}\ ] ] ( 0,0 ) of a complex number @xmath98 in the complex plane .
if @xmath98 @xmath36 @xmath99 e@xmath100 , then @xmath101 @xmath36 @xmath102 e@xmath103 [ fig - sqrt],title="fig:",width=188 ] ( 48,36 ) ( 20.5,26.5)(0,0)i ( 33,13)(0,0)1 ( 2,26)(0,0)@xmath98 ( 28,30)(0,0)@xmath101 ( 16 , 3)(0,0)@xmath104 ( 20.5,19.5)(0,0)@xmath105 ( 31,23)(0,0)@xmath106 _ proof . _ first we consider the behavior for @xmath16 @xmath66 @xmath107 . in this case , @xmath108 @xmath36 @xmath109 @xmath52 @xmath110 with @xmath109 @xmath36 @xmath111 @xmath54 @xmath112
. thus asymptotically @xmath70 @xmath66 @xmath113 @xmath52 @xmath114 , which is solved by the first limit ( [ u - asymptotic ] ) as is seen by direct computing . on the other hand
, we calculate by @xmath115 @xmath36 @xmath116 @xmath54 @xmath117 that @xmath108 @xmath66 @xmath118 as @xmath16 @xmath66 @xmath0 . hence , @xmath70
@xmath66 @xmath119 @xmath52 @xmath120 , which is solved by the second limit in ( [ u - asymptotic ] ) .
note that the complex conjugate of @xmath121 in ( [ a - s - def ] ) equals the radial operator @xmath122 of the waves of spin @xmath123 , @xmath124 @xmath36 @xmath122 .
in particular , @xmath125 @xmath36 @xmath126 @xmath36 0 , and thus @xmath127 the advantage of the transformation in theorem [ satz - potential ] is the fact that the wronskian of two solutions of the differential equation @xmath70 @xmath36 0 is constant .
in particular , the behavior of two _ independent _ solutions at the boundaries of the domain of definition determines the reflexion and transmission coefficients of the potential @xmath128 , cf . @xcite . by the peeling theorem of newman and penrose @xcite ,
for a null field of spin weight @xmath31 the ingoing and outgoing solutions differ in magnitude by the factor @xmath129 .
accordingly , the solutions @xmath41 show the asymptotic behavior @xmath130 this is in accordance with the asymptotic behavior of @xmath89 , using the transformation ( [ u - s - def ] ) and @xmath84 @xmath36 @xmath131 . [ satz - tunneling ] the only massless particles that can tunnel out of a non - extremal kerr black hole are photons with the discrete frequencies @xmath132 their wave numbers @xmath133 at @xmath16 @xmath36 @xmath0 are given by @xmath134 depending on the angular quantum number @xmath39 @xmath32 @xmath40 .
the amplitudes of a photon wave with the asymptotic behavior ( [ u - asymptotic ] ) is determined by the conservation law @xmath135 _ proof . _ a kerr black hole with mass @xmath0 and rotation parameter @xmath1 is non - extremal if and only if @xmath136 . for @xmath137 @xmath36 @xmath138 as in ( [ omega - m ] ) we compute straightforwardly the functions @xmath139
@xmath36 @xmath140 as @xmath141 inserting @xmath138 into the expression for @xmath96 in ( [ k - def ] ) yields @xmath142 ^ 2 + { |s|-1 - \lambda - 4 { \mathrm{i } } s \omega_m m \over m^2 - a^2 } .\ ] ] by theorem [ satz - potential ] a massless particle of spin - weight @xmath31 is given as a wave function @xmath89 satisfying @xmath70 @xmath36 0 . since the wkb approximation is applicable if @xmath108 is varying slowly enough , @xmath143 we may use it for a solution @xmath144 in a region close enough to the event horizon at @xmath16 @xmath36 @xmath145 and far enough from the turning points @xmath108 @xmath36 0 , see @xcite 8.2 , say @xmath146 @xmath147 @xmath16 @xmath147 @xmath148 .
in this regime , the transmission coefficient @xmath149 is given by @xmath150 \label{transmission - coefficient}\ ] ] the integral diverges if @xmath108 has a pole at @xmath16 @xmath36 @xmath145 of order not smaller than 2 , but it attains a finite value for a pole of order 1 since @xmath151 there are three terms in @xmath108 which may result in a pole of order 2 at @xmath16 @xmath36 @xmath145 , namely @xmath152 the third term , the imaginary part , vanishes ( in the non - extremal case , where @xmath145 @xmath153 @xmath0 ) if and only if @xmath154 @xmath36 0 or @xmath31 @xmath36 @xmath71 .
the first two terms , the real part , vanish at @xmath16 @xmath36 @xmath145 if and only if @xmath155 @xmath36 @xmath156 @xmath54 @xmath157 @xmath158 .
since @xmath108 is never purely real and negative , a pole both in the real part as well as in the imaginary part of @xmath108 is also a pole of re@xmath159 , cf .
figure [ fig - sqrt ] .
thus in ( [ q - s - poles ] ) the real and the imaginary terms of @xmath108 resulting in a pole of order 2 have to vanish simultaneously at @xmath16 @xmath36 @xmath145 , i.e. @xmath154 @xmath36 0 and @xmath63 @xmath36 1 .
this means that the integral @xmath160 exists only for photons with frequencies @xmath161 .
in particular , with ( [ k - m ] ) we obtain @xmath162 @xmath36 @xmath163 .
this leads to the functions @xmath164 @xmath165 now let @xmath89 be a solution of @xmath70 @xmath36 0 with the asymptotic behavior ( [ u - asymptotic ] )
. then @xmath166 solves @xmath167 @xmath36 0 and is therefore a wave of spin @xmath123 , too .
by @xmath168 @xmath36 @xmath121 it follows that @xmath169 @xmath36 0 , i.e. @xmath170 is a wave with spin @xmath171 .
hence the wronskian of the two solutions of @xmath172 @xmath36 0 is given asymptotically by @xmath173 & \begin{array}{c } \\[-4.25ex ] \phantom{\scriptstyle r \to \infty}\\[-1.5ex ] \longrightarrow\\[-1.5ex ] { \scriptstyle r \to \infty } \\[-1.5ex ] \end{array } & 2 { \mathrm{i}}\omega \ , ( |a_{\mathrm{out}}|^2 - |a_{\mathrm{in}}|^2 ) , \nonumber \\ * \left[u_s , u_{-s}^*\right ] & \begin{array}{c } \\[-4.25ex ] \phantom{\scriptstyle r \to m}\\[-1.5ex ] \longrightarrow\\[-1.5ex ] { \scriptstyle r \to m } \\[-1.5ex ] \end{array } & 2 { \mathrm{i } } \ { \mathrm{re}}\ ,
k\ ( |b_{\mathrm{out}}|^2 - |b_{\mathrm{in}}|^2 ) , \label{[u , u]-asymptotic}\end{aligned}\ ] ] since the wronskian is constant for an equation of the form @xmath174 @xmath36 0 , the assertion follows .
an extremal kerr black hole is given by the maximally possible angular momentum @xmath27
@xmath36 @xmath0 . in this case
, we have @xmath145 @xmath36 @xmath175 @xmath36 @xmath0 , and the functions @xmath84 and @xmath140 in ( [ rho - delta ] , [ k - def ] ) simplify to @xmath140 @xmath36 @xmath176 @xmath54 @xmath117 , @xmath84 @xmath36 @xmath158 . according to ( [ q - s - def ] ) this yields @xmath177
now @xmath108 has a pole of order 4 at @xmath178 . to apply a similar consideration as in the proof of theorem [ satz - tunneling ]
, we have to arrange the nominators to obtain a zero of order 3 . for convenience
, we again consider the imaginary and the real part of @xmath108 separately .
first we note that @xmath115 must vanish to avoid a pole of order 3 , i.e. @xmath115 @xmath36 0 , or equivalently @xmath179 with this , @xmath140 simplifies to @xmath140 @xmath36 @xmath180 , and therefore @xmath181 i.e. , the imaginary part of @xmath108 has a pole of order 1 .
regarding the real part , we insert our derived expression for @xmath140 to obtain @xmath182 at @xmath16 @xmath36 @xmath0 , the nominator vanishes if and only if @xmath47 @xmath36 @xmath183 @xmath52 @xmath184 , or with ( [ omega - m - extremal ] ) , @xmath185 therefore , massless spin waves only can tunnel out of an extremal kerr black hole under a restrictive condition on the separation constant @xmath47 . since @xmath184 , is integer for integer spin @xmath31
, @xmath39 has to be even to guarantee that @xmath186 and thus @xmath47 is an integer .
( remember that @xmath47 has to be integer for integer spin @xmath31 . ) on the other hand , @xmath39 also has to be even such that @xmath47 is a square of an odd - half integer for @xmath63 @xmath36 @xmath187 .
( note : @xmath184 @xmath36 @xmath188 , and @xmath47 @xmath36 @xmath189 ; since @xmath190 @xmath36 0 or 1 mod 4 , @xmath191 @xmath36 1 or 3 mod 4 ; this is only a square number if @xmath191 @xmath36 1 mod 4 , because any square number @xmath192 @xmath36 0 or 1 mod 4 . )
for @xmath1 @xmath36 0 and 0 @xmath147 @xmath193 @xmath194 @xmath195 , the kerr - newman spacetime represents a static charged black hole , the reissner - nordstrm black hole . a general master equation like teukolsky s equation governing massless spin waves in the reissner - nordstrm geometry and leading to radial and angular equations like ( [ radial equation ] ) , ( [ angular equation ] )
could not be derived to date .
nonetheless , the basic equations governing electromagnetic and gravitational waves ( @xmath63 @xmath36 1 , 2 ) can be transformed to @xmath196 @xmath36 0 , where @xmath197 @xmath36 @xmath198 @xmath52 @xmath199 @xmath52 @xmath200 with @xmath201 and @xmath202 .
\label{g - s - rn}\ ] ] here the separation constant @xmath192 is given by @xmath203 @xmath36 @xmath204 , the indices @xmath205 and @xmath206 take the two values @xmath207 @xmath36 1,2 , but @xmath208 , and the constants @xmath209 are given by @xmath210 these relations can be readily deduced from equations ( 244 ) and ( 245 ) in @xcite , 5 p. 244 . by the same procedure as in the proof of theorem [ satz - potential ]
the spin - weighted wave equations can be transformed to @xmath211 @xmath36 0 with @xmath121 @xmath36 d@xmath212d@xmath213 @xmath52 @xmath108 and @xmath108 @xmath36 @xmath200 @xmath54 @xmath214 @xmath54 @xmath215 .
but since @xmath216 @xmath36 @xmath217 @xmath52 @xmath218 , @xmath108 has a pole of order not smaller than 2 as long as @xmath219 . hence for waves of spin - weight @xmath63 @xmath36 1 or 2 no tunneling out of a reissner - nordstrm black hole can occur
for a particle of mass @xmath220 and electrical charge @xmath221 let be given the constant @xmath222 @xmath36 @xmath223 , such that @xmath224 is the compton wavelength
. then it is described by the dirac equation in a general spacetime , cf .
@xcite eq .
( 16 ) ( for @xmath225 ) . in the kerr - newman geometry
it can be reduced to the complex radial differential equation @xmath226 @xmath36 0 , where @xmath197 @xmath36 @xmath198 @xmath52 @xmath199 @xmath52 @xmath200 with @xmath227 and @xmath228 ( note that @xmath72 is related to the usual radial function @xmath41 by ( [ radial transformation ] ) ) , and the angular equation @xmath229 , with @xmath230 here @xmath231 and @xmath232 .
the constants @xmath47 , @xmath48 @xmath32 @xmath49 are separation constants obtaining discrete values depending on the boundary conditions of @xmath43 .
for details see @xcite , for the kerr case ( @xmath30 ) see @xcite , 104 eqs .
( 123 ) , ( 124 ) . to analyze the potential
, a transformation analogous to the proof of theorem [ satz - potential ] has to be done , where now by @xmath233 , @xmath234 cf .
( [ g-s ] ) .
this yields a complex potential @xmath108 differing from ( [ q - s - def ] ) only by terms depending on @xmath222 .
as long as @xmath47 @xmath235 @xmath236 , these terms are of order @xmath237 , and the proof of theorem [ satz - potential ] in essence can be applied .
hence no tunneling of electrons out of a kerr - newman black hole can occur , unless for the case of extremal rotation .
if @xmath47 @xmath36 @xmath236 the situation changes , since now the terms by which @xmath108 differs from the massless case , are of order @xmath238 .
first we notice that @xmath47 @xmath52 @xmath239 @xmath36 @xmath240 . therefore , with @xmath84 @xmath36 @xmath241 , @xmath242 @xmath36 1 , @xmath243 @xmath36 @xmath244 , we obtain @xmath245 with these equations the terms deciding the occurence of a pole of order 2 at the horizon @xmath16 @xmath36 @xmath145 are determined by @xmath246 to avoid a pole of order 2 , the imaginary part has to vanish .
this implies @xmath247 ( even in the extremal case @xmath0
@xmath36 @xmath248 @xmath52 @xmath193 , for then @xmath84 @xmath36 @xmath158 and @xmath0 @xmath36 @xmath175 ) , or @xmath249 the real part @xmath250 @xmath251 re@xmath108 @xmath36 @xmath252 then vanishes if and only if @xmath145 @xmath36 @xmath175 @xmath36 @xmath0 , which is equivalent to @xmath195 @xmath36 @xmath248 @xmath52 @xmath193 .
this is the extremal case , and we see that then @xmath140 @xmath36 @xmath253 @xmath254 , i.e. @xmath255 or @xmath256 therefore , @xmath108 has a pole of order 2 at the event horizon , such that no tunneling may occur for @xmath47 @xmath36 @xmath236 .
in the present paper we consider properties of spin - weighted waves in various kerr - newman spacetimes . we analyze the singular potential barriers of the respective wave equations emerging at the event horizon .
it turns out that the potential barriers are repulsive enough to prevent spin waves with spin @xmath63 @xmath36 0 , @xmath187 , 2 from tunneling out of the black hole , but weak enough to admit the tunneling of photons in case of non - extremal rotation . in particular ,
if @xmath1 denotes the angular momentum of a kerr - newman spacetime and @xmath2 its electrical charge , then the following table shows the occurence of tunneling ( ` @xmath257 ' ) , its impossibility ( ` ' ) and the present ignorance ( ` ? ' ) with respect to the various kinds of spins @xmath31 .
@xmath258 spin $ |s|$ & $ 0 $ & $ \frac12 $ & $ 1 $ & $ 2 $ \\[.5ex ] \hline $ a=0 $ & -- & -- & -- & -- \\ $ 0<|a|<m$ , $ q=0 $ & -- & -- & $ \times$ & -- \\ $ |a|=m$ , $ q=0 $ & $ \times$ & $ \times$ & $ \times$ & $ \times$ \\ $ 0<|a| , |q|<m$ & -- & -- & ?
& ? \\ \hline \hline \end{tabular}% } \ ] ] the case @xmath259 , i.e. charged rotating kerr - newman black holes , still remains unknown .
the crucial role that rotation plays in the known cases suggests that photons may tunnel out of _ any _ rotating black hole .
a very strong hint supporting this conjecture is the analysis of one of the authors @xcite , where any electromagnetic wave _ in the outer space _ of a rotating kerr - newman black hole can gain some energy from the hole .
however , we still have no such hints for gravitons .
for the case of extremal rotation , @xmath260 , any particle can tunnel out of the black hole .
because of the cosmic censorship hypothesis ( [ ccc ] ) , @xmath2 vanishes necessarily in this case .
there is only the restriction on the separation constant @xmath47 given by ( [ lambda - sm - extremal ] ) to be an even integer .
whether such an integer exists at all depends on the sturm - liouville equation for @xmath43 , ( [ angular equation ] ) or ( [ angular dirac ] ) , respectively .
mathematically , in case of @xmath30 or @xmath261 a necessary and sufficient condition for the occurence of the tunnel phenomenon is the vanishing of @xmath262 , the factor of the first derivative in the radial wave equations , and of the function @xmath140 , cf .
equations ( [ g - s ] ) , ( [ k - def ] ) for massless waves of spin @xmath31 , and ( [ g - s - kn ] ) , ( [ k - def - kn ] ) for spin-@xmath187 waves of mass @xmath220 @xmath36 @xmath263 .
this condition yields a potential with an order-1 pole at the event horizon @xmath16 @xmath36 @xmath145 , see figure [ fig - pot ] . the real and imaginary part of @xmath108 for @xmath264 , @xmath30 , @xmath265 , @xmath266 , @xmath161 ; also plotted is the constant @xmath267 .
the tics on the horizontal @xmath16-axis are in units of @xmath0 .
note the poles at @xmath145 @xmath36 @xmath268 and @xmath175 @xmath36 @xmath269 . ]
remarkably , the real part of a potential with a pole of order one is attractive from the outer space of the black hole and repulsive from the interior , but the imaginary part tends to @xmath270 from the outside and to @xmath271 from the interior .
a wave packet inside the black hole with nonvanishing outgoing amplitude @xmath94 then reaches the outer space with nonvanishing amplitude @xmath92 ; if , e.g. , @xmath94 is known at @xmath178 and @xmath272 , the outside amplitude is given by ( [ conservation - law ] ) with @xmath273 . where does this all lead us to ?
notably , the tunnel phenomenon studied in this paper is a semiclassical phenomenon .
this implies that the black hole gravitational field has to be strong enough compared to the typical energy of the wave field such that it does not perturb the curvature essentially , i.e. such that it is a test field .
the tunnel phenomenon seems to be related to another semiclassical phenomenon , the hawking radiation .
but whereas the hawking radiation is a black hole thermodynamical effect occuring in particular for static black holes , the tunneling only exists in case of rotation . as such it resembles the classical superradiance effect , but this affects only , and all , particles with integer spin .
a complete theory of quantum gravity must include the tunnel phenomenon and should explain it .
especially the puzzles concerning gravitons and photons in charged rotating black holes should be solved . to date , the most promising candidate for a theory of quantum gravity is m - theory . it has made strong progress in the last few years , not only with respect to black hole phenomena .
one of the most remarkable aspects in this context certainly is the deduction of the notion of entropy of a black hole @xcite .
however , to our knowledge the analysis of rotating black holes , be they uncharged or charged , in m - theory is still missing . to conclude , rotation in general relativity over and over again reveals surprising phenomena , a fact which is not only demonstrated by @xcite , @xcite , or @xcite .
succeeding in integrating rotation in m - theory would give deep insight . | the potentials of spin - weighted wave equations in various kerr - newman black holes are analyzed .
they all form singular potential barriers at the event horizon . applying the wkb approximation it is shown that no particle can tunnel out of the interior of a static black hole .
however , photons inside a non - extremely rotating kerr black hole may tunnel out into the outer space , whereas neutrinos , electrons , and gravitons may not .
if the rotation is extremal , any particle may tunnel out , under restrictive conditions .
it is unknown whether photons and gravitons may tunnel out if the black hole is charged and rotating . |
atomtronics is an emerging area of investigation in which analogues of electronic circuits and devices are constructed using ultra - cold bosonic atoms rather than electrons as in conventional electronics @xcite .
the conventional way to construct an atomtronic device is to use cold atoms trapped in an optical lattice , which has a description in terms of either the mott - hubbard model transferred from condensed matter physics @xcite for fermionic atoms , and the bose - hubbard model for bosonic atoms @xcite .
these models can represent either insulating behaviour , in the mott insulator regime , or conducting behaviour , in the superfluid regime . in this work we consider only bosonic atoms . shortly after the realisation of trapped bose - einstein condensates ( bec ) , jaksch _ et al . _
@xcite showed that the bose - hubbard model can provide an accurate description of bosonic atoms trapped in a deep optical lattice .
the basics of this model have been used to investigate a wide variety of atomtronic devices @xcite , including one with diode - like behaviour @xcite , a single - atom switching transistor @xcite , and one which uses a triple well configuration to mimic a field effect transistor @xcite .
more recent proposals include circuits of diodes and transistors @xcite , and an atomtronic battery @xcite .
proposed atomtronic transistors with the idea of combining them into gates for quantum computers @xcite .
we have no doubt that there will be more proposals in the future , and what we will show here is that a full quantum description of both the dynamics and the initial state needs to be taken into account to guarantee accurate descriptions of the dynamics of such devices .
various theoretical methods have been used to analyse these devices up until the present . for small numbers of atoms and sites , direct quantum calculations using a master equation
are often possible .
et al . _ developed a quantum master equation to treat systems which interact with sources and sinks , based largely on methods which have been extremely successful in quantum optics . with this method
, they have analysed electronic diodes , field - effect transistors , bipolar junction transistors , and an analogy to a logic gate @xcite .
the master equation was then solved numerically , giving fully quantum solutions .
this method allows for the investigation of systems with relatively small numbers of atoms and lattice sites , with the authors typically treating @xmath0 or @xmath1 sites , each with @xmath2 or @xmath3 atoms .
used coupled nonlinear schrdinger equations in one dimension to calculate the phase evolution of eigenstates , with two distinguishable atoms in a triple well @xcite . for the small systems considered by these authors , their methods are perfectly adequate . in the mesoscopic regime ,
when the numbers of atoms can make density matrix calculations complicated , the one - body schrdinger equation can be used for non - interacting atoms , or the mean - field gross - pitaevskii equations ( gpe ) for interacting atoms .
both of these have been used to model coherent atomic transport in a three - well potential @xcite and each has disadvantages .
the one - body schrdinger equation , being linear , can not include the interactions required for the mott insulator regime .
the gpe approach can not describe any quantum statistical features , such as the actual quantum states or any entanglement properties .
the three - well coherent transport model has previously been analysed using the fully quantum positive - p representation @xcite , which was used to note the differences caused by different initial quantum states on the dynamics , the entanglement properties @xcite and the quantum steering properties @xcite . in what follows
we base ourselves on the approach taken by milburn _
et al . _
@xcite , generalisng this to four wells @xcite , and using the fully quantum positive - p phase space representation .
we consider this to be the most suitable approach here because it is exact , allows for an easy representation of mesoscopic numbers of atoms , and can simulate different quantum states @xcite .
just as importantly , the positive - p calculations scale linearly with the number of sites and can in principle deal with any number of atoms . these are powerful advantages when we wish to consider mesoscopic numbers of atoms in arbitrary numbers of potential wells .
one disadvantage of the positive - p representation is that the integration has a tendency to diverge for high collisional nonlinearities @xcite , although it often converges for long enough to show marked differences from mean - field predictions @xcite and also allows for the calculation of quantum correlations @xcite .
the four - well system is as shown in the schematic of fig .
[ fig : setup ] , where the circles represent the wells or lattice sites , each of which contains a single atomic mode .
the @xmath4 are bosonic annihilation operators for atoms in mode @xmath5 , the @xmath6 are the ground state single - atom energies of the wells , and @xmath7 represents the coupling between the wells .
we assume that there is no coupling between well @xmath1 and wells @xmath2 and @xmath0 .
we also assume that any atoms at each site are initially in their ground state .
the basic idea is that atoms from the outside wells will tunnel into the centre well and from there to well @xmath1 .
because of the wave nature of the condensed atoms , we expect that the initial phase difference , @xmath8 between the atomic modes at sites @xmath2 and @xmath0 will affect the rate of tunnelling into the middle and hence into well @xmath1 .
we would hence have a type of phase sensitive gate , where the phase of one mode mode can be used to control the occupation of another . and
@xmath0 have nearest neighbour couplings , and well @xmath1 is coupled with well @xmath3 .
the @xmath4 are bosonic annihilation operators for atoms in mode @xmath5 , the @xmath6 are the ground state single - atom energies of the wells , and @xmath7 represents the coupling between the wells .
@xmath8 represents a phase - shift which is applied to mode @xmath0 at the beginning of the evolution , when the populations of @xmath2 and @xmath0 will be equal , with @xmath3 and @xmath1 initially being empty . ]
following the usual procedures @xcite , with @xmath9 as the s - wave collisional term , we write the hamiltonian as @xmath10 where @xmath5 runs from @xmath2 to @xmath1 . starting from this hamiltonian ,
our first step is to find the semi - classical mean - field equations in the gross - pitaevskii approach .
we will use the solutions of these for comparison purposes , since it is well known that they are not always accurate for the bose - hubbard model , even for calculation of the mean fields @xcite . using the variables @xmath11 to represent the mean - fields , we find @xmath12 for the numbers in each well as a function of dimensionless time , @xmath13 .
the initial phase difference , @xmath8 , between wells @xmath2 and @xmath0 is zero , and the numbers in these two wells are equal .
wells @xmath3 and @xmath1 are initially empty .
we see that atoms pass through the middle well and populate the fourth .
the parameters used in this and all subsequent plots are @xmath14 , @xmath15 , @xmath16 , @xmath17 , and @xmath18 . ] to solve the full quantum equations , we use the positive - p representation @xcite , which allows for exact solutions of the dynamics arising from the hamiltonian of eq . [
eq : ham ] in terms of normally - ordered operator averages , in the limit of the average of an infinite number of trajectories of stochastic differential equations in a doubled phase - space .
this method also allows for the representation of different quantum states in the initial conditions @xcite .
the positive - p representation does have the disadvantage that it can suffer from divergence problems in cases with a largish colliisional nonlinearity , often limiting its utility to short time dynamics @xcite . in this letter
, it is exactly this regime we are analysing , so that the positive - p representation is perfectly adequate . following the standard methods @xcite ,
the set of it stochastic differential equations @xcite are found as @xmath19 where the @xmath20 are standard gaussian noises with @xmath21 and @xmath22 .
as always , averages of the positive - p variables represent normally ordered operator moments , such that , for example , @xmath23 .
in order to analyse and compare the effects of different initial states , we begin with a total of @xmath24 atoms , evenly distributed between wells @xmath2 and @xmath0 and analyse the classical evolution given by eq . [ eq : gpe ] .
the main quantity of interest is @xmath25 , the number of atoms in well @xmath1 as a function of time . with
the initial phases equal , we see a @xmath26 transfer of atoms into the fourth well at certain times , as shown in fig . [
fig : classzero ] .
when we set @xmath8 to @xmath27 , this fourth well remains unpopulated , showing that the device can act semi - classically as a phase - dependent gate . with @xmath28 , as seen in fig .
[ fig : classpihalf ] , we see that a little less than half the atoms appear in the fourth well . having
thus benchmarked the device , we will now proceed to see how it performs with initial coherent and fock states in wells @xmath2 and @xmath0 . for the numbers in each well for the same parameters and initial conditions as fig .
[ fig : classzero ] except that @xmath28 .
we see that the maximum population of well four is approximately half that seen in fig .
[ fig : classzero ] . ] the initial quantum states are modelled as in ref .
@xcite , with the positive - p representation equations being numerically integrated over a large number of stochastic trajectories .
the immediate difference is that , because we can calculate close approximations to normally - orderd operator expectation values , we are also able to calculate the standard deviations about the average solutions for the atom numbers .
we begin with coherent states with the same average occupation numbers as used classically .
the important differences are that the initial physical number distribution is now poissonian ( close to gaussian as the number of quanta increases ) and quantum noise is included in the calculations .
we immediately find that the average solutions for the intensities are the same as the semi - classical mean - field solutions , for both @xmath29 and @xmath30 .
the only difference is that there is a distribution about these solutions , as shown by the lines plotted at plus and minus one standard deviation .
thus the semi - classical treatment gives the most likely solutions , although we must remember that most individual experiments will give something different .
how important this is in the case of initial coherent states will depend on how precisely we wish a particular atomtronic device to function . and the solutions are the average of @xmath31 stochastic trajectories of the positive - p equations . ] and @xmath32 are equal .
the initial conditions are the same as for fig .
[ fig : czero ] , but with @xmath30 , and the solutions are the average of @xmath33 stochastic trajectories of the positive - p equations . ] when we begin the simulations with fock states of definite atom number in wells @xmath2 and @xmath0 , we find that the device loses its phase dependence completely .
the average results are almost identical to those found semi - classically for @xmath28 , and do not change irrespective of the initial phase .
this is not unexpected since fock states can be roughly thought of as having indeterminate phase . as can be seen in fig .
[ fig : fzero ] , however , the standard deviation around the mean solution is significant . in practice
this means that the device would be totally inappropriate as a phase sensitive gate if the initially occupied wells are in number states . as this is usually thought of as the natural state for mott insulators , this would be problematic . and the solutions are the average of @xmath34 stochastic trajectories of the positive - p equations .
the solutions for initial fock states are identical for all values of @xmath8 . ]
although we have considered the two extreme cases of coherent and fock states here , it is worth noting that there are other possible initial states . as an example
, the collisional nonlinearity of condensed atoms has been shown to give a crescent " state , where the wigner function is stretched in the phase space @xcite .
this initial state has been previously shown to affect the dynamics of molecular association @xcite and would be expected to have an effect with our device as well .
the advantage of the positive - p representation is that it allows us to consider arbitrary initial quantum states , as well as the calculation of any correlations that can be expressed as normally - ordered operator expectation values .
in conclusion , we have shown that the positive - p representation is a useful tool for analysing the performance of atomtronic devices where mesoscopic numbers of sites and atoms are involved .
it is easily extended to much greater numbers of both than we have considered here .
we have also shown that mean - field claculations can be very misleading for such systems and it is therefore important to undertake full quantum calculations .
we have considered two markedly different initial states , showing that the coherent states , as expected , give results close to the semi - classical predictions . on the other hand , if the atoms are initially in fock states , the semi - classical predictions are almost useless . as shown previously ,
correlations between the atomic modes at each site can also be calculated with this representation , which will open the door for fully quantum atomtronics in which use can be made of properties such as entanglement and epr steering .
this will definitely open up regimes that are not readily accessible with standard electronics .
this research was supported by the australian research council under the future fellowships program ( grant i d : ft100100515 ) , the new zealand marsden fund and a rutherford discovery fellowship from the royal society of new zealand . | it is known that a semi - classical analysis is not always adequate for atomtronics devices , but that a fully quantum analysis is often necessary to make reliable predictions . while small numbers of atoms at a small number of sites are tractable using the density matrix , a fully quantum analysis is often not straightforward as the system becomes larger .
we show that the fully quantum positive - p representation is then a viable calculational tool .
we postulate an atomtronic phase - gate consisting of four wells in a bose - hubbard configuration , for which the semi - classical dynamics are controllable using the phase of the atomic mode in one of the wells .
we show that the quantum predictions of the positive - p representation for the performance of this device have little relation to those found semi - classically , and that the performance depends markedly on the actual quantum states of the initially occupied modes .
we find that initial coherent states lead to closest to classical dynamics , but that initial fock states give results that are quite different .
a fully quantum analysis also opens the door for deeply quantum atomtronics , in which properties such as entanglement and epr ( einstein - podolsky - rosen ) steering become valuable technical properties of a device . |
recently , theoretical @xcite-@xcite and experimental @xcite,@xcite-@xcite attention has been focused on spatiotemporal instabilities in one - dimensional reactive flows . among these pattern - forming instabilities
are the differential flow ( difi)@xcite@xcite@xcite@xcite , turing@xcite@xcite , and the physically distinct flow - distributed oscillation ( fdo)@xcite@xcite@xcite@xcite@xcite mechanisms .
two of these , difi and fdo , necessarily involve a flow , while turing and difi necessarily involve the differential transport of activator and inhibitor species .
instabilities in a flowing medium may be absolute or convective.@xcite@xcite@xcite@xcite . in the first case
, a localized disturbance grows with time and spreads both upstream and downstream . in the convective case ,
on the other hand , a localized disturbance can not propagate upstream , and so the effect of a temporary localized perturbation is eventually washed downstream and out of the system if there is a downstream boundary .
however , persistent disturbances upstream can have a large effect on the downstream behavior .
this leads to the possibility of noise - sustained structures@xcite@xcite or patterns which are controlled primarily by the _ upstream _ boundary conditions .
we are insterested here in this latter case , where the upstream boundary is crucial to the control of the pattern .
fdo is a convective mechanism of pattern formation whereby an open flow maps the temporal dynamics of an oscillating medium , whose phase is set at the upstream boundary , onto space . in the limit of vanishing diffusion , the resulting stationary@xcite@xcite@xcite@xcite@xcite , travelling@xcite@xcite@xcite and
pulsating@xcite@xcite waves are simple kinematic phase waves@xcite , making fdo conceptually the simplest of the pattern - forming mechanisms , although it was discovered later than the others .
the turing instability , by contrast , was initially conceived of as an absolute instability of a stationary reaction - diffusion medium . in a flow system , however , turing @xcite and difi @xcite patterns can also be generated and controlled by means of the upstream boundary condition under convectively unstable conditions . since an open flow with a fixed upstream boundary is equivalent , via a galilean transformation , to a stationary medium with a moving boundary@xcite@xcite@xcite@xcite , the physical ideas of fdo and other boundary - driven convective instabilities are also applicable to growing media . in developmental biology
an fdo mechanism driven by an oscillator or `` segmentation clock '' at the growing tip of an embryo leads to the formation of somites @xcite , the precursors of vertebrae and body segments during early embryogenesis@xcite@xcite@xcite@xcite ( the best - studied examples are chick and mouse . ) . quite generally , the issue of pattern formation on a growing domain is vitally important to developmental biology.@xcite recent laboratory experiments @xcite@xcite@xcite in turing or hopf unstable media also make use of a moving boundary that mimics a flow .
by contrast , a packed bed reactor ( pbr ) is a flow reactor in which the inlet , not the medium , is fixed in the laboratory frame of reference . in the experiments of @xcite@xcite@xcite@xcite@xcite the reactor
is fed by the outlet of a continuous stirred tank reactor ( cstr ) which can be made to oscillate , generating travelling waves in the pbr - tube , or remain at a fixed point , leading to stationary waves .
an extensive comparison of the parameter ranges for the production of stationary waves by means of various instabilities was made by satnoianu et al.@xcite , who suggested that all of these waves be viewed as variants of a general mechanism called `` flow and diffusion- distributed structures '' ( fds ) . in reference
@xcite , travelling waves and combinations of differential flow and diffusion were also considered .
travelling waves were refered to as difi waves while stationary waves were referred to as fds . our goal is to clarify the relationships among the convectively driven fdo and differential transport ( turing and difi ) modes in an open flow .
we develop a general linear stability analysis for convective modes driven by boundary perturbations .
we illustrate the relationships visually by plotting solutions of the dispersion relations .
our approach differs from that of @xcite and @xcite in several ways .
first , we choose to focus on patterns driven convectively by the upstream boundary condition , distinguishing them from absolute instabilities .
we do this because the possibility of convective instability embodies much of the new behavior that is possible with a flow ( or growth ) as opposed to a stationary medium .
accordingly , we treat the dispersion relation for small disturbances differently , taking the real frequency , set by the boundary condition , as the independent variable and examining the _ spatial _ behavior of the resulting disturbance rather than examining the temporal behavior of an imposed spatial perturbation .
we consider a mode unstable if it grows with downstream _ distance _ in response to a constant or periodic driving at the boundary .
this approach resembles that of @xcite and @xcite . within this approach
we find it useful to distinguish wave modes not by whether they are stationary or travelling as in @xcite but by other criteria including the phase velocity and the relative phase between oscillations of the activator and inhibitor .
we find that the fdo and difi mechanisms are closely related to each other , both being related to an underlying hopf instability , while the turing mechanism gives rise to a distinct set of modes .
the two sets of modes are apparent as two distinct peaks in the spatial growth rate at different perturbation frequencies .
the difi / fdo modes can be either travelling or stationary , while turing modes are stationary only in the case of zero flow velocity .
in the zero - flow case ( the original case in which the turing mechanism was considered ) , the instability is absolute and therefore not controlled by the boundary .
however , it has been observed that turing patterns can be generated in a system with nonzero flow @xcite@xcite@xcite in which case they are advected along with the flow , i.e. , stationary in the co - moving frame . in this case
the instability can be convective and a turing mode with a particular wavelength can be selected by imposing a periodic perturbation at the inflow .
we find that in the presence of simultaneous differential flow and differential diffusion ( relevant to the packed - bed reactor ) some of the distinguishing features of turing modes are modified , but the essential picture of two separate peaks remains unchanged . at the end of the paper
, we describe some nonlinear simulations which help to illustrate some of the linear results and relationships we describe .
we find that the linear results give a rather good insight into the nature of the fully nonlinear solutions , at least in the case where the nonlinearity is not very strong .
we consider the reaction - diffusion - advection ( rda ) equations describing the transport and chemical kinetics of an activator and an inhibitor species .
the chemical medium is defined by the local or batch reactor dynamics together with transport terms .
we wish to consider several forms of differential transport , so we allow each species to have its own flow velocity and diffusion coefficient .
the rda equations are : @xmath0 our aim is to analyze the pattern - forming instabilities , so we shall assume that the local kinetics has a stable or unstable fixed point , and linearize the equations for small perturbations of the uniform fixed point state . for convenience
, we shall use units in which the flow velocity of species b is unity .
linearizing near the fixed point @xmath1 , transforming the units to ones where @xmath2 and defining the velocity and diffusion ratios @xmath3 and @xmath4 respectively , and @xmath5 gives : @xmath6 where the matrix @xmath7 is the jacobian of the local kinetic system evaluated at the fixed point and @xmath8 , @xmath9 are the perturbations . a complex exponential solution@xmath10 represents a travelling wave in which the concentrations of both species oscillate .
is that of @xcite , chosen for later convenience .
@xmath11 represents the spatial growth rate , while -@xmath12 is the inverse wavelength or real wavenumber . ]
the relative amplitude and phase are determined by the complex amplitudes @xmath13 and @xmath14 ( a real solution is formed from ( [ complexexp ] ) and its complex conjugate . ) substitution into ( [ rdalin ] ) gives@xmath15 which can be combined to give the dispersion relation@xmath16 where @xmath17 and @xmath18 are respectively the trace and determinant of the jacobian .
two particular cases of differential transport were studied in previous work .
the case @xmath19 @xmath20 is relevant to @xcite , in which there is differential diffusion due to the immobilization of one of the species , but the flow velocities are the same , since it is actually the boundary that moves relative to the medium . on the other hand , satnoianu et al.@xcite considered the case @xmath21 which may be a good approximation in a flow system when one of the species is immobilized and the other moves freely .
the pure fdo case @xmath22 has no differential transport . with the appropriate changes of variables and restrictions on the transport ratios
, the above dispersion relation reduces to the ones given in the previous references @xcite@xcite@xcite@xcite@xcite for particular cases .
we wish to consider more general forms of differential transport , for both theoretical and experimental reasons .
first , varying the two transport ratios independently allows a fuller understanding of the effects of the two types of differential transport and their interaction .
second , we wish to allow the possibility of experiments in which the transport coefficients are related in ways other than those previously considered .
we wish to analyze the steady - state response of the system to a sinusoidal forcing of the inflow boundary at a constant amplitude . in general , in the linear approximation , this gives rise to a travelling wave with a complex wave number .
the frequency @xmath23 will be taken to be purely real , reflecting the constant amplitude of the forcing . however , the convective dynamics of the medium may cause the disturbance to grow or damp with the downstream distance , so that @xmath24 may be complex . thus , we consider the real @xmath23 as an independent variable and solve the dispersion relation numerically for the complex @xmath24 .
the dispersion relation is quartic in @xmath24 and so has in general four solutions .
each is associated with an eigenvector @xmath25 which can be found by substituting the solution @xmath24 back into ( [ doubledisp ] ) . in this way we can find @xmath26 and @xmath27 the ratio @xmath28 , which in general is complex , gives information about the relative amplitude and phase of oscillations in the two species ( an example is discussed below )
we will see that in general the four solutions comprise two pairs , of which only one pair is relevant to the system s behavior near the upstream boundary .
the two solutions of a pair together give one physical oscillation mode with an arbitrary phase . to illustrate the physical meaning of the dispersion relation , we consider first the simplest case of pure fdo , in which @xmath30 it can easily be verified that in this case the dispersion relation ( [ dispersion ] ) relation does not depend on the jacobian matrix elements separately , but only on the trace and determinant . the pair of equations ( [ doubledisp ] )
can then be diagonalized completely by changing coordinates to the eigenbasis of the jacobian , and the quartic dispersion relation factorizes into two quadratic ones as derived in @xcite , one for each eigenvector .
the quadratic dispersion relations depend on the jacobian eigenvalues , which are given by @xmath31 and are complex conjugates if @xmath32 . for the sake of simplicity we consider the particular case with @xmath33 whose eigenvalues and eigenvectors are @xmath34 and @xmath35 in general
, the relative amplitude of the two components is complex for an oscillatory system , indicating a phase difference between the the two components . in this particular case , the phase difference is @xmath36 .
a plot of the four solutions @xmath37 ( @xmath38 ) then has the form shown in figure [ purefdo ] .
there are two solutions associated with each of the two eigenvectors , of which one has a much larger real component .
both solutions are necessary in order to satisfy a boundary value problem in which boundary conditions are specified at @xmath39 and at some downstream point @xmath40
. however , it has been argued@xcite that for quite general boundary conditions at a far - away downstream boundary , it is the solution with the _ less _ positive growth rate that predominates near the upstream boundary ( @xmath41 ) . as an example , consider adjusting the coefficients @xmath42 and @xmath43 in the general solution @xmath44 ( with @xmath45 ) so as to satisfy either dirichlet or neumann boundary conditions at @xmath40 . in either of these cases , @xmath43 is smaller than @xmath42 by a factor of order @xmath46 .
therefore , when there is a clear separation between the pairs of solutions , the one with lower growth rate dominates everywhere but close to the downstream boundary , which we take to be far away compared to the growth or damping length scales@xmath47(@xmath48 ) .
we therefore focus attention on the lower solutions with smaller growth rates .
these two solutions are associated with the two eigenvectors of the jacobian , so we may label them @xmath49 .
they are complex conjugate mirror images of each other under reflection through the vertical axis ( @xmath50 ) and represent the same physical wave solution , namely@xmath51 = \operatorname{re}\left [ e^{-i\omega t}e^{k_{-}x}\mathbf{u}_{-}\right ] = % \begin{pmatrix } \cos(\omega t+\operatorname{im}k_{+}x)\\ \sin(\omega t+\operatorname{im}k_{+}x ) \end{pmatrix } e^{(\operatorname{re}k_{+})x}.\label{physicalsolution}%\ ] ] because of this reflection symmetery , it will be convenient in the remainder of the paper to plot only one solution , with the understanding that the reflected complex conjugate is also present .
figure [ purefdo ] shows the growth rates and wave numbers for all four solutions .
note that : 1 ) @xmath12 has a zero at the natural frequency @xmath52 .
when @xmath12=0 , the phase velocity@xmath53 has a corresponding pole , as will be seen in figures [ difi1]-[difituring ] .
disturbances at precisely the natural frequency result in growing uniform oscillations of the medium ( this is essentially the batch hopf mode ) while perturbations faster or slower than @xmath54 give downstream or upstream travelling waves , respectively @xcite@xcite .
stationary waves occur for @xmath55 .
as shown in @xcite , the sharpness of the growth rate peak depends on the dimensionless quantity @xmath56 .
increasing the value of @xmath57 ( or , equivalently , reducing the flow velocity ) makes the peak sharper and narrower and reduces the gap between the upper and lower solutions , until , at the threshold of absolute instability , the growth rate curve develops a cusp and the upper and lower solutions cross . when the growth rate curves of the two solutions cross , it is no longer correct to view the solution in the bulk as being determined primarily by the upstream boundary condition it is also strongly influenced by downstream conditions .
this is the signature of an absolute rather than a convective instability .
the peak of the growth rate occurs precisely at the frequency defined by the imaginary part of the jacobian eigenvalue . if the fixed point is not hopf unstable but instead has eigenvalues @xmath58 with @xmath59 , then the picture is qualitatively the same , except that the peak remains below the horizontal axis .
thus all perturbations are damped in this case , but the most slowly damped ones are at the natural frequency .
the pure fdo case can be viewed as a baseline for the physical interpretation of the dispersion relations and their solution curves in the presence of differential transport .
differential transport will modify the shapes of the curves , and will make the eigenvectors @xmath23-dependent and no longer coincident with those of the jacobian
. finally , note that in the case when the fixed point is an unstable node rather than a focus , the jacobian has distinct real eigenvalues and eigenvectors instead of complex conjugate pairs .
peak growth rates for the modes along both eigenvectors then occur at @xmath60@xcite the above results are qualitatively universal for any system with a hopf instability .
[ ptb ] purefdo.eps with the pure fdo case as a comparison , we now examine the effects of differential transport and the convective , boundary driven manifestations of difi and turing instabilities .
typical results are shown in figs .
[ difi1]-[difituring2 ] using the fitzhugh - nagumo - like @xcite ( fn ) model ( [ fneqns]-[fnjac ] ) for the local dynamics .
the key features of the relevant solutions in the fdo case are that @xmath61 has a growth peak and the associated phase velocity has a pole at positive @xmath23 , while @xmath62 has a peak and pole on the opposite side , @xmath63 .
a brief summary of the effects of differential transport on the dispersion relation solutions is as follows : \1 .
the primary effect of differential flow is to displace and distort the positive-@xmath23 peak of @xmath64(and its mirror image in @xmath62 ) .
depending on the details of the model , the peak may be shifted to the left , right , upward or downward .
the pole in the phase velocity may also be shifted left or right .
if sufficiently strong , differential flow can raise the peak growth rate from negative to positive , thus giving an instability even for a stable fixed point .
this is precisely what happens in difi . for @xmath65
, the peak growth rate may occur quite far from the pole of the phase velocity ; thus the strongest instability is to a travelling wave solution rather than to a uniform oscillation .
the most important effect of differential diffusion in the absence of differential flow ( @xmath66 ) is to alter the shape of the negative-@xmath23 tail of the @xmath61 solution ( or , equivalently , the positive tail of @xmath62 ) . for fast inhibitor diffusion ( @xmath67 ) the negative tail can develop first an inflection point and then a second growth rate peak .
the modes within this second peak are distinguished by the following features , confirming their interpretation as turing patterns imposed by the boundary condition and advected with the flow : a ) their phase velocities are all very close to unity ( in units where the flow velocity is 1 ) showing that they are stationary in the comoving frame .
b ) the amplitude ratio @xmath68 from the associated eigenvector is almost purely real , meaning that , contrary to the situation in self - sustained oscillations , there is no phase lag between the two species ( the activator and inhibitor are almost exactly in phase or @xmath69 out of phase ) .
for the most general case of differential transport , then , the dispersion relation has either one or two peaks , which we can identify as hopf / fdo / difi and turing peaks respectively .
changing @xmath70 can alter the shape of the fdo peak and conversely , @xmath71 can alter the turing peak , but they generally retain a separate identity , and for the most part exist in a see - saw relation
. we now illustrate these statements using examples based on particular models for the form of the jacobian .
the examples we present in figures [ difi1]-[difituring2 ] and the nonlinear simulations used a form of the fitzhugh - nagumo ( fn ) model@xcite , for which the local dynamics is given by @xmath72 and the jacobian is @xmath73 where @xmath74 is a control parameter .
@xmath42 and @xmath43 play the roles of activator and inhibitor , respectively .
a hopf bifurcation occurs at @xmath75 ; the fixed point is unstable for @xmath76 .
we also studied another model using the simpler jacobian@xmath77 with control parameter @xmath78 and hopf instability for @xmath79 , hereafter referred to as the @xmath78-model . for the @xmath78-model with @xmath79 ,
both species are autocatalytic , but one inhibits the other . in most cases ,
qualitatively similar results were obtained for both models .
when the results for the @xmath78-model differ from theose of the fn model , we describe them verbally .
figures [ difi1 ] and [ difi2 ] show typical effects of differential flow on the fdo peak .
differential flow with either fast inhibitor or fast activator transport shifts the position and height of the peak growth rate .
it can also shift slightly the location of the pole in the phase velocity ( i.e. , the zero of @xmath12 ) but usually this shift is less pronounced than the shift of the peak in @xmath11 . in the case shown in figures
[ difi1 ] and [ difi2 ] , the peak height and location are both apparently monotonic in @xmath80 near @xmath81 ; the peak shifts upward and to the right as @xmath80 decreases .
this is not universal , however . in some cases
( see for example figures [ puredifi ] and [ difituring ] below ) , the peak height has its minimum when @xmath81 , so that faster flow of _ either _ species raises the height of the instability peak . in one case
we examined using the marginally stable @xmath78-model with @xmath82 , the peak shifts to the right both for @xmath83 and @xmath84 . what has been referred to as the differential flow instability ( difi )
can be understood a special case in which a growth rate peak whose maximum is less than zero in the absence of differential flow is shifted above zero when @xmath85 , thus creating a convective instability even though the fixed point of the local system is stable .
an example of this is shown in figure [ puredifi ] .
physically , the shifting of the peak relative to the phase velocity pole means that in the presence of differential flow the fastest - growing mode is a travelling wave with some finite velocity , rather than a uniform oscillation . from figure [ difi2 ]
it is evident that , while the amplitude ratio @xmath68 of the two species remains constant in the pure fdo case @xmath81 , differential flow modifies both their amplitude and phase relations in a frequency - dependent manner .
[ ptb ] difi1.eps [ ptbptb ] difi2.eps [ ptbptbptb ] puredifi.eps figures [ turing1 ] and [ turing2 ] show the effect of differential diffusion without differential flow @xmath86 .
a family of solution curves is shown for @xmath87 ( i.e. , equal diffusion or fast inhibitor diffusion ) .
this is the case that renders a turing instability possible in a stationary medium .
the growth rate peak is distorted somewhat relative to that of the pure fdo case .
this effect was more pronounced in some other examples we studied . in one case ,
fast inhibitor diffusion lowered and broadened the fdo peak slightly while fast activator diffusion raised and sharpened the peak significantly . in any case , however , the distortion of the fdo peak is rather less salient than the growth of a second peak at a different frequency .
when this second peak rises above zero , the modes contained within it have two important features : first , their phase velocity is close to 1 . the phase velocity curves in fig .
[ turing1 ] flatten out at @xmath88 for the range of amplified frequencies in the second peak .
this means that , in the co - moving frame , the waves are stationary .
second , for frequencies within the range of the second peak , the imaginary part of the amplitude ratio @xmath89 is very small , indicating a lack of phase lag between the two components . both of these observations are consistent with the turing instability caused by differential diffusion .
because diffusion is directionally symmetric , a mechanism driven by differential diffusion can not cause a phase lag between the two components .
turing patterns are reflection - symmetric , and stationary in the co - moving frame . in view of these observations
we attribute the second peak to turing - like modes and refer to it as a turing peak . in this example
it is quite clearly separated from the fdo peak , the latter being characterized by strongly frequency dependent phase velocities and a non - zero , imaginary component @xmath89 of the amplitude ratio .
there is a range of frequencies between the two peaks for which there are only damped modes . in some cases
, however , the two peaks can grow broader and almost merge , so that as the driving frequency changes , the resulting waves change continuously from an fdo - like to a turing - like character .
even in such cases , the turing modes are distinguishable by means of their near - unity phase velocities and nearly real amplitude ratios .
[ ptb ] turing1.eps [ ptbptb ] turing2.eps in figures [ difituring ] and [ difituring2 ] , we examine the interaction between differential flow and differential diffusion , allowing both differential transport modes to operate simultaneously , as in refs .
@xcite@xcite@xcite . here
we plot the dispersion solutions for a constant value of @xmath70 as @xmath80 varies .
@xmath70 is such that a well - defined turing peak exists for @xmath81 .
we observe that setting @xmath85 shifts the fdo peak as we expect . in this case , unlike that of figure [ difi1 ] but similar to fig .
[ puredifi ] , the peak grows higher for either fast activator or fast inhibitor flow . on the other hand ,
the turing peak is lowered for any @xmath85 .
the two peaks appear to have a see - saw relation .
differential flow has other effects on the modes within the turing peak .
their velocity begins to depart from unity and is less uniform across the peak , and the amplitude ratio is no longer purely real . in these senses ,
the turing modes begin to lose their turing - like character in the presence of differential flow , even though one can still perceive two separate peaks in the growth rate .
[ ptb ] difituring1.eps [ ptbptb ] difituring2.eps
we now show the results of some nonlinear simulations of the fitzhugh - nagumo flow system in order to illustrate the application of the dispersion relations to experiments . we choose to simulate the fn model with @xmath90 , @xmath91 , @xmath81 and @xmath92 . as in the examples of figures [ turing1 ] and [ turing2 ]
, the dispersion relation shows both an fdo and a turing peak . the natural oscillation frequency ( the pole in the phase velocity ) is @xmath93 . for comparison with the simulations ,
we plot both solutions @xmath49 on the same axes , for physical frequencies @xmath94 these plots are shown in figs .
[ turingexdisp1]-[turingexdisp3 ] .
figure [ multiplots ] shows a series of simulations with different driving frequencies .
the boundary conditions for these simulations were given by @xmath95 where @xmath96 .
the plots in the left column of this figure show the space - time patterns of the waves generated by the boundary perturbation .
the dotted white line in each plot represents the trajectory of a point co - moving with the flow .
this allows the phase velocities of the waves to be compared readily with the flow velocity .
the plots in the right column show both dynamical variables @xmath97 and @xmath98 as functions of position for a single time .
the latter plots allow an examination of the waveforms , including the phase shifts between activator and inhibitor .
the dispersion relation ( fig . [ turingexdisp1 ] ) predicts a positive growth rate at @xmath55 , and , accordingly , a constant ( zero - frequency ) perturbation indeed gives rise to growing stationary waves which saturate at a finite amplitude . at @xmath99 , below the natural frequency
, the dispersion relation shows that both @xmath61 and @xmath62 have positive growth rates . @xmath62
( the solid curves in fig
. [ turingexdisp1 ] ) gives waves with a phase velocity of @xmath100 , i.e. , downstream waves moving slower than the flow velocity , while @xmath61 gives waves with a negative phase velocity , i.e. , upstream travelling waves . near the boundary ,
a superposition of both waves occurs , but the upstream waves have a much larger growth rate and they dominate at positions farther downstream , crowding out the other mode entirely and reaching a nonlinear saturated amplitude . at @xmath101
, only @xmath61 has a positive growth rate , giving waves with an positive ( downstream ) phase velocity faster than the flow speed .
a small admixture of the other mode @xmath62 can be discerned near the boundary , but it decays rapidly with downstream distance .
@xmath102 falls within the gap between the fdo and turing peaks .
thus there are no growing modes at this frequency and the disturbance decreases with downstream distance .
@xmath103 , however , lies within the turing peak of the @xmath62 solution .
as predicted by the dispersion relation , the resulting waves have a velocity nearly equal to the flow velocity and there is almost no phase difference between the two species .
[ ptb ] turingexdisp1.eps [ ptbptb ] turingexdisp3.eps [ ptbptbptb ] multiplots.eps the amplitude and phase characteristics of the nonlinear waveforms may also be compared with the predictions of the linear dispersion relation , and in this case the agreement is quite close . in the complex exponential solution eq .
[ complexexp ] , the modulus of the ratio @xmath104 gives the ratio of the peak amplitudes of the oscillations of the two dynamical variables , while the argument of @xmath68 gives the relative phase . for frequencies within the fdo peak ,
our nonlinear simulations show that , as one expects from figure [ turingexdisp3 ] , the ratio of the peak amplitudes of @xmath98 and @xmath97 is slightly larger than unity , while the phase shift is approximately @xmath105 . for @xmath103 , on the other hand , the linear dispersion relation gives a very small phase shift for the relevant @xmath62 solution , and an amplitude ratio slightly larger than @xmath106 .
the phase shift is indeed almost zero and the @xmath98 amplitude is indeed smaller than that of @xmath97 .
the actual amplitude ratio in the saturated waveform is approximately @xmath107 , close to the prediction of the linearized analysis .
finally , we note that at frequencies where more than one mode is present , one can be selected by manipulating the driving function itself so as to align it with one eigenvector or the other . as an example , consider @xmath99 , a frequency at which both solution branches exhibit positive growth rates . here
, we find that for the faster - growing @xmath61 mode the complex amplitude ratio is @xmath108 while for the other mode @xmath109 . the driving function used in fig .
[ multiplots ] excites both of these modes and a superposition is seen near the upstream boundary . by tuning the driving function to be @xmath110 however
, we can excite mostly the @xmath61 mode so that the upstream travelling waves appear almost uncontaminated .
conversely , by choosing the amplitude and phase of the driving to align with the other eigenvector , we excite mostly the other , @xmath62 mode .
eventually , however , the other , faster - growing mode begins to appear , possibly through nonlinear effects or through the small admixture still present in the boundary condition , and the @xmath61 mode eventually wins in the asymptotic downstream region .
simulation results which show this selection effect are plotted in figure [ selectionplot ] .
[ ptb ] selectionplot.eps
we have examined the linear stability analysis relevant to the convective growth of spatiotemporal patterns in a reactive flow system , excited by a small sinusoidal perturbation of a fixed point at the inflow boundary .
we examined the real and imaginary parts of the wavenumber as functions of the real boundary forcing frequency .
they represent the downstream growth rate and periodicity of the disturbances caused by a boundary perturbation .
we found that the growth rate as a function of frequency exhibits at most two physically distinct peaks corresponding to two types of waves , one or both of which may be present and convectively unstable .
we found that , in addition to the phase velocity , the complex ratio @xmath68 of activator and inhibitor concentrations , which encodes the relative amplitude and phase of oscillations in the activator and inhibitor concentrations , provides an additional criterion for distinguishing the types of modes .
turing - like modes are distinguished from fdo - like modes by the lack of a phase lag between the two components and by phase velocities that are close to the flow velocity , so that in the co - moving frame they are stationary patterns .
viewing the different types of modes as belonging to peaks in the growth rate , we saw the close relationship between fdo and the differential flow instability .
one can be viewed as a continuous deformation of the other .
a primary effect of differential flow is to shift the fdo peak .
what has been referred to as the differential flow instability ( the appearance of a travelling wave instability in a medium which is neither hopf nor turing unstable ) can be interpreted as a case in which a sub - threshold fdo peak is shifted sufficiently to bring it above zero and to create unstable modes ( fig .
[ puredifi ] ) .
we now comment on experimental verifications of the present predictions .
the presence of two peaks could be seen in an experiment in which the perturbation frequency at the inflow is the control parameter . frequencies for which the growth rate is positive will result in sustained waves , while the waves will die out and fail to propagate if the growth rate is negative .
based on our results we expect that growing waves will appear within at most two frequency ranges .
the phase velocities can also be measured and compared with our general findings .
two types of experiments may be envisaged , in which oscillatory driving is implemented differently .
the first type @xcite@xcite@xcite makes use of a linearly growing , light - sensitive reaction - diffusion system .
the effective moving boundary is provided by a moving mask which extinguishes the reaction on the illuminated side of a moving line .
the illumination at the moving boundary can be modulated periodically , resulting in an oscillatory perturbation . in these experiments ,
differential diffusion @xmath20 is achieved by immobilizing one species on a gel , but differential flow is absent , @xmath81 .
the second type of experiment is conducted in a packed bed reactor ( pbr ) , which is by the outlet of a continuous stirred tank reactor ( cstr).@xcite@xcite@xcite@xcite@xcite@xcite the cstr may be manipulated to be stationary or to oscillate slower or faster than the medium in the pbr .
this leads to stationary , upstream and downstream moving waves @xcite@xcite@xcite@xcite@xcite . by packing the pbr with ion - exchanger beads that immobilize either activator or inhibitor ,
conditions of simultaneous differential diffusion and differential flow may be obtained .
@xcite@xcite modelled this by setting @xmath111 it is a challenge to devise experiments in which the flow ratio , diffusion ratio and driving frequency can be varied independently .
while the phase velocities of travelling waves can easily be measured , verification of other properties of the waves may present experimental challenges .
verification of the predicted phase relationships would require simultaneous measurements of both activator and inhibitor concentrations . | motivated by recent experiments and models of biological segmentation , we analyze the excitation of pattern - forming instabilities in convectively unstable reaction - diffusion - advection systems , occuring by constant or periodic forcing at the upstream boundary .
such boundary - controlled pattern selection is a generalization of the flow - distributed - oscillation ( fdo ) mechanism that can be modified to include differential diffusion ( turing ) and differential flow ( difi ) modes .
our goal is to clarify the relationships among these mechanisms in the general case where there is differential flow as well as differential diffusion .
we do so by analyzing the dispersion relation for linear perturbations and showing how its solutions are affected by differential transport .
we find a close relationship between difi and fdo modes , while the turing mechanism gives rise to a distinct set of unstable modes .
finally , we illustrate the relevance of the dispersion relations using nonlinear simulations and we discuss experimental implications of our results . |
rich clusters of galaxies are the largest gravitationally bound structures in the universe and should confine a representative fraction of its mass .
therefore the study of their dynamical properties and radiation content should allow , amongst other things , interesting cosmological conclusions on the relative amounts of visible and dark baryonic matter , and of nonbaryonic matter ( @xcite ) .
another basic characteristic , due to energetic particle confinement , is the ratio of nonthermal to thermal energy in these objects . to a significant extent
that ratio should be pre - determined during the epoch of early starburst activity and thus preserve the energetic history of cluster formation .
the necessary confinement of the nonthermal particle components is intimately related to the existence of strong and chaotic magnetic fields in the intracluster medium ( icm ) , and we shall propose physical mechanisms for their early generation as well as for their present fluctuations . in principle ,
detailed ab initio simulations of the dynamics of cluster formation under the dominant gravitational influence of the dark matter component ( @xcite ) should establish the overall cosmological framework for the present considerations .
we rather start in a complementary way with the discussion of a simplified model of cluster formation and of chemical enrichment of the intracluster gas .
it has the advantage that it directly allows a discussion of the physical processes of nonthermal particle production and confinement .
the main part of the paper concerns a proposal of cluster magnetic field generation in terms of galactic winds due to early starbursts and their amplification effect on magnetic fields drawn out from the progenitors of today s cluster galaxies into intracluster space .
it is argued that due to these dynamical processes there is no need for the operation of a dissipative turbulent dynamo in the icm .
the ongoing cluster accretion naturally leads to a strong fluctuating part of the intracluster magnetic fields .
a detailed discussion of the nonthermal radiation from galaxy clusters will be given in a separate paper ( @xcite ) .
we shall be concerned here with rich clusters , i.e. conglomerates with typically more than 100 member galaxies .
they have typical radii @xmath2 few mpc and baryonic masses @xmath3 .
many such clusters are rather evolved and contain predominantly early type s0 and e - galaxies , at least in their inner parts .
examples for bright and relatively nearby clusters of this type are the perseus and the coma clusters with distances @xmath4 .
the perseus cluster is the brightest cluster in soft x - rays .
the large x - ray luminosity is due to the very hot ( @xmath5k ) , massive ( @xmath6 ) , and metal - rich ( @xmath7_{\rm cl } \simeq 0.35 [ { \rm fe}]_{\odot}$ ] ) icm gas ( @xcite ) . as a consequence
the gas pressures are extremely high , with @xmath8 ranging from @xmath9 .
the metallicity of the icm gas , for instance in terms of the fractional icm iron mass , is correlated with the total optical luminosity in the e and s0 galaxies of rich clusters ( @xcite ) .
the correlation supports the view that early starbursts due to galaxy - galaxy interactions of gas - rich progenitors have produced a large number of core collapse supernovae due to massive stars ( for simplicity referred to here as sne ) .
they should have heated the originally present interstellar gas and generated violent galactic winds which removed the interstellar medium , leaving gas - poor e and s0 galaxies behind .
this mass loss should have led to the observed strong chemical enrichment of the icm gas .
we also conjecture that the ionizing radiation , the winds , and the large - scale shocks from these early galaxy mergers - together with the hard radiation from agns - strongly heated the remaining primordial icm gas , and thus prevented further galaxy formation . a quantitative discussion of the dynamical prerequisites for galactic winds and of the total number of sne in clusters is given by vlk et al .
( ( * ? ? ?
* hereafter referred to as paper i ) ) .
the total number of sne since galaxy formation in the cluster , roughly a hubble time @xmath10 ago , is then given by @xmath11}_{\odot } \times m_{\rm cl}}{\delta m_{\rm fe}}\;,\ ] ] where @xmath12 is the amount of iron produced per event . in such starbursts
we dominantly expect core collapse sne from massive progenitor stars to occur , with @xmath13 on average . for the perseus cluster
this implies @xmath14 .
the corresponding total energy input into the interstellar medium is @xmath15 , where @xmath16 is the average hydrodynamic energy release per sn in units of @xmath17 . assuming the early starbursts to occur at a typical redshift of @xmath18 due to the merging of gas - rich progenitors in an overdense protocluster environment ( @xcite ) , with a duration of @xmath19
, we obtain @xmath20 where @xmath21 is taken as 1/(30yr ) , and @xmath22 denotes the number of galaxies in the perseus cluster . as an example we can compare to the archetypical contemporary starburst galaxy @xmath23 .
it has a current sn rate @xmath24 , a wind velocity @xmath25 , and a mass - loss rate of @xmath26 ( @xcite ) .
the starburst nucleus of m82 is characterized by the following values for the interstellar gas temperature @xmath27 , gas density @xmath28 , and thermal gas pressure @xmath29 at the base of the wind : @xmath30 , @xmath31 , and @xmath32 ( @xcite ) .
since the thermal icm gas pressure in the perseus cluster is @xmath33 , it is clear that an object like m82 could readily drive a wind even against the _ present - day _ icm pressure . at the galaxy formation epoch
the icm pressure should have been much smaller than this value . in
an expanding galactic wind flow the sn - heated gas will cool adiabatically to quite small temperatures .
however it will be reheated in the termination shock , where the ram pressure of the wind adjusts to the icm pressure .
much beyond this point the ejected galactic gas is expected to slowly exchange energy and some metal - rich material with the unprocessed icm gas .
cluster formation also implies the production of a strong nonthermal component of relativistic particles .
they will be accelerated during the early phase - and presumably also in later accretion events that have shock waves associated with them . in the supernova remnants ( snrs )
the main acceleration should occur at the outer shock , with a very high efficiency of @xmath34 ( @xcite ; @xcite ) , the rest of @xmath35 going into the thermal gas ( @xmath36 ) ( @xcite ) , and radiation ( @xmath37 percent ) .
however , since the particles are ultimately removed from the galaxies in a strong galactic wind ( see paper i and section 3 .
below ) , they will cool adiabatically like the thermal gas and transfer their energy to the kinetic energy of the wind flow . as a consequence
the original snr - accelerated particles should constitute a negligible fraction of the present - day nonthermal particle content of the cluster . but this is not the end of the story . at distances
@xmath38 from the galaxies , fresh particle acceleration will occur at the strong galactic wind termination shocks ( fig .
[ fig1 ] ) .
we estimate the overall acceleration efficiency in these shocks to be again of the order of 10 percent . over the early phase of galaxy formation , using our estimate for the total number of sne , this should result in a total gas internal energy in the postshock region @xmath39 erg and a nonthermal energy @xmath40 erg for a system like the perseus cluster , ultimately driven by star formation and the subsequent sn explosions .
since the confinement time in the cluster ( see next subsection ) exceeds the cluster lifetime , the energy spectrum of these particles in the icm is basically the same as the source spectrum at the termination shocks . in cosmic ray ( cr )
parlance these are _ cosmological _ crs .
since the galaxies are distributed across the cluster quasi - uniformly , this should originally also be true for the nonthermal particle population .
the ensuing gravitational contraction / accretion of the cluster will subsequently energize crs and thermal gas at least adiabatically or , more likely , shock accelerate / irreversibly heat both components so that finally the total energy @xmath41 of energetic particles should reach at least the adiabatic value @xmath42 in the cluster ; @xmath43 now denotes the total present internal energy of the icm gas ( paper i ) .
it is worthwhile to compare the expected nonthermal energy with the thermal energy content of the cluster galaxies . assuming the stars internally to be in virial equilibrium and , for purposes of estimate , all of them to have a solar mass and radius , then @xmath44 . for a total mass of about @xmath45 contained in the _ galaxies _ of the perseus cluster
this gives a total thermal energy in stars @xmath46 , and thus @xmath47 .
this means that the nonthermal icm energy should be at least as large as the total thermal energy content of all the stars in all the galaxies together .
it has been argued more recently that , apart from star formation and overall gravitational contraction , also individual giant radio galaxies should have injected large and in fact comparable amounts of nonthermal particles during the life time of a cluster ( @xcite ) .
this is no doubt an important additional possibility .
a weakness of this argument consists in the fact that per se it is predicated on statistical knowledge about the luminosity function for active galaxies in clusters in general , and not on direct observations of the individual cluster to which it is applied .
energetic particle confinement has been discussed in a number of papers in recent years ( paper i , berezinsky et al .
1997 , colafrancesco & blasi 1998 ) .
we review here this important point , also in the light of the magnetic field structure to be discussed in the next section
. the large - scale magnetic field in the icm gas may be quite chaotic and not well connected over distances exceeding typical intergalactic distances .
thus energetic particles may not readily escape from the cluster due to such topological characteristics .
however , already a consideration of pure pitch angle diffusion along straight magnetic field lines with superposed turbulent fluctuations gives important insights into the confinement properties of galaxy clusters .
standard quasilinear theory yields a spatial diffusion coefficient @xmath48 along the large scale field @xmath49 due to a power spectrum @xmath50 of magnetic field fluctuations with wavelength @xmath51 in the following form : @xmath52 where @xmath53 is the gyro radius of a particle with momentum @xmath29 , @xmath54 , and @xmath55 denotes the wavenumber of the field fluctuations .
let us assume a relative fluctuation field strength of order unity at the inter - galaxy distance @xmath56 , i.e. a totally turbulent field @xmath57 on this scale , and a power law form of @xmath58 .
then the diffusion time across the cluster @xmath59 for @xmath60 , for @xmath61 and @xmath62 , respectively ( paper i ) .
also @xmath63 for nuclear collisions in the icm gas .
therefore - except at subrelativistic energies with their prevailing coulomb losses - up to these energies cr hadrons should accumulate in the cluster since the galaxy formation epoch , and that is what we called _ cosmological _ crs before .
the situation is different for relativistic electrons , which suffer radiative losses .
electrons , and the nonthermal radiation from galaxy clusters are discussed extensively in the paper by atoyan & vlk ( 2000 ) .
the accumulated cosmological cr protons and nuclei will produce high energy @xmath64 rays from inelastic pp - collisions on the intracluster gas which , in particular , lead to @xmath65-production and subsequent decay .
[ fig2 ] shows the @xmath65-decay energy fluxes expected from the hadronic crs of the coma cluster with a differential energy spectrum @xmath66 , with @xmath67 and an upper cutoff energy @xmath68 .
the solid and the dashed curve assume @xmath69 and @xmath70 erg , respectively , in an icm with a gas density of @xmath71 .
also the egret upper limit by sreekumar et al .
( @xcite ) is shown .
the observed size of the radio emission produced by high - energy electrons in coma is about half a degree .
the brightness of the _ hadronic _ tev emission should be significantly higher in the central region of the cluster , since the icm gas density strongly increases towards the center .
detection of an extended and weak ( @xmath720.1 crab ) tev flux by current instruments is problematic , but such fluxes are quite accessible for future imaging atmospheric cherenkov telescope arrays like h.e.s.s . , veritas , and cangaroo iii ( see fig.2 ) .
the magnetic field strengths in the icm of rich clusters , which may be as large as @xmath73 as inferred by faraday rotation measurements ( @xcite ) , are not easily explained by a contemporary dissipative mechanism because present - day turbulent dynamo effects in such a large - scale system should be extremely slow .
this problem is compounded by the extraordinary smallness of the expected intergalactic seed fields .
therefore we suggest a field configuration that is due to the early formation history of galaxy clusters as discussed above ; it should be preserved in its essential features to this day ( @xcite ) .
this field should even be still in a state of development at the present epoch .
the model derives from the violent early galactic winds which accompany the starbursts responsible for the predominance of the early type galaxies in rich clusters . in a general form
the ejection of galactic fields has been discussed by kronberg ( @xcite . for the generation of the general intergalactic magnetic field , related arguments have been advanced independently by kronberg et al .
( @xcite ) , assuming very early formation of dwarf galaxies with wind outflows at redshifts @xmath74 .
we assume first of all that the gas - rich progenitors whose mergers supposedly constitute the building blocks for the e and s0 galaxies , can be specifically pictured as protospirals that had already generated galactic magnetic fields of @xmath75 g strength .
this should indeed be possible within a time of @xmath76 yr or less , i.e. a time scale of the order of a rotation period of our galaxy or even shorter .
the explanation derives from fast turbulent dynamo action that invokes boyancy effects due to crs that inflate the magnetic flux tubes together with magnetic reconnection over spatial scales of order @xmath77 ( @xcite ) . in starburst galaxies
also systematic dynamo effects might play an important role ( section 3.1 ) . in a second stage ,
i.e. during the galaxy mergers , the resulting supersonic galactic winds will extend these fields from the interacting galaxies to almost intergalactic distances . in the final and by far longest stage that lasts until now
, the fields should be recompressed by the contraction of the cluster to its present size .
in addition , the continuing accretion of subclusters and individual galaxies constantly perturbs this field , keeping the fluctuations around this large scale field at a high level .
the icm fields do not reconnect on the intergalactic scale in a hubble time .
consequently there is no need for a continuous regeneration of these fields since their formation .
however , this also implies that a topologically connected overall icm field will on average not be formed either and that the icm field is chaotic on a scale smaller or equal to the present intergalactic distance .
in detail we draw on arguments which we have used in the past for the field configuration in a galactic wind from our own galaxy ( @xcite ) , see also fig .
they are based on estimates of the relative amount of field line reconnection vs. the extension of galactic field lines by a wind to `` infinity '' ( @xcite ) .
the basic result is that the rates of reconnection - and thus of the formation of `` parker bubbles '' leaving the galaxy by their boyancy and allowing the generation of the disk magnetic field - and the rates of extension of this field into the galactic halo by the pressure forces of the wind are roughly equal . thus both effects occur with about equal probability . for the cluster galaxies
this means that magnetic energy can be generated on the large scale of the wind at the expense of the thermal and nonthermal enthalpies produced in the starburst .
the geometry of the field should roughly correspond to straight field lines out to meridional distances @xmath78 of the order of the starburst ( sb ) radius , @xmath79 kpc in the protogalactic disk , and spherically diverging field lines beyond that .
the slow rotation of the system should then lead to an azimuthal field component , decreasing with the wind distance @xmath80 , which dominates at large distances over the radial component .
however , in contrast to the familiar situation in the solar wind equatorial plane , the axis of rotation is rather parallel than perpendicular to the flow at the base of the wind , and thus the dominance of the azimuthal field component is by no means as drastic as in the case of a stellar wind ( fig .
[ fig3 ] ) .
the wind becomes supersonic at about the same critical distance , @xmath81 .
far beyond this critical point the mass velocity @xmath82 becomes constant and the density falls off @xmath83 . choosing a present average baryon density @xmath84 , i.e. assuming most of the baryonic matter to be in the form of intergalactic gas , the mean density at the formation stage of the early type galaxies ( at redshifts @xmath85 ) was then @xmath86 . with a present icm density @xmath87
, we have @xmath88 due to gravitational compression of the icm gas . with a dominant thermal gas pressure the corresponding adiabatic pressure increase @xmath89 amounts to @xmath90 ; for the following we shall use @xmath91 . the wind termination shock distance @xmath92 is then given by @xmath93 , where @xmath94 is the wind mass density upstream of the shock . to estimate the wind characteristics we assume a quasi - steady state and a strong starburst for which we may disregard gravity , magnetic forces and cr pressure gradients in the overall energy balance equation ( e.g. @xcite ) .
it reads in this case : @xmath95_s \simeq \frac{\gamma}{\gamma -1}\frac { p_{\rm base}}{\rho_{\rm base}},\ ] ] where in addition @xmath96 is assumed at the base @xmath97 of the wind .
the critical point @xmath98 is given by @xmath99 .
approximately @xmath100 , where @xmath101 is the asymptotic wind speed . beyond the critical point at @xmath102
the wind achieves spherical symmetry so that from mass conservation @xmath103 if the icm pressure is small compared to the base pressure then we can assume that the wind is highly supersonic at the shock distance which implies @xmath104 .
this should be true for m82-like objects - although they should have a considerably larger scale @xmath105 kpc - which have @xmath106 .
in this case we can assume the wind pressure to evolve adiabatically to lowest order , so that @xmath107 as a result , taking into account that in a flow dominated by the thermal gas the adiabatic index @xmath108 @xmath109^{1/2 } \left(\frac{p_{\rm base}}{p_{\rm cl}}\right)^{1/2 } \simeq 1.27~ \left(\frac{p_{\rm base}}{p_{\rm cl}}\right)^{1/2 } \simeq 400\ ] ] for the above starburst parameters .
therefore @xmath110 the wind bubble containing the shock - heated wind gas will have a radius still exceeding @xmath92 . beyond the bubble , part of the external gas will be shocked by the rapidly expanding bubble gas .
this shock heated gas will exchange energy with the cold ambient gas by heat conduction and instabilities . to some extent such exchanges
will also take place with the bubble gas .
even though the volume of at least initially unmagnetized icm gas may be large enough so that the icm gas mass exceeds the mass associated with galaxies by a factor of a few - as observed - it may therefore be that some wind bubbles touch .
thus we should consider whether the magnetic fields of the bubbles can reconnect with each other to produce field lines that pervade the entire cluster .
let us assume that the fastest rate of reconnection proceeds with a speed between 1 and 10 percent of the alfv@xmath111n speed @xmath112 , to be specific , with @xmath113 ( @xcite ) .
then the present - day reconnection time across a termination shock scale is @xmath114 greater than a hubble time , even for a present - day icm magnetic field as high as @xmath1 g. therefore , many bubble fields might not yet be reconnected and much of the field structure could well remain topologically disconnected until today .
the field strength @xmath115 in the early wind bubbles should be of the order of @xmath116 or somewhat larger , if the field in the bubble increases in the decelerating postshock flow .
the subsequent and still ongoing overall cluster contraction / accretion compresses the field to lowest order isotropically @xmath117 , with the scale factor @xmath118^{1/3}/(1+z),\ ] ] where @xmath119 and @xmath120 . for a present mean baryon number density @xmath121
, we obtain @xmath122 . from these estimates ,
we obtain for the present - day icm field : @xmath123 .
therefore the present - day icm magnetic field should have a mean strength of the order of @xmath0 g , from @xmath124 `` primordial '' seed fields , and should be randomly directed on an intergalactic scale .
although smaller by about one order of magnitude than estimated from faraday rotation measurements , such fields need not necessarily be unrealistic , considering that observations might emphasize regions of high magnetic fields .
indeed the simplest inverse compton interpretation of the euv excess and the excess x - ray flux in the coma cluster requires such low field strengths ( @xcite ) .
an additional possibility is that the `` initial '' fields for such starburst galaxies might be an order of magnitude stronger than assumed .
apart from a fast dynamo whose strength could be directly proportional to the star formation rate ( @xcite ) , we could invoke a systematic field amplification through the commencing galactic outflow that `` combs '' the field outward .
it might occur on very short time scales of about @xmath125 yr .
empirically this field amplification is suggested by the statistical time - independence of the radio synchrotron to far infrared emission ratio in starburst galaxies ( @xcite ) which can hardly be understood otherwise than through a field that increase almost simultaneously in strength with the star formation rate .
a final argument is that the large scale shear deformations induced by the later accretion of large subclusters may amplify the field even further .
thus we can not exclude @xmath75 g fields ; the force balance certainly allows them .
the wind bubbles and the associated magnetic fields should at some stage decouple from the galaxies they emanated from , simply by magnetic reconnection which is fastest near the galaxies : the alfv@xmath111n velocity is approximately independent of the meridional distance s in the wind , whereas the distance between oppositely directed field lines that emanate from the galaxy is obviously the smaller the smaller s is .
thus , after the termination of the starburst , the magnetized bubbles and the stellar component of the remaining early type galaxies should acquire independent identities and their dynamics should decouple .
the topologically disconnected structure of the mean magnetic field in the cluster has also some bearing on the evolution of magnetic field strengths during the development of cooling flows towards the cluster center ( @xcite ) . instead of building up global magnetic pressure gradients and tension forces such a subsonic and sub - alfv@xmath111nic
flow allows optimal internal segregation of high and low field regions on galaxy - galaxy seperation scales .
thus we should expect that the field strength towards the cluster center increases by less than by isotropic compression , @xmath126 , even though at the compressed spatial scales reconnection will be more effective .
the two - thirds law should hold only for the overall cluster gas compression discussed earlier .
a question of direct importance for the interpretation of faraday rotation measurements is the scale of reversal changes of these fields .
field reversals are a natural consequence of the suggested field structure , given the field pattern in the interstellar media of the starburst galaxies .
if the milky way can serve as a guide , then this pattern is determined by the 100 pc scale of the parker instability ( parker 1966 ) . from the radius @xmath127 to distances @xmath128 this scale projects like the ratio @xmath129 . using the above estimates this implies a field reversal scale @xmath130 kpc in clusters , rather well in line with observational estimates which indicate reversal scales of @xmath131 to @xmath132 kpc ( @xcite ; @xcite ) . from the foregoing arguments there
is hardly any need for a contemporary `` turbulent icm dynamo '' .
however , the ongoing accretion will perturb this mean intracluster field randomly in space and time , maintaining a turbulent magnetic fluctuation field that develops smaller and smaller spatial scales .
due to the topology of the mean field , the largest turbulent scale is given by the distance between galaxies .
the accretion will probably span the range from single field galaxies falling into the cluster to accreting massive subclusters .
as long as this accretion process remains important , one has to expect that the cluster rings with it . at the largest scale
the relative magnetic fluctuation level should approach unity ( see subsection 2.3 ) .
aharonian , f. a. , hofmann , w. , konopelko , a. k. , & vlk , h. j. 1997 , astropart .
phys . , 6 , 343
; 369 arnaud , m. , rothenflug , r. , boulade , o. , vigroux , l. , & vangioni - flam , e. 1992 , a&a , 254 , 49 atoyan , a. m. , & vlk , h. j. 1999 , apj in the press berezhko , e.g. , & vlk , h.j .
2000 , apj submitted berezinsky , v. s. , blasi , p. , & ptuskin , v. s. 1997 , , 487 , 529 bhringer , h. 1996 , extragalactic radio sources , r. ekers et al . , 1996 ,
iau , 357 breitschwerdt d. , mckenzie j. f. , & vlk h. j. 1993 , a&a , 269 , 54 breitschwerdt d. , 1994 , habilitationschrift , univ .
heidelberg , 158 bibitem[clarke et al .
1999]clarke99 clarke , t.e . , kronberg , p.p . ,
bhringer , h. 1999 , proc . of the ringberg workshop : `` diffuse thermal and relativistic plasma in galaxy clusters '' , ringberg castle , germany , april 19 - 23 , 1999 , eds .
h. bhringer , l. ferretti & p. schcker , mpe report 271 , 82 colafrancesco , s. , & blasi , p. 1998 , astropart .
phys . 9 , 227 dolag , k. m. , bartelmann , m. , & lesch , h. 1999 , a&a 348 , 351 dorfi , e. a. 1993 , galactic high - energy astrophysics , high - accuracy timing and positional astronomy , j. van paradijs and h. m. maitzen , lecture notes in physics 418 , berlin : springer , 1993 , 43 drury , l. oc . , markiewicz , w. j. , & vlk , h. j. 1989 , a&a , 225 , 179 enlin , t. a. , biermann p. l. , kronberg , p. p. , wu , x .- p .
1997 , , 477 , 560 fabian , a.c .
1994 , araa 32 , 277 fusco - femiano , r. , dal fiume , d. , feretti , l. , giovannini , g. , grandi , p. , matt , g. , molendi , s. , & santangelo , a. 1999 , , 513 , l21 | we propose a mechanism for the early generation of the mean intracluster magnetic field in terms of magnetized galactic winds .
these winds are the result of starburst phases of the cluster galaxies , assumed to produce the predominant population of early type galaxies in mergers of gas - rich progenitors . after further cluster contraction
typical field strengths are @xmath0 gauss .
this estimate may increase to the level of @xmath1 gauss if more extreme galactic parameters , and subsequent shear amplification of the field are considered .
the topology of the field is one of almost unconnected wind bubbles with parker - type spiral field configurations over scales of the distance between galaxies .
further cluster accretion , that continues chaotically in space and time up to the present , will perturb these `` large - scale '' mean fields on smaller or at best comparable spatial scales .
the small scale fields in the resulting turbulent fluctuation spectrum should be able to confine relativistic particles over times longer than the age of the universe .
the nonthermal particle content of galaxy clusters should therefore also have a `` cosmological '' hadronic component generated during the early starburst phase of the member galaxies . already by itself
it implies a nonthermal energy fraction of about 10 percent for the intracluster gas which should then be detectable by future gamma - ray telescopes . |
game theory has recently been used as an effective tool to model and solve many security problems in computer and communication networks . in a noncooperative matrix game between an attacker and a defender ,
if the payoff matrices are assumed to be known to both players , each player can compute the set of nash equilibria of the game and play one of these strategies to maximize her expected gain ( or minimize its expected loss ) .
however , in practice , the players do not necessarily have full knowledge of each other s payoff matrix .
for repeated games , a mechanism called fictitious play ( fp ) can be used for each player to learn her opponent s motivations . in a fp process
, each player observes all the actions and makes estimates of the mixed strategy of her opponent . at each stage , she updates this estimate and plays the pure strategy that is the best response ( or generated based on the best response ) to the current estimate of the other s mixed strategy .
it can be seen that in a fp process , if one player plays a fixed strategy ( either of the pure or mixed type ) , the other player s sequence of strategies will converge to the best response to this fixed strategy .
furthermore , it has been shown that , for many classes of games , such a fp process will finally render both players playing a nash equilibrium ( ne ) .
specifically , we examine a two - player game , where an attacker ( denoted as player @xmath0 or @xmath1 ) and a defender ( denoted as player @xmath2 or @xmath3 ) participate in a discrete - time repeated nonzero - sum matrix game . in a general
setting , the attacker has @xmath4 possible actions and the defender has @xmath5 posssible actions to choose from .
for example , when @xmath6 , the attacker s actions could be to attack one node in a two - node network , and those of the defender are to defend one of these two nodes .
players do not have access to each other s payoff function .
they adjust their strategies based on each other s actions which they observe . in a stochastic fp process
, each player makes a maximum likelihood estimation of her opponent s mixed strategy .
as will be seen later on , this will result in a time - varying update of the opponent s empirical frequency , where the weight of the action at time step @xmath7 is @xmath8 . in a practical repeated security game , however , we notice a couple of possible complications .
first , players may not have the exact and synchronized time steps .
second , each player may want to adjust the weight of the other s current action to converge either faster or more accurately to the equilibrium .
a more flexible scheme to update the estimate of the mixed strategy may be needed in such situations .
motivated by these practical considerations , we examine in this paper a time - invariant frequency update mechanism for fictitious play . also , as a side note , such a time - invariant update mechanism will allow us to use the analysis tools applicable only to time - invariant systems .
security games have been examined extensively in a large number of papers , see for example , @xcite .
surveys on applications of game theory to network security can be found in @xcite , @xcite .
relevant literature on fictitious play can be found in @xcite .
a comprehensive exposition of learning in games can be found in @xcite .
the rest of this paper is organized as follows . in section [ sec : tifp ] , we provide an overview of the static game and the standard stochastic fp process , and then introduce the stochastic fp with time - invariant frequency update .
the analysis for fp with time - invariant frequency update is given in section [ sec : analysis ] . in section [ sec : atifp ] , we introduce an adaptive algorithm based on the time - invariant fp process .
next , simulation results are given in section [ sec : sim ] .
finally , some concluding remarks will end the paper .
in this section , we present first an overview of a two - player static games , then the concept of stochastic fictitious play with time - varying frequency update ( tvfu - fp ) @xcite , and finally the concept of stochastic fictitious play with time - invariant frequency update ( tifu - fp ) .
while we introduce both classical version and stochastic version of static games , we restrict ourseves to only stochastic fictitious play in subsections [ ss : tvfp ] and [ ss : tifp ] and in the rest of the paper .
we consider here static security games , where each player @xmath9 , has two possible actions ( or pure strategies ) .
we use @xmath10 , to denote the action of @xmath11 .
let @xmath12 be the simplex in @xmath13 , i.e. , @xmath14 each @xmath10 takes value in the set of ( two ) vertices of @xmath12 : @xmath15^t$ ] for the first action , and @xmath16^t$ ] for the second action . in a static game ,
player @xmath11 selects an action @xmath10 according to a mixed strategy @xmath17 . the ( instant ) payoff for player
@xmath11 is is used to indicate those of other players , or the opponent in this case . ]
@xmath18 , where @xmath19 is the payoff matrix of @xmath11 , and @xmath20 is the entropy of the probability vector @xmath21 , @xmath22 . the weighted entropy @xmath23 with @xmath24
is introduced to boost mixed strategies . in a security game
, @xmath25 signifies how much player @xmath26 wants to randomize its actions , and thus is not necessarily known to the other player . also , for @xmath27 ( referred to as classical fp ) , the best response mapping can be set - valued , while it has a unique value when @xmath28 ( referred to as stochastic fp ) . for a pair of mixed strategy @xmath29 ,
the utility functions are given by the expected payoffs : @xmath30 + \tau_i h(p_i ) \\ & = & p_i^t m_i p_{-i } + \tau_i h(p_i ) .
\end{aligned}\ ] ] now , the _ best response _
mappings @xmath31 are defined as : @xmath32 if @xmath28 , the best response is unique as mentioned earlier , and is given by : @xmath33 where the soft - max function @xmath34 is defined as @xmath35 note that @xmath36 , and thus the range of the soft - max function is just the interior of the simplex . finally , a ( mixed strategy ) nash equilibrium is defined to be a pair @xmath37 such that for all @xmath17 @xmath38 we can also write a nash equilibrium @xmath39 as the fixed point of the best response mappings : @xmath40 from the static game described in subsection [ static ] , we define the discrete - time tvfu - fp as follows .
suppose that the game is repeated at times @xmath41 .
the empirical frequency @xmath42 of player @xmath11 is given by @xmath43 using induction , we can prove the following recursive relation : @xmath44 from the equations of discrete - time tvfu - fp ( [ em_freq ] ) , ( [ rec_em_freq ] ) , the continuous - time version of the iteration can be written down as follows @xcite : @xmath45 in tvfu - fp , players take the maximum likelihood estimate of the mixed strategy of their opponent ( [ em_freq ] ) , ( [ rec_em_freq ] ) . in tifu - fp , the estimates of the mixed strategies will be calculated in a time - invariant manner as follows : @xmath46 where @xmath47 is a constant and @xmath48 . for each player , this is basically the exponential smoothing formula used in time series analysis ( see for example @xcite ) .
we will prove that with this formulation , at time @xmath7 , @xmath49 will be a weighted average of all the actions up to present of player @xmath26 where more recent actions have higher weights .
suppose that the payoff matrices of player @xmath0 and player @xmath2 are , respectively , @xmath50 [ as : payoff ] based on a realistic security game , we can make the following assumptions : * @xmath51 : when the defender defends , the payoff of the attacker will be decreased if it attacks .
* @xmath52 : when the defender does not defend , the payoff of the attacker will be increased if it attacks . *
@xmath53 : when the attacker attacks , the payoff of the defender will be decreased if it does not defend .
* @xmath54 : when the attacker does not attack , the payoff of the defender will be increased if it does not defend .
given payoff matrix @xmath19 , coefficient @xmath28 , @xmath55 .
update the estimated frequency of the opponent using ( [ tiest1 ] ) , ( [ tiest2 ] ) . compute the best response using ( [ best_response ] ) .
( note that the result is always a completely mixed strategy . )
randomly play an action @xmath56 according to the best response mixed strategy @xmath57 .
in tifu - fp , both players employ algorithm [ alg : tifp ] .
the mean dynamic of the evolution of tifu - fp can be written as : @xmath58 note that equations ( [ est_mean_dynamic ] ) are just evolution of the _ estimated frequencies _ ; the _ empirical frequencies _ still evolve in a time - varying manner : @xmath59 the mean dynamic of empirical frequencies then can be written as @xmath60
we start the analysis with the following result for the static games given in subsection [ static ] .
the static @xmath2-player @xmath2-action game in [ static ] with assumption [ as : payoff ] and @xmath61 admits a unique nash equilibrium .
-player @xmath2-action game in [ static ] with assumption [ as : payoff ] and @xmath61 - best response mappings.,title="fig:",width=302 ] + in what follows , let @xmath62 , @xmath63 , @xmath64 , and @xmath65 .
we first use the brouwer fixed point theorem ( see for example @xcite ) to prove the existence of a nash equilibrium , then use monotonicity of @xmath66 and @xmath67 to prove that the fixed point is unique . here
we write @xmath66 as a scalar - valued function of the only independent variable @xmath68 using the fact that @xmath69 , or @xmath70 .
function @xmath67 is defined similarly .
also , as @xmath71 , a pair @xmath72 completely specifies the estimated frequencies of the players . as seen from equation ( [ best_response_mapping ] )
, a nash equilibrium @xmath73 is a fixed point of the best response mapping : @xmath74 it suffices to write this mapping as @xmath75 , where @xmath76 .
specifically , the mapping @xmath77 can be detailed as : @xmath78 \right\ } } } { e^ { \left\ { \frac{1}{\tau_1 } [ a r^1_2 + b(1-r^1_2 ) ] \right\}}+e^ { \left\ { \frac{1}{\tau_1 } [ c r^1_2 + d(1-r^1_2 ) ] \right\}}}.\end{aligned}\ ] ] it can be seen that @xmath79 .
similarly , we have @xmath80 thus @xmath77 is a transformation from @xmath81 ^ 2 $ ] to @xmath81 ^ 2 $ ] , which is a compact convex set . as both mappings @xmath82 and @xmath83 that constitute @xmath77 are continuous , @xmath77 is also a continuous transformation . using the brouwer fixed point theorem
, there exists at least one fixed point @xmath84 such that @xmath85 , which is a nash equilibrium of the static game .
now we examine the derivatives of @xmath66 and @xmath67 with respect to their own independent variables : @xmath86 \beta^1_1(r_2 ) \beta^2_1(r_2),\\ \nonumber \frac { d\beta^1_2(r_1)}{d r^1_1 } & = & \frac{1}{\tau_2 } [ ( e - f)+(h - g)]\beta^2_1(r_1 ) \beta^2_2(r_1).\end{aligned}\ ] ] from assumption [ as : payoff ] , @xmath87 and @xmath88 .
thus @xmath66 is strictly decreasing in @xmath89 , and @xmath67 is strictly increasing in @xmath90 . now
suppose that there exist two distinct nash equilibria , @xmath91 and @xmath92 .
obviously , @xmath93 , otherwise we will have @xmath94 , and these two points coincide . without loss of generality , assume that @xmath95 . as @xmath67 is strictly increasing in @xmath90 , we have that @xmath96 .
however , @xmath66 is strictly decreasing in @xmath89 , so @xmath97 , which is contradictory to the initial assumption .
thus the nash equilibrium is unique .
we illustrate in figure [ fig : sfp_best_response ] the curves @xmath66 and @xmath67 with the values of @xmath98 , @xmath99 , @xmath100 , and @xmath101 as shown .
the intersection of these two curves is the nash equilibrium of the static game .
we present here two propositions for tifu - fp : the first shows the weights of each player s actions in the estimated frequency , and the second shows the relationship between estimated frequencies and empirical frequencies . for @xmath102 , the estimated frequencies in tifu - fp constructed using ( [ tiest1 ] ) , ( [ tiest2 ] ) will satisfy @xmath103 where @xmath55 . this result can be proved using induction . in tifu - fp ,
the empirical frequencies are related to the estimated frequencies calculated using ( [ tiest1 ] ) , ( [ tiest2 ] ) through the following equation : @xmath104 this result can be proved by writing the actions of player @xmath11 at times @xmath105 in terms of the estimated frequencies at times @xmath106 . [ thm : conv ] consider a tifu - fp with assumption [ as : payoff ] and @xmath61 .
the mean dynamic given in equations ( [ est_mean_dynamic ] ) is asymptotically stable if and only if @xmath107\left[(e - f)+(h - g)\right]}{\tau_1 \tau_2 } \bar{r}^1_1 \bar{r}^2_1 \bar{r}^2_1 \bar{r}^2_2 + 1 } .\ ] ] as can be seen in equations ( [ est_mean_dynamic ] ) , this is a deterministic nonlinear discrete - time time - invariant system .
we linearize the system at the fixed point and examine stability properties of the linearized system using techniques described in standard textbooks for nonlinear systems ( e.g. , @xcite ) . using the mean dynamic ( [ est_mean_dynamic ] ) , where @xmath108 it can be seen that a pair @xmath109 that satisfies @xmath110 , is a fixed point of the system .
consider the jacobian matrix @xmath111 we have that @xmath112 recall that @xmath113 , where @xmath114 \\ \frac{1}{\tau_1 } [ c r^1_2 + d(1-r^1_2 ) ] \\ \end{array } \right).\ ] ] thus @xmath115 \right\ } } } { e^ { \left\ { \frac{1}{\tau_1 } [ a r^1_2 + b(1-r^1_2 ) ] \right\}}+e^ { \left\ { \frac{1}{\tau_1 } [ c r^1_2 + d(1-r^1_2 ) ] \right\}}}.\ ] ] then @xmath116 \beta^1_1(r_2 ) \beta^2_1(r_2),\ ] ] @xmath117 \beta^1_1(r_2 ) \beta^2_1(r_2).\ ] ] at the fixed point @xmath109 , we can write @xmath118 \bar{r}^1_1 \bar{r}^2_1.\ ] ] similarly , @xmath119\bar{r}^1_2 \bar{r}^2_2 .\ ] ] using the conditions for local stability , @xmath120 , where @xmath121 are eigenvalues of the jacobian matrix , we finally have the condition in equation ( [ conv_tifufp ] ) .
although this theorem only mentions the asymptotic stability of the estimated frequencies ( of the mean dynamic ) , once these estimated frequencies converge to the nash equilibrium , the best responses will also converge to the nash equilibrium , and so will the empirical frequencies in the long run .
in this section we examine an adaptive fp algorithm ( hereafter referred to as afp ) based on fp with time - invariant frequency update , where the step size @xmath47 is piecewise constant and decreased over time . for the specific implementation shown in algorithm [ alg : atifp ] ,
the step size is either kept fixed or halved , based on the variance of empirical frequency in the previous time window . given payoff matrix @xmath19 , coefficient @xmath25 , @xmath55 , initial step size @xmath122 , minimum step size @xmath123 , and window size @xmath124 .
update the estimated frequency of the opponent , @xmath125 , using ( [ tiest1 ] ) , ( [ tiest2 ] ) .
compute the best response mixed strategy @xmath57 using ( [ best_response ] ) . randomly play an action @xmath126 according to the best response mixed strategy @xmath57 , such that the expectation @xmath127=\beta_i(r_{-i}(k))$ ] . compute the standard deviation of the estimated frequencies ( @xmath128 ) in the time window @xmath129 $ ] ( using an unbiased estimator ) : @xmath130 decrease step size : @xmath131 and @xmath132 . keep step size @xmath47 constant
we present in this section some simulation results for tifu - fp and afp where the payoff matrices and entropy coefficients are chosen to be @xmath133 the nash equilibrium of the static game is @xmath134 and @xmath135 .
the local stability threshold ( the rhs of equation ( [ conv_tifufp ] ) ) is @xmath136 . for simplicity , in the graphs shown here ,
we only plot the first component of each frequency vector .
some simulation results for the mean dynamic of tifu - fp ( equations ( [ est_mean_dynamic ] ) ) are given in figures [ fig : tisfp_est_freq_025 ] and [ fig : tisfp_est_freq_026 ] .
when @xmath137 , the estimated frequencies are shown in figure [ fig : tisfp_est_freq_025 ] .
the simulation results show that both estimated frequencies and empirical frequencies ( not presented here due to space limitations ) converge to the ne as expected .
when @xmath138 , however , the estimated frequencies do not converge anymore .
these simulations thus confirm the theoretical result in theorem [ thm : conv ] .
it is also worth noting that the empirical frequencies in the case @xmath139 still converge to the ne .
, @xmath136.,title="fig:",width=302 ] + , @xmath136.,title="fig:",width=302 ] + unlike the mean dynamic , a stochastic tifu - fp process ( generated with algorithm [ alg : tifp ] ) exhibits significant random fluctuations .
the graph in figure [ fig : tisfp_s_est_freq_001 ] shows the estimated frequencies of such a process where we choose @xmath140 .
however , the empirical frequencies ( whose graph is not shown here due to space limitations ) still converge to the ne .
.,title="fig:",width=302 ] + + some simulation results for adaptive fp are shown in figures [ fig : atisfp_emp_freq ] and [ fig : atisfp_step_size ] .
the payoff matrices and entropy coefficients are the same as those in [ ss : simtifp ] .
initial and minimum step sizes are chosen to be @xmath141 and @xmath142 , respectively .
the time window for updating the step size is @xmath143 steps .
the evolution of the empirical frequencies are depicted in figure [ fig : atisfp_emp_freq ] , which shows that adaptive fp converges faster than the stochastic fp with time - varying frequency update ( tvfu - fp ) ( figure [ fig : tvsfp_emp_freq ] ) .
we however remark that it is possible to incorporate a decreasing coefficient into the step size in tvfu - fp ( which is originally @xmath8 ) to make the tvfu - fp process converge faster @xcite .
the update of the step size in adaptive fp is shown in figure [ fig : atisfp_step_size ] .
note that when compared to the step size @xmath8 in tvfu - fp , the step sizes in adaptive fp are higher in the beginning and smaller afterwards , resulting in aggressive convergence first and less fluctuation in the stable phase .
in this paper , we have introduced a time - invariant scheme to estimate the frequency of the opponent s actions in a two - player two - action fictitious play process .
we have proved local stability of the unique nash equilibrium for the mean version of this fp dynamic .
this frequency update scheme , when used adaptively , allows players to converge faster to the nash equilibrium .
for this two - player two - action fp , conditions for global stability , if they exist , are yet to be found .
also , having more than two possible actions for each player is an intriguing research extension .
t. alpcan and t. baar . a game theoretic approach to decision and analysis in network intrusion detection . in _ proc . of the 42nd ieee conference on decision and control _
, pages 25952600 , maui , hi , december 2003 .
t. alpcan and t. baar . a game theoretic analysis of intrusion detection in access control systems . in _ proc . of the 43rd ieee conference on decision and control _ , pages 15681573 , paradise island , bahamas , december 2004 .
k. c. nguyen , t. alpcan , and t. baar .
stochastic games for security in networks with interdependent nodes . in _ proc . of intl . conf . on game theory for networks
( gamenets 2009 ) _ ,
istanbul , turkey , may 2009 . | we study two - player security games which can be viewed as sequences of nonzero - sum matrix games played by an attacker and a defender .
the evolution of the game is based on a stochastic fictitious play process , where players do not have access to each other s payoff matrix .
each has to observe the other s actions up to present and plays the action generated based on the best response to these observations . in a regular fictitious play process , each player makes a maximum likelihood estimate of her opponent s mixed strategy , which results in a time - varying update based on the previous estimate and current action . in this paper
, we explore an alternative scheme for frequency update , whose mean dynamic is instead time - invariant . we examine convergence properties of the mean dynamic of the fictitious play process with such an update scheme , and establish local stability of the equilibrium point when both players are restricted to two actions .
we also propose an adaptive algorithm based on this time - invariant frequency update . |
photonic metamaterials are trials to invent and produce novel electromagnetic ( em ) states by modifying periodic structures in which constituents have the same optical constants with the bulk materials .
the trials started at gigahertz , and has proceeded to terahertz and optical frequencies .
the rapid progress was reviewed in @xcite .
the strategy is significantly different from nano - scale materials science in which quantum size effect on electronic states is a key . in metamaterials ,
novel effective em states are explored by controlling the geometrical structures .
the effective em states are solutions of maxwell equations for a homogenized system obtained by coarse - graining and represent averaged em fields approximately .
recent metamaterial optics usually employs a retrieval way to evaluate effective optical constants @xcite .
the procedure is briefly explained as follows : when one knows complex transmissivity and reflectivity ( called @xmath0 parameters ) for a finitely thick object , one can numerically calculate ( or retrieves ) the effective permittivity and permeability by substituting the @xmath0 parameters into the transmission and reflection formulas , which hold for a finitely thick material of effective optical constants .
the way was first introduced in 2002 @xcite and has been called retrieval way . the retrieval way is algorithmic and not rigorous compared with the effective media description , which has been studied over one century @xcite since maxwell - garnet @xcite .
the effective media descriptions are usually derived for specified structures by assuming analytical models and the long wavelength limit .
the difference between the effective media description and recent retrieval way was already described in the introduction of @xcite . in short
, the effective media description is derived analytically for ideally large objects under the long wavelength approximation , whereas the retrieval way transforms the @xmath0 parameters into effective optical constants in a practical way . in the retrieval way
, there are two distinct points from the effective media description : ( i ) the interface of metamaterials play a crucial role because the @xmath0 parameters are sensitive to the surface layers ; ( ii ) the description using effective optical constants turned out to be approximately valid for several artificial metamaterials beyond the long wavelength limit @xcite . by use of the retrieval way , effective magnetic responses and negative refraction at optical frequencies have been demonstrated experimentally @xcite .
those metamaterials exhibit novel optical states , which are exceptions to the statement that permeability is unity at optical frequencies @xcite . a simple theoretical analysis presented a possibility of super - resolution by a thin metal film @xcite and the sub - diffraction - limited image was shown experimentally @xcite .
it is expected that near - em - field enhancement is induced to compensate the loss at resonance with exploiting thin metallic layers .
another experimental result on transmission was reported which supports the enhancement of em fields by stratified thin metals @xcite .
it was also shown that the enhancement is associated with high magnetic field in thin metals @xcite .
besides , it was recently argued that negative refractive response appears in stratified stacks composed of metal - dielectric several layers @xcite .
thus , the magnetic response in metallic thin films attracts great interest , connecting to novel optical phenamena . ) .
blue bold line indicates the unitcell of smdm .
incident light sheds on @xmath1 plane with @xmath2 polarization ( _ i.e. _ @xmath3 ) and incident angle @xmath4 . the thickness of surface layer is denoted by _
d_. in this configuration , the two products of @xmath5 and @xmath6 are determined from two incident angles .
details are given in sec .
[ fig1],width=283 ] in this paper , stratified metal - dielectric metamaterials ( smdm ) are specified to clarify the effective magnetic response at visible frequencies .
typical optical properties of smdm have been shown in an earlier publication @xcite . here , the effective magnetic response is focused on and explored with modifying the structure of unitcell .
figure [ fig1 ] shows the schematic drawing of smdm and the coordinate configuration .
the metal is set to be ag and the dielectric is mgf@xmath7 .
the thickness of metal and dielectric are 15 and 60 nm , respectively .
the periodicity is then 75 nm .
the thickness of ag layers was set to ensure using dielectric constant of bulk silver in literature @xcite . in sec .
[ tcrm ] , two - complex reflectivity method ( tcrm ) are described .
numerical results are shown in sec .
[ result ] , and the implication is explored to elucidate the origin of effective magnetic response .
more discussion and comparison with other types of magnetism reported so far are given in sec .
[ discussion ] .
in this study , bulk smdm is analyzed which is thick enough to eliminate transmission , and full components of @xmath8 and @xmath9 tensors are extracted by tcrm .
although the scheme of tcrm was reported already @xcite , the full description is provided in the following . since the structure of smdm is uniaxial , effective tensors of permittivity @xmath8 and permeability @xmath9
are assumed to be diagonal and uniaxial : @xmath10 the optical axis is set to be the @xmath11 axis , and then the two relations of @xmath12 and @xmath13 hold in eq .
( [ eps_mu ] ) . in linear and axial media ,
electric and magnetic flux densities are expressed using permittivity and permeability tensors as @xmath14 where @xmath15 and @xmath16 are permittivity and permeability in vacuum , respectively .
first , the boundary conditions of maxwell equations are taken into account . in fig .
[ fig1 ] , incident wavenumver vector @xmath17 is in the @xmath18 plane and then refracted wave has the wavevector @xmath19 in the @xmath18 plane .
we here introduce normalized refracted wavevector @xmath20 where @xmath21 is wavenumber of light in vacuum , and can write @xmath22 as @xmath23 then , complex reflectivity @xmath24 under @xmath2 polarization ( that is , @xmath3 ) in fig .
[ fig1 ] is easily derived from the maxwell boundary conditions @xcite ; the equation for @xmath24 is sometimes called fresnel formula @xcite .
after simple modification , the next equation is obtained : @xmath25 where @xmath4 is incident angle , and @xmath26 denotes the refractive index in vacuum .
in particular , @xmath27 is refractive index along the @xmath11 axis .
next , the equation of dispersion is included in tcrm . when the em fields are monochromatic plane waves and proportional to @xmath28 , the two of the maxwell equations are @xmath29 note that the fields @xmath30 , @xmath31 , @xmath32 , and @xmath33 in eqs .
( [ kxe ] ) and ( [ kxh ] ) are independent of @xmath34 .
when the solution of plane wave is assumed , it is generally allowed that the @xmath8 and @xmath9 tensors depend on the frequency @xmath35 and wavevector @xmath19 ; that is , the @xmath8 and @xmath9 tensors have frequency and spatial dispersions ( @xmath36 and @xmath37 ) @xcite .
however , it is for the present assumed that the @xmath8 and @xmath9 tensors are independent of * k*. in solid crystals , the @xmath9 tensor is set to be unity at optical frequencies and independent of * k * , and the spatial dispersion of @xmath8 tensor is often so weak that it is frequently negligible @xcite .
equations ( [ d = ee ] ) and ( [ kxh ] ) yield eq .
( [ pre_e_eq ] ) , and eq . ( [ e_eq ] ) is derived with help of eqs .
( [ b = mh ] ) and ( [ kxe ] ) : @xmath38 \label{e_eq } \\ & = & -\hat{\textbf{k } } \times \left [ \mu^{-1}% \left(\hat{\textbf{k } } \times \textbf{e}\right)\right ] . \label{n_e_eq}\end{aligned}\ ] ] equation ( [ n_e_eq ] ) is obtained by using @xmath39 and the relation @xmath40 the determinant of coefficient matrix for * e * in eq .
( [ n_e_eq ] ) has to be zero to have nontrivial solution ; therefore , eq .
( [ n_e_eq ] ) yields a general constraint between @xmath8 and @xmath9 for a given @xmath41 .
the constraint condition is equation of dispersion and can be regarded as generalization of fresnel equation .
historically , fresnel equation was derived under the condition of @xmath42 , gave the equation concerning @xmath8 for a given @xmath41 , and has been usually employed to explain optical phenomena , _
e.g. _ birefringence , in solid crystals @xcite . under _ s _ polarization , eq .
( [ n_e_eq ] ) is modified using eq .
( [ eps_mu ] ) and ( [ k - hat ] ) : @xmath43 \nonumber \\ & = & % % % 2nd line \begin{pmatrix } 0 \\ \mu_x^{-1 } \hat{k}_z^2 e_y + \mu_z^{-1 } \hat{k}_x^2 e_y \\ 0 \end{pmatrix}. \nonumber \end{aligned}\ ] ] we thus obtain the equation of dispersion under @xmath2 polarization , @xmath44 where @xmath45 in fig . [ fig1 ] .
combining eqs .
( [ kz_s ] ) and ( [ disper_s ] ) , we obtain the next equation : @xmath46 when one knows the value of complex reflectivity @xmath24 by optical measurement or computation , eq .
( [ master_eq ] ) is equation for @xmath5 and @xmath6 .
two different angles ( @xmath47 ) yield @xmath48 , and enable to evaluate @xmath5 and @xmath6 uniquely . by permutating the configuration @xmath49
, we can figure @xmath50 and @xmath51 .
then , incident light sheds on the @xmath52 plane and has the polarization of @xmath53 .
the plane of incidence is the @xmath1 plane . in this second step ,
( [ master_eq ] ) is modified as @xmath54 where @xmath55 is complex reflectivity on the @xmath56 plane and @xmath57 is incident angle in the @xmath1 plane . in evaluating @xmath50 and @xmath51 similarly to @xmath5 and @xmath6
, it is required that the object should be semi - infinitely thick along the @xmath58 axis .
since the relations of @xmath13 are assumed for smdm , @xmath59 is evaluated in the second step ; succeedingly , the rest components of @xmath60 , @xmath61 , and @xmath62 are uniquely obtained except for the sign .
the sign of @xmath63 is set to be @xmath64 at 800 nm because the @xmath63 spectrum keeps constant and does not show any resonant behavior around the wavelength .
the sign of @xmath63 at different wavelength is taken to avoid discontinuous jump as shown in fig .
[ fig2 ] .
complex reflectivity @xmath24 was computed exactly , whereas @xmath55 was evaluated quite precisely , that is , within a - few - percent numerical error . the numerical way to compute @xmath55 was reported earlier @xcite .
the precision in the computation for complex reflectivities ensures that the full components of tensors are determined within a - few - percent numerical error . and 15@xmath65 .
the surface thickness @xmath66 of smdm is 20 nm .
( a ) : effective permittivity .
upper solid line denotes re(@xmath60 ) and upper dotted line does im(@xmath60 ) .
lower two lines represent @xmath61 likewise .
( b ) : effective permeability .
lines indicate @xmath63 and @xmath62 with the notation similar to ( a ) .
( c ) : effective refractive index along the @xmath58 and @xmath11 axes .
the real parts are shown with solid lines and the imaginary parts with dotted lines .
[ fig2],width=283 ]
figure [ fig2 ] presents the full components of ( a ) @xmath8 and ( b ) @xmath9 tensors of smdm of surface thickness @xmath67 nm . the tensors are extracted by tcrm with the two angles of 0@xmath65 and 15@xmath65 .
other sets of two incident angles result in the same spectra of @xmath60 , @xmath61 , and @xmath63 within the numerical error .
this consistency indicates that @xmath60 , @xmath61 , and @xmath63 are independent of incident angle and consequently wavevector * k*. in tcrm analysis for smdm using the angle pairs of ( @xmath68,@xmath4 ) and ( @xmath68,@xmath57 ) , it is easily verified from eq .
( [ master_eq ] ) that @xmath62 depends only on the @xmath69 .
the three components of @xmath60 , @xmath61 , and @xmath63 could in principal depend on the incident angles ; however , they have been found to be independent of incident angles and therefore to be well - defined for smdm . on the other hand , the component @xmath70 is exceptionally evaluated for each non - zero @xmath4 .
thus , all the components of @xmath8 and @xmath9 tensors are determined self - consistently and in a well - defined manner at visible frequencies in fig .
[ fig2 ] .
the @xmath60 component has typical dispersion of drude metal ( upper solid and dotted lines in fig .
[ fig2](a ) : the real and imaginary parts of @xmath60 , respectively ) , and the effective plasma frequency @xmath71 , which is defined by re@xmath72 = 0 $ ] , is located at @xmath73 nm . on the other hand , the @xmath61 spectrum exhibits typical dielectric properties for the incident light of @xmath74 on the @xmath18 or @xmath52 planes ( lower solid and dotted lines in fig .
[ fig2](a ) : the real and imaginary parts of @xmath61 , respectively ) . concerning permeability components , @xmath63 is close to 1 at 360800 nm ( upper solid and dotted lines in fig .
[ fig2](b ) : the real and imaginary parts of @xmath63 , respectively ) , while @xmath62 goes down to 0.57 at 477 nm and shows strong diamagnetism ( lower solid and dotted lines in fig .
[ fig2](b ) : the real and imaginary parts of @xmath62 , respectively ) .
refractive indexes along the @xmath58 and @xmath11 axes are displayed in fig .
[ fig2](c ) ; the upper solid and dotted lines respectively denote re(@xmath75 ) and im(@xmath75 ) , and the lower solid and dotted lines respectively stand for re(@xmath76 ) and im(@xmath76 )
. the imaginary parts of @xmath75 and @xmath76 are small below @xmath77 , and correspond to efficient em - wave propagation in smdm . under normal incidence on the @xmath1 plane with @xmath2 polarization .
solid line indicates forward propagation only , that is , incident and refractive components .
dashed line represents the electric field calculated by the effective refractive index @xmath76 in fig .
[ fig2](c ) .
vertical dotted line indicates the interface .
[ fig3],width=283 ] in fig .
[ fig3 ] , we test the validity of description by effective refractive index at @xmath77 .
the interface of vacuum and smdm is set at @xmath78 and incident light illuminates on the @xmath1 plane normally .
solid line denotes electric field of incident light in vacuum ( @xmath79 ) ; it represents refractive component of electric field in smdm ( @xmath80 ) .
dashed line presents the effective field profile calculated by using @xmath76 in fig .
[ fig2](c ) and reproduces the averaged field profile in smdm approximately .
thus , tcrm extracts averaged em response in smdm .
as was already confirmed in @xcite , the effective description has relevant physical meanings and works fairly well in the visible range of interest . as shown in fig .
[ fig2 ] , effective diamagnetic response appears in the @xmath62 spectrum .
this implies that magnetic induction @xmath81 and magnetic field @xmath82 behave in a different way .
figure [ fig4 ] displays the actual distribution of ( a ) @xmath81 and ( b ) @xmath82 as image plot .
arrows denote the vectors at dotted points .
the wavelength is at 477 nm , where re(@xmath62 ) is minimun . in vacuum
( @xmath83 ) , the incident light travels with incident angle 15@xmath65 ; only incident component is shown for simplicity in fig .
[ fig4 ] . in smdm ( @xmath84 ) , both forward and backward components are displayed simultaneously because reflectance @xmath85 is not small ( @xmath86 ) and the backward components connected to inner multi - reflection is not negligible .
the computation for @xmath81 at each point is carried out from eq .
( [ kxe ] ) , and on the other hand magnetic field @xmath82 is calculated from eq .
( [ kxh ] ) .
figure [ fig4 ] obviously shows that the directions of @xmath81 and @xmath82 are opposite in metallic layer ( @xmath87 nm ) .
dielectric layers behave like paramagnetic materials . besides , the @xmath82 component in the metallic layer is far more intense than the @xmath82 in vacuum and dielectric layers . in this way
, the thin metallic layers are found to be diamagnetic along the @xmath11 axis .
the diamagnetic nature does not come from electronic spin states of silver but from the superimposed em fields peculiar to mesoscopic structure of smdm .
the backward component is indeed responsible for the local diamagnetism .
thus , the effective diamagnetic @xmath62 response is found to originate from the thin metallic layers . and ( b ) @xmath82 at 477 nm , the minimum of re(@xmath62 ) .
the region of vacuum ( @xmath83 ) shows incident light only , for simplicity ; the incident angle is @xmath88 .
the region of smdm ( @xmath84 ) contains both forwardly and backwardly propagating components .
arrows indicate the vectors at typical points ( dots ) . in particular , the directions of @xmath81 and @xmath82 in metallic layer ( @xmath87 nm ) are clearly opposite .
[ fig4],width=302 ] the diamagnetic response of metallic layer in smdm is confirmed numerically in fig .
it is also easily derived from maxwell equation . in fig .
[ fig4 ] , eqs .
( [ kxe ] ) and ( [ kxh ] ) in a metallic layer are explicitly written as @xmath89 where sgn(@xmath90 ) takes @xmath91 or @xmath92 , and denotes the forward and backward propagation for the @xmath11 axis . from the computation for actual em distribution , @xmath93
is found handreds times larger than @xmath94 ; _ i.e. _ @xmath95 .
consequently , the first term in the left hand side of eq .
( [ kxh_comp ] ) is negligible to a good approximation .
then , deviding eq .
( [ kxe_comp ] ) by eq .
( [ kxh_comp ] ) , the sign of the real part of @xmath96 turns out to be determined by a factor @xmath97 : @xmath98 at the wavelength in fig .
[ fig4 ] , @xmath99 directly results in @xmath100 in the metallic layer .
thus , the diamagnetic nature of thin metallic layers is confirmed from the simple calculation .
all the spectra are obtained by tcrm with the two angles of 0@xmath65 and @xmath101 . however ,
as described in the text , @xmath62 depends only on the @xmath4 in fig .
the spectrum at 15@xmath65 is the same with that in fig .
[ fig2](b ) .
the spectra at @xmath102 , 45@xmath65 , and 60@xmath65 are offset for visibility by 0.5 , 1.0 , and 1.5 , respectively .
horizontal dotted lines indicate re(@xmath62 ) @xmath103 1.0 for each spectrum .
[ fig5],width=283 ] figure [ fig5 ] shows the dependence of @xmath62 on incident angle .
the @xmath62 spectra in fig .
[ fig5 ] vary for incident angles and indicates spatial dispersion .
note that @xmath62 at off - resonance above 540 nm is almost independent of @xmath4 .
the spatial dispersion is a resonant behavior .
figure [ fig6 ] displays the effect of structure of unitcell on the effective @xmath62 .
figures [ fig2][fig5 ] are the results regarding the surface thickness @xmath104 nm ; the spectra is shown again in fig .
[ fig6 ] . from bottom to top , the thickness @xmath66 varies from 0 to 30 nm , respectively . by changing the surface thickness , the structure of unitcell in smdm varys ; see fig . [ fig1 ] . smdm with @xmath105 and 10 nm show diamagnetic property in a wide range .
the increase of surface thickness @xmath66 gradually reduces diamagnetism and finally eliminates the diamagnetism at @xmath106 nm . since the structure of unitcell directly affects the phase of em wave in smdm , it is crucially significant how to set the unitcell in realizing the effective diamagnetism .
smdms of different unitcells are different metamaterials .
figure [ fig6 ] is the explicit indication .
it is also consistent with the conclusion in a recent numerical study @xcite , showing that the interfaces play a vital role in photonic slabs of finite thickness , in extracting effective constants by the retrieval way .
spectra of smdm with various unitcells : the thickness @xmath66 of surface layers from 0 to 30 nm .
all the spectra are computed at the incident angle of 15@xmath65 , and are presented without vertical offset .
the spectrum of @xmath67 nm in fig .
[ fig5 ] is displayed again for comparison .
[ fig6],width=283 ] as is shown in figs .
[ fig2 ] and [ fig6 ] , the effective diamagnetic response emerges at about @xmath77 , comes from the stratified metallic layers in smdm .
in order to observe the diamagnetic resonance in smdm , it is essential how to select the structure of unitcell . as the @xmath77 shifts with varying the ratio of metal in the unitcell
, the diamagnetic resonance follows the shift of @xmath77 .
it strongly suggests that the diamagnetic resonance is magnetic part of the effective plasma resonance . the spatially dispersive behavior in fig .
[ fig5 ] also shows a typical characteristic of resonance .
from computational and theoretical studies , it was reported that effective magnetism takes place in the systems composed of dielectric @xcite and metallic rods @xcite , results from the mie resonances peculiar to the rods , and is spatially dispersive @xcite .
these magnetic responses stem from the local resonance due to geometrical shape of constituents .
the effective diamagnetism in smdm also inherited from the the local diamagnetic nature of thin metals , which was shown in sec .
[ origin ] .
there is a remarkable characteristic to the diamagnetism of smdm , which is absent in the systems of the local mie resonance . in the mie systems
it is known that the resonances always induce not a small amount of absorption loss , whereas the diamagnetic resonance is almost free from loss as seen in fig .
[ fig2](c ) .
although small loss [ _ i.e. _ small im(@xmath76 ) ] exists , it appears irrespective of the diamagnetism ; indeed , the small loss also exists in the non - magnetic smdm of surface thickness @xmath106 nm in fig .
[ fig6 ] @xcite .
this kind of magnetic resonance with low loss has not been found so far to our knowledge and is a new class of magnetism in metamaterials .
consequently , it is expected that low - loss , effective magnetic resonance at visible frequencies will be realized based on this magnetism .
thick enough smdm have been focused on in this paper .
of course , this system is not easily fabricated in reality .
however , it is a good touchstone to examine how the effective optical responses emerge in periodic metal - dielectric systems . in the present analysis
, there is no evidence to suggest negative refraction in smdm , which was lately argued in finite systems @xcite .
if negative refraction takes place in finitely thick smdm , it would be due to an interference effect in finite systems .
the full components of effective tensors of @xmath8 and @xmath9 in smdm has been numerically extracted by tcrm in a well - defined way . as a main result , the effective strong diamagnetic resonance along the optical axis has been clarified .
the resonance is spatially dispersive .
it has been revealed that stratified thin metals in smdm are the source of effective diamagnetism .
the diamagnetism of thin metals results not from local mie resonances but from the em states realized in smdm .
the diamagnetic resonance is induced always near the effective plasma frequency ; it can therefore be regarded as a magnetic component of the collective excitation .
the diamagnetic nature of thin metals is experimentally feasible , and it will be a good test to measure nonlinear optical responses via diamagnetic resonance .
the author thanks t. ishihara and s. g. tikhodeev for stimulating discussion .
this study was partially supported by research foundation for opto - science and technology and by the information synergy center , tohoku university in numerical implementation . ,
s. schultz , p. marko , and c. m. soukoulis , `` determination of effective permittivity and permeability of metamaterials from reflection and transmission coefficients , '' phys .
b * 65 * , 195401 ( 2002 ) . , g. daguanno , n. mattiucci , m. j. bloemer , d. de ceglia , m. centini , c. sibilia , n. akozbek , m. g. cappeddu , m. fowler , and j. w. haus , `` negative refraction and sub - wavelength focusing in the visible range using transparent metallodielectric stacks , '' opt .
express * 15 * , 508523 ( 2007 ) .
, h. jiang , b. gralak , s. enoch , g. tayeb , and m. lequime , `` toward @xmath92 effective index with one - dimensional metal - dielectric metamaterial : a quantitative analysis of the role of absorption losses , '' opt .
express * 15 * , 77207729 ( 2007 ) .
, g. tayeb , s. enoch , d. maystre , and b. gralak , `` photonic crystal lens : from negative refraction and negative index to negative permittivity and permeability , '' phys . rev .
lett . * 97 * , 073905 ( 2006 ) . | it is numerically demonstrated that effectively strong diamagnetic resonance emerges at visible frequencies in stratified metal - dielectric metamaterials .
the effective optical constants are extracted by two - complex reflectivity method .
it is clarified that the effective diamagnetic response originates from local diamagnetism at stratified thin metals .
the effective diamagnetism is crucially sensitive to the sturucture of unitcell .
the effective diamagnetic response is always associated with effective plasma frequency and is therefore regarded as a magnetic component of the collective excitation . |
in the early seventies , toomre ( 1970 ) and toomre & toomre ( 1972 ) , presented numerical models which demonstrated that strong tidal forces between interacting galaxies could result in features like plumes , shells , rings , tidal tails , and bridges .
they also proposed that such strong collisions between galaxies would lead to orbital decay and eventual merging . in 1977 , toomre identified a series of galaxies that he believed represented galaxies at different stages of merging ( `` the toomre sequence '' ) and proposed that the end product of such merging could be an elliptical galaxy .
this hypothesis came to be known as the `` merger hypothesis . ''
hibbard & van gorkom ( 1996 ) presented h-@xmath1 , r - band , and 21 cm observations of several galaxy systems from the toomre sequence and showed that these galaxies most likely do form an evolutionary chain .
three stages in the merging sequence were identified : ( i ) the early stage , in which the disks are well separated and only marginally disrupted , ( ii ) the intermediate stage , which exhibits distinct nuclei embedded in luminous material , and ( iii ) the late stage , in which tidal appendages are seen to emerge from a single nucleus . the purpose of this study is to better understand and interpret the evolutionary sequence as described by hibbard & van gorkom ( 1996 ) .
this project was initially part of a multiwavelength project , in collaboration with h. deeg and c. muoz - tun of the instituto de astrofisica de canarias in spain ( deeg et al .
we have picked seven galaxies : arp 213 , arp 78 , arp 135 , arp 31 , arp 279 , arp 263 and arp 158 , which we believe are in various stages of the evolutionary sequence . in this first paper
we report on neutral hydrogen ( h@xmath0i ) observations of arp 158 .
arp 158 ( ngc 523 ) is located at @xmath1 = 01@xmath2 25@xmath3 20.8@xmath4 and @xmath5 = + 3401@xmath6 29@xmath7 ( j2000 ) .
it is described in zwicky s catalogue of selected compact galaxies and post - eruptive galaxies ( 1971 ) as `` post - eruptive , blue compact with three compact knots connected by a bright bar , fan - shaped jets and matrix . '' the characteristics of the galaxy are given in table 1 .
arp s image ( figure 1 ) shows that there are three visible knots , with a central linear structure .
( without appropriate kinematic information , it is impossible to determine whether this is a bar .
) there is also a bright tail extending to the west and an extensive , faint diffuse tail to the south - east .
only the western and eastern knots are easily visible in this image ; the third knot is located approximately in the center of the linear structure .
the westernmost `` nucleus '' was examined by chincarini & heckathorn ( 1973 , hereafter referred to as c&h ) and was claimed to be a foreground star . from a close visual exam of arp s image , this knot appears to be an unresolved point source , and resembles other stars in the image .
dahari ( 1985 ) , however , argued that his measured redshift for this knot was similar to that of the south - east one and thus could not be a foreground star . without confirming spectra , we can not determine the nature of this knot , or establish its connection with arp 158 .
for the purposes of discussion , we will refer to it as the western knot / star .
high sensitivity observations of the 21 cm emission of arp 158 were made during may 14 and may 17 , 1999 by m. iyer , c. simpson and b. malphrus with the very large array ( vla ) in its spectral line mode .
the galaxy was observed with the d - array configuration for a total of five hours , including move time and calibration time , using a 128 channel spectrometer with a total bandwidth of 6.25 mhz . on - line hanning
smoothing was not used , so the velocity resolution is 1.2 times the channel separation ( wilcots et .
al , 1997 ) . for our observations , with a channel separation of 48.8 khz ( 10.3 km s@xmath8 ) , this results in a velocity resolution of 12.3 km s@xmath8 .
the observations were made at a central velocity of 4758 km s@xmath8 and the data were collected for a single polarization .
the observational parameters are given in table 2 .
standard calibration and editing procedures were performed using the aips data reduction package available from the nrao .
the continuum sources 0119 + 321 ( j2000 ) and 0137 + 331 ( j2000 ) were used to calibrate the phase and amplitude response of the receiving system .
the data from each day were edited and calibrated independently before being combined in the _
uv_-plane to produce the combined data set .
the data reduction for the may 17 data set was complicated owing to solar interference .
the d - array configuration is very susceptible to solar contamination , particularly at short baselines . to rectify this problem ,
all the data points from baselines less than 1k@xmath9 ( 210 m ) for the may 17 data were not used in the imaging process .
this was approximately 13% of the total data for both the days combined and reduces by @xmath10 7% the sensitivity of our maps , especially to large scale structures .
narrow band interference was also present in channels 99 and 100 but as these channels were not included in the calibration and did not include any line emission , they did not interfere with our reduction or analysis .
after calibration , the aips task uvlin ( cornwell et .
al , 1992 ) was used to subtract the continuum emission .
the task makes linear fits to the visibility data in a range of channels without h@xmath0i line emission and then subtracts the requisite values from all the channels . in our dataset ,
channels 6 to 33 ( 5376 km s@xmath8 and 5088 km s@xmath8 ) and channels 102 to 119 ( 4354 km s@xmath8 and 4174 km s@xmath8 ) were line - free and were used to perform the continuum subtraction .
a fast - fourier transform was then applied to this _ uv _ data to produce a cube ( @xmath1 , @xmath5 , v ) of images , one at each observed frequency ( velocity channel ) . during imaging , each channel
was cleaned ( hgbom 1974 ; clark 1980 ) to reduce the effect of the sidelobes produced by non - gaussian features of the synthesized beam ( the `` dirty '' beam ) .
we tested various weighting schemes ( robust parameter in the imagr task in aips ) during the imaging process to increase the resolution , but because of the lack of baselines less than 1k@xmath9 , the increase in resolution was not worth the corresponding decrease in sensitivity .
the higher resolution images did not reveal any new morphological or kinematic features , and with the lowered sensitivity , low column density features were lost completely .
hence , in order to maximize the sensitivity to low surface brightness features , we used the natural weighting scheme , resulting in a beamsize ( fwhm ) of 47.3 @xmath7 @xmath11 42.2 @xmath7 .
the data cube was cleaned down to 0.98 mjy per beam , corresponding to 1@xmath12 as determined by the statistical analysis of a signal - free channel map .
line emission signal was present in channels from 41 to 89 , approximately 4500 km s@xmath8 to 5000 km s@xmath8 , which were then integrated to produce moment maps representing the h@xmath0i integrated column density ( 0th moment ) , the temperature - weighted mean velocity ( 1st moment ) , and the velocity dispersion ( 2nd moment ) .
throughout this paper we use a value of h@xmath13 equal to 75 km s@xmath8 mpc@xmath8 and a heliocentric velocity of 4758 km s@xmath8 ( nasa ipac extragalactic database , ned ) for arp 158 .
these values imply a distance of 63 mpc .
the results from the derived quantities are shown in table 3 , and are discussed below .
figure 2 is a grayscale image of the h@xmath0i integrated flux density .
we can see that there are three distinct h@xmath0i concentrations which we denote as a , b , and c. figure 3 represents the optical image from the second generation blue digitized sky survey ( dss2 ) overlaid with h@xmath0i contours .
the central optical knot is more easily visible in this image . for the entire distribution
, we have detected a total h@xmath0i mass of 6.5 @xmath11 10@xmath14 m@xmath15 .
system a , which has the highest beam - smoothed h@xmath0i column density ( n@xmath16 = 13.0 @xmath11 10@xmath17 atoms @xmath18 ) coincides with the optical knot system and seems to be centered to the west of the western knot / star . system b ( n@xmath16 = 7.4 @xmath11 10@xmath17 atoms @xmath18 ) appears to coincide with the diffuse tail to the south - east . there is also a very interesting gas knot to the north - west of the galaxy .
this is system c ( n@xmath16 = 1.2 @xmath1110@xmath17 atoms @xmath18 ) .
note that the field of view for arp s image ( figure 1 ) does not include this area .
although this system is faint in h@xmath0i it does appear in four contiguous channels and we believe it to be a valid detection . using separate integration maps , we have calculated h@xmath0i masses associated with each of our defined systems ( see section on kinematics ) , as well as their flux - weighted systemic velocities ( table 3 ) , and an estimate of their velocity widths ( fwhm ) . system a has the highest h@xmath0i mass ( 2.9@xmath11 10@xmath14 m@xmath15 ) and the h@xmath0i masses associated with systems b and c are approximately 1.9 @xmath11 10@xmath14 m@xmath15 and 0.1 @xmath11 10@xmath14 m@xmath15 , respectively .
using the optical angular size ( 2.5@xmath6 @xmath11 0.7@xmath6 ) catalogued in ned , the optical dimensions for arp 158 are 46 @xmath11 13 kpc .
the total h@xmath0i distribution is spread well beyond the optical image .
we measure an h@xmath0i extent of 70 @xmath11 33 kpc ( not deconvolved ) , measured out to 2 @xmath11 10@xmath19 atoms @xmath18 .
looking at figure 3 , there is faint optical emission located north and very slightly east of the h@xmath0i .
the northernmost part of the h@xmath0i contour might overlay the southernmost part of the optical emission , but the h@xmath0i is not obviously coincident with this weak optical emission .
a deeper optical image is needed to establish its nature and a spectrum would be required to establish any physical connection to arp 158 .
figure 4 shows the intensity - weighted velocity map .
the isovelocity contours are very different for systems a and b , and there is an area of steep gradient between the two which overlies the bright optical emission in the system .
there also appear to be three distinct kinematic areas coinciding with systems a , b , and c. the apparent kinematical axes of systems a and b , as indicated by the isovelocity contours , are oriented in different directions .
the position angle of the kinematic axis for system a , measured from north through east , is approximately 25 on the western side , rotating to a position angle of approximately 90 on the eastern side .
this almost 70 kink in the isovels in system a shows that this is a fairly abrupt change . for system
b , the position angle is approximately 150 . the channel maps ( figure 5 ) for arp 158 reveal that there is emission in every channel between systems a and b , showing that there is gas extending between them which is superposed along the line - of - sight in the moment maps
. the extended emission could be either a bridge or a tail .
it could be that systems a and b were two different galaxies that are now interacting . in this case
, we could consider the extended emission to be a bridge .
we do not believe this to be the case .
optically , systems a and b do not have separate nuclei .
the distinct optical knots only seem to coincide with system a , whereas system b does not have much of a stellar component associated with it .
it seems more likely that b is a tidal tail that formed due to the interaction that deformed the optical components of the system .
the center point in each of the channel maps is moving smoothly to the west going to higher velocities .
the contours for system b show a smooth run in velocity , with a few widely separated isovels indicating a shallow observed gradient in velocity .
it also exhibits a sharp cut - off in velocity space at the westernmost end .
thus system b , which optically resembles a tail , also stretches out in the h@xmath0i and has a narrow spread in velocity .
all these properties are consistent with this feature being a tidal tail ( hibbard & van gorkom , 1996 ) , perhaps seen edge - on , leading to the superposition of gas we see in the moment maps . however
, the shallow velocity gradient we observe could also be due to gas moving mostly transverse to our line of sight .
because of the superposition of different kinematic components of this complex object , we have used the channel maps in conjunction with position - velocity diagrams ( figure 6 ) to define ( somewhat subjective ) velocity ranges for systems a , b , and c , as well as for the gas `` tail '' connecting a and b. to produce these position - velocity diagrams , we used the aips task xsum to ( a ) sum along the declination at the central right ascension and ( b ) sum along the right ascension at the central declination . using our velocity ranges
, we then integrated the appropriate channel maps for each of these components separately to produce moment maps for each object and the gas tail .
the velocity field for system a is shown in figure 7a , which was created by integrating only the channels from 4503.0 km s@xmath8 to 4694.2 km s@xmath8 in the cleaned cube .
system c , which overlaps with a in velocity space , appears in the channel maps from 4566.7 - 4598.6 km s@xmath8 .
figure 7b , showing only system b , was created by integrating the cube from 4907.0 km s@xmath8 to 5002.8 km s@xmath8 .
figure 7c , showing the connecting gas tail , is from the integration of 4704.8 km s@xmath8 to 4885.7 km s@xmath8 .
for all the three parts of figure 7 , the isovels are 10 km s@xmath8 apart .
the kinematic axes of systems a and b are more clearly visible in these figures , and are oriented in different directions .
the isovels in the connecting tail appear to broaden moving from system a southeastward toward system b. using our h@xmath0i flux - weighted systemic velocites , we find that systems a and b are separated by almost 343 km s@xmath8 , whereas systems a and c are separated by about 33 km s@xmath8 .
the velocity widths ( figure 8) for systems a , b and c are approximately 256 km s@xmath8 , 76 km s@xmath8 and 30 km s@xmath8 .
the velocity widths as well as the h@xmath0i masses were measured using only the channels defined for each object . except for system c , which is clearly separate
, please note that these quantities are dependent upon our velocity ranges for each object .
there is an unresolved continuum source , nvss j012521 + 340128 ( condon et al . 1998 ) , at the position of arp 158 ( figure 9 ) , with a catalogued flux density of 15.4 mjy .
it is located in the area between systems a and b , roughly coincident with the start of the connecting gas tail .
this could either be a background source , unconnected to the arp 158 system , or a source within the system .
as it is not uncommon for continuum emission to be associated in some way with interacting systems , it is highly likely that the continuum source is connected with arp 158 .
the idea that radio continuum emission was related to galaxies undergoing an interaction was initially conceived due to the connection of three radio sources with optically strange galaxies ( baade & minkowski , 1954 ) . in either case , there is a possibility of absorption of the continuum emission by the h@xmath0i in arp 158 .
as only the line - free channels were used to estimate the continuum intensity for subtraction from our data cube , we have confidence that our continuum subtraction was not affected by any absorption that may be present .
there could however , be an issue with the measured h@xmath0i intensity at the position of the source .
if the level is low between systems a and b , it could be that there really is less h@xmath0i or that there in a reduction in the h@xmath0i intensity due to absorption .
there is nothing obvious in the spectrum of the peak pixel of the continuum source to indicate the presence of absorption , and the integrated flux of the continuum source made from channels with and without line emission were comparable .
we also attempted to check for absorption in the h@xmath0i line at the location of the continuum source by comparing an `` off - source '' spectrum through the cube to an `` on - source '' spectrum .
the on - source spectrum is run at the position of the continuum source , while the off - source spectrum is obtained from an area of emission away from the continuum source .
the two spectra are then subtracted , and if absorption is present , an absorption line profile may then be revealed ( england & gottesman 1990 ) .
however , the success of this method depends on the two spectra being from kinematically similar areas , since the off - source spectrum is supposed to represent what the h@xmath0i line `` should '' look like ( for no absorption ) .
unfortunately , even the closest off - source spectra we looked at are located in areas with h@xmath0i kinematics that are very different than the kinematics in the on - source area .
this prevents us from directly determining the presence or absence of absorption at the location of the continuum source .
however , since the measured intensity of the continuum source was the same in maps made with and without h@xmath0i emission , any absorption would have to be less than a few times the noise in our channel maps . this would indicate that the h@xmath0i is optically thin .
in the optical image for arp 158 , there are bright components which include the knots and the linear structure , as well as a diffuse component ( the south - eastern tail ) .
there is still debate on the true identity of the westernmost nucleus , i.e. whether or not it is a foreground star . through this discussion , however , we assume the westernmost nucleus to be a foreground star as suggested by c&h . on the basis of the optical and h@xmath0i data , we believe arp 158 to be an intermediate stage merger , consistent with the model proposed by hibbard & van gorkom ( 1996 ) . using arp s image , we estimate that the knots and the linear structure extend to approximately 27 kpc .
the bright tail , defined as the central knot to the beginning of the diffuse tail , is about 12 kpc .
the extensive , faint diffuse tail to the south - east spreads out approximately another 26 kpc . from their electrograph at the cassegrain focus of the 82 inch struve reflector at the mcdonald observatory , c&h s measurements are fairly consistent with ours .
they estimated the optical extent of the bright components to be approximately 21 kpc , with the diffuse components extending to about 4152 kpc .
there are a few small optical condensations in the faint south - east tail ; they may be h@xmath0ii regions ( c&h ) , which would not be unusual in a tidal tail ( hibbard & van gorkom 1996 ) .
c&h presumed that the eastern knot and the central knot were two separate galactic nuclei from two interacting galaxies .
we have estimated their projected separation to be approximately 6 kpc .
bernlhr ( 1993 ) calculated a difference of 100 km s@xmath8 between the optical eastern and central knots , which is in agreement with that of c&h .
the eastern and central knots lie within the east part of the h@xmath0i system a and the west part of the connecting `` tail . '' from their spectra , c&h assumed a similar composition and mass - to - light ratio for the knots , and attributed the optical brightness of the eastern knot to a higher central star density . using this information
, they calculated the mass ratio of the two nuclei to be 5:1 .
this mass ratio implies that the gravitational interaction between the two nuclei is weak . to get substantial disruption in a weak encounter , the objects need to pass very close to each other in a prograde encounter .
since arp 158 is very disrupted , the encounter must have been close . according to a. toomre as reported by c&h , the best proof that such an encounter had occurred was the optical tail in the south - east direction . by asssuming the galaxy to be completely spherical and calculating the ratio of the major and minor axes of the central knot , c&h estimated arp 158 s inclination at 25 .
if this is a reasonable estimate , then the system is seen somewhat edge - on .
although this inclination estimate can not be considered very accurate since the system is disturbed , the flat distribution of the h@xmath0i in our images is further indication that arp 158 is probably an edge - on system .
the superposition of the tail is expected if the system is edge - on and/or the plane of interaction is along the line of sight .
this superposition is clearly seen in both the channel maps and the position - velocity diagrams .
the low spatial resolution of our data precludes any in - depth inclination analysis , so the question of the true nature of its h@xmath0i morphology remains a question . based on our h@xmath0i data for arp 158 , system b , associated with the faint diffuse south - east optical extension , and the h@xmath0i tail together
have h@xmath0i features that resemble a tidal feature .
it is not clear if system c is , or has been , involved in this interaction .
we do not see a stellar component associated with system c nor is it very well resolved .
system c could be a tidal dwarf formed as a result of the interaction , a nearby galaxy , or perhaps a counterarm to the `` tail . ''
the lack of any apparent gas connecting it to the rest of the system would argue against it being a counterarm . on the other hand , its narrow spread in velocity and its location , almost directly opposite system b , are consistent with system c being a counterarm .
toomre also noted the optical similarity between ngc 520 ( arp 157 ) and arp 158 ( c&h ) .
ngc 520 , classified as an intermediate stage merger by hibbard & van gorkom ( 1996 ) , consists of two parent nuclei separated by @xmath10 6 kpc ( from images kindly supplied by j. hibbard ) .
there is also a tail to the south which extends to @xmath10 23 kpc .
there is a plume which emanates from this tail , bends east , and then stretches to the dwarf galaxy ugc 957 at a distance of @xmath10 52 kpc . like
ngc 520 , arp 158 shows distinct nuclei immersed in luminous material , which is consistent with an intermediate stage merger .
in addition , the tails in the two systems extend to almost the same length and the separation between the parent nuclei for ngc 520 and the separation between the eastern and central nuclei for arp 158 are comparable .
furthermore , the h@xmath0i for arp 158 extends beyond the optical image , and this is also true for ngc 520 .
the total hi mass for ngc 520 was measured to be 7.1 @xmath1110@xmath14 m@xmath15 ( hibbard & van gorkom 1996 ) , roughly similar to our measurement of arp 158 , which is 6.5 @xmath1110@xmath14 m@xmath15 .
we believe arp 158 to be a mid - stage merger . optically , it has a very disturbed morphology , with bright knots connected by a linear structure , and a faint diffuse tail .
the h@xmath0i is spread out far beyond the optical image , which is not unusual for interacting systems .
there are three distinct hi concentrations , which are kinematically quite different .
system a coincides with the optical knot system and system b appears to coincide with the diffuse south - east tail .
system b also stretches out in the h@xmath0i and with its narrow spread in velocity , has features coincident with those of a tidal tail .
system c is a previously undetected entity .
there is no obvious optical counterpart to this h@xmath0i system , other than weak optical emission located just north of it .
it would require a deeper optical image and spectra to verify any association of the optical emission with system c. there is a continuum source in the vicinity of the galaxy .
the westernmost optical knot also needs to be studied further to determine whether it is associated with arp 158 or if it is a foreground star .
there are distinctive similarities between arp 158 and ngc 520 , identified as a mid - stage merger , which lends credence to our belief that arp 158 is also an intermediate stage merger .
this research has made use of the nasa / ipac extragalactic database ( ned ) which is operated by the jet propulsion laboratory , calafornia institute of technology , under contract with the national aeronautics and space administration .
this work has also made use of nasa s _ skyview _ facility located at nasa goddard space flight center to obtain optical images from the second digitized sky survey .
the second palomar observatory sky survey ( poss - ii ) was made by the california institute of technology with funds from the national science foundation , the national aeronautics and space administration , the national geographic society , the sloan foundation , the samuel oschin foundation , and the eastman kodak corporation .
the compressed files of the `` palomar observatory - space telescope science institute digital sky survey '' of the northern sky , based on scans of the second palomar sky survey are copyright ( c ) 1993 - 2000 by the california institute of technology and are distributed herein by agreement .
all rights reserved .
produced under contract no . nas5 - 26555 with the national aeronautics and space administration .
& 523 + ugc no . &
00979 + vv no . &
783 + right ascension ( j2000 ) & 01@xmath2 22@xmath3 20.80@xmath4 + declination ( j2000 ) & @xmath20 01@xmath6 29.0@xmath7 + morphological type & pec + distance ( mpc ) & 63 + photographic magnitude & 13.5 + optical dimensions ( arcmin ) & 2.5 @xmath11 @xmath21 + date & 14 and 17 may 1999 + no . of visibilities & 65344 + bandwidth ( mhz ) & 6.25 + no . of channels & 127 + channel separation ( khz ) & 48.8 + heliocentric velocity ( km s@xmath8 ) & 4758 + beam ( fwhm ) ( arcsec ) & 47.3 @xmath11 42.2 + rms noise in channel maps ( mjy / beam ) & 0.98 + equivalent brightness temperature ( k ) & 0.29 + arp 158 & 9.7 & & 6.5 @xmath11 10@xmath14 & & + system a & 5.9 & 13.0 @xmath11 10@xmath17 & 2.9@xmath11 10@xmath14 & 4614 & 256.4 + system b & 2.4 & 7.4 @xmath11 10@xmath17 & 1.9 @xmath11 10@xmath14 & 4957 & 76.9 + system c & 0.9 & 1.2 @xmath11 10@xmath17 & 0.1 @xmath11 10@xmath14 & 4581 & 30.7 + tail & 2.6 & 8.1 @xmath11 10@xmath17 & 1.9 @xmath11 10@xmath14 & & + | we present 21 cm observations of arp 158 .
we have performed a study of the neutral hydrogen ( h@xmath0i ) to help us understand the overall formation and evolution of this system .
this is a disturbed system with distinct optical knots connected by a linear structure embedded in luminous material .
there is also a diffuse spray to the southeast .
the h@xmath0i seems to be made up of three distinct , kinematically separate systems .
arp 158 bears a certain optical resemblance to ngc 520 ( arp 157 ) , which has been identified as a mid - stage merger . from our 21 cm observations of arp 158
, we also see a comparable h@xmath0i content with ngc 520 .
these similarities suggest that arp 158 is also an intermediate stage merger . |
cataclysmic variables ( cvs ) are close binary systems transferring matter from a low - mass dwarf secondary to a white dwarf .
the transferred matter forms an accretion disk . thermal instability od the disk caused by partial ionization of hydrogen results outbursts in dwarf novae ( dne ) , a subclass of cvs .
tidal instability of the disk caused by the 3:1 resonance with the orbiting secondary is considered to develop an eccentric ( or flexing ) disk in su uma - type dwarf novae , a subclass of dne .
this eccentric disk is responsible for superhumps , which have periods a few percent longer than the orbital period [ see e.g. @xcite ; @xcite ; @xcite ] .
the enhanced mass accretion by tidal instability causes long - lasting superoutbursts [ thermal tidal instability ( tti ) model : @xcite ; @xcite ] .
although there had been intensive discussions whether superoutbursts are a result of tidal instability or an enhanced mass - transfer from the secondary [ e.g. @xcite , @xcite , @xcite ] , recent detailed analyses of high - precision kepler observations have favored the tti model as the only viable model for ordinary su uma - type dwarf novae ( @xcite ; @xcite ; @xcite ) .
[ for general information of cvs , dne , su uma - type dwarf novae and superhumps , see e.g. @xcite ] .
this paper is one of series of papers @xcite , @xcite , @xcite , @xcite , @xcite and @xcite .
these papers originally intended to clarify the period variations of superhumps in su uma - type dwarf novae and first succeeded in identifying superhump stages ( stages a , b and c ) in @xcite . among them , stage a superhumps have recently been identified to be reflect the precession of the eccentric disk at the radius of the 3:1 resonance , and have been one of the most promising tools in determining the mass - ratios in su uma - type dwarf novae and in following the terminal evolution of cvs @xcite .
continuing this project , we report observations of superhumps and associated phenomena in su uma - type dwarf novae whose superoutbursts were observed in 20142015 . in this paper
, we report basic observational materials and discussions in relation to individual objects .
general discussion related to wz sge - type dwarf novae , which are a subclass of su uma - type dwarf novae with infrequent , large amplitude superoutbursts , will be given as in a planned summary paper by @xcite .
starting from @xcite , we have been intending these series of papers to be also a source of compiled information , including historical , of individual dwarf novae since there have been no compiled publication since @xcite .
the material and methods of analysis are given in section [ sec : obs ] , observations and analysis of individual objects are given in section [ sec : individual ] , including some discussions particular to the objects , the general discussion is given in section [ sec : discuss ] and the summary is given in section [ sec : summary ] .
the data were obtained under campaigns led by the vsnet collaboration @xcite . for some objects
, we used the public data from the aavso international databasehttp://www.aavso.org / data - download@xmath2 . ] .
the majority of the data were acquired by time - resolved ccd photometry by using 30cm - class telescopes located world - wide , whose observational details will be presented in future papers dealing with analysis and discussion on individual objects of interest .
the list of outbursts and observers is summarized in table [ tab : outobs ] .
the data analysis was performed just in the same way described in @xcite and @xcite and we mainly used r softwarehttp://cran.r - project.org/@xmath2 . ] for data analysis . in de - trending the data , we used both lower ( 15th order ) polynomial fitting and locally - weighted polynomial regression ( lowess : @xcite ) .
the times of superhumps maxima were determined by the template fitting method as described in @xcite .
the times of all observations are expressed in barycentric julian days ( bjd ) .
the abbreviations used in this paper are the same as in @xcite : @xmath3 means the orbital period and @xmath4 for the fractional superhump excess . following @xcite , the alternative fractional superhump excess in the frequency unit @xmath5 has been introduced because this fractional superhump excess can be directly compared to the precession rate .
we therefore used @xmath6 in referring the precession rate .
we used phase dispersion minimization ( pdm ; @xcite ) for period analysis and 1@xmath7 errors for the pdm analysis was estimated by the methods of @xcite and @xcite .
we present evidence for an su uma - type dwarf nova by presenting period analysis and averaged superhump profile if the paper provides the first solid presentation of individual objects as such .
the resultant @xmath8 , @xmath9 and other parameters are listed in table [ tab : perlist ] in same format as in @xcite .
the definitions of parameters @xmath10 and @xmath9 are the same as in @xcite . and @xmath11 ) for the stages b and c given in the table sometimes overlap because of occasional observational ambiguity in determining the stages . ]
comparisons of @xmath0 diagrams between different superoutbursts are also presented whenever available , since this comparison was one of the main motivations in of these series papers ( cf .
@xcite ) . combined @xmath0 diagrams also help identifying superhump stages particularly when observations are insufficient . in drawing combined @xmath0 diagrams
, we usually used @xmath12 for the start of the superoutburst , which usually refers to the first positive detection of the outburst .
this epoch usually has an accuracy of @xmath11 d for well - observed objects , and if the outburst was not sufficiently observed , we mentioned in the figure caption how to estimate @xmath13 in such an outburst .
we also present @xmath0 diagrams and light curves especially for wz sge - type dwarf novae , which are not expected to undergo outbursts in the near future .
in all figures , the binned magnitudes and @xmath0 values are accompanied by 1@xmath7 error bars , which are omitted when the error is smaller than the plot mark .
we used the same terminology of superhumps summarized in @xcite .
we especially call attention to the term `` late superhumps '' .
although this term has been used to refer to various phenomena , we only used the concept of `` traditional '' late superhumps when an @xmath10.5 phase shift is detected [ @xcite ; see also table 1 in @xcite for various types of superhumps ] , since we suspect that many of the past claims of detections of `` late superhumps '' were likely stage c superhumps [ cf .
@xcite ; note that the kepler observation of v585 lyr also demonstrated this persistent stage c superhumps without a phase shift @xcite , and most recently it is confirmed in another kepler cv by @xcite ] .
early superhumps are double - wave humps seen during the early stages of wz sge - type dwarf novae , and have period close to the orbital periods ( @xcite ; @xcite ; @xcite ) . we are going to discuss this phenomenon in the planned paper @xcite .
we used the period of early superhumps as approximate orbital period @xcite .
as in @xcite , we have used coordinate - based optical transient ( ot ) designations for some objects , such as apparent dwarf nova candidates reported in the transient objects confirmation page of the central bureau for astronomical telegramshttp://www.cbat.eps.harvard.edu / unconf / tocp.html@xmath2 . ] and listed the original identifiers in table [ tab : outobs ] . for objects detected in the catalina real - time transient survey ( crts ;
@xcite)http://nesssi.cacr.caltech.edu / catalina/@xmath2 . for the information of the individual catalina cvs ,
see @xmath14http://nesssi.cacr.caltech.edu / catalina / allcv.html@xmath2 . ]
transients , we preferably used the names provided in @xcite . if these names are not yet available , we used the international astronomical union ( iau)-format names provided by the crts team in the public data releasehttp://nesssi.cacr.caltech.edu / datarelease/@xmath2 . ] since asas - sn detectors have relatively poor angular resolutions ( 7.5 arcsec / pixel ) , we provided coordinates from our own astrometry and astrometric catalogs for asas - sn cvs .
we used sdss , the initial gaia source list ( igsl , @xcite ) and guide star catalog 2.3.2 .
the coordinates used in this paper are j2000.0 .
questions have been raised to our surveys how to select the periods among the aliases and what is the uncertainty .
such question are natural if one only sees pdm diagrams .
we should note that pdm ( and most of other period finding algorithms ) assumes `` uncorrelated '' ( in time ) observations and defines the statistics .
the actual data have more information , such as the superhump timing data .
even if the pdm result shows strong aliases , we can resolve the alias problem if we have sufficiently long continuous observations ( continuous data produce no aliases ) .
our period selection is mostly based on this principle , when alias selection is inconclusive by the pdm analysis only .
the second approach is to examine the trends of @xmath0 values against the trial periods .
a period longer than the actual one produces systematically decreasing @xmath0 values within each night , and our measurements of the superhump maxima have typical errors of 0.001 d , which is usually sufficient to select one - day aliases when multiple superhumps were detected each night . if the case is not , we describe the uncertainty of the selection .
we show an example how our method works by using the data of fi cet ( subsection [ obj : ficet ] ) . a pdm analysis over a wider range of periods ( figure [ fig : ficetshpdm2 ] ) gives an impression that there are many period candidates and a period of 0.057 d and 0.060 d are equally acceptable .
if we rely on statistics assuming temporarily `` uncorrelated '' observations , this would give equal significance to these periods .
an example for fi cet is shown in figure [ fig : ficetoccomp ] .
scatter in the @xmath0 values and systematic trends within each night are apparent for trial period 0.06033 d , which is rejected .
although we do not show similar figures for other objects due to the limitation of space , we made similar analysis for objects when ambiguities in period selection remained .
readers may be interested in the result of least absolute shrinkage and selection operator ( lasso ) ( @xcite ; @xcite ) analysis , which has been proven to be very effective in detecting rapidly varying periods in unevenly sampled data ( e.g. @xcite ; @xcite ; @xcite ; @xcite ) .
the result is shown in figure [ fig : ficetlasso ] , and the impression is so different from the pdm result ( figure [ fig : ficetshpdm2 ] ) .
however , we did nt widely used this method in selecting the aliases since our model used in lasso analysis also assumed temporarily uncorrelated observations , and the suppression of the aliases is simply a result of highly non - linear characteristics of compressed sensing . this figure ( if compared to classical figures ) would give misleading impression that there is no possibility for aliases and we have not used this type of figure in this paper .
( 88mm,70mm)fig1.eps ( 88mm,70mm)fig2.eps ( 88mm,70mm)fig3.eps ccccl subsection & object & year & observers or references & i d + [ obj : kxaql ] & kx aql & 2014 & rpc & + [ obj : nncam ] & nn cam & 2014 & dpv & + [ obj : v342cam ] & v342 cam & 2014 & oku , kis , ioh & + [ obj : oycar ] & oy car & 2014 & spe & + & & 2015 & spe & + [ obj : ficet ] & fi cet & 2014 & hac & + [ obj : zcha ] & z cha & 2014 & hac & + [ obj : yzcnc ] & yz cnc & 2014b & aavso & + [ obj : v337cyg ] & v337 cyg & 2014 & imi & + [ obj : v503cyg ] & v503 cyg & 2014 & kis , dpv & + & & 2014b & rpc & + [ obj : bcdor ] & bc dor & 2015 & hac , coo & + [ obj : v660her ] & v660 her & 2014 & ioh & + [ obj : cthya ] & ct hya & 2015 & spe , kis & + [ obj : lyhya ] & ly hya & 2014 & hac & + [ obj : mmhya ] & mm hya & 2014 & mdy , ioh & + [ obj : rzlmi ] & rz lmi & 2014 & mic & + [ obj : aylyr ] & ay lyr & 2014 & ous , aka & + [ obj : v453nor ] & v453 nor & 2014 & hac & + [ obj : dtoct ] & dt oct & 2014 & hac & + [ obj : uvper ] & uv per & 2014 & ioh , dpv , imi , mdy , kis , & + & & & nov , rpc & + [ obj : hypsc ] & hy psc & 2014 & oku , ioh , imi , kis & + + + + [ cols="^,^,^,^,<",options="header " , ] ( 160mm,110mm)fig151.eps in @xcite , we showed from the kepler data of v1504 cyg and v344 lyr that superhumps start to develop following precursor outbursts .
this finding was examined in more detail using the kepler data of the same objects in @xcite , @xcite . the same phenomenon was also confirmed in the kepler data of the background dwarf nova of kic 4378554 and v516 lyr @xcite .
these findings have strengthened the universal application of the tti model to various su uma - type dwarf novae . in this paper , we have observed similar evolution of superhumps following the precursor outburst in z cha .
although only one superhump was recorded , superhumps likely started to grow just before the final rise to the main superoutburst also in cy uma .
an @xmath0 analysis of su uma also suggested that superhumps likely started to appear around the maximum of the precursor outburst .
all of these findings are in good agreement with what the tti model predicts ( cf .
@xcite ) . in recent years , an increasing number of am cvn - type objects [ for recent reviews of am cvn - type objects , see e.g. @xcite ; @xcite ] have been recorded in outbursts and superhumps were detected during outbursts : cr boo ( observation in atypical state in @xcite , more regular superoutburst in @xcite ) ; v803 cen ( observation in atypical state in @xcite ) ; kl dra @xcite ; v406 hya @xcite ; sdss [email protected] ( @xcite ; @xcite ) ; yz lmi ( partial coverage in @xcite ) ; ptf1 [email protected] @xcite ; cp eri ( @xcite ; @xcite ) ; sdss [email protected] @xcite ; css j045019.7@xmath16093113 @xcite . in @xcite
, we also studied superhumps and outburst patterns in an am cvn star cr boo and found that the basic @xmath0 variation of superhumps in am cvn - type superoutburst is the same as in hydrogen - rich systems .
also detected stage a superhumps in cr boo , first time in am cvn - type objects with typical supercycles .
in this paper , we added new examples of am cvn - type systems ptf1 j071912 , sdss j090221 ( see also @xcite ) and sdss j173047 . during the present survey , two further objects asassn-14ei and
asassn-14mv have been identified as am cvn - type objects showing superoutbursts and multiple rebrightenings (; for spectroscopic identifications , see @xcite , see also vsnet - alert 18160 ) .
asassn-14cn was also found to be an eclipsing am cvn - type object in outburst ( vsnet - alert 17879 ; superhumps are yet to be detected ) .
asassn-14fv was also found to be an outbursting am cvn - type object ( @xcite for spectroscopic identification ; superhumps are yet to be detected ) .
these data suggest that about 8% of objects showing dwarf nova - type outbursts are am cvn - type objects .
this fraction is much larger than the hitherto known statistics ( about 1% in rkcat edition 7.21 @xcite ) .
am cvn - type objects may be more populous than have been considered .
there have also been an increasing number of ei psc - type objects ( cvs containing but depleted hydrogen with orbital period below the period minimum ) : v485 cen ( @xcite ; @xcite ) ; ei psc ( @xcite ; @xcite ; @xcite ) ; crts j102842.9@xmath16081927 ( @xcite ; @xcite ; @xcite ) ; crts j112253.3@xmath16111037 @xcite ; sbs 1108@xmath15574 ( @xcite ; @xcite ; @xcite ) ; css j174033.5@xmath15414756 ( ) .
candidate systems include crts j233313.0@xmath16155744 @xcite .
we have added another object of this class , v418 ser , in this paper . in all well - observed cases ,
the development of superhumps in these systems follow the same pattern as in hydrogen - rich systems . although the number of these objects is smaller than am cvn - type objects , these object may be more abundant than have been considered . in this paper , we studied long - term variation of supercycles in mm hya ( subsection [ obj : mmhya ] ) and cy uma ( subsection [ obj : cyuma ] ) . in mm hya ,
variations of the supercycle in the range of 330 d and 386 d were recorded . in cy uma , a sudden decrease of the supercycle from 362(3 )
d to 290(1 ) d was observed in 2003 . in both systems , the supercycle tended to be constant for several to ten years , and there was a tendency of a sudden switch to a different period .
similar systematic variations of cycle lengths were studied in ss cyg - type dwarf novae [ cf .
ar and , uu aql , ru peg : @xcite ] .
@xcite suggested the solar - type activity as a possible cause , but the phenomenon still remained a puzzle . in recent years , @xcite recorded variations of supercycles from 43.6 to 59.2 d in er uma .
the variations of the supercycle in er uma was an order similar to those in the two systems studied in this paper .
although there was a possibility that a disk tilt , which is supposed to produce negative superhumps ( @xcite ; @xcite ) , could affect the outburst properties , @xcite could not find a correlation between the appearance of negative superhumps and the supercycle length . since a disk tilt is less likely to occur in systems with lower mass - transfer rate like mm hya and cy uma supposing a theoretical interpretation by @xcite
, we consider it less likely that a disk tilt can explain the variations of the supercycle commonly seen in a variety of dwarf novae .
the mechanism needs to be sought further .
in addition to results of observations superhumps of the objects studied in this paper , the major findings we obtained can be summarized as follows . *
the contribution of various surveys in detecting su uma - type dwarf novae have dramatically changed in the last ten years .
the advantage of a wide - field survey on nightly basis ( such as asas - sn survey ) has become obvious . *
thanks to the increase of samples of su uma - type dwarf novae , we could clarify the distribution of orbital periods in su uma - type dwarf novae , which approximate the cv population with short orbital periods .
there is a sharp cut - off at a period of 0.053 d , which is considered to be the period minimum .
there is a high concentration of objects just above the period minimum , which can be interpreted as the `` period spike '' .
the distribution monotonically decreases towards the longer period , and there is little indication of the period gap . *
we observed a precursor outburst in z cha and found that superhumps developed just following the precursor outburst .
@xmath0 analyses in cy uma and yz cnc also support this finding .
the agreement of the time of growing superhumps with the rising branch following the precursor outburst has been confirmed in all su uma - type systems so far studied .
this finding provides a strong support to the thermal - tidal disk instability model which predicts that a superoutburst develops as a result of the development of superhumps .
* we detected possible negative superhumps in z cha . during the phase of negative superhumps ,
the outburst cycle apparently lengthened .
this finding seems to support a suggestion that a disk tilt suppresses normal outbursts .
* we studied secular variations of the quiescent brightness and the amplitude of orbital humps throughout one supercycle in z cha .
the quiescent brightness decreased as the system approached the next superoutburst .
there was no enhanced orbital humps just before the superoutburst .
* we studied long - term trends in supercycles in mm hya and cy uma and found systematic variations of supercycles of @xmath120% .
this degree and characteristics os variations are similar to those recorded in other ss cyg - type dwarf novae and the su uma - type system er uma .
a disk tilt is unlikely a common source of this variation . *
the wz sge - type object asassn-15bp showed a phase jump of superhumps during the plateau phase . * a sizable number of am cvn - type objects ( ptf1 j071912 , sdss j090221 , sdss j173047 ) were studied in this paper and four more am cvn - type objects ( asassn-14cn , asassn-14ei , asassn-14fv and asassn-14mv ) were found in this period .
this number suggests that about 8% of objects showing dwarf nova - type outbursts are am cvn - type objects .
* we have added another ei psc - type object v418 ser in this paper .
these object may be more abundant than have been considered . *
css j174033 , an ei psc - type object , showed a similar type of superoutbursts in 2012 and 2014 , comprising of a dip and the second plateau phase .
this finding suggests that the same type of superoutbursts tend to be reproduced in ei psc - type objects as in hydrogen - rich systems . * ot j213806 , a wz sge - type object ,
exhibited a remarkably different peak brightness and @xmath0 diagrams between the 2010 and 2014 superoutbursts .
the fainter superoutburst in 2014 was shorter and the difference was most striking in the later part of the plateau phase .
* master j085854 showed two post - superoutburst rebrightenings .
this object had a rather exceptionally short superhump period among the objects showing multiple rebrightenings .
* four deeply eclipsing su uma - type dwarf novae were identified ( asassn-13cx , asassn-14ag , asassn-15bu , nsv 4618 ) .
asassn-14id also showed shallow eclipses .
this work was supported by the grant - in - aid `` initiative for high - dimensional data - driven science through deepening of sparse modeling '' ( 25120007 ) from the ministry of education , culture , sports , science and technology ( mext ) of japan .
the authors are grateful to observers of vsnet collaboration and vsolj observers who supplied vital data .
we acknowledge with thanks the variable star observations from the aavso international database contributed by observers worldwide and used in this research .
we are also grateful to the vsolj database .
this work is deeply indebted to outburst detections and announcement by a number of variable star observers worldwide , including participants of cvnet and baa vss alert .
the ccd operation of the bronberg observatory is partly sponsored by the center for backyard astrophysics .
we are grateful to the catalina real - time transient survey team for making their real - time detection of transient objects available to the public .
r. modic acknowledge the bradford robotic telescope for the detection of the 2014 outburst of qz vir .
the work by a. sklyanov is partially performed according to the russian government program of competitive growth of kazan federal university . | continuing the project described by @xcite , we collected times of superhump maxima for 102 su uma - type dwarf novae observed mainly during the 20142015 season and characterized these objects .
our project has greatly improved the statistics of the distribution of orbital periods , which is a good approximation of the distribution of cataclysmic variables at the terminal evolutionary stage , and confirmed the presence of a period minimum at a period of 0.053 d and a period spike just above this period .
the number density monotonically decreased toward the longer period and there was no strong indication of a period gap .
we detected possible negative superhumps in z cha .
it is possible that normal outbursts are also suppressed by the presence of a disk tilt in this system .
there was no indication of enhanced orbital humps just preceding the superoutburst , and this result favors the thermal - tidal disk instability as the origin of superoutbursts .
we detected superhumpsin three am cvn - type dwarf novae .
our observations and recent other detections suggest that 8% of objects showing dwarf nova - type outbursts are am cvn - type objects .
am cvn - type objects and ei psc - type object may be more abundant than previously recognized . ot j213806 , a wz sge - type object , exhibited a remarkably different feature between the 2010 and 2014 superoutbursts .
although the 2014 superoutburst was much fainter the plateau phase was shorter than the 2010 one , the course of the rebrightening phase was similar .
this object indicates that the @xmath0 diagrams of superhumps can be indeed variable at least in wz sge - type objects .
four deeply eclipsing su uma - type dwarf novae ( asassn-13cx , asassn-14ag , asassn-15bu , nsv 4618 ) were identified .
we studied long - term trends in supercycles in mm hya and cy uma and found systematic variations of supercycles of @xmath120% . |
an essential precondition for a correct operation of the nervous system consists in well - ordered neural connections between different cell sheets .
an example , which has been explored both experimentally and theoretically in detail , is the formation of ordered projections between retina and tectum , a part of the brain which plays an important role in processing optical information @xcite . at an initial stage of ontogenesis , retinal ganglion cells have random synaptic contacts with the tectum . in the adult animal , however , a so - called _ retinotopic _ projection is realized : neighboring cells of the retina project onto neighboring cells of the tectum . a detailed analytical treatment of hussler and von der malsburg described these ontogenetic processes in terms of self - organization @xcite .
in that work retina and tectum were treated as one - dimensional discrete cell arrays .
the dynamics of the connection weights between retina and tectum were assumed to be governed by the so - called hussler equations . in ref . @xcite we generalized these equations of evolution to _ continuous _ manifolds of _ arbitrary geometry _ and _ dimension_. furthermore , we performed an extensive synergetic analysis @xcite near the instability of stationary uniform connection weights between retina and tectum .
the resulting generic order parameter equations served as a starting point for analyzing retinotopic projections between euclidean manifolds in ref .
our results for strings turned out to be analogous to those for discrete linear chains , i.e. our model included the special case of hussler and von der malsburg @xcite .
additionally , we could show in the case of planar geometries that superimposing two modes under suitable conditions provides a state with a pronounced retinotopic character .
+ in this paper we apply our general model @xcite again to projections between two - dimensional manifolds .
now , however , we consider manifolds with _ constant positive curvature_. typically , the retina represents approximately a hemisphere , whereas the tectum has an oval form @xcite .
thus , it is biologically reasonable to model both cell sheets by spherical manifolds . without loss of generality
we assume that the two cell sheets for retina and tectum are represented by the surfaces of two unit spheres , respectively .
thus , in our model , the corresponding continuously distributed cells are represented by unit vectors @xmath0 and @xmath1 .
every ordered pair @xmath2 is connected by a positive connection weight @xmath3 as is illustrated in figure [ kugel ] .
the generalized hussler equations of ref .
@xcite for these connection weights are specified as follows @xmath4 the first term on the right - hand side describes cooperative synaptic growth processes , and the other terms stand for corresponding competitive growth processes . the total growth rate is defined by @xmath5 where @xmath6 denotes the global growth rate of new synapses onto the tectum , and is the control parameter of our system .
the cooperativity functions @xmath7 , @xmath8 represent the neural connectivity within each manifold .
they are assumed to be positive , symmetric with respect to their arguments , and normalized .
the integrations in ( [ hslerkugel ] ) and ( [ gro ] ) are performed over all points @xmath9 on the manifolds , where @xmath10 represent the differential solid angles of the corresponding unit spheres .
note that the factors @xmath11 in eq .
( [ hslerkugel ] ) are twice the measure @xmath12 of the unit sphere , which is given by @xmath13 if the global growth rate of new synapses onto the tectum @xmath6 is large enough , the long - time dynamics is determined by a uniform connection weight .
however , we shall see within a linear analysis in section [ linanalys ] that this stationary solution becomes unstable at a critical value of the global growth rate .
therefore , we have to perform a nonlinear synergetic analysis , in section [ nonlinanalys ] , which yields the underlying order parameter equations in the vicinity of this bifurcation . as in the case of euclidean manifolds ,
we show that they have no quadratic terms , represent a potential dynamics , and allow for retinotopic modes . in section [ 11retino ]
we include the influence of higher modes upon the connection weights , which leads to recursion relations for the corresponding amplitudes .
if we restrict ourselves to special cooperativity functions , the resulting recursion relations can be solved analytically by using the method of generating functions . as a result of our analysis
we obtain a perfect one - to - one retinotopy if the global growth rate @xmath6 is decreased to zero .
the cells of retina and tectum , which are assumed to be continuously distributed on unit spheres , are represented by their unit vectors @xmath0 and @xmath1 , respectively .
the two cell sheets are connected by positive connection weights @xmath3 . ] according to the general reasoning in ref .
@xcite we start with fixing the metric on the manifolds and determine the eigenfunctions of the corresponding laplace - beltrami operator .
afterwards , we expand the cooperativity functions with respect to these eigenfunctions and perform a linear analysis of the stationary uniform state . for the time being we neglect the distinction between retina and tectum , because the following considerations are valid for both manifolds .
using spherical coordinates , we write the unit vector on the sphere as @xmath14 .
the laplace - beltrami operator on a manifold reads quite generally @xcite @xmath15 for the sphere the components of the covariant tensor @xmath16 are @xmath17 with this the determinant of the covariant metric tensor reads @xmath18 and the components of the contravariant metric are given by @xmath19 whence the laplace - beltrami operator for the sphere takes the well - known form @xmath20 its eigenfunctions are known to be given by spherical harmonics
@xmath21 : @xmath22 with @xmath23 and @xmath24 they are @xmath25-fold degenerate and form a complete orthonormal system on the unit sphere : @xmath26 the argument of the cooperativity functions @xmath27 is the scalar product @xmath28 which takes values between @xmath29 and @xmath30 . therefore the cooperativity functions can be expanded in terms of legendre functions @xmath31 , which form a complete orthogonal system on this interval @xcite : @xmath32 then the expansion of the cooperativity functions read @xmath33 where @xmath34 denote the respective expansion coefficients . using the legendre addition theorem @xcite @xmath35 we arrive , for each manifold , at the expansion @xmath36 note that the normalization of the cooperativity functions and the orthonormality relations ( [ kuor1 ] ) lead to the constraints @xmath37 .
the initial state of ontogenesis with randomly distributed synaptic contacts is described by the stationary uniform solution of the generalized hussler equations , @xmath38 .
its stability is analyzed by linearizing the hussler equations ( [ hslerkugel ] ) with respect to the deviation @xmath39 .
the resulting linearized equations read @xmath40 with the linear operator @xmath41 \nonumber \\ & & \hspace{-0.8cm}-\frac{1}{8\pi}\int\ ! d\omega_{r ' } \left[v(\hat t,\hat r ' ) + \int\ ! d\omega_{t''}\!\int\ !
d\omega_{r''}\ , c_t(\hat t\cdot\hat t '' ) \ , c_r(\hat r ' \cdot\hat r'')\ , v(\hat t'',\hat r '' ) \right]+\int\ !
d\omega_{r'}\ , c_t(\hat t\cdot \hat t ' ) \ , c_r(\hat r\cdot \hat r')\ , v(\hat t',\hat r ' ) \,.\end{aligned}\ ] ] to solve eq .
( [ gen43 ] ) , we have to consider the eigenvalue problem of the linear operator ( [ loperator ] ) .
it has the eigenfunctions @xmath42 and the spectrum of eigenvalues reads @xcite : @xmath43 by changing the uniform growth rate @xmath6 in a suitable way , the real parts of some eigenvalues ( [ kuew1042 ] ) become positive and the system can be driven to the neighborhood of an instability . which eigenvalues ( [ kuew1042 ] ) become unstable in general depends on the respective values of the given expansion coefficients @xmath44 , @xmath45 . if we assume monotonically decreasing expansion coefficients @xmath44 , @xmath45 , @xmath46 the maximum eigenvalue in ( [ kuew1042 ] ) is given by @xmath47 .
thus , the instability occurs when the global growth rate reaches its critical value @xmath48 . at this
instability point all nine modes with @xmath49 and @xmath50 , @xmath51 become unstable , where we have introduced the index @xmath52 for the unstable modes .
in this section we specialize the generic order parameter equations of ref .
@xcite to unit spheres .
we observe that the quadratic term vanishes and derive selection rules for the appearance of cubic terms .
furthermore , we essentially simplify the calculation of the order parameter equations by taking into account the symmetry properties of the cubic terms .
we show that the order parameter equations represent a potential dynamics , and determine the underlying potential .
the linear stability analysis motivates treating the nonlinear hussler equations ( [ hslerkugel ] ) near the instability by decomposing the deviation @xmath53 in unstable and stable contributions , @xmath54 using einstein s sum convention the expansion of the unstable modes reads @xmath55 and , correspondingly , the contribution of the stable modes is given by @xmath56 note that the summation in ( [ sexp ] ) is performed over all parameters @xmath57 except for @xmath49 , i.e. from now on the parameters @xmath57 stand for the stable modes alone . with the help of the slaving principle of synergetics @xcite the original high - dimensional system can be reduced to a low - dimensional one which only contains the unstable amplitudes .
the resulting order parameter equations read @xcite @xmath58 they contain , as usual , a linear , a quadratic , and a cubic term of the order parameters .
the corresponding coefficients can be expressed in terms of the expansion coefficients @xmath44 , @xmath45 of the cooperativity functions ( [ ku1 ] ) and integrals over products of the eigenfunctions @xmath59 : @xmath60 the quadratic coefficients read @xmath61 whereas the cubic coefficients are @xmath62 i_{1,1\,l}^{m^u , m^u { } ' m } \ , i_{1,1\,l}^{m^u , m^u { } ' m } - \frac{1}{4\sqrt{\pi } } \ , \left[\delta_{l0}\,\delta_{m0 } \delta_{m^u m^u { } ' } \ , \left ( 1 + f_l^r \right ) i_{1,1\,l}^{m^u , m^u { } ' m } \right.\right .
\nonumber \\ & & \left .
+ \delta_{l0}\,\delta_{m0}\delta_{m^u m^u { } ' } \ , \left ( 1+f_l^t\right ) \ , i_{1,1\,l}^{m^u , m^u { } ' m } \right ] \right\ } h_{ll}^{mm , m^u { } '' m^u { } '' m^u { } ' '' m^u { } ' '' } \ , \ , .
\label{bb}\end{aligned}\ ] ] note that eq .
( [ bb ] ) involves a summation over all stable modes @xmath63 .
as is common in synergetics , the cubic coefficients ( [ bb ] ) consist in general of two parts , one stemming from the order parameters themselves and the other representing the influence of the center manifold @xmath64 on the order parameter dynamics according to @xmath65 here the center manifold coefficients @xmath66 are defined by @xmath67 \ , .
\label{qres}\end{aligned}\ ] ] the order parameter equations contain the following integrals : @xmath68 .
the first integral is obtained by the orthonormality relation ( [ kuor1 ] ) and @xmath69 yielding @xmath70 .
integrals over three and four spherical harmonics can be calculated with the help of the following relation @xcite : @xmath71 where @xmath72 represent the clebsch - gordan coefficients @xcite . applying ( [ kuyprod ] ) to integrals over three spherical harmonics leads to @xmath73 for @xmath74 it follows @xmath75 as the clebsch - gordan coefficients @xmath76 vanish if the sum @xmath77 is odd @xcite , we obtain @xmath78 .
thus , the quadratic contribution ( [ aa ] ) to the order parameter equations ( [ ope ] ) vanishes , by analogy with euclidean manifolds @xcite .
furthermore , non - vanishing integrals ( [ 1046 ] ) can only occur for @xmath79 and @xmath80 .
for @xmath79 we obtain from the clebsch - gordan coefficients @xcite the result @xmath81 for @xmath80 we find , correspondingly , the nonvanishing integrals @xmath82
@xmath83 furthermore , the integrals @xmath84 follow from @xmath85 integrals over four spherical harmonics can also be calculated with the help of ( [ kuyprod ] ) , and the result is @xmath86 specialyzing ( [ sphere38 ] ) to @xmath87 and taking into account ( [ 3fachint ] ) leads to @xmath88 .
thus , we obtain the selection rule that the nonvanishing integrals @xmath89 fulfill the condition @xmath90 . the detailed evaluation yields for those the respective values @xmath91 to simplify the calculation of the cubic coefficients ( [ bb ] ) in the order parameter equations ( [ ope ] ) , we perform some basic considerations which lead to helpful symmetry properties . to this end
we start with replacing @xmath92 by @xmath93 . using eq .
( [ kuyminusm ] ) we obtain @xmath94 .
corresponding symmetry relations can also be derived for the other terms in ( [ bb ] ) .
therefore , we conclude that the order parameter equation for @xmath95 is obtained from that of @xmath96 by negating all indices @xmath97 and @xmath92 with unchanged factors .
thus , instead of explicitly calculating nine order parameter equations , it is sufficient to restrict oneself determining the order parameter equations for @xmath98 , @xmath99 , @xmath100 , and @xmath101 .
the remaining five order parameter equations follow instantaneously from those by applying the symmetry relations . with this
the order parameter equations result in @xmath102 with the abbreviations @xmath103 , @xmath104 and @xmath105 the respective coefficients in ( [ uequ ] ) read @xmath106 the first term proportional to @xmath107 describes the influence of the order parameters themselves , while the other terms stand for the contributions of the center manifold . to investigate how the complex order parameter equations contribute to the one - to - one retinotopy , we transform them to real variables according to @xmath108 then the equations of evolution for the real variables @xmath109 read @xmath110 note that the real order parameter equations ( [ kureellopgu0])([kureellopgu8 ] ) follow according to @xmath111 from the potential @xmath112(u_1 ^ 2+u_2 ^ 2+u_3 ^ 2+u_4 ^ 2)-\frac{\beta_7}{2}(u_1 ^ 2+u_2 ^ 2)(u_3 ^ 2+u_4 ^ 2)\nonumber\\ & & \hspace{-1.1cm}-\beta_3 ( u_5 ^ 2+u_6 ^ 2)(u_7 ^ 2+u_8 ^ 2)-\frac{\beta_8}{4}\left[(u_1 ^ 2 + u_2 ^ 2)^2+(u_3 ^ 2+u_4 ^ 2)^2\right]+\frac{\beta_9}{4}(u_5 ^ 2+u_6 ^ 2)^2+\frac{\bar\beta_9}{4}(u_7 ^ 2+u_8 ^ 2)^2 \ , .
\label{kupot}\end{aligned}\ ] ] naturally , a complete analytical determination of all stationary states of the real order parameter equations ( [ kureellopgu0])([kureellopgu8 ] ) is impossible .
however , we are able to demonstrate that certain stationary states admit for retinotopic modes .
to this end we consider the special case @xmath113 .
then the equations ( [ kureellopgu0 ] ) , ( [ kureellopgu3 ] ) , and ( [ kureellopgu4 ] ) for the non - vanishing amplitudes @xmath114 , @xmath115 , @xmath116 reduce to @xmath117 due to the relation @xmath118 one obtains constant phase - shift angles , i.e. it holds @xmath119 .
therefore , the system of three coupled differential equations can be reduced to two variables . to this end
we introduce the new variable @xmath120 which leads to @xmath121 the stationary solution , which corresponds to a coexistence of the two modes , is given by @xmath122 where we used the relation @xmath123 following from ( [ sphere41 ] ) . demanding real amplitudes
@xmath114 , @xmath124 leads to the coexistence condition @xmath125 furthermore , we require stability for this state . therefore we consider the corresponding potential @xmath126 , which can be read off from ( [ kupot ] ) and ( [ kuneuvariab ] ) : @xmath127 stable states correspond to a minimum of @xmath128 , which leads to the conditions @xmath129 the inequalities ( [ kukoexungl1 ] ) , ( [ kukoexungl2 ] ) can be summarized according to @xmath130 if they are valid , both the @xmath114- and the @xmath124-mode coexist .
if we set @xmath131 , without loss of generality , the solution reads in complex variables according to ( [ kuneuvariab ] ) @xmath132 thus , the unstable part ( [ uexp ] ) is given by @xmath133\,.\ ] ] using the legendre addition theorem ( [ kuaddtheorem ] ) reduces ( [ sphere66 ] ) to @xmath134 with @xmath135 .
thus , the unstable part is minimal , if @xmath1 and @xmath0 are antiparallel , i.e. the distance of the corresponding points on the unit sphere is maximum .
decreasing of the angle between @xmath1 and @xmath0 leads to increasing values of @xmath136 , and the maximum occurs for parallel unit vectors .
this justifies calling the mode ( [ kupl1 ] ) retinotopic .
now we investigate whether the generalized hussler equations ( [ hslerkugel ] ) describe the emergence of a perfect one - to - one retinotopy between two spheres . to this end
we follow the unpublished suggestions of ref .
@xcite and treat systematically the contribution of higher modes .
because the legendre functions form a complete orthogonal system ( [ wagnerportho ] ) , ( [ wagnerpvollst ] ) for functions defined on the interval @xmath137 $ ] , their products can always be written as linear combinations of legendre functions .
this motivates that the influence of higher modes upon the connection weights , which obey the generalized hussler equations ( [ hslerkugel ] ) , can be included by the ansatz @xmath138 where the amplitudes @xmath139 are time dependent . inserting ( [ wagneransatz ] ) into the generalized hussler equations ( [ hslerkugel ] ) and performing the integrals over the respective unit spheres leads to @xmath140 + \sum_{l=0}^{\infty}(2l+1)z_l p_l(\sigma)\sum_{l'=0}^{\infty}(2l'+1)z_{l'}f_{l'}^t f_{l'}^r [ p_{l'}(\sigma)-z_{l ' } ] \,.\label{wagnerstern}\ ] ] the products of legendre functions occuring in ( [ wagnerstern ] ) can be reduced to linear combinations of single legendre functions according to the standard decomposition @xcite @xmath141 with the coefficients @xmath142 thus
, contributions to the polynomial @xmath143 only occur iff the relation @xmath144 is fulfilled . furthermore , using the orthonormality relation ( [ wagnerportho ] ) yields the following recursion relation for the amplitudes @xmath139 : @xmath145-(2l+1)z_l(z_0 ^ 2 + 3f_1^t f_1^r z_1 ^ 2 ) + \sum_{l'=0}^{\infty}(2l'+1)z_{l'}\left[\sum_{l''=0}^l ( 2l''+1)z_{l''}f_{l''}^t f_{l''}^r \right.\nonumber\\ & & \left.\times\sum_{k=0}^{l''}a_{l',l'',k}\delta_{k,(l'+l''-l)/2}+\sum_{l''=l'+1}^{\infty}(2l''+1 ) z_{l''}f_{l''}^t f_{l''}^r \sum_{k=0}^{l'}a_{l',l'',k}\delta_{k,(l'+l''-l)/2 } \right]\ , .
\label{128}\end{aligned}\ ] ] note that eq .
( [ 128 ] ) can not be solved analytically for arbitrary expansion coefficients @xmath146 , @xmath45 of the cooperativity functions .
therefore , we restrict ourselves from now on to a special case . for simplicity
we assume that the expansion of the cooperativity functions ( [ kuct ] ) breaks down after the first order : @xmath147\,,\qquad c_r(\hat r\cdot\hat r ' ) = \frac{1}{4\pi}[1 + 3 f_1^r p_1(\hat r\cdot\hat r')]\,.\ ] ] with this choice the recursion relation ( [ 128 ] ) for @xmath79 reduces to @xmath148 where we have used again the abbreviation @xmath104 . for @xmath149 , by taking into account ( [ wagnera ] ) , we obtain @xmath150 the long - time behavior of the system corresponds to its stationary states .
they are determined by @xmath151 from ( [ wagnerz0 ] ) , whereas ( [ wagner3stern ] ) leads to a nonlinear recursion relation for the amplitudes @xmath139 with @xmath149 . however , by introducing the variable @xmath152 this nonlinear recursion relation can be formally transformed into the linear one @xmath153 thus , solving the nonlinear recursion relation ( [ wagner3stern ] ) amounts to solving the linear recursion relation ( [ wagnerrek ] ) for @xmath154 in such a way that the self - consistency condition ( [ wagnerualpha ] ) is fulfilled . to determine the amplitudes @xmath154
we calculate their generating function @xmath155 where we have the normalization @xmath156 multiplying both sides of ( [ wagnerrek ] ) with @xmath157 and summing over @xmath158 leads to an inhomogeneous nonlinear partial differential equation of first order for the generating function : @xmath159 at first , we consider the homogeneous equation corresponding to ( [ wagnerinh ] ) : @xmath160 it is solved by the method of separating variables , yielding @xmath161 where @xmath162 is an integration constant .
afterwards , we determine a particular solution of the inhomogeneous equation ( [ wagnerinh ] ) by using the method of varying constants . using the ansatz
@xmath163 leads to the differential equation @xmath164 which is solved by using @xcite : @xmath165\ln[2\sqrt{x^2 - 2ux+1}+2(x - u)]}{\sqrt{x^2 - 2ux+1}}\,.\ ] ] thus , the complete solution @xmath166 of eq .
( [ wagnerinh ] ) reads as follows : @xmath167\ln[2\sqrt{x^2 - 2ux+1}+2(x - u)]}{\sqrt{x^2 - 2ux+1}}\,.\ ] ] furthermore , using the normalization condition ( [ wagnerex0 ] ) fixes the integration constant to @xmath168\ln(2 - 2u)$ ] .
thus , the generating function is finally given by @xmath169\ln\displaystyle{\frac{\sqrt{x^2 - 2ux+1}+x - u}{1-u}}}{\sqrt{x^2 - 2ux+1}}\,.\ ] ] we now determine the unknown amplitudes @xmath154 . from the mathematical literature
it is well - known that the recursion relation ( [ wagnerrek ] ) holds both for the legendre functions of first kind @xmath170 and second kind @xmath171 , respectively @xcite .
thus , we expect that the generating function ( [ wagnererzeugende ] ) can be represented as a linear combination of the generating functions of the legendre functions of both first and second kind , which are given by @xcite and @xcite : @xmath172 indeed , taking into account the explicit form of the legendre function of second kind for @xmath79 @xcite @xmath173 the generating function ( [ wagnererzeugende ] ) decomposes according to @xmath174q_0(u)\}e_p(x , u)-[z_1(u)-u]e_q(x , u)\,.\ ] ] inserting ( [ pgen ] ) , ( [ qgen ] ) and performing a comparison with ( [ wagnererz ] ) then yields the result @xmath175q_0(u)\}p_l(u)-[z_1(u)-u]q_l(u)\,.\ ] ] thus , the amplitudes @xmath154 turn out to be linear combinations of @xmath170 and @xmath171 . to fix the yet undetermined amplitude @xmath176 in the expansion coefficients of ( [ wagnerzl ] ) , we have to take into account the boundary condition that the sum in the ansatz ( [ wagneransatz ] ) has to converge .
the legendre functions of first and second kind @xmath170 and @xmath171 for @xmath177 .
we have @xmath178 , whereas @xmath171 diverges for @xmath179 .
important for the boundary condition of @xmath154 is the different behavior for increasing values of @xmath180 : @xmath170 diverges according to ( [ wagnerpldiv ] ) , whereas @xmath171 converges to zero . ] because the legendre functions @xmath181 do not vanish with increasing @xmath180 , we must require @xmath182 the series of legendre functions of first kind @xmath170 with fixed @xmath177 diverges for @xmath183 according to @xcite @xmath184 the legendre functions of second kind @xmath171 , however , converge to zero ( see figure [ pqlegendre ] ) .
thus , performing the limit @xmath183 in eq .
( [ wagnerzl ] ) , we obtain @xmath185q_0(u)=0\,.\ ] ] from the explicit form @xcite @xmath186 it follows that @xmath176 is fixed according to @xmath187 with this we obtain that the result ( [ wagnerzl ] ) finally reads @xmath188 which is not valid only for @xmath149 but also for @xmath79 due to
( [ wagnerex0 ] ) . inserting ( [ wagnerzlq ] ) into ( [ wagneransatz ] )
yields the following solution for the connection weight : @xmath189 using the identity @xcite @xmath190 and ( [ wagnerq0 ] ) , we obtain for the connection weight @xmath191 note that integrating ( [ 117 ] ) over the unit sphere leads to @xmath192 i.e. the total connection weight coincides with the measure ( [ kumass ] ) .
+ on the other hand we have to take into account that the self - consistency condition ( [ wagnerualpha ] ) yields an explicit relation between the variable @xmath52 and the control parameter @xmath6 . indeed
, we infer from ( [ wagnerualpha ] ) and ( [ wagnerz1 ] ) the following transcendental relation between @xmath6 and @xmath52 @xmath193\,,\ ] ] which is depicted in figure [ alphavonu]a .
\a ) a ) relation ( [ wagnera2 g ] ) between the control parameter @xmath6 and the variable @xmath52 .
b ) the connection weight for different values of the control parameter @xmath6 . for decreasing values of @xmath6 the connection weight around @xmath194 is growing . in the limiting case @xmath195 the connection weight @xmath196 becomes dirac s delta function ( [ kudeltafkt]).,title="fig : " ] b )
a ) relation ( [ wagnera2 g ] ) between the control parameter @xmath6 and the variable @xmath52 .
b ) the connection weight for different values of the control parameter @xmath6 . for decreasing values of @xmath6 the connection weight around @xmath194 is growing . in the limiting case
@xmath195 the connection weight @xmath196 becomes dirac s delta function ( [ kudeltafkt]).,title="fig : " ] the limiting value of ( [ wagnera2 g ] ) for @xmath197 is determined with the help of the expansion @xcite @xmath198 and reads @xmath199 thus , we conclude that the case @xmath200 corresponds to the instability point @xmath48 , which was obtained from the linear stability analysis in section [ linanalys ] .
correspondingly , using again ( [ wagnera2 g ] ) , we observe that the connection weight ( [ 117 ] ) coincides in the limit @xmath197 with a uniform distribution : @xmath201 another biological important special case is @xmath179 , where we obtain from ( [ wagnera2 g ] ) @xmath202 furthermore , considering the limit @xmath179 in ( [ 117 ] ) for @xmath203 , we obtain @xmath204 on the other hand , integrating ( [ 117 ] ) for @xmath179 over @xmath205 yields @xmath206 therefore , we conclude that the connection weight @xmath207 becomes in this limit dirac s delta function : @xmath208 thus , decreasing the control parameter @xmath6 means that the projection between two spheres becomes sharper and sharper ( see figure [ alphavonu]b ) . a perfect one - to - one retinotopy
is achieved for @xmath209 when the uniform and undifferentiated formation of new synapses onto the tectum is completely terminated .
in this series of three papers we have analyzed in detail the self - organized formation of retinotopic projections between manifolds of different geometries . applying our generalized hussler equations @xcite to euclidean manifolds @xcite , and to spheres in the present paper , led to remarkably analogous results .
both for one - dimensional strings and for spheres we have furnished proof that our generalized hussler equations describe , indeed , the emergence of a perfect one - to - one retinotopy .
furthermore , we have shown in both cases that the underlying order parameter equations follow from a potential dynamics and do not contain quadratic terms . however ,
in contrast to strings , spherical manifolds represent a more adequate description for retina and tectum .
therefore , the present paper represents an essential progress in the understanding of the ontogenetic development of neural connections between retina and tectum .
goodhill and l.j .
richards , trends neurosci . *
22 * , 529 ( 1999 ) a.f .
hussler and c. von der malsburg , j. theoret . neurobiol .
* 2 * , 47 ( 1983 ) m. gmann , a. pelster , and g. wunner , _ self - organized formation of retinotopic projections between manifolds of different geometries part 1 : the general model ; _ eprint : physics/0607253 h. haken , _ synergetics , an introduction _ , third edition , springer , berlin ( 1983 ) h. haken , _ advanced synergetics _ , springer , berlin ( 1983 ) m. gmann , a. pelster , and g. wunner , _ self - organized formation of retinotopic projections between manifolds of different geometries part 2 : euclidean manifolds ; _ eprint : physics/0607259 m. gmann , _ self - organization between manifolds of euclidean and non - euclidean geometry by cooperation and competition _ , universitt stuttgart , ph.d .
thesis ( 2006 ) ; + internet : www.itp1.uni-stuttgart.de/publikationen/guessmann_doktor_2006.pdf h. kleinert , _ path integrals in quantum mechanics , statistics , polymer physics and financial markets _ , 4th ed .
world scientific , singapore ( 2006 ) i. s. gradshteyn and i. m. ryzhik , _ table of integrals , series , and products _ , 4th ed .
academic press , new york ( 1965 ) v. heine , _ group theory in quantum mechanics _ , dover , new york ( 1993 ) w. wagner and c. von der malsburg , private communication c. cohen - tannoudji , b. diu , and f. lalo , _ quantum mechanics _ , vol .
2 , wiley - interscience publication , new york ( 1977 ) g. b. arfken and h. j. weber , _ mathematical methods for physicists _
, 5th ed . academic press ,
london ( 2001 ) | we follow our general model in ref . @xcite and analyze the formation of retinotopic projections for the biologically relevant situation of _ spherical _ geometries . to this end
we elaborate both a linear and a nonlinear synergetic analysis which results in order parameter equations for the dynamics of connection weights between two spherical cell sheets .
we show that these equations of evolution provide stable stationary solutions which correspond to retinotopic modes
. a further analysis of higher modes furnishes proof that our model describes the emergence of a perfect one - to - one retinotopy between two spheres . |
evaluating the amount of work extraction from reservoirs is an important issue in thermodynamics and nonequilibrium statistical mechanics . here , we focus on the work extraction by using information , where two distinct formalisms exist .
one formalism concerns the measurement - feedback processes between two systems : a celebrated example is maxwell s demon . in this case , the generalizations of the second law and the fluctuation theorem have been obtained by including the _ mutual information _
@xcite between the system and the controller ( demon ) @xcite , and it has been applied to biochemical sensing @xcite .
we call this avenue of research as the measurement - feedback ( mf ) formalism . the other formalism concerns the work extraction from bit sequences called _ information reservoirs _ , where the shannon entropy of a long bit sequence is consumed to extract work ( see also fig .
[ f : f : mjrevisit](a ) ) .
a second - law - like inequality stronger than the conventional second law has been obtained in this formalism @xcite .
we call this as the information - reservoir ( ir ) formalism .
although both of these two formalisms address the fundamental link between information and thermodynamics , the relation between the mf and the ir formalisms has not been fully understood . in previous researches ,
an important observation is that the ir formalism gives a stronger inequality than the mf formalism in some setups @xcite .
in contrast , in this paper we show that the mf formalism is also applicable to ir models , and leads to a stronger second - law - like inequality than the ir formalism .
the central idea here is the additive decomposition of the total entropy production , which was first investigated in the context of a single measurement and feedback @xcite , and then extended to general markov jump processes @xcite .
this decomposition assigns the _ partial entropy production _ to each transition path , for which a generalized second law is obtained .
in addition , we show that the mandal - jarzynski ( mj ) model , a prominent ir model , can be regarded as a contracted model of an autonomous mf model even at the level of stochastic trajectories . in this special case ,
the newly obtained inequality is shown to be equivalent to the inequality with the mutual - information flow in terms of the mf formalism @xcite .
our result would lead to the comprehensive understanding of the link between information and thermodynamics .
s : partial we briefly review the notion of the partial entropy production introduced in ref .
@xcite , which is an extension of the mf formalism . throughout this paper ,
we consider markov jump processes with discrete states and continuous time , and focus on the stationary states in the presence of a single heat bath .
we normalize the inverse temperature @xmath0 as 1 for simplicity .
the transition from a state @xmath1 to another state @xmath2 is written as @xmath3 , whose transition rate is denoted by @xmath4 . the stationary distribution @xmath5 satisfies the stationary condition @xmath6 we define the total entropy production @xmath7 with time interval @xmath8 as @xcite @xmath9 where we denoted the number of jumps by @xmath10 , and the @xmath11-th jump occurs from @xmath12 to @xmath13 at time @xmath14 ( @xmath15 ) .
the state at time @xmath16 is written as @xmath17 .
@xmath18 is the heat absorption by the system from the heat bath with the transition @xmath3 , and @xmath19 represents the stochastic shannon entropy at the state @xmath2 .
we note that the local detailed balance condition @xcite is assumed .
we now define the partial entropy production @xmath20 for a set of transitions , written as @xmath21 , which is a subset of all possible transitions .
we first define the partial probability flux at the state @xmath2 as @xmath22 .
we then define the partial entropy production @xmath23 as @xmath24 here , we defined @xmath25 , where @xmath26 takes @xmath27 if @xmath28 and takes @xmath29 otherwise .
we also defined @xmath30 with @xmath31 .
it is crucial that the partial entropy production @xmath20 is an additive decomposition of the total entropy production @xmath32 where @xmath33 is the complement of @xmath21 .
the partial entropy production satisfies the fluctuation theorem : @xmath34 which leads to the generalized second law for the partial entropy production : @xmath35 it has been shown that the fluctuation theorem for autonomous measurement and feedback processes is obtained as a special case of eq .
( [ partial - ineq ] ) @xcite .
we further discuss the significance of this inequality in sec .
[ s : power ] .
s : mj we here explain the mj model which extracts work from a bit sequence @xcite . in particular , we consider a simplified version of the mj model introduced in ref .
the simplified mj model is a composite system of a bit sequence and a weight with the coupling reaction @xmath36 ( see fig .
[ f : f : mjrevisit](a ) ) . here , 0 and 1 represent
states of a bit , @xmath37 and @xmath38 represent ( relative ) positions of the weight on ladder - like energy levels with energy interval @xmath39 .
the weight plays the role of a work storage .
the bit sequence moves in one direction , from left to right , with the poisson process with the rate @xmath40 .
the probability that the incoming bit is 1 is @xmath41 .
when a new bit comes into the central site , the state of the weight is instantly labeled as @xmath38 ( @xmath37 ) corresponding to the state of the incoming bit 0 ( 1 ) , and the higher ( lower ) energy level is labeled as @xmath37 ( @xmath38 ) ( see fig .
[ f : f : mjrevisit](b ) ) . under this labeling ,
the possible states of the composite system are only @xmath42 and @xmath43 .
there are two possible transitions between @xmath37 and @xmath38 ( see fig .
[ f : f : mjrevisit](c ) ) .
one is the stochastic jump induced by the thermal bath with rates @xmath44 and @xmath45 , which satisfy @xmath46 .
this is the coupling reaction of the up - down of the weight and the flip of a bit such that @xmath36 .
we denote the set of these transitions by @xmath47 . the other
is induced by the bit stream with rates @xmath48 and @xmath49 ; from the above - mentioned rule of the state labeling , the stream of the bit sequence also changes the states @xmath37 and @xmath38 .
if @xmath41 is small ( i.e. , the entropy of the bit sequence is low ) , the weight can be raised on average by increasing the entropy of the bit sequence .
the stationary probability of @xmath37 is given by @xmath50 , where @xmath51 and @xmath52 . since @xmath53 is the probability that the outgoing bit is 1 , the average energy gain per single incoming bit is given by @xmath54 , which we regard as work extraction .
we note that the entropy change in the work storage is irrelevant to our argument , because the fluctuation of the state of the work storage does not affect the procedure of autonomous work extraction .
this type of work storages for autonomous work extraction have been discussed in refs .
s : power we now apply the fluctuation theorem with the partial entropy production , discussed in sec . [
s : partial ] , to the simplified mj model .
we set @xmath21 in sec .
[ s : partial ] to @xmath47 in sec .
[ s : mj ] ( the red arrows in fig .
[ f : f : mjrevisit](c ) ) .
equality ( [ partial - ft ] ) and inequality ( [ partial - ineq ] ) are then written as @xmath55 and @xmath56 by using the stationary condition @xmath57 , inequality ( [ mj - ineq - mid ] ) reduces to a simple form : @xmath58 the left - hand side of ( [ restrict - mj ] ) represents the entropy change induced by the heat bath .
we next extend our foregoing argument to general markov jump processes with states @xmath59 . in line with ref .
@xcite , we divide pairs of transitions into two groups , which we refer to as _ thermal transitions _ and _ informational transitions_. we denote these sets of transitions by @xmath47 and @xmath60 , respectively ( see fig . [
f : mj - gen ] ) . the work is extracted during the thermal transitions . by setting @xmath47 to @xmath21 , eq .
( [ partial - ineq ] ) reduces to @xmath61 where we defined @xmath62 .
we now show the hierarchical structure of three upper bounds on @xmath63 with the simplified mj model : the conventional second law , an inequality derived from the ir formalism @xcite , and the newly derived inequality ( [ restrict - mj ] ) .
first , we compare the conventional second law and the inequality obtained from the ir formalism . to clarify the thermodynamic energy cost of the one - way stream of the bit sequence , we model the bit sequence by non - interacting particles with two components @xmath29 and @xmath27 ( fig .
[ f : f : tape - bath ] ) , where the transition rates are the same as @xmath64 .
particles in the bath stochastically come and go to the reaction site , which corresponds to the yellow cage of the mj model .
the ratio of the particles @xmath29 and @xmath27 in the bath is @xmath65 , and therefore the gain ( loss ) of the chemical potential with the transition @xmath66 ( @xmath67 ) is @xmath68 . although the kinetics of the particle - bath model is equivalent to that of the bit sequence , the entropy production in the particle - bath and that in the bit sequence are different as shown below .
the conventional second law for the total system with the particle - bath is given by @xmath69 on the other hand , the second law for the total system with the bit sequence has been derived from the ir formalism , which is given by @xcite @xmath70 we refer to this inequality as the mj inequality . subtracting the left - hand side ( lhs ) of ( [ mj - ineq ] ) from the lhs of ( [ mj-2nd ] ) , we obtain @xmath71 where @xmath72 represents the kullback - leibler divergence that is non - negative @xcite
. inequality ( [ 2nd - vs - mj ] ) confirms that the mj inequality gives a stronger bound than the conventional second law .
their difference @xmath73 characterizes an extra cost to move the bit sequence in one direction . in the particle - bath case ,
the outgoing particle from the reaction site is in state @xmath27 with probability @xmath74 , while a particle in the particle bath is in state @xmath27 with probability @xmath41 .
thus , when a particle leaves the reaction site to the particle bath , the entropy production in the particle bath is calculated as @xmath75 .
in contrast , in the bit sequence case , the ratio of bit 1 in the bit sequences on the left side of the cage ( i.e. , the particle bath releasing the incoming bit ) and that on the right side of the cage ( i.e. , the particle bath receiving the outgoing bit ) are different .
therefore , a bit on the right side of the cage is in state 1 with probability @xmath74 , and when a bit leaves the reaction site to the bath , the entropy production in the bath is calculated as @xmath76 .
the difference between the foregoing two entropy productions is equal to @xmath73 .
this implies that the strength of the mj inequality comes not from the use of the bit sequence carrying information but from the one - directional operation , which inevitably needs the additional energy cost .
we now compare two inequalities ( [ mj - ineq ] ) and ( [ restrict - mj ] ) ( i.e. , the mj inequality and the inequality derived from the partial entropy production method ) . subtracting the lhs of ( [ restrict - mj ] ) from the lhs of ( [ mj - ineq ] )
, we get @xmath77 here , @xmath78 characterizes the dissipation corresponding to the fact that the time evolution is not quasi - static @xcite . the left - hand side of the mj inequality ( [ mj - ineq ] )
is just characterized by the entropy change between the initial state ( i.e. , the state of the incoming bit ) and the final state ( i.e. , the state of the outgoing bit ) .
therefore , the mj inequality does not take into account the dissipation due to the free relaxation of a bit , while inequality ( [ restrict - mj ] ) can take it into account .
such dissipation is unavoidable , as autonomous information engines inevitably involve relaxation processes , because of the very fact that they are autonomous .
this is the physical reason why the partial entropy production method gives a stronger bound of the work extraction .
more generally , we can show that inequality ( [ restrict - mj - gen ] ) is a stronger inequality than the generalized version of the mj inequality for the general ir formalism @xcite .
the setup for the generalized mj inequality is the same as that for inequality ( [ restrict - mj - gen ] ) . the general version of the mj inequality ( [ mj - ineq ] ) is written as @xcite @xmath79 where we defined @xmath80 and @xmath81 .
we then compare two inequalities ( [ mj - gen ] ) and ( [ restrict - mj - gen ] ) .
subtracting the lhs of ( [ restrict - mj - gen ] ) from lhs of ( [ mj - gen ] ) , we obtain @xmath82 where we defined a positive constant @xmath83 as @xmath84 for each informational transition @xmath85 .
therefore , inequality ( [ restrict - mj - gen ] ) is stronger than ( [ mj - gen ] ) .
we next consider an autonomous measurement - feedback ( amf ) model @xcite , and show the equivalence of the simplified mj model and the amf model .
this equivalence allows us to directly compare the mj inequality and the second - law - like inequality with the mutual information flow .
the amf model consists of a fluctuating particle on the tilted one - dimensional discrete lattice and the delta - function - like wall ( see fig . [
f : info - rat](a ) ) .
the position on the lattice is labeled as @xmath86 in mod 2 , and the delta - function - like walls are at the right side of 0 or 1 , which we label as @xmath87 or @xmath88 , respectively .
the energy difference of the particle between a site and its right neighbor is @xmath39 .
the transition rate of the position of the particle due to the thermal diffusion is given by @xmath89 with @xmath46 .
the transition rate of the positions of the walls is given by @xmath90 which are equal to @xmath91 and @xmath92 , respectively .
we note that @xmath93 is regarded as the chemical potential difference that drives the switching of the walls . owing to the chemical driving ,
the particle climes up against the potential bias when @xmath94 .
we stress that the particle and the walls do not directly change the energy , but exchange only information . in this sense ,
the amf model is a typical example of an autonomous information - driven engine @xcite .
analogous to maxwell s demon , the walls measure the position of the particle , and change their own position depending on the measurement outcome , which is the feedback control .
the amf model is a bipartite system with two variables @xmath95 , and thus we can define the entropy production associated only with the particle as @xmath96 , where @xmath97 is the heat absorbed by the subsystem @xmath98 , and @xmath99 is the change in the stochastic shannon entropy of @xmath98 . due to the presence of the autonomous measurement and feedback , @xmath100 can be negative on average and does not satisfy the fluctuation theorem in the conventional form .
however , by taking into account the flow of stochastic mutual information @xmath101 @xcite , we obtain the modified fluctuation theorem in the stationary state @xcite : @xmath102 here we defined the mutual information flow by @xmath103 , where @xmath104 with @xmath105 and kronecker delta @xmath106 , and @xmath107 .
we stress that eq .
( [ infoft ] ) is a special case of eq .
( [ partial - ft ] ) : if we apply eq .
( [ partial - ft ] ) to the amf model by setting all transitions of @xmath98 to @xmath21 , we straightforwardly obtain eq .
( [ infoft ] ) . by applying jensen s inequality to eq .
( [ infoft ] ) , the generalized second law with the mutual information flow is obtained as @xcite @xmath108 where @xmath109 represents the stationary entropy production rate of @xmath98 .
this inequality is stronger than the conventional second law of the composite system .
we now consider the contraction of the amf model by contracting two states @xmath110 and @xmath111 to a single state @xmath38 , and @xmath112 and @xmath113 to @xmath37 ( see fig .
[ f : info - rat](b ) ) .
this contraction means that the absolute positions of the particle and walls are replaced by the relative positions . since the amf model has the two types of transition paths corresponding to the thermal diffusion of the particle and the chemical switching of the walls , its contracted model also has two types of transition paths between @xmath37 and @xmath38 , which we refer to as the thermal transition and the informational transition .
then , by denoting the transition rates of the thermal and informational transitions by @xmath114 and @xmath115 respectively , the contracted amf model is regarded equivalent to the simplified mj model introduced in sec .
[ s : mj ] .
correspondingly , each stochastic trajectory in the amf model is contracted to that in the simplified mj model , where the transition rates are kept unchanged : @xmath116 we denote as @xmath117 and @xmath118 the probabilities that trajectory @xmath119 from time 0 to @xmath120 and its contraction are realized in the amf model and the simplified mj model , respectively .
we then obtain @xmath121 where the factor @xmath122 comes from the contraction of the initial state : if the initial state after the contraction is @xmath38 ( @xmath37 ) , the initial state before the contraction is either @xmath110 or @xmath111 ( @xmath112 or @xmath113 ) with probability @xmath122 .
we next show that eq .
( [ infoft ] ) directly reduces to eq .
( [ mj - ft ] ) , and correspondingly , inequality ( [ infoineq ] ) reduces to inequality ( [ restrict - mj ] ) .
therefore , for the case of the simplified mj model we can see the direct application of the mf formalism to an ir model without invoking the partial entropy production method .
we first note that @xmath123 holds for @xmath124 or @xmath125 , where every line is just zero if @xmath126 , @xmath127 .
we then obtain @xmath128 for any trajectory @xmath119 of the amf model .
equality ( [ s - equiv ] ) implies that the two fluctuation theorems , eqs .
( [ infoft ] ) and ( [ mj - ft ] ) , are equivalent , and thus inequalities ( [ infoineq ] ) and ( [ restrict - mj ] ) are also equivalent .
this clarifies the physical meaning of inequality ( [ restrict - mj ] ) from the viewpoint of the mf formalism : the bit sequence plays the equivalent role to the chemical fuel that drives the autonomous measurement and feedback .
we have established the fundamental relation between the mf and the ir formalisms .
inequality ( [ restrict - mj ] ) derived from the partial entropy production method provides a stronger inequality than that derived from the ir formalism ( [ mj - ineq ] ) .
moreover , we have demonstrated that the contraction of the amf model , to which the fluctuation theorem with the mutual information flow is applicable , is equivalent to the simplified mj model .
this allows us the direct comparison of the inequality with the mutual information flow and the mj inequality .
we have also shown the hierarchy of the three second - law - like inequalities for the simplified mj model : the conventional second law ( [ mj-2nd ] ) , the mj inequality ( [ mj - ineq ] ) , and the inequality ( [ restrict - mj ] ) with the partial entropy production .
an important insight is that the more source of dissipation is taken into account , the stronger inequality we obtain . from this perspective
, the reason why inequality ( [ restrict - mj ] ) gives the strongest bound is the following : it specifies the entropy production only associated with the thermal transitions , and it takes into account the dissipation induced by the free relaxation of the bit
. in general , the partial entropy production method would be useful to investigate stronger second - law - like inequalities for a variety of nonequilibrium dynamics .
in addition , we discuss a possible extension of our result to quantum cases .
recently , the refinement of quantum thermodynamics has been intensively studied from several perspectives @xcite .
these studies have revealed novel and fundamental aspects of the second law of thermodynamics .
in addition , both of the mf formalism @xcite and the ir formalism @xcite for quantum systems has already been discussed .
unifying these two formalisms in light of quantum thermodynamics would be an interesting future problem .
the authors thank sosuke ito and kyogo kawaguchi for fruitful discussions and helpful comments .
ns is supported by grant - in - aid for jsps fellows number 26 - 7602 .
ts is supported by jsps kakenhi grant no .
25800217 and no .
22340114 , by kakenhi no .
25103003 fluctuation ` & ` structure " , and by platform for dynamic approaches to living system from mext , japan .
99 szilard l. szilard , z. phys . * 53 * , 840 ( 1929 ) . touchette h. touchette and s. lloyd , phys .
rev . lett . * 84 * , 1156 ( 2000 ) .
allahverdyan a. e. allahverdyan , d. janzing , and g. mahler , j. stat .
p09011 ( 2009 ) .
su2010 t. sagawa and m. ueda , phys .
lett . * 104 * , 090602 ( 2010 ) .
toyabe s. toyabe , t. sagawa , m. ueda , e. muneyuki , and m. sano , nat . phys .
* 6 * , 988 ( 2010 ) .
su2012 t. sagawa and m. ueda , phys .
lett . * 109 * , 180602 ( 2012 ) .
hsp j. horowitz , t. sagawa , and j. m. r. parrondo , phys .
lett . * 111 * , 010602 ( 2013 ) .
hartich d. hartich , a. c. barato , and u. seifert , j. stat .
mech . p02016 ( 2014 ) .
jordan2014 j. m. horowitz and m. esposito , phys . rev .
x * 4 * , 031015 ( 2014 ) .
hs j. m. horowitz and h. sandberg , new j. phys . *
16 * , 125007 ( 2014 ) .
ss n. shiraishi and t. sagawa , phys .
e * 91 * , 012130 ( 2015 ) .
siks n. shiraishi , s. ito , k. kawaguchi , and t. sagawa , new j. phys .
* 17 * , 045012 ( 2015 ) .
bo s. bo , m. d. giudice , and a. celani , j. stat .
p01014 ( 2015 ) .
is s. ito and t. sagawa , nat .
* 6 * , 7498 ( 2015 ) .
hartich2 d. hartich , a. barato , and u. seifert , arxiv:1509.02111 ( 2015 ) .
mj d. mandal and c. jarzynski , proc .
ac . sci . * 109 * , 11641 ( 2012 ) .
refri d. mandal , h. t. quan , and c. jarzynski , phys .
lett . * 111 * , 030602 ( 2013 ) .
deffner s. deffner and c. jarzynski , phys .
x * 3 * , 041003 ( 2013 ) .
strasberg p. strasberg , g. schaller , t. brandes , and c. jarzynski , phys . rev .
e * 90 * , 062107 ( 2014 ) .
merhav n. merhav , j. stat .
p06037 ( 2015 ) .
strasberg2 p. strasberg , j. cerrillo , g. schaller , and t. brandes , phys .
e * 92 * , 042104 ( 2015 ) .
boyd a. b. boyd , d. mandal , and j. p. crutchfield , arxiv:1507.01537 ( 2015 ) .
campisi m. campisi , p. talkner , and p. hnggi , phys .
lett . * 102 * , 210401 ( 2009 ) .
goldstein s. goldstein , t. hara , and h. tasaki , arxiv:1303.6393 ( 2013 ) .
brandao f. brando , m. horodecki , n. ng , j. oppenheim , and s. wehner , proc .
sci . * 112 * , 3275 ( 2015 ) .
esposito m. esposito , m. a. ochoa , and m. galpern , phys .
* 114 * , 080602 ( 2015 ) . | there have been two distinct formalisms of thermodynamics of information : one is the measurement - feedback formalism , which concerns bipartite systems with measurement and feedback processes , and the other is the information reservoir formalism , which considers bit sequences as a thermodynamic fuel . in this paper , we derive a second - law - like inequality by applying the measurement - feedback formalism to information reservoirs , which provides a stronger bound of extractable work than any other known inequality in the same setup . in addition , we demonstrate that the mandal - jarzynski model , which is a prominent model of the information reservoir formalism , is equivalent to a model obtained by the contraction of a bipartite system with autonomous measurement and feedback .
our results provide a unified view on the measurement - feedback and the information - reservoir formalisms . // |
in this article we consider markov processes that are stopped when reaching the boundary of a given set @xmath1 .
these processes appear in a wide range of applications , such as population genetics @xcite , finance @xcite , neuroscience @xcite , physics @xcite and engineering @xcite .
the vast majority of the papers in the literature deal with fully observed stopped processes and assume the parameters of the model are known . in this paper
we address problems when this is not the case .
in particular , bayesian inference for the model parameters is considered , when the stopped process is observed indirectly via data .
we will propose a generic simulation method that can cope with many types of partial observations . to the best of our knowledge
, there is no previous work in this direction .
an exception is @xcite , where maximum likelihood inference for the model parameters is investigated for the fully observed case . in the fully observed case ,
stopped processes have been studied predominantly in the area of rare event simulation . in order to estimate the probability of rare events related to stopped processes , one needs to efficiently sample realisations of a process that starts in a set @xmath0 and terminates in the given rare target set @xmath1 before returning to @xmath0 or getting trapped in some absorbing set .
this is usually achieved using importance sampling ( is ) or multi - level splitting ; see@xcite and the references in those articles for an overview .
recently , sequential monte carlo ( smc ) methods based on both these techniques have been used in @xcite . in @xcite
the authors also prove under mild conditions that smc can achieve same performance as popular competing methods based on traditional splitting .
sequential monte carlo methods can be described as a collection of techniques used to approximate a sequence of distributions whose densities are known point - wise up to a normalizing constant and are of increasing dimension .
smc methods combine importance sampling and resampling to approximate distributions .
the idea is to introduce a sequence of proposal densities and to sequentially simulate a collection of @xmath4 samples , termed particles , in parallel from these proposals .
the success of smc lies in incorporating a resampling operation to control the variance of the importance weights , whose value would otherwise increase exponentially as the target sequence progresses e.g. @xcite . applying smc in the context of
fully observed stopped processes requires using resampling while taking into account how close a sample is to the target set .
that is , it is possible that particles close to @xmath1 are likely to have very small weights , whereas particles closer to the starting set @xmath0 can have very high weights . as a result
, the diversity of particles approximating longer paths before reaching @xmath1 would be depleted by successive resampling steps . in population genetics , for the coalescent model @xcite ,
this has been noted as early as in the discussion of @xcite by the authors of @xcite .
later , in @xcite the authors used ideas from splitting and proposed to perform the resampling step only when each sample of the process reaches intermediate level sets , which define a sequence of nested sets from @xmath0 to @xmath1 .
the same idea appeared in parallel in ( * ? ? ?
* section 12.2 ) , where it was formally interpreted as an interacting particle approximation of appropriate multi - level feynman - kac formulae . naturally , the choice of the intermediate level sets is critical to the performance of such a scheme .
that is , the levels should be set in a `` direction '' towards the set @xmath1 and so that each level can be reached from the previous one with some reasonable probability @xcite .
this is usually achieved heuristically using trial simulation runs .
also more systematic techniques exist : for cases where large deviations can be applied in @xcite the authors use optimal control and in @xcite the level sets are computed adaptively on the fly using the simulated paths of the process .
the contribution of this paper is to address the issue of inferring the parameters of the law of the stopped markov process , when the process itself is a latent process and is only partially observed via some given dataset . in the context of bayesian inference one often needs to sample from the posterior density of the model parameters , which can be very complex .
employing standard markov chain monte carlo ( mcmc ) methods is not feasible , given the difficulty one faces to sample trajectories of the stopped process .
in addition , using smc for sequential parameter inference has been notoriously difficult ; see @xcite . in particular , due to
the successive resampling steps the simulated past of the path of each particle will be very similar to each other .
this has been a long standing bottleneck when static parameters @xmath5 are estimated online using smc methods by augmenting them with the latent state .
these issues have motivated the recently introduced particle markov chain monte carlo ( pmcmc ) @xcite .
essentially , the method constructs a markov chain on an extended state - space in the spirit of @xcite , such that one may apply smc updates for a latent process , i.e. use smc approximations within mcmc . in the context of parameter inference
for stopped process this brings up the possibility of using the multi - level smc methodology as a proposal in mcmc . this idea to the best of our knowledge has not appeared previously in the literature .
the main contributions made in this article are as follows : * when the sequence of level sets is fixed _ a priori _ , the validity of using multi - level smc within pmcmc is verified .
* to enhance performance we propose a flexible scheme where the level sets are adapted to the current parameter sample .
the method is shown to produce unbiased samples from the target posterior density .
in addition , we show both theoretically and via numerical examples how the mixing of the pmcmc algorithm is improved when this adaptive strategy is adopted .
this article is structured as follows : in section [ sec : problem - formulation ] we formulate the problem and present the coalescent as a motivating example . in section [ sec : stopping_time ] we present multi - level smc for stopped processes . in section
[ sec : pmcmc ] we detail a pmcmc algorithm which uses multi - level smc approximations within mcmc .
in addition , specific adaptive strategies for the levels are proposed , which are motivated by some theoretical results that link the convergence rate of the pmcmc algorithm to the properties of multi - level smc approximations . in section [ sec : numerical ] some numerical experiments for the the coalescent are given .
the paper is concluded in section [ sec : summary ] .
the proofs of our theoretical results can be found in the appendix .
the following notations will be used . a measurable space
is written as @xmath6 , with the class of probability measures on @xmath7 written @xmath8 . for @xmath9 , @xmath10
the borel sets are @xmath11 . for a probability measure
@xmath12 we will denote the density with respect to an appropriate @xmath13-finite measure @xmath14 as @xmath15 . the total variation distance between two probability measures @xmath16
is written as @xmath17 . for a vector @xmath18
, the compact notation @xmath19 is used ; if @xmath20 @xmath19 is a null vector . for a vector @xmath21
, @xmath22 is the @xmath23norm .
the convention @xmath24 is adopted .
also , @xmath25 is denoted as @xmath26 and @xmath27 is the indicator of a set @xmath1 .
let @xmath7 be a countable ( possibly infinite dimensional ) state - space , then @xmath28 denotes the class of stochastic matrices which possess a stationary distribution .
in addition , we will denote as @xmath29 the @xmath30-dimensional vector whose @xmath31 element is @xmath32 and is @xmath33 everywhere else . finally , for the discrete collection of integers we will use the notation @xmath34 .
let @xmath5 be a parameter vector on @xmath35 , @xmath36 with an associated prior @xmath37 .
the stopped process @xmath38 is a @xmath39valued discrete - time markov process defined on a probability space @xmath40 , where @xmath41 is a probability measure defined for every @xmath42 such that for every @xmath43 , @xmath44 is @xmath45measurable . for simplicity
will we will assume throughout the paper that the markov process is homogeneous .
the state of the process @xmath38 begins its evolution in a non empty set @xmath0 obeying an initial distribution @xmath46 and a markov transition kernel @xmath47 .
the process is killed once it reaches a non - empty target set @xmath48 such that @xmath49 .
the associated stopping time is defined as @xmath50 where it is assumed that @xmath51 and @xmath52 , where @xmath53 is a collection of positive integer values related to possible stopping times .
in this paper we assume that we have no direct access to the state of the process . instead the evolution of the state of the process generates a random observations vector , which we will denote as @xmath54 .
the realisation of this observations vector is denoted as @xmath55 and assume that it takes value in some non empty set @xmath56 .
we will also assume that there is no restriction on @xmath1 depending on the observed data @xmath55 , but to simplify exposition this will be omitted from the notation . in the context of bayesian inference we are interested in the posterior distribution
: @xmath57 where @xmath58 is the stopping time , @xmath59 is the prior distribution and @xmath60 is the un - normalised complete - data likelihood with the normalising constant of this quantity being : @xmath61 the subscript on @xmath5 will be used throughout to explicitly denote the conditional dependance on the parameter @xmath5 .
given the specific structure of the stopped processes one may write @xmath60 as @xmath62 where @xmath63 is the likelihood of the data given the trajectory of the process . throughout
, it will be assumed that for any @xmath42 , @xmath64 , @xmath60 admits a density @xmath65 with respect to a @xmath66finite measure @xmath67 on @xmath68 and the posterior and prior distributions @xmath69 admit densities @xmath70
respectively both defined with respect to appropriate @xmath66finite dominating measures .
note that ( [ eq : target ] ) is expressed as an inference problem for @xmath71 and not only @xmath5 .
the overall motivation originates from being able to design a mcmc that can sample from @xmath72 , which requires one to write the target ( or an unbiased estimate of it ) up - to a normalizing constant @xcite .
still , our primary interest lies in bayesian inference for the parameter and this can be recovered by the marginals of @xmath72 with respect to @xmath5 . as it will become clear in section [
sec : pmcmc ] the numerical overhead when augmenting the argument of the posterior is necessary and we believe that the marginals with respect to @xmath73 might also be useful by - products .
the framework presented so far is rather abstract , so we introduce the coalescent model as a motivating example . in figure
[ fig : coalgraph ] we present a particular realisation of the coalescent for two genetic types @xmath74 .
the process starts at epoch @xmath75 when the most recent common ancestor ( mrca ) splits into two versions of itself . in this example @xmath1
is chosen to be the mcra and the process continues to evolve by split and mutation moves . at the stopping point ( here @xmath76 )
we observe some data @xmath55 , which corresponds to the number of genes for each genetic type .
( 1,1 ) ( .5,.75)(0,1)0.25 ( 0.25,.75)(1,0).5 ( 0.75,.75)(0,-1)0.75 ( 0.25,.75)(0,-1)0.25 ( 0.125,.5)(1,0)0.25 ( 0.125,.5)(0,-1)0.5 ( 0.375,.5)(0,-1)0.5 ( -0.1,0.0)(0.1,0)12(1,0)0.02 ( -0.1,0.2)(0.1,0)12(1,0)0.02 ( -0.1,0.5)(0.1,0)12(1,0)0.02 ( -0.1,0.5)(0.1,0)12(1,0)0.02 ( -0.1,0.6)(0.1,0)12(1,0)0.02 ( -0.1,0.75)(0.1,0)12(1,0)0.02 ( 0.56,.90)@xmath1 ( 0.76,0.62)@xmath77 ( -0.05,0.21)@xmath77 ( 0.1,-0.05)@xmath78 ( 0.35,-0.05)@xmath1 ( 0.73,-0.05)@xmath78 ( 1.1,0.0)@xmath79 stop @xmath80 ( 1.1,0.2)@xmath81 mutation ( 1.1,0.5)@xmath82 split ( 1.1,0.6)@xmath83 mutation ( 1.1,0.75)@xmath75 split in general we will assume there are @xmath30 different genetic types .
the latent state of the process @xmath84 is composed of the number of genes of each type @xmath85 at epoch @xmath86 of the process and let also @xmath87 .
the process begins by default when the first split occurs , so the markov chain @xmath88 is initialised by the density @xmath89 and is propagated using the following transition density : @xmath90 where @xmath91 is defined in section [ sub : notations ] . here
the first transition type corresponds to individuals changing type and is called mutation , e.g. @xmath77 at @xmath92 in figure [ fig : coalgraph ] .
the second transition is called a split event , e.g. @xmath93 in the example of figure [ fig : coalgraph ] . to avoid any confusion
we stress that in figure [ fig : coalgraph ] we present a particular realisation of the process that is composed by a sequence of alternate split and mutations , but this is not the only possible sequence . for example , the bottom of the tree could have be obtained with @xmath78 being the possible mcra and a sequence of two consecutive splits and a mutation .
the process is stopped at epoch @xmath94 when the number of individuals in the population reaches @xmath95 .
so for the state space we define : @xmath96 and for the initial and terminal sets we have : @xmath97 the data is generated by setting @xmath98 , which corresponds to the counts of genes that have been observed . in the example of figure [ fig : coalgraph ]
this corresponds to @xmath99 .
hence for the complete likelihood we have : @xmath100\label{eq : gamma_y}\ ] ] as expected , the density is only non - zero if at time @xmath94 @xmath101 matches the data @xmath55 exactly .
our objective is to infer the genetic parameters @xmath102 , where @xmath103 and @xmath104 and hence the parameter space can be written as @xmath105 . to facilitate monte carlo inference
, one can reverse the time parameter and simulate backward from the data .
this is now detailed in the context of importance sampling following the approach in @xcite . to sample realisations of the process for a given @xmath42 , importance sampling is adopted but with time reversed .
first we introduce a time reversed markov kernel @xmath106 with density @xmath107 .
this is used as an importance sampling proposal where sampling is performed backwards in time and the weighting forward in time .
we initialise using the data and simulate the coalescent tree backward in time until two individuals remain of the same type .
this procedure ensures that the data is hit when the tree is considered forward in time .
the process defined backward in time can be interpreted as a stopped markov process with the definitions of the initial and terminal sets appropriately modified . for convenience
we will consider the reverse event sequence of the previous section , i.e we posed the problem backwards in time with the reverse index being @xmath108 .
the proposal density for the full path starting from the bottom of the tree and stopping at its root can be written as @xmath109 with reference to we have @xmath110 then the marginal likelihood can be obtained @xmath111 in @xcite the authors derive an optimal proposal @xmath112 with respect to the variance of the marginal likelihood estimator .
for the sake of brevity we omit any further details . in the current setup
where there is only mutation and coalescences , the stopped - process can be integrated out @xcite , but this is not typically possible in more complex scenarios . a more complicated problem , including migration , is presented in section [ sec : migration ] .
finally , we remark that the relevance of the marginal likelihood above will become clear later in section [ sec : pmcmc ] as a crucial element in numerical algorithms for inferring @xmath5 .
in this section we shall briefly introduce generic smc without extensive details .
we refer the reader for a more detailed description to @xcite . to ease exposition ,
when presenting generic smc , we shall drop the dependence upon parameter @xmath5 .
smc algorithms are designed to simulate from a sequence of probability distributions @xmath113 defined on state space of increasing dimension , namely @xmath114 .
each distribution in the sequence is assumed to possess densities with respect to a common dominating measure : @xmath115 with each un - normalised density being @xmath116 and the normalizing constant being @xmath117 .
we will assume throughout the article that there are natural choices for @xmath118 and that we can evaluate each @xmath119 point - wise .
in addition , we do not require knowledge of @xmath120 smc algorithms approximate @xmath121 recursively by propagating a collection of properly weighted samples , called particles , using a combination of importance sampling and resampling steps . for the importance sampling part of the algorithm , at each step @xmath122 of the algorithm , we will use general proposal kernels @xmath123 with densities @xmath124 , which possess normalizing constants that do not depend on the simulated paths .
a typical smc algorithm is given in algorithm [ alg : generic - smc - algorithm ] and we obtain the following smc approximations for @xmath125 @xmath126 and for the normalizing constant @xmath117 : @xmath127 initialisation , @xmath128 : for @xmath129 1 .
sample @xmath130 .
2 . compute weights @xmath131 for @xmath132 , for @xmath129 , 1 . resampling : sample index @xmath133 , where @xmath134 .
2 . sample @xmath135 and set @xmath136 .
3 . compute weights @xmath137 in this paper we will use @xmath138 to be the multinomial distribution .
then the resampled index of the ancestor of particle @xmath85 at time @xmath122 , namely @xmath139 , is also a random variable with value chosen with probability @xmath140 . for each time @xmath122
, we will denote the complete collection of ancestors obtained from the resampling step as @xmath141 and the randomly simulated values of the state obtained during sampling ( step 2 for @xmath142 ) as @xmath143 .
we will also denote @xmath144 the concatenated vector of all these variables obtained during the simulations from time @xmath145 .
note @xmath146 the is a vector containing all @xmath147 simulated states and should not be confused with the particle sample of the path @xmath148 .
furthermore , the joint density of all the sampled particles and the resampled indices is @xmath149 the complete ancestral genealogy at each time can always traced back by defining an ancestry sequence @xmath150 for every @xmath151 and @xmath142 . in particular , we set the elements of @xmath150 using the backward recursion @xmath152 where @xmath153 . in this context
one can view smc approximations as random probability measures induced by the imputed random genealogy @xmath154 and all the possible simulated state sequences that can be obtained using @xmath155 .
this interpretation of smc approximations was introduced in @xcite and will be later used together with @xmath156 for establishing the complex extended target distribution of pmcmc . for different classes of problems one can find a variety of enhanced smc algorithms
; see e.g. @xcite . in the context of stopped processes ,
a multi - level smc implementation was proposed in @xcite and the approach was illustrated for the coalescent model of section [ sec : coal_model ] .
we consider a modified approach along the lines of section 12.2 of @xcite which seems better suited for general stopped processes and can provably yield estimators of much lower variance relative to vanilla smc .
introduce an arbitrary sequence of @xmath157nested sets @xmath158 with the corresponding stopping times denoted as @xmath159 note that the markov property of @xmath160 implies @xmath161 .
the implementation of multi - level smc differs from the generic algorithm of section [ sec : smc_algo ] in that between successive resampling steps one proceeds by propagating in parallel trajectories of @xmath162 until the set @xmath163 is reached for each @xmath164 . for a given @xmath164
the path @xmath162 is `` frozen '' once @xmath165 , until the remaining particles reach @xmath163 and then a resampling step is performed .
more formally denote for @xmath128 @xmath166 where @xmath167 is a realisation for the stopping time @xmath168 and similarly for @xmath169 we have @xmath170 multi - level smc is a smc algorithm which ultimately targets a sequence of distributions @xmath171 each defined on a space @xmath172 where @xmath173 , @xmath174 and @xmath175 are finite collections of positive integer values related to the stopping times @xmath176 respectively . in the spirit of generic smc
define intermediate target densities @xmath177 w.r.t to an appropriate @xmath13-finite dominating measure @xmath178 .
we will assume there exists a natural sequence of densities @xmath179 obeying the restriction @xmath180 so that the last target density @xmath181 coincides with @xmath182 in ( [ eq : unnormalised_likeli ] ) .
note that we define a sequence of @xmath183 target densities , but this time the dimension of @xmath119 compared to @xmath184 grows with a random increment of @xmath185 .
in addition , @xmath181 should clearly depend on the value of @xmath5 , but this suppressed in the notation .
@xmath186the following proposition is a direct consequence of the markov property : [ prop : markov_level]assume @xmath51 . then the stochastic sequence defined @xmath187 forms a markov chain taking values in @xmath188 defined . in addition , for any bounded measurable function @xmath189 , then @xmath190
. the proof can be found in ( * ? ? ?
* proposition 12.2.2 , page 438 ) , ( * ? ? ? * proposition 12.2.4 , page 444 ) and the second part is due to @xmath191 .
we will present multi - level smc based as a particular implementation of the generic smc algorithm .
firstly we replace @xmath192 with @xmath193 respectively .
contrary to the presentation of algorithm [ alg : generic - smc - algorithm ] for multi - level smc we will use a homogeneous markov importance sampling kernel @xmath194 , where @xmath195 , @xmath196 by convention and @xmath112 is the corresponding density w.r.t .
@xmath14 . to compute the importance sampling weights of step 3 for @xmath142 in algorithm [ alg : generic - smc - algorithm ] we use instead :
@xmath197 and for step 2 at @xmath128 : @xmath198 to simplify notation from
herein we write @xmath199 and given @xmath183 , for any @xmath169 we have @xmath200 where again we have suppressed the @xmath5-dependance of @xmath201 in the notation .
we present the multi - level smc algorithm in algorithm [ alg : multilevel - smc - algorithm ] .
note here we include a procedure whereby at each stage @xmath122 , particles that do not reach @xmath163 before time @xmath202 are rejected by assigning them a zero weight , whereas before it was hinted that resampling is performed when all particles reach @xmath163 .
similar to , it is clear that the joint probability density of all the random variables used to implement a multi - level smc algorithm with multinomial resampling is given by : @xmath203 where @xmath204 is defined similarly to @xmath146 . finally , recall by construction @xmath205 so the approximation of the normalizing constant of @xmath60 for a fixed @xmath5 is @xmath206 initialisation , @xmath128 : for @xmath129 1 . for @xmath207 :
1 . sample @xmath208 .
2 . if @xmath209 set @xmath210 , @xmath211 and go to step 2 .
compute weights @xmath212 for @xmath132 , for @xmath129 , 1 . resampling : sample index @xmath133 , where @xmath134 .
2 . for @xmath213
: 1 . sample @xmath208 .
2 . if @xmath214 set @xmath215 , @xmath216 and go to step 3 . 3 . set @xmath217 .
4 . compute weights @xmath218 we will begin by showing how the levels can be set for the coalescent example of section [ sec : coal_model ]
. we will proceed in the spirit of section [ sec : like_comp ] and consider the backward process so that the `` time '' indexing is set to start from the bottom of the tree towards the root .
we introduce a a collection of integers @xmath219 and define @xmath220 clearly we have @xmath221 , @xmath222 and @xmath223 .
one can also write the sequence of target densities for the multi - level setting as : @xmath224 the major design problem that remains in general is that given _ any _ candidates for @xmath225 , how to set the spacing ( in some sense ) of the @xmath226 and how many levels are needed so that good smc algorithms can be constructed .
that is , if the @xmath226 are far apart , then one can expect that weights will degenerate very quickly and if the @xmath226 are too close that the algorithm will resample too often and hence lead to poor estimates .
for instance , in the context of the coalescent example of section [ sec : coal_model ] , if one uses the above construction for @xmath226 the importance weight at the @xmath122-th resampling time is @xmath227 now , in general for any @xmath228 and @xmath183 it is hard to know beforehand how much better ( or not ) the resulting multi - level algorithm will perform relative to a vanilla smc algorithm .
whilst @xcite show empirically that in most cases one should expect a considerable improvement , there @xmath5 is considered to be fixed . in this case
one could design the levels sensibly using offline heuristics or more advanced systematic methods using optimal control @xcite or adaptive simulation @xcite , e.g. by setting the next level using the median of a pre - specified rank of the particle sample .
what we aim to establish in the next section is that when @xmath5 is varies as in the context of mcmc algorithms , one can both construct pmcmc algorithms based on multi - level smc and more importantly easily design for each @xmath5 different sequences for @xmath226 based on similar ideas .
particle markov chain monte carlo ( pmcmc ) methods are mcmc algorithms , which use all the random variables generated by smc approximations as proposals . as in standard
mcmc the idea is to run an ergodic markov chain to obtain samples from the distribution of interest .
the difference lies that in order use the simulated variables from smc , one defines a complex invariant distribution for the mcmc on an extended state space .
this extended target is such that a marginal of this invariant distribution is the one of interest .
this section aims on providing insight to the following questions : 1 .
is it valid in general to use multi - level smc within pmcmc ? 2 .
given that it is , how can we use the levels to improve the mixing of pmcmc ?
the answer to the first question seems rather obvious , so we will provide some standard but rather strong conditions for which multi - level pmcmc is valid . for the second question we will propose an extension to pmmh that adapts the level sets used to @xmath5 at every iteration of pmcmc .
@xcite introduces three different and generic pmcmc algorithms : particle independent metropolis hastings algorithm ( pimh ) , particle marginal metropolis hastings ( pmmh ) and particle gibbs samplers . in the remainder of the paper we will only focus on the first two of these . 1 .
sample @xmath229 from using the multi - level implementation of algorithm [ alg : generic - smc - algorithm ] detailed in section and compute @xmath230 .
sample @xmath231 .
2 . set @xmath232 and @xmath233 3 . for @xmath234
: 1 . propose a new @xmath235 and @xmath236 as in step 1 and compute @xmath237 2 .
accept this as the new state of the chain with probability @xmath238 if we accept , set @xmath239 and @xmath240 .
otherwise reject , @xmath241 and @xmath242 .
we will begin by presenting the simplest generic algorithm found in @xcite , namely the particle independent metropolis hastings algorithm ( pimh ) . in this case @xmath5 and @xmath183 are fixed and pimh is designed to sample from the pre - specified target distribution @xmath243 also considered in section [ sec : smc_algo ] . although pimh is not useful for parameter inference it is included for pedagogic purposes .
one must bear in mind that pimh is the most basic of all pmcmc algorithms .
as such it is easier to analyse but still can provide useful intuition that can be used later in the context of pmmh and varying @xmath5 .
pimh is presented in algorithm [ alg : pimh ] .
it can be shown , using similar arguments to @xcite , that the invariant density of the markov kernel above is exactly ( see the proof of proposition [ prop : stop_within_mcmc ] ) @xmath244 where @xmath245 is as in and as before we have @xmath246 and @xmath247 for every @xmath248 .
note that @xmath249 admits the target density of interest , @xmath250 as the marginal , when @xmath251 and @xmath252 are integrated out .
we commence by briefly investigating some convergence properties of pimh with multi - level smc . even though the scope of pimh is not parameter inference
, one can use insight on what properties are desired by multi - level smc for pimh when designing other pmcmc algorithms used for parameter inference .
we begin with posing the following mixing and regularity assumption : [ assump : a1 ] for every @xmath42 and @xmath253 there exist a @xmath254 such that for every @xmath255 : @xmath256 there exist a @xmath257 such that for @xmath222 and every @xmath258 : @xmath259 the stopping times are finite , that is for @xmath222 there exist a @xmath260 such that @xmath261 assumption ( a[assump : a1 ] ) is rather strong , but are often used in the analysis of these kind of algorithms @xcite because they simplify the proofs to a large extent .
recall that @xmath42 and @xmath253 are fixed .
we proceed by stating the following proposition : [ prop : conv_rate ] assume ( a[assump : a1 ] ) . then for @xmath262 algorithm [ alg : pimh ] generates a sequence @xmath263 that for any @xmath264 , @xmath265 , @xmath42 satisfies : @xmath266 the proof can be found in the appendix .
the following remarks are generalised and do not always hold , but provide some intuition for the ideas that follow .
the result shows intrinsically that as the supremum of the sum of the stopping times with respect to @xmath267 gets smaller , so does the convergence rate increase .
this can be also linked to the variance of the estimator of @xmath230 , which is well known to increase linearly with @xmath183 ( * ? ? ?
* theorem 12.2.2 , pages 451 - 453 ) .
shorter stopping times will typically yield lower variance and hence better mcmc convergence properties . on the other hand
often @xmath268 will be larger for longer @xmath183 and longer stopping times ( proposition [ prop : conv_rate ] is derived for a fixed @xmath183 ) .
in addition , sampling a stopped process is easier using a higher number of levels . in summary ,
the tradeoff is that although it is more convenient to use more auxiliary variables for simulating the process , these will slow down the mixing of pmcmc . in practice one balances this by trying to use a moderate number of levels for which most the particles to reach @xmath1 .
this tradeoff serves as a motivation for developing flexible schemes to vary @xmath269 with @xmath5 in the pmcmc algorithm presented later in section [ sec : adaptive_strat ] . in the remainder of this section
we will focus on using a multi - level smc implementation within a pmmh algorithm .
given the commentary in section [ sec : stopping_time_smc ] and our interest in drawing inference on @xmath42 , it seems that using multi - level smc within pmcmc should be highly beneficial .
recall can be expressed in terms of densities as : @xmath270 and let the marginal density given by @xmath271 for the time being we will consider the case when @xmath183 is fixed . in the context of our stopped markov process , we propose a pmmh algorithm targeting @xmath272 in algorithm [ fig : stop_within_mcmc ] .
1 . sample @xmath273 .
given @xmath274 sample @xmath275 using multi - level smc and compute @xmath276 .
sample @xmath231 .
2 . set @xmath277 and @xmath278 .
3 . for @xmath234 : 1 . sample @xmath279 ; given @xmath280 propose a new @xmath281 and @xmath236 as in step 1 and compute @xmath282 .
2 . accept this as the new state of the chain with probability @xmath283 if we accept , set @xmath284 and @xmath285 .
otherwise reject , @xmath241 and @xmath286 .
we will establish the invariant density and convergence of this algorithm , under the following assumption : [ assump : a5 - 6 ] for any @xmath42 and @xmath253 we define the following sets for @xmath145 : @xmath287 and @xmath288 . for any @xmath42 we have that @xmath289 .
in addition the ideal metropolis hastings targeting @xmath290 using proposal density @xmath291 is irreducible and aperiodic .
this assumption contains assumptions 5 and 6 of @xcite modified to our problem with a simple change of notations .
we proceed with the following result : [ prop : stop_within_mcmc ] assume ( a[assump : a5 - 6 ] ) ; then for any @xmath262 : 1 .
the invariant density of the procedure described in algorithm [ fig : stop_within_mcmc ] , is on the space @xmath292 and has the representation @xmath293 where @xmath294 is as in and @xmath295 as in .
in addition , admits @xmath290 as a marginal .
algorithm [ fig : stop_within_mcmc ] generates a sequence @xmath296 such that @xmath297 where @xmath72 is as in .
the proof of the result is in the appendix .
the result is based on theorem 4 of @xcite .
note that algorithm [ fig : stop_within_mcmc ] presented in a generic form of a `` vanilla '' pmmh algorithm , so it can be enhanced using various strategies .
for example , it is possible to add block updating of the latent variables or backward simulation in the context of a particle gibbs version @xcite . in the next section ,
we propose a flexible scheme that allows to set a different number of levels after a new @xmath298 @xmath186is proposed .
the remaining design issue for pmmh is how to tune multi - level smc by choosing @xmath183 and @xmath267 .
whilst , for a fixed @xmath42 , one could solve the problem with preliminary runs , when @xmath5 varies this is not an option .
in general the value of @xmath5 should dictate how small or large @xmath183 should be to facilitate an efficient smc algorithm .
hence , to obtain a more accurate estimate of the marginal likelihood and thus an efficient mcmc algorithm , we need to consider adaptive strategies to propose randomly a different number of levels @xmath183 and levels sequence @xmath267 for each @xmath299 sampled at every pmmh iteration @xmath85 . to ease exposition
we will assume that @xmath269 can be expressed as functions of an arbitrary auxiliary parameter @xmath300 .
given @xmath299 is a random variable , the main questions we wish to address is how to perform such an adaptive strategy consistently .
an important point , is the fact that since we are interested in parameter inference , it is required that the marginal of the pmmh invariant density is @xmath290
. this can be ensured ( see proposition [ prop : adap_stop ] ) by introducing at each pmmh iteration , the parameters that form the level sets @xmath301 as an auxiliary process , which given @xmath299 is conditionally independent of @xmath302 . this way we define an extended target for the mcmc algorithm , which includes @xmath183 and @xmath267 in the target variables .
it should be noted that this scheme is explicitly different from proposition 1 of @xcite , where the mcmc transition kernel at iteration @xmath85 is dependent upon an auxiliary process .
here one just augments the target space with more auxiliary variables .
1 . sample @xmath273 . given @xmath274 : sample @xmath303 from @xmath304 , then @xmath305 using multi - level smc and compute @xmath276 . sample @xmath231 .
2 . set @xmath306 and @xmath278 .
3 . for @xmath234 : 1 .
sample @xmath279 ; sample @xmath307 from @xmath308 and @xmath309 @xmath236 as in step 1 and compute @xmath282 .
2 . accept this as the new state of the chain with probability @xmath283 if we accept , set @xmath310 and @xmath285 .
otherwise reject , @xmath241 and @xmath286 .
consider now that it is possible at every pmmh iteration @xmath85 to simulate the auxiliary process @xmath300 defined upon an abstract state - space @xmath311 .
let this with associated random variable @xmath300 , be distributed according to @xmath312 , which is assumed to possess a density with respect to a. @xmath66finite measure @xmath313 written as @xmath314 .
as hinted by the notation @xmath312 should depend on @xmath5 and @xmath300 is meant be used to determine the sequence of levels @xmath267 for each @xmath299 in pmmh .
this auxiliary variable will induce for every @xmath42 : * a random number of level sets @xmath315 . *
a sequence of level sets @xmath316 with @xmath317 .
we will assume that for any @xmath42 , proposition [ prop : markov_level ] and will hold @xmath318almost everywhere , where this time @xmath183 should be replaced by @xmath319 .
this implies that for every @xmath42 we have : @xmath320 where the expression holds @xmath318 almost everywhere . in algorithm
[ fig : stop_within_mcmc1 ] we propose a pmmh algorithm , which at each step @xmath85 uses @xmath299 to adapt the levels @xmath321 . for algorithm
[ fig : stop_within_mcmc1 ] we present the following proposition that verifies varying the level sets in this way is theoretically valid : [ prop : adap_stop ] assume ( a[assump : a5 - 6 ] ) and hold .
then , for any @xmath262 : 1 .
the invariant density of the procedure in algorithm [ fig : stop_within_mcmc1 ] is defined on the space @xmath322 and has the representation @xmath323 where @xmath294 is as in and @xmath295 is as in .
in addition , admits @xmath290 as a marginal .
the generated sequence @xmath324 satisfies @xmath325 where @xmath72 is as in .
the proof is contained in the appendix .
we are essentially using an auxiliary framework similar to @xcite . as in
we included @xmath73 in the target posterior , when we were primarily interested in @xmath5 , this time we augment the target posterior with @xmath300 and the smc variables @xmath326 , which is a consequence of using pmcmc .
the disadvantage is that as the space of the posterior increases it is expected that the mixing of the algorithm will be slower .
this could be improved if we have opted @xmath73 and @xmath300 to be dependent on each other given @xmath5 , but this would need additional assumptions for the structure of @xmath60 .
in addition , in many applications the parameters @xmath300 that determine @xmath267 appear naturally and @xmath300 often is low dimensional .
also , in most applications it might seem easier to find intuition on how to construct and tune @xmath312 than computing the level sets directly from @xmath5 .
for example , for the coalescent model of section [ sec : coal_model ] with the mutation matrix @xmath327 is fixed , one can envisage for a larger value of @xmath328 coalescent events are less likely and more level sets closer together are needed compared to smaller values of @xmath328 .
we will illustrate the performance of pmmh using numerical examples on two models from population genetics . the first one deals with the coalescent model of section [ sec : coal_model ] when a low dimensional dataset is observed .
this is meant as an academic / toy example suitable for comparing different pmmh implementations .
the second example is a more realistic application and deals with a coalescent model that allows migration of individual genetic types from one sub - group to another @xcite . in both cases
we will illustrate the performance of pmmh implemented with a simple intuitive strategy for adapting the level sets .
we will use a known stochastic matrix @xmath327 with all entries equal to @xmath329 . in this example
@xmath330 with and the dataset is @xmath331 .
the parameter - space is set as @xmath332 $ ] and a uniform prior will be used .
for @xmath106 we will use the optimal proposal distributions provided by @xcite .
the pmmh proposal @xmath333 in algorithm [ fig : stop_within_mcmc1 ] is a log normal random walk , i.e. we use @xmath334 with @xmath335 .
we will compare pmmh when implemented with a simple adaptive scheme for @xmath183 and when @xmath183 is fixed . in the latter case we set @xmath336 .
when an adaptive strategy is employed we will sample each time @xmath183 directly using a multinomial distribution defined on @xmath337 with weights proportional to @xmath338 . in both cases given
@xmath183 we place the levels almost equally spaced apart .
the adaptive and normal versions were run with @xmath339 for @xmath340 iterations . in each case
the algorithm took approximately @xmath341 , @xmath342 , @xmath343 hours to complete when implemented in matlab and run on a linux workstation using a intel core 2 quad q9550 cpu at 2.83 ghz .
the results are shown in figure [ fig : plots_noadap ] and [ fig : plots_adap ] .
we observed that when we varied the number of levels , this allowed the sampler to traverse through a bigger part of the state space compared to when a fixed number of levels is used . as a result
the estimated pdf of the adaptive case manages to include a second mode that is not seen in the non adaptive case . in the fixed levels case
we see a clear improvement with increasing @xmath344 , although the difference in the mixing between @xmath345 and @xmath346 is marginal . in the adaptive case
the sampler performed well even with lower values of @xmath344 .
level sets is used .
left : estimated pdf of @xmath328 for @xmath339 .
centre : the trace plot for @xmath345 .
right : autocorrelation plots for @xmath339 .
the average acceptance ratio was @xmath347 , @xmath348 and @xmath349 respectively.,title="fig:",scaledwidth=25.0% ] level sets is used .
left : estimated pdf of @xmath328 for @xmath339 .
centre : the trace plot for @xmath345 .
right : autocorrelation plots for @xmath339 .
the average acceptance ratio was @xmath347 , @xmath348 and @xmath349 respectively.,title="fig:",scaledwidth=25.0% ] level sets is used .
left : estimated pdf of @xmath328 for @xmath339 .
centre : the trace plot for @xmath345 .
right : autocorrelation plots for @xmath339 .
the average acceptance ratio was @xmath347 , @xmath348 and @xmath349 respectively.,title="fig:",scaledwidth=25.0% ] . far left : estimated pdf of @xmath328 for @xmath339 .
central left : the trace plot for @xmath345 .
central right : histogram of number of levels in the posterior for @xmath345 .
far right : autocorrelation function plots for @xmath339 .
the average acceptance ratio was @xmath349 , @xmath350 and @xmath351 respectively.,title="fig:",scaledwidth=25.0% ] . far left : estimated pdf of @xmath328 for @xmath339 .
central left : the trace plot for @xmath345 .
central right : histogram of number of levels in the posterior for @xmath345 .
far right : autocorrelation function plots for @xmath339 .
the average acceptance ratio was @xmath349 , @xmath350 and @xmath351 respectively.,title="fig:",scaledwidth=25.0% ] . far left : estimated pdf of @xmath328 for @xmath339 .
central left : the trace plot for @xmath345 .
central right : histogram of number of levels in the posterior for @xmath345 .
far right : autocorrelation function plots for @xmath339 .
the average acceptance ratio was @xmath349 , @xmath350 and @xmath351 respectively.,title="fig:",scaledwidth=25.0% ] . far left : estimated pdf of @xmath328 for @xmath339 .
central left : the trace plot for @xmath345 .
central right : histogram of number of levels in the posterior for @xmath345 .
far right : autocorrelation function plots for @xmath339 .
the average acceptance ratio was @xmath349 , @xmath350 and @xmath351 respectively.,title="fig:",scaledwidth=25.0% ] the model is similar to the one as described in section [ sec : coal_model ] .
the major difference is that this time genetic types are of classified into sub - groups within which most activity happens .
in addition , individuals are allowed to migrate from one group to another .
we commence with a brief description of the model and refer the interested reader to @xcite for more details . as in section
[ sec : coal_model ] we will consider the process forward in time .
let @xmath352 be the number of groups and the state at time @xmath86 be composed as the concatenation of @xmath352 groups of different genetic types as : @xmath353 the process under - goes split , mutation and migration transitions as follows : @xmath354 where @xmath355 with @xmath356 and @xmath357 is a vector with a zero in every element except the @xmath358 -th one . similarly to the simpler model of section [ sec : coal_model ]
the transition probabilities are parameterised by the mutation parameter @xmath328 , mutation matrix @xmath327 and a migration matrix @xmath359 .
the latter is a symmetric matrix with zero values on the diagonal and positive values on the off - diagonals .
finally the data is generated when at time @xmath94 the number of individuals in the population reaches @xmath95 , and @xmath360 .
as for the model described in section [ sec : coal_model ] one can reverse time and employ an backward sampling forward weighting importance sampling method ; see @xcite for the particular implementation details .
in our example we generated data with @xmath361 , @xmath362 and @xmath363 .
this is quite a challenging set - up . as in the previous example we set the mutation matrix @xmath327 to be known and uniform and we will concentrate on inferring the @xmath364 .
independent gamma priors with shape and scale parameters equal to @xmath32 were adopted for each of the parameters .
we implemented pmmh using @xmath339 and a simple adaptive scheme for @xmath183 .
we allow @xmath365 and use approximately equal spacing between the levels .
we choose each @xmath183 with probability proportional to @xmath366 the proposals for the parameters were gaussian random - walks on the log - scale .
the algorithm was implemented in c / c++ was run for @xmath340 iterations , which took approximately 3 , 6 and 12 hours to complete . whilst the run - time is quite long it can be improved by at least one order of magnitude if the smc is implemented on graphical processing units ( gpu ) as in @xcite . for the dataset plotted in figure [ fig : plots2 ] ( left )
the results are plotted in figures [ fig : plots2 ] ( right ) and [ fig : plots1 ] .
the auto - correlation and trace plots indicate that the sampler mixes reasonably well for every @xmath344 .
these results in this example are encouraging as to the best of our knowledge bayesian inference has not been attempted for this class of problems .
we expect that practitioners with insight in the field of population genetics can come with more sophisticated mcmc proposals or adaptive schemes for the level sets , so that the methodology can be extended to realistic applications . in the resulting posterior for @xmath345.,title="fig:",scaledwidth=25.0% ] in the resulting posterior for @xmath345.,title="fig:",scaledwidth=25.0% ] .
top row : estimated pdfs for @xmath328 , @xmath367 , @xmath368 , @xmath369 ( from left to right ) .
middle : trace plots for @xmath345 .
bottom : autocorrelation function plots .
the acceptance rate was @xmath370 respectively.,title="fig:",scaledwidth=25.0% ] .
top row : estimated pdfs for @xmath328 , @xmath367 , @xmath368 , @xmath369 ( from left to right ) .
middle : trace plots for @xmath345 .
bottom : autocorrelation function plots .
the acceptance rate was @xmath370 respectively.,title="fig:",scaledwidth=25.0% ] .
top row : estimated pdfs for @xmath328 , @xmath367 , @xmath368 , @xmath369 ( from left to right ) .
middle : trace plots for @xmath345 .
bottom : autocorrelation function plots .
the acceptance rate was @xmath370 respectively.,title="fig:",scaledwidth=25.0% ] .
top row : estimated pdfs for @xmath328 , @xmath367 , @xmath368 , @xmath369 ( from left to right ) .
middle : trace plots for @xmath345 .
bottom : autocorrelation function plots .
the acceptance rate was @xmath370 respectively.,title="fig:",scaledwidth=25.0% ] + .
top row : estimated pdfs for @xmath328 , @xmath367 , @xmath368 , @xmath369 ( from left to right ) .
middle : trace plots for @xmath345 .
bottom : autocorrelation function plots .
the acceptance rate was @xmath370 respectively.,title="fig:",scaledwidth=25.0% ] .
top row : estimated pdfs for @xmath328 , @xmath367 , @xmath368 , @xmath369 ( from left to right ) .
middle : trace plots for @xmath345 .
bottom : autocorrelation function plots .
the acceptance rate was @xmath370 respectively.,title="fig:",scaledwidth=25.0% ] .
top row : estimated pdfs for @xmath328 , @xmath367 , @xmath368 , @xmath369 ( from left to right ) .
middle : trace plots for @xmath345 .
bottom : autocorrelation function plots .
the acceptance rate was @xmath370 respectively.,title="fig:",scaledwidth=25.0% ] .
top row : estimated pdfs for @xmath328 , @xmath367 , @xmath368 , @xmath369 ( from left to right ) .
middle : trace plots for @xmath345 .
bottom : autocorrelation function plots .
the acceptance rate was @xmath370 respectively.,title="fig:",scaledwidth=25.0% ] + .
top row : estimated pdfs for @xmath328 , @xmath367 , @xmath368 , @xmath369 ( from left to right ) .
middle : trace plots for @xmath345 .
bottom : autocorrelation function plots .
the acceptance rate was @xmath370 respectively.,title="fig:",scaledwidth=25.0% ] .
top row : estimated pdfs for @xmath328 , @xmath367 , @xmath368 , @xmath369 ( from left to right ) .
middle : trace plots for @xmath345 .
bottom : autocorrelation function plots .
the acceptance rate was @xmath370 respectively.,title="fig:",scaledwidth=25.0% ] .
top row : estimated pdfs for @xmath328 , @xmath367 , @xmath368 , @xmath369 ( from left to right ) .
middle : trace plots for @xmath345 .
bottom : autocorrelation function plots .
the acceptance rate was @xmath370 respectively.,title="fig:",scaledwidth=25.0% ] .
top row : estimated pdfs for @xmath328 , @xmath367 , @xmath368 , @xmath369 ( from left to right ) .
middle : trace plots for @xmath345 .
bottom : autocorrelation function plots .
the acceptance rate was @xmath370 respectively.,title="fig:",scaledwidth=25.0% ]
in this article we have presented a multi - level pmcmc algorithm which allows one to perform bayesian inference for the parameters of a latent stopped processes . in terms of methodology the main novelty of the approach
is that uses auxiliary variables to adaptively compute the level sets with @xmath5 .
the general structure of this auxiliary variable allows it to incorporate the use of independent smc runs with less particles to set the levels . in the numerical examples we demonstrated that the addition auxiliary variables slow down the convergence of pmcmc , but this seemed a reasonable compromise in terms of performance compared when
fixed number of level sets were used .
the proposed algorithm requires considerable amount of computation , but to the authors best knowledge for such problems there seems to be a lack of alternative approaches .
also , recent developments gpu hardware can be adopted to speed up the computations even by orders of magnitude as in @xcite .
there are several extensions to the work here , which may be considered .
firstly , the scheme that is used to adapt the level sets relies mainly on intuition .
we found simple adaptive implementations to work well in practice . in the rare events
literature one may find more systematic techniques to design the level sets , based upon optimal control @xcite or simulation @xcite .
although these methods are not examined here , they can be characterised using alternative auxiliary variables similar to the ones in proposition [ prop : adap_stop ] , so the auxiliary variable framework we use is quite generic .
in addition , we emphasise that within a pmcmc framework one may also include multi - level splitting algorithms instead of smc , which might appeal practitioners familiar with multi - level splitting .
secondly , one could seek to use these ideas within a smc sampler framework of @xcite as done in @xcite .
as noted in the latter article , a sequential formulation can improve the sampling scheme , sometimes at a computational complexity that is the same as the original pmcmc algorithm .
in addition , this article focuses on the pmmh algorithm , so clearly extensions using particle gibbs and block updates might prove valuable for many applications . finally , from a modelling perspective
, it may be of interest to apply our methodology in the context of hidden markov models . in this context
, one has @xmath371 with @xmath372 being the conditional likelihood of the observations
. it would be important to understand , given a range of real applications , the feasibility of statistical inference , combined with the development of our methodology .
an investigation of the effectiveness of such a scheme when applied to queuing networks is currently underway .
we thank arnaud doucet and maria de iorio for many conversations on this work .
the first author was supported by a ministry of education grant .
the second author was kindly supported by the epsrc programme grant on control for energy and sustainability ep / g066477/1 .
much of this work was completed when the first author was employed by the department of mathematics , imperial college london .
[ proof of proposition [ prop : conv_rate ] ] the result is a straight forward application of theorem 6 of @xcite which adapted to our notation states : @xmath373\wedge\mathbb{e}_{\psi_{\theta}}\left[1\wedge\frac{\hat{z}_{p}(\xi)}{\hat{z}_{p}(\xi)}\bigg|\xi\right]\right)\bigg)^{i}\bigg],\ ] ] where the conditional expectation is the expectation w.r.t .
the smc algorithm ( i.e. @xmath374 ) and the outer expectation is w.r.t .
the pimh target ( i.e. @xmath375 ) .
we also denote the estimate of the normalizing constant as @xmath376 with @xmath377 denoting which random variables generate the estimate .
now , clearly via ( a[assump : a1 ] ) @xmath378^{\tau_{n}+\tau_{n-1}}\ ] ] with the convention that @xmath379 .
thus , it follows that @xmath380^{\bar{\tau}_{n}+\bar{\tau}_{n-1}}\leq\bigg[\frac{1}{\rho\varphi}\bigg]^{2\sum_{n=1}^{p}\bar{\tau}_{n}}\ ] ] and we obtain : @xmath381 note that by assumption @xmath382 and thus we have @xmath383\bigg)^{i}\ ] ] given ( * ? ? ?
* theorem 7.4.2 , equation ( 7.17 ) , page 239 ) and the fact that @xmath60 is defined to be strictly positive in ( a[assump : a1 ] ) we have that the smc approximation @xmath376 is an unbiased estimate of the normalizing constant @xmath384 @xmath385=z_{p},\label{eq : smc_unbiased}\ ] ] and we can easily conclude .
[ proof of proposition [ prop : stop_within_mcmc ] ] the proof of parts 1 . and 2 .
follows the line of arguments used in theorem 4 of @xcite , which we will adapt to our set - up .
the main difference lies in the multi - level construction and second statement regarding the marginal of @xmath386 .
for the validity of the multi - level set - up we will rely on proposition [ prop : markov_level ] .
suppose we design a metropolis hastings kernel with invariant density @xmath386 and use a proposal @xmath387 .
then @xmath388 where we denote the normalising constant of the posterior in as : @xmath389 therefore the metropolis - hastings procedure to sample from @xmath390 will be as in algorithm [ fig : stop_within_mcmc ] . alternatively using similar arguments one we may write @xmath391 summing over @xmath251 and using the unbiased property of the smc algorithm in equation it follows that @xmath392 admits @xmath393 as a marginal , so the proof of part 1 .
is complete . [
proof of proposition [ prop : adap_stop ] ] the proof is the similar as that of proposition [ prop : stop_within_mcmc ] . for the proof of the first statement of part 1 .
one repeats the same arguments as for proposition [ prop : stop_within_mcmc ] with difference being in the inclusion of @xmath394 for @xmath395 and @xmath396 . for the second statement , to get the marginal of @xmath386 , re - write the target as : @xmath397 let @xmath398 denote the marginal of @xmath392 obtained in proposition [ prop : stop_within_mcmc ] .
using and the conditional independence of @xmath300 and @xmath326 , then for the marginal of @xmath399 w.r.t @xmath300 , @xmath326 , @xmath251 we have that @xmath400 where the summing over @xmath251 and integrating w.r.t .
@xmath326 is as in proposition [ prop : stop_within_mcmc ] . for part 2 .
note that the conditional density given @xmath251 and @xmath300 and @xmath5 of @xmath401 is @xmath402 hence the sequence @xmath403 satisfies the required property as direct consequence theorem 1 in @xcite and assumption ( a[assump : a5 - 6 ] ) .
lee , a. , yau , c. , giles , m. , doucet , a. & holmes c.c .
( 2010 ) on the utility of graphics cards to perform massively parallel implementation of advanced monte carlo methods , _ j. comp .
. statist .
_ , * 19 * , 769789 . | in this article we consider bayesian parameter inference associated to partially - observed stochastic processes that start from a set @xmath0 and are stopped or killed at the first hitting time of a known set @xmath1 .
such processes occur naturally within the context of a wide variety of applications .
the associated posterior distributions are highly complex and posterior parameter inference requires the use of advanced markov chain monte carlo ( mcmc ) techniques .
our approach uses a recently introduced simulation methodology , particle markov chain monte carlo ( pmcmc ) @xcite , where sequential monte carlo ( smc ) @xcite approximations are embedded within mcmc .
however , when the parameter of interest is fixed , standard smc algorithms are not always appropriate for many stopped processes . in @xcite the authors introduce smc approximations of multi - level feynman - kac formulae , which can lead to more efficient algorithms .
this is achieved by devising a sequence of nested sets from @xmath0 to @xmath1 and then perform the resampling step only when the samples of the process reach intermediate level sets in the sequence .
naturally , the choice of the intermediate level sets is critical to the performance of such a scheme . in this paper
, we demonstrate that multi - level smc algorithms can be used as a proposal in pmcmc .
in addition , we propose a flexible strategy that adapts the level sets for different parameter proposals .
our methodology is illustrated on the coalescent model with migration .
+ * key - words * : stopped processes , sequential monte carlo , markov chain monte carlo @xmath2department of statistics & applied probability , national university of singapore , singapore , 117546 , sg .
+ e-mail:`[email protected] ` + @xmath3department of electrical engineering , imperial college london , london , sw7 2az , uk .
+ e-mail:`[email protected] ` |
let @xmath1 be an oriented surface with finitely many punctures .
suppose that @xmath1 is endowed with a negatively curved riemannian metric and that in a neighborhood of each puncture the metric is asymptotically modeled " on a surface of revolution obtained by rotating the curve @xmath8 , for some @xmath9 , about the @xmath10-axis in @xmath11 ( where @xmath12 may depend on the puncture ) .
the results in this paper allow us to conclude that the geodesic flow on @xmath6 mixes exponentially fast . before stating the hypotheses precisely , we recall some facts about the metric on a surface @xmath13 of revolution for the function @xmath14 .
this surface is negatively curved , incomplete and the curvature can be expressed as a function of the distance to the cusp point @xmath15 where @xmath16 .
denote by @xmath17 the induced riemannian path metric and @xmath18 the riemannian distance to the cusp : @xmath19 then for @xmath20 , the gaussian curvature on @xmath13 has the following asymptotic expansion in @xmath21 , as @xmath22 : @xmath23 our main theorem applies to any incomplete , negatively curved surface with singularities of this form .
more precisely , we have : [ t = main ] let @xmath24 be a closed surface , and let @xmath25 .
suppose that the punctured surface @xmath26 carries a @xmath27 , negatively curved riemannian metric that extends to a complete distance metric @xmath28 on @xmath24 .
assume that the lift of this metric to the universal cover @xmath29 is geodesically convex .
denote by @xmath30 the distance @xmath31 , for @xmath32 .
assume that there exist @xmath33 such that the gaussian curvature @xmath34 satisfies @xmath35 and @xmath36 for @xmath37 and all @xmath38 .
then the geodesic flow @xmath39 is exponentially mixing : there exist constants @xmath40 such that for every pair of @xmath41 functions @xmath42 , we have @xmath43 for all @xmath44 , where @xmath45 denotes the riemannian volume on @xmath6 ( which is finite ) normalized so that @xmath46 . the regularity hypotheses on @xmath47 are not optimal .
see corollary [ c = main ] in the last section for precise formulations .
theorem [ t = main ] has a direct application to the dynamics of the weil - petersson flow , which is the geodesic flow for the weil - petersson metric @xmath48 of the moduli spaces @xmath4 of riemann surfaces of genus @xmath49 and @xmath50 punctures , defined for @xmath51 . for a discussion of the wp metric and properties of its flow , see the recent , related work @xcite . as a corollary , we obtain the following result , which originally motivated this study .
[ c = wp ] the weil - petersson geodesic flow on @xmath52 mixes exponentially fast when @xmath53 or @xmath54 .
wolpert shows in @xcite that the hypotheses of theorem [ t = main ] are satisfied by the wp metric on @xmath4 , for @xmath55 .
mixing of the wp flow ( for all @xmath56 ) had previously been established in @xcite . for @xmath57 ,
the conclusions of corollary [ c = wp ] do _ not _ hold @xcite : for every @xmath58 , there exist compactly supported , @xmath59 test functions @xmath47 such that the correlation between @xmath60 and @xmath61 decays at best polynomially in @xmath62 .
the geodesic convexity assumption in theorem [ t = main ] can be replaced by a variety of other equivalent assumptions .
for example , it is enough to assume that @xmath63 , where @xmath64 is a convex function ( as is the case in the wp metric ) .
alternatively , one may assume a more detailed expansion for the metric in the neighborhood of the cusps .
for example , the assumptions near the cusp are satisfied for a surface of revolution for the function @xmath65 , where @xmath66\to { { \mathbb r}}_{\geq 0}$ ] is @xmath27 , with @xmath67 and @xmath68 .
one can easily formulate further perturbations of this metric outside the class of surfaces of revolutions for which the hypotheses of theorem [ t = main ] hold near @xmath69 . to simplify the exposition and reduce as much as possible the use of unspecified constants
, we will assume in our proof that @xmath70 , so that @xmath71 has only one cusp . in a landmark paper @xcite
, dolgopyat established that the geodesic flow for any negatively - curved compact surface is exponentially mixing .
his techniques , building in part on earlier work of ruelle , pollicott and chernov , have since been extracted and generalized in a series of papers , first by baladi - valle @xcite , then avila - gouzel - yoccoz @xcite , and most recently in the work of arajo - melbourne @xcite , upon which this paper relies . ultimately , the obstructions to applying dolgopyat s original argument in this context are purely technical , but to overcome these obstructions in any context is the heart of the matter .
the solution to the analogous problem in the billiards context exponential mixing for sinai billiards of finite horizon has only been recently established @xcite . to prove exponential mixing using the symbolic - dynamical approach of dolgopyat , baladi - valle @xmath72 , one constructs a section to the flow with certain analytic and symbolic dynamical properties . in sum ,
one seeks a surface @xmath73 transverse to the flow @xmath7 in the three manifold @xmath6 on which the dynamics of the return map can be tightly organized . in particular , we seek a return time function @xmath74 defined on a full measure subset @xmath75 , with @xmath76 for all @xmath77 and so that the dynamics of @xmath78 on @xmath79 are hyperbolic and can be modeled on a full shift on countably many symbols . for @xmath7 to be exponentially mixing , the function @xmath13 must be constant along stable manifolds , have exponential tails and satisfy a non - integrability condition ( uni ) ( which will hold automatically if the flow @xmath7 preserves a contact form , as is the case here ) . whereas in @xcite and @xcite the map @xmath80 is required to be piecewise uniformly @xmath81 , the regularity of @xmath80 is relaxed to @xmath82 in @xcite .
this relaxation in regularity might seem mild , but it is crucial in applications to nonuniformly hyperbolic flows with singularities .
the reason is that the surface @xmath83 is required to be saturated by leaves of the ( strong ) stable foliation @xmath84 for the flow @xmath7 .
the smoothness of the foliation @xmath84 thus dictates the smoothness of the surface @xmath83 which then determines the smoothness of @xmath80 ( up to the smoothness of the original flow @xmath7 ) .
_ even in the case of contact anosov flows in dimension 3 , _ the foliation @xmath84 is no better than @xmath82 , for some @xmath85 , unless the flow is algebraic in nature . while it has long been known that this @xmath82 regularity condition holds for the stable and unstable foliations of contact anosov flows in dimension 3 , this is far from the case for singular and nonuniformly hyperbolic flows , even in low dimension . in the context of this paper ,
the geodesic flow @xmath7 is not even complete , and the standard singular hyperbolic theory fails to produce @xmath7-invariant foliations @xmath86 and @xmath84 , let alone foliations with @xmath82 regularity .
the flows @xmath7 considered here , while incomplete , bear several resemblances to anosov flows .
most notably , there exist @xmath87-invariant stable and unstable cone fields that are defined _ everywhere _ in @xmath6 .
the angle between these cone fields tends to zero as the basepoint in @xmath6 approaches the singularity .
the action of @xmath87 in these cones is strongly hyperbolic , with the strength of the hyperbolicity approaching infinity as the orbit comes close to the singularity .
the key observation in this paper is that by changing the riemannian metric on @xmath6 _ and _ performing a natural time change in @xmath7 one obtains a volume - preserving _ anosov _ flow on a _ complete _ riemannian manifold of finite volume .
this time change does not change orbits and has a predictable effect on stable and unstable bundles .
one can apply all of the known machinery for anosov flows to this rescaled flow , and transferring the information back to the original flow , one concludes that @xmath7 possesses invariant stable and unstable foliations @xmath86 and @xmath84 that are locally uniformly @xmath82 .
this gives the crucial input in constructing the section @xmath83 and return function @xmath13 defined above . in the setting of weil - petersson geometry
, one can summarize the results of this time change : in the exceptional case @xmath88 , the weil - petersson geodesic flow , when run at unit speed _ in the teichmller metric _ is ( like the teichmller flow ) an anosov flow . for @xmath89 ,
the wp flow is not anosov , even when viewed in the teichmller metric ( or an equivalent riemannian metric such as in @xcite ) , but it might be fruitful to study the flow from this perspective .
we remark here that hamenstdt @xcite has recently constructed measurable orbit equivalences between the wp and teichmller geodesic flows for all @xmath90 .
a different approach , using anisotropic function spaces , has been employed by liverani to establish exponential mixing for contact anosov flows in arbitrary dimension , even when the foliations @xmath86 and @xmath84 fail to be @xmath41 @xcite .
this method is more holistic ( though no less technical ) as the arguments take place in the manifold itself ( not a section ) and avoid symbolic dynamics
. it would be interesting to attempt to import this machinery to the present context .
this is the approach employed in the recent work of baladi , demers and liverani on sinai billiards in @xcite mentioned above .
this paper is organized as follows . in section [ s = prelim ]
we recall some facts about geodesic flows and basic comparison lemmas for odes . in section [ s = deltaregularity ] , we establish ( under the hypotheses of theorem [ t = main ] ) @xmath91 regularity for the functions @xmath92 measuring distance to the cusps in @xmath1 .
the arguments there bear much in common with standard proofs of regularity of busemann functions in negative curvature , but additional attention to detail is required to obtain the correct order estimates on the size of the derivatives of the @xmath92 . in section
[ s = revolution ] we establish basic geometric properties of the surfaces considered here , in close analogy to properties of surfaces of revolution .
these results refine some known properties of the weil - petersson metric .
section [ s = global ] addresses the global properties of the flow @xmath7 .
here we construct a new riemannian metric on @xmath6 , which we call the @xmath93 metric , in which @xmath6 is complete .
rescaling @xmath7 to be unit speed in the @xmath93 metric , we obtain a new flow @xmath94 which we prove is anosov , with _ uniform _ bounds on its first three derivatives ( in the @xmath93 metric ) .
we derive consequences of this , including ergodicity of @xmath7 and existence and @xmath82 regularity of @xmath7 invariant unstable and stable foliations @xmath84 and @xmath86 . in the final section ( section [ s = expmix ] )
, we construct the section @xmath83 to the flow and return time function @xmath13 satisfying the hypotheses of the arajo - melbourne theorem . in essence
this is equivalent to constructing a young tower for the return map to @xmath83 and is carried out using standard methods . here the properties of geodesics established in section [ s = revolution ] come into play in describing the dynamics of the return map of the flow to the compact part of @xmath6 .
we thank scott wolpert , sebastien gouzel , carlangelo liverani and curtis mcmullen for useful conversations , and viviane baladi and ian melbourne for comments on a draft of this paper .
let @xmath1 be an oriented surface endowed with a riemannian metric . as usual @xmath95 denotes the inner product of two vectors and @xmath96 is the levi - civita connection defined by the riemannian metric .
it is the unique connection that is symmetric and compatible with the metric .
the surface @xmath1 carries a unique almost complex structure compatible with the metric .
we denote this structure by @xmath97 ; for @xmath98 , the vector @xmath99 is the unique tangent vector in @xmath100 such that @xmath101 is a positively oriented orthonormal frame . the covariant derivative along a curve @xmath102 in @xmath1 is denoted by @xmath103 , @xmath104 or simply @xmath105 if it is not necessary to specify the curve ;
if @xmath106 is a vector field along @xmath107 that extends to a vector field @xmath108 on @xmath1 , we have @xmath109 . given a smooth map @xmath110 , we let @xmath111 denote covariant differentiation along a curve of the form @xmath112 for a fixed @xmath113 .
similarly @xmath114 denotes covariant differentiation along a curve of the form @xmath115 for a fixed @xmath116 .
the symmetry of the levi - civita connection means that @xmath117 for all @xmath116 and @xmath113 .
a geodesic segment @xmath118 is a curve satisfying @xmath119 , for all @xmath120 . throughout this paper , all geodesics
are assumed to be unit speed : @xmath121 .
the riemannian curvature tensor @xmath13 is defined by @xmath122})c\ ] ] and the gaussian curvature @xmath123 is defined by @xmath124 where @xmath125 is an arbitrary unit vector . for @xmath126
, we represent each element @xmath127 in the standard way as a pair @xmath128 with @xmath129 and @xmath130 , as follows
. each element @xmath127 is tangent to a curve @xmath131 with @xmath132 .
let @xmath133 be the curve of basepoints of @xmath134 in @xmath1 , where @xmath135 is the standard projection .
then @xmath136 is represented by the pair @xmath137 regarding @xmath138 as a bundle over @xmath1 in this way gives rise to a natural riemannian metric on @xmath139 , called the _ sasaki metric_. in this metric , the inner product of two elements @xmath140 and @xmath141 of @xmath142 is defined : @xmath143 this metric is induced by a symplectic form @xmath144 on @xmath138 ; for vectors @xmath140 and @xmath141 in @xmath142 , we have : @xmath145 this symplectic form is the pull back of the canonical symplectic form on the cotangent bundle @xmath146 by the map from @xmath139 to @xmath146 induced by identifying a vector @xmath147 with the linear function @xmath148 on @xmath149 . for @xmath150 let @xmath151 denote the unique geodesic @xmath151 satisfying @xmath152 . the geodesic flow @xmath153 is defined by @xmath154 wherever this is well - defined .
the geodesic flow is always defined locally .
the _ geodesic spray _ is the vector field @xmath155 on @xmath139 ( that is , a section of @xmath138 ) generating the geodesic flow . in the natural coordinates on @xmath138 given by the connection
, we have @xmath156 , for each @xmath126 .
the spray is tangent to the level sets @xmath157 .
henceforth when we refer to the geodesic flow @xmath7 , we implicity mean the restriction of this flow to the unit tangent bundle @xmath6 .
since the geodesic flow is hamiltonian , it preserves a natural volume form on @xmath6 called the liouville volume form .
when the integral of this form is finite , it induces a unique probability measure on @xmath6 called the _ liouville measure _ or _
liouville volume_. consider now a one - parameter family of geodesics , that is a map @xmath158 with the property that @xmath159 is a geodesic for each @xmath160 .
denote by @xmath161 the vector field @xmath162 along the geodesic @xmath163 .
then @xmath164 satisfies the _ jacobi equation : _ @xmath165 in which @xmath105 denotes covariant differentiation along @xmath166 .
since this is a second order linear ode , the pair of vectors @xmath167 uniquely determines the vectors @xmath161 and @xmath168 along @xmath169 . a vector field @xmath164 along a geodesic
@xmath166 satisfying the jacobi equation is called a _ jacobi
field_. the pair @xmath170 corresponds in the manner described above to the tangent vector at @xmath171 to the curve @xmath172 , which is @xmath173 .
thus [ prop : key ] the image of the tangent vector @xmath174 under the derivative of the geodesic flow @xmath175 is the tangent vector @xmath176 , where @xmath164 is the unique jacobi field along @xmath166 satisfying @xmath177 and @xmath178 .
computing the wronskian of the jacobi field @xmath179 and an arbitrary jacobi field @xmath164 shows that @xmath180 is constant .
it follows that if @xmath181 for some @xmath182 , then @xmath183 for all @xmath113 .
similarly if @xmath184 and @xmath181 for some @xmath182 , then @xmath185 and @xmath183 for all @xmath113 ; in this case we call @xmath164 a
_ perpendicular jacobi field_. if @xmath186 is a variation of geodesics giving rise to a perpendicular jacobi field , then we call @xmath186 a _ perpendicular variation of geodesics_. the space of all perpendicular jacobi fields along a unit speed geodesic @xmath166 corresponds to the orthogonal complement @xmath187 ( in the sasaki metric ) to the geodesic spray @xmath188 at the point @xmath189 .
to estimate the norm of the derivative @xmath87 on @xmath190 , it suffices to restrict attention to vectors in the invariant subspace @xmath191 ; that is , it suffices to estimate the growth of perpendicular jacobi fields along unit speed geodesics . because @xmath1 is a surface , the jacobi equation ( [ e = jacobi ] ) of a perpendicular jacobi field along a unit speed geodesic segment can be expressed as a scalar ode in one variable .
given such a geodesic @xmath118 , any perpendicular jacobi field @xmath164 along @xmath166 can be written in the form @xmath192 , where @xmath193 satisfies the _ scalar jacobi equation : _ @xmath194 to analyze solutions to ( [ e = scalarjacobi ] ) it is often convenient to consider the functions @xmath195 and @xmath196 which satisfy the riccati equations @xmath197 and @xmath198 , respectively . in the next subsection , we describe some techniques for analyzing solutions to these types of equations .
we will use a few basic comparison lemmas for solutions to ordinary differential equations .
the first is standard and is presented without proof : [ l = comparison]*[basic comparison ] * let @xmath199 \to { { \mathbb r}}$ ] be @xmath41 , and let @xmath200 \to { { \mathbb r}}$ ] be a solution to the ode @xmath201 suppose that @xmath202\to{{\mathbb r}}$ ] are @xmath41 functions satisfying @xmath203 . then the following hold : * if @xmath204 for all @xmath205 $ ] , then @xmath206 for all @xmath207 $ ] . *
if @xmath208 for all @xmath205 $ ] , then @xmath209 for all @xmath210 $ ] . *
if @xmath211 for all @xmath205 $ ] , then @xmath212 for all @xmath207 $ ] . *
if @xmath213 for all @xmath205 $ ] , then @xmath214 for all @xmath210 $ ] .
we will have several occasions to deal with riccati equations of the form @xmath215 on an interval @xmath216 $ ] ( where typically @xmath217 for some geodesic segment @xmath166 ) .
since the curvature of the surfaces we consider is not bounded away from @xmath218 , most of the odes we deal with will have unbounded coefficients .
this necessitates reproving some standard results about solutions .
a key basic result is the following .
[ l = monotone2]*[existence of unstable riccati solutions ] * suppose @xmath219 \to { { \mathbb r}}_{>0}$ ] is a @xmath41 function satisfying @xmath220 .
then there exists a unique solution @xmath221 to the riccati equation @xmath222 for @xmath223 $ ] satisfying @xmath224 on @xmath225 $ ] and @xmath226 .
moreover , if @xmath227 \to { { \mathbb r}}_{>0}$ ] is any @xmath41 function satisfying @xmath228 and @xmath229 , for all @xmath210 $ ] , then @xmath230 , for all @xmath207 $ ] .
let @xmath231 be a function satisfying the hypotheses of the lemma .
then there is a function @xmath232 \to { { \mathbb r}}_{>0}$ ] such that @xmath233 and @xmath234 for all @xmath235 $ ] and @xmath236 as @xmath237 .
observe that @xmath238 for @xmath239 .
now fix a decreasing sequence @xmath240 in @xmath241 . for each @xmath242 let @xmath243 be the solution to ( [ e = riccsimple ] ) on @xmath244 $ ] with @xmath245 .
we can apply lemma [ l = comparison ] to equation ( [ e = riccsimple ] ) on the interval @xmath244 $ ] with @xmath246 and @xmath247 .
this gives us @xmath248 for @xmath249 .
we can also apply lemma [ l = comparison ] on this interval with @xmath250 for @xmath251 and @xmath252 .
this gives us @xmath253 , for @xmath254 .
the sequence of solutions @xmath243 is thus increasing , positive and bounded above by @xmath255 .
it follows that the function @xmath256 is a solution to ( [ e = riccsimple ] ) , is positive on @xmath257 $ ] , is bounded above by @xmath255 , and thus satisfies @xmath258 .
it remains to show that @xmath259 is the only solution of ( [ e = riccsimple ] ) with the desired properties .
suppose @xmath260 is another such solution .
since the graphs of two solutions of ( [ e = riccsimple ] ) can not cross , we may assume that @xmath261 for @xmath262 .
but then @xmath263 \leq 0\ ] ] for @xmath239 .
since @xmath264 as @xmath237 , this is possible only if @xmath265 for @xmath262 .
we call the solution of the riccati equation defined by the previous lemma the _ unstable solution _ on @xmath225 $ ] .
[ l = monotone2.5]*[comparison of unstable riccati solutions ] * for @xmath266 , let @xmath267 \to { { \mathbb r}}_{>0}$ ] be a @xmath41 function satisfying @xmath268 and let @xmath269 \to { { \mathbb r}}_{>0}$ ] be the unstable solution .
suppose @xmath270 for all @xmath223 $ ] .
then @xmath271 for all @xmath272 $ ] .
suppose @xmath273 for some @xmath274 $ ]
. then we can apply lemma [ l = comparison ] to the equation @xmath275 with @xmath276 and @xmath277 to obtain @xmath278 for all @xmath279 $ ] .
it now suffices to show that if there is @xmath280 $ ] such that @xmath281 for all @xmath282 $ ] , then we must have @xmath283 for all @xmath282 $ ] .
but if @xmath284 on @xmath285 $ ] we have @xmath286 for @xmath287 .
since @xmath288 as @xmath237 , this is possible only if @xmath289 for @xmath290 .
[ l = monotone3 ] let @xmath291 \to { { \mathbb r}}_{>0}$ ] be a @xmath41 function satisfying @xmath292 , and let @xmath221 be the unstable solution .
let @xmath293 . 1 .
if @xmath294 for all @xmath295 $ ] , then @xmath296 for all @xmath295 $ ] .
if @xmath297 for all @xmath295 $ ] , then @xmath298 for all @xmath295 $ ] .
suppose @xmath299 , @xmath300 and @xmath301\qquad\text{for all $ t \in ( 0,t_2]$.}\ ] ] then there exists @xmath302 such that @xmath303 \qquad\text{for all $ t \in ( 0,t_2]$.}\ ] ] 1 and 2 .
these follow from lemma [ l = monotone2.5 ] because @xmath304 is the unstable solution for @xmath305 \3 .
choose @xmath306 such that @xmath307 .
then , @xmath308 for @xmath309 .
it follows from parts 1 and 2 of this lemma that for each @xmath310 $ ] we have @xmath311 $ ] , for all @xmath312 $ ] .
consequently , @xmath313 $ ] , for all @xmath314 $ ] .
now choose @xmath315 .
we then have @xmath316 for @xmath317 .
we conclude that @xmath318 $ ] , for @xmath319 .
suppose @xmath1 satisfies the hypotheses of theorem [ t = main ] , with @xmath70 . before considering the global properties of the metric on @xmath1
, we introduce local coordinates about the puncture @xmath320 and study the behavior of geodesics that remain in this cuspidal region during some time interval . in this section and the next , we thus assume that the punctured disk @xmath321 has been endowed with an incomplete riemannian metric , whose completion is the closed disk @xmath322 .
assume that the lift of this metric to @xmath323 is _ geodesically convex : _ that is , any two points in @xmath323 can be connected by a unique geodesic in @xmath323 .
let @xmath28 be the riemannian distance metric on @xmath322 and for @xmath324 , let @xmath325 .
for @xmath326 , we denote by @xmath327 the set of @xmath328 with @xmath329 .
assume that that there exists @xmath68 such that for all @xmath330 the curvature of the riemannian metric satisfies : @xmath331 and @xmath332 for @xmath37 .
the main result of this section establishes regularity of the function @xmath21 and estimates on the size of its derivatives .
we also introduce a function @xmath107 that measures the geodesic curvature of the level sets of @xmath21 and establish some of its properties .
the results in this section establish in this incomplete , singular setting the standard regularity properties of busemann functions for compact , negatively curved manifolds ( see , e.g. @xcite ) in particular , busemann functions for a @xmath59 metric are @xmath333 .
the main techniques are thus fairly standard but require some care in the use of comparison lemmas for odes .
to avoid tedium , we have described many calculations in detail but have left others to the reader .
[ p = c ] the cusp distance function @xmath21 is @xmath91 .
let @xmath334 , and let @xmath335 be the geodesic curvature function defined by @xmath336 then : 1 .
@xmath337 = cjv$ ] .
2 . for any vector field @xmath338 : @xmath339 .
3 . @xmath340 .
4 . @xmath341 .
5 . @xmath342 .
[ l = cuspdist ] the function @xmath21 satisfies : @xmath343 and @xmath344 , for @xmath345 .
this follows from the facts : @xmath346 , @xmath347 , and @xmath348 , @xmath349 , proved in proposition [ p = c ] .
we prove first that @xmath21 is @xmath91 , in several steps .
* step 0 : @xmath21 is continuous .
* we realize the universal cover of the punctured disk @xmath350 as the strip @xmath351 with the deck transformations @xmath352 , @xmath353 .
endow @xmath351 with the lifted metric , which is geodesically convex by assumption , and lift @xmath21 to a function @xmath354 . by assumption ,
the completion of @xmath350 is @xmath355 , and so the completion @xmath356 in this metric is the union of @xmath357 $ ] with a single point @xmath358 .
since @xmath351 is negatively curved and geodesically convex , it is in particular @xmath359 .
the @xmath359 property is preserved under completion , and so @xmath360}$ ] is also @xmath359 .
thus for every for every @xmath361 $ ] , there is a unique unit - speed geodesic from @xmath362 to @xmath358 .
this projects to a ( unique ) geodesic in @xmath363 from @xmath364 to @xmath365 .
fix @xmath366 with lift @xmath361 $ ] , and let @xmath367\to \overline { { \mathbb d}}$ ] be the unit - speed geodesic from @xmath365 to @xmath364 found by the previous argument .
it has the property that @xmath368 for every @xmath369 $ ] .
let @xmath370 be a sequence of times in @xmath371 , and define a sequence of functions @xmath372 \to { { \mathbb r}}_{>0}$ ] by @xmath373 .
the @xmath374 are convex , @xmath375 away from @xmath376 , and @xmath377 for all @xmath378 . for every @xmath379 $ ] and all @xmath380
, we have @xmath381 this follows from the triangle inequality .
since @xmath382 is cauchy , it converges ( locally uniformly in @xmath357 $ ] ) to a continuous , convex function @xmath383 .
moreover @xmath384 is the distance @xmath385 .
it follows that @xmath386 is continuous ( and convex ) , and so @xmath21 is continuous .
* step 1 : @xmath21 is @xmath41 .
* let @xmath387 be the corresponding sequence of radial vector fields on @xmath357 $ ] .
[ l = vn ] fix @xmath379 $ ] . for all
@xmath380 sufficiently large , we have : @xmath388 thus @xmath389 is a cauchy sequence in the local uniform topology .
this is a standard argument in negative curvature ( in fact nonpositive curvature suffices ) .
this uses that @xmath377 for all @xmath378 .
this lemma implies that @xmath21 is @xmath41 .
let @xmath390 be the local uniform limit of the @xmath389 : by definition , @xmath391 .
let @xmath392 be the projection of @xmath390 to @xmath321 .
it remains to show that @xmath134 is @xmath375 , which implies that @xmath21 is @xmath91 .
* step 2 : @xmath134 is @xmath41 .
* [ l = visc1 ] there exists @xmath393 , such that for every @xmath394 with @xmath395 , the following holds . for every vector field @xmath338 ,
@xmath396 exists and is continuous in a neighborhood of @xmath397)$ ] .
moreover : @xmath398 for all @xmath399 $ ] , where @xmath259 is the positive solution to the riccati equation @xmath400 given by lemma [ l = monotone2 ] , satisfying @xmath401 .
fix @xmath393 ( we will specify how small it must be later ) . fix @xmath364 with @xmath402 , and denote by @xmath166 the geodesic @xmath403 . for each @xmath378 , define a perpendicular , radial variation of geodesics @xmath404 by the properties : @xmath405 , @xmath406 and @xmath407 for all @xmath408 with @xmath409 ( and @xmath116 belonging to a small , fixed neighborhood of @xmath365 ) .
let @xmath410 be defined by @xmath411 ; then for @xmath409 , and @xmath116 sufficiently small , we have @xmath412 it follows that for any @xmath413 , if @xmath378 sufficiently large and @xmath414 , we have @xmath415.\ ] ] we have already shown ( working on the universal cover ) that in a neighborhood of @xmath166 , we have @xmath416 and @xmath417 uniformly on compact sets .
let @xmath418 be the limiting variation of geodesics , which satisfies @xmath419 , and @xmath420 .
at this point we have shown that @xmath166 is @xmath41 , with @xmath421 .
note that @xmath422 , for all @xmath378 , @xmath409 .
since @xmath423 , it suffices to show that @xmath424 the proof that @xmath425 is immediate : since @xmath426 is a variation of geodesics , we have @xmath427 .
we now show that @xmath428 .
let @xmath429 be the scalar jacobi field associated with the perpendicular variation @xmath426 : @xmath430 on the one hand , @xmath431 while on the other hand , @xmath432 writing @xmath433 , we thus have @xmath434 .
we prove that @xmath435 converges to @xmath436 , the positive solution to ( [ e = zetaeq ] ) .
to see this , we first establish uniform upper and lower bounds for @xmath435 , for @xmath437 .
the @xmath243 satisfy the riccati equation : @xmath438 with @xmath439 , for all @xmath116 .
now @xmath440 by ( [ e = uniformtn ] ) we thus have : @xmath441,\ ] ] if @xmath442 is sufficiently small , @xmath443 , and @xmath378 is sufficiently large .
we show that there exists @xmath444 such that , for @xmath378 sufficiently large , we have @xmath445.\ ] ] to see the lower bound , let @xmath446 . then @xmath447 . on the other hand , when @xmath448 , we have @xmath449 .
this is larger than @xmath450 provided that @xmath451 . but
this will hold if @xmath452 .
the upper bound is similarly obtained . by lemma [ l = comparison ] , @xmath453 for all @xmath409 , which proves ( [ e = zetamu ] ) .
we now use lemma [ l = comparison ] to prove that for some large but fixed @xmath454 : latexmath:[\[\label{e = zetamzetan } and @xmath443 . for @xmath380
, we have , since @xmath456 : @xmath457 subtracting the odes for @xmath458 and @xmath243 , we have for @xmath459 : @xmath460 @xmath461 @xmath462 since @xmath463 ) , @xmath464 , and @xmath465 .
writing @xmath466 , we have that @xmath467 satisfies the ode @xmath468 fix @xmath378 and let @xmath469 .
then @xmath470 , and @xmath471 evaluated at @xmath472 is @xmath473 we claim that if @xmath454 is sufficiently large , then @xmath474 . to see this fix @xmath299 such that
@xmath475 then ( [ e = crudekbounds ] ) and ( [ e = zetamu ] ) imply that @xmath476 clearly this is @xmath477 for @xmath409 if @xmath454 is sufficiently large .
since @xmath478 , for all @xmath479 , this implies by lemma [ l = comparison ] that @xmath480 for all @xmath437 ; a similar argument shows that @xmath481 , and hence if @xmath454 is sufficiently large and @xmath380 , then for all @xmath437 ( [ e = zetamzetan ] ) holds .
thus @xmath482 tends to @xmath365 as @xmath370 , with @xmath408 fixed .
thus the @xmath435 converge , and since they satisfy ( [ e = zetan ] ) , their limit @xmath436 satisfies ( [ e = zetaeq ] ) .
we obtain that the functions @xmath483 converge locally uniformly , and hence @xmath484 exists and is continuous . since @xmath485 , @xmath486 , and @xmath487 ,
we obtain ( [ e = jvstuff ] ) by taking a limit and setting @xmath488 . in light of lemma
[ l = visc1 ] , we define a function @xmath489 as follows .
for each @xmath490 , let @xmath491\to { { \mathbb r}}_{>0}$ ] be the positive solution to ( [ e = zetaeq ] ) given by lemma [ l = monotone2 ] .
we then set @xmath492 .
it follows immediately from lemma
[ l = visc1 ] that for every @xmath490 , we have @xmath493 * step 3 : @xmath494 is @xmath41 . * to prove that @xmath21 is @xmath375 , it thus suffices to show that @xmath494 is @xmath41 .
equation ( [ e = zetaeq ] ) implies that @xmath495 exists and is continuous .
it remains to show that @xmath496 exists and is continuous .
we fix @xmath393 as above and let @xmath497 , and we reintroduce the variations of geodesics @xmath404 defined by the properties : @xmath498 , @xmath499 and @xmath500 for all @xmath116 . as
above , write @xmath501 , and let @xmath418 be the limiting variation of geodesics .
we observe that lemma [ l = visc1 ] also implies that @xmath502 and @xmath503 , locally uniformly , where @xmath504 and @xmath505 .
the convergence follows from the formulae : @xmath506
@xmath507 thus the variation @xmath166 is @xmath81 on @xmath321 , and satisfies @xmath508 .
we record here a lemma , which follows easily from these formulae , combined with ( [ e = deltadiff ] ) , ( [ e = crudekbounds ] ) and ( [ e = zetamzetan ] ) .
[ l = jratio ] for all @xmath409 , we have @xmath509 , and for all @xmath443 , we have @xmath510 . as above , let @xmath435 be the solution to ( [ e = zetan ] ) , and let @xmath436 be the solution to ( [ e = zetaeq ] ) .
note that @xmath511 . to prove that @xmath496 exists and is continuous , we show that @xmath512 converges uniformly to @xmath513 , the unique bounded solution to @xmath514 which satisfies @xmath515 , for all @xmath116 .
since @xmath516 and @xmath517 locally uniformly , this will imply that @xmath518 exists and is continuous .
[ l = partialzetadiff ] there exists @xmath302 such for all @xmath519 and all @xmath443 , we have @xmath520 ) with respect to @xmath116 , we obtain : @xmath521 note that since @xmath439 , for all @xmath378 , we have that @xmath522 , for all @xmath378 . to simplify notation , fix @xmath116 , and write @xmath523 , @xmath524 , and @xmath525 .
then @xmath526 , for all @xmath378 . from equations ( [ e = zetamu ] ) and ( [ e = zetamzetan ] )
, we have @xmath527 and @xmath528
. then equation ( [ e = partialzetan ] ) gives @xmath529 we first claim there exists @xmath530 such that @xmath531 , for all @xmath409 .
let @xmath532 .
then @xmath533 , whereas @xmath534 evaluated at @xmath535 gives @xmath536 .
then @xmath537 if and only if @xmath538 ; dividing through by @xmath539 , and recalling that @xmath540 , we are reduced to showing : @xmath541 which holds if and only if @xmath542 . since @xmath543 , @xmath544 , and @xmath545 , for @xmath378 sufficiently large ,
this will hold provided that @xmath454 and @xmath378 are sufficiently large .
we conclude that latexmath:[\[\label{l = wnbound } we next claim that there exists @xmath302 such that for @xmath380 sufficiently large and @xmath443 , we have @xmath547 ) of lemma [ l = partialzetadiff ] . for @xmath380 ,
subtracting the corresponding equations in ( [ e = partialzetanshort ] ) , we obtain : @xmath548 we claim that there exists @xmath299 such that for @xmath380 , and @xmath549 , we have @xmath550 ) .
let @xmath551 be given so that ( [ e = epsilonclaim ] ) holds for @xmath380 , and @xmath437 .
let @xmath552 .
then @xmath553 , and ( [ l = wnbound ] ) implies that @xmath554 @xmath555 @xmath556 @xmath557 for some @xmath299 , since @xmath558 , by ( [ e = zetamu ] ) , and @xmath559 , by lemma [ l = jratio ] .
this shows that ( [ e = wndiff ] ) holds at @xmath560 , provided @xmath378 is sufficiently large .
we claim that there exists @xmath302 such that for all such @xmath561 , we have @xmath562 , for @xmath437 .
we prove the upper bound ; the lower bound is similar .
we will employ lemma [ l = comparison ] . to this end , let @xmath563 .
note that @xmath564 , whereas evaluating @xmath471 at @xmath565 , we get @xmath566 to satisfy the hypotheses of lemma [ l = comparison ] , we require that @xmath567 whenever @xmath565 , which is implied by : @xmath568 or : @xmath569 since @xmath570 ( by [ e = crudekbounds ] ) , we see that this will hold ( for all @xmath378 sufficiently large ) if @xmath571 .
this establishes ( [ e = wndiff ] ) .
we return to the proof of the claim that there exists an @xmath299 such that for @xmath380 , and @xmath443 the inequality ( [ e = epsilonclaim ] ) holds .
the proof amounts to adding and subtracting terms within the left hand side of ( [ e = epsilonclaim ] ) , varying one at a time the multiplied quantity in each term .
the added and subtracted terms are grouped in twos and the absolute value of the difference in each pair is bounded above .
to illustrate , consider the difference appearing on the left hand side of ( [ e = epsilonclaim ] ) .
the first two terms appearing in that difference , coming from ( [ e = wnminuswm ] ) , are : @xmath572 @xmath573 the quantity @xmath574 can be bounded , and the remaining term @xmath575 can be further decomposed , as follows .
first , using ( [ e = zetamu ] ) to bound @xmath576 , the assumption that @xmath577 together with ( [ e = uniformtn ] ) to bound @xmath578 , and the fact from lemma [ l = jratio ] that @xmath579 , we have that @xmath580 second , to deal with the remaining term @xmath581 , we write : @xmath582 @xmath583 and bound each term separately in a similar way to give a bound on the initial quantity @xmath584 of order @xmath585 .
the same procedure is used to bound the remaining part of the difference appearing in ( [ e = epsilonclaim ] ) , which is : @xmath586 @xmath587 in all , we use the following bounds : * @xmath21 is comparable to @xmath113 , by ( [ e = uniformtn ] ) ; * @xmath543 , by ( [ e = zetamu ] ) ; * @xmath588 , by ( [ e = zetamzetan ] ) ; * @xmath589 , for @xmath437 , by lemma [ l = jratio ] ; * @xmath590 , since @xmath591 and @xmath592 ; * @xmath593 , which uses the bounds on @xmath594 and @xmath595 , as well as ( [ e = uniformtn ] ) and lemma [ l = vn ] .
the details are left to the patient reader .
this finishes the verification of the claim in ( [ e = epsilonclaim ] ) , and thus the proof of lemma [ l = partialzetadiff ] .
to finish the proof that @xmath494 is @xmath41 , note that equation ( [ e = partialzetadiff ] ) can then be re - expressed using ( [ e = zetamu ] ) : latexmath:[\[\label{e = wndiff2 } for @xmath443 .
recalling that @xmath502 , we conclude that @xmath597 locally uniformly in @xmath408 .
the bounds @xmath598 from ( [ l = wnbound ] ) become in the limit @xmath599 . but @xmath600 , and so we conclude that latexmath:[\[\label{e = jvnubound } * step 4 : @xmath494 is @xmath81 . * a very similar proof to the one in step 3 ( with more terms to estimate , but using , in addition to the previously obtained bounds , the bound @xmath602 ) gives that @xmath494 is @xmath81 with @xmath603 .
one obtains this estimate as in the previous step by bounding @xmath604 , @xmath605 , @xmath606 , and @xmath607 .
each of these is controlled by a differential equation , whose solutions can be estimated using a double variation of geodesics @xmath608 .
the key point , illustrated by the previous computations , is that because @xmath34 has the expected " order of derivatives with respect to @xmath21 , any quantity obtained by solving a first - order linear differential equation derived from the riccati equation with coefficients expressed in terms of these derivatives will have the expected " order in @xmath21 as well .
thus , @xmath609 for @xmath610 implies that @xmath611 , for @xmath612 .
this completes the proof that @xmath21 is @xmath91 .
we now turn to items 1 - 5 .
\1 . @xmath613 $ ] by the symmetry of the levi - civita connection .
but @xmath614 .
\2 . for arbitrary @xmath338 , we have @xmath615 , giving the first conclusion : @xmath616 .
the second conclusion follows from the first and the fact that @xmath97 is parallel .
\3 . note that @xmath617 .
it then is equivalent to prove that @xmath618 .
fix @xmath619 and let @xmath620 . along this geodesic
, @xmath621 satisfies the equation ( [ e = zetaeq ] ) . on the other hand @xmath622 .
part 3 of lemma [ l = monotone3 ] implies the desired result . \4 . the desired estimate @xmath623 is equivalent to @xmath624 because @xmath625 and @xmath626 . but
( [ e = jvnubound ] ) gives that @xmath624 .
the estimate @xmath627 is equivalent to @xmath603 .
this estimate was obtained in step 4 above .
we continue to work locally in @xmath321 with a metric satisfying ( [ e = kassump ] ) and ( [ e = nablakassump ] ) . in this section
we establish properties of geodesics in this cuspidal region .
the theme of this section is that metrics of this form inherit many of the geodesic properties of a surface of revolution for a profile function @xmath14 , with @xmath68 . in @xmath11 , coordinates on this surface are @xmath628 , \,\phi\in [ 0,2\pi].\ ] ] as remarked in the introduction , if @xmath21 denotes the distance to the cusp @xmath629 on this surface , then @xmath630 and ( [ e = kassump ] ) holds .
other properties are : * * area : * the area of the region @xmath631 is @xmath632 . *
* clairaut integral : * let @xmath169 be a geodesic segment in the surface of revolution , and let @xmath633 be the angle between @xmath634 and the foliation @xmath635 .
then the function @xmath636 is constant .
we establish here in sections [ s = area]-[s = qc ] the analogues of these properties in our setting .
we also establish in section [ s = jacobi ] some coarse invariance properties of positive jacobi fields in @xmath321 .
fix @xmath393 , and for @xmath637 , denote by @xmath638 the disk @xmath639 .
let @xmath107 be the geodesic curvature function defined in the previous section , @xmath640 the arclength element and @xmath641 is the area element defined by the metric . let @xmath418 be the radial unstable variation of geodesics described in the previous section , defined by the properties * @xmath508 , for all @xmath116 , * @xmath642 , for all @xmath408 , and * @xmath643 , for all @xmath116 . the arclength element is found by differentiating @xmath418 with respect to @xmath116 : @xmath644 using part 3 of proposition [ p = c ] , we estimate @xmath645 by @xmath646 and so @xmath647 . we obtain that : @xmath648 the volume of the region @xmath638 is obtained by integrating @xmath649 over the region @xmath650 ; t\in [ 0,t_0]$ ] , where @xmath651 is the length of the curve @xmath652 .
it follows that @xmath653 for @xmath656 , we define @xmath657 by : @xmath658 thus the vectors @xmath659 with @xmath660 and @xmath661 point directly at the cusp @xmath365 that is , the geodesics that they determine hit the cusp in finite time and the vectors with @xmath660 and @xmath662 point away from @xmath365 .
the functions @xmath663 $ ] satisfy @xmath664 ; in the example of the surface of revolution mentioned in the beginning of the section we have @xmath665 and @xmath666 , where @xmath667 is the angle between @xmath659 and the foliation @xmath635 , measured from the direction pointing into the cusp .
recall the definition @xmath668 from ( [ e = cdef ] ) .
we study here how the functions @xmath669 and @xmath107 vary along a geodesic in @xmath321 .
[ l = derivs ] let @xmath670\to { { \mathbb d}}^\ast$ ] be a geodesic segment , and for @xmath671 $ ] , write @xmath672 , and @xmath673 .
then : 1 . @xmath674 ; 2 .
@xmath675 ; 3 .
@xmath676 ; this is a straightforward verification . from the definitions
, we have @xmath677 , and @xmath678 .
similarly , @xmath679 . we apply conclusion 3 of proposition [ p = c ] to get the final estimate .
we next prove that there is a clairaut - type integral for geodesic rays in @xmath680 .
recall that for @xmath681 , @xmath327 denotes the set of @xmath328 with @xmath682 .
[ p = clairaut ] if @xmath683 is sufficiently small , then there exists @xmath684 , such that for every geodesic segment @xmath685\to { { \mathbb d}}^\ast(\delta_0)$ ] , the following quasi - clairaut formula holds , for all @xmath686 $ ] : @xmath687 first note that the statement is trivially true if @xmath688 ; hence we may assume @xmath689 .
let @xmath690\to { { \mathbb d}}^\ast(\delta_0)$ ] , where @xmath683 will be specified later . as in the previous section , write @xmath691 the first main ingredient in the proof of proposition [ p = clairaut ] is the following lemma .
[ l = convexd ] for every @xmath692 , the function @xmath21 is convex along @xmath151 and strictly convex if @xmath693 , at any time @xmath113 where @xmath694 we have @xmath695 so that by lemma [ l = derivs ] , we have @xmath696 , which is positive , by proposition [ p = c ] . returning to the proof of proposition [ p = clairaut ] , let @xmath697 . we first calculate : @xmath698 and so @xmath699 .
thus there is a constant @xmath454 such that @xmath700 .
fixing @xmath701 , we have @xmath702 thus @xmath703 since @xmath21 is convex and @xmath704 , we have @xmath705 for all @xmath706 $ ] , and there is at most one @xmath707 $ ] where @xmath708 vanishes .
it follows easily that @xmath709 , which implies the conclusion .
[ p = leave ] every unit - speed geodesic in @xmath321 that enters the region @xmath327 leaves the region in time @xmath710 .
for @xmath166 a geodesic segment in @xmath321 , we consider solutions to the riccati equation : @xmath711 which is defined on a time interval containing @xmath365 .
the next lemma shows that there is a cone condition " on initial data that is preserved by solutions to ( [ e = zeta ] ) .
we use this in the next section to construct an invariant cone field for solutions to ( [ e = zeta ] ) in @xmath1 .
[ l = cone ] for every @xmath712 there exists @xmath393 such that the following holds for every geodesic segment @xmath670\to { { \mathbb d}}^\ast(\delta_0)$ ] .
let @xmath259 be a solution to ( [ e = zeta ] ) along @xmath166 . 1 . if @xmath713 , then @xmath714 for all @xmath715 $ ] , and 2 . if @xmath716 , then @xmath717 , for all @xmath715 $ ] .
we establish the lower bound first . to this end
. then @xmath719 . to show that @xmath720 , it suffices by lemma [ l = comparison ] to show that @xmath721 equivalently , it suffices to show that @xmath722 . since @xmath723 , this clearly will hold if @xmath683 is sufficiently small .
the upper bound is proved similarly .
now consider the surface @xmath1 with one puncture , satisying the hypotheses of theorem [ t = main ] .
let @xmath21 be the distance to the cusp . for @xmath393 , denote by @xmath725 the convex @xmath683-neighborhood of the cusp . in this section ,
we modify the function @xmath21 outside of a neighborhood @xmath726 and use the modified function @xmath727 to construct a @xmath87-invariant conefield on @xmath728 .
we also use the modified function @xmath727 to construct a new riemannian metric on @xmath6 , called the @xmath93 metric , that makes @xmath6 complete . having done this , we consider the flow @xmath94 on @xmath6 given by rescaling @xmath7 to have unit speed in the @xmath93 metric .
we prove that this flow is anosov in the @xmath93 metric and preserves a smooth , finite volume .
this allows us to conclude that @xmath7 is ergodic and has smooth invariant stable and unstable foliations on which @xmath7 acts with bounded distortion . in this subsection , we prove the following key technical result , which we will use to prove that a rescaled version of @xmath7 is anosov .
[ p = cones4]*[cones ] * for every @xmath413 sufficiently small , if @xmath729 is sufficiently small , then the following holds .
there exists @xmath730 , an extension @xmath731 of @xmath732 and a function @xmath733 $ ] satisfying @xmath734 for @xmath735 , @xmath736 for all @xmath38 , and such that the following holds
. let @xmath737\to s$ ] be a geodesic segment in @xmath1 .
then : 1 .
if @xmath738 is a solution to ( [ e = ricplain ] ) below then @xmath739\implies u(t ) \in \left[\frac{\chi(\gamma(t))}{\bar\delta(\gamma(t ) ) } , \frac{r+\epsilon}{\bar\delta(\gamma(t))}\right],\ ] ] for all @xmath740 $ ] , 2 .
if @xmath738 is a solution to ( [ e = ricplain ] ) below then @xmath741 \implies u(t ) \in \left[- \frac{r+\epsilon}{\bar\delta(\gamma(r ) ) } , -\frac{\chi(\gamma(t))}{\bar\delta(\gamma(t))}\right],\ ] ] for all @xmath740 $ ] . the proof is broken into a few steps .
[ l = cone2 ] for every @xmath413 , there exist @xmath743 and for every @xmath393 sufficiently small , a continuous function @xmath744 with the following properties : 1 . for every @xmath38
, we have @xmath745 .
2 . for every @xmath746
, we have @xmath747 .
3 . let @xmath737\to s$ ] be a geodesic segment in @xmath71 , and let @xmath738 be any solution to @xmath748 suppose that @xmath749 .
then @xmath750 , for all @xmath671 $ ] .
given @xmath751 , we choose @xmath681 sufficiently small according to lemma [ l = cone ] .
let @xmath752 be an upper bound on the curvature on @xmath1 , and let @xmath753 we fix @xmath754 very small ( to be specified later ) .
let @xmath755\to { { \mathbb r}}_{>0}$ ] be the affine function satisfying @xmath756 and define @xmath757 by : @xmath758 by construction , @xmath759 satisfy conditions 1 and 2 .
we check invariance of the condition @xmath760 ; to this end , let @xmath737\to s$ ] be a geodesic , and suppose that @xmath738 is a solution to ( [ e = ricplain ] ) satisfying @xmath761 . by breaking @xmath166 into pieces if necessary
, we may assume that one of the following holds : 1 .
* @xmath762\subset s\setminus \cn(\delta_0)$ ] , 2 .
* @xmath762\subset \cn(\delta_0 ) \setminus \cn(\mu\delta_0)$ ] , or 3 .
* @xmath762\subset \cn(\mu\delta_0)$ ] .
cases 1 and 3 are pretty trivial . in case 1 ,
@xmath763 , and the fact that @xmath764 implies that the condition @xmath765 is invariant . in case 3 , @xmath766 , and we apply lemma [ l = cone ] . in case 2 , we will apply lemma [ l = comparison ] to the function @xmath767 . differentiating @xmath768
, we have @xmath769 lemma [ l = comparison ] implies that @xmath770 for all @xmath671 $ ] provided that @xmath771 and @xmath772 , for all @xmath113 .
the latter is equivalent to : @xmath773 since @xmath774 , if @xmath775 and @xmath683 are sufficiently small , inequality ( [ e = u0comp ] ) will hold provided that @xmath776 since @xmath777 $ ] and @xmath778 , inequality ( [ e = etacomp ] ) holds automatically when @xmath779 . for @xmath780 , inequality ( [ e = etacomp ] ) will hold provided that for all @xmath781 $ ] , we have : @xmath782 since @xmath783 , we are reduced to proving the inequality @xmath784 to verify this , it suffices to show that the correct inequality holds at the endpoints @xmath785 and @xmath652 ; this is easily verified provided @xmath683 and @xmath786 are sufficiently small .
fix @xmath751 , and let @xmath393 and @xmath787 be given by lemma [ l = cone2 ] .
let @xmath788 .
since @xmath789 is compact , we may assume that @xmath683 ( and hence @xmath729 ) is small enough that @xmath790 fix @xmath791 close enough to @xmath792 that @xmath793 and @xmath794 .
we extend @xmath21 to a @xmath91 function @xmath795 satisfying @xmath796 for @xmath797 and @xmath798 , for @xmath799 .
we may do this so that @xmath800 , and @xmath801 in @xmath726 .
we also denote by @xmath727 the function on @xmath724 defined by @xmath802 , which is constant on the fibers of @xmath724 .
thus if @xmath803 , and @xmath804 , we have : @xmath805 and we will at times write these expressions interchangeably .
let @xmath806 be the lower cone function given by lemma [ l = cone2 ] .
define @xmath807 by @xmath808 . as with @xmath727
, we will lift @xmath809 to a function on @xmath6 and write @xmath810 .
our choice of @xmath811 ensures that the following lemma holds [ l = chi ] for all @xmath38 , we have @xmath812 .
there exists @xmath813 such that for all @xmath38 , we have @xmath814 .
the first assertion follows easily from the fact that @xmath815 and part 1 of lemma [ l = cone2 ] . if @xmath816 , then @xmath817 , and the conclusion holds with @xmath818 .
if @xmath819 , then @xmath820 , @xmath821 and @xmath822 ; thus @xmath823 .
we conclude by setting @xmath824 . using the modified cuspidal distance function @xmath727
, we now can define an upper edge to an invariant cone field for solutions to ( [ e = ricplain ] ) .
[ l = cone3 ] let @xmath727 be defined as in section [ ss = bardelta ] .
let @xmath737\to s$ ] be a geodesic segment in @xmath1 , and let @xmath738 be any solution to ( [ e = ricplain ] ) with @xmath826 .
then @xmath827 , for all @xmath671 $ ] .
this is a straightforward application of lemma [ l = comparison ] , using only the facts that @xmath828 , and @xmath829 , for all @xmath38 .
lemmas [ l = cone2 ] , [ l = chi ] and [ l = cone3 ] can be applied as well to the flow @xmath830 to obtain invariant negative cones for solutions to the equation ( [ e = ricplain ] ) .
one can do this using the same functions @xmath831 satisfying both ( 1 ) and ( 2 ) in the conclusion of proposition [ p = cones4 ] this completes the proof of the proposition .
define a new riemannian metric on @xmath724 by @xmath832 for @xmath833 .
the @xmath93 metric on @xmath6 is comparable ( i.e. bi - lipschitz equivalent ) to the induced sasaki metric for the so - called ricci metric on @xmath1 .
( the ricci metric on @xmath1 is obtained by scaling the original metric by @xmath834 . )
we briefly explain .
define a metric @xmath835 on @xmath1 by conformally rescaling the original metric , as follows : @xmath836 this is comparable to the ricci metric , since @xmath834 is comparable to @xmath837 .
we claim that the metric on @xmath6 induced by the sasaki metric for @xmath835 is comparable to @xmath838 .
here is a crude sketch of the proof .
the unit tangent bundle @xmath6 for the original metric is clearly not the @xmath835 unit tangent bundle , but angles remain the same , and so @xmath839 angular distance in the vertical fibers of @xmath6 coincides with @xmath835 angular distance . on the other hand , @xmath840-distance in the horizontal fibers of @xmath139 ( with respect to the original connection )
is the original distance scaled by @xmath841 .
thus the formulas are comparable .
as it is more convenient to work with the @xmath93 metric , we will not pursue here further the properties of the @xmath840 metric on @xmath1 , but one can prove that ( for @xmath683 sufficiently small ) it is complete , negatively curved with pinched curvature , and of finite volume . in the case where the original metric is the wp metric , the @xmath840 metric is comparable to the teichmller metric , which is the hyperbolic metric .
we will not be using the riemannian properties of the @xmath93 metric beyond completeness and finite volume .
let @xmath842 be the riemannian distance on @xmath724 induced by @xmath843 .
@xmath842 is complete . by the hopf - rinow theorem
, it suffices to show that any @xmath93-geodesic is defined for all time .
the only way in which a geodesic in @xmath6 can stop being defined is for its projection to @xmath1 to hit the cusp .
but the projection to @xmath1 of a @xmath93-geodesic is a curve that has speed @xmath21 when it is at distance @xmath21 from the cusp in the geometry of our riemannian metric @xmath844 on @xmath1 .
it is clear that such a curve can not reach the cusp in finite time .
if @xmath24 is a vector field on @xmath1 , then @xmath24 has two well - defined lifts @xmath845 and @xmath846 to vector fields on @xmath139 , the _ horizontal _ and _ vertical _ lifts , respectively .
they are defined by @xmath847 for @xmath848 .
the following formulas for lie brackets of such lifts are standard ; see @xcite .
let @xmath24 and @xmath849 be arbitrary vector fields on @xmath1 . then * @xmath850_u = 0 $ ] * @xmath851_u = ( 0 , \nabla_xy)$ ] * @xmath852_u = ( [ x , y ] , - r(x , y)u)$ ] recall the definitions of @xmath853 and @xmath854 . to simplify notation , and since the calculations that follow are only interesting in the thin part @xmath855 where @xmath856 , we will write @xmath857 , and @xmath21 for their barred counterparts in what follows .
[ l = brackets ] denote by @xmath858 the horizontal and vertical lifts , respectively , of @xmath134 and @xmath859 . then for @xmath860
, we have : 1 . @xmath861_u = [ v^h , jv^v]_u = [ v^h , v^v]_u = 0 $ ] , 2 . @xmath862_u = ( 0 , cjv)$ ] 3 . @xmath863_u = ( cjv , -r(jv , v)u)$ ] 4 .
@xmath864_u = ( 0 , -cv)$ ] this is a direct application of the previous lemma and the fact that @xmath337 = cjv$ ] from proposition [ p = c ] .
observe that @xmath865 , and @xmath866 .
we have : let @xmath24 and @xmath849 be arbitrary vector fields on @xmath1 with @xmath867 , and denote by @xmath868 their horizontal and vertical lifts
. then @xmath869 and @xmath870 in particular , the @xmath93 connection is summarized in table 1 .
* table 1 : * @xmath871 , for @xmath872 , where @xmath24 is the row vector field , + and @xmath849 is the column vector field .
the proof is a calculation using lemma [ l = brackets ] and koszul s formula : @xmath873,z\rangle_\star - \langle [ x , z],y\rangle_\star - \langle[y , z],x\rangle_\star.\end{array}\ ] ] the details can be found in @xcite .
let @xmath874 be defined as above
. then 1 .
@xmath875 , 2 .
@xmath876 , and @xmath877 , 3 .
@xmath878 , and @xmath879 4 .
@xmath880 , and @xmath881 1 . to differentiate a function @xmath882 on @xmath6 with respect to @xmath883 at @xmath884
, we parallel translate @xmath885 along the geodesic @xmath169 through @xmath886 tangent to @xmath134 to obtain @xmath887 , and then differentiate the function @xmath888 with respect to @xmath113 at @xmath889 .
since @xmath166 is a geodesic tangent to @xmath134 , the angle between @xmath890 and @xmath134 remains constant , and so @xmath891 and @xmath892 are both constant .
thus their derivatives are both zero . 2 .
let @xmath893 be the parallel translate of @xmath885 along the integral curve of the vector field @xmath859 through @xmath886 .
then @xmath894 and @xmath895 3 . to compute the derivative @xmath896 at @xmath885
, we differentiate @xmath897 at @xmath889 in the fiber over @xmath886 . thus @xmath898 and @xmath899 4 .
to compute the derivative @xmath900 at @xmath885 , we differentiate @xmath901 at @xmath889 in the fiber over @xmath886 .
the calculations are similar to those in 3 .
[ p = starnabla ] let @xmath24 be any vector field on @xmath6 with @xmath902 .
then @xmath903 in particular : 1 .
@xmath904 2 .
@xmath905 3 .
@xmath906 4 .
@xmath907 the proof is just a calculation . to see 1 , for example ,
observe that @xmath908 @xmath909 where @xmath910 .
thus @xmath911 the other formulas are proved similarly ; see @xcite for the details . as above
, let @xmath155 be the geodesic spray ; i.e. the generator of the geodesic flow on @xmath724 . define a new flow @xmath94 on @xmath724 with generator @xmath912
one might ask first whether this flow is complete ; that is , is it defined for all time @xmath913 , for each @xmath803 ?
note that the original flow @xmath7 is not complete , since it is the geodesic flow of an incomplete manifold .
the completeness of @xmath94 follows from the completeness of @xmath724 in the @xmath93-metric defined above , and the following lemma .
the vector field @xmath914 is @xmath375 , and there exists a constant @xmath530 such that for every @xmath803 , @xmath915 the flow @xmath916 preserves a finite measure @xmath917 on @xmath918 that is equivalent to liouville volume for the original metric : @xmath919 . by definition of the @xmath93 metric
, we have @xmath920 . since @xmath802 ,
the derivatives of @xmath727 have no vertical component , and corollary [ l = cuspdist ] gives that @xmath921 .
a unit horizonal vector in the @xmath93 norm is of the form @xmath922 , where @xmath923 .
thus @xmath924 for @xmath610 . proposition
[ p = starnabla ] implies @xmath925 has magnitude @xmath926 in the @xmath93 metric . a similar calculation taking higher covariant derivatives of the formulas in proposition [ p = starnabla ] and using the facts that @xmath927 , @xmath928 , @xmath929 , and ( [ e = diffdelta ] ) gives that @xmath930 for @xmath610 .
combining ( [ e = diffdotvarphi ] ) and ( [ e = diffdelta ] ) , we obtain that @xmath931 similarly , we obtain that @xmath932 , for @xmath933 .
let @xmath934 be the canonical one form on the tangent bundle @xmath935 with respect to the original metric .
then @xmath936 , for all @xmath113 , and @xmath937 on @xmath6 .
we have that : @xmath938 since @xmath939 .
thus @xmath94 preserves the smooth measure @xmath917 defined by @xmath940 . to see that @xmath941 , we use the expression for @xmath641 from ( [ e = volume ] ) and integrate : @xmath942 the flow @xmath94 is a time change of @xmath7 ;
that is , it has the same orbits , but they are traversed at a different speed , depending on the distance to the singular locus . indeed , defining the cocycle @xmath943 by the implicit formula @xmath944 we have that @xmath945 , for all @xmath803 , @xmath913 .
this gives an alternate way to see the completeness of the flow @xmath946 : the function @xmath727 clearly remains positive along orbits of @xmath946 for all time .
[ t = accelanosov ] the flow @xmath94 is an anosov flow in the @xmath93-metric .
that is , there exists a @xmath947-invariant , continuous splitting of the tangent bundle : @xmath948 and constants @xmath530 , @xmath949 such that for every @xmath803 , and every @xmath44 : * @xmath950 , and * @xmath951 . from theorem [ t = accelanosov ]
we obtain several important properties of both @xmath94 and @xmath7 .
the first is ergodicity . since volume preserving anosov flows
are ergodic , the flows @xmath7 and @xmath94 have the same orbits , and the @xmath93 volume is equivalent to ( i.e. has the same zero sets as ) the original volume on @xmath6 , we obtain : [ c = ergodic ] the flow @xmath94 is ergodic with respect to the invariant volume @xmath917 .
consequently , @xmath7 is ergodic with respect to volume . in the next corollary
we obtain a splitting of @xmath190 , invariant under @xmath87 .
@xmath87 has an invariant singular hyperbolic splitting @xmath952 with @xmath953 and @xmath954 given by intersecting @xmath955 and @xmath956 with the smooth , @xmath87-invariant bundle @xmath191 .
since the weak stable and unstable distributions of a @xmath375 anosov flow in dimension @xmath957 are @xmath82 , for some @xmath958 , we also obtain : the distributions @xmath955 and @xmath956 are @xmath82 , for some @xmath958 .
the distributions @xmath959 and @xmath960 are also @xmath82 , when measured in the @xmath93 metric .
thus in the compact part @xmath961 , the distributions @xmath959 and @xmath960 are uniformly @xmath82 .
if @xmath962 , then the vector @xmath963 points directly away from the cusp .
it is not difficult to see that the unstable manifold @xmath964 consists of the restriction of the vector field @xmath134 to the circle @xmath965 . for these vectors , the unstable bundles @xmath966 and @xmath959 coincide .
finally we obtain the key bounds on distortion for the flow @xmath7 that will be used to prove exponential mixing .
[ c = distortion ] for @xmath804 and @xmath833 denote by @xmath967 and @xmath968 the @xmath93- norm of the restriction of @xmath969 to @xmath970 and @xmath966 , respectively .
similarly define @xmath971 and @xmath972 using the bundles @xmath960 and @xmath959 .
there exist @xmath973 , @xmath974 and @xmath975 such that for every @xmath833 : 1 .
if @xmath976 and @xmath977 , then for all @xmath44 : @xmath978 and @xmath979 2 . if @xmath980 and @xmath981 , then for all @xmath44 : @xmath982 and @xmath983 the results for @xmath94 are standard properties of anosov flows . for @xmath7 , we need only note that the map induced by @xmath7 between any two @xmath984 manifolds on the same orbit is just the composition of the map induced by @xmath94 between the corresponding @xmath985 manifolds with projections along flow lines at both ends between the @xmath984 and @xmath985 manifolds .
these latter projections are uniformly @xmath81 .
it is not hard to see that the stable and unstable bundles @xmath966 and @xmath970 are not jointly integrable .
it follows that the anosov flow @xmath94 is mixing with respect to the measure @xmath917 .
this leads to the question : is @xmath94 exponentially mixing ( if , for example , @xmath683 is chosen small enough in the construction ) ? by standard arguments in smooth dynamics , to prove that @xmath94 is an anosov flow , it suffices to find nontrivial cone fields @xmath986 and @xmath987 over @xmath724 and constants @xmath530 , @xmath988 with the properties : * @xmath989 , and @xmath990 ; * @xmath991 , and @xmath992 ; and * for all @xmath44 , and all @xmath993 and @xmath994 , we have @xmath995 the derivative of @xmath94 restricted to @xmath191 has a component in the @xmath155 direction of @xmath996 owing to the time change .
we have : @xmath997 our strategy to find the cone fields is summarized in two steps . 1 .
use the properties of @xmath998 previously obtained in lemma [ l = cone2 ] to define the perpendicular components ( i.e. in @xmath191 ) of @xmath999 .
2 . using a bound on the shear term " @xmath1000 in the @xmath93 norm ,
we then define the components of @xmath1001 in the @xmath1002 direction .
we carry out these steps in the following sections . here
we fix @xmath44 and study the action of @xmath1004 at @xmath1005 .
the derivatives of @xmath1006 do not enter into these calculations ; we are essentially establishing properties of the original flow @xmath1007 ( as measured in the @xmath93-metric ) .
[ p = yzstuff ] for any @xmath803 , and any real numbers @xmath1008 satisfying @xmath1009 $ ] , the following holds . for @xmath1010 , define @xmath1011 and @xmath1012 by : @xmath1013 then : 1 .
@xmath1014 $ ] , for all @xmath1015 , and 2 . for every @xmath44 : @xmath1016 for @xmath1017 , let @xmath1018 . then , since @xmath1019 is a perpendicular jacobi field , the definition of @xmath1011 , @xmath1012 implies that @xmath1020 .
in particular , @xmath1021 .
suppose that @xmath1022 $ ]
. then @xmath1023\ ] ] holds for @xmath488 , and proposition [ p = cones4 ] implies that ( [ e = yzcone ] ) holds for all @xmath1017 .
we conclude that @xmath1024 $ ] , for all @xmath1017 .
turning to the second item in the proposition , we have that @xmath1025 which gives that @xmath1026 and so @xmath1027 since @xmath1028 , for @xmath1029 , and @xmath1030 , we have @xmath1031 we make the substitution @xmath1032 and use the fact that @xmath1033 to obtain the conclusion .
we define invariant stable and unstable cones @xmath987 and @xmath986 . the angle between @xmath986 and @xmath987 will be uniformly bounded in the @xmath93-metric , as will be the angle between either of them and @xmath914 .
we establish the properties of @xmath986 in detail ; the analogous properties for @xmath987 are obtained by the same proof , reversing the direction of time .
fix @xmath1034 to be specified later , and let @xmath1035 \ , \ & \ , |x| \leq b |y|\}\cup\{0\},\ ] ] and @xmath1036 \,\&\ ,
|x| \leq b |y|\}\cup\{0\}.\ ] ] note that since @xmath809 is bounded below away from @xmath365 , if @xmath1037 , then @xmath1038 is uniformly comparable to both @xmath1039 and @xmath1040 .
if @xmath1034 is sufficiently large , then @xmath991 , and there exists @xmath530 such that @xmath1041 for all @xmath1042 . recall that @xmath1043 , and @xmath1044 .
applying @xmath969 to @xmath1045 and using ( [ e = dpsi ] ) , we get @xmath1046 @xmath1047 @xmath1048 @xmath1049 @xmath1050 where in the last expression we ve used the abbreviation @xmath1051 .
assume that @xmath1052 and without loss of generality that @xmath1053 .
this implies that @xmath1054 $ ] and @xmath1055 .
proposition [ p = yzstuff ] implies that for any @xmath44 : @xmath1056 @xmath1057 @xmath1058 , and @xmath1059 $ ] .
now fix @xmath1060 , and let @xmath1061 .
we want to show that @xmath1062 ; i.e. that @xmath1063 $ ] and @xmath1064 . from the previous discussion
, we have @xmath1065 $ ] , and setting @xmath1066 , we also have : @xmath1067 thus we want choose @xmath1068 such that @xmath1069 , which holds if @xmath1070 a similar argument works for @xmath1071 .
finally if @xmath1052 , then @xmath1072 is uniformly comparable to @xmath1073 ; since @xmath1073 grows exponentially on the order @xmath1074 , so does @xmath1075 .
as mentioned in the introduction , to prove exponential mixing of @xmath7 , we will construct a young tower a special section to the flow whose return times have exponential tails . since orbits of @xmath7 spend only a bounded amount of time in the cuspidal region , ensuring exponential tails for the return time
is not difficult .
fix @xmath1076 sufficiently small , and denote by @xmath1077 the circle @xmath1078 .
then @xmath1079 lifts to two distinguished circles @xmath1080 in the unit tangent bundle : @xmath1081 then @xmath1082 is a closed leaf of the unstable foliation @xmath86 for @xmath7 , and @xmath1083 is a closed leaf of the stable foliation . in a neighborhood @xmath338 of @xmath1082 in @xmath6 ,
there is a well - defined projection @xmath1084 along local leaves of the weak - stable foliation @xmath1085 for @xmath7 .
since the foliation @xmath1085 is @xmath82 , the map @xmath1086 is a @xmath82 fibration . 1 .
@xmath1094 , where @xmath1095 denotes lebesgue measure on unstable leaves .
2 . for each @xmath1096
, there exists @xmath1097 such that @xmath1098 .
3 . define @xmath1099 by @xmath1100 . for each @xmath1092
there is a diffeomorphism @xmath1101 such that for all @xmath1102 : @xmath1103 4 .
@xmath1104 is a uniform contraction : @xmath1105 .
@xmath1106 is uniformly @xmath82 : @xmath1107 for all @xmath1108 .
@xmath1109 for all @xmath1110
. 7 . for each @xmath58
, we have @xmath1111 ; moreover , there exists @xmath413 such that @xmath1112 8 .
( uni condition ) for @xmath1113 , let @xmath1114 ( where defined ) and let @xmath1115 be the set of inverse branches of @xmath1116 , which satisfy @xmath1117 , for all @xmath1118 .
then there exists @xmath1119 such that , for all @xmath1120 , there exist @xmath1121 and @xmath1122 such that @xmath1123 a recent result of arajo - melbourne @xcite shows that conditions ( 1)(8 ) imply exponential mixing of @xmath7 . for @xmath1124 $ ] ,
define @xmath1125 to be the set of of @xmath1126 functions @xmath1127 such that @xmath1128 , where @xmath1129 in the language of @xcite , conditions ( 1)-(8 ) in theorem [ t = tower ] imply that we can express the ergodic flow @xmath7 as the natural extension of a @xmath82 skew product flow satisfying the uni condition .
see the discussion in @xcite after remark 4.1 .
theorem 3.3 in @xcite then applies to give that @xmath7 is exponentially mixing for a suitable function space of observables , in particular those that are @xmath375 .
a standard mollification argument gives exponential mixing for observables in @xmath1132 ( see remark 3.4 in @xcite ) .
the construction is carried out in two parts .
first , in subsection [ ss = sections ] , we isolate those orbits that leave the thick part of @xmath6 and travel deeply into the cusp .
these orbits are easily described on a topological level using the quasi - clairaut relation developed in section [ s = qc ] .
we give a precise description of the first return map to the thick part for these orbits .
next , in subsection [ ss = thick ] , we analyze the orbits beginning in @xmath1082 and decompose into pieces visiting the thin part in a controlled way .
we combine these analyses to obtain the desired decomposition of @xmath1133 in theorem [ t = tower ] .
let @xmath1136 be the torus consisting of all unit tangent vectors to @xmath1 with footpoint in @xmath1137 : @xmath1138 this torus is transverse to the vector field @xmath155 , except at the two circles @xmath1139 let @xmath1140 and @xmath1141 be the laminations of @xmath1142 obtained by intersecting leaves of the weak foliations @xmath1143 and @xmath1085 with @xmath1142 . on @xmath1144 , the laminations @xmath1140 and @xmath1141 are transverse foliations with @xmath792-dimensional leaves .
each lamination @xmath1140 and @xmath1141 has exactly one closed leaf , the curves @xmath1082 and @xmath1083 respectively , which are also unstable and stable manifolds for @xmath7 . for @xmath1146
we define two open subsets @xmath1147 and @xmath1148 of @xmath1142 as follows : @xmath1149 and @xmath1150 where @xmath654 and @xmath655 are defined by ( [ e = abdef ] ) .
note that @xmath1151 , and @xmath1152 , for all @xmath1146 .
proposition [ p = clairaut ] implies that if @xmath1155 is sufficiently small , then for all @xmath1156 , there is a well - defined first return map @xmath1157 for the flow @xmath7 , with a local inverse @xmath1158 .
these maps , where defined , are @xmath375 and preserve the foliations @xmath1140 and @xmath1141 .
fix @xmath1087 and recall that @xmath1159 . fix @xmath260 small , and let @xmath1160 .
the image of @xmath1161 under @xmath1162 is the union of two infinite rays spiraling into the unique closed stable manifold @xmath1083 in @xmath1142 .
fix another point @xmath1163 ( for example , @xmath1164 ) , and fix two points @xmath1165 , to the left and right , respectively , of @xmath1166 in @xmath1167 with respect to some fixed orientation .
let @xmath1168 , and let @xmath1169 .
the unstable manifold @xmath1170 contains an infinite ray from @xmath1171 , spiraling into @xmath1082 from the left and cutting @xmath1172 infinitely many times .
let @xmath1173 be the first intersection point of this ray with @xmath1172 ; it lies to the right of @xmath1171 , and to the left of @xmath1174 .
the points @xmath1175 define a closed curve @xmath1176 in @xmath1142 , consisting of the piece of @xmath1170 connecting @xmath1171 to @xmath1173 and the subinterval of @xmath1177 from @xmath1171 to @xmath1173 .
similarly , let @xmath1178 be the first intersection of the infinite ray @xmath1179 spiraling into @xmath1082 from the right , and let @xmath1180 be the curve constructed analogously .
the two curves @xmath1176 and @xmath1180 bound a cylindrical region @xmath1181 in @xmath1142 , which is depicted in figure 2 .
let @xmath1182 be the projection along leaves of @xmath1141 , which is simply the restriction of the projection @xmath1086 previously defined to the domain @xmath1183 .
then @xmath1184 is @xmath82 and maps the boundary curves @xmath1176 and @xmath1180 onto @xmath1082 .
the map @xmath1184 is a diffeomorphism when restricted to the interior of any interval of @xmath1141 that begins and ends in @xmath1172 and makes one revolution around @xmath1181 .
we define the section @xmath1185 similarly : it is bounded by two curves @xmath1186 and @xmath1187 , where @xmath1186 is the union of two segments of @xmath1188 and @xmath1189 , and @xmath1187 is the union of segments of @xmath1190 and @xmath1191 .
see figure 3 . by construction
, we have that @xmath1192 , @xmath1193 , and : @xmath1194 let @xmath1198 be the projection along leaves of @xmath1140 , which is the restriction of the center - unstable @xmath1199 to @xmath1200 .
the fibers of @xmath1201 are pieces of @xmath1140-unstable manifold .
let us examine the return time function for the flow on the fibers of @xmath1201 .
let @xmath1202 be a small neighborhood of @xmath1203 defined by flowing @xmath1203 under @xmath7 in a small time interval .
for @xmath1204 , let @xmath1205 be the smallest time @xmath44 satisfying @xmath1206 : @xmath1207 thus @xmath1208 , for all @xmath1209 .
let @xmath1210 and let @xmath1211 be a closed interval .
we say that @xmath1212 is a _ fundamental interval _ if the endpoints of @xmath1212 lie on the same leaf of the @xmath1141 foliation and the interior of @xmath1212 contains no points on that leaf . 1 .
@xmath1214 , and the restriction of @xmath1215 to the interior of @xmath1212 is a @xmath82 diffeomorphism , whose inverse has uniformly bounded distortion .
2 . for any @xmath1216
, we have @xmath1217 , where @xmath1218 denotes the derivative of the restriction of @xmath1219 to @xmath1212 . for a point @xmath1224 ,
let @xmath1225 be the return time of @xmath659 for @xmath7 to @xmath1181 .
note that @xmath1226 and @xmath1227 ( by proposition [ p = clairaut ] ) .
let @xmath1228 be the point where @xmath1229 .
we write @xmath1230 for @xmath1231 , respectively .
lemma [ l = derivs ] implies that @xmath1232 , and so @xmath1233 thus , since @xmath1234 by proposition [ p = c ] , we have @xmath1235 for some @xmath974 , and @xmath1236 then , since distortion is bounded on small intervals by corollary [ c = distortion ] , we have @xmath1237 for some @xmath958 , since @xmath68 .
[ l = ibcompare ] there exist @xmath974 and @xmath1238 such that the following holds . if @xmath1068 is a @xmath1239-interval in @xmath1240 with @xmath1241 , and @xmath1242 is a fundamental interval , then @xmath1243
. in particular , if @xmath1244 is sufficiently small then @xmath1245 .
let @xmath1253 and @xmath34 be given .
let @xmath1254 be small enough such that for all @xmath1255 , if @xmath1256 , then @xmath1257 ( because @xmath94 is anosov , this holds for the restriction of the @xmath93 metric to @xmath34 , which is then comparable to the original metric , since @xmath34 is compact ) . since @xmath7 is ergodic ( by corollary [ c = ergodic ] ) there exists @xmath848 whose backward orbit is dense in @xmath6 and such that @xmath1258}(u)\cap \ct_{in}(\delta_2,\eta/2 ) \neq \emptyset.\ ] ] cover the compact set @xmath34 with a finite collection of @xmath1259-balls . fix @xmath1260 such that @xmath1261 has length @xmath1262 .
if @xmath1263 is sufficiently large , then @xmath1264}(u)$ ] meets all of the @xmath1259-balls , and thus meets every @xmath1265-center unstable manifold .
this implies the conclusion , with @xmath1266 .
we assume @xmath1155 and @xmath1197 are very small , so that by corollary [ c = distortion ] distortion is at most @xmath1271 on unstable intervals of length @xmath1272 : @xmath1273 , \forall t>0.\ ] ] denote by @xmath1274 the circumference of @xmath1275 , which is less than @xmath792 if @xmath1155 is small enough , and without loss of generality assume @xmath1276 .
let @xmath1277 be the return function defined by ( [ e = deft_1 ] ) .
fix @xmath1278 the thick part defined by @xmath1279 note that @xmath1280 .
lemma [ l = meets ] implies that there exists @xmath1248 such that for any @xmath1281 , there exists @xmath1282 and @xmath1251 $ ] such that @xmath1283 .
let @xmath1284 be the maximum time needed for a piece of unstable interval to double in length under @xmath7 .
let @xmath1285 .
denote by @xmath1286 the singular set consisting of all vectors @xmath833 with @xmath1287 and such that @xmath1288 hits the cusp in time @xmath1289 ; that is : @xmath1290}(z^s).\ ] ] we begin by chopping @xmath1275 into a collection @xmath1291 of intervals of length in @xmath1292 .
we say that an open interval @xmath1293 is _ active gap interval at time @xmath1042 _ if @xmath1294}(g)\cap z^{cs}= \emptyset$ ] and @xmath1295 .
thus @xmath1291 consists of active gap intervals at time @xmath365 .
if ( a ) happens first , we chop @xmath1300 into two new gap intervals , @xmath1306 and @xmath1307 , so that @xmath1308 .
we say that @xmath1306 and @xmath1307 are _ born _ and become _ active _ at time @xmath113 and write @xmath1309 . note that @xmath1310 , where @xmath1174 is the point where the interval @xmath1300 is cut .
since distortion is bounded by @xmath1271 on intervals of length @xmath1311 , we have : @xmath1312,\;\hbox { and } \frac{|g_i|}{|g|}\in \left [ \frac38 , \frac23 \right ] , i=1,2.\ ] ] we say that @xmath1300 is _ inactive _ in the time interval @xmath1313 , and that @xmath1314 are inactive in the time interval @xmath1315 .
if ( b ) happens first , then @xmath1300 gives birth to two _ border intervals _ @xmath1316 and @xmath1317 and two gap intervals @xmath1314 as follows .
let @xmath1318 be the unique point satisfying @xmath1319 .
let @xmath1320 and @xmath1321 be the components of @xmath1322 that lie to the left and right of @xmath659 , respectively . for @xmath612 , let @xmath1323 be the largest interval satisfying : for @xmath612 , we define the _ birthday _ of @xmath1328 and @xmath1329 to be @xmath1330
. let @xmath1331 to be the first time such that @xmath1332 is an active interval .
note that since @xmath1333 meets @xmath1240 and @xmath1334 contains a fundamental interval , we have that @xmath1335 $ ] . similarly ,
any fundamental interval in @xmath1336 will return to @xmath1181 in time at most @xmath1337 .
* @xmath1306 and @xmath1307 are gap intervals that are born at time @xmath1339 $ ] , respectively . for @xmath612
, there exists @xmath1340 $ ] such that @xmath1329 is active at time @xmath1331 .
we say that @xmath1329 is _ inactive _ in the time period @xmath1341 and _ dormant _ during the period @xmath1342 .
* @xmath1316 and @xmath1317 are border intervals that are born at time @xmath1343 $ ] , respectively . for @xmath612 ,
the set @xmath1344 is a countable union of fundamental intervals , and for any @xmath1345 , we have @xmath1346 $ ] .
* since each @xmath1347 contains exactly one fundamental interval ( and no more ) , it has bounded length when it first returns to @xmath1348 : when this forward image is projected onto @xmath1349 it covers at least once , but not more than twice .
thus , assuming that @xmath1350 etc .
are small enough , we have that if @xmath1351 is the smallest time such that @xmath1352 , then @xmath1353 * since each @xmath1347 is contained in two fundamental intervals , lemma [ l = ibcompare ] implies that @xmath1354 ; since distortion is bounded by @xmath1271 on intervals of length @xmath1311 , we have : latexmath:[\[\label{e = ratio2 } starting with the intervals in @xmath1291 and applying the algorithm to all active gap intervals , we obtain for any time @xmath1042 , three disjoint collections of disjoint intervals @xmath1356 , @xmath1357 and @xmath1358 , the active , border and dormant intervals .
the set @xmath1356 consists of the gap intervals that are active at time @xmath113 , the set @xmath1357 consists of border intervals @xmath1068 with birth time @xmath1359 , and @xmath1358 are the gap intervals that are dormant at time @xmath113 .
let @xmath1360 be the collection of all gap intervals active or dormant at time @xmath113 .
let @xmath1369 .
then @xmath1300 is either active or dormant at time @xmath1370 . since an interval can not be active for more than time @xmath1371 and
can not be dormant for more than time @xmath1372 , it follows that @xmath1300 is inactive at time @xmath1373 .
it follows that @xmath1300 has a unique _ ancestor _ in @xmath1374 ; that is , there exists @xmath1375 such that @xmath1376 and @xmath1377 gives birth in the time interval @xmath1378 $ ] .
now suppose @xmath1375 .
then @xmath1377 will become active within time @xmath1337 and some point @xmath1379 will intersect @xmath1380 within time @xmath1381 .
thus during the time period @xmath1378 $ ] , the interval @xmath1377 will divide finitely many times , and at least one active piece will intersect @xmath1380 . the number of times this division can occur is uniformly bounded . as the gap evolves in the time interval @xmath1378 $ ] , it gives birth to new gaps according to rule ( a ) or ( b ) above .
the number of times that case ( a ) can apply between two occurrences of case ( b ) is bounded : if a gap @xmath1382 is produced by rule ( b ) , then by ( [ e = boundedreturn ] ) , we have @xmath1383 , where @xmath1384 is the first time @xmath1382 returns to @xmath1181 . in @xmath1348 ,
the derivative @xmath87 is bounded above , and so any active interval meeting @xmath1348 can divide a bounded number of times before some descendent meets @xmath1380 ( which happens within time @xmath1385 ) .
thus the number of times ( a ) can apply within two occurrences of ( b ) is uniformly bounded .
we conclude that @xmath1386 , with @xmath1387 , where @xmath1388 , and @xmath1389 .
moreover , there exists @xmath299 , independent of @xmath1390 such that @xmath1391 .
combined with ( [ e = ratio1 ] ) and ( [ e = ratio2 ] ) , this implies that there exists @xmath1392 such that @xmath1393 thus @xmath1394 ; since @xmath1395 , we obtain the conclusion .
we create a countable collection @xmath1399 of intervals @xmath1400 as follows : we decompose each @xmath1401 into a countable union @xmath1402 such that for each @xmath1110 , @xmath1403 is a fundamental interval .
then we set @xmath1404 note that @xmath1405 , and so conclusion ( 1 ) of theorem [ t = tower ] holds .
we extend the definition of @xmath1406 to intervals in @xmath1399 in the natural way : if @xmath1407 , we set @xmath1408 . for @xmath1409 , let @xmath1410 then @xmath1411 is the the minimal time @xmath1412 such that @xmath1413 .
we now define the return time function @xmath1421 .
recall the projection @xmath1422 along the leaves of @xmath1141 .
the fibers of @xmath1184 are local @xmath1141 manifolds . over each point
@xmath1423 there lies a unique point @xmath1424 such that @xmath1425 , for some small value of @xmath1426 .
let @xmath1427 .
[ l = barsigma ] the function @xmath1428 is uniformly @xmath82 , and @xmath1429 is a @xmath82 manifold .
the map @xmath1430 is a @xmath82 diffeomorphism .
the manifold @xmath1431 is @xmath82 foliated by local @xmath84-leaves : @xmath1432 and the projection @xmath1433 along these local leaves is a @xmath82 submersion .
we define @xmath1434 by @xmath1435 ; it has the property that @xmath1436 .
lemma [ l = barsigma ] implies that for each @xmath1096 , there exists a unique @xmath1097 namely , @xmath1437 such that @xmath1438 , giving conclusion ( 2 ) . for @xmath1439
, we define @xmath1440 to be the inverse of the map @xmath1441 .
this is well - defined , because @xmath1442 since @xmath1443 is a fundamental interval .
since @xmath1184 is a submersion , and the map @xmath1444 is a diffeomorphism from @xmath1400 onto its image , the composition is a diffeomorphism from @xmath1400 to @xmath1133 .
thus its inverse @xmath1440 is a diffeomorphism .
this establishes conclusion ( 3 ) .
note that @xmath1445 , and so conclusion ( 4 ) holds .
conclusion ( 5 ) follows from the facts that @xmath7 has bounded distortion and the map @xmath1086 is uniformly @xmath82 .
indeed note that the map @xmath1446 can also be expressed in the following way .
we fix some point @xmath1447 and consider the image @xmath1448 under the constant time flow @xmath1449 .
this is a piece of unstable manifold that meets @xmath1431 .
the map @xmath1446 is just the composition @xmath1450 of this flow with the center - stable projection @xmath1451 , defined in the beginning of the section .
this latter projection is a uniformly @xmath82 submersion and a local diffeomorphism when restricted to local unstable manifolds , since the foliation @xmath1085 is uniformly @xmath82 .
thus @xmath1104 is uniformly @xmath82 .
let s examine the map @xmath1452 .
again fix a point @xmath1447 , and consider the image @xmath1453 , which is a piece of unstable manifold meeting @xmath1181 at the point @xmath1454 .
it follows that there is a uniformly bounded @xmath1455 function @xmath1456 such that @xmath1457 sends the piece of unstable manifold @xmath1458 to @xmath1459 .
then @xmath1460 , and so @xmath1461 .
thus @xmath1462@xmath1463 @xmath1464 the derivatives @xmath1465 and @xmath1466 are uniformly bounded . since @xmath1467
, we obtain that there exists a uniform constant @xmath974 such that @xmath1468 , for all @xmath1110 .
this gives conclusion ( 6 ) .
since the function @xmath12 is bounded , lemma [ l = r_0bound ] implies that for each @xmath58 , we have @xmath1469 since @xmath1470 , this gives conclusion ( 7 ) of theorem [ t = tower ] , where @xmath413 is chosen so that @xmath1471 . finally we verify that the uni condition in conclusion ( 8) holds
this is a direct consequence of the fact that @xmath7 preserves a contact @xmath792-form @xmath934 , which implies that the foliations @xmath84 and @xmath86 are not jointly integrable .
the details are carried out in lemma 12 of @xcite ( in the axiom a context ) and lemma 4.2 and corollary 4.3 of @xcite ( close to the current context ) .
v. arajo and i. melbourne , _ exponential decay of correlations for nonuniformly hyperbolic flows with a @xmath82 stable foliation , including the classical lorenz attractor , _ preprint .
arxiv:1504.04316 ( 2015 ) . | we establish exponential mixing for the geodesic flow @xmath0 of an incomplete , negatively curved surface @xmath1 with cusp - like singularities of a prescribed order . as a consequence
, we obtain that the weil - petersson flows for the moduli spaces @xmath2 and @xmath3 are exponentially mixing , in sharp contrast to the flows for @xmath4 with @xmath5 , which fail to be rapidly mixing . in the proof
, we present a new method of analyzing invariant foliations for hyperbolic flows with singularities , based on changing the riemannian metric on the phase space @xmath6 and rescaling the flow @xmath7 . |
an old problem in von neumann algebra theory is the question of whether each separable von neumann algebra @xmath1 is singly generated .
a single generator @xmath18 leads to two self - adjoint generators @xmath19 and any pair @xmath20 of self - adjoint generators yields a single generator @xmath21 .
thus the single generation problem has an equivalent formulation as the existence of two self - adjoint generators .
earlier work in this area solved all cases except for the finite von neumann algebras , @xcite .
here there has been progress in special situations , @xcite , but a general solution is still unavailable .
recently junhao shen , @xcite , introduced a numerical invariant @xmath22 , and was able to show that single generation for @xmath23 factors was a consequence of @xmath24 .
he proved that @xmath25 for various classes of @xmath26factors , giving a unified approach to the single generation of @xmath26factors with cartan masa , with property @xmath27 and those factorising as tensor products of @xmath26factors .
his work settled some previously unknown cases as well as giving a unified approach to various situations that had been determined by diverse methods .
it should be noted that 0 is the only value of shen s invariant that is currently known .
if strictly positive values are possible , then corollary [ scaling.nonsingle ] guarantees examples of separable @xmath26factors which are not singly generated . in this paper ,
our purpose is to undertake a further investigation of this invariant , and to relate it to a quantity @xmath28 which counts the minimal number of generators for @xmath3
. a related quantity @xmath29 counts the minimal number of self - adjoint generators , and there is a parallel invariant @xmath30 to @xmath31 which has a similar definition ( given below ) but which restricts attention to self - adjoint generating sets .
the contents of the paper are as follows .
the second section gives the definitions of @xmath22 and @xmath32 in terms of generating sets and finite decompositions of 1 as sums of orthogonal projections .
this is a slightly different but equivalent formulation of the original one in @xcite .
these are related by the inequalities @xmath33 , although it is shown subsequently that @xmath34 for all @xmath26factors @xmath3 .
the main result of the third section is that the relation @xmath35 for @xmath26factors @xmath3 implies generation by @xmath6 self - adjoint elements .
the case @xmath36 is of particular interest since single generation is then a consequence of @xmath37 .
the fourth section relates the generator invariant of a @xmath26factor @xmath3 to that of a compression @xmath38 . if @xmath39 , then @xmath40 . up to isomorphism
, @xmath41 can be uniquely defined as @xmath38 for any projection @xmath42 with @xmath39 , @xmath43 . in a standard way
, @xmath41 can be defined for any @xmath10 as @xmath44 where @xmath45 is any integer greater than @xmath46 , and @xmath47 is a projection of trace @xmath48 . in this more general situation , the scaling formula @xmath49 for @xmath10 also holds .
the subsequent section contains some consequences of the scaling formula , and also establishes it for the related invariant @xmath30 .
this requires a more indirect argument since the method of passing between generating sets for @xmath3 and for @xmath41 does not preserve self - adjointness and so can not apply to @xmath50 although it is suitable for @xmath51 .
the equality @xmath52 is also established in this section .
the sixth section is concerned with finite index inclusions @xmath53 of @xmath26factors .
the main results are that @xmath54 and that @xmath55 , where @xmath56 is the basic construction and @xmath57 denotes the index @xmath58 $ ] . a standard result of subfactor theory is that @xmath3 is the basic construction @xmath59 for an index @xmath57 inclusion @xmath60 , so two of these basic constructions scale @xmath61 by @xmath62 .
this suggests the formula @xmath63 , but this is still an open problem .
section [ free ] concentrates on free group factors and their generalisations , the interpolated free group factors . for @xmath64 $ ]
, the formula @xmath65 is established , where @xmath66 is a fixed constant in the interval @xmath67 $ ] .
this leads to two possibilities , depending on the value of @xmath66 .
if @xmath68 , then @xmath69 is singly generated for all @xmath70 , while if @xmath71 , then the free group factors are pairwise non - isomorphic , being distinguished by the generator invariant .
the paper concludes with a discussion of voiculescu s modified free entropy dimension @xmath72 , where @xmath73 is a finite generating set for @xmath3 .
a quantity @xmath74 is introduced in the second section on the way to defining @xmath31 .
the main results of the last section are the inequalities @xmath75 for general finite generating sets , and the stronger form @xmath76 for generating sets of self - adjoint elements .
these have the potential for providing lower bounds for @xmath31 . finally , a word on notation . for a subset @xmath73 of a von neumann algebra @xmath3
, @xmath77 will denote the von neumann algebra generated by @xmath73 .
it is not assumed that @xmath77 automatically contains the identity of @xmath3 .
for example , @xmath78 for a projection @xmath42 .
the main focus of the paper is on @xmath26factors .
however we define the generator invariant and establish basic results in the context of diffuse finite von neumann algebras with a fixed faithful trace @xmath79 , which is normalised with @xmath80 .
diffuse finite von neumann algebras will be denoted by @xmath1 , while @xmath3 is reserved for @xmath81 factors .
[ prelim.defp ] let @xmath82 be a finite von neumann algebra with fixed trace .
let @xmath83 ( or @xmath84 when the underlying algebra is unclear ) denote the collection of all finite sets of mutually orthogonal projections in @xmath1 which sum to @xmath85 .
an important subclass of @xmath83 is the collection @xmath86 of those @xmath87 which consist of projections of equal trace .
note that in the context of a finite factor @xmath3 , @xmath88 can be diagonalised in the sense that there exist matrix units @xmath89 in @xmath3 satisfying @xmath90 for all @xmath91 . in this factor
context , the elements of @xmath86 are referred to as the _ diagonalisable _ elements of @xmath83 .
. then @xmath93 _ refines _
@xmath87 , written @xmath94 , when every @xmath95 is a sum of elements of @xmath93 .
the sets @xmath87 and @xmath93 are _ equivalent _ , written @xmath96 , when there is a unitary @xmath97 with @xmath98 . note that in a @xmath26factor , @xmath96 if , and only if , the multiset of the traces of the elements of @xmath87 is the same as the multiset of the traces of the elements in @xmath93 .
the ordering @xmath94 defined above is chosen so that the map @xmath99 below is order preserving .
this will be established in lemma [ prelim.refine ] .
[ prelim.defi ] let @xmath1 be a finite von neumann algebra . given @xmath100 and @xmath101 , define @xmath102 for a finite subset @xmath103 and @xmath101 , define @xmath104 and @xmath105 on occasion , the notation @xmath106 and @xmath107 will emphasise the algebra @xmath1 and hence the choice of trace .
the definition of @xmath108 given above is formally different from that of shen ( * ? ? ?
* definition 2.1 ) , in that shen only considers families from @xmath86 and performs the limiting procedure in a slightly different order .
nevertheless , the resulting invariant @xmath109 defined in definition [ prelim.defg ] below agrees with ( * ? ? ?
* definition 2.1 ) in the case of diffuse von neumann algebras . before proceeding ,
a few elementary observations are recorded .
[ prelim.remarksi ] 1 .
the inequality @xmath110 holds for all finite subsets @xmath73 in @xmath1 and all @xmath101 .
2 . if @xmath73 is a finite subset of @xmath1 , then define @xmath111 .
the equality @xmath112 holds for all @xmath101.[prelim.remarksi.star ] 3 .
[ prelim.remarksi.commutant ] let @xmath101 be a set of @xmath113 projections .
the estimate @xmath114 is valid for each @xmath115 . in particular , if @xmath116 , then @xmath117 4 . if @xmath118 and @xmath119 then @xmath120 writing @xmath121 in this way is often useful in calculations , an example being ( * ? ? ?
* theorem 4.1 ) .
5 . for any @xmath122 and @xmath101 , @xmath123 6 .
if @xmath73 is a finite generating set of @xmath1 , then so also is @xmath124 and , furthermore , @xmath125 this follows from the preceding remark .
take @xmath126 and @xmath127 for each @xmath128 to obtain @xmath129 .
now let @xmath130 and @xmath131 so that ( [ prelim.remarksi.1 ] ) and item [ prelim.remarksi.star ] above combine to give @xmath132 [ prelim.refine ] let @xmath1 be a finite von neumann algebra .
consider a finite subset @xmath103 and @xmath92 with @xmath94 .
then @xmath133 take @xmath100 and two pairs of orthogonal projections @xmath134 and @xmath135 in @xmath1 . if @xmath136 , then @xmath137 the result now follows by induction .
in many applications it will be useful to know that the infimum defining @xmath138 in ( [ prelim.defi.1 ] ) can be taken through @xmath86 . recall that a separable diffuse abelian von neumann algebra @xmath139 is isomorphic to @xmath140 $ ] and , if @xmath139 is equipped with a trace , then this isomorphism can be chosen so that the trace is given by @xmath141 on @xmath140 $ ] .
[ prelim.diagonalisable ] let @xmath1 be a diffuse von neumann algebra and let @xmath73 be a finite subset of @xmath1 .
for each @xmath142 , @xmath143 fix @xmath144 and find @xmath145 with @xmath146 let @xmath139 be a diffuse abelian subalgebra of @xmath1 with @xmath147 , and
let @xmath148 by approximating the family @xmath149 by rational numbers with common denominator @xmath150 , there exists @xmath151 such that if @xmath152 , then there are @xmath153 with @xmath154 for each @xmath91
. choose projections @xmath155 in @xmath139 with @xmath156 and @xmath157 for each @xmath91 .
let @xmath158 , so @xmath159 .
let @xmath160 .
thus @xmath161 whenever @xmath162 .
hence @xmath163 now @xmath164 by inequality ( [ prelim.diagonalisable.1 ] ) , the estimate @xmath159 and the choice of @xmath165 in ( [ prelim.diagonalisable.21 ] ) .
finally , refine @xmath166 to find a family @xmath167 in @xmath86 with @xmath150 elements .
lemma [ prelim.refine ] gives @xmath168 by ( [ prelim.diagonalisable.20 ] ) .
since this can be done for any @xmath152 , and @xmath169 was arbitrary , the result follows .
these preliminaries allow the _ generator invariant _ of a diffuse finite von neumann algebra to be defined .
[ prelim.defg ] if @xmath1 is a finitely generated diffuse finite von neumann algebra , define the _ generator invariant _ @xmath109 by @xmath170 and the _ hermitian , ( or self adjoint ) , generator invariant _ ,
@xmath171 by @xmath172 if @xmath1 is not finitely generated , then define @xmath173 .
[ prelim.selfremark ] by item [ prelim.remarksi.sa ] of remarks [ prelim.remarksi ] , the inequalities @xmath174 hold for any diffuse finite von neumann algebra @xmath1 . in theorem [ selfad.main ] it will be shown that @xmath175 for all @xmath26factors @xmath3 .
theorem 4.1 of @xcite states that if a @xmath26factor @xmath3 has @xmath176 , then it is generated by a projection and a hermitian element .
it is subsequently remarked that the same proof shows that a @xmath26factor @xmath3 with @xmath177 is singly generated .
the theorem below strengthens this result to consider @xmath26factors which may not be singly generated , with shen s remark arising from the case @xmath36 .
the basic idea is the same combinatorical counting argument of @xcite which dates back through @xcite and @xcite to work in the 1960 s by douglas , pearcy and wogen , @xcite .
a good account of this material can be found in the book by topping , @xcite . recall that if @xmath89 are matrix units for the @xmath178 matrices , @xmath179 , then the self adjoint elements @xmath180 and @xmath181 generate @xmath179 . [ number.hermitians ]
let @xmath3 be a separable @xmath182 factor and @xmath142 .
if @xmath183 , then @xmath3 is generated by @xmath6 hermitian elements .
suppose that @xmath183 for some @xmath142 .
there exists @xmath184 such that @xmath185 for all @xmath186 . by lemma [ prelim.diagonalisable ]
, there is a finite set @xmath187 of generators for @xmath3 , some @xmath186 and a diagonalisable family @xmath188 such that @xmath189 choose a set of matrix units @xmath89 in @xmath3 with @xmath190 for each @xmath91 .
define a set of triples @xmath191 so the definition of @xmath121 gives @xmath192 for each @xmath193 , let @xmath194 be the set of triples @xmath195 with @xmath196 and let @xmath197 be the set of triples @xmath198 with @xmath199
. then @xmath200 by the choice of @xmath186 , @xmath201 from ( [ number.hermitians.1 ] ) and ( [ number.hermitians.3 ] ) .
decompose @xmath202 into a partition @xmath203 with each @xmath204 and find , for each @xmath13 , injections @xmath205 , written as @xmath206 .
this really defines @xmath45 maps indexed by @xmath13 , but as the domains @xmath207 and ranges @xmath194 are disjoint , these are regarded as a single map from @xmath208 to @xmath209 .
define self
adjoint operators @xmath210 for @xmath211 . if @xmath211 and @xmath212 , then @xmath213 appears exactly once in the set @xmath214 as @xmath194 is disjoint from its transpose on the first two variables and the map @xmath205 is an injection . for such @xmath215 , @xmath216 as @xmath217 , equation ( [ number.hermitians.10 ] ) and the definition of @xmath202 in ( [ number.hermitians.20 ] )
imply that @xmath218 for each @xmath219 .
thus the set @xmath220 generates the @xmath26factor @xmath3 .
finally note that @xmath221 so that @xmath222 for any @xmath223 by the spectral theorem . in this way @xmath3
is generated by the @xmath6 hermitian elements @xmath224 as required .
[ number.remark ] in corollary [ ngenerators ] , it will be shown that if @xmath3 is generated by @xmath6 hermitian elements , then @xmath225 , which almost gives a converse to this result .
the remaining gap is summarised in the following question .
let @xmath3 be a @xmath26factor . if @xmath226 for some @xmath142 , is @xmath3 generated by @xmath6 hermitians ?
this section examines the behaviour of @xmath227 under amplifications and compressions .
theorem [ scaling.formula ] establishes that @xmath228 for all @xmath26factors @xmath3 and @xmath10 .
note that it is a consequence of equation ( [ scaling.1 ] ) that @xmath3 is finitely generated if and only if @xmath41 is finitely generated for all @xmath10 .
this result is deduced in lemma [ scaling.finitegen ] from the lemmas that will be needed to prove equation ( [ scaling.1 ] ) .
[ scaling.second ] let @xmath3 be a separable @xmath26factor and let @xmath142 .
then @xmath229 assume that @xmath230 as otherwise the lemma is vacuous .
let @xmath144 .
by lemma [ prelim.diagonalisable ] , there is a finite generating set @xmath73 for @xmath3 , some @xmath231 and a diagonalisable @xmath232 such that @xmath233 .
find matrix units @xmath234 in @xmath3 with @xmath190 .
define @xmath235 for @xmath236 .
the family @xmath237 is a system of matrix units in @xmath3 .
consider @xmath238 , a projection of trace @xmath239 so @xmath240 is a representative of @xmath241 .
the von neumann algebra @xmath240 is generated by @xmath242 using induction on the equation @xmath243 see for example ( * ? ? ?
* lemma 5.2.1 ) .
consider @xmath244 , a family of orthogonal projections in @xmath240 with sum @xmath245 , so @xmath93 is a diagonalisable element of @xmath246 . for @xmath128 , @xmath247 and @xmath236
the relation @xmath248 is equivalent to @xmath249 since @xmath250 let @xmath251 be the number of quintuples @xmath252 with this property .
since each projection in @xmath93 has trace @xmath253 in @xmath240 , it follows that @xmath254 whereas @xmath255 therefore @xmath256 from which the lemma follows .
[ scaling.second.remark ] it does not appear to be possible to obtain @xmath257 using the methods of lemma [ scaling.second ] , since a self adjoint generating set @xmath73 leads to a generating set @xmath242 for @xmath240 which is not necessarily self adjoint . nevertheless inequality ( [ scaling.second.remark.1 ] ) is true , as will be established in corollary [ selfad.scaling ] from the scaling formula of theorem [ scaling.formula ] and theorem [ selfad.main ] .
the next lemma provides one inequality in ( [ scaling.1 ] ) and also the parallel inequality for the hermitian generator invariant .
[ scaling.first ] let @xmath3 be a separable @xmath26factor and @xmath43 .
then @xmath258 assume that @xmath41 is finitely generated , otherwise there is nothing to prove . fix a projection @xmath42 of trace @xmath46 so that @xmath38 is a representative of @xmath41 and let @xmath73 be an arbitrary finite set of generators for @xmath38 .
for @xmath144 , find orthogonal projections @xmath259 such that @xmath260 the factor @xmath261 arises in ( [ scaling.first.1 ] ) as @xmath262 for @xmath263 .
let @xmath264 be the maximal integer such that @xmath265 and find a family of orthogonal projections @xmath266 in @xmath3 such that : a. @xmath267 ; b. @xmath268 , for @xmath269 ; c. @xmath270 . in this way @xmath271 with @xmath272
possibly the zero projection .
let @xmath273 and find partial isometries @xmath274 such that : 1 .
@xmath275 , for each @xmath276 ; 2 .
@xmath277 , for @xmath269 ; 3 .
@xmath278 is a subprojection of @xmath279 of the form @xmath280 for some @xmath281 and some @xmath282 .
it will now be shown that @xmath283 generates the @xmath26factor @xmath3 .
since @xmath73 generates @xmath38 , it follows that @xmath284 .
the relations @xmath285 for @xmath286 , then imply that @xmath287 and so @xmath288 generates @xmath3 . the final step is to estimate @xmath289 .
let @xmath290 be a set of mutually orthogonal projections with sum @xmath291 that refines @xmath292 such that @xmath293 thus @xmath294 and @xmath295 .
extend @xmath290 to a family @xmath296 of projections .
then @xmath297 and the sum of the orthogonal projections in @xmath298 is @xmath85 so @xmath299 .
fix @xmath300 and note that if @xmath301 for some @xmath302 and @xmath303 , then @xmath304 ( as otherwise @xmath305 ) and @xmath306 ( as otherwise @xmath307 ) ; when these conditions hold @xmath308 ( as @xmath309 ) .
thus @xmath310 and lemma [ prelim.refine ] and ( [ scaling.first.1 ] ) imply that @xmath311 finally @xmath312 the inequality @xmath313 then follows , as @xmath144 was arbitrary and @xmath73 was an arbitrary finite generating set for @xmath38
. only minor modifications are necessary for the hermitian case .
assume that each element of @xmath73 is self - adjoint , and replace the partial isometries @xmath314 in by the self - adjoint elements @xmath315 and @xmath316 . then the same arguments lead to the inequality @xmath317 for any finite generating set @xmath73 consisting of self - adjoint elements
. the factor of 4 , which was not present in the counterpart inequality , arises from the estimate @xmath318 this change reflects the replacement of @xmath314 by @xmath315 and @xmath316 in the generating set @xmath288 of .
[ scaling.finitegen ] let @xmath3 be a @xmath26factor and let @xmath10 .
then @xmath41 is finitely generated if , and only if , @xmath3 is finitely generated . without loss of generality ,
suppose that @xmath43 , otherwise replace @xmath46 by @xmath319 .
lemma [ scaling.first ] shows that if @xmath41 is finitely generated so too is @xmath3 .
conversely , if @xmath3 is finitely generated , choose @xmath142 with @xmath320 .
lemma [ scaling.second ] shows that @xmath241 is finitely generated .
let @xmath321 .
since @xmath322 and @xmath323 is finitely generated , another application of lemma [ scaling.first ] shows that @xmath41 is finitely generated .
[ scaling.formula ] let @xmath3 be a separable @xmath26factor .
for each @xmath10 , @xmath324 this formula will be established by considering successively the cases @xmath325 , @xmath326 and @xmath327 . by lemma [ scaling.finitegen ] ,
@xmath3 is infinitely generated if , and only if , @xmath41 is infinitely generated .
assume then that both @xmath3 and @xmath41 are finitely generated .
for @xmath142 and any separable @xmath26factor @xmath3 , the equation @xmath328 is a consequence of the inequalities of lemma [ scaling.second ] and lemma [ scaling.first ] .
let @xmath329 be a rational and apply ( [ scaling.formula.2 ] ) twice .
this gives @xmath330 and @xmath331 this proves the theorem for rational @xmath46 . for arbitrary @xmath43 ,
consider @xmath322 such that @xmath332 .
then @xmath333 by applying lemma [ scaling.first ] twice and the rational case above .
hence @xmath334 .
if @xmath335 , then @xmath336 this deals with all possible cases and so completes the proof .
this section contains an initial collection of deductions from the scaling formula .
results regarding the free group factors and estimates involving free products are reserved to section [ free ] .
note that there is currently no example for which the hypothesis of the first corollary is known to hold .
[ scaling.nonsingle ] if there exists a separable @xmath26factor @xmath3 such that @xmath337 , then there exist separable @xmath26factors which are not singly generated .
if the separable @xmath26factor @xmath3 satisfies @xmath338 , then this is already an example with no finite set of generators .
thus the additional assumption that @xmath339 can be made , in which case @xmath340 by theorem [ scaling.formula ] .
it is immediate from the definition of @xmath341 that any singly generated factor has @xmath342 , and so @xmath41 is not singly generated for sufficiently small values of @xmath46 .
[ scaling.fundamentalgroup ] any finitely generated @xmath26factor @xmath3 with non trivial fundamental group @xmath343 has @xmath344 and is singly generated .
if @xmath345 , then @xmath346 by theorem [ scaling.formula ] as @xmath347 . since @xmath230 , it follows that @xmath344 so @xmath3 is singly generated by theorem [ number.hermitians ] or @xcite .
the following lemma leads to the upper bound for @xmath227 in remark [ number.remark ] , and which will finally be established in corollary [ ngenerators ] . a simple modification yields the odd integer case of corollary [ ngenerators ] without the need for further work . [ selfad.ngenerators ]
let @xmath3 be a @xmath26factor which is generated by @xmath45 hermitian elements for some @xmath142
. then @xmath348 .
let @xmath349 be @xmath45 hermitian elements generating @xmath1 .
let @xmath144 and let @xmath139 be a masa containing @xmath350 . since @xmath351 $ ] , there exists some diagonalisable @xmath352 with @xmath353 .
then @xmath354 from item [ prelim.remarksi.commutant ] of remarks [ prelim.remarksi ] . for each @xmath355 ,
the estimate @xmath356 gives @xmath357 and the result follows . the next lemma is the easy direction of theorem [ selfad.main ] . [ selfad.easy ]
let @xmath3 be a @xmath26factor .
then @xmath358 suppose that @xmath3 is finitely generated , as otherwise the inequality is vacuous .
let @xmath144 .
use lemma [ prelim.diagonalisable ] to find a self adjoint set of generators @xmath349 for @xmath3 and a diagonalisable set of projections @xmath116 with @xmath359 .
take @xmath360 for some @xmath361 and choose a diagonalisable refinement @xmath362 of @xmath87 . for @xmath363
let @xmath364 then let @xmath365 and @xmath366 .
since @xmath367 for each @xmath368 , the finite set @xmath369 generates @xmath3 .
moreover , @xmath370 since the regions in @xmath371 depending on @xmath288 , @xmath372 and @xmath373 are disjoint . this gives the estimate @xmath374 as @xmath375 for all @xmath368 , item [ prelim.remarksi.commutant ] of remarks [ prelim.remarksi ] gives @xmath376 .
hence @xmath377 as @xmath94 .
the lemma follows .
the generator invariant can now be related to the hermtian generator invariant .
[ selfad.main ] let @xmath3 be a @xmath26factor
. then @xmath378 by lemma [ selfad.easy ] , it suffices to prove that @xmath379 for a finitely generated @xmath26factor @xmath3 .
let @xmath144 and choose @xmath380 with @xmath381 such that @xmath382 write @xmath383 so that @xmath43 . the scaling formula ( theorem [ scaling.formula ] ) gives @xmath384 by the generation theorem ( theorem [ number.hermitians ] ) , @xmath41 is generated by @xmath6 hermitian elements so that @xmath385 by lemma [ selfad.ngenerators ] .
lemma [ scaling.first ] on scaling by small @xmath46 implies that @xmath386 this proves the theorem .
the scaling for the generator invariant can now be extended to the hermitian case , as stated in remark [ scaling.second.remark ] . [ selfad.scaling ]
if @xmath3 is a separable @xmath26factor and @xmath10 , then @xmath387 this follows directly from theorems [ scaling.formula ] and [ selfad.main ] .
the next corollary gives the partial converse to the generation theorem which was indicated in remark [ number.remark ] .
[ ngenerators ] if @xmath45 is the minimal number of hermitian generators of a finitely generated @xmath26factor @xmath3 , then @xmath388 lemma [ selfad.ngenerators ] and theorem [ selfad.main ] give @xmath389 which yields the second inequality . if @xmath390 , then @xmath391 so that the generation theorem ( theorem [ number.hermitians ] ) gives @xmath392 hermitian generators for @xmath3 , contradicting the minimality of @xmath45 .
this gives the first inequality .
[ ngenerators.group ] if @xmath298 is a countable discrete i.c.c . group generated by @xmath45 elements , then @xmath393 .
if @xmath298 is generated by @xmath394 , then there are hermitian elements @xmath395 in the @xmath26factor @xmath396 with @xmath397 for all @xmath91 by the spectral theorem , since each @xmath398 is normal .
the previous corollary completes the proof .
the following result can be found in @xcite .
it dates back to douglas and percy @xcite and a proof can also be found in ( * ? ? ? * chapter 15 ) .
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ if a von neumann algebra @xmath1 is generated by @xmath113 self
adjoint elements , then @xmath399 is generated by two self
adjoint elements , equivalent to being singly generated .
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ in the case of @xmath26factors , this can be strengthened .
[ scaling.matrices ] if @xmath3 is a @xmath26factor generated by @xmath400 self adjoint elements , then @xmath401 is generated by two self
adjoint elements , equivalent to being singly generated . by corollary [ ngenerators ]
@xmath402 the scaling formula ( theorem [ scaling.formula ] ) gives @xmath403 the result follows from the generation theorem ( theorem [ number.hermitians ] ) or @xcite .
corollary [ mingen ] gives a formula describing the generator invariant in terms of the minimal number of generators required to generate compressions of the von neumann algebra , based on the following : let @xmath3 be a @xmath26factor .
write @xmath404 for the minimal number of generators of @xmath3 if @xmath3 is finitely generated and let @xmath405 if @xmath3 is not finitely generated .
the quantity @xmath406 has a similar definition in terms of the number of self - adjoint generators .
the following simple estimates will be used below .
1 . @xmath407 and @xmath408.[mingen.easy.1 ] 2 .
@xmath409.[mingen.easy.2 ] [ mingen ] let @xmath3 be a separable @xmath26factor .
then @xmath410 and @xmath411 by lemma [ scaling.finitegen ] , assume that @xmath3 is finitely generated .
the first step is to establish ( [ mingen.1 ] ) .
the scaling formula ( theorem [ scaling.formula ] ) and the estimate [ mingen.easy.1 ] above give @xmath412 for all @xmath113 .
for @xmath144 find @xmath184 with @xmath413 .
for @xmath186 , find @xmath142 with @xmath414 the scaling formula gives @xmath415 so the generation theorem ( theorem [ number.hermitians ] ) ensures that @xmath416 is generated by @xmath6 hermitians .
mingen.easy.2 ] preceding the corollary gives @xmath417 so that @xmath418 this gives equation ( [ mingen.1 ] ) . for ( [ mingen.2 ] ) , note that the estimate [ mingen.easy.2 ] above gives @xmath419 the result follows from ( [ mingen.1 ] ) and theorem [ selfad.main ] .
the last result of this section records that the previous corollary is part of a general phenomenon regarding invariants of @xmath26factors which scale and are bounded by the numbers of generators involved .
the proof is omitted .
suppose that @xmath420 is an invariant of a @xmath26factor which satisfies @xmath421 for all @xmath10 .
if there exist constants @xmath422 and @xmath423 with @xmath424 for all separable @xmath26factors @xmath3 , then @xmath425
this brief section examines the generator invariant for finite index inclusions of @xmath26factors .
recall from @xcite that if @xmath426 is a unital inclusion of @xmath26factors and @xmath427 is the orthogonal projection from @xmath428 onto @xmath429 , then the basic construction @xmath430 is the von neumann subalgebra of @xmath431 generated by @xmath3 and @xmath427 and @xmath432 , where @xmath433 is the usual modular conjugation operator on @xmath428 given by extending the map @xmath434 .
this last equation holds for any von neumann subalgebra of a @xmath26factor @xmath3 , and implies that @xmath430 is a factor precisely when the same is true for @xmath1 .
recall also that @xmath435 .
in this situation , @xmath1 is said to be a _ finite index _
subfactor if @xmath430 is a type @xmath26factor , and one formulation of the index @xmath58 $ ] is given by @xmath58={\textrm{tr}}(1)$ ] , where @xmath436 is the unique trace on @xmath430 normalised with @xmath437 .
[ subfactor.bcon ] suppose that @xmath426 is a unital inclusion of @xmath26factors with @xmath58<\infty$ ] .
then @xmath438 note that in this lemma , @xmath439 is computed with respect to the trace @xmath440 , where @xmath441 $ ] , since this is the trace on the basic construction algebra @xmath430 , normalised to take the value @xmath85 at the identity .
take @xmath144 , a finite generating set @xmath73 for @xmath3 and ( by lemma [ prelim.diagonalisable ] ) a collection of projections @xmath442 with @xmath233 . since @xmath1 is also a factor
, there exists a unitary @xmath443 such that @xmath444 .
choose a refinement @xmath93 of @xmath445 in @xmath446 with @xmath447 for some @xmath448 .
the inequality @xmath449 is implied by item [ prelim.remarksi.commutant ] of remarks [ prelim.remarksi ] .
since @xmath450 also generates @xmath3 , it follows that @xmath451 proving the result .
[ subfactor.subfactor ] let @xmath426 be a finite index unital inclusion of @xmath26factors .
then @xmath452 recall from ( * ? ? ?
* lemma 3.1.8 ) , that given a finite index inclusion @xmath426 , there exists a subfactor @xmath453 with @xmath454=[m : n]$ ] and @xmath455 .
the result follows immediately from the previous lemma .
note that if there is a @xmath26factor @xmath3 with @xmath337 , then the previous lemma does not hold for infinite index subfactors as there is always a copy of the hyperfinite @xmath26factor @xmath456 inside any @xmath26factor .
let @xmath426 be a finite index unital inclusion of @xmath26factors .
then @xmath344 if , and only if , @xmath457 .
suppose that @xmath344 .
by lemma [ subfactor.bcon ] , it follows that @xmath458 .
now @xmath459 , so @xmath460 , where @xmath441 $ ] .
the scaling formula ( theorem [ scaling.formula ] ) immediately gives @xmath457 .
the reverse direction is lemma [ subfactor.subfactor ] above .
more generally , suppose that @xmath426 is a finite index unital inclusion of @xmath26factors and write @xmath441 $ ] .
the isomorphism @xmath459 and the scaling formula lead to the equality @xmath461 furthermore , lemmas [ subfactor.bcon ] and [ subfactor.subfactor ] give @xmath462 it is then natural to pose the following question .
suppose that @xmath426 is a finite index unital inclusion of @xmath26factors and write @xmath441 $ ] .
is it the case that @xmath463
the main result in this section is theorem [ scaling.freegroup ] below , which is obtained by using the quadratic scaling of the generator invariant and a similar property of the interpolated free group factors @xmath464 of the first author and rdulescu @xcite . for @xmath12 and @xmath465 , the quadratic scaling formula for the interpolated free group factors
states that @xmath466 from ( * ? ? ?
* theorem 2.4 ) .
[ scaling.freegroup ] there exists a constant @xmath467 such that @xmath468 for all @xmath469 $ ] . for @xmath470 write @xmath471 .
equation ( [ scaling.freegroup.10 ] ) above and theorem [ scaling.formula ] combine to give @xmath472 take @xmath473 and @xmath474 in ( [ scaling.freegroup.1 ] ) to obtain @xmath475 for all @xmath476 .
hence , there is a constant @xmath477 with @xmath478 .
the estimate @xmath479 follows from corollary [ ngenerators.group ] as @xmath480 and @xmath481 is certainly generated by two elements .
it remains to extend ( [ scaling.freegroup.2 ] ) to the @xmath482 case . since the fundamental group of @xmath483 is @xmath484 , @xcite , or more easily since @xmath485 , (
* corollary 5.2.3 ) , it follows that @xmath486 is either @xmath487 or @xmath488 , by corollary [ scaling.fundamentalgroup ] .
suppose that @xmath68 so that @xmath489 for each @xmath490 . consideration of the chain @xmath491 gives @xmath492 by shen s main technical theorem , ( * ? ? ?
* theorem 5.1 ) .
conversely , if @xmath492 , then the relation @xmath493 follows from the isomorphism @xmath494 this is essentially in @xcite and a proof can also be found in ( * ? ? ?
* chapter 15 ) . hence @xmath492
if and only if @xmath68 and the result follows . as an immediate consequence , there is a direct link between the free group isomorphism problem and the generator invariant .
[ freegroups.cor ] if @xmath495 , then all the interpolated free group factors @xmath496 are pairwise non - isomorphic for @xmath12 .
since @xmath497 $ ] has generator invariant @xmath487 , the previous theorem extends to include the case @xmath498 .
the next results give some estimates regarding the generator invariant and free products .
reverse inequalities to either theorem [ freeproduct.general ] or theorem [ freeproduct.hyperfinite ] would immediately combine with corollary [ freegroups.cor ] to show the non isomorphism of the free group factors .
lemma [ freeproduct.amaldiffuse ] originates in a conjugacy idea used repeatedly by shen to obtain ( * ? ? ?
* theorem 5.1 ) .
[ freeproduct.amaldiffuse ] let @xmath3 and @xmath1 be @xmath26factors containing a common diffuse von neumann subalgebra @xmath499 .
then @xmath500 given @xmath144 , choose finite subsets @xmath501 and @xmath502 and families @xmath503 and @xmath504 with @xmath505 , @xmath506 and @xmath507 by refining if necessary , it may be assumed that @xmath87 and @xmath93 are equivalent since proposition [ prelim.refine ] ensures that the estimate ( [ freeproduct.amaldiffuse.1 ] ) is unaffected by refinement .
since @xmath499 is diffuse , choose @xmath508 equivalent to @xmath87 and @xmath93 and unitaries @xmath509 and @xmath510 with @xmath511 and @xmath512 .
then @xmath513 is a finite generating set for @xmath514 and @xmath515 from which the result follows .
[ freeproduct.general ] let @xmath3 and @xmath1 be @xmath26factors .
then @xmath516 choose masas @xmath517 and @xmath518 .
then @xmath519 and @xmath520 , and so @xmath521 where @xmath522 .
now use lemma [ freeproduct.amaldiffuse ] twice to obtain @xmath523 the result follows as theorem [ scaling.freegroup ] gives @xmath524 .
the remainder of this section examines free products with finite hyperfinite von neumann algebras . in @xcite ,
jung proves that for a fixed hyperfinite von neumann algebra @xmath93 with a fixed faithful normal tracial state @xmath525 , voiculescu s modified free entropy dimension @xmath72 is the same for all finite sets @xmath73 that generate @xmath93 .
write @xmath526 for this quantity .
the definition of the modified free entropy dimension will be given in the next section , which discusses free entropy dimension in conjunction with the generator invariant . here only the value of @xmath526 is needed . following @xcite , given @xmath527 ,
decompose @xmath93 over its centre to obtain @xmath528 where @xmath445 is diffuse or @xmath529 , the sum on the right is either empty , finite or countably infinite and each @xmath530 .
the trace @xmath525 is given by @xmath531 where * @xmath532 and @xmath533 is a faithful normal trace on @xmath445 , if @xmath534 ; * @xmath535 and @xmath536 if @xmath537 ; * @xmath538 is the tracial state on the @xmath539 matrices @xmath540 and each @xmath541 .
then , from @xcite , @xmath542 furthermore , as also noted in @xcite , this quantity agrees with the ` free dimension number ' for @xmath93 defined in earlier work of the first author @xcite . in this work
it was shown ( theorem 4.6 of @xcite ) that if @xmath543 $ ] is equipped with the usual trace @xmath141 , then @xmath544 , where @xmath545 .
[ freeproduct.hyperfinite ] let @xmath3 be a @xmath26factor and @xmath93 a hyperfinite von neumann algebra with a fixed faithful normalised trace @xmath525 .
then @xmath546 choose a masa @xmath517 so that @xmath139 is isomorphic to @xmath140 $ ] with the trace on @xmath139 ( coming from @xmath547 ) being given by @xmath141 .
the discussion preceding the theorem gives @xmath548 where @xmath549 and @xmath139 is a masa in @xmath496 . by proposition [ freeproduct.amaldiffuse ] and theorem [ scaling.freegroup ] , @xmath550 exactly as required .
the next two corollaries are obtained by taking @xmath93 to be successively the @xmath551 matrices with the usual normalised trace and to be @xmath552 , with the group trace .
the results follow from calculating @xmath553 and @xmath554 from jung s formula ( [ freeproduct.freeentropy ] ) .
[ freeproduct.mn ] let @xmath3 be a @xmath26factor and @xmath555 .
then @xmath556 [ freeproduct.zn ] let @xmath3 be a @xmath26factor and @xmath555 . then @xmath557
the objective in this section is to relate the generator invariant to voiculescu s modified free entropy dimension by proving the inequalities @xmath558 when @xmath73 is a finite generating set in a finite von neumann algebra @xmath3 and , under the extra assumption that @xmath73 consists of self
adjoint elements , @xmath559 for certain sets @xmath73 of operators , these inequalities give lower bounds on @xmath560 .
these seem to be the only such lower bounds that are currently known .
consider the case of a @xmath561-operator @xmath373 , introduced by the first author and haagerup in @xcite . by
@xcite each such @xmath373 , including the quasi nilpotent dt
operator @xmath202 , generates @xmath562 .
the operator @xmath373 is constructed by realising @xmath562 as generated by a semicircular element @xmath563 together with a free copy of @xmath140 $ ] , using projections from @xmath140 $ ] to cut out an `` upper triangular '' part of @xmath563 , which is the quasi nilpotent dt
operator @xmath202 , and then adding an operator from @xmath140 $ ] to get @xmath373 . using projections from @xmath140
$ ] , one easily sees that @xmath564 . on the other hand , since , by @xcite , @xmath565 , ( [ entropy.nonsa ] ) gives @xmath566 similarly , free semi
circular elements @xmath567 generating @xmath562 satisfy @xmath568 from @xcite , so that ( [ entropy.sa ] ) gives @xmath569 .
the reverse inequality follows by taking a sufficiently fine family of projections in either @xmath570 or @xmath571 see the proof of lemma [ selfad.ngenerators ] so that @xmath572 the central question in the theory of microstates free entropy dimension is that of invariance : if one has two finite generating sets @xmath573 and @xmath574 for the same @xmath26factor must @xmath575 ?
certain conditions on the factor , such as having a cartan masa ( @xcite ) and non - primeness ( @xcite ) , imply that all finite generating sets have free - entropy at most one .
independently jung @xcite and hadwin and shen @xcite have shown that the existance of a finite generating set @xmath73 for a @xmath26factor with @xmath576 and certain additional technical conditions , which are subtly different in each approach , imply that all finite generating sets @xmath288 for @xmath3 have @xmath577 . before discussing these conditions further and establishing inequalities
( [ entropy.nonsa ] ) and ( [ entropy.sa ] ) , the prevailing notation in this area is outlined .
the definition of the modified free - entropy dimension given below is jung s covering number reformulation from @xcite .
let @xmath3 be a @xmath26factor and continue to write @xmath79 for the faithful normal trace on @xmath3 normalised by @xmath80 . for @xmath231 write @xmath179 for the @xmath178 matrices , equipped with the trace @xmath578 normalised with @xmath579 . for a finite subset @xmath580 , @xmath581 , @xmath582 and @xmath583 define the microstate space @xmath584 to be the set of all @xmath45-tuples @xmath585 of @xmath178-matrices whose @xmath586-moments approximate those of @xmath587 up to order @xmath264 within a tolerance of @xmath588 and whose norms are bounded by @xmath456 .
this means that @xmath589 for all @xmath91 and @xmath590 for all @xmath591 , @xmath592 and @xmath593 . when all the @xmath594 s are self
adjoint , it makes no difference to the definition of @xmath72 whether all the @xmath595 s are required to be self
adjoint see for example the beginning of section 3 of @xcite .
given @xmath144 , the covering number @xmath596 of a metric space @xmath288 is the minimal cardinality of an @xmath597-net for @xmath288 . one easy estimate , used in the sequel , is @xmath598 where @xmath599 is the maximal number of disjoint open @xmath600-balls which can be found in @xmath288 . in @xcite , jung defines , for @xmath361 , @xmath601 and @xmath583 , @xmath602 where the metric on @xmath584 is that obtained from the euclidian norm on @xmath603 given by @xmath604 then define @xmath605 and @xmath606 jung shows in corollary 24 of @xcite , that the modified free entropy dimension @xmath72 is given by @xmath607 in @xcite , jung made a detailed analysis of the rate of convergence in the limit - superior in ( [ entropy.defdelta ] ) .
for @xmath71 , say that the generating set @xmath73 of @xmath3 to be _ @xmath66-bounded _ if , for some @xmath608 , there exists some constant @xmath609 and @xmath610 such that @xmath611 for all @xmath612 .
corollary 1.4 of @xcite demonstrates that this is equivalent to the original definition of @xmath66-boundedness in @xcite and that if @xmath73 is @xmath66-bounded , then @xmath613 .
furthermore , @xmath3 is said to be _
strongly@xmath85bounded _ if it has a generating set @xmath73 that is @xmath85-bounded and if there exists a self - adjoint element @xmath18 belonging to the @xmath586-algebra generated by @xmath73 that has finite free entropy .
theorem 3.2 of @xcite shows that if @xmath3 is strongly@xmath85bounded , then every finite set of generators for @xmath3 is @xmath85bounded .
the alternative approach of hadwin and shen in @xcite is to examine coverings by unitary orbits of balls . for @xmath614 as above and @xmath144 , define @xmath615 to be the minimum cardinality of a set @xmath616 such that for every @xmath617 there is some unitary @xmath618 and @xmath619 with @xmath620 , i.e. the microstate space @xmath621 is covered by @xmath622 orbit @xmath597-balls . here
@xmath623 is defined to be the @xmath45-tuple @xmath624 .
then define @xmath625 and the _ upper free - orbit dimension _
@xmath626 by @xmath627 the main theorem of @xcite ( section 3 , theorem 1 ) is that if @xmath73 is a finite set of generators for a @xmath26factor with @xmath628 , then every finite set of generators @xmath288 for this factor also has @xmath629 .
therefore , it makes sense to say a @xmath26factor has zero upper free orbit - dimension when it has a finite generating set @xmath73 with @xmath628 .
the connection between the upper free orbit - dimension and voiculescu s modified free entropy dimension is immediate from results of szarek @xcite which show that there is an absolute constant @xmath630 such that the groups @xmath631 of unitary matrices in @xmath179 equipped with the metric induced from the operator norm satisfy @xmath632 given a finite generating set @xmath633 for a @xmath26factor @xmath3 , take @xmath634 .
provided @xmath635 and @xmath588 is sufficiently small , the estimate @xmath636 follows .
therefore @xmath637 and @xmath638 the free orbit - dimension @xmath639 of @xcite is defined by @xmath640 the relationships between the approaches of @xcite and @xcite follow from the definition of @xmath641 and ( [ entropy.freeorbit.1 ] ) . by hypothesis there is a set of self - adjoint generators @xmath73 for the @xmath26factor with @xmath628 . if there is no element of finite entropy in the @xmath586-algebra generated by @xmath73 , adjoin one .
this new set of generators @xmath288 also has zero upper free orbit - dimension by ( * ? ? ?
* section 3 , theorem 1 ) .
the definition of @xmath644 ensures that @xmath645 from equation ( [ entropy.freeorbit.1 ] ) . that is @xmath288 is @xmath85-bounded , and so @xmath3 is strongly @xmath85-bounded . in this section
the upper free - orbit dimension is used as it enables us to show , in corollary [ entropy.freegroup ] , that no set of generators @xmath73 for an interpolated free group factor can have @xmath646 . the strongly-@xmath85-bounded approach to this
would only rule out the existance of such generators @xmath73 which also have an element of finite entopy in the @xmath586-algebra they generate . before proceeding to the main result of this section ,
recall the following lemma , due to voiculescu .
formally , a proof can be constructed by copying the ideas of the proof of proposition 1.6 of @xcite . [ entropy.proj ] suppose that @xmath349 is a finite subset of @xmath3 with @xmath505 .
fix @xmath647 and pairwise orthogonal projections @xmath648 in @xmath3 with sum @xmath85 and @xmath649 for each @xmath91 . given
@xmath581 and @xmath361 , there exists @xmath650 , @xmath651 such that if @xmath652 for some @xmath653 with @xmath654 , then there exist pairwise orthogonal projections @xmath655 in @xmath179 each with @xmath656 ( so that @xmath657 ) satisfying @xmath658 [ entropy.boundedness ] let @xmath3 be a finite von neumann algebra and @xmath501 a finite generating set for @xmath3 . in general , @xmath660 and in the case that each @xmath128 is self - adjoint , @xmath661 .
furthermore , if @xmath646 , then @xmath628 .
let @xmath633 be an @xmath45-tuple which generates @xmath3 .
fix @xmath144 and suppose that @xmath662 .
use lemma [ prelim.diagonalisable ] to find @xmath663 with @xmath664 and each @xmath649 .
let @xmath583 and write @xmath665 .
take @xmath581 and @xmath361 with @xmath666 .
let @xmath667 and @xmath668 be the constants obtained by applying lemma [ entropy.proj ] to @xmath73 and @xmath669 .
let @xmath653 be divisible by @xmath291 and fix a family of pairwise orthogonal projections @xmath670 in @xmath179 with each @xmath671 .
for each @xmath672 write @xmath673 for the set of pairs @xmath674 with @xmath675 . as @xmath676 , it follows that @xmath677 following the approach of @xcite , define the projection @xmath93 from @xmath603 into @xmath603 by @xmath678 when each @xmath679 , the sets @xmath673 are invariant under the adjoint operation so in this case @xmath93 restricts to give a projection from @xmath680 into @xmath680 .
the range @xmath681 is a @xmath682dimensional subspace of @xmath683 . under the additional assumption that @xmath679 for each @xmath368 , the range @xmath684 is a @xmath685dimensional subspace of @xmath686 .
let @xmath288 be the subset @xmath687 consisting of all elements @xmath688 with @xmath689 if @xmath73 does not consist of self - adjoint elements , and let @xmath690 when each @xmath679 .
volume considerations give @xmath691 in the first case and @xmath692 in the second .
the simple estimate ( [ entropy.coverpack ] ) combines with ( [ entropy.sizet ] ) to give @xmath693 given any point @xmath694 , let @xmath655 be the orthogonal projections in @xmath179 with each @xmath695 from lemma [ entropy.proj ] .
now each @xmath696 can be written @xmath697 . since @xmath698 it follows that @xmath699 from the standing assumption that @xmath666 .
take a unitary @xmath618 such that @xmath700 for each @xmath701 .
then @xmath702 is the @xmath45-tuple with @xmath703 in the @xmath368-entry .
thus @xmath704 provided @xmath588 is taken small enough , @xmath705 , so that @xmath706 lies in the ball @xmath288 . if in addition , @xmath707 , it follows at most @xmath708 orbit @xmath597-balls are required to cover @xmath621 .
that is @xmath709 for @xmath583 , @xmath666 , @xmath588 sufficiently small and @xmath653 which is divisible by @xmath291 .
thus @xmath710 this holds for each @xmath662 and so we can replace @xmath711 by @xmath560 in the inequality above to obtain @xmath712 in general , and if @xmath646 , then @xmath628 .
the next corollary follows from proposition [ entropy.boundedness ] and ( * ? ? ?
* section 3 , theorem 1 ) .
it shows in particular that @xmath716 for every generating set @xmath73 of an interpolated free group factor by @xcite .
l@l ken dykema & roger smith & allan sinclair + department of mathematics & school of mathematics + texas a&m university & university of edinburgh + college station , tx 77843 & edinburgh eh9 3jz + usa & uk + ` [email protected] ` & ` [email protected] ` + ` [email protected] ` + + stuart white + department of mathematics + university of glasgow + university gardens + glasgow g12 8qw + uk + ` [email protected] ` | in 2005 , junhao shen introduced a new invariant , @xmath0 , of a diffuse von neumann algebra @xmath1 with a fixed faithful trace , and he used this invariant to give a unified approach to showing that large classes of @xmath2 factors @xmath3 are singly generated .
this paper focuses on properties of this invariant .
we relate @xmath4 to the number of self - adjoint generators of a @xmath2 factor @xmath3 : if @xmath5 , then @xmath3 is generated by @xmath6 self - adjoint operators , whereas if @xmath3 is generated by @xmath6 self - adjoint operators , then @xmath7 .
the invariant @xmath8 is well - behaved under amplification , satisfying @xmath9 for all @xmath10 .
in particular , if @xmath11 for any particular @xmath12 , then the free group factors are pairwise non - isomorphic and are not singly generated for sufficiently large values of @xmath13 .
estimates are given for forming free products and passing to finite index subfactors and the basic construction .
we also examine a version of the invariant @xmath14 defined only using self - adjoint operators ; this is proved to satisfy @xmath15 .
finally we give inequalities relating a quantity involved in the calculation of @xmath16 to the free - entropy dimension @xmath17 of a collection of generators for @xmath3 . |
star formation is the fundamental process that shapes galaxies into different classes .
although the majority of stars in the local universe are found in spheroidal systems , most of the star formation is contributed by disc galaxies of the later types ( beyond sb ) .
the key ingredient for star formation is cold ( _ star - forming _ ) gas , which is present almost exclusively in disc galaxies .
however , the amount of cold gas currently available in galaxy discs appears rather scant .
@xcite estimated that disc galaxies have current star formation rates ranging from a few to about @xmath4 .
thus , considering a typical gaseous mass of a few @xmath5 , the gas consumption time scale ( i.e. the time needed to exhaust the gas fuel with a constant star formation rate ) is always of the order of a few gigayears .
this result , known as the gas - consumption dilemma @xcite , suggests the need for continuous accretion of cold gas onto galaxy discs ( e.g. * ? ? ?
several pieces of evidence show that disc galaxies should collect fresh gas from the environment in order to sustain their star formation . in the milky way
, the star formation rate ( sfr ) in the solar neighbourhood appears to have remained rather constant in the last @xmath6 ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) , suggesting a continuous replenishment of the gas supply .
simple ( closed - box ) models of chemical evolution for our galaxy predict too few metal deficient _ g - dwarf _ stars than observed .
these observations are easily explained by accounting for infall of fresh unpolluted gas ( e.g. * ? ? ?
* ; * ? ? ?
observations of damped lyman alpha systems show almost no evolution in the neutral gas content of structures in the universe @xcite .
@xcite pointed out that this constancy of gas density can be explained assuming a rate of gas replenishment proportional to the universal sfr density .
@xcite converged to a similar result when comparing the evolution of the molecular gas depletion rate with that of the cosmic star formation history ( sfh ) .
finally , the derivations of sfhs for galaxies with different stellar masses consistently show that late type systems do have a rather constant sfr throughout the whole hubble time ( e.g. * ? ? ?
these findings , other than being the signature of _ downsizing _ in cosmic structures , point at a continuous infall of gas onto galaxies of late hubble types . the way gas accretion into galaxies takes place is still a matter of debate . the classical picture states that thermal instabilities should cause the cooling of the hot coronae that surround disc galaxies and be the source of cold gas infall .
however , recent studies have shown that , due to a combination of buoyancy and thermal conduction , hot coronae turn out to be remarkably stable and thermal instability does not appear to be a viable mechanism for gas accretion @xcite . on the other hand , accretion may take place in the form of cold flows but the importance of this process is expected to drop significantly for redshift @xmath7 .
observations of local gas accretion at 21-cm emission seem to show too little gas around galaxies in the form of hi ( high - velocity ) clouds to justify an efficient feeding of the disc star formation @xcite . however , uv absorption towards quasars and halo stars point to the possibility that ionised gas at temperatures between a few 10@xmath8 and a few @xmath9 could fill the gap between expectations and data . in this paper , we estimate gas accretion in galactic discs indirectly by comparing basic physical properties of galaxies today .
our model relies on the existence of a law relating the star formation rate density ( sfrd ) and the gas surface density ( @xmath10 ) holding at every redshift , i.e. the kennicutt - schmidt ( k - s ) law @xcite .
this approach has similarities with the model of that we describe in detail in section [ sec : discussion ] .
the paper is organized as follows .
section [ sec : model ] describes our method . in section [ sec : data ]
we apply our model to the milky way disc and to a sample of external discs . in section
[ sec : discussion ] we discuss our results .
section [ sec : conclusions ] sums up .
the evolution of gas and stellar densities ( @xmath10 and @xmath11 ) in a galactic disc as a function of the lookback time @xmath12 is described by the following equations @xcite : @xmath13 where @xmath14 is the star formation rate density , @xmath15 is the contribution from stellar feedback and @xmath16 is the inflow / outflow rate from / to the external environment .
@xmath14 gives the rate at which stars are formed or , with a sign inversion , the rate at which gas is consumed , whilst @xmath15 represents the gas density per unit time returned by stellar evolution to the gas reservoir in the disc .
we treat the stellar feedback with the instantaneous recycling approximation ( i.r.a . ) , meaning that the fractional mass that stars return to the ism at each time is assumed to be constant and equal to the return factor @xmath17 .
equations ( [ eq : tinsley1])-([eq : tinsley2 ] ) modify as follows : @xmath18 in section [ sec : nonira ] we discuss the effect of a delayed return . the parameter @xmath17 in eqs .
( [ eq : tinsley1wr])-([eq : tinsley2wr ] ) depends on the imf and the fraction @xmath19 of mass that a star of mass @xmath20 returns to the ism , called the _ stellar initial versus final mass relation_. following the prescription of @xcite we assume that high - mass stars ( @xmath21 m@xmath22 ) all leave a @xmath23 m@xmath22 remnant .
low - mass stars are assumed to leave remnants with masses between @xmath24 and @xmath25 m@xmath22 , with a linear dependency from the initial mass ( see chapter 2 of * ? ? ? * for details ) . assuming a salpeter imf , our fiducial value for the recycling parameter is @xmath26 .
assuming kroupa @xcite or chabrier @xcite imfs would make little difference , giving @xmath27 and @xmath28 ( or @xmath29 for a flatter high - mass slope ) respectively . with this value
the gas consumption timescales are extended by a factor @xmath30 ( see * ? ? ?
* ) . the third term on the rhs of eq . ( [ eq : tinsley2 ] ) is the inflow / outflow term . the combination of eqs .
( [ eq : tinsley1 ] ) and ( [ eq : tinsley2 ] ) leads to an expression for this term : @xmath31 in the following we often refer to this term as the accretion ( or infall ) rate .
equations ( [ eq : tinsley1 ] ) and ( [ eq : tinsley2 ] ) can be applied to a galactic disc assuming that they are valid at each radius , i.e. that the disc can be divided into annuli that evolve independently .
this assumption is equivalent to say that stars that form at a certain radius remain at that radius for the lifetime of the galactic disc . in section [ sec : migration ]
we discuss the effect of stellar migration and show that our main results do not change significantly . as a consequence of this assumption we can integrate eq . ( [ eq : tinsley1 ] ) from @xmath32 ( the lookback time at the disc formation ) to @xmath33 to get the current stellar surface density .
clearly , in order to solve the above system one needs the @xmath14 as a function of lookback time and radius , i.e. the sfh as a function of radius . in the milky way , the sfh is known with fairly good precision only in the solar neighborhood ( e.g. * ? ? ?
* ; * ? ? ?
in external galaxies there have been pioneering attempts to estimate the sfh as a function of galactic radius ( e.g. * ? ? ? * ; * ? ? ?
* ) . in general , however
, a precise determination of @xmath34 is beyond the capability of the current data .
therefore , for the time being , we must content ourselves with using simple parametrisations for the shape of the sfh . in the following section
we describe how we estimate the trend of the sfh at each radius in a galactic disc .
we consider only the disc , leaving aside the bulge / bar , which may have formed and evolved differently as it shows a markedly different sfh ( e.g. * ? ? ?
we assume that the sfh in a galaxy disc can be described by a dimensionless and positive function @xmath35 .
this allows us to write the star formation rate density as : @xmath36 where @xmath37 is the radial distribution of the current star formation rate density , which we derive from the data . the simplest expression for
@xmath38 one can assume is a first order polynomial with time : @xmath39 \frac{t}{t_{\rm form } } \quad ( t_{\rm form}\geq t \geq 0)\ ] ] where @xmath12 is the lookback time and @xmath32 is the age of the stellar disc . with this definition , @xmath40 is the value of the function @xmath38 at @xmath41 and sets the steepness of the sfh .
a value of @xmath40 equal to unity represents a constant sfr over the disc lifetime , @xmath42 describes a sfr that increases ( decreases ) with increasing lookback time . would imply a lower formation time at that radius ; this occurance does not arise with our choice of parameters . ]
the dependence of @xmath43 on the radius @xmath44 allows us to describe different shapes of sfh within the disc . in section [ sec : fs ] we discuss the effect of taking different forms for @xmath38 , we anticipate that our main results remain unchanged .
we define the _ global _
@xmath43 for a galactic disc as : @xmath45 where the integration is performed out to the maximum radius @xmath46 ; the denominator is just the current sfr of the disc .
this definition allows us to write an expression for the global sfr analogous to eqs .
( [ eq : sfrd ] ) and ( [ eq : f1 ] ) , @xmath47\ ] ] evolution of the average sfr density of the universe , the points are taken from .
the thick black curve shows a fit with our function @xmath48 described in the text .
, scaledwidth=50.0% ] as an illustration of the meaning of our parameter @xmath43 in fig .
[ fig : sfhuniverse ] we fit the average sfr density of the universe with eq .
( [ eq : sfr ] ) .
the data points are taken from . here
we take @xmath49 gyr . for the other parameters we infer @xmath50 and @xmath51 as the best - fit values .
this fit shows that the sfr density of the universe as a whole is described by a value of @xmath43 much larger than unity , as expected given that the sfr density in the past was much higher than now .
the shape of the universal sfr density is produced by the combined contribution of galaxies with very different sfhs .
massive red - sequence galaxies dominate the overall sfr at high - z , whilst late - type galaxies become relatively more and more important as time passes ( e.g. * ? ? ?
* ; * ? ? ?
although our parametrisation is only suitable for galaxies with non - negligible current sfr , taking @xmath43 as the ratio of @xmath52 , clearly red sequence galaxies must have @xmath53 . the above assumptions for the sfrd ( eqs .
[ eq : sfrd ] and [ eq : f1 ] ) allow us to integrate eq .
( [ eq : tinsley1wr ] ) from @xmath32 to 0 to obtain : @xmath54 where @xmath55 is the surface density profile of the stellar disc that we derive from the observed surface brightness profile .
[ eq : gammar ] reveals that our parameter @xmath40 is ultimately a ratio between the stellar density and the current sfrd , given a certain formation time for the disc , which is of the order of the hubble time and can be assumed roughly equal for all discs .
a @xmath40 close to 1 means that the current sfr density extended to an hubble time produces as many stars as observed at that location in the disc .
integrating @xmath40 over the disc and making used of eq .
( [ eq : gammadef ] ) we find an expression for the global @xmath43 as : @xmath56 where @xmath57 and @xmath58 are respectively the current stellar mass and star formation rate of the entire disc .
clearly from eq .
( [ eq : gamma ] ) , one can also describe @xmath43 as proportional to the ratio between the average past sfr and the current sfr .
this latter is the inverse of the so - called scalo @xmath59-parameter @xcite to which @xmath43 is closely related ( see section [ sec : things ] ) .
we also note that any modification of eq .
( [ eq : f1 ] ) with power law terms of @xmath60 would produce the same dependencies for @xmath43 as eq .
( [ eq : gammar ] ) .
we discuss these points further in section [ sec : fs ] .
the main ingredient of our model is the relation between the rate of star formation and the gas density .
the empirical star formation law is broadly accepted to be a power - law relation between the `` cold '' gas density and the star formation rate density @xcite .
this relation has been tested on global ( i.e. averaging gas and star formation on the whole disc ) and local scales @xcite considering the molecular gas or the sum of atomic and molecular gas .
remarkably , it appears to hold across various orders of magnitudes , from quiescent discs to star - busting galaxies @xcite .
the dependence on gas metallicity is difficult to investigate due to the related dependency of the x@xmath61 factor @xcite .
however , the compilation of azimuthal averages of sfrd and @xmath10 presented by @xcite for the things sample , which includes galaxies with very different masses and presumably different metallicities , seems to show that the effect is limited ( see * ? ? ? * and section [ sec : others ] ) .
thus , in the following we assume the following star formation law : @xmath62 with @xmath63 and @xmath64 if @xmath10 is measured in @xmath65 and sfrd in @xmath66 @xcite .
note that the above value of @xmath67 takes into account the correction for the presence of helium ( a factor 1.36 in mass ) , not applied in @xcite .
deviations from the above relation are observed at column densities lower than @xmath68 , referred to as the density threshold .
below these densities the law steepens and the scatter increases making it more difficult to describe @xcite . in the context of this paper
, it may have an effect in the outer parts of the discs ( see sections [ sec : mw ] and [ sec : others ] ) .
@xcite have recently pointed out that the star formation law expressed in terms of volume density is more general than the one expressed in terms of surface density .
this is very relevant to understand the physical origin of the law and to perform numerical simulations .
however , for our investigation a surface - density law is far more practical . clearly , the choices are equivalent if in the region of interest , i.e. within the star forming disc , the scale - height of the gaseous disc does not change much as a function of radius . the validity of this condition ( within a factor 2 ) is well established in our galaxy , as the disc appears to be flaring significantly only beyond @xmath69 .
the above assumptions on the sfh and the star formation law allow us to infer the local gas accretion rate using eq .
( [ acc_prima ] ) . if the star formation law is described by eq .
( [ eq : sflaw ] ) , the accretion rate density can be written as : @xmath70 where @xmath37 is the current total gas density , @xmath67 and @xmath71 are the parameters of the star formation law and @xmath38 is the functional form of the star formation history .
the rhs of eq .
( [ eq : gar ] ) can also be expressed in terms of @xmath72 ( using the k - s law ) instead of @xmath37 .
this equation is general and it could be used if the
_ exact _ shape of the @xmath34 were known .
if @xmath38 is given by eq .
( [ eq : f1 ] ) , the gas accretion rate density becomes : @xmath73 } \nonumber \\ & & -\frac{{\rm sfrd}^{1/n}(r,0)}{n\ , a^{1/n}\,t_{\rm form } } \left(\gamma(r)-1\right ) \left[1+\left(\gamma(r)-1\right)\frac{t}{t_{\rm form}}\right]^\frac{1-n}{n}\end{aligned}\ ] ] or , equivalently in terms of the gas density : @xmath74 } \nonumber \\ & & -\frac{\sigma_{\rm gas}(r,0)}{n\ , t_{\rm form } } \left(\gamma(r)-1\right ) \left[1+\left(\gamma(r)-1\right)\frac{t}{t_{\rm form}}\right]^\frac{1-n}{n}\end{aligned}\ ] ] the above equations give the rate at which fresh gas must be added to the disc at a certain radius and as a function of time to produce the star formation density given by @xmath75 .
as shown above , the shape of the sfh is regulated by the value of @xmath40 . for @xmath76 ,
sfr and accretion are constant in time .
more in general , when @xmath43 is close to 1 the second term on the rhs of eq .
( [ eq : garf1_1 ] ) is small leading to @xmath77 .
therefore the gas needed is directly proportional to the gas consumed by the star formation and this proportionality is simply @xmath78 . for @xmath79 the second term in eqs .
( [ eq : garf1_1 ] ) and ( [ eq : garf1_2 ] ) becomes gradually more negative and less accretion is needed .
high values of @xmath40 may result in a negative @xmath80 , hence at that time and radius , the amount of gas available in the disc exceeds what is needed by star formation
. this would be the signature of an outflow of gas .
we define the global accretion rate @xmath81 as the gas mass accreted per unit time over the whole galactic disc , obtained integrating eqs .
( [ eq : garf1_1 ] ) or ( [ eq : garf1_2 ] ) over the disc out to our maximum radius @xmath46 .
it is evident that the global accretion rate will have dependencies similar to those of the local one and the above considerations apply . in particular , for galaxies with @xmath43 of order unity , the global infall rate closely follows the @xmath82 at every time @xmath83 .
it is worth noting that @xmath84 represents the gas surface density that has to be added to ( or removed from ) the disc per unit time to produce the reconstructed @xmath75 .
if we are considering a single annulus , @xmath84 refers generically to the external environment including the adjacent annuli .
therefore , @xmath84 coincides with the accretion ( infall ) rate at that radius only if there are no radial flows in the gaseous disc .
this is because , with this method we do not trace the location of the gas infall but rather the radius at which such gas turns into stars .
we discuss this issue a bit further later on . on the other hand
, @xmath81 represents the total mass needed to be accreted by the galaxy disc from outside , at the time @xmath12 , to sustain the sfr at the reconstructed rate . note however that also in this case the definition of _ outside _ the disc is merely _ beyond _ the maximum radius @xmath46 assumed for the disc , which we took to be 5 times the scale - length of the stellar disc .
in this section we apply the method described above to real galaxy discs .
we impose the following _ boundary conditions _ :
i ) the current stellar density profile @xmath55 , ii ) the current sfr density profile @xmath37 . the latter could be in principle derived from the current gas density profile using the star formation law ( section [ sec : sflaw ] ) .
however here we restrict our analysis to cases where the current sfrd is known from independent estimates .
we first consider the milky way and then a sample of spiral galaxies with known sfrd profiles .
current star formation rate density of the milky way disc estimated using pulsars ( open squares ) , and supernova remnants ( filled squares ) .
these data , originally normalized to the solar neighbourhood value , have been fitted with a function described in the text ( solid line ) and rescaled so that the integral over the disc gives sfr(0)@xmath85 .
the shaded region shows the @xmath37 one would get from the total ( neutral + molecular ) gas density using the k - s law . the lower and upper boundaries are due to the different determinations of the hi surface densities from and respectively . , scaledwidth=50.0% ]
we assume that the stellar disc of the milky way is exponential with a scale length @xmath86 and a maximum extension of @xmath87 .
the normalization of this profile is chosen so that the total stellar mass of the disc is @xmath88 m@xmath22 .
the present distribution of the star formation rate density with respect to the solar neighborhood , @xmath89 , has been estimated using several methods . in fig .
[ fig : sfrdmw ] we show the estimates coming from the distribution of pulsars ( open squares , * ? ? ? * ) and supernova remnants .
we fit these points with a function ( solid line in fig .
[ fig : sfrdmw ] ) of the form : @xmath90 where @xmath91 is an exponent defining the tapering towards the centre and @xmath92 is a scale radius such that the peak of the distribution is located at @xmath93 . the central sfr density @xmath94 is normalized in order to have a global current star formation rate of the milky way @xmath95 ( see table 1 in * ? ? ?
the same normalization is then applied to the data points .
the shaded region in fig .
[ fig : sfrdmw ] shows , as a consistency check , the sfrd inferred from the current total gas density using the k - s law .
the upper and lower boundaries refer to two different determinations of the neutral gas density by and .
the latter is not determined for @xmath96 .
the molecular gas density is taken from .
on the whole the agreement between the direct determinations of sfrd and that inferred by the k - s law is very good . in the outer parts
the k - s predicts a larger sfr than observed , likely due to the low densities and the steepening of the law .
we quantify this effect in section [ sec : others ] .
some recent determinations of the current sfr of the milky tend to give values around @xmath97 . however , fig .
[ fig : sfrdmw ] shows that only a @xmath98 is self - consistent with the derivation of the @xmath37 from the gas surface density by inverting the k - s law , so we prefer to adopt this value throughout the paper . in the following we further fix the age of the stellar disc @xmath99 or , equivalently , @xmath100 . and @xmath101 .
] note that taking a larger value for @xmath32 , suggested for instance by the investigation of would strengthen our results .
steepness of the sfh as a function of radius for the milky way disc , parametrised by @xmath40 ( see section [ sec : sfh ] ) .
the horizontal line shows the global value obtained for the whole disc .
, scaledwidth=50.0% ] we reconstruct the sfh of the milky way disc as described in section [ sec : sfh ] .
we find a steepness of @xmath102 for the whole disc .
the behavior as a function of radius is shown in fig.[fig : gammar ] .
the parameter @xmath40 appears to be a decreasing function of the radius @xmath44 . recalling that @xmath40 is the ratio between the initial and the current sfrds , this trend shows ,
as expected , that star formation in the central regions must have been faster than in the outer disc , and it has spread from inside out .
note that our approach is fully axi - symmetric and it does not account for the presence of a bar , thus our results for @xmath103 should be taken with some caution
. corresponds to a complete replenishment and thus a constant sfr .
[ fig : sfh+accmw],scaledwidth=50.0% ] kpc ( _ short dashed line _ ) , @xmath104 kpc ( _ solid line _ ) and at @xmath105 kpc ( _ long dashed line _ ) versus time and redshift .
[ fig:3radii],scaledwidth=50.0% ] the sfr(t ) of the galaxy reconstructed using the linear temporal dependency of eq .
( [ eq : f1 ] ) is shown in fig .
[ fig : sfh+accmw ] ( top panel ) ( @xmath102 ) .
we remind that the key parameters to obtain this curve are the current sfr and the current stellar mass of the disc ( eq . [ eq : gamma ] ) . changing these parameters and the formation time of the disc ( @xmath106 )
would produce rather intuitive variations in the value of @xmath43 and the steepness of the sfh .
extreme values can be obtained by taking @xmath107 , @xmath108 and @xmath109 on the one hand and @xmath110 , @xmath111 and @xmath112 on the other , corresponding to @xmath113 and @xmath114 respectively .
having derived @xmath40 we can proceed to estimate the gas infall rate at each radius and its global value by integrating over the whole disc surface using eq .
[ eq : garf1_1 ] .
[ fig : sfh+accmw ] shows the global gas accretion rate necessary to maintain the star formation in the milky way disc .
the accretion rate is positive at every redshift starting from @xmath115 at @xmath32 and decreasing to @xmath116 at the current time .
it follows very closely the shape of the sfr .
these results are in agreement with earlier determinations based on chemical evolution models ( e.g. * ? ? ?
* ; * ? ? ?
in the same figure , we also plot the sfr that our galaxy would have if it evolved like a closed system starting from the same initial conditions .
a closed system is only able to produce @xmath117 of stars from @xmath41 to @xmath33 , virtually leaving no remnant gas mass at present . on the contrary
, our model with gas infall predicts a present total gas mass of @xmath118 , comparable to what observed .
the bottom panel of fig .
[ fig : sfh+accmw ] shows the gas accretion efficiency defined as the ratio of the accretion rate over the net gas - consumption including feedback , @xmath119 .
for the whole disc of the milky way we obtain @xmath120 at @xmath33 .
@xmath121 has a mild dependence on time , reaching a value of about 0.9 at @xmath1 .
this shows that the efficiency of gas accretion is rather high but not enough to keep the sfr constant ( @xmath122 ) .
thus , gas accretion does not feed the discs indefinitely , however it delays the running - out of fuel significantly beyond the ( closed - box ) gas - consumption timescale .
the value of the current accretion rate and efficiencies are rather uncertain as they strongly depend on the assumed zero - points .
for instance , leaving all the other parameters unchanged and taking @xmath123 would give a current accretion of only @xmath124 .
this large difference is totally localized to the current time , indeed the efficiency of accretion remains always above 0.5 from the disc formation to @xmath125 plunging then to 0.1 at @xmath2 , see also the case of ngc5055 in section [ sec : things ] .
gas accretion profile at three epochs : lookback time = 3 gyr , 1.5 gyr , and now .
note the very recent development of an inner depression and of a peak in the gas accretion distribution , now at about @xmath126 from the galactic centre .
, width=321 ] we now turn to the local infall rate @xmath84 . for visualization purposes ,
we split the galaxy in three representative zones : the central part ( between @xmath127 kpc ) , the solar circle ( @xmath128 kpc ) and the outer disc ( @xmath129 kpc ) . for each region
we calculate the local gas accretion rate integrating over the annulus surface and we plot the results in fig .
[ fig:3radii ] .
we find three different regimes : i ) in the central region the gas accretion declines very steeply , ii ) in the solar circle the gas accretion declines by a factor 1.7 , iii ) in the outer disc it increases slightly reaching its maximum at @xmath2 .
the behaviour visible in fig .
[ fig:3radii ] reflects the simple fact that the ratio between the stellar density and gas density decreases strongly with radius in the galaxy .
the same kind of trends are visible in the sfh reconstructed for the same annuli , given that as seen above , @xmath130 .
[ fig : peak ] shows the gas accretion profile as a function of @xmath44 in the milky way s disc now and at two epochs in the recent past .
the current profile is negative for @xmath131 , peaks at about @xmath132 and then falls further out .
the negative value in the central regions shows that there the amount of gas is sufficient to sustain the star formation and the model would predict outflow ( of about @xmath133 ) .
this inner feature is of recent formation as 3 gyrs ago the peak of accretion was right in the centre .
so we conclude that the bulk of gas accretion is moving away from the inner disc . as a consequence star formation also moves out although with a delay with respect to the accretion .
note that the shape of the accretion profile at @xmath134 is remarkably similar to the current @xmath37 ( see fig.[fig : sfrdmw ] ) , hinting at a delay between accretion and star formation of a gyr or so .
interestingly , this is of the order of the gas depletion timescale .
we stress again that with our approach we are able to trace the locations in the disc where the new gas begins to form stars .
the infall itself could have happened somewhere else , realistically more further out and the infalling gas could have quiescently flown to inner radii before starting forming stars and becoming _ detectable _ by our method .
the ( disc - integrated ) parameter @xmath43 ( steepness of the sfh ) for the sample of 21 disc galaxies of @xcite as a function of the stellar mass ( _ top panel _ ) and galaxy type ( _ bottom panel _ ) .
the star indicates the milky way .
a common time of disc formation of @xmath135 gyr and the sfh - function @xmath48 have been assumed for all galaxies .
the error - bars show the interval allowed by two different methods of integration described in the text .
the dashed curve in the top panel shows the value of @xmath43 derived from the scalo @xmath59-parameter for the sloan survey @xcite .
, title="fig:",width=321 ] the ( disc - integrated ) parameter @xmath43 ( steepness of the sfh ) for the sample of 21 disc galaxies of @xcite as a function of the stellar mass ( _ top panel _ ) and galaxy type ( _ bottom panel _ ) .
the star indicates the milky way .
a common time of disc formation of @xmath135 gyr and the sfh - function @xmath48 have been assumed for all galaxies .
the error - bars show the interval allowed by two different methods of integration described in the text .
the dashed curve in the top panel shows the value of @xmath43 derived from the scalo @xmath59-parameter for the sloan survey @xcite .
, title="fig:",width=321 ] in this section we extend the application of our method to a sample of external galaxies .
we consider galaxies from the things survey @xcite for which radial profiles of gas density and sfrd are available @xcite .
as for the milky way , we assume a common return factor @xmath136 , an age of the discs of @xmath137 gyr and the linear @xmath48 described in section [ sec : sfh ] . we derive the global values of @xmath43 both using eq.([eq : gamma ] ) and by integrating @xmath40 ( eq . [ eq : gammar ] ) over the disc . in the former case the values of stellar masses and sfrs were taken from table 4 of @xcite , in the latter we used their sfr densities available online .
the two methods led , in some cases , to significantly different values for @xmath43 .
this is partially due to the different range of integration , @xcite integrate out to @xmath138 while we integrate out to 5 scale - lengths ( when the profiles extend this far , i.e. for roughly for half of the galaxies ) for consistency with the milky way analysis .
[ fig : gammaleroy ] shows the derived values for the global @xmath43 as a function of the stellar mass ( top ) and hubble type ( bottom ) , the black star shows the milky way .
the bars show the ranges allowed by the two methods described above .
note that for some galaxies also the stellar masses resulting from our integrations differ from those of @xcite .
the values of @xmath43 inferred for these galaxies are all included between 0 and 11 , with one outlier at @xmath139 , ngc2841 , a massive sb galaxy with rather low current sfr ( shown only in the top plot ) .
there is a clear trend with stellar mass and hubble type , showing that big galaxies had much larger sfrs in the past . on the other hand , low values of @xmath43 , typical for smaller galaxies
betray a disc continuously forming through the acquisition of fresh gas .
as mentioned , we can relate our @xmath43 to the scalo @xmath59-parameter .
we use the recent determination of @xmath59 versus stellar mass relation from the sdss @xcite and convert it into @xmath43 .
we plot this relation as a dashed curve in the top panel of fig.[fig : gammaleroy ] .
the two parametrisations agree remarkably well , considering that they are derived in completely different ways .
our values tend to be slightly above the curve for masses lower than 10.5 in the log and below for larger masses .
this is likely due to the shape adopted for the sfh , see discussion in section [ sec : fs ] .
note that , in this respect , fig .
[ fig : gammaleroy ] is analogous to fig . 3 in @xcite .
for the above disc galaxies we have performed the same analysis as for the milky way .
we derived the radial profiles of @xmath40 and @xmath140 and integrated the latter to derive the global accretion rates .
[ fig : sfhsleroy ] shows the results for three galaxies representative of three classes of stellar masses and galaxy types ( see labels in fig .
[ fig : gammaleroy ] ) . in general ,
small late - type galaxies tend to have a flatter sfh and an efficiency of gas accretion close to 1 , see the case of ngc4449 ( left panels ) .
galaxies of intermediate types have a shallower sfr and efficiency gradually decreasing with time .
finally , massive discs with low current star formation rates seem to have exhausted their capability of acquiring significant amount of gas from the environment .
they now appear to be consuming their remnant gas before moving to the red sequence ( e.g. * ? ? ?
in the same fig . [
fig : sfhsleroy ] we also compare our results with the average sfhs derived using the sloan survey by @xcite , shown as open circles .
panter et al .
divide their galaxies in bins of current stellar masses , those relevant here are @xmath141 , @xmath142 , @xmath143 , respectively applicable for ngc5055 , ngc628 , and ngc4449 .
the agreement between our sfrs and their average sfrs is very good , except for the galaxies in the intermediate mass class as our method tends to find steeper sfhs .
note however that this may also depend on the way the stellar masses are estimated .
we used the estimate of @xcite who simply multiply the 3.6-@xmath144 surface brightness by a constant mass - to - light ratio of 0.5 .
in general , the analysis of these 21 things galaxies gives results similar to those found for the milky way s disc .
gas accretion proceeds almost parallel to the sfr and it remains significantly important throughout a large fraction of the lifetime of a typical disc galaxy .
the efficiency of accretion seems to decrease fast for large galaxies , while small late - type discs remain quite efficient in collecting cold star - forming gas from the environment .
the current ( @xmath2 ) values of gas infall that we derive are quite low for most galaxies in the sample and a few also have negative values showing that the available gas is more than sufficient for star formation to proceed .
thus , it would appear that these disc galaxies do not need gas accretion to sustain their current star formation ( see e.g. ngc5055 ) .
this is potentially a very interesting result but unfortunately , as shown for the milky way , these values are very uncertain because they come from an extrapolation .
moreover , they depend on parameters like the stellar masses that have intrinsic uncertainties and other parameters that are simply assumed ( see eq . [ eq : garf1_1 ] ) . for instance
if one increases the formation time of the disc or decrease the return factor the value of the current accretion rate may increase substantially . what remains indisputable
is the fundamental relation between star formation and accretion and the fact that the latter @xmath145 be a major player in the evolution of galaxies from their formation to the present .
there is no way a disc galaxy could become as we see it today without substantial gas infall at @xmath146 .
in the previous sections we have shown how a simple comparison between the stellar density in galactic discs and their current sfrd ( or gas density ) implies the need for a large amount of gas infall during their lifetimes .
this need is rooted in the existence of the kennicutt - schmidt law and in the assumption that this does not significantly evolve with time . under this assumption
we were able to derive the accretion rates as a function of galactocentric radius and time for a number of disc galaxies including the milky way .
here we discuss the limitations of our approach and the implications of our results . treating stellar feedback with the i.r.a .
saves computing time and it allows us to write simple analytic expressions for the equations of the model , however the delayed return from stars with mass @xmath147 may be an important effect that needs to be quantified .
we added this ingredient using the approximation that a star with mass @xmath20 returns a fraction @xmath19 of its mass to the interstellar medium at a time @xmath148 ( see * ? ? ? * ) .
the stellar lifetime ( @xmath149 ) are taken from : @xmath150 given that @xmath151 m@xmath152 gyr stars below @xmath153 m@xmath22 do not contribute to the feedback .
the difference between the i.r.a .
and the delayed return is that the term @xmath15 in eqs .
( [ eq : tinsley1 ] ) and ( [ eq : tinsley2 ] ) becomes : @xmath154 where @xmath155 is the imf .
the integration is performed from @xmath156 , which is the mass of a star with a ms time equal to the age of the disc at that time @xcite ; note that with @xmath12 we indicate , as usual , lookback time . in order to calculate @xmath40 with delayed feedback , we have numerically intergrated eq .
( [ eq : tinsley1 ] ) from the present to @xmath32 using eqs .
( [ eq : sfrd ] ) and ( [ eq : f1 ] ) for @xmath75 . at each radius @xmath44
, the integration has been repeated iteratively until we matched the observed stellar surface brightness @xmath55 .
then the global @xmath43 was calculated using eq .
( [ eq : gamma ] ) .
finally the accretion rate as a function of radius and time was derived using eq .
( [ eq : garf1_1 ] ) .
[ fig : nonira ] shows the comparison between the global sfr and accretion rate for the milky way with i.r.a . and with delayed stellar feedback .
the values of current sfr and stellar mass are kept fixed in both cases . as expected the global gamma with delayed feedback
is lower , @xmath157 ( 2.6 for a chabrier imf ) , than that with i.r.a .
this is because if the mass return to the ism is delayed then fewer stars need to form in order to have the same stellar mass at the present time .
the gas accretion is higher at the beginning due to the fact that less gas is available from stellar recycling and becomes slightly lower later on when feedback from low mass stars becomes important .
the bottom panel of fig .
[ fig : nonira ] shows the accretion efficiency calculated with and without delayed feedback .
accretion with delayed feedback tends to have a slighly larger efficiency .
however the shape of the two curves is exactly the same and the difference in the normalization is minor . in conclusion ,
including delayed feedback in our calculations makes little difference and it goes in the direction of increasing the efficiency of gas accretion .
_ top panel : _ sfr and global gas accretion rate , versus time and redshift for the disc of the milky way with instantaneous recycling approximation ( thin solid lines ) and delayed feedback ( thick dashed lines ) .
_ bottom panel _ : gas accretion efficiency as a function of time with instantaneous recycling approximation ( thin solid line ) and delayed feedback ( thick dashed line ) . , width=321 ] the reconstruction of the star formation rate density as a function of time requires the parametrisation of the sfh . in section
[ sec : sfh ] we have chosen a simple first order polynomial .
we also point out that a more general power - law would give similar dependencies for @xmath40 . here
we consider three different parametrisations and show that our results do not change significantly .
we take the following functional forms : i ) a piecewise - linear function of time @xmath158 that linearly increases to a maximum sfr at @xmath159 and remains constant at higher lookback times ; ii ) a triangular function of time @xmath160 that linearly increases to a maximum sfr at @xmath159 and then decreases to @xmath58 at @xmath41 ; iii ) an exponential function of this form : @xmath161 where @xmath162 is the exponential @xmath163-folding time of the sfr . with these choices
we encompass both monotonic and non - monotonic functions . in ii ) @xmath43 is defined as the ratio between @xmath164 . in table
[ tab : fts ] , we show the values of the global @xmath43 , and the present accretion rate @xmath165 and accretion efficiency @xmath166 obtained with the four types of @xmath48 for the milky way .
we took @xmath167 ( @xmath168 ) .
for the exponential function , often used for early type galaxies ( e.g. * ? ? ?
* ) , we found a rather long scale - time of @xmath169 and a shape of the sfh very similar to that obtained with @xmath170 .
even though these three functions depict very different trends for the sfh , the disc - integrated @xmath43 varies only by @xmath171% and the values of the present accretion rate and @xmath166 change by no more than 13 % .
note that the inverse of the scalo @xmath59-parameter would correspond to a piece - wise linear functional form @xmath158 for @xmath172 .
all the above values are obtained using the i.r.a . using
the delayed return with different shapes of the sfh can produce different effects .
for instance an exponential sfh of the type @xmath173 maximizes the gas consumption timescales , which can increase up to a factor @xmath0 at the earliest times ( see * ? ? ? * ) . .
[ tab : fts ] the effect of using different parametrisations for the sfh described in the text .
@xmath43 is always defined as the ratio between the initial over the current sfr except for the triangular function where it is the ratio between the sfr at @xmath168 over the current value . [
cols="<,<,<,<",options="header " , ] showed that the dominant effect of spiral waves in galaxy discs is to churn the stars in a manner that preserves the overall angular momentum distribution .
stars move both inwards and outwards without any significant radial spreading of the disc or increase in non - circular motions . in their fig .
13 they plotted the final distribution of home radii for six different radial bins for our galaxy .
the distributions are roughly gaussian for every radial bins , with the largest dispersion @xmath174 at about @xmath175 .
so as to give an upper limit on this effect we infer the new stellar profile @xmath176 as the average of that modified by inward radial motion with parameter @xmath177 kpc and that of an outward motion with parameter @xmath178 kpc .
we found that the changes in @xmath40 are significant for some radii but there is basically no effect on the global value of @xmath43 .
we conclude that both migration and random redistribution of stars over the disc do not affect our global results .
it is a matter of debate whether galaxies had their reservoir of gas in place from the beginning or they have been harvesting the gas they needed throughout their lives .
we find that the second must be the case for all late type galaxies . a concentration of gas as large as the current disc stellar masses would imply an extremely high sfr at earlier times , a sudden formation of most of the stars and a steeply declining star formation at later times .
this is how a red sequence galaxy forms and evolves .
galaxies like the milky way or spirals of later types clearly did not have this kind of star formation history .
the evidence comes from the reconstruction of the star formation histories from the color magnitude diagrams ( e.g. * ? ? ?
* ) , from chemical evolution models ( e.g. * ? ? ?
* ) , and from cosmology ( e.g. * ? ? ?
our study further supports this point making use of the kennicutt - schmidt law .
a possible way out would be to invoke a variation of the k - s law with redshift or with metallicity as proposed by e.g.
. however , to reconcile the observations with negligible accretion at @xmath146 the variations in the k - s law should be very large .
if one required the initial gas mass of the milky way disc to be as large as the stellar mass today , the k - s law would imply an initial sfr of @xmath179 . at that rate ,
the stellar disc would form in a gyr and the subsequent evolution would be passive ( similar shape to the closed - box shown in fig .
[ fig:3radii ] ) . therefore to keep the current sfr at the observed value
the transformation of gas into stars should have been an order of magnitude less efficient than the k - s law would predict . moreover
, this efficiency should have evolved with the galaxy itself in order to keep the star formation roughly constant in time .
this is in contradiction with the fact that today galaxies with very different masses and thus different disc metallicities roughly follow the same k - s law @xcite .
thus , it seems unlikely that any realistic evolution of the k - s law can account for the evolution of discs without conspicuous gas accretion at @xmath146 . despite the lack of metallicity information in our analysis ,
we have tried to quantify the above effect by assuming a different star formation law for low density environments . to this aim , we broke the star formation law into two parts : we assumed that the standard k - s is valid only down to a break surface density of 10 @xmath65 in hydrogen , below which the law takes a different form .
we took this form from the work of @xcite who studied a compilation of extremely faint ( and presumably metal poor ) dwarf galaxies and found that a normalization of @xmath180 and a slope of @xmath181 are suitable for these low density environments .
this parametrisation makes star formation less efficient at low gas densities .
we repeated all the above calculations and found that the general result is , as expected , that more accretion is required , because star formation needs more gas to occur at the same rate .
however , the differences are not big
. for instance the amount of gas in the milky - way disc at all times increases by a factor in the range @xmath182 and the amount of accretion today only by @xmath183 .
for the other galaxies the differences can , in some cases , go up to a factor @xmath184 .
the reason for this is that the second term on the r.h.s .
( [ eq : garf1_1 ] ) does not change by more than a factor 2.3 below the break surface density .
another important result of our investigation is that gas accretion moves in time from the central to the outer parts of the discs . in other words ,
the inner parts dominate the accretion at earlier times , whilst the outer parts dominate at later times , see fig .
[ fig:3radii ] .
this inside - out evolution is also responsible for the production of metallicity gradients .
the progressive shift of accretion may be related to the angular momentum of the infalling material and it would be interesting to check it against the results of ab - initio hydrodynamical simulations .
gaseous and stellar mass as a function of the lookback time in the disc of the milky way.,width=321 ] our model also predicts the distributions of stars and gas in the disc at all times .
we found that the stellar disc of the milky way remains roughly exponential , at least in the inner parts , with a scale - length slightly increasing with time from 2.7 to @xmath185 .
the gaseous disc , initially exponential , develops an inner depression with time , as stars are formed in the central region and the accretion becomes more and more efficient in the outer parts .
[ fig : massevolution ] shows the evolution of the stellar and gaseous masses in the milky way disc .
the cumulative stellar mass increases slowly with time .
roughly @xmath186 of the stars formed at @xmath146 as a consequence of gas accretion towards the inner disc . on the other hand
, the gas mass remained almost constant and always rather low .
interestingly , the ratio between our initial and current gas mass for the milky way is the same as that obtained by comparing the hi mass in galaxies at @xmath2 with that of damped ly@xmath187 systems ( e.g. * ? ? ?
* ; * ? ? ?
we stress again that the gas total mass in fig .
[ fig : massevolution ] is calculated within @xmath188 , more gas may lay at larger radii and flow inwards with time . in a recent
paper built a model for the evolution of the milky way that has aspects in common to our investigation .
they consider the growth of a disc that goes through an initial merger phase that lasts about @xmath189 and then follows a quiescent evolution .
this second phase is analogous to the phase we are considering in this paper .
assume the variation of the scalelength of the disc with time as the inverse of the hubble parameter , so by a factor about 3 in the last @xmath190 .
this assumption together with the k - s law , used to derive the sfr , determine the amount of gas accretion onto the disc .
in contrast to this approach , we derive the @xmath75 from the normalization at redshift zero and use the k - s law to derive the amount of gas available at any time and the amount of gas needed ( gas accretion ) to reproduced that @xmath75 . then from this we obtain the evolution of the stellar and gaseous discs . from the above it is clear that the methods used in the present paper and in are rather complementary and it is remarkable that we attain a very similar evolutionary picture for the galaxy disc .
the main difference is the evolution of the stellar - disc scalelength , which evolves qualitatively in the same way but quantitatively much less in our treatment , i.e. our disc has initially @xmath191 as opposed to @xmath192 in their model ( see their figure 1 ) .
the sfrs in the solar neighbourhood are slightly different but both comparable with the determination of @xcite .
finally , they also find that the accretion rate closely follows the sfr at every time and at present is about @xmath193 , slightly higher than ours .
our analysis also shows similarities to the work of . these authors choose a different form of the star formation law and derive chemical and spectrophotometric properties of the milky - way disc . in their case , the accretion - rate is fixed and assumed to take an exponential law .
globally , all these descriptions give a consistent picture of the evolution of a galactic disc , reinforcing the validity of the different parametrizations .
our treatment does not include metallicity information but it is clear that the infalling gas is bound to be at low metallicity .
the need for metal - deficient accretion has been shown and confirmed throughout the last decades starting from pioneering works such as and @xcite .
more recently it has become clear that one needs two epochs of gas infall : one at the early epochs and one slower and still ongoing , which is the one we are describing here @xcite .
the metallicity of this gas should be fairly low , of order @xmath194 or less ( e.g. * ? ? ?
the origin of this accretion is still matter of debate and it is beyond the scope of this paper . as mentioned , several numerical investigations show that cooling of the hot corona may provide a reasonable amount of gas accretion onto the milky - way disc , although at a relative low rate @xcite .
these simulations roughly reproduce also the kinematics of this gas @xcite and the radial distribution of the accretion @xcite .
however , the whole mechanism of direct cooling via thermal instability has been recently challenged both by analytical analyses @xcite and by recent grid - based simulations @xcite and the matter is still not settled .
our accretion is smooth but this is obviously an artifact of our assumptions . in principle
the main contribution to gas infall could be minor mergers with gas rich satellites , which would give presumably a more bursty sfh . in this sense ,
our analysis is still applicable if one averages all quantities over a time larger than the typical starburst timescale .
current hi observations seem to show that the amount of gas that can be realistically brought into large galaxies by gaseous satellites in the local universe is rather low @xcite .
however , these estimates are still uncertain and only future hi surveys will allow us to determine the accretion from mergers with the required accuracy .
a recent analysis of the contribution of satellites to the gas accretion in the milky way shows that they would also not reproduce the distribution of sfr observed today @xcite .
finally , we found that accretion and star formation proceed together .
this is clear from eq .
[ eq : gar ] and from figs .
[ fig : sfh+accmw ] and [ fig : sfhsleroy ] .
this finding may just highlight the obvious point that the more a galaxy accretes gas the more stars it forms .
however , the regularity of this link seems to point at a more fundamental interplay between star formation and accretion ( see also the discussion in * ? ? ?
in a series of papers it has been suggested that gas accretion in discs is fundamentally related to star formation because it is produced by supernova ( _ positive _ ) feedback .
supernovae eject ( high - metallicity ) gas clouds out of galactic discs mixing it with the surrounding hot corona and causing prompt cooling of the latter .
this implies an accretion that follows closely the star formation in the disc and proceeds from inside out given the distribution of cloud orbits .
@xcite applied this model to the milky way and found a global accretion rate of @xmath195 .
they also found that the current accretion profile is an increasing function of the radius out to a peak at @xmath196 followed by a sudden drop further out . in this paper , we found a similar shape but with a peak located closer in , see fig .
[ fig : peak ] .
however , as mentioned , our approach traces the location where the new gas forms stars .
so it is conceivable that the infall took place at slightly larger radii , and then it flowed quiescently in .
it is compelling that two completely independent methods as these produce such similar accretion patterns and global gas accretion rates .
in this paper we have proposed a simple method to derive the amount of gas needed for the star formation to proceed in a galactic disc , using the kennicutt - schmidt law .
we found that in typical disc galaxies ( sb or later types ) most star - forming gas is not in place in the disc at the time of formation but needs to be slowly acquired from the surrounding environment .
we derived the gas accretion rate as a function of time for the milky way and other 21 disc galaxies from the things sample .
we parametrized the sfh with a simple polynomial function with a parameter ( @xmath43 ) that defines the slope of the sfh , @xmath197 stands for a flat sfh .
we summarize our results as follows : 1 . the disc of the milky way as a whole ( not only the solar neighborhood ) formed stars in the past a rate that was @xmath198 times larger than now ; 2 .
the steepness of the sfh is a function of galaxy mass and hubble type , with late type galaxies having nearly flat or inverted sfh ; 3 .
gas accretion is tightly linked to star formation , a constant ratio between the two is maintained for a large fraction of the life of a galaxy , pointing at a reciprocal interplay ; 4 .
galaxy discs must have experienced accretion of large fraction of their mass ( in gaseous form ) after @xmath1 , unless the k - s law had a strong evolution with @xmath199 ; 5 .
gas accretion progressively moves from the inner to the outer regions of galaxy discs , as a consequence star formation also proceeds inside - out ; 6 .
the efficiency of gas accretion is less than 1 and decreases with time in all galaxies expect some late - type dwarfs , as a consequence the gas consumed by star formation is not completely replenished and the galaxy eventually stops forming stars ; 7 .
the present - time accretion rates we derive are very uncertain , but even so there are indications that some large spirals may not need ongoing gas accretion .
an obvious development of this study would be to incorporate metallicity .
our results on the amount of gas infall needed for the galaxy are broadly in agreement with the amount of infalling material needed by chemical evolution models .
our independent determination of the infall rate as a function of time and radius could be taken as an input in chemical evolution models to obtain a better understanding of the accretion mechanisms that feed the star formation in galaxy discs .
we thank andrew hopkins and donatella romano for very helpful advise and stimulating discussions .
we are grateful to an anonymous referee for helpful and constructive comments .
ff was supported by the prin - miur 2008sptacc .
mt was supported for this research through a stipend from the international max planck research school ( imprs ) for astronomy and astrophysics at the universities of bonn and cologne , and by the sfb 956 `` conditions and impact of star formation '' by the deutsche forschungsgemeinschaft ( dfg ) .
m. , emsellem e. , krajnovi d. , mcdermid r. m. , serra p. , alatalo k. , blitz l. , bois m. , bournaud f. , bureau m. , davies r. l. , davis t. a. , de zeeuw p. t. , khochfar s. , kuntschner h. , lablanche p .- y . , 2011 , mnras , 416 , 1680 r. , halloin h. , kretschmer k. , lichti g. g. , schnfelder v. , strong a. w. , von kienlin a. , wang w. , jean p. , kndlseder j. , roques j. , weidenspointner g. , schanne s. , hartmann d. h. , winkler c. , wunderer c. , 2006 , nature , 439 , 45 jr .
r. c. , calzetti d. , walter f. , helou g. , hollenbach d. j. , armus l. , bendo g. , dale d. a. , draine b. t. , engelbracht c. w. , gordon k. d. , prescott m. k. m. , w. r. m. , thornley m. d. , bot c. , brinks e. , de blok e. , et al .
2007 , apj , 671 , 333 p. , chengalur j. n. , briggs f. h. , colless m. , de propris r. , pracy m. b. , de blok w. j. g. , fujita s. s. , ajiki m. , shioya y. , nagao t. , murayama t. , taniguchi y. , yagi m. , okamura s. , 2007 , mnras , 376 , 1357 d. r. , dalcanton j. j. , williams b. f. , gilbert k. m. , skillman e. d. , seth a. c. , dolphin a. e. , mcquinn k. b. w. , gogarten s. m. , holtzman j. , rosema k. , cole a. , karachentsev i. d. , zaritsky d. , 2011 , arxiv e - prints t. k. , martin d. c. , barlow t. a. , foster k. , friedman p. g. , morrissey p. , neff s. g. , neill j. d. , schiminovich d. , seibert m. , bianchi l. , donas j. , heckman t. m. , lee y .- w . , madore b. f. , milliard b. , rich r. m. , szalay a. s. , yi s. k. , 2009 , apj , 696 , 1834 | we show how the existence of a relation between the star formation rate and the gas density , i.e. the kennicutt - schmidt law , implies a continuous accretion of fresh gas from the environment into the discs of spiral galaxies .
we present a method to derive the gas infall rate in a galaxy disc as a function of time and radius , and we apply it to the disc of the milky way and 21 galaxies from the things sample . for the milky way , we found that the ratio between the past and current star formation rates is about @xmath0 , averaged over the disc , but it varies substantially with radius . in the other disc galaxies there is a clear dependency of this ratio with galaxy stellar mass and hubble type , with more constant star formation histories for small galaxies of later type .
the gas accretion rate follows very closely the sfr for every galaxy and it dominates the evolution of these systems .
the milky way has formed two thirds of its stars after @xmath1 , whilst the mass of cold gas in the disc has remained fairly constant with time . in general , all discs have accreted a significant fraction of their gas after @xmath1 .
accretion moves from the inner regions of the disc to the outer parts , and as a consequence star formation moves inside - out as well . at @xmath2
the peak of gas accretion in the galaxy is at about @xmath3 from the centre .
[ firstpage ] galaxy : evolution galaxies : star formation galaxies : ism galaxies : evolution |
the knowledge of the properties of highly compressed and heated hadronic matter is an important issue for the understanding of astrophysics such as the physical mechanism of supernovae explosions and the physics of neutron stars @xcite .
hic provide the unique opportunity to explore highly excited hadronic matter , i.e. the high density behavior of the nuclear eos , under controlled conditions ( high baryon energy densities and temperatures ) in the laboratory @xcite .
important observables have been the nucleon collective dynamics @xcite and the dynamics of produced particles such as pions and kaons @xcite
. however , the reaction dynamics is a rather complex process which involves the nuclear mean field ( eos ) and binary @xmath2-body collisions . in the presence of the nuclear medium the treatment of binary collisions represents a non - trivial topic .
the nn cross sections for elastic and inelastic processes , which are the crucial physical parameters here , are experimentally accessible only for the free space and not for @xmath2-body scattering at finite baryon density .
recent microscopic studies , based on the @xmath3-matrix approach , have shown a strong decrease of the elastic nn cross section @xcite in the presence of a hadronic medium .
these in - medium effects of the elastic nn cross section considerably influence the hadronic reaction dynamics @xcite .
obviously the question arises whether similar in - medium effects of the _ inelastic _ nn cross sections may affect the reaction dynamics and , in particular , the production of particles ( pions and kaons ) .
since microscopic results are not available , we discuss here in a simple phenomenological way possible density modifications of the inelastic nn cross sections and their influences on particle multiplicities , rapidity distributions and ratios .
we find a strong dependence of the yields and rapidity distributions on the in - medium modifications of the inelastic cross sections , but on the other hand , this effect is only moderate for particle ratios such as @xmath0 , and almost vanishes for @xmath1 .
therefore such ratios turn out to be robust observables in determining the nuclear eos and , particularly , the isovector channel of the nuclear mean field @xcite .
in this chapter we briefly discuss the transport equation by concentrating on the treatment of the cross sections , which are the important parameters of the collision integral .
the theoretical description of hic is based on the kinetic theory of statistical mechanics , i.e. the boltzmann equation @xcite .
the relativistic semi - classical analogon of this equation is the relativistic boltzmann - uehling - uhlenbeck ( rbuu ) equation @xcite @xmath4 f(x , k^ { * } ) = \frac{1}{2(2\pi)^9 } \nonumber\\ & & \times \int \frac{d^3 k_{2}}{e^{*}_{{\bf k}_{2 } } } \frac{d^3 k_{3}}{e^{*}_{{\bf k}_{3 } } } \frac{d^3 k_{4}}{e^{*}_{{\bf k}_{4 } } } w(kk_2|k_3 k_4 ) \left [ f_3 f_4 \tilde{f}\tilde{f}_2 -f f_2 \tilde{f}_3\tilde{f}_4 \right ] \label{rbuu } \end{aligned}\ ] ] where @xmath5 is the single particle distribution function . in the collision term
the short - hand notation @xmath6 for the particle and @xmath7 and the hole - distribution is used .
the collision integral exhibits explicitly the final state pauli - blocking while the in - medium scattering amplitude includes the pauli - blocking of intermediate states .
( 8.,7.3 ) ( 3.0,0.3 ) the dynamics of the lhs of eq.([rbuu ] ) , the drift term , is determined by the mean field . here
the attractive scalar field @xmath8 enters via the effective mass @xmath9 and the repulsive vector field @xmath10 via kinetic momenta @xmath11 and via the field tensor @xmath12 .
the in - medium cross sections enter into the collision integral via the transition amplitude @xmath13 with @xmath3 the in - medium scattering matrix element . in the kinetic equation ( [ rbuu ] ) one should use both physical quantities , the mean field ( eos ) and the collision integral ( cross sections ) according to the same underlying effective two - body interaction in the medium , i.e. the in - medium t - matrix ; @xmath14 , respectively @xmath15 .
however , in most practical applications phenomenological mean fields and cross sections have been used . in these models adjusting the known bulk properties of nuclear matter around the saturation point one tries to constrain the models for supra - normal densities with the help of heavy ion reactions @xcite .
medium modifications of the nn cross section are usually not taken into account which works , in comparison to experimental data , astonishingly well @xcite .
however , in particular kinematics regimes a sensitivity of dynamical observables such as collective flow and stopping @xcite or transverse energy transfer @xcite to the elastic nn cross section has been observed .
( 8.,7.3 ) ( 3.0,0.3 ) fig .
[ fig1 ] shows the energy dependence of the in - medium neutron - proton @xmath16 cross section at fermi momenta @xmath17 , corresponding to @xmath18 ( @xmath19 is the nuclear matter saturation density ) as found in relativistic dirac - brueckner ( db ) calculations @xcite .
the presence of the medium leads to a substantial suppression of the cross section which is most pronounced at low laboratory energy @xmath20 and high densities where the pauli - blocking of intermediate states is most efficient . at larger @xmath20
asymptotic values of 15 - 20 mb are reached . however , not only the total cross section but also the angular distributions are affected by the presence of the medium .
the initially strongly forward - backward peaked @xmath21 cross sections become much more isotropic at finite densities @xcite which is mainly do to the pauli suppression of soft modes ( @xmath22-exchange ) and correspondingly of higher partial waves in the t - matrix @xcite .
obviously one expects similar in - medium effects for the inelastic nn cross sections mainly due to pauli - blocking of intermediate scattering states and in - medium modified matrix elements .
such microscopic studies for inelastic processes are very rare or still in development @xcite . however , to explore the sensitivity we use here a rather simple parametrization which assumes a reduction of the inelastic nn cross section with increasing baryon density , in line with that of fig .
[ fig1 ] for the elastic one which has previously been used in ref .
this can be achieved by assuming a factorization of the effective matrix element of the form @xmath23 with @xmath24 the vacuum matrix element ( taken from experimental free scattering data ) and @xmath25 a density dependent function depending on a quenching parameter @xmath26 .
a very simple parametrization of the function @xmath27 is shown in fig .
[ fig2 ] ( bottom panel ) for different values of the parameter @xmath26 .
the effect on the inelastic nn cross section is shown on the top of fig .
[ fig2 ] for the choice @xmath28
. a significant reduction of the effective inelastic nn cross section ( by a factor of 2 ) is observed with respect to that of the free case at a given baryon density @xmath29 .
we have applied the parametrization ( [ meff ] ) in the collision integral of the transport equation ( [ rbuu ] ) and analyzed the transport calculations in terms of particle production . at energies below 1.6 agev pions and kaons
are produced , where the second ones do not significantly affect the reaction dynamics due to their small production cross sections .
in particular , the kaons are treated perturbatively and can be produced from @xmath30 and @xmath31 .
here @xmath32 stands for a nucleon or @xmath33 resonance and @xmath34 for a hyperon .
pions are created via the decay of the @xmath35 resonance . for more details
we refer to @xcite . for the nuclear mean field the @xmath36 parametrization of the non - linear walecka model @xcite
is adopted here with a compression modulus of @xmath37 mev and an effective dirac mass of @xmath38 ( @xmath39 is the bare nucleon mass ) .
the momentum dependence enters via the relativistic treatment through the vector components of the baryon self energy . in order to keep the discussion transparent
, we do not apply here any isovector components of the baryon self energy , as it has been previously done in ref .
this implies no effective mass splitting between protons and neutrons , and on the other hand , between the different isospin states of the @xmath33 resonance ( @xmath40 ) and of the hyperons ( @xmath41 ) .
furthermore , pions are propagated in a coulomb field and kaons do not experience any potential .
( 3.0,0.3 ) we start the discussion with the time evolution of the multiplicities of the @xmath33 resonances and of the pions , as can be seen in fig .
the time evolution of the multiplicity of produced @xmath33-resonances is shown with their maximum around 15 fm / c which corresponds to the time of maximum compression .
due to the finite lifetime these resonances decay into pions ( and nucleons ) according @xmath42 ( some of these pions are re - absorbed in the inverse process , i.e. @xmath43 ) .
this mechanism continues until all resonances have decayed leading to a final constant pion yield for times @xmath44 fm / c ( the so - called freeze - out time ) .
after the freeze - out the pions can be measured experimentally .
the experimental pion multiplicity is schematically shown in fig .
[ fig2 ] by the gray band for central au+au collisions @xcite .
we observe an essential reduction of the pion multiplicity using the effective inelastic nn cross sections , in line with the previous fig .
[ fig2 ] . as an important result the calculations with the effective inelastic nn cross section describe the experimental data reasonably well for a quenching parameter of about @xmath28 . however , for a more realistic comparison one should use more microscopic calculations of the inelastic cross section , as it has been done in ref . @xcite .
such a progress is under study .
the strong in - medium dependence of the inelastic cross sections is shown also in the rapidity distributions of pions and kaons in fig .
[ fig4 ] .
( 3.0,0.3 ) the rapidity distribution of kaons is even more stronger affected by the density behavior of the inelastic cross section with respect to that of pions .
this is due to the fact that the leading channels for kaon production are @xmath45 and pionic ones , which are both reduced when the in - medium dependent @xmath46 are used .
the question arises whether particle ratios are influenced by in - medium effects of the inelastic cross section .
an answer on this question is of major importance , particularly for kaons and less for pions , since particle rations have been widely used in determining the nuclear eos at supra - normal densities .
relative ratios of kaons between different colliding systems have been used in determining the isoscalar sector of the nuclear eos , see refs .
more recently , the @xmath0- and @xmath1-ratios have been used in exploring the high density behavior of the symmetry energy , i.e. the isovector part of the nuclear mean field @xcite .
.in - medium dependence of the @xmath0- and @xmath1-ratios for central ( @xmath47 fm ) au+au collisions at 1.48 agev incident energy .
[ cols="^,^,^",options="header " , ] table [ table1 ] shows the @xmath0- and @xmath1-ratio for central ( @xmath47 fm ) au+au collisions at 1.48 agev incident energy using both , the free and in - medium inelastic cross sections .
the pionic ratio moderately depends on the density dependence of the inelastic cross section , but the @xmath1-ratio is not particularly affected by the in - medium effects . a possible reason for the different behavior between the pionic and kaonic ratios may be the fast pre - equilibrium emission of the kaons .
in particular , kaons are created very early during the formation of the high density phase and are emitted from the compression region without undergoing any interaction with the hadronic environment .
therefore one obviously expects a direct relation between the high density effects of the inelastic cross sections and the @xmath1-ratio .
pions , on the other hand , are strongly interacting with the medium via secondary re - absorption processes and thus are emitted from different stages of a collision which may influence the final @xmath0-ratio @xcite .
it turns out that the strangeness ratio presents a robust observable to investigate the isovector character of the nuclear matter eos .
we have investigated the role of the density dependence of the inelastic cross section on particle production in intermediate energy heavy ion collisions within a covariant transport equation of a boltzmann type . since microscopic studies on the in - medium behavior of inelastic cross sections are still rare , we have used here a simple phenomenological in - medium dependence of the inelastic cross sections .
we have applied the transport equation to au+au collisions at intermediate relativistic energies below the kaon threshold energy .
our studies have shown a strong sensitivity of the particle multiplicities and rapidity distributions of pions and kaons .
in particular , a reduction by a factor of @xmath2 for pions has been seen when the in - medium effects in the inelastic cross section are accounted for .
consequently , the kaon ( @xmath48 ) yields decrease more than 50 % .
this is due to the asumption of a reduction of the inelastic cross section at high densities . as an interesting
finding , the multiplicities of @xmath49 and @xmath50 are influenced in such a way that their ratio is independent on the density dependence of the inelastic cross sections .
this may be due to the long mean free path of the @xmath48 , whereas the pionic ratio , due to their strong secondary interaction processes with the hadronic environment , have shown a moderate dependence on the density behavior of the inelastic cross sections .
certainly more systematic studies are necessary to investigate better the mechanism which leads to a moderate relation between the in - medium dependence of the inelastic cross sections and the pionic ratio .
a comparison with more experimental data would be helpful in determining more precisely the density dependence of the inelastic cross sections , which will be an object of future studies . on this level of investigations
we conclude that the @xmath1-ratio represents a robust observable in determining the nuclear matter eos at supra - normal densities .
+ _ acknowledgment _ this work is supported by bmbf , grand 06lm189 .
one of the authors ( v.p . ) would like to thank h.h .
wolter and m. di toro for the warm hospitality during her short stays at their institutes .
www99 t. gaitanos , c. fuchs , h.h .
wolter , phys .
lett . * b609 * ( 2005 ) 241 ; + e. santini , t. gaitanos , m. colonna , m. di toro , nucl . phys . *
a756 * ( 2005 ) 468 ; + t. gaitanos , c. fuchs , h.h .
wolter , prog .
nucl . phys .
* 53 * ( 2004 ) 45 . | the in - medium modification of nucleon - nucleon ( nn ) cross sections is investigated by means of particle production in heavy ion collisions ( hic ) at intermediate energies .
in particular , the density dependence of the _ inelastic _ cross sections considerably affects the pion and kaon yields and their rapidity distributions .
however , the @xmath0- and @xmath1-ratios depend only moderately on the in - medium behavior of the inelastic cross sections .
it turns out that particle ratios seem to be robust observables in determining the nuclear equation of state ( eos ) and , particularly , its isovector sector . |
in this section , we will consider the disk open string whose boundary is attached to the @xmath0 d3-brane , and compute the disk open string amplitudes by using the semi - classical technique .
we concentrate on the exponent of the correction in the large @xmath33 limit .
we postpone the integral on the moduli space and the one - loop determinant to future works .
we find that the exponent of the correction agree with the matrix model result .
actually , there are perturbative corrections to the expression .
we might not be allowed to retain the non - perturbative correction calculated in this section , since it is much smaller than the perturbative corrections .
this may be justified by the supersymmetry but we leave it to future works .
it is convenient to use the coordinate system of @xcite .
the metric of @xmath39 is expressed as @xmath40,\qquad l^2=\alpha'\sqrt{\lambda},\\ 0\le \eta \le \frac{\pi}{2},\qquad 0\le \psi \le
2\pi,\qquad 0\le \rho,\qquad 0\le \theta\le \pi,\qquad 0\le \phi \le 2\pi . \end{aligned } \label{metric}\end{aligned}\ ] ] the @xmath0 d3-brane worldvolume @xcite is expressed in this coordinate system as @xmath41 the electric field on the d3-brane worldvolume is excited and takes the value @xmath42 where we identify @xmath43 and @xmath44 as worldvolume coordinates .
now let us consider the disk string worldsheet whose boundary is attached to the d3-brane expressed by eqs .
let the worldsheet coordinates be @xmath45 .
the coordinate @xmath46 is the radial coordinate of the disk .
@xmath47 corresponds to the boundary , while @xmath48 to the center of the disk .
the other coordinate @xmath49 is the angular coordinate of the disk . the string worldsheet action is written as @xmath50 where @xmath51 is the induced metric and @xmath52 is the gauge field . in this letter , we consider the following special ansatz . @xmath53 since the center of the worldsheet @xmath48 is one point and should be mapped to one point in spacetime , the condition @xmath54 is imposed .
meanwhile the boundary value of @xmath55 is denoted by @xmath56 .
this boundary of the string is attached to the d3-brane , and it gives a constraint on the boundary value of @xmath44 as @xmath57 . putting this ansatz into the action ,
we obtain ( prime `` @xmath58 '' denotes the @xmath59 derivative ) @xmath60 the constant shift of the boundary action is fixed so that @xmath61 at @xmath62 where the boundary of the worldsheet shrinks to a point .
if we fix the boundary value @xmath63 , the bulk action has the lower bound @xmath64 this bound is saturated when @xmath65 this configuration actually satisfies the equations of motion derived from the bulk action .
this configuration is shown in figure [ fig - bubble ] . when we derive the bound , we assume the boundary @xmath56 is fixed .
however this is not the true boundary condition ; the boundary of the string worldsheet can move along the d3-brane . in this sense
the configuration is not a stationary point of the action .
though the configuration is not a solution , it is still useful to evaluate the path - integral ; it is `` the bottom of the trough '' @xcite .
we explain here how to evaluate the path - integral using the configuration .
we want to evaluate the path - integral @xmath67 with the correct boundary condition determined by the configuration of the d3-brane .
this integral can be rewritten as @xmath68 in the path - integral @xmath69 the boundary value of @xmath55 is fixed to @xmath63 .
hence is the saddle point of this integral .
the path - integral @xmath69 can be evaluated by the point . @xmath70
the @xmath63 integral in is written as @xmath71= \exp[-t(\pi/2)]-\exp[-t(0)],\label{t}\end{aligned}\ ] ] where @xmath72 is defined as a solution of @xmath73 since @xmath74 is proportional to @xmath75 , we can take @xmath72 as @xmath76 the first term on the right - hand side of eq . is @xmath77=\exp[-{\widetilde{s}}(\pi/2)+o(\log \sqrt{\lambda } ) ] = ( \text{powers of } \lambda).\end{aligned}\ ] ] thus this term captures the perturbative corrections . the second term on the right - hand side of eq .
is the exponentially small term that we want to see here .
it is written as @xmath78=\exp\left[-{\widetilde{s}}(0)+o(\log \sqrt{\lambda})\right ] \cong \exp\left[-\sqrt{\lambda}\left(\sqrt{1+\kappa^2}-1\right)\right ] .
\label{disk - result}\end{aligned}\ ] ] this is the main result of this letter .
the result agrees with the matrix model result .
this small non - perturbative effect can be understood qualitatively as follows .
there are two forces acting on the string end point : the string tension pulling the end point inside , and the electric force pushing the end point outside .
the string tension is always larger than the electric force , and there is no stationary point other than the constant map .
however , as the string worldsheet becomes larger and larger , the difference of these two forces becomes smaller and smaller . actually when the worldsheet is large enough ( @xmath79 )
, the two forces almost cancel each other and the worldsheet boundary can be moved almost freely without increasing or decreasing the action .
in other words , however large the worldsheet becomes , the action remains finite .
the correction is the result of this effect .
this kind of exotic effect is not present for a flat d - brane in the flat space . in this case
, the action diverges as the worldsheet becomes larger .
thus the contribution to the amplitudes is zero .
therefore the dbi action does not capture this effect since it is based on the small curvature approximation .
there is an intuitive explanation why the configuration with @xmath80 produces the term @xmath20 .
it is proposed in @xcite how the eigenvalue distribution of the gaussian matrix model can be seen in @xmath1 .
figure [ eigen - string ] represents the configuration in the picture of @xcite .
when @xmath63 is close enough to @xmath81 , this picture around center is similar to figure [ eigen - matrix ] ; the fundamental string looks like @xmath30 and the d3-brane look like @xmath28 around the center , is supposed to be different from the one used in the circular fundamental wilson loop @xcite .
one can distinguish these two , for example , by the value of the action .
our string attached to the d3-brane has the action @xmath82 .
on the other hand the on - shell action of the string which represent the fundamental wilson loop is @xmath83 . ] . from the similar point of view , one may guess a branched d3-brane configuration , shown in figure [ branched - d3 ] , contributes as another monomial in the symmetric wilson loop . calculating this contribution
is an interesting future problem . '' of @xcite , while vertical direction is the radial direction of @xmath84 fiber .
the d3-brane is represented by the red solid vertical line .
the fundamental string is represented by the blue dash - dotted line.,width=302 ] i would like to thank gordon w. semenoff for useful discussions .
i am also grateful to nadav drukker for careful reading of the manuscript and useful comments .
this work was supported in part by the european research training network contract 005104 `` forcesuniverse . ''
j. m. maldacena , `` the large n limit of superconformal field theories and supergravity , '' adv . theor .
* 2 * ( 1998 ) 231252 [ http://arxiv.org/abs/hep-th/9711200[arxiv:hep-th/9711200 ] ] .
rey and j .- t .
yee , `` macroscopic strings as heavy quarks in large n gauge theory and anti - de sitter supergravity , '' eur .
j. * c22 * ( 2001 ) 379394 [ http://arxiv.org/abs/hep-th/9803001[arxiv:hep-th/9803001 ] ] .
j. m. maldacena , `` wilson loops in large n field theories , '' phys .
* 80 * ( 1998 ) 48594862 [ http://arxiv.org/abs/hep-th/9803002[arxiv:hep-th/9803002 ] ] .
n. drukker and b. fiol , `` all - genus calculation of wilson loops using d - branes , '' jhep * 02 * ( 2005 ) 010 [ http://arxiv.org/abs/hep-th/0501109[arxiv:hep-th/0501109 ] ] .
s. a. hartnoll and s. prem kumar , `` multiply wound polyakov loops at strong coupling , '' phys
* d74 * ( 2006 ) 026001 [ http://arxiv.org/abs/hep-th/0603190[arxiv:hep-th/0603190 ] ] .
s. yamaguchi , `` wilson loops of anti - symmetric representation and d5-branes , '' jhep * 05 * ( 2006 ) 037 [ http://arxiv.org/abs/hep-th/0603208[arxiv:hep-th/0603208 ] ] .
j. gomis and f. passerini , `` holographic wilson loops , '' jhep * 08 * ( 2006 ) 074 [ http://arxiv.org/abs/hep-th/0604007[arxiv:hep-th/0604007 ] ] .
d. rodriguez - gomez , `` computing wilson lines with dielectric branes , '' nucl .
phys . * b752 * ( 2006 ) 316326 [ http://arxiv.org/abs/hep-th/0604031[arxiv:hep-th/0604031 ] ] .
k. okuyama and g. w. semenoff , `` wilson loops in n = 4 sym and fermion droplets , '' jhep * 06 * ( 2006 ) 057 [ http://arxiv.org/abs/hep-th/0604209[arxiv:hep-th/0604209 ] ] . s. a. hartnoll and s. prem kumar , `` higher rank wilson loops from a matrix model , '' jhep * 08 * ( 2006 ) 026 [ http://arxiv.org/abs/hep-th/0605027[arxiv:hep-th/0605027 ] ] .
s. a. hartnoll , `` two universal results for wilson loops at strong coupling , '' http://arxiv.org/abs/hep-th/0606178[arxiv:hep-th/0606178 ] .
b. chen and w. he , `` on 1/2-bps wilson-t hooft loops , '' http://arxiv.org/abs/hep-th/0607024[arxiv:hep-th/0607024 ] .
s. giombi , r. ricci , and d. trancanelli , `` operator product expansion of higher rank wilson loops from d - branes and matrix models , '' http://arxiv.org/abs/hep-th/0608077[arxiv:hep-th/0608077 ] .
tai and s. yamaguchi , `` correlator of fundamental and anti - symmetric wilson loops in ads / cft correspondence , '' http://arxiv.org/abs/hep-th/0610275[arxiv:hep-th/0610275 ] .
j. gomis and f. passerini , `` wilson loops as d3-branes , '' http://arxiv.org/abs/hep-th/0612022[arxiv:hep-th/0612022 ] .
n. drukker , s. giombi , r. ricci , and d. trancanelli , `` on the d3-brane description of some 1/4 bps wilson loops , '' http://arxiv.org/abs/hep-th/0612168[arxiv:hep-th/0612168 ] .
tai , `` two polyakov loop correlators from d5-branes at finite temperature , '' http://arxiv.org/abs/hep-th/0701006[arxiv:hep-th/0701006 ] .
s. yamaguchi , `` bubbling geometries for half bps wilson lines , '' http://arxiv.org/abs/hep-th/0601089[arxiv:hep-th/0601089 ] .
o. lunin , `` on gravitational description of wilson lines , '' jhep * 06 * ( 2006 ) 026 [ http://arxiv.org/abs/hep-th/0604133[arxiv:hep-th/0604133 ] ] .
n. drukker , `` 1/4 bps circular loops , unstable world - sheet instantons and the matrix model , '' jhep * 09 * ( 2006 ) 004 [ http://arxiv.org/abs/hep-th/0605151[arxiv:hep-th/0605151 ] ] .
j. gomis and t. okuda , `` wilson loops , geometric transitions and bubbling calabi- yau s , '' http://arxiv.org/abs/hep-th/0612190[arxiv:hep-th/0612190 ] .
j. k. erickson , g. w. semenoff , and k. zarembo , `` wilson loops in n = 4 supersymmetric yang - mills theory , '' nucl .
phys . * b582 * ( 2000 ) 155175 [ http://arxiv.org/abs/hep-th/0003055[arxiv:hep-th/0003055 ] ] .
n. drukker and d. j. gross , `` an exact prediction of n = 4 susym theory for string theory , '' j. math . phys .
* 42 * ( 2001 ) 28962914 [ http://arxiv.org/abs/hep-th/0010274[arxiv:hep-th/0010274 ] ] .
s. r. coleman , `` the uses of instantons . ''
lecture delivered at 1977 int .
school of subnuclear physics , erice , italy , jul 23-aug 10 , 1977 .
d. berenstein , r. corrado , w. fischler , and j. m. maldacena , `` the operator product expansion for wilson loops and surfaces in the large n limit , '' phys . rev .
* d59 * ( 1999 ) 105023 [ http://arxiv.org/abs/hep-th/9809188[arxiv:hep-th/9809188 ] ] . | in the ads / cft correspondence , an @xmath0 d3-brane with electric flux in @xmath1 spacetime corresponds to a circular wilson loop in the symmetric representation or a multiply wound one in @xmath2 super yang - mills theory . in order to distinguish the symmetric loop and the multiply wound loop , one should see an exponentially small correction in large t hooft coupling .
we study semi - classically the disk open string attached to the d3-brane .
we obtain the exponent of the term and it agrees with the result of the matrix model calculation of the symmetric wilson loop .
[ cols=
" < " , ] however , what we really want to do is to take the limit @xmath3 with @xmath4 kept finite . in this case , @xmath5 and @xmath6 are in the same order in @xmath7 ( linear in @xmath7 ) .
therefore we can not ignore the term @xmath6 when searching the saddle point in large @xmath7 .
the function @xmath8 looks like figure [ vy ] ( b ) .
this function is not flat at all in the region @xmath9 .
thus the expression is not valid in this case .
we find the minimum at a certain point @xmath10 instead .
let us evaluate the integral by the point @xmath11 .
@xmath12 is the solution of the equation @xmath13 then eq .
is solved as @xmath14 here we add the argument to @xmath15 in order to remember that it depends on @xmath16 . as a result
, we can approximate the integral in the large @xmath7 limit as @xmath17=\exp[2n(\kappa\sqrt{1+\kappa^2}+\sinh^{-1}\kappa ) ] .
\label{mk}\end{aligned}\ ] ] this is the same result as in @xcite . for later convenience , we define the quantity @xmath18 as @xmath19 next , we turn to the calculation of @xmath20 . this expectation value is written as @xmath21+(k-1)y_1+y_2\right).\label{int2}\end{aligned}\ ] ] as the same way as above , let us define the effective potential @xmath22 by @xmath23\right ) .
\label{eff2}\end{aligned}\ ] ] then eq . is written as @xmath24 actually , in the large @xmath7 limit , eq . leads to the following expression of @xmath25 .
@xmath26 next let us turn to the @xmath27 integral .
we perform the @xmath28 integral first by the saddle point @xmath29 . as for @xmath30 integral , the approximation like eq .
is valid since the situation is similar to eq . with @xmath31 . as a result ,
eq .
becomes @xmath32 so far we only take large @xmath7 limit with @xmath4 and @xmath33 kept finite .
thus , the expression is valid for finite @xmath33 .
now we take the large @xmath33 limit with @xmath34 kept finite . in this limit
, we can use the saddle point approximation at @xmath35 . the integral become @xmath36 = \exp\left[-{{\cal f}}_{(k)}-\sqrt{\lambda}\sqrt{1+\kappa^2}+\sqrt{\lambda}\right].\end{aligned}\ ] ] the final result of this section is @xmath37 .
\label{matrix - result}\end{aligned}\ ] ] this is actually exponentially small in the large @xmath33 limit . in the next section , we will compare this exponent to the worldsheet non - perturbative correction .
let us make a comment on the eigenvalue distribution here . in the large @xmath7 and large @xmath33 limit
, the eigenvalue distribution at the saddle point looks like figure [ eigen - matrix ] .
this will be compared to the brane configuration in the ads side of the calculation . .
the black thick line denotes the semi - circle distribution of eigenvalues @xmath38.,width=377 ] |
is considered the prototype for a class of evolved binaries with peculiar properties .
the effective temperatures and luminosities of the members of this class suggest that they are in the post - asymptotic giant branch ( post - agb ) phase of their evolution , but their evolutionary path is thought to be severely affected by the presence of a close companion ( see * ? ? ? * for a review ) .
like the other members of its class , hr 4049 shows a significant infrared ( ir ) excess and a time - variable optical and ultraviolet ( uv ) deficit @xcite , which suggests the presence of a massive circumbinary disk .
in addition , its photosphere shows a severe depletion in refractory elements , but it has roughly solar abundances in volatiles .
this unusual depletion pattern is generally attributed to the formation of dust ( incorporating refractory elements ) in a circumbinary disk , followed by the re - accretion of the depleted gas onto the star @xcite .
since the circumbinary disk plays an important role in determining the properties of hr 4049 , it has been studied extensively ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
while the ir excess is a clear indication of the presence of dust in a stable disk around the system , it has been difficult to determine the nature of the dust in the disk .
indeed , it is not even clear whether the dust in the disk is oxygen - rich or carbon - rich . while the ir spectrum shows the clear presence of gas that is typically associated with oxygen - rich environments ( e.g. co@xmath0 , h@xmath0o , oh ; @xcite ; see also ( * ? ? ? * paper i hereafter ) ) there is no trace of corresponding oxygen - rich dust features such as silicates or oxides .
instead , the spectral energy distribution ( sed ) of the disk resembles a 1150 k black body down to sub - millimeter wavelengths @xcite , while also showing strong emission features due to polycyclic aromatic hydrocarbons ( pahs ) . this pah emission however
does not originate from the disk , but from what appear to be bipolar lobes @xcite .
the 1150 k black body sed can be reproduced by a wall model " , in which the circumbinary disk is vertically extended and the dust is very optically thick , effectively producing a radiating inner wall at a temperature of @xmath5 k , with a scale height of @xmath63 au at the inner wall distance of 10 au @xcite .
because of the high opacity , the dust beyond the inner rim is cold and does not contribute to the ir excess .
however , the wall model has been challenged by recent interferometric observations by @xcite that show a more extended distribution of material emitting strongly in the ir than the wall model can accommodate . instead , these authors suggest emission from optically thin dust with smooth opacity profiles . in paper
i , we present a detailed analysis of the molecular emission in the mid - ir spectrum and conclude that neither of these models is fully consistent with the properties of the molecular gas ( in particular co@xmath0 , h@xmath0oand co ) . indeed ,
while we find that the gas originates from a radially extended disk , we also determine that it is very optically thick across most of the mid - ir .
radiative trapping by this optically thick gas then results in a warmer and more homogeneous temperature structure than previously considered for the disk .
additional clues to the geometry and properties of the disk can be inferred from observations in the near - ir at high spectral resolution .
@xcite presented such observations , examining three regions of the near - ir spectrum of hr 4049 at 2.3 , 3.0 and 4.6 @xmath1 m .
they detect many co , oh and h@xmath0olines and identify distinct components in the system .
they propose that the absorption in the co overtone originates from gas in keplerian rotation along the inner rim of the disk in the wall model .
furthermore , they suggest that the gas is slowly streaming out over the edge of the wall and over the disk , causing the more complex emission - absorption line profiles in the 4.6 @xmath1mregion . finally , they found evidence for a cold gas component .
clearly , the near - ir observations contain a lot of information about the properties , geometry and kinematics of the gas disk .
it is thus important to investigate whether we can reconcile these data with the disk model inferred from mid - ir observations .
therefore , we re - examine the near - ir data presented by @xcite .
this paper is organized as follows . in section [ sec : observations ] we briefly describe the observational data .
we present our analysis of the spectrum in section [ sec : analysis ] , discuss our results in section [ sec : discussion ] and present our conclusions in section [ sec : conclusion ] .
the data we discuss in this paper are high - resolution ( @xmath7 ) observations in the near - ir ( @xmath8 m ) carried out with the phoenix spectrograph @xcite from the national optical astronomy observatory ( noao ) mounted on gemini south .
these observations primarily targeted co fundamental and overtone lines and were described and discussed previously by @xcite .
a large number of telluric lines are present in the near - ir ; fig .
[ fig : overtone_fits ] shows an illustrative atmospheric radiance spectrum from hitran @xcite on the web .
additional observations of a hot star ( without stellar lines in this wavelength range ) at the same airmass as hr 4049 were used to divide out these lines .
this telluric correction is generally adequate ; however , some residuals remain due to imperfect cancellations , especially in wavelength ranges where there are a lot of lines ( e.g. near 2.319 @xmath1 m ; see fig .
[ fig : overtone_fits ] ) . in most cases
though , the residuals are significantly smaller than the depth of the lines we study here .
the telluric lines in the hot star spectrum were also used to achieve a very accurate wavelength calibration , with residuals of typically 0.25 km s@xmath9 . for more details about these observations or the data reduction aspects ,
we refer to @xcite . for each individual data segment
, we then performed a heliocentric velocity correction and normalized the data to the continuum .
next , we used a weighted mean ( adopting the signal - to - noise , s / n , ratios provided by * ? ? ?
* ) to merge all the individual segments into a single , final spectrum .
finally , we rebinned the resulting spectrum onto a constant resolution wavelength grid for later comparison to our model spectra .
we begin our analysis of these data with the co overtone lines in the 2.3 @xmath1mregion .
there are several good reasons for this .
first , in this range , the spectrum reveals simple , pure absorption profiles with relatively little blending between the lines . since they are intriniscally much weaker , these lines will furthermore have a much lower optical depth than the fundamental lines for a given column density , thus offering the best prospects to reliably determine the temperatures and column densities .
finally , since the gas in this region of the spectrum is absorbing continuum radiation , its location is constrained : it must be directly along the line of sight to the dusty disk and/or the central star ; note however that the stellar continuum will only contribute @xmath625% of the flux at this wavelength .
we will then use our results from the 2.3 @xmath1mrange to better interpret the emission - absorption spectrum at 4.6 @xmath1 m . the combined spectrum of hr 4049 in the 2.3 @xmath1mrange is shown in fig . [ fig : overtone_fits ] .
we remind the reader that @xcite used these data to produce an excitation diagram for co , and found that the absorption originates from two distinct layers of co gas in local thermodynamic equilibrium ( lte ) : a warm layer ( @xmath10 k , @xmath11 10@xmath12 @xmath13 ) and a cold one ( @xmath14 k , @xmath15 10@xmath16 @xmath13 ) .
we first used these parameters to create a plane - parallel slab model in which 1200 k black body radiation ( representing the dust ) is absorbed by the warm layer of co gas which is in turn absorbed by the cold layer of co gas .
we used the co line list from @xcite and adopted a gaussian line profile with a width of 10 km s@xmath9 .
this corresponds to the measured full width at half maximum ( fwhm ) for the overtone lines ; note that @xcite suggest 16 km s@xmath9 ; however , this results in lines that are too broad ( see also sect . [
sec : discussion_kinematics ] ) . when we compared this model to the observations ( see fig .
[ fig : overtone_fits ] ) , we found that the model reproduces neither the depths nor the relative strength ratios of the individual rovibrational lines .
this discrepancy could be due to the use of a different line list ; at any rate , it warrants an independent analysis .
[ [ excitation - diagram ] ] excitation diagram + + + + + + + + + + + + + + + + + + first , we constructed our own excitation diagram for the co first overtone @xmath17 lines .
we fit a gaussian to each co line and integrated these to determine the equivalent width ( @xmath18 ) .
then we determined the population of each level using @xmath19 where @xmath20 is the column density ( in @xmath13 ) , @xmath21 is the oscillator strength for a rovibrational transition ( obtained from the co line list by * ? ? ?
* ) , @xmath22 is the central wavelength of the transition and both @xmath18 and @xmath22 are measured in @xcite .
we present the resulting excitation diagram in fig .
[ fig : co_excitation ] .
qualitatively , our results are similar to those obtained by @xcite , in the sense that our excitation diagram is consistent with two layers of optically thin co in lte a warm and a cold one .
however , we find very different values for the temperatures and column densities of the co gas : our `` hot '' layer has a temperature of 970 @xmath23 40 k and a column density of ( 1.19 @xmath23 0.04)@xmath24 @xmath13 .
the temperature we determine is nearly twice that found by @xcite and the column density is two orders of magnitude larger . for the `` cold '' layer ,
we find a temperature of 40 @xmath23 10 k and a column density of ( 6.1 @xmath23 0.5)@xmath25 @xmath13 . here , the temperature and column density are relatively similar to those determined by @xcite
. lines indicated by diamonds , the @xmath26 points indicated by crosses and the @xmath27co points indicated by squares .
the gray triangles indicate the points on the excitation diagram presented by @xcite .
the solid black lines indicate the best fits to the different components .
the dashed line shows the extrapolation of the lines originating from the ground state to those from the first vibrationally excited level.,scaledwidth=50.0% ] if we extrapolate the hot component to higher energies , we find that it also fits the overtone lines originating from the first vibrationally excited state ( i.e. the @xmath26 lines ) relatively well .
this suggests that the co may be in or close to vibrational lte . for the @xmath27co lines ,
we find a column density of ( 5.5 @xmath23 0.6)@xmath28 10@xmath16 @xmath13 .
this yields a @xmath29c/@xmath27c ratio of 2.1 @xmath23 0.9 , which is a little lower the value of 6@xmath30 reported by @xcite .
[ [ model - spectra ] ] model spectra + + + + + + + + + + + + + as an independent check , we modeled the co absorption at 2.3 @xmath1musing the same methods employed to build the spectrafactory database @xcite .
for these models , we began once again with co line lists from ( * ? ? ?
* including the @xmath27co , c@xmath16o and c@xmath12o isotopologues ) . from these
, we calculated optical depth profiles assuming a population in lte and a gaussian intrinsic line profile with a width of 0.5 km s@xmath9(consistent with the @xmath31 value derived by * ? ? ?
note though that since the lines are optically thin , the precise value of the intrinsic line width in our approach does not matter much as long as the line width is much smaller than a resolution element .
we tested a few different model configurations .
initially , we attempted to fit a model consisting of a single layer of lte gas absorbing a 1200 k black body , however , we found that this model did not reproduce all of the co absorption and found strong residuals , especially for the low @xmath32 lines .
thus , it seems that at least two layers of co gas were required ; this is consistent with the results from the excitation diagrams .
we thus modified our model to include two layers of gas in a slab geometry absorbing a background dust continuum . in our fit to the co absorption , we used gas temperatures between 50 and 1000 k in increments of 100 k and column densities between 10@xmath33 and 10@xmath34 @xmath13 in increments of log@xmath20 = 0.2 for each of the gas layers .
we varied log[/ ] from 0 to 2 ; log[/ ] and log[/ ] from 0 to 3 in increments of 0.2 .
we also varied the radial velocity ( @xmath35 ) between -30 and -40 km s@xmath9 in increments of 1 km s@xmath9 and smoothed our line profiles to yield lines with widths between 2 and 20 km s@xmath9 in increments of 2 km s@xmath9 to represent the observed line broadening ( due to e.g. the rotation of material in a disk ) .
we compared each model to the entire 2.3 @xmath1 m spectral region and calculated @xmath36 , the reduced @xmath37 statistic for each model . in table
[ table : nirfits ] , we present the parameters from our best fit model and in fig . [
fig : overtone_fits ] , we compare this model to the spectrum . we note that this model reproduces the co absorption very well . in addition , the resulting best - fit temperatures ( 50 k and 900 k ) agree well with those we determined from our excitation diagram ( 40 k and 970 k ) and also the column densities are comparable ( log@xmath38 versus log@xmath39 for the cold layer and log@xmath40 versus log@xmath41 in the hot ) ; this corresponds to a maximum optical depth of @xmath42 .
we find a @xmath36 of 3.54 for our best model however , indicating significant residuals .
l c c & model & excitation diagram + temperature ( k ) & 900 @xmath23 50 & 970 @xmath23 40 + [ 1.1ex ] log @xmath20 ( co ) & 19.2 @xmath23 0.1 & 19.08 @xmath23 0.01 + [ 1.1ex ] log @xmath20 ( h@xmath0o ) & 19.4 @xmath23 0.1 & @xmath43 + [ 1.1ex ] temperature ( k ) & 50 @xmath23 25 & 40 @xmath23 10 + [ 1.1ex ] log @xmath20 ( co ) & 18.2 @xmath23 0.1 & 18.7@xmath44 + [ 1.1ex ] @xmath29c@xmath45c & 1.6@xmath46 & 2.1 @xmath23 0.9 + [ 1.1ex ] @xmath33o@xmath47o & 16@xmath48 & @xmath43 + [ 1.1ex ] @xmath33o@xmath49o & 16@xmath50 & @xmath43 + [ 1.1ex ] & & velocity analysis + @xmath35 ( km s@xmath9 ) & -33.0 @xmath23 0.5 & -32.7 @xmath23 0.2 + @xmath51 ( km s@xmath9 ) & 10 @xmath23 1 & 11.0 @xmath23 0.5 + [ table : nirfits ] note that our model does not only reproduce the ground state @xmath17 transitions , but also the @xmath26 lines ( see e.g. the r18 , r24 and r28 lines in fig .
[ fig : overtone_fits ] ) , again suggesting that the co gas is near vibrational lte .
note however that the critical density for the vibrational levels of co is @xmath52 @xmath53 @xcite , which is much higher than any realistic estimate of the density in this environment .
as described in paper i , the disk is extremely gas rich and significant radiative trapping is occurring .
the gas may thus be radiatively thermalized .
the residual spectrum still exhibits several additional absorption features ; some of these are telluric ( see fig . [
fig : overtone_fits ] ) , however , others are consistent with h@xmath0oat the radial velocity of hr 4049 .
these features are much shallower than the co lines which makes them much less obvious .
we thus recalculated our model for this region with the addition of h@xmath0ousing the line list from ( * ? ? ? * including the h@xmath54o and h@xmath55o isotopologues ) .
we began with the assumption that h@xmath0owas in the same layers as the co ( keeping the same layers reduced the number of free parameters ) and varied the column density in each layer between 10@xmath33 and 10@xmath34 @xmath56 .
the best fit for these models indicated an absence of h@xmath0oin the cold molecular layer .
given the temperature of this cold layer , water may not exist in the gas phase here .
thus , we removed h@xmath0ofrom our cold co layer and only included it in the hot layer .
like the hot co , we find a high column density for h@xmath0oin this region of the spectrum ( log @xmath20 = 19.4 ) , but still corresponding to optically thin lines ( @xmath57 ) .
armed with a good characterization of the absorbing gas in the 2.3 @xmath1mregion , we now turn our attention to the lines near 4.6 @xmath1 m .
we show the spectrum of hr 4049 in this range in fig .
[ fig : co_fundamental ] .
in contrast to the pure absorption lines in the co overtone region , the spectrum here is dominated by emission bands ( of both co and h@xmath0o ) , many of which show a superposed absorption component as well .
note that some of these absorption features are not just self - absorption of the emission lines , but actually absorb continuum radiation as well .
such a spectrum , dominated by emission , is somewhat surprising at first sight given our results from above .
the overtone lines at 2.3 @xmath1mrevealed the presence of a hot gas in front of the continuum emission from the star and dust disk .
all other things being equal , the same gas should also produce absorption in the fundamental lines since they originate from the same ( ground ) state . in fact , since the fundamental lines are intrinsically @xmath6100 times stronger than the overtone lines and since the @xmath61200 k black body dust emission is almost twice as strong at 4.6 @xmath1mthan it is at 2.3 @xmath1 m , we would expect much stronger , saturated absorption lines , as we show in fig .
[ fig : co_fundamental ] . using the same parameters as for the 2.3 @xmath1mregion
, we find that in this range , both the co and h@xmath0olines would be very optically thick ( @xmath58 for co and @xmath59 for h@xmath0o ) and thus appear as broad , saturated lines .
this is in stark contrast with the observations .
thus , the emission must originate from gas that is not residing in the same line of sight as the gas detected in the overtone and must cover a much larger area to completely fill in the expected absorption .
@xcite interpreted this in the framework of the wall model and suggested that the gas in absorption was along the inner wall of the disk , while the gas in emission originated from above the cold dust surface
. however , this is not a viable explanation anymore given our current understanding of the disk from interferometric and mid - ir spectroscopic observations : the disk is radially extended , and is warm over a large radial distance ( * ? ? ?
* paper i ) . furthermore , the gas emitting in the co fundamental must be quite warm as well , as evidenced by the high @xmath32 lines and the presence of @xmath60 and @xmath61 transitions .
one way to explain these observations is by considering that the line of sight that we probe at 2.3 @xmath1mis such that the co is not absorbing the dust emission , but rather the stellar continuum . given our line depths at 2.3 @xmath1mand that about 25% of the flux at those wavelengths originate from the star , this could certainly not be ruled out .
the total flux we observe is then the sum of this absorbing line of sight and the total emission ( gas + dust ) of the disk . at the gas temperatures we determined ,
the gas in the disk would not emit much at 2.3 @xmath1 m , thus the main effect would be veiling of the absorption lines by the dust . at 4.6 @xmath1mhowever ,
the starlight contributes only 5% to the continuum , so the depths of the absorption lines would be much more reduced due to veiling . at the same time , the @xmath61000 k gas would also emit very efficiently at 4.6 @xmath1mand any remaining absorption would be more easily filled in by emission from hot gas .
however , this scenario does not work .
if the co gas in the overtone lines would be absorbing only the stellar radiation , our line of sight through the disk would only probe material at a zero radial velocity since any gas motions would be perpendicular to the line of sight . a detailed analysis of the overtone line profiles
( see sect .
[ sec : discussion_kinematics ] ) shows that the higher @xmath32 lines have much broader line profiles and thus at least a fraction of the absorption originates from gas with a significant radial velocity ; thus , this gas must be absorbing non - stellar radiation .
therefore , we must conclude that the gas in the overtone region is absorbing dust emission as well . to explain the emission in the fundamental region
, there must then be a significant amount of hot gas located in lines of sight that do not intersect the dust emission . at this point , there only seem two plausible locations for that gas . first , the gas could originate from above ( and/or below ) the disk mid - plane , while most of the dust providing the continuum emission would be located near the mid - plane . as noted by @xcite ,
the dust settling time for a gas - rich disk is short typically 150 years for the case of hr 4049 ; thus , the dust distribution is most likely concentrated near the mid - plane . at the same time , if the disk is in hydrostatic equilibrium , the high gas temperatures imply that the disk must still have a large scale height .
it is thus possible that the dust emission is predominantly originating from the mid - plane while most of the gas emission is coming from above the mid - plane .
however , there must still be enough ( small ) dust grains mixed in with the gas at appreciable vertical distances from the mid - plane to explain the phase - dependent extinction @xcite .
furthermore , if the gas is being heated by the dust , the dust must be able to intercept a sufficient amount of stellar radiation , which would not be possible if all the dust is located at the mid - plane .
a second possibility is that there is hot gas located inside the dust disk , in the dust - free central cavity .
since the inner boundary of the disk is determined by the sublimation temperature of the dust and since the dissociation temperature for co ( and h@xmath0o ) is much larger than the dust sublimation temperature , some gas must indeed be located inside the disk .
some support for an origin inside the disk can be found in the larger widths of the emission components compared to the absorption ( by hot gas ) in the overtone ( see sect . [
sec : discussion_kinematics ] ) , pointing to larger velocities for the emitting component than for the absorbing gas .
finally , we should also briefly discuss the absorption components in the emission lines .
while most emission lines show some evidence for an absorption component , it is most pronounced in those lines involving the ground vibrational state .
furthermore , the low @xmath32 lines show the strongest absorption , including absorption of the continuum radiation .
thus , this absorption component must originate from a much cooler component ( such as the cold component we observe in absorption at 2.3 @xmath1 m ) and can not be due to self - absorption by the gas ( which would not result in continuum absorption ) . since some of the absorption lines are slightly blue shifted compared to the emission component , this cold gas may be in an outflow , possibly related to the bipolar lobes .
we also investigate the kinematic properties of the gas . as noted in sect .
[ sec : analysis2 ] , the lines at 2.3 @xmath1mappear narrower than the velocity width reported by @xcite .
when we fitted all the overtone lines from the ground state with gaussians , we determined that the average full - width at half maximum velocity was @xmath62 km s@xmath9rather than their @xmath63 km s@xmath9 .
as an additional check , we also incorporated the line width as a parameter in our full model fit which confirmed this result .
note that this also agrees with @xcite , who determined that the low @xmath32 lines were unresolved when observed with an instrumental resolution of 10 km s@xmath9 .
km s@xmath9 * ? ? ?
the lower plot compares the @xmath64 as a function of lower energy .
the blue dot - dashed horizontal line indicates the weighted mean of all the @xmath51 measurements and the blue dotted lines indicate the uncertainties on this measurement .
the red dashed line indicates a linear fit between these parameters , with the dotted red lines showing the uncertainties in the fit .
error bars on the velocity measurements are determined with a monte carlo technique.,scaledwidth=50.0% ] to further examine the kinematics of the gas in the disk , we fit unconstrained gaussian profiles to only the isolated co overtone transitions from the ground vibrational state ( i.e. those that are not blended with lines from hot bands , other isotopologues or h@xmath0olines ) and examined the resulting radial velocities and line widths . we find that on average , the lines have radial velocities of -32.7 @xmath23 0.2 km s@xmath9and a range between -35.5 and -31.1 km s@xmath9 . we also find an average line width of 11.0 @xmath23 0.5 km s@xmath9and a range between 8.5 and 13.4 km s@xmath9 .
interestingly , both the radial velocity and width measurements show some clear variations as a function of the lower state energy of the lines involved ( see fig .
[ fig : overtone_v ] ) .
[ [ radial - velocities ] ] radial velocities + + + + + + + + + + + + + + + + + in fig . [
fig : overtone_v ] , we recover a trend previously noted by ( * ? ? ? * their fig . 10 , but there for the absorption component in the _ fundamental _ lines ) : the low @xmath32 lines are blue shifted compared to the higher @xmath32 lines which are closer to the system velocity .
this was attributed to a slow outflow of gas from the system . to compare both absorption characteristics
, we show the profiles of the overtone and fundamental lines that exhibit absorption for lines originating from the same level in fig .
[ fig : fundamental_v_overtone ] .
again , we selected only the lines which had relatively clean absorption profiles with as little blending as possible in both the fundamental and the overtone .
it is clear that the centers of the absorption lines do not correspond well to each other ; the @xmath65 lines are the only ones with a clear overlap . in addition , the variations in the fundamental lines are different from the overtone : for the overtone , the low @xmath32 lines are at the system velocity and the largest deviations from the system velocity are found for the r5 and r10 lines ; for the fundamental , it is only the lowest @xmath32 lines that are blue shifted .
these observations are hard to explain consistently .
for instance , the absorption in the fundamental could on first sight be explained by a cold outflow from the system .
indeed , with the deepest absorption ( the r2 line ) originating from the @xmath66 level , this blue shifted gas must have @xmath67 .
however , at such temperatures , there would still be a significant population in the @xmath65 state and we would expect to also see the same blue shifted absorption quite strongly in the r3 line but this is not the case .
that points to a low - density gas that is not even in rotational lte and may be more characteristic of interstellar gas than circumstellar gas . in the overtone lines ,
the case for a systematic effect is even less clear ( see fig . [
fig : overtone_v ] ) , with some low @xmath32 lines at the system velocity and others blue shifted .
it is thus not clear whether this represents a true effect ( e.g. we underestimate blending by other lines ) or rather points to instrumental or calibration issues . for completeness
, we note that these shifts are only seen in the co fundamental of the main isotopologue ; there are no discernible trends in the radial velocities of the absorption components of the hot bands and other isotopologues of co. [ [ line - widths ] ] line widths + + + + + + + + + + + an interesting trend becomes obvious when looking at the widths of the lines as a function of the lower state energy ( bottom panel of fig .
[ fig : overtone_v ] ) : overtone absorption lines originating from higher energies are broader than the lower energy lines .
a linear fit yields a slope of @xmath68 km s@xmath9 per cm@xmath9 .
since only hot gas can contribute to the lines at higher energies and since cold gas dominates the lower energy lines , we conclude that the hotter gas has a larger rotational velocity than the cooler gas , consistent with keplerian motion in a disk that is hotter on the inside and cooler on the outside .
note that thermal broadening is much smaller than the effect we measure here ( at @xmath70 k , @xmath71 km s@xmath9 ) .
we will discuss this effect further below .
we also note that in fig .
[ fig : fundamental_v_overtone ] , the emission component is clearly broader than the overtone absorption .
if the broadening mechanism is rotation in a keplerian disk and the overtone absorption begins at the interior of the dust disk , this once more suggests that the emission component originates on the interior of the dust disk .
it is clear from our analysis above that the co ( and h@xmath0o ) gas probed in the near - ir traces a very complex , gas - rich environment . as pointed out by @xcite , hydrostatic gas pressure in a gas - rich disk
will not only puff up the disk in the vertical direction , but it will also act in radial direction and tend to spread out the disk . indeed , also the near - ir observations point to such a radially extended disk , in agreement with our conclusions from the mid - ir observations in paper i. as we suggest in sect .
[ sec : discussion_kinematics ] , the gas appears to have some temperature variation in which the gas closest to the binary system is hotter than the gas further away .
if the gas and dust are mixed in , then this should also be true of the dust .
if the dust were gray and in equilibrium with the stellar radiation field , the temperature in the disk would vary from 520 k to 370 k at 15 and 30 au respectively ( using a stellar temperature of 7500 k and a stellar radius of 31 r@xmath4 as described by * ? ? ? * ) or 640 and 450 at 10 and 20 au .
the dust must be hotter than this to allow the @xmath72 k gas component to appear in absorption ; indeed , it must be hotter than the gas for the gas to appear in absorption . in paper
i , we suggest that the optically thick gas in the disk traps and re - radiates the stellar radiation , thus the temperatures of the gas and dust in the disk will increase significantly compared to equilibrium values and they will become more homogeneous .
the question thus remains : what type of temperature distribution can the disk have while still reproducing the sed we observe in hr 4049 ? as described by @xcite , the sed can be reproduced by the sum of equally weighted black bodies with temperatures between 880 and 1325 k. as well , when they assigned weights for the black bodies based on a power law distribution , they found that the temperatures could vary between 730 and 1238 k. in general , the maximum dust temperature possible is @xmath61300 k. if the inner rim of the dust component of the disk is determined by the sublimation temperature of the dust , this puts constraints on the type of dust in the disk . for instance , this excludes dust species with high dust condensation temperatures such as e.g. metallic iron or alumina . instead , we should expect the dust to be composed primarily of dust with a sublimation temperature close to 1300 k. we can use our measurements to estimate the ( radial ) size of the gas disk .
if we assume that the overtone absorption originates from gas in keplerian rotation in the disk , the gas velocity at a distance @xmath73 from the center of mass is @xmath74 where @xmath75 is the gravitational constant , @xmath76 is the total mass of the binary system ( the stars are much closer to each other than the disk is to the stars so they can be treated as a point mass ) .
since we find @xmath77 of @xmath78 km s@xmath9and @xmath79 of @xmath80 km s@xmath9 , we conclude that @xmath81 , compatible with the results by @xcite , who determined that the outer radius was @xmath82 based on their model fits to the interferometric observations and mid - ir sed .
the value for @xmath83 is hard to determine , but is most likely between 10 au @xcite and 15 au @xcite ; this would then correspond to outer radii of @xmath84 and @xmath85 au and disk surface areas of @xmath86 and @xmath87 au@xmath88 respectively .
the lower values compare especially well to what we determined for the co@xmath0gas in the mid - ir where we found an emitting area of 1300 au@xmath88 ( assuming an inclination angle of 60@xmath89 , paper i ) .
thus , the circumbinary disk of hr 4049 is a radially extended , gas - rich disk .
however , it is not a flat disk .
indeed , the high temperatures of the gas ensure that hydrostatic pressure will still hold up the disk to appreciable vertical scale heights as well , but less than the @xmath90 that was determined by @xcite . from @xmath77 of the disk
we are able to estimate a mass of the binary system . using an inclination angle of 60@xmath91 , we determine a deprojected @xmath77 of 15 @xmath23 4 km s@xmath9at the inner radius ; since this is the fwhm velocity , we find a tangential velocity of @xmath92 km s@xmath9for the gas in keplerian rotation at the inner radius . adopting an inner radius of 15 @xmath93 au , we find a total mass of @xmath94 m@xmath4 for the binary .
if we then apply the mass function for the primary determined by ( * ? ? ?
* @xmath95 = 0.158 @xmath23 0.004 ) , we calculate a mass of @xmath96 m@xmath4 for the primary and @xmath97 m@xmath4 for the secondary . if we use an inner radius of 10 au , we determine smaller masses for the binary and individual stars ( @xmath98 ) .
the mass we find for the primary at an inner radius of 15 au is close to the @xmath99 m@xmath4 estimated by @xcite .
it is thus interesting to contemplate the implications of such a low mass for the evolution of the primary .
when a star is on the red giant branch ( rgb ) , it burns hydrogen in a shell and adds helium to the core . for a low - mass star , with a degenerate core
, the temperature for helium ignition is attained when the core mass reaches 0.45 m@xmath4 .
unless part of the core was removed during the common envelope phase , this suggests that either hr 4049 never ignited helium in its core or that it had a non - degenerate core on the rgb , which can ignite helium at lower masses .
however , this latter scenario would require a much higher initial mass for hr 4049 , and would suggest that it has lost a great deal of material which is currently missing from the system .
when low - mass stars terminate their evolution without igniting helium in their cores , they will end their lives as helium white dwarfs .
however , hr 4049 is still a giant star .
this suggests that something else may be occurring in this system .
theoretical predictions show that hydrogen shell flashes occur on helium white dwarfs which have sufficient hydrogen atmospheres remaining ; these flashes are also thought to return the white dwarf to giant sizes for very short periods of time ( @xmath610@xmath100 yr , * ? ? ?
the short duration of these flashes is inconsistent with the observational history of hr 4049 .
perhaps it has lost most of its envelope , but has kept enough for it to sustain a longer - term hydrogen burning shell . to our knowledge , no evolutionary models of such an object have been attempted .
based on our new analysis and re - interpretation of near - ir high - resolution spectoscopic data of hr 4049 , we find that the circumbinary disk surrounding the system is hot and radially extended , consistent with the results from the mid - ir observations described in paper iand by the interferometry described by @xcite .
we find evidence that the fundamental co emission originates from within the central , dust - free cavity of the disk . in the absence of dust ( and thus ignoring radiation pressure ) , viscous dissipation will cause this gas to end up accreting onto one of the stars in the binary system . the gas in the circumbinary disk must play a significant role in determining the physical properties and geometry of the disk .
future radiative transfer modeling should elucidate these effects . the mass of the system and the individual components are both very low and consistent with mass loss .
in particular , the total mass of the system would correspond to a main sequence lifetime on the order of 10 gyr ; such an age is incompatible with the abundances of the system . therefore , the system must have lost a considerable amount of mass during its evolution .
we acknowledge the support from the natural sciences and engineering research council of canada ( nserc ) .
we would like to thank ken hinkle and sean brittain for their help in providing us with the gemini - phoenix data .
this research has also made use of nasa s astrophysics data system bibliographic services and the simbad database , operated at cds , strasbourg , france .
, joyce r.r .
, et al . , 2003 , in : guhathakurta p. ( ed .
) , _ society of photo - optical instrumentation engineers ( spie ) conference series _ , vol .
4834 of _ society of photo - optical instrumentation engineers ( spie ) conference series _ , pp . | hr 4049 is a peculiar evolved binary which is surrounded by a circumbinary disk .
mid - infrared observations show that the disk is rich in molecular gas and radially extended . to study the properties of this disk , we re - analyzed a set of near - infrared observations at high spectral resolution obtained with gemini - phoenix .
these data cover absorption lines originating from the first overtone of co and from h@xmath0oin the 2.3 @xmath1 m region as well as more complex emission - absorption profiles from h@xmath0oand the fundamental mode of co near 4.6 @xmath1 m . by using an excitation diagram and from modeling the spectrum
, we find that most of the co overtone and h@xmath0oabsorption originates from hot gas ( @xmath2 k ) with high column densities , consistent with the mid - infrared data .
the strong emission in the wavelength range of the co fundamental furthermore suggests that there is a significant quantity of gas in the inner cavity of the disk .
in addition , there is a much colder component in the line of sight to the disk .
a detailed analysis of the overtone line profiles reveals variations in the line widths which are consistent with a radially extended disk in keplerian rotation with hotter gas closer to the central star .
we estimate the mass of the primary to be @xmath3 m@xmath4 and discuss the implications for its evolutionary status . |
geometrical frustration plays an important role in determining ground states and phase transitions in magnetic systems .
a triangular lattice in two - dimension in particular is one of the simplest systems to study the effect of geometrical frustration .
lufe@xmath0o@xmath1 is a member of rfe@xmath0o@xmath1 family of compounds , where r can be y , ho , er , tm , yb , and lu . @xcite these materials all have hexagonal layered structure , in which fe ions form a triangular lattice within each bilayer .
@xcite since the average charge valence of fe in this compound is + 2.5 , this system is expected to exhibit charge order behavior similar to fe@xmath3o@xmath1 @xcite or half doped manganites .
@xcite however , due to the geometrical frustration introduced by the triangular lattice , understanding charge order in this material is not straightforward .
@xcite previous electron and x - ray diffraction studies have shown that charge ordering sets in below @xmath2 300k , and anomalous dielectric dispersion was observed in this temperature range .
@xcite in particular , ikeda and coworkers argued that the observed pyroelectric signal below charge ordering temperature indicates charge order driven ferroelectricity .
@xcite this result has been drawing much attention,@xcite since this would be the first such observation of ferroelectricity with electronic origin .
in addition , it was observed that the pyroelectric signal shows an unusual step around the spin ordering temperature , and a large magnetodielectric response under low magnetic fields was also observed in lufe@xmath0o@xmath1 at room temperature,@xcite which prompted further interests in this compound as a possible multiferroic ( or magnetic ferroelectric ) material .
although whether the magnetic and ferroelectric order parameters are coupled in lufe@xmath0o@xmath1 is not clear at the moment , lufe@xmath0o@xmath1 exhibits quite interesting magnetic properties , as a result of the geometrical frustration arising from the triangular lattice .
most of earlier studies of the magnetism in rfe@xmath0o@xmath1 have been focused on yfe@xmath0o@xmath1 .
tanaka et al . first reported that fe spins order below 220 k based on their mssbauer experiments .
@xcite in their studies of transport properties , they also observed that there are two distinct transitions at 240 k and 225 k , and the former corresponds to verwey - like charge ordering accompanied by magnetic ordering .
@xcite this was corroborated in the x - ray study of nakagawa and coworkers , in which first order structural phase transitions were observed around these temperatures .
@xcite recently , ikeda et al . reported that more than two transitions exist in yfe@xmath0o@xmath1 based on their x - ray powder diffraction studies .
@xcite they also argued that the transition at 250 k corresponds to charge and spin ordering .
however , it was also realized that the oxygen non - stoichiometry in yfe@xmath0o@xmath1 can cause significant changes in its magnetic properties , while lufe@xmath0o@xmath1 is believed to be free from such oxygen non - stoichiometry problems .
@xcite in their comprehensive magnetization and neutron scattering work on lufe@xmath0o@xmath1 , iida and coworkers were able to elucidate unusual magnetic properties of this compound .
@xcite specifically , they found that the system does not show any long range three - dimensional magnetic order down to 4.2 k. instead , they argued that the system at low temperatures consists of ferrimagnetic clusters of various sizes , based on their thermoremanent magnetization measurements .
the ferrimagnetism in this case arises due to the mixture of s=2 and s=5/2 spins . in recent neutron scattering experiments ,
however , sharp magnetic bragg peaks were observed , suggesting existence of long - range magnetic order.@xcite therefore , the nature of the ground state of lfo is still not understood well . in this paper , we report our comprehensive study of magnetic properties of lufe@xmath0o@xmath1 using ac susceptibility , dc magnetization and specific heat .
we have observed two magnetic transitions : the high temperature transition occurs at @xmath2236k , and corresponds to the previously observed ferrimagnetic transition.@xcite the signature of this transition is also observed in our specific heat measurements .
in addition to this ferrimagnetic transition , we observe an unusual magnetic transition at a lower temperature , which shows relaxational behavior similar to that of a spin - glass phase .
this paper is organized as follows . in the next section
, we will explain our sample preparation and characterization in detail . in sec .
iii , our experimental results from magnetic susceptibility and specific heat measurements are presented . in sec .
iv , we will discuss the implication of the observed results , and possible connection with the charge order and ferroelectricity .
lufe@xmath0o@xmath1 ( lfo ) single crystals were grown using the travelling solvent floating zone method at brookhaven national laboratory following the method reported in ref .
our experiments were done using the crystals from the same batch without any special annealing procedure .
the chemical composition of one of the crystals was examined with electron probe microanalysis ( epma)with beam size less than 1 micron .
the lu / fe ratio was analyzed at 25 randomly selected points on the sample surface .
the average lu / fe ratio was [email protected] , and the mean deviation from the average value was less than 1@xmath5 .
this result shows that the lu and fe is homogeneously distributed with almost stoichiometric ratio .
the oxygen contents of two other pieces were studied using x - ray photoemission spectroscopy ( xps ) , revealing that the oxygen content in one sample was higher than the other sample , suggesting that there is a small but finite oxygen non - stoichiometry issue in lufe@xmath0o@xmath1 .
it turns out that the magnetic and structural properties of lufe@xmath0o@xmath6 depends very sensitively on the oxygen stoichiometry
. detailed study of phase diagram is still in progress , but we made sure that all the samples studied in this work show the same magnetic properties .
this ensures that the variation of @xmath7 among the samples studied here is very small .
the largest piece with a shape of a rectangular parallelepiped ( @xmath8 mm ) was used in our magnetization studies .
dc magnetization and ac susceptibility measurements were done using quantum design mpms squid magnetometer .
specific heat measurements on the same sample were carried out using thermal relaxation method on quantum design ppms .
in fig . [ fig : fig.1](a ) , we show the temperature dependence of the thermo - magnetization of lufe@xmath0o@xmath1 obtained with 10 oe field applied parallel to the crystallographic c axis which is perpendicular to the hexagonal planes .
a sharp peak appears in the magnetization curve at a temperature of @xmath2 236k , below which the field - cooled ( fc ) data begin to diverge from the zero - field - cooled ( zfc ) data . in fig .
[ fig : fig.1 ] ( b ) , thermo - magnetization obtained in a field perpendicular to the c axis is shown .
note that the magnetization in this direction is 2 orders of magnitude smaller than that shown in panel ( a ) .
this small magnetization can be entirely accounted for by the possible sample misalignment with respect to the field direction .
this also illustrates that the easy axis is along the c - axis , and the ising anisotropy is very large .
the non - zero zfc magnetization at low temperature in this case is probably due to the small residual field in the magnetometer .
figure [ fig : fig.2 ] shows the temperature dependence of the specific heat , @xmath9 , of the same sample used in the magnetization study .
one can identify two features in this curve .
the high temperature feature above 300k is relatively broad and has maximum at @xmath2 330k .
this feature is probably related to the 3d charge order observed in previous electron and x - ray diffraction studies .
@xcite the low temperature peak emerges below @xmath2250k and has a cusp at @xmath2237k .
the peak position of this low temperature feature is very close to the peak in magnetic susceptibility , suggesting that this feature is related to the magnetic phase transition .
= 1 oe was applied and the magnetization was measured .
the real and imaginary part of the susceptibility are shown in part ( a ) and ( b ) , respectively.,width=297 ] figure [ fig : fig.3 ] shows the real and imaginary part of the ac susceptibility as a function of temperature .
the different curves correspond to different driving frequencies of the ac field .
the amplitude of ac field was kept constant at @xmath10=1 oe .
a well - defined peak is observed for the real part of the susceptibility @xmath11 at 236 k and the low - temperature tail of this peak decreases with increasing frequency
. the imaginary part of the susceptibility , @xmath12 , appears below @xmath2 240 k , and consists of two peaks .
the high temperature component , appearing as a shoulder , is located at @xmath2 237 k and grows as frequency increases , while the peak position remains the same . on the other hand
, the low temperature peak grows and shifts to higher temperature with increasing frequency .
such a behavior is commonly observed in spin glass systems . for a spin glass system , with decreasing temperature
the spin dynamics become sluggish , so that it takes longer time for a spin to relax , and the maximum relaxation time increases accordingly .
when an external ac magnetic field with a driving frequency @xmath13 is applied to a spin glass system , if the maximum relaxation time @xmath14@xmath15 is longer than @xmath16 , the system will not be able to keep up with the oscillating field and become out of equilibrium .
therefore , one can define the freezing temperature , @xmath17 , as the temperature , at which @xmath18 . as a result , @xmath17 is a function of driving frequency @xmath19 .
experimentally @xmath20 ) can be determined either from the maximum of @xmath21 ) or from the inflection point of @xmath22.@xcite since the maximum of @xmath11 is difficult to identify due to the second peak located at 236k , we use the inflection point of @xmath23 to determine @xmath17 . the maximum relaxation time and @xmath24
can be modeled with conventional critical slowing down @xcite @xmath25 where @xmath26 is the spin - glass transition temperature , @xmath27 is the dynamical exponent , @xmath28 is the usual critical exponent for the correlation length and @xmath14@xmath29 is the microscopic flipping time of the fluctuating spins .
the scaling of the ac susceptibility is shown in fig .
[ fig : fig.4 ] , and the best fit to eq . ( [ equ : n ] ) yields @xmath30 k , @xmath31 and @xmath32 s. the value of @xmath33 is very close to the microscopic spin flip time @xmath34 seconds in other spin glass systems .
@xcite the value of @xmath35 is within the range of well - known spin - glasses such as cumn(4.6 at.% ) ( @xmath35=5.5 ) @xcite and cdcr@xmath0(in)s@xmath1 ( @xmath36 ) .
@xcite this value of @xmath35 is also close to the value obtained from numerical simulations in three - dimensional ( 3d ) ising spin glasses .
@xcite this scaling analysis indicates that the low temperature phase is quite possibly a spin - glass phase .
thus , taken together with the heat capacity and the susceptibility data , lfo seems to undergo a continuous phase transition from a curie paramagnetic phase to ferrimagnetically ordered phase at @xmath2 236k and then to a reentrant spin - glass phase below 228k .
figure [ fig : fig.5 ] shows the magnetization ( m ) versus field ( h ) curves at different temperatures after cooling under a 200 oe magnetic field . at low temperatures ,
since the zero - field magnetization is non - zero , there is thermal remanent magnetization ( trm ) , which almost vanishes above @xmath2232k . at high temperatures
, there is no trm and the slope of the m vs h curve is largely determined by linear susceptibility , which shows maximum at 236 k ( see fig .
[ fig : fig.1 ] ) . at low temperatures ,
the magnetization initially shows negative slope due to the relaxational behavior of spin - glass phase .
nonlinear susceptibility is a valuable tool to study the critical behavior of a magnetic phase transition . in particular ,
the spin - glass susceptibility is believed to be directly proportional to the non - linear susceptibility.@xcite if a system undergoes a sg phase transition , the magnetization m can be expanded by odd order of h. @xmath37 the data in fig .
[ fig : fig.5 ] ( above 50 oe ) were fitted using eq .
( [ equ : m ] ) , and the fitting results are shown as solid lines .
the constant offset @xmath38 , linear susceptibility @xmath39 and the third order nonlinear susceptibility @xmath40 obtained from the fits are shown in fig .
[ fig : fig.6 ] .
@xmath38 starts to grow below @xmath2235k and then increases rapidly below @xmath2230k .
the temperature at which @xmath38 increases rapidly is close to @xmath26 , suggesting that @xmath38 behaves like the spin - glass order parameter .
when the temperature dependence of @xmath38 below 230k is fitted using @xmath41 , best fits are obtained with @xmath42 k and @xmath43 , as shown by the solid line in fig .
[ fig : fig.6](a ) .
this critical exponent @xmath44 is consistent with that of canonical spin - glass systems.@xcite the linear susceptibility peak in @xmath39 at @xmath2236k ( fig .
[ fig : fig.6](b ) ) was also observed in dc magnetization and heat capacity , and this indicates the existence of a magnetic phase transition at @xmath45k .
if we fit the linear susceptibility for @xmath46 to @xmath47 , @xmath48 is obtained , and the fitting result is plotted as the solid line in fig .
[ fig : fig.6](b ) .
as temperature is lowered , the third order nonlinear susceptibility @xmath40 shows a negative minimum at 236k , and then it abruptly changes sign at 234k and then show a broad positive peak at @xmath2230k .
this behavior is reminiscent of the recent observation that @xmath40 near a ferromagnetic transition diverges negatively and positively as @xmath49 is approached from above and below , respectively.@xcite therefore our observation of the zero - crossing of @xmath40 around 234k seems to suggest that this high temperature magnetic transition is described as ferrimagnetic ( fm ) transition .
this assignment is also consistent with earlier studies.@xcite the behavior of @xmath40 around 230k , however , is quite unusual for a spin - glass system .
it was observed that the non - linear susceptibility ( @xmath40 ) is negative with a cusp at the paramagnet to spin glass transition in ref .
25 and 30 .
we observe broad positive peak around @xmath26 , which may be due to the reentrant nature of the spin - glass phase in lfo .
in fact , similarly unusual behavior have been observed in other reentrant spin - glass systems.@xcite our nonlinear susceptibility data also supports the phase diagram suggested by ac susceptibility data .
although spins are considered `` frozen '' in the spin glass phase , due to the slow dynamics , the system simply does not reach the equilibrium state within the experimental time scale . as a result ,
the spin glass system exhibits non - equilibrium phenomena .
one such example is aging .
when a spin - glass system is cooled below @xmath26 , the spin - glass domain grows . since this domain growth occurs logarithmically in time , it is customary to define the relaxation rate @xmath50.@xcite in fig .
[ fig : fig.7 ] , we show our data for @xmath51 as a function of @xmath52 .
note that the sample was cooled down to 0.87@xmath26 @xmath2 200 k in the absence of magnetic field . after waiting for a certain time
( @xmath53=1000s , 5000s and 10000s ) without external field , the magnetization was recorded as a function of time after a 10 oe magnetic field was applied .
as can been seen from the figure , @xmath54 at which the maximum relaxation rate occurs increases with increasing @xmath53 , and in fact it is almost equal to @xmath53 .
this kind of aging behavior illustrates non - equilibrium dynamics of domain growth , and has been observed in other spin glass systems.@xcite another interesting example of non - equilibrium dynamics of spin - glass system is the so - called memory effect . in order to show this effect , we have measured temperature dependence of @xmath55 in two distinct routes .
the first , @xmath56 , was obtained by cooling in 10 oe magnetic field from 300 k down to 50 k at a constant cooling rate of 2 k / min and then heating back continuously at the same rate . in the second route ,
m was recorded on cooling in 10 oe at the same rate from 300 k to 50 k with two halts at @xmath57 k for 72000 s and at @xmath58 k for 48000 s. during the halts , the external field is turned off to let the magnetization relax . after each halt
, m shows a clear deviation from the reference as illustrated in fig .
[ fig : fig.8 ] , due to aging . after reaching 50 k ,
the sample temperature is increased continuously at 2 k / min rate in @xmath59 oe . during the reheating ,
the system exhibits a step - like feature at both @xmath60 and @xmath61 .
the jump at @xmath60 is not very pronounced , but clear jump in @xmath55 around @xmath61 is clearly visible .
this suggests that the system somehow remembers the history of halts during cooling .
exceeding the halt points , @xmath62 recovers to the reference value and the system is called rejuvenated .
such aging , memory and rejuvenation behavior was observed in other spin glass systems as well.@xcite these observations also suggest that the low temperature phase of lfo is consistently described as a spin - glass . in fig .
[ fig : fig.9 ] , ac susceptibility at 10 hz driving frequency is plotted as a function of temperature for different external static magnetic fields h. as can be seen in the figure , @xmath11 is suppressed by the magnetic field . as the field increases , the main peak of @xmath11 decreases and a double - peak feature emerges .
the ferrimagnetic phase transition temperature determined from specific heat and nonlinear susceptibility ( @xmath2236 k ) is quite close to the position of the high temperature peak in @xmath11 , which slightly increases with increasing field .
the low temperature peaks in @xmath11 and @xmath12 correspond to the spin - glass transition . as the field increases ,
the low temperature peak in @xmath11 decreases and finally disappears under @xmath2 1 t ( as shown in the inset ) .
the peak in @xmath12 also shifts to lower temperature with increasing field . under very high external field
( above 1 t ) , the spin - glass transition seems to be completely suppressed .
hz and @xmath10=1 oe .
each curve was obtained under different applied static magnetic field of @xmath63 .
the inset shows the data obtained with @xmath64 t.,width=278 ] figure [ fig : fig.10 ] shows the field versus temperature phase diagram . due to the large uncertainties associated with determining transition temperature of two nearby phase transitions , the error bars at low field is relatively large .
however , the field dependence of the pm - fm transition is very weak in this region , while it is clear that the transition temperature gradually increases with increasing field at high field .
the spin - glass transition temperature also exhibits substantial field dependence .
the sg - fm transition temperature is suppressed rapidly as a small field is applied .
in addition , a threshold field @xmath65 oe is observed , below which @xmath26 does not show a systematic change with the change of the field . according to the mean field theory
, there exists a phase boundary in @xmath66 phase diagram called de almeida - thouless ( at ) line,@xcite whereby a spin - glass phase can only exist under this boundary ( in the low field region ) .
the at line is given by @xcite @xmath67 here @xmath63 is the external magnetic field and @xmath68 is the field - dependent glass transition temperature.@xcite our data fits this relation very well as shown in fig . [
fig : fig.10 ] , in which a linear relationship between @xmath26 and @xmath69 is clearly illustrated .
we have presented a variety of experimental evidences showing that the lfo sample goes through re - entrant spin glass transition around 228 k. however , this observation is quite puzzling in several aspects .
the first is the microscopic origin of the spin - glass behavior .
conventionally , disordered spin arrangements or interactions ( random site or random bond ) are necessary to produce magnetic frustration required for spin - glass behavior .
however , this system , lfo , is considered highly stoichiometric , and in addition , there exists charge ordering below 300 k , which implies that the arrangement of fe@xmath70 and fe@xmath71 spins are regular .
therefore , this system seems to possess only geometrical frustration as a necessary ingredient for spin glass behavior . unless spin - glass behavior can arise from pure geometrical frustration
, one must find the missing `` disorder '' in this system to explain the observed spin - glass behavior . before discussing the disorder effect in this system , it is useful to examine magnetic interactions present . in lfo , both the fe@xmath70 and fe@xmath71 ions are in their high spin configuration , with the spin angular momentum s=2 and s=5/2 , respectively .
the exchange interactions between the fe@xmath70-fe@xmath70 and fe@xmath71-fe@xmath71 are presumably antiferromagnetic through the superexchange path via the oxygen ions .
however , the magnetic interaction between the fe@xmath70 and fe@xmath71 ions requires further consideration .
note that the fe@xmath70 is in @xmath72 configuration , and the hund s rule dictates that the extra electron in this ion , compared to the fe@xmath71 ( @xmath73 ) ion should point in the opposite direction of the rest of the `` @xmath73 '' electrons .
therefore , one can expect the interaction between the fe@xmath70 and fe@xmath71 to be ferromagnetic based on a kinetic energy argument , analogous to the double exchange mechanism in manganites.@xcite however , since this compound is insulating at all temperatures , such `` extra '' electron can not be mobile , but presumably resides in a resonance state between the two neighboring fe ions .
one of the important degrees of freedom in this system , often overlooked , is the orbital degrees of freedom . since the fe ions all have trigonal bipyramidal crystal field environment , the 3d orbitals will split into three levels .
the highest energy level will be @xmath74 , and there will be two doubly degenerate orbitals ( @xmath75 and @xmath76 ; @xmath77 and @xmath78 ) at lower energies . following the hunds rule , one can easily see that fe@xmath71 ion has @xmath73 configuration and is isotropic .
however , fe@xmath70 ion has @xmath72 configuration and the lowest energy orbitals will have orbital degeneracy . in order to break this orbital degeneracy , there could be a cooperative jahn - teller distortion of fe@xmath70 ions , once the charge order sets in .
taken together , it is conceivable that two neighboring fe@xmath70 and fe@xmath71 ions form a `` dimer '' , sharing a minority spin electron .
in fact , such bond dimerization scenario has been considered in their study of mixed valence b - site fe ions in fe@xmath3o@xmath1 by seo and coworkers.@xcite such bond dimerization will break the orbital degeneracy , and make the dimer spins form a highly frustrated kagome lattice , which may be responsible for the observed spin glass behavior .
further structural investigation of this system is required to address this conjecture .
another important issue is the oxygen non - stoichiometry .
if there exists oxygen non - stoichiometry , this will necessarily affect the ratio of fe@xmath70 and fe@xmath71 spins , leading to disorder in charge order , and possibly disorder in the exchange interactions .
in fact , our preliminary studies suggest that the oxygen contents seem to change magnetic properties quite dramatically , possibly explaining different magnetic behavior of @xmath79 compounds reported in the literature .
another possibility is the charge - ordering model itself .
if the charge ordering at 300 k is not of second - order kind , and has some relaxational component , such as charge glass type ordering , then this will naturally affect the spin ordering part .
for example , one can imagine charge - ordered domains exhibiting relaxor type behavior and influence dielectric and magnetic behavior at lower temperatures .
our preliminary x - ray scattering experiments also suggest that the charge ordering in this system is quite complicated and has nontrivial temperature dependence .
another surprising aspect of the spin - glass transition is its large temperature scale .
almost all known spin - glass systems occur at low temperatures .
this is due to the fact that the lower critical dimension of the spin glass transition is believed to be between 2 and 3 .
since a 3d sg system is very close to the lower critical dimension , its ordering temperature is very close to absolute zero . in that sense
this observation of spin glass behavior at temperatures above 200 k is not only unusual , but very surprising for a quasi 2d system .
again , if the glassy nature of the system arises from the charge sector , this may provide a natural explanation .
the magnetic properties of @xmath79 single crystals were investigated with dc magnetization and ac susceptibility . based on the dynamic scaling of ac susceptibility and the behavior of non - linear susceptibility , it is suggested that @xmath79 goes through first a ferrimagnetic ordering at 236 k , and then subsequently goes through a reentrant spin - glass transition at @xmath80 k. typical properties of spin glass system , such as aging , memory , and rejuvenation have been also observed in this low temperature phase .
the field dependence of the spin glass transition temperature is described well by the well - known de almeida - thouless theory .
it was also observed that the ferrimagnetic transition temperature shows quite sizable field dependence . in order to understand the origin of the spin glass behavior in this compound , possibilities based on the frustration , oxygen non - stoichiometry , and glassy charge ordering
have been discussed .
we would like to thank david ellis , s. m. shapiro , g. xu , j. brittain , a. gershon and h. zhang for invaluable discussions .
the work at university of toronto was supported by natural sciences and engineering research council of canada , canadian foundation for innovation , ontario innovation trust , and early researcher award by ontario ministry of research and innovation .
the work at brookhaven was supported by the u. s. department of energy , office of science . | we have carried out a comprehensive investigation of magnetic properties of lufe@xmath0o@xmath1 , using ac susceptibility , dc magnetization and specific heat .
a magnetic phase transition around @xmath2236 k was observed with dc magnetization and specific heat measurements , which is identified as a paramagnetic to ferrimagnetic transition based on the nonlinear susceptibility data . upon further cooling below this temperature
, we also observed highly relaxational magnetic behavior : the dc magnetization exhibits history and time dependence , and the real and imaginary part of the ac susceptibility shows large frequency dependence .
dynamic scaling of the ac susceptibility data suggests that this low temperature phase can be described as a reentrant spin glass phase .
we also discuss magnetic field dependence of the spin glass transition and aging , memory and rejuvenation effect below the glass transition temperature around 228 k. |
central supermassive black holes are a common feature to galaxies today , but which came first , the black hole or the galaxy ?
conventional thinking would suggest that the the first generation of stars evolved into black holes , which have subsequently settled to the centers of their host galaxies , merged , and accreted gas .
but this idea , in which central black holes form inside pre - existing galaxies , has recently earned some scrutiny .
first , the discovery of increasingly high redshift quasars requires a surprisingly early formation of the black holes ( see , for example , @xcite and @xcite ) .
second , a large quasar sample shows no evidence of black holes growing in mass with decreasing redshift @xcite .
so we are left to consider the possibility that either the central black holes formed before their host galaxies , or they grew to maturity very quickly within them . either way , they have grown little since the quasar epoch .
the most distant known quasar lies at @xmath0 , with a central black hole of mass @xmath1 @xcite . in the @xmath2cdm cosmology observed by wmap @xcite , with @xmath3 , @xmath4 , and @xmath5
, this redshift corresponds to a time when the universe was only @xmath6 old . for the present work , we will take this as the time to beat : @xmath7 billion solar masses in @xmath8 million years .
in the past year , two separate hst studies have cited evidence for intermediate mass black holes ( imbhs ) in the centers of globular clusters : a @xmath9 hole in m15 @xcite , and a @xmath10 hole in andromeda s g1 cluster @xcite .
this is the lastest and strongest evidence for imbhs , but there is additional evidence , and good theoretical motivation as well ; see @xcite for a comprehensive review .
imbhs are widely believed to be a necessary growth stage for smbhs . in section [ flowchart ] of this paper , we will review the major proposed routes to the formation of a smbh , all of which include an imbh phase , from which time the question is simply one of growth .
we start in section [ flowchart ] with a flowchart of avenues leading to the creation of a smbh . in section [ timescales ]
, we examine the timescales for each needed process .
we conclude in section [ conclusions ] by discussing how realistic each avenue is in light of these timescales .
there are essentially four proposed families of models leading to the formation of imbhs , and two or three ways to grow them .
these approaches are depicted in figure [ flow ] and discussed in turn below .
the black holes may be _ primordial _ , in which case they formed from primordial density variations before big bang nucleosynthesis .
primordial black holes ( pbhs ) have been studied extensively , and the most direct mechanism for their creation is the collapse of gaussian random density fluctuations @xcite .
these holes come from horizon scale ( or smaller ) modes , and therefore their masses are determined by their time of formation . in the radiation dominated early universe , @xmath11 but in order to preserve the successful bbn prediction of light element abundances , there must be no significant rate of pbh formation once nucleosynthesis begins , and therefore the pbhs are capped at intermediate mass .
in addition , @xcite have pointed out that , given a small scalar spectral index @xmath12 @xmath13 was recently observed in the cmb @xcite @xmath12 pbhs from density inhomogeneities should only have formed in quantities too small to be of interest .
a more promising , and perhaps inevitable mechanism for forming pbhs also exists , in which the collapse is triggered by `` color holes '' at the quark - hadron phase transition @xcite .
however , because this occurred at @xmath14 , these pbhs would be smaller than @xmath15 by eq .
[ pbh_time ] , and would remain as collisionless dark matter today , rather than collecting into larger black holes .
( interestingly , @xcite shows evidence for such pbhs in the microlensing of distant quasars , in numbers comparable to that needed to account for dark matter . ) \2 .
normal _ population iii stars _ formed at @xmath16 , evolved to black holes , and merged at the center of their small dark matter halos .
this is perhaps the least exotic way to create imbhs , and at this point there is very strong support for the process in the form of numerical simulations of structure formation @xcite .
these simulations include the relevant atomic and molecular processes in the first gas clouds , particularly cooling by rotation of molecular hydrogen , superimposed on cdm halo evolution .
they find that @xmath17 stars of @xmath16 form in halos of @xmath18 , engaging @xmath19 of the system s baryonic matter .
because the cooling of population iii stars is hindered by the lack of metals , these first stars would be more massive than those allowed by fragmented star formation today .
@xcite have shown that these massive stars will evolve into black holes containing a significant fraction of the star s initial mass ( this fraction varies with the stellar mass , and is of order @xmath20 ) , unless the stars are in the mass range @xmath21 , in which case they are completely disrupted when they go supernova .
given the small initial size of such a cosmologically young halo ( @xmath22 ) , the holes sink to the center and presumably merge into a single imbh there .
_ supermassive stars _ may have been the first baryonic objects to form .
if so , they would have evolved rapidly into imbhs .
supermassive stars ( smss ) , as a class of objects , span from @xmath23 to @xmath24 , although the first generation of them would reside at the lower end of that range ( @xcite provides a comprehensive introduction on the topic ) .
a @xmath25 sms has a lifetime of @xmath26 years ( sms lifetimes range down to 10 years ) , at the end of which , it undergoes relativistic collapse to a black hole in a matter of seconds @xcite .
this collapse sends @xmath27 of the star s original mass into the black hole , with the remaining @xmath28 forming an accretion disk @xcite .
the radius of a @xmath25 sms , @xmath29 , is around @xmath28 of the virial radius for the earliest halos containing gas cool enough to collapse .
it is therefore reasonable to suspect that these stars should appear in early halos , especially since population iii stars produce ultraviolet radation which splits hydrogen molecules , thus destroying the only coolant which could collapse smaller stars .
the collapsing gas might produce enough radiation pressure to prevent fragmentation @xcite .
another possibility is that the first stars formed dense clusters which merged into one or more smss .
however , whether or not supermassive stars ended the cosmological dark ages remains a matter of speculation .
if _ dense clusters _ of stars emerged from early star forming regions , then the cluster stars may have merged with each other to create massive stars ( say , several hundred solar masses ) , which then evolved to massive seed black holes .
this idea has been mapped out recently by @xcite , and we will not repeat it here . however , for the sake of timescales , which are the focus of this paper , this scenario is nearly the same as the population iii star scenario discussed above , because the merging occurs during the lifetime of the cluster s massive stars .
each of the paths in figure [ flow ] sees a phase with an imbh imbedded in a halo , at which point accretion and halo mergers take over to achieve heavier black holes .
the merger scenario merges central black holes hierarchically along with the host halos , growing both together .
it is clear that merging imbhs and smbhs are real , interesting phenomena deserving study .
for example , @xcite have simulated the whole process , and shown that coalescing smbhs remove the steep cusps that cdm simulations generate , thus matching observed galaxy rotation curves .
but it remained to be shown here whether or not mergers are prevalent enough to sufficiently increase black hole mass .
gas accretion is another major way to grow imbhs into smbhs .
the belief that quasars are powered by accretion at the eddington luminosity gives this idea credibility .
but over what fraction of cosmic history did these holes accrete ?
some authors suppose that eddington limited accretion is continuous since the black hole first formed @xcite , and thus its mass grows rapidly .
others have proposed scenarios in which accretion is negligible over the history of the bh @xcite , such that @xmath30 holes that formed in the first episode of star formation still populate our galactic bulge today .
another possibility is that the bh feeds on dark matter , in which case the halo and the imbh evolve together .
@xcite proposed a model in which the halo and hole grow self - similarly , with both analytical and numerical results @xcite supporting their idea .
this model is particularly interesting because it also simultaneously explains both the spectral index for the galaxy power spectrum , @xmath31 , and the @xmath32 relationship observed to be universal in galaxy centers @xcite .
all of these methods certainly contribute to the growth of massive black holes . in the next section
we will use timescales to help constrain which methods could realistically dominate the process of forming smbhs in less than a gigayear .
first , we examine the evolution to the imbh phase , as depicted in figure [ flow ] . if the bh is primordial , then it can be up to @xmath33 in the very early universe .
its growth by merger or accretion then awaits the formation of matter clumps to surround the pbh .
these could begin to arrive as early as photon decoupling , so timescales alone will not strongly constrain this possibility unless the pbh in question started at @xmath34 , as it would if it formed during the quark - hadron phase transition . in this case the bh seed has a very long way to grow , which is barely possible by accretion , as will be shown in subsection [ accretion ] .
the simulations of population iii star formation of @xcite produce some very useful information .
the earliest simulated star forming region occurs at @xmath35 , with stars forming at @xmath36 per comoving @xmath37 , increasing to @xmath38 at @xmath39 , and holding constant thereafter @xcite . molecular hydrogen coolant forms in these clouds in @xmath40 .
by @xmath41 , the gas pressure has diminished and gas clumps begin to contract .
halos are @xmath42 to @xmath43 by this time , and the largest virialized star forming clouds are @xmath33 .
the gas takes an approximately isothermal density profile , with constant density inside a core radius of @xmath44 .
the virial temperature is a few thousand kelvins .
a simple application of the virial theorem gives a virial radius of @xmath45 , and the corresponding circular orbital speed is @xmath46 .
this , then , is the starting point for population iii stars leading to imbhs , with star formation beginning in earnest at @xmath41 . to convert redshifts to times in a @xmath2cdm cosmology with @xmath47 @xcite , @xmath48 and with wmap values inserted ,
this is @xmath49 so gas collapses into the first stars at @xmath50 .
these stars collapse in the free fall time , @xmath51 which is about @xmath52 at the densities observed in simulations by @xcite .
they evolve to black holes ( or disruptive supernovae ) in a few million years , leaving @xmath53 remaining to grow into a quasar power source . using the dynamical friction formula from @xcite , quoted here from @xcite
, we can find the time needed for these population iii remnant bhs to sink to the center of their host halo .
treating the bhs as point particles , @xmath54 where @xmath55 is called the coulomb logarithm , which characterizes the distance range perpendicular to the path of motion over which the frictional encounters occur ( for a recent study of this parameter , see @xcite ) . using the values mentioned above for the parameters in this equation , and choosing @xmath56 ( this choice is common and does nt affect the outcome much because of the logarithm ) , the dynamical friction timescale relevant here is at least @xmath57 , bringing the total down to @xmath58 available for bh growth .
the coalescence time of these holes is not well known , and could be long . if the star formation yields one or more supermassive stars , rather than multiple normal population iii stars , then the evolution to a bh is faster than for normal stars @xcite . the dynamical friction time to merge smaller stellar bhs
is also avoided with smss , so in the sms case , we have @xmath53 remaining to grow a smbh . in the case of an early - forming dense star cluster
in which runaway collisions merge the stars , the timing remains essentially the same .
one waits until the cluster stars form , merge , and evolve to one or more bhs .
the time available to become supermassive is still at most @xmath59 .
we consider the case of a satellite subhalo halo falling into a larger halo .
each halo is modeled as a singular isothermal sphere , so that @xmath60 and orbital velocities within a halo are constant .
we seek a formulation that tracks the friction until the satellite s central bh reaches the new center .
smaller halos formed earlier in the cdm hierarchy , when the universe was denser , and therefore they are more tightly bound .
so as a subhalo sinks , we assume it holds itself together , until it gets close to the center of the larger halo . there , which halo an orbiting particle belongs to becomes ambiguous .
this occurs roughly when a particle s distance from its original host subhalo is on order of the separation between the two halo centers .
this separation distance is @xmath61 , a radial coordinate extending from the center of the larger halo .
the internal speed within an isothermal halo is @xmath62 the square root in this equation is helpful because it reduces the dependence that velocities have on a halo size @xmath63 , which is generally not well known .
the mass which can be associated with the sinking satellite is then @xmath64 where @xmath65 is the mass of the satellite halo residing inside a distance equal to @xmath61 around it , @xmath66 is the satellite subhalo s central bh mass , and @xmath67 is the orbital velocity inside the satellite . choosing to use @xmath61 in this way assumes that the isothermal satellite halo extends to distances larger than some initial infall radius : @xmath68 .
the angular momentum of the satellite in the larger halo is @xmath69 , where @xmath70 is the orbital velocity within the larger halo .
dynamical friction induces a torque @xmath71 . following the standard derivation presented in @xcite , the statement of @xmath72 leads to the equation @xmath73 we next define 3 constants as follows : @xmath74 , @xmath75 , and @xmath76 . inserting the satellite mass from eq .
[ m_subhalo ] , we have @xmath77 eq . [ diffeq ] is a first - order , linear , ordinary differential equation , but the closed - form solution is prohibitively long and complicated due to the @xmath78 factor in the second term .
a numerical solution readily parameterizes the sinking orbit @xmath79 , and a root can be found for the time @xmath80 at which @xmath81 .
one needs only to specify a starting radius for the satellite halo , @xmath82 .
an example orbit decay curve computed in this way is given in figure [ roft ] .
this dynamical friction model approximately doubles the time spent spiralling inward , relative to the chandrasekhar formula for an infalling point particle , eq .
[ chandra_df ] .
@xcite found the same result using both @xmath83-body simulations and the theory of linear response .
a subsequent paper by @xcite gives an analytical approximation for nfw halos @xcite , also finding that the decay time is increased by a factor of a few if the satellite has less than @xmath84 of the larger halo s mass . in the mass range of interest for the present work ,
all these models essentially agree ; however , since we wish to follow length and mass scales ranging down to a central bh , we have chosen to use the method described above .
we first consider the viability of black hole merging as a means to grow smbhs .
as halos merge over cosmic history , their central bhs sink to the new center by dynamical friction , and subsequently coalesce into a larger bh .
the usual way to track mergers is by @xmath83-body simulation .
we begin with such a simulation , and calculate the total mass of subhalos that can merge into one large halo between the epoch of the first massive bhs and the early quasar epoch .
we subject this calculation to the constraint that the merging subhalos must sink to the center in the time allowed between the two epochs .
this can become a very complicated calculation , because the halo population is continuously changing over a long period of cosmic time . to make the problem tractable
, we make a series of major approximations . we describe them along the way , and discuss them afterward .
we demonstrate first that with our assumptions , mergers are _ not _ sufficient to grow smbhs , regardless of how the bhs originally formed .
second , we show that a more accurate calculation to replace our simplifying assumptions would only strengthen that conclusion .
the argument begins with the numerical simulations of @xcite .
they follow a @xmath85 halo today back in time to all its subhalos , presenting halo mass spectra at multiple redshifts . for our purposes ,
the results they present at @xmath86 and @xmath87 are most important , approximating the time of the first quasars ( @xmath0 ) and the first imbhs in halos ( after @xmath41 ) .
we ll address the impact of these approximations in redshift and total mass after the calculation .
the halo functions are nearly power law in the mass range of interest . for @xmath88 in @xmath89 units , at @xmath86 @xmath90 and at @xmath87 @xmath91 consider the largest halo in existence as the period of interest begins .
imagine ( conservatively ) that all other subhalos are available to merge into this one , and ( also conservatively ) that all of these subhalos contain central bhs of @xmath92 of the parent galaxy s luminous mass , as observed in large galaxies today @xcite . according to wmap , the ratio of baryonic matter to total matter is @xmath93 @xcite ,
so conservatively we may say that a central bh has @xmath94 of the host subhalo s mass .
then every subhalo massive enough to dynamically sink to the center in the available time does sink , leading to a more massive combined halo .
if all the imbhs from all the combined subhalos coalesce quickly ( again , highly conservative ) , then the new smbh mass is @xmath94 of the total mass of all subhalos with mass @xmath95 , the minimum satellite mass needed for dynamical friction to sink a subhalo by the early quasar era .
however , during the merging process , the largest halo is gaining mass and the number of massive halos is increasing . to evade this problem
, we suppose that the halo mass spectrum at the _ end _ of the merging process ( at @xmath86 ) applies throughout the merging process .
our calculations show that in most cases , this assumtion conservatively exaggerates the final smbh mass by more than an order of magnitude . using eq .
[ dndm5 ] , the total mass of the quasar s smbh is @xmath96 the lower limit , @xmath97 , is determined by dynamical friction as described in section [ df ] .
we follow the infalling halo mass , and transition to the ( smaller ) infalling bh mass as the two holes close in , achieving results closely consistent with other models that include mass loss but not a central bh @xcite . in the dynamical friction calculation
, three parameters must be chosen .
we take the internal velocity of the larger halo to be @xmath98 , noting that for a fixed dynamical sinking time ( fixed by the time available for bh growth ) , the final smbh mass obeys @xmath99 .
so choosing a larger velocity would only very slightly reduce the final smbh mass by reducing the number of subhalos which sink in the allowed time .
the other needed quantities are radii : @xmath100 , the initial radius from which the satellite halo begins its descent , and @xmath101 , the size of the satellite subhalo .
the former is roughly the impact parameter of the colliding subhalos , and the latter indicates the subhalo s concentration factor .
both radii will clearly vary from one encounter to the next , so we choose several representative values .
the results demonstrate that any reasonable ( in fact , conservative ) choice of these numbers leads to the same conclusion , that merging black holes take too long . as shown earlier , the timescale between the first imbh and a quasar s smbh depends on how the bh formed .
a primordial bh has the age of the universe at the time of the most distant quasar s emission known today , or @xmath102 .
population iii stars and supermassive stars have @xmath59 .
if the population iii stars merged within their original halo first , then they have @xmath103 to grow .
we fix the sinking time at the appropriate value and compute the total bh mass that can be assembled by halo mergers , neglecting the time needed for coalescence .
our results are listed in table [ tab_merger ] , which lists a maximum smbh mass for various choices of @xmath101 and @xmath100 .
no choice brings @xmath66 up near @xmath104 , the value needed to explain the most distant known quasars . .smbh
masses achieved through halo mergers by the early quasar epoch .
@xmath101 is a characteristic size for the satellite subhalos , and @xmath100 is the radius from which they are assumed to begin their infall .
@xmath105 is the time allowed for growth .
[ cols="^,^,^,^",options="header " , ] one is led to conclude that mergers alone are not sufficient to grow the largest smbhs in the oldest quasars .
note that although the argument above relies on many simplifying assumptions , all but one tend to overestimate the resulting smbh mass , so the smbh masses in table [ tab_merger ] should be considered extreme upper limits .
our assumptions are summarized as follows .
all halos contain central bhs at @xmath94 of the halo mass ( this may seriously overestimate the number of bhs ) .
all relevant merging is completed by @xmath86 , instead of today , so the number of infalling bhs is again exaggerated .
\3 . in the @xmath2cdm world
, going from redshift @xmath106 to @xmath107 allows over @xmath108 for merging , so the number of more massive subhalos available to merge is artificially large .
this overestimates the final smbh mass , because larger infalling subhalos will dynamically sink faster .
infall radii begin at @xmath109 , which for some mergers represents a head start of multiple orders of magnitude .
the choice of an isothermal halo model probably exaggerates the central density , artificially speeding up the action of dynamical friction near the center .
smbh coalescence is assumed to occur arbitrarily quickly , but in reality takes at least @xmath110 years , and probably much longer @xcite .
the choice to trace a @xmath85 halo today back in time tends to _ _ under__estimate the number of progenitor subhalos . extrapolating this mass linearly from @xcite , we find that increasing it by @xmath111 times would increase @xmath112 at @xmath86 by a factor of @xmath17 while approximately maintaining its slope . in this case , the largest halo at @xmath86 would be @xmath113 , so halo speeds should be comparable to those in m87 today , the giant elliptical at the center of the virgo galaxy cluster . radial velocities for globular clusters around m87 @xcite range from @xmath114 to @xmath115 .
these changes , using @xmath116 , allow a smbh to become almost @xmath117 in @xmath59 from an initial radius @xmath118 , which is a factor of @xmath119 greater than our original result .
so if we attempt to correct for this seventh assumption , we approach ( but do not reach ) the @xmath104 hole that quasar observations require .
but this is only true if we take literally the six highly conservative assumptions above .
consider first gas accretion , followed by feeding on dark matter .
_ gas accretion _ is probably the simplest and most successful way to explain the growth of an imbh into a smbh , or even a large stellar mass bh into a smbh .
@xcite show that if the black hole accretes at the eddington limit continuously , from birth through detection as a quasar , it achieves the needed growth .
they give the eddingtion luminosity , @xmath120 and introduce a radiative efficiency @xmath121 and a fraction of eddington output @xmath122 .
the resulting @xmath123-folding time is @xmath124 which equals a @xmath125-folding timescale @xmath126 for a fiducial value of @xmath127 and continuous eddington - limited accretion ( @xmath128 ) , one can calculate the initial bh mass needed to grow to @xmath104 in the alloted time . if the seed bh is primordial , it might accrete steadily from approximately the beginning of time . in this case
, it has @xmath102 to grow ; that s @xmath129 orders of magnitude in mass by eq .
[ 10-fold ] .
so the initial seed mass must have been @xmath130 , which , interestingly , could have been formed during the cosmological quark - hadron phase transition @xcite . if the seed bh is a normal population iii stellar remnant , or a supermassive star s remnant , then it had @xmath59 .
it s initial mass was then @xmath131 , which is completely plausible for the first generation of stars .
one star of @xmath132 would suffice @xcite . in @xmath103 , the seed
bh mass needs to be @xmath133 .
this applies for the case where population iii stars must sink to the center of their parent halos and merge before beginning to grow significantly by gas accretion . in the case where the seed bh formed from the evolution of normal population iii stars , the mass of the seed star is suggestively close to what one would expect .
thus this route is possible , but only if eddington limited accretion is maintained during the entire imbh growth process .
if the seed hole evolved from a more massive sms , then there is time to spare , and the accretion rate could have dropped for some of the growth period .
the success of this quick exponential growth calculation seems to indicate that baryon accretion is the most realistic mechanism for imbh growth . when the bh is small , the eddington mass accretion rate is not prohibitively large , and by @xmath0 we have very strong evidence that the quasar is accreting at its eddington rate @xcite .
so it is reasonable to speculate that it has been accreting at this rate all along .
however it is worth noting that the central smbhs in modern galaxies appear to grow by a factor of 10 on timescales of @xmath134 @xcite .
this is very slow compared to eddington limited growth , requiring @xmath135 .
this dramatic drop in accretion rate needs explaining , especially if we are to claim that no similar drop has ever happened before @xmath136 in all of cosmic history , as the notion of smbh growth by gas accretion onto population iii stellar corpses would mandate .
this problem is only alleviated if the first stars were supermassive , or the seed bhs were primordial and significantly larger than @xmath137 . _ accreting dark matter _ is another possibility .
@xcite have proposed an interesting way to account for imbh growth in which the hole grows proportionally to the dark matter halo as matter falls in . as mentioned earlier in section [ flowchart ] , this assumption leads to the observed galaxy power spectrum and the observed @xmath138 relation . in numerical simulations @xcite ,
dark matter was found to form a self - similar region surrounding the bh , which does grow in proportion to the bh growth at the center .
the formation of this region proceeds by gravitational interactions between clumps of dark matter ( `` particles '' in the simulations ) which must have formed earlier in the cdm universe .
the bh grows whenever the particles cross inside the schwarzschild radius .
( the bh growth rate is thus understated in the sense that @xmath139 was used instead of the larger capture radius . ) in the absence of angular momentum , the black hole grows rapidly : @xmath140 .
but for realistic departures from pure spherically symmetric infall , in which the dark matter has angular momentum to help it resist falling into the hole , the growth of the bh is only linear in time : @xmath141 ( although the @xmath138 relation is recovered either way ) .
the simulations were not performed in standard units , and therefore can not be immediately applied to relevant halo sizes and cosmic timescales .
here we simply note that the simulated bh growth is slow ( e.g. , linear in time , as compared with exponential in time for gas accretion ) .
one therefore needs an unusually large imbh to start from , which requires an unusually large sms as its origin .
in this paper we have surveyed the state of knowledge of smbh formation and growth in light of a difficult new data point to satisfy : a quasar at @xmath0 whose smbh was @xmath104 when the universe was only @xmath142 old .
we have extended the calculations of others on smbh growth via mergers and accretion . in the case of smbh growth by mergers ,
we follow the output from a cdm simulation in order to track progenitor subhalos .
we stipulate that a subhalo must sink to the center by dynamical friction after a merger within the time available for smbh growth .
this limits the mass of subhalos and therefore bhs which could have coalesced in time , thus constraining the final smbh mass .
we find that in any realistic case , mergers are incapable of growing a bh with the needed speed .
a primordial black hole that formed during the quark - hadron phase transition has accreted at the eddington rate continuously ever since baryons have been able to cluster until at least the quasar epoch .
each of these options involves seed stars at the upper end of their predicted mass range .
so although the eddington limited gas accretion scenario is adequate to grow the holes in the time available , it is not entirely satisfying , because it requires heavy gas infall which is never significantly diminished before the quasar era , despite the relative gas suffocation which has evidently happened since .
one can avoid imposing this requirement by starting with a larger bh : \4 . a single sms , or the merged remains of a dense cluster of stars , or a pbh formed well after the quark - hadron phase transition , yielded an imbh much larger than @xmath145 .
it accreted through the quasar epoch , but did _ not _ need to maintain the maximum accretion rate during that time .
one other option , wherein the bh grows in proportion to its host halo by eating dark matter , is expected to yield slower bh growth . as such ,
normal population iii stellar seeds would be inadequate , although a particularly massive sms seed remains a possibility .
we conclude by suggesting that _ some _ unconventional mechanism is needed to realistically beat the clock for the early formation of smbhs .
bh growth through mergers is too slow , and growth by accreting nonstop at the eddington limit is probably too contrived .
the remedy appears to require one of the following : ( a ) pbhs from unusually large overdensity pockets , ( b ) population iii starbursts well in excess of those predicted by current simulations , or ( c ) supermassive stars . | the high redshifts of the most distant known quasars , and the best estimates of their black hole masses , require that supermassive black holes ( smbhs ) must have formed very early in history .
several mechanisms for creating and growing these holes have been proposed . here
we present an evaluation of the timescales needed for various critical processes in order to discriminate between the proposed scenarios .
we find in particular that mergers alone are not able to grow the black holes at a sufficient rate .
accretion models offer a solution and we use accretion timescales to constrain the manner in which the black hole was first formed .
this analysis implies , but does not require , the action of some unconventional process . |
during the last decade considerable effort has been devoted to simulational and experimental verification of the mode coupling theory ( mct ) of glassy dynamics and the glass transition @xcite .
the consensus that emerged from this work is that mct describes in a satisfactory way `` weakly '' supercooled liquids ( _ i.e. _ it describes the first few decades of slowing down on approaching the glass transition ) . in particular , mct has been quite successful when applied to concentrated colloidal suspensions @xcite , the colloidal glass @xcite , and gelation @xcite transitions .
notably , less effort has been devoted to the foundations of the mode coupling theory ( see , however , refs .
this is somewhat surprising in view of mct s several well - known problems .
the most important , fundamental problem is the uncontrolled nature of the basic mct approximation : factorization of a complicated _ time - dependent _ pair - density ( _ i.e. _ four - particle ) correlation function .
recently , we proposed an extension of mct for dynamics of colloidal suspensions and the colloidal glass transition @xcite .
our theory includes , in an approximate way , time - dependent pair - density fluctuations .
it relies upon a factorization approximation that is similar to that used in mct , but is applied at a level of a memory function for the time - dependent pair - density correlation function .
the theory predicts an ergodicity breaking transition similar to that of mct , but at a higher density .
thus it partially solves another well - known mct problem : overestimation of so - called dynamic feedback effect and the resulting underestimation of the colloidal glass transition density . here ,
for a simpler , toy model , we go further : we completely avoid using factorization approximation for any _ dynamic _ correlation function .
we only assume gaussian factorization of _ static _ correlations @xcite . it should be noted that a frequently used approach to glassy dynamics is to start from a set of fluctuating hydrodynamics equations which are supplemented by a quadratic free energy implying gaussian static density fluctuations @xcite .
we argue that the analysis presented here has implications for such models .
since the approach is technically quite involved , we state the main results immediately : we derive an essentially exact expression for the time - integrated memory function for a single brownian particle moving among immobile obstacles .
we compare this expression with one derived from mct and show that the latter includes a subset of the former s terms : only explicitly positive terms from the exact series ( _ i.e. _ the terms that always increase the effective friction felt by the mobile particle ) are included within mct .
this is the origin of mct s overestimation of the dynamic feedback effect .
an approximate re - summation of the exact series suggests that , under the assumption of gaussian static fluctuations , the mobile particle s motion is always diffusive .
in contrast , mct predicts that the mobile particle becomes localized at high enough obstacle density .
this result has important consequences for models used to study glassy dynamics .
we show here that , if static correlations are gaussian , a single mobile particle is never localized by immobile obstacles .
this suggests that a similar _ fully mobile _ system ( _ i.e. _ gaussian static correlations and all particles diffusing ) can not undergo an ergodicity breaking transition . in other words
, the ergodicity breaking transition predicted for such a system by a mode coupling theory is , most probably , an artifact of the factorization approximation . note that this does _ not _ mean that mct is qualitatively wrong for a system with complicated many - particle static correlations ( like , _
e.g. _ , the hard sphere system ) .
it can be argued that terms that cut - off mct s localization transition ( _ i.e. _ terms that are neglected in mct ) are canceled by other terms that originate from non - gaussian static correlations .
indeed , empirical success of mct for colloidal systems suggests that this might be the case .
it is at present unclear how to describe this remarkable cancellation .
we consider one spherical brownian particle diffusing between @xmath0 immobile , spherically symmetric obstacles .
the particle interacts with the obstacles via a potential @xmath1 .
the obstacles are mechanically identical to the mobile particle .
we assume that the initial joint probability distribution for the mobile particle and the obstacles is given by the equilibrium canonical distribution at temperature @xmath2 .
the time evolution of the system is described by a generalized smoluchowski equation : @xmath3 with the initial condition @xmath4 here @xmath5 denotes the position of the mobile particle and @xmath6 denote positions of the obstacles .
furthermore , @xmath7 is the diffusion coefficient of the mobile particle in the absence of the obstacles , and @xmath8 is the force acting on it . finally , the second line in eq .
( [ nsm ] ) defines the @xmath9-particle generalized smoluchowski operator @xmath10 .
our theory starts from the memory function representation of the mobile particle density correlation function , @xmath11 , @xmath12 here @xmath13 is the fourier transform of the mobile particle density , @xmath14 and @xmath15 denotes the average over the equilibrium probability distribution , @xmath16 .
note the equilibrium distribution stands to the right of the quantity being averaged , and all operators act on it as well as on everything else . to derive the memory function representation we start from an expression for the laplace transform , @xmath17 , of the time derivative of @xmath18 : @xmath19 using standard projection operator manipulations
we rewrite ( [ fktder ] ) in the following form : @xmath20 here @xmath21 is a projected current density of the mobile particle , @xmath22 @xmath23 , and @xmath24 is a projection operator on the mobile particle density subspace , @xmath25 note that in eq . ( [ p1 ] ) we introduced a summation convention : we sum over all repeated wavevectors appearing in adjacent _ ket _ , @xmath26 , and _ bra _ , @xmath27 .
next we define the one - particle irreducible evolution operator , @xmath28 and we use the same standard projection operator manipulations to re - write the current - current correlation function appearing in eq .
( [ fktmr ] ) in the following form : @xmath29 combining eqs .
( [ fktmr ] ) and ( [ fktmr2 ] ) we derive the memory function representation for @xmath30 : @xmath31 comparing eq .
( [ fktm ] ) with the standard form of the memory function representation , @xmath32 we identify the irreducible memory function : @xmath33 where @xmath34 .
next we define @xmath35 : the part of the joint density of the mobile particle and the obstacles that is orthogonal to the mobile particle s density , @xmath36 we use the following identity that is exact for two - particle additive interactions : @xmath37 where @xmath38 denotes the kernel of the inverse integral operator . using eq .
( [ vertex1 ] ) and an analogous identity for @xmath39 we can express the irreducible memory function ( [ m1irr ] ) in terms of the pair - density correlation function : @xmath40 in turn , this correlation function can be subjected to the same manipulations as the mobile particle density correlation function @xmath11 .
re - tracing the steps between eqs .
( [ fktdef ] ) and ( [ fktm ] ) we arrive at the following : @xmath41 here @xmath42 is the two - particle projected current density , @xmath43 and @xmath44 is the two - particle irreducible evolution operator , @xmath45 in eqs .
( [ j2][o2 ] ) @xmath46 , with @xmath24 defined by eq .
( [ p1 ] ) , and @xmath47 being projection operator on @xmath35 : @xmath48 combining eqs .
( [ m1irr ] ) , ( [ vertex1 ] ) and ( [ f2ktm2 ] ) we can obtain an expression for the memory function in terms of the autocorrelation function of the two - particle projected current density : @xmath49 note that to get expression ( [ m1irr2 ] ) we resorted to a technical approximation that is similar to the first enskog approximation used to approximately invert the boltzmann collision operator @xcite .
this minor , technical approximation is not required in one dimension where eq .
( [ m1irr2 ] ) is exact .
it is clear this procedure can be continued _
ad infinitum _ : the two - particle projected current correlation function can be expressed in terms of the three - body density correlation function , _ etc . _
( note that for higher order densities , @xmath50 , @xmath51 , it is advantageous to use ordered multiplets of wavevectors , @xmath52 ) .
this is somewhat akin to the well - known continuous fraction expansion .
it is different from it in that at each step a new function depending on a larger number of variables is introduced .
the resulting expressions simplify greatly if we assume gaussian factorization of static correlations .
note that operators @xmath53 remove contributions involving fewer than @xmath54 independent connections between two groups of particles @xcite .
thus , for example , for the correlations of the projected @xmath55-particle density we get @xmath56 furthermore , for the density - current correlations of the type ( [ vertex1 ] ) we get @xmath57 where @xmath58 is the number density , @xmath59 , and @xmath60 is the direct correlation function , @xmath61 . using the above described procedure , under the assumption of static gaussian fluctuations , we get the following expression for the time - integrated ( _ i.e. _ @xmath62 ) memory function at @xmath63 @xmath64 where @xmath65 denotes all permutations of the wavevectors indices . eq .
( [ mseries ] ) is the main result of this note . under the assumption of static gaussian correlations ,
this expression is essentially exact .
the only , technical approximation is the one similar to the first enskog approximation .
expression ( [ mseries ] ) should be compared with that resulting from the mode coupling theory .
the latter approach starts from expression ( [ m1irr ] ) for the memory function , rewrites it in terms of the pair density correlation function using ( [ vertex1 ] ) and ( [ f2kt ] ) , and then resorts to a factorization approximation @xcite .
this procedure gives the following expression for the memory function : @xmath66 using eq .
( [ mmct ] ) we can obtain a series expression for mct s time - integrated memory function .
we notice that @xmath67 and iterate ( [ mmct ] ) to get @xmath68 where we wrote the resulting series in a form similar to ( [ mseries ] ) .
it should be noted that from each term of ( [ mseries ] ) the mct expression ( [ mmctseries ] ) includes only these terms that are explicitly positive definite .
we suggest that the difference between the two expressions is the origin of mct s well - known overestimation of so - called dynamic feedback effect . to investigate this last issue further we attempt to re - sum the expression ( [ mseries ] ) .
to this end we resort to an approximation which simplifies angular dependence of individual terms in series ( [ mseries ] ) .
this approximation will be motivated in detail elsewhere @xcite .
after the approximation we obtain @xmath69 where @xmath70 denotes dimensionality of the space .
one should note that eq .
( [ mseriesapp ] ) reproduces exactly the first two terms in series ( [ mseries ] ) .
also , eq . ( [ mseriesapp ] ) is exact in one dimension . using ( [ mseriesapp ] )
we obtain the following simple formula for the mobile particle s diffusion coefficient : @xmath71 it is clear that , for a generic interaction potential , the mobile particle s motion is always diffusive ; _ i.e. _ the mobile particle is never localized .
one should also note a striking resemblance between ( [ d ] ) and deem and chandler s result @xcite for the diffusion coefficient of a single particle in gaussian random media ( note that @xmath72 is the effective mobile particle - obstacle interaction potential @xcite ) .
comparison with numerical simulations of diffusion in gaussian random media showed that the latter result is extremely accurate .
in contrast , for a generic interaction potential , mct s equation ( [ mmct ] ) ( supplemented by the memory function representation for @xmath30 ) predicts that at sufficiently high obstacle density mobile particle s diffusion coefficient vanishes and the particle gets localized .
it should be emphasized that the absence of the localization transition within the full theory and the resulting stark discrepancy with mct is independent of the approximation leading to eq .
( [ mseriesapp ] ) .
the diffusion coefficient in @xmath70 dimensions is bounded from below by the @xmath73 diffusion coefficient @xcite . in one dimension the series ( [ mseries ] ) can be re - summed without difficulty .
the result is given by formula ( [ d ] ) with @xmath73 .
we showed here that , under the assumption of gaussian factorization of static fluctuations , a single diffusing particle is never mobilized by immobile obstacles . in contrast , mct predicts localization at a high enough obstacle density .
one should notice that in some cases , _
e.g. _ for a hard sphere diffusing among immobile hard spheres , we expect the mobile particle to get localized at sufficiently high obstacle density .
our result suggests that this localization is intimately connected to non - gaussian character of _ static _ density fluctuations .
furthermore , mct implicitly assumes gaussian factorization but it neglects terms that cut off the localization transition .
it would be of great interest to investigate whether in a more accurate theory these terms are canceled by terms originating from non - gaussian character of static density fluctuations .
finally , our results suggest that theories based on gaussian density correlations are not suitable for description of the glass transition .
99 w. kob , _ cond - mat/0212344 _ ( to appear in les houches proceedings ) .
w. gtze , j. phys . : condens .
matter * 11 * , a1 ( 1999 ) .
see also proceedings of international discussion meetings on relaxations in complex systems , k.l ngai ( ed . ) : j. non - cryst .
solids * 131 - 133 * ( 1991 ) , _ ibid . _
* 172 - 174 * ( 1994 ) , _ ibid . _
* 235 - 237 * ( 1998 ) , _ ibid . _
* 307 - 310 * ( 2002 ) . for a recent overview emphasizing connections between diffusional and viscoelastic properties
see g. ngele , j. phys .
matter * 15 * , s407 ( 2003 ) .
cates , _ cond - mat/0211066_. k.a .
dawson , curr . opin .
colloid interface sci .
* 7 * , 218 ( 2002 ) .
liu and i. oppenheim , physica a * 247 * , 183 ( 1997 ) and references therein .
e. zaccarelli , g. foffi , f. sciortino , p. tartaglia , and k.a .
dawson , europhys .
lett . * 55 * , 157 ( 2001 ) . for an elegant diagrammatic analysis that leads to mode coupling - like equations , applied to a lattice model ,
see s.j . pitts and h.c .
andersen , j. chem .
* 114 * , 1101 ( 2001 ) .
g. szamel , phys .
lett . * 90 * , 228301 ( 2003 ) .
note that a fully consistent approach would include _ both _ gaussian factorization of static density fluctuations and a gaussian form of the structure factor , _
i.e _ @xmath74 with @xmath58 being the number density , @xmath75 and @xmath76 being the fourier transform of the potential . in the following we only assume gaussian factorization and
implicitly use the exact structure factor ; this is in the spirit of other approaches @xcite .
for a recent example see k. miyazaki and d.r .
reichman , phys . rev . * 66 * , 050501 ( 2003 ) . a similar approach based on a diffusion equation for a fluctuating density field has been proposed by k. kawasaki ( k. kawasaki , physica a * 208 * , 35 ( 1994 ) ) .
p. rsibois , m. de leener , _ classical kinetic theory of fluids _
( wiley , new york , 1977 ) . for
a discussion in a slightly different language see b. cichocki , physica a * 142 * , 245 ( 1984 ) .
there are several ways to derive mct . here
we outline the original projection operator method ( see , _
e.g. _ , w. gtze , in _ liquids , freezing and glass transition _ , j.p .
hansen , d. levesque , and j. zinn - justin , eds .
( north - holland , amsterdam , 1991 ) ) .
g. szamel ( to be published ) .
m. deem and d. chandler , j. stat . phys . * 76 * , 911 ( 1994 ) . in the fully consistent approach @xcite @xmath77 and deem and chandler s result @xcite is recovered .
de masi , p.a .
ferrari , s. goldstein , and w.d .
wick , j. stat .
phys . * 55 * , 787 ( 1989 ) . | we consider a toy model for glassy dynamics of colloidal suspensions : a single brownian particle diffusing among immobile obstacles .
if gaussian factorization of _ static _ density fluctuations is assumed , this model can be solved without factorization approximation for any _ dynamic _ correlation function .
the solution differs from that obtained from the ideal mode coupling theory ( mct ) .
the latter is equivalent to including only some , positive definite terms in an expression for the memory function .
an approximate re - summation of the complete expression suggests that , under the assumption of gaussian factorization of static fluctuations , mobile particle s motion is always diffusive .
in contrast , mct predicts that the mobile particle becomes localized at a high enough obstacle density .
we discuss the implications of these results for models for glassy dynamics . |
silicon is an indirect gap semiconductor with six minima ( valleys ) in the conduction band . in the case of ( 001)-oriented wafers , the long axes of two of the fermi ellipsoids are perpendicular to the @xmath1 plane of a @xmath2 interface ( with @xmath3 ) , while the long axes of the other four have an in - plane orientation ( @xmath4 ) .
a quantum well is formed at the interface by applying a positive voltage @xmath5 to the gate , and the electron motion in @xmath6 direction is quantized .
the difference in @xmath7 causes a splitting of the energy spectrum of bound states into two independent ladders of levels : the twofold - degenerate ladder of eigenenergies @xmath8 and a fourfold - degenerate ladder @xmath9 . due to the higher effective mass in the @xmath6-direction , the lowest energy level in the potential
well is @xmath10 .
the self - consistent calculations ( see e.g @xcite ) predicts that the excited levels @xmath0 and @xmath11 are very close on energy scale and become occupied ( cross the fermi energy @xmath12 ) at the carrier concentration @xmath13 . while @xmath14 increases linearly with @xmath5 , the fourfold - degenerate @xmath0 stays `` pinned '' to the fermi energy and @xmath15 is almost constant even for the highest possible gate voltages . in both cases
the difference between the fermi energy and the eigenenergies of excited states is at most a few percent of @xmath16 , the difference between the fermi energy and the ground level , for all gate voltages .
in our experiments we have employed the hall bar samples of russian provenance , with 200 nm thick gate oxide and the top mobilities @xmath17 above @xmath18 at liquid helium temperature .
the concentration of electrons in the inversion layer is related to the gate voltage by @xmath19 with the threshold voltage close to @xmath20 .
the samples are 0.25 mm wide and 0.5 mm long with the distance between potential leads 0.625 mm .
the highest concentration of carriers @xmath21 is reached for @xmath22 .
this is well above @xmath23 , the onset voltage of the occupation of excited subbands . according to our numerical modeling of the electronic structure @xcite , @xmath24 while @xmath25 at the highest concentration .
the samples exhibit the previously unreported features , which turned out to be important in experiments with high carrier concentration .
the metallic ( aluminium ) gate of samples overlaps to the thicker oxide outside the lithographically defined hall bar shape .
the thin molybdenium layer separates the aluminum gate and the silicon oxide .
the mo layer can be seen by cross - sectional electron microscopy ( fig .
[ sample ] ) and allows identification of a step in the thickness of oxide layers . therefore , a parallel channel is formed at the sample edges .
this explains two series of magnetoresistance oscillations at high concentration of carriers we reported in our previous publications ( see @xcite and references therein ) .
an attempt to attribute the second series to the sdh oscillations of electrons from the second subband turned out to be wrong , the ratio of periods of the two series of sdh oscillation was 9 : 2 , i.e. it corresponds to the ratio of the two oxide layers . to suppress the effect of parallel edge - channels , approximately @xmath26 of the gate metal was lithographically removed on the sample edges .
after this procedure the second series of sdh oscillations , corresponding to the sample edge channels , disappeared .
the high mobility of samples was not reduced by this procedure .
the magnetoresistance of si(001 ) mosfets with high density of electrons in the inversion layer was measured as a function of the magnetic field @xmath27 up to 11 t for a series of fixed gate voltages in the bath of pumped @xmath28he . the standard ac technique was used with a frequency 13 hz .
the measuring current @xmath29 yields the current density @xmath30 . here
we report results obtained for two samples .
the sample a has a top mobility @xmath31 at 0.45 k. the magnetoresistance and the hall resistance of this sample are presented in figs .
[ resa ] and [ halla ] for a dense set of gate voltages .
the sample b magnetoresistance ( the top mobility @xmath32 is shown in fig .
[ resb ] only for selected gate voltages @xmath5 . for this sample
we also show the current - density and temperature dependence of the magnetoresistance at the maximum gate voltage @xmath33 ( see figs .
[ current ] and [ temperature ] ) .
anomalous behavior was observed for all gate voltages corresponding to occupied excited subbands .
first , a novel series of sdh oscillations appeared , most pronounced at the highest gate voltage 120 v where the amplitude of sdh oscillations of the first subband is very small .
the oscillations are periodic in @xmath34 .
their period is almost independent of @xmath5 and close to @xmath35 for both samples .
second , a strong negative magnetoresistance and the nonlinear field dependence of the hall resistance accompany the novel oscillations at high carrier concentrations .
both the novel oscillations and the negative magnetoresistance are most pronounced at the lowest temperatures ( @xmath36 0.4 k ) and the smallest density of current ( @xmath37a ) used in our measurements .
the influence of the current density on the observed effect is illustrated in fig .
[ current ] .
the current was first increased by an order of magnitude and then by two orders of magnitude .
both amplitudes of the novel oscillations and the negative magnetoresistance are gradually suppressed , while the amplitude of sdh oscillations from the main series remain unchanged .
an independent measurement , in which the current density was kept low and the bath temperature was increased to 4.2 k , confirmed that the suppression of amplitudes of the novel oscillations and of the negative magnetoresistance is really due to the overheating of 2d electron gas in the inversion layer .
again , the amplitude of oscillations from the main series are unchanged .
we attribute the observed anomalies to the large difference between the electron structure of the ground subband and excited subbands .
the twofold - degenerate ground subband contains many landau levels even for highest magnetic fields .
the electron mobility in this subband quickly decreases with increasing gate voltage .
the roughness of si / sio interface dominates the electron scattering .
the dingle temperature is well above the electron layer temperature as witnessed by the temperature independent amplitude of the sdh oscillations . the quasiclassical picture of the electric conductivity seems to be relevant .
the occupation of excited subbans is very small and only a small number of landau levels should be occupied even at relatively week @xmath27 .
the electrons in the subband attached to the level @xmath11 have an effective mass 0.19 , the same as the electrons in the ground subband . with the difference @xmath38 of a few mev at high concentration limit
, the first landau level can cross the fermi energy at relatively week @xmath39 .
then the subband is emptied and the density of states of electrons belonging to the twofold - degenerate ladder is halved .
this can contribute to reduction of the scattering rate and to the negative magnetoresistance .
on the other hand , we believe that the fourfold - degenerate subband is not emptied by increasing @xmath27 and that the level @xmath0 stays pinned to the fermi energy also in the presence of the magnetic field .
we suggest that it is responsible for the novel oscillations the period of which does not depend on @xmath5 .
the landau level separation in this subband is smaller due to the larger cyclotron effective mass @xmath400.43 .
therefore , the amplitude of oscillations should be more sensitive to the temperature , in agreement with the experimental data .
the experiments also indicate that the dingle temperature of these electrons is rather low . assuming the fourfold degeneracy
, the concentration of electrons corresponding to @xmath41 would be close to @xmath42 .
two very different groups of electrons contribute to the magnetoresistance of a sample with high electron concentrations : the electrons from the ground subband and the more mobile electrons from the excited subbands . while the transport mediated by the ground state electrons should by described quasiclassically ,
the quantization of electron states is important for excited subbands . therefore , the deviations from the standard quasiclassical model of transport by two groups of carries should be anticipated . with increasing temperature
the energy @xmath43 becomes comparable with the separation between landau levels .
the electron - electron scattering increases , the importance of a small group of electrons from the excited subband decreases and the observed anomalies are suppressed .
this work was supported by avoz 1 - 010 - 914 and in part by the grand agency of the czech republic under grants no .
202/01/0754 and 202/96/036 . | we present an experimental study of electron transport in inversion layers of high - mobility si(001 ) samples with occupied excited subbands .
the second series of oscillations , observed in addition to the main series of shubnikov - de hass oscillations , is tentatively attributed to the occupation of a subband associated with the @xmath0 level . besides
, a strong negative magnetoresistance and nonlinear field dependence of the hall resistance accompany the novel oscillations at high carrier concentrations .
the heating of the 2d electron layers leads to suppression of the observed anomalies . , , , , and si mosfet , magnetoresistance , hall effect 73.40.qv , 73.50.jt |
non - local quantum correlations or entanglement between space - separated parties is one of the most fertile and thought - generating properties of quantum mechanics .
recently it has become a very useful resource for many of the applications in quantum information theory and this has led to a lot of work devoted to understanding how it can be quantified and manipulated .
bipartite pure state entanglement is almost completely understood , while many questions are still open for the mixed state case . for pure states ,
the schmidt decomposition @xcite has proven to be a very useful tool , since it allows to write any pure state shared by two parties a and b in a canonical form , where all the information about the non - local properties of the state is contained in the positive schmidt coefficients .
the non - local properties of quantum states can be also specified by means of other quantities invariant under the action of local unitary transformations .
an interesting type of these invariants are given by polynomial combination of the coordinates of the state in a product basis , and the relation between these invariants and the schmidt coefficients is well known .
some novel aspects , compared to the bipartite case , appear for entangled systems of more than two parties . in this work we study the canonical forms of three - qubit pure states , extending the results of bipartite systems .
first we analyze the forms proposed for generalizing the schmidt decomposition for three - qubit pure states .
then , we relate one of these decompositions to the polynomial invariants studied in @xcite .
we give a one - to - one correspondence between a canonical form for a three - qubit pure state and a complete set of polynomial invariants describing its entanglement properties .
we also classify the different types of canonical forms by means of the minimal number of local bases product states ( lbps ) , i.e. the minimal number of non - local parameters , needed for the specification of a state . for any three - qubit pure state we give its decomposition with the minimal number of lbps and the procedure that has to be applied in order to build it .
finally we indicate how to generalize the results to systems of @xmath0-qubits , where many difficulties arise .
the schmidt decomposition has been a very useful tool for the study of entanglement properties of bipartite systems . for a generic bipartite pure state @xmath1 it reads [ bischmidt ]
|=_i=1^l _ i|ii , _ i0 , where @xmath2 , @xmath3 , being @xmath4 orthonormal vectors in each subsystem , and @xmath5 are the schmidt coefficients
. it would be very interesting to find for three - qubit pure states a canonical decomposition generalizing the features of the schmidt decomposition .
however , the trivial generalization is not possible @xcite and it is not evident how to extend the schmidt decomposition to the case of @xmath0-party systems ( @xmath6 ) .
indeed several forms have been proposed ( see for instance @xcite ) . in recent work
@xcite we gave a generalization of the schmidt decomposition for three - qubit pure states , in the sense that the coefficients of this decomposition carry all the information about the non - local properties of the state , and do so minimally and unambiguously , i.e. the decomposition is not superflous . starting from a generic state shared by three parties , a , b and c , [ state ] | in this paper we analyze the canonical forms into which any pure three - qubit state can be cast .
the minimal forms , i.e. the ones with the minimal number of product states built from local bases , are also presented and lead to a complete classification of pure three - qubit states . this classification is related to the values of the polynomial invariants under local unitary transformations by a one - to - one correspondence . |
the cosmological principle ( cp ) is one the most fundamental hypothesis upon which the concordance model based . in this work ,
we discuss the validity of the cosmological isotropy with different compilations of type ia supernovae ( sne ) , namely the union2.1 @xcite ) and jla data sets @xcite , using a hemispherical comparison method , hence determining whether the cosmological isotropy actually holds in large angular scales , and whether such hypothesis is not only a mathematical simplification , but a valid assumption .
we test the isotropy of the universe expansion by mapping the @xmath0 and @xmath1 parameters through the celestial sphere , so that an opposite hemisphere comparison is performed following ref .
( see also ref . ) .
each pair of these hemispheres is well defined by the healpix pixelization scheme @xcite , such that we fit @xmath0 and @xmath1 by minimising the following quantity @xmath3 where the set @xmath4 contains the observational information of the sne data , i.e. , redshift , distance moduli and associated uncertainty of the _ i - th _ object , respectively , where @xmath5 is the distance modulus given by a specific cosmological model according to @xmath6 } + 42.38 - 5\log_{10}(h ) \;,\ ] ] where @xmath7 , @xmath8 , and @xmath9 is the adimensional luminosity distance , whose arguments are the redshift @xmath10 , in addition to the set of cosmological parameters @xmath11 which describe the underlying cosmological model , @xmath9 is given by a cosmographic expansion up to second order , where @xmath12 . ] .
furthermore , we quantify the angular non - uniformity of the data sets using the method named sigma - map , as performed in ref .
( see also ref . ) , which is based upon the two - point angular correlation function of the cosmic objects distribution computed inside each assigned hemisphere .
in other words , this estimator constructs a pixelised map in which its colour ranges from blue , when the actual distribution of sne is less correlated than the mean value expected in a random catalogue , to red , in the case when the correlation is larger .
in addition , we analyse the anisotropies of the cosmological parameters and the angular sne distribution not only in the pixel space , but in the multipole space as well , so that @xmath13 represents the quantity scanned through the celestial sphere , such as the @xmath0 and @xmath1 parameters . ] , and @xmath14 is the angular power spectrum of the hubble- , q- and sigma - map .
since we are interested in large scale angular correlations , we limit our analyses to @xmath15 .
the statistical significance of the hubble and q - maps analyses is estimated with two different approaches . in the first approach ,
the galactic coordinates of each sne is fixed , yet the set @xmath16 is shuffled ( hereafter _ shuffle _ test ) .
the second approach also keeps the original @xmath16 of each object yet the sne positions are isotropically redistributed on the celestial sphere ( hereafter the _ mc _ test ) . hence , we can test whether the directional dependence of these parameters are statistically significant in its amplitude as well as in its direction .
the results of the sigma - map analyses are shown in fig .
[ fig1 ] ( pixel space ) and fig .
[ fig2 ] ( multipole space ) for both sne data sets .
it is possible to note that they present a preferred direction on the celestial sphere , as discussed on the description of fig .
[ fig1 ] , and that the both sne catalogues are highly inconsistent with a perfectly isotropic distribution , since the analyses performed in multipole space present much higher @xmath17 s than their average values obtained by the mcs . moreover , the hubble- and q - map results are featured in fig .
[ fig3 ] ( [ fig4 ] ) for the union2.1(jla ) compilations , for the analyses performed in pixel space , whereas fig
. [ fig5 ] ( [ fig6 ] ) refer to the analyses carried out in multipole space for the union2.1 ( jla ) compilations as well .
+ we note that the direction @xmath18 obtained for the union2.1 hubble - map is consistent with the bulk flow motion direction estimated in ref . , that is , @xmath19 km / s towards @xmath20 , as well as many works which probed the cosmological isotropy with a similar approach @xcite .
moreover , the anisotropy of the @xmath0 can possibly explain the tension between the @xmath0 determinations @xcite ) from low-@xmath10 standard candles @xcite and planck cmb temperature @xcite .
it was found that the maximal @xmath0 variance through the celestial sphere is consistent with their values , and that its direction is consistent with the bulk flow motion as well .
this reinforces the idea that such anisotropy arises as a local effect , instead of an intrinsic cosmological anisotropy .
we also evaluate the strength of the correlation between these maps , finding a negligible correlation between the hubble- and q - maps with the sigma - map of the union2.1 data set ( @xmath21 and @xmath22 , respectively ) , even though the correlation is moderate in the jla analyses : @xmath23 and @xmath24 , respectively .
thus , we conclude that the anisotropy detected on the hubble and q - maps in the jla data is possibly explained by the incompleteness of the sample in terms of sky coverage , while the anisotropy pointed by the union2.1 sne is most likely of local origin .
+ the results of the statistical significance are depicted , in multipole space , in figs .
[ fig5 ] and [ fig6 ] for the union2.1 and jla case , respectively .
it is possible to note that the union2.1 hubble - map present mild disagreement with the _ mc _ and _ shuffle _ tests specially in the lower @xmath25 ( @xmath26 ) , and that the q - map strongly disagrees with both realisations except for the dipole case , thus showing significant evidence for anisotropy in this analysis .
nevertheless , this signal can be ascribed to the limited constraining power of the union2.1 data on this parameter , besides the degeneracy with the hubble parameter . for the jla catalogue
, there is a better agreement between the real data and the _ shuffle _ realisations in both hubble- and q - maps , except for the dipole contributions , whereas the data presents stronger disagreement with the _ mc _ runs specially in the q - map analysis .
this result shows , once again , the fact that the angular non - uniformity of the jla sample indeed biases the hubble and q - map analyses .
we have shown that the anisotropy detected on the @xmath0 mapping with the sne sample can be attributed to the bulk flow motion due to its proximity of reported directions in the literature , and the @xmath1 significant anisotropy probably arises due to the limitation of this data set . on the other hand ,
the jla directional analyses show a significant dependence with its celestial coverage , then biasing the hubble and q - map results .
therefore , we conclude that there is no significant violation of the cosmological isotropy with the latest sne data in the @xmath27 range , albeit next - generation surveys such as lsst and euclid may improve this test with the greater precision and much larger data sets that they shall provide .
we thank roy maartens , ribamar reis and ivan soares ferreira for helpful discussions .
we acknowledge the healpix package for the derivation of our analyses .
this work is supported by cnpq , faperj and capes . | we investigate the validity of the cosmological principle by constraining the cosmological parameters @xmath0 and @xmath1 through the celestial sphere .
our analyses are performed in a low - redshift regime in order to follow a model independent approach , using both union2.1 and jla type ia supernovae ( sne ) compilations .
we find that the preferred direction of the @xmath0 parameter in the sky is consistent with the bulk flow motion of our local universe in the union2.1 case , while the @xmath1 directional analysis seem to be anti - correlated with the @xmath0 for both data sets .
furthermore , we test the consistency of these results with monte carlo ( mc ) realisations , finding that the anisotropy on both parameters are significant within @xmath2 confidence level , albeit we find a significant correlation between the @xmath0 and @xmath1 mapping with the angular distribution of sne from the jla compilation .
therefore , we conclude that the detected anisotropies are either of local origin , or induced by the non - uniform celestial coverage of the sne data set . |
currently the standard picture of galaxy formation involves the collapse of baryonic matter in hierarchically clustering dark matter halos and the subsequent building of big galaxies from small ones via merging processes e.g. , @xcite .
while recent cosmological simulations can adequately reproduce many global properties of galaxies and their correlations , the details are still very much dependent on the gas physics and stellar feedback involved ( see e.g. , @xcite ) . additionally , most , if not all , galaxies harbor supermassive black holes in their center @xcite .
correlations have been recently detected between black hole masses , galaxy masses , and central velocity dispersions in galaxies @xcite .
these correlations are strong evidence that black holes in galactic nuclei are linked to the dynamical history of their host galaxies . @xcite and @xcite
demonstrate how this is consistent with the framework of semi - analytic models that follow the formation and evolution of galaxies in a cold dark matter - dominated universe .
they assume supermassive black holes are formed and fueled during major mergers , qualitatively explaining many aspects of the observed evolution of galaxies , including the observed relation between bulge luminosity , velocity dispersion , and central black hole mass .
as already discussed by @xcite , such a scenario requires the formation of galactic nuclei containing at least two black holes , depending on the black hole merger rate relative to the galaxy merger rate . however , there is very little observational evidence for massive black hole binaries @xcite .
this conflict between theory and observations has become known as the `` sinking black hole problem '' . as an alternative to minimally impacting stellar dynamical processes ,
@xcite and @xcite have proposed mechanisms which lead to rapid decay of massive black hole orbits and subsequent black hole mergers in galactic centers .
also , @xcite offered the solution that gas accretion could dominate the orbital decay in the intermediate phase of the sinking black hole problem when dynamical friction becomes inefficient .
however , as we will discuss later , dynamical friction , as laid out by @xcite , is not sufficiently effective by itself to lead to rapid coalescence of black hole binaries .
if there are no quick mergers , multiple black hole nuclei could lose black holes through slingshot ejections @xcite .
once a binary system becomes hard , the high orbital velocities of the black holes allow further hardening through close encounters and three - body interactions with stars .
such processes will evacuate field stars from the surroundings of the binary , therefore it can be argued that the stellar scatterings can not produce rapid coalescence .
the preceding argument assumes that the center of mass of the binary does not move with respect to the stellar system .
however , we will show that even with a fairly symmetrical initial setup the binary gains some linear momentum .
this introduces a wandering motion which exceeds the expectations from equipartition .
the wandering of the binary guarantees an adequate supply of stars for binary hardening and rapid coalescence through purely stellar dynamical processes .
our new computational method allows us to study in detail three - body interactions of a black hole binary with field stars .
although one may argue that the perturbing mass of the field stars is small compared to the black hole mass and should have negligible impact , there are many stars , and each encounter can lead to changes in binding energy and eccentricity of the black hole binary .
in fact , our models show that the black hole binary keeps a rather high eccentricity due to the encounters .
thus high eccentricity will speed up gravitational radiation mergers very efficiently , and is , as noted by @xcite and @xcite , a way to expedite massive black hole mergers in a purely stellar dynamical way .
the correct theoretical prediction of the frequency of black hole mergers in galactic environments will be important in the search for gravitational waves .
the merging of supermassive black holes of @xmath2 to @xmath3 in the nuclei of merging galaxies and protogalaxies can be detected with high signal - to - noise at redshifts from @xmath4 @xcite by the laser interferometer space antenna ( lisa ) @xcite .
previous attempts to quantify this prediction have been made by either solving the perturbed two and three - body problem in simplified models @xcite , direct @xmath0-body models @xcite , or a combination of the two @xcite .
simulating binary black hole hardening is extremely challenging , algorithmically and computationally .
since the mass differences between the black holes and the stars is so large , high particle numbers are required in order to model the relaxation processes around the black holes accurately .
the simulations have used softened particles on special purpose computers @xcite or a hierarchical hybrid code in which all forces involving the black hole particles are keplerian @xcite .
these schemes used particle numbers in the order of @xmath5 . in this paper , we describe a new hybrid field - particle code which treats all particles with orbits crossing the central regions of the system with a high precision direct @xmath0-body method appropriate for collisional stellar dynamics .
all other particles are integrated using a field method . in order to adapt both parts of the hybrid code to each other , the field method ( approximating the potential exerted by a set of particles by a series expansion , referred to here as `` scf '' ) had to be upgraded to a fourth order hermite integrator .
this integration also uses the time derivative of the potential , as in modern direct @xmath0-body codes . in the following sections some details of the sinking black hole problem are introduced .
section [ sec : es - intro ] introduces the integration software used for the numerical experiments described in this paper .
section [ sec : es - testing ] is devoted to a comparison between the new collisional code with a well used workhorse simulator in this field called _ nbody6 _ @xcite , using its parallel implementation _
nbody6++ _ @xcite .
the application of the code to the sinking binary black hole problem is reported in section [ sec : blkh - runs ] .
a numerical simulation of the hardening phase ( until the massive black holes start to radiate gravitational waves ) must be able to accurately follow three - body encounters .
for this reason , the keplerian potential should not be softened in the denser parts of the system .
computationally , the central part of the system would best be treated using a collisional integrator .
the code must be able to integrate encounters leading to large angle deflections in an efficient way , while requiring neither too much computing time nor introducing energy errors .
the overall @xmath0-body integration does not need to be symplectic , but should keep the energy error as low as possible . on the other hand , in a system showing a core halo structure , the bulk of the stars in the halo move under the influence of the mean field of the whole cluster .
the halo part of the central galactic cluster can be integrated with a mean field method . in the new method ( which we will refer to as eurostar ) , both the collisional code nbody6++ @xcite and the scf method @xcite are merged to optimize large-@xmath0 collisional @xmath0-body simulations .
the star cluster , which is assumed to be in equilibrium , is divided into two sections by applying a critical angular momentum criterion .
as shown in figure [ fig : decomp - scheme ] , this allows for distinction between particles orbiting solely in the halo and ones which have trajectories leading through the core of the system . in a stationary gravitational point mass system
, two - body relaxation leads to an exchange between halo particles and core particles in a divided cluster . in a system of
more than @xmath5 particles , only a few particles cross the core halo border per dynamical time .
this is very fortunate because it allows us to integrate the orbits of the halo particles with a collisionless method and the core particles with a collisional code .
exchanges of particles cause energy conservation problems since the contribution of a particle to the main potential would be changed from keplerian to a sample point of a mean field in eurostar .
this means that switching of particles from the core to the halo part of the integrator and vice versa is not permitted .
nbody6++ integrates trajectories of point masses in the core of the system .
it is an aarseth - type direct force integrator applying the hermite integration scheme .
nbody6++ gains its efficiency by implementing an ahmad - cohen neighbor scheme and individual block time steps @xcite .
close interactions between particles are treated by regularization of the equations of motion @xcite .
nbody6++ scales well on parallel computer systems and can also be used with the grape special purpose computer @xcite .
the hermite scheme requires one to compute @xmath6 and @xmath7 at each time step where , @xmath8 .
\label{eq : direktfdot}\end{gathered}\ ] ] the relative distance between particles @xmath9 and @xmath10 is given by @xmath11 .
accordingly , the relative velocity is @xmath12 .
the extra effort of computing two direct force quantities allows one to approximate the particle s orbit to fourth order . by storing @xmath6 and @xmath7 from the previous time step it is possible to interpolate the next two higher derivatives and to apply a predictor / corrector scheme @xcite .
the scf method qualifies for the collisionless part of a spherical system since then the basis functions are given analytically .
this allows one to implement the hermite scheme for scf , which makes scf an ideal far force extension to nbody6++ .
its drawback , however , is that this method restricts the input systems to have an approximately spherical particle distribution around the coordinate center .
the possible asphericity depends on the number of spherical harmonics used for the potential expansion . in order to have better convergence , one - parameter basis functions for @xmath13 and @xmath14
are used @xcite : @xmath15 the @xmath16 are called ultraspherical or gegenbauer polynomials .
the spherical harmonics are given by the @xmath17 .
once the @xmath18 for a certain set of particles are known , an analytic expression of the potential and the density is found . due to the truncation of the expansion
they represent the mean field and mean density .
this means that the force at each position and its derivative can be computed using the following expressions , @xmath19 since equations ( [ eq : gradphi ] ) and ( [ eq : dotgradphi ] ) lead to a significant modification of the scf scheme , we provide detailed form of these expressions in the appendix .
as a first test for the new method , we have followed the last stages of an ongoing merger between two galaxies , each containing a supermassive black hole . for the initial setup , it is assumed that the stellar systems have already arranged themselves into a spherical system .
the two , formerly central , supermassive black holes are moving through the stellar component with a speed on the order of the relative velocity between the two initial galaxies . in the present simulations ,
the stellar component is a realization of a plummer model .
the density and potential of the spherically symmetric plummer model are given by @xcite : @xmath20 the quantity @xmath21 describes the total mass of the system and @xmath22 is the gravitational constant .
with the plummer radius chosen to be @xmath23 , the half mass radius of this system is at a radius of @xmath24 in the model units of the simulations .
the total mass of the system @xmath21 is set to unity .
the gravitational constant @xmath22 is set to unity as well , conforming to the model units described by heggie & mathieu ( 1986 ) .
the stellar system is centered around the origin .
the black hole particles contain 1% of the system s total mass .
their initial positions are at @xmath25 , and their initial velocities are 13.6% of the circular velocity at their initial radii .
the black hole orbits are analyzed during the simulation assuming the orbit can be approximated by the classical two - body problem .
the binding energies and eccentricities of the black hole orbits are computed from their relative distances and velocities , assuming a keplerian potential . once the black holes become bound
, their two - body attraction is the most important force .
the eccentricity of the binary @xmath26 and the binding energy @xmath27 are computed as follows , using the definition @xmath28 .
the vectors @xmath29 and @xmath30 denote the position vectors of the black holes @xmath31 and @xmath32 . from this
it follows that @xcite , @xmath33 this method of analysis provides a sensitive measure for the moment when the black holes become bound to one other .
furthermore , this way of analyzing the data also offers a precise tool for following the hardening of the binary . since this sinking binary problem is the first application of eurostar ,
its results are compared with those of the fully collisional code nbody6++ . for the comparison
runs , 16384 particles were simulated .
figure [ fig : esnb6bin - epshh ] shows the results for the two comparative runs .
the plot on the left hand side in figure [ fig : esnb6bin - epshh ] shows the eccentricity of the binary as a function of the simulated time in model units .
the plot on the right hand side shows the two - body binding energy as a function of time .
the binary becomes bound after 10 time units , in both cases .
the fully collisional method and the hybrid code eurostar show slightly different sinking rates for the binary at the beginning of the simulation .
these differences result from the different density of collisional particles around the black holes in both codes . while the black holes in the fully collisional run suffer small angle encounters with every particle in the system , this is not possible in the hybrid code .
naturally , all particles treated by the mean field method can only interact with the system through changes in the mean potential .
after the binary has become bound , the hardening process is driven by the stars which have a small enough impact parameter such that they have an encounter timescale smaller than the orbital timescale of the massive binary .
therefore , the hardening depends more on the two - body encounters with neighboring particles , which have large orbital velocities . with increasing simulation time , the binary locks into an oscillating motion around the center of mass of the stellar component .
this motion does not extend beyond of the dense galactic core . effectively , the differences in the density of the collisional particles between the two methods vanish after the binary has become bound .
figure [ fig : esnb6bin - epshh ] reflects this by showing a parallel evolution of eccentricity and binding energy for times larger than 10 time units in the simulation .
this new code is intended for raising the total particle number for collisional simulations of spherical @xmath0-body systems .
hence the evolution of one , two or several massive bodies in a dense stellar cluster appears to be an ideal problem for eurostar .
this is why we are addressing here the problem of a sinking massive black hole binary in galactic centers .
another useful potential application os the dynamics of globular clusters .
the particles representing the stellar component are distributed according to plummer s model with @xmath34 .
the total mass of the stars is fixed at @xmath35 , while the black hole particles carry @xmath36 each , so @xmath37 .
this is a fairly high mass for the black holes compared to the total mass of the stellar system , since ferrarese and merritt @xcite found the black hole mass in bulges to be smaller than that .
however our simulations start at a situation resembling the final stage of a galactic merger , which means we are concentrating on the innermost part of the allready spherical system .
the black hole particles are initially placed symmetrically about the center of mass of the stellar component .
their initial radii are @xmath38 , their initial velocities are 13.6% of the circular velocity at this radius . in the given model units ,
this represents starting points for the black holes at @xmath25 and @xmath39 .
the center of mass of the stellar component is at the origin .
the mass factor between a stellar particle and a black hole particle is : @xmath40 for 131072 particles , @xmath41 for 65536 particles , and @xmath42 for 32768 particles . in order to have a statistical basis for analysis
, we compare the results from five runs with 32768 particles , two runs with 65536 particles , and three runs with 131072 particles .
not all runs reached the @xmath43 time unit mark due to time step scheduling problems caused by accuracy problems in very close encounters between stars and a black hole particle .
the regularization methods implemented in eurostar are identical to the ones suitable for open or globular cluster simulations .
the extreme situation in the late stages of the sinking binary black hole problem may cause the chain algorithm to fail @xcite .
this problem can be solved by applying different regularization methods . however , up to the point of failure , the simulations conserved the total energy with relative errors below @xmath44 . in all runs ,
the binary becomes bound at approximately @xmath45 time units .
the parameters of the hybrid code have been adjusted in the following way : the scf part uses the parameter @xmath46 for the basis functions . with this choice ,
the basis functions represent a plummer model to zeroth order , which is in accordance with the models used by @xcite .
also , this choice ensures an optimal representation of the actual potential by the expansion method . to allow flexibility in the expansion
, seven basis functions are used for the radial direction and five ( @xmath47 $ ] , @xmath48 $ ] ) for the angular expansions .
the nbody6++ part uses @xmath49 for the irregular time steps and @xmath50 for the regular time steps .
furthermore , the ahmad - cohen neighbor scheme @xcite has been modified in such a way that the search radius for neighbors is enhanced by a factor of @xmath51 , @xmath52 , and @xmath53 for interactions with the black holes in the runs with 32768 , 65536 , and 131072 particles respectively .
.number of sample points in the bins used for analyzing the motion of the binary .
the bins for the evolution time @xmath54 in @xmath0-body units are centered around @xmath55 .
plots using bins for the total particle number @xmath56 have the number of sample points given in the row labeled `` total '' . [ cols= " > ,
> , > , > , > " , ] this means we find a center of mass motion for the binary which exceeds the expected value from brownian motion .
however , the motion is enhanced by larger factors than proposed by @xcite or by @xcite . a more detailed discussion of this result remains for future work .
following @xcite , dynamical friction becomes inefficient as the driving forces behind the binary black hole orbit decay after it becomes hard . in order to put constraints on this
, we estimate the influence of dynamical friction on the decay of the binary in the simulation . from @xcite ,
we take the following expression , which is derived in @xcite .
@xmath57 equation ( [ eq : dynbinney ] ) describes the deceleration of a particle with mass @xmath58 under the influence of weak encounters with surrounding particles having a uniform mass @xmath59 . in the special case of a plummer model , the integral in equation ( [ eq : dynbinney ] )
can be expressed in terms of the escape velocity @xmath60 @xcite , @xmath61 the quantity @xmath62 defines the number density of the stars at the position @xmath63 and @xmath64 . taking the limit of a continuous system with @xmath65 ,
the term @xmath66 becomes @xmath67 .
the integral over @xmath68 in equation ( [ eq : myplum ] ) can then be solved in a closed form . if the motion of the black holes is determined by their self interaction plus a frictional force term , this friction can be linked to the decay of the angular momentum @xmath69 as follows : @xmath70 where @xmath71 is the radial acceleration of the two body motion of the black holes , @xmath72 is the dynamical friction acting on each black hole , and @xmath73 is the two - body velocity of the black holes . with equations ( [ eq : dynbinney ] ) and ( [ eq : myplum ] ) we can evaluate the impact of dynamical friction on the orbital angular momentum according to @xmath74 the gravitational constant @xmath22 is unity in our model units , and the black holes have mass @xmath75 each .
we use the mean orbital velocities for @xmath76 . in order to evaluate @xmath77
, we use the mean separation of the black holes for @xmath78 , which introduces only a very small error in a plummer model . assuming a linear behavior for @xmath79 , we find @xmath80 .
using this to estimate the angular momentum from the slopes @xmath32 of the linear fits in figure [ fig : llalltime ] , we find @xmath81 .
this result shows that the usual assumption of large @xmath82 does not hold .
both the possibility of a linear fit for the evolution of @xmath69 and the small hardening rate @xmath83 indicate that mainly dynamical friction causes the shrinking of black hole orbits in our simulations .
as stated before , the collisional simulations which include black hole particles do not reach the observed mass contrast in galactic nuclei . in order to transform simulation units to physical units , a system size in parsec or a stellar mass in units of solar mass has to be chosen . setting the gravitational constant @xmath22 ,
all remaining units can be rescaled @xcite . in the following ,
a run with 65536 particles is scaled to a physical stellar system .
since this work focuses on the dynamics of galactic nuclei , the physical mass of the supermassive objects motivates the following choices , @xmath84 the masses are chosen so that the total mass of the system is @xmath85 and the mean mass of a particle is @xmath86 .
this choice means that every stellar particle with mass @xmath59 represents a compact star cluster with the order of @xmath5 particles .
the chosen mass for the black hole particle has approximately the same mass as the central black hole of m31 @xcite .
the conversion between physical units and @xmath0-body units follows @xmath87 for simulated quantities . by choosing the central velocity dispersion to be @xmath88 , we find @xmath89 and @xmath90 , @xmath91 in a plummer model , the half mass radius @xmath92 is related to the plummer radius @xmath93 by @xmath94 @xcite . since @xmath95 ,
the half mass radius in model units is @xmath96 .
therefore , the initial model for the simulated decay of a black hole in the galactic center is a plummer sphere with a half mass radius of 272.23 pc .
the initial distance between the black holes is 355.39 pc .
they become bound after approximately 40 million years .
the total simulated time is approximately 190 million years . at the end of the simulation the black hole distances vary from 1 pc at apocenter to 0.2 pc at pericenter .
the semi - major axis of the first bound orbit is 21 pc .
this scaling allows us to compare our results with @xcite .
we find that our smallest average orbits at the end of the simulation are not yet small enough that gravitational radiation , according to their estimates , would dominate the evolution time scale .
however , at the end of our simulations evolution is still dominated by dynamical friction and not by long evolution time scales for hard binaries as proposed by @xcite .
their estimate for the gravitational radiation shrinking time scale assumes circular orbits for the binary . with eccentricities of roughly 0.85 for the black hole binaries in our runs
, we expect gravitational radiation to be efficient and coalescence in roughly @xmath97 years after our simulations stopped .
we have created a new @xmath0-body hybrid code by merging a high accuracy direct hermite integrator of the standard type @xcite with a collisionless @xmath0-body method which approximates the potential of a given particle distribution by a series expansion @xcite .
the scf method has been completely rewritten to include a computation of the time derivative of the gravitational force and a fourth order hermite integrator .
we have used this code to model a galactic nucleus containing two massive black holes with up to 128k single particles .
the evolution of the binary black hole is followed from an initial phase , to a phase driven by standard dynamical friction where the binary is bound , and then further hardened by three - body encounters with single stars . in that hardening phase , we take full advantage of the regularized three - body integration developed by @xcite and @xcite .
the method proves to work well , and reproduces standard expectations , such as the chandrasekhar dynamical friction in the initial phase . in the final hardening phase due to
three - body encounters , we find that the eccentricity of the black hole binary maintains a fairly large value ( around 0.85 ) .
this is very interesting because it decreases the time scale for gravitational radiation merger of binary black holes dramatically , thus increasing our chances of detecting gravitational radiation from such events with lisa .
due to computational limitations , however , our particle numbers are still not large enough to fully describe the real physical situation .
any further scaling is problematic , and so further work with improved hardware and software must be done .
we study in detail the motion of a black hole binary in the center of a galaxy .
we find that the wandering motion does not decay with increasing particle number as expected .
the mechanism exciting these anomalous motions is unclear .
if they exist in simulations with realistic particle numbers , they will solve the problem of feeding the black holes with fresh stellar dynamical material raised by @xcite .
the authors would like to thank s. aarseth , d. heggie , w. sweatman , c. theis , c. boily , d. merritt , m. milosavljevi , f. cruz , h. baumgardt , g. hensler , l. hernquist , h. s. zhao , p. ghavamian and e. barnes for fruitful help and discussion .
this project is funded by _
deutsche forschungsgemeinschaft _ ( dfg ) project sp 345/9 - 1,2 and sonderforschungsbereich ( sfb ) 439 funded at the university of heidelberg , nsf grant ast 00 - 71099 , nasa grants nag 5 - 7019 , nag 5 - 6037 , and nag 5 - 9046 . technical help and computer resources
are provided by _
nic _ in jlich , _ hlrs _ in stuttgart , _ tracs _ and _ epcc _ in edinburgh , _ zib _ in berlin , _ ssc _ in karlsruhe , university of heidelberg , rutgers university , and university of kiel , and by the pittsburgh supercomputer center and the san diego supercomputer center .
the authors thank the aspen center for physics , the institute of astronomy and the lorenz center at leiden university for hospitality .
the sources for eurostar are available from the authors or via ` http://www.physics.rutgers.edu/~marchems/ ` aarseth , s. 1996 , in p. hut and j. makino ( eds . ) , _ dynamical evolution of star clusters _
, pp 161170 , international astronomical union aarseth , s. j. 1993 , in g. contopoulos , n. k. spyrou , and l. vlahos ( eds . ) , _ galactic dynamics and @xmath0-body simulations _ , vol
. 433 of _ lecture notes in physics _ ,
pp 365417 , springer
verlag , thessaloniki aarseth , s.j .
1999 , , 127 aarseth , s.j . , hnon , m. , wielen , r. 1974 , , 183 abramowitz , m. and stegun , i. a. ( eds . ) 1972 , , dover , new york , 9 .
edition ahmad , a. and cohen , l. 1973 , , 349 armitage , p.
j. and natarajan , p. 2002
, , l9 begelman , m. c. , blandford , r. d. , and rees , m. j. 1980 , , 307 binney , j. and tremaine , s. 1987 , , princeton university press , princeton , 1 .
edition boccaletti , d. and pucacco , g. 1996 , , vol . 1 of _ astronomy and astrophysics library _ , springer verlag , berlin , heidelberg , new york , 1 .
edition chandrasekhar , s. 1943 , , 255 chatterjee , p. , hernquist , l. and loeb a. 2002 , , in press clutton - brock , m. 1973 , , 55 danzmann , k. 2000 , , proceedings of the h0.1 symposium of cospar scientific commission h , held during the 32nd cospar scientific assembly , ed . s. vitale , pergamon press , p. 1129
diaferio , a. , kauffmann , g. , colberg , j.m . , white , s.d.m .
1999 , , 537 ferrarese , l. , merritt , d. 2000 , , l9 gebhardt , k. , bender , r. , bower , g. , dressler , a. , faber , s.m . , filippenko , a.v . ,
green , r. , grillmair , carl , ho , l.c . , kormendy , j. , lauer , t.r . ,
magorrian , j. , pinkney , j. , richstone , d. , tremaine , s. 2000 , , l13 gould , a. and rix , h .- w .
2000 , , l29 haehnelt , m.g . ,
kauffmann , g. 2000 , , l35 halpern , j.p . ,
eracleous , m. 2000 , , 647 heggie , d. c. and mathieu , r. m. 1986 , in p. hut and s. l. w. mcmillan ( eds . ) , _ the use of supercomputers in stellar dynamics _ , pp 233235 , new york hernquist , l. and ostriker , j. p. 1992
, , 375 hernquist , l. , sigurdsson , s. and bryan , g. l. 1995 , , 717 hills , j. g. 1992 , , 1955 holley - bockelmann , k. , mihos , j.c . , sigurdsson , s. and hernquist , l. 2001 , * 567 * , 187 kauffmann , g. , colberg , j.m . ,
diaferio , a. , white , s.d.m .
1999a , , 529 kauffmann , g. , colberg , j.m .
, diaferio , a. , white , s.d.m .
1999b , , 188 kauffmann , g. , haehnelt , m.g .
2000 , , 576 kormendy , j. and richstone , d. 1995 , , 581 kustaanheimo , p. and stiefel , e. 1965 , , 204 lehto , h.j . , valtonen , m.j .
1996 , , 207 magorrian , j. , tremaine , s. , richstone , d. , bender , r. , bower , g. , dressler , a. , faber , s. m. , gebhardt , k. , green , r. , grillmayr , c. , kormendy , j. , and lauer , t. 1998 , , 2285 makino , j. 1997 , , 58 makino , j. , fukushige , t. , okumura , s. k. , and ebisuzaki , t. 1993 , , 303 merritt , d. 2000a , in f. combes ( ed . ) , _ galaxies and the universe _ , asp - conference series , pp 221 , paris merritt , d. 2000b , , astro - ph/0012264 merritt , d. and ferrarese , l. 2001 , , l30 merritt , d. and quinlan , g. d. 1998 , , 625 mikkola s. , aarseth s.j .
1993 , mikkola s. , aarseth s.j .
1996 , mikkola s. , aarseth s.j .
1998 , , 309 mikkola , s. and valtonen , m. j. 1992 , , 115 navarro , j.f . , steinmetz , m. 2000 , , 477 peebles , p. j. e. 1993 , , princeton university press , princeton , 1 .
edition phinney , e.s .
2000 , , 46.01 plummer , h. c. 1911 , , 460 quinlan , g. d. 1996 , , 35 quinlan , g. d. and hernquist , l. 1997 , , 533 richstone , d. , ajhar , e.a .
, bender , r. , bower , g. , dressler , a. , faber , s.m . ,
filippenko , a.v
. , gebhardt , k. , green , r. , ho , l.c .
, kormendy , j. , lauer , t.r .
, magorrian , j. , tremaine , s. 1998 , , 14 sigurdsson , s. , he , b. , melhem , r. and hernquist , l. 1997 , 378 spitzer , l. j. 1987 , , princeton university press , princeton , nj , 1 .
edition spurzem , r. 1999 , , 407 spurzem , r. and baumgardt , h. 2001 , , ari
preprint no .
68 , submitted to spurzem , r. and kugel , a. 1999 , , astro - ph/9906155 , to appear in procs . of molecular dynamics on parallel computers , workshop of the john von neumann - institute for computing ( nic ) jlich , 1999 , world scientific , singapore sugimoto , d. , makino , j. , taiji , m. , and ebisuzaki , t. 1995 , in _ proceedings of the first aizu international symposium on parallel algorithms / architecture synthesis _ , pp 3844 , ieee computer society valtonen , m. j. , mikkola , s. , heinmki , p. , and valtonen , h. 1994 , , 69 zhao , h. 1996 , , 488
throughout the computation for the forces and force derivatives in our scf - scheme several special functions have to be tabulated .
recurrence relations provide a very efficient means of calculating these functions .
the following recursion relations have been applied to compute ultraspherical polynomials and their derivatives . as starting values for @xmath98
the recurrence formulae for the gegenbauer or ultraspherical polynomials obey the relation @xmath99 the expressions for higher values are given by : @xmath100 from that the first derivative can be computed as : ( @xcite , equation ( 22.7.22 ) and table 22.7 . )
@xmath101 for practical reasons and higher accuracy the second derivative polynomial is computed using equation ( [ eq : ultrasp ] ) : @xmath102 [ kap : gradphidot ] because the particle track is approximated by using the hermite scheme , one has to find forces and the first force derivative simultaneously .
an approximation using two timesteps for the first force derivative introduces errors to the second and third derivative of the forces .
all particles move within a time dependent potential ; therefore , the first derivative has a term describing the change of the potential and a term describing the change of force depending on the particle s orbit . @xmath103 with the help of the orbit integration for the single particle in a given static potential case , equation ( [ eq : scfadot ] ) evaluates to : @xmath104 the evalutation of the first term on the right hand side of equation ( [ eq : scfadot ] ) is given in section [ kap : gradphidottim ] .
the derivatives with respect to the spatial coordinates in equation ( [ eq : scfadotorb ] ) can be found in section [ kap : gradphidotorb ] .
because all positions and velocities of the dataset are time - dependent , the partial derivatives with respect to @xmath54 apply only to the coefficients @xmath18 .
these are implemented as the variables @xmath105 , @xmath106 , @xmath107 , and @xmath108 , from which the partial derivative can be formed : @xmath109 , \end{split } \\
\frac{\partial}{\partial t } \ , p_{lm}(\cos(\vartheta_k ) ) = - \frac{\partial \vartheta_k}{\partial t } \ , \sin(\vartheta_k ) \ , \frac{\partial p_{lm}(\cos(\vartheta_k))}{\partial \cos(\vartheta_k ) } , \\
\frac{\partial}{\partial t } \ , \cos(m \varphi_k ) = - m \frac{\partial \varphi_k}{\partial t } \ , \sin(m \varphi_k ) , \\
\frac{\partial}{\partial t } \ , \sin(m \varphi_k ) = m \frac{\partial \varphi_k}{\partial t } \ , \cos(m \varphi_k ) .
\end{gathered}\ ] ] in the coefficient computation section the standard leap frog integrator provided by @xcite is extended by two additional variables , which are computed by using the recursion relations in section [ kap : rekursion ] .
in order to account for the change of force due to the particle orbit one has to calculate the nine partial derivatives in equation ( [ eq : scfadotorb ] ) . these nine derivatives will now be listed . in order to save some space the second derivative of @xmath110
is given first : @xmath111 \ , \frac{c_{n - 1}^{(\omega + 1)}(\xi)}{c_n^{(\omega)}(\xi ) } \\ & \quad + 16 \ , \biggl ( \frac{r^{\frac{1}{\alpha}}}{\alpha r } \biggr)^2 \ , \frac{\omega ( \omega + 1)}{(1 + r^{\frac{1}{\alpha}})^4 } \ , \frac{c_{n - 2}^{(\omega + 2)}(\xi)}{c_n^{(\omega)}(\xi ) } \label{eq : scfdtwophinltil } \end{split}\ ] ] the nine derivatives can be implemented as follows : the radial derivative for the radial acceleration becomes : @xmath112 , \\
\intertext{with : } g_{lm } & = n_{lm } \ , \sum\limits_{n = 0}^{\infty } \ , \tilde{a}_{nl } \frac{\partial^2}{\partial r^2 } \tilde{\phi}_{nl}(r ) \ , \sum\limits_k \ ,
m_k \tilde{\phi}_{nl}(r_k ) \ ,
p_{lm}(\cos(\vartheta_k ) ) \ , \cos(m \varphi_k ) , \\ h_{lm } & = n_{lm } \ , \sum\limits_{n = 0}^{\infty } \ , \tilde{a}_{nl } \frac{\partial^2}{\partial r^2 } \tilde{\phi}_{nl}(r ) \ , \sum\limits_k \ , m_k \tilde{\phi}_{nl}(r_k ) \ , p_{lm}(\cos(\vartheta_k ) ) \ , \cos(m \varphi_k).\end{aligned}\ ] ] the radial deriavtive for the acceleration in @xmath113 direction becomes : @xmath114 .
\end{split}\ ] ] the radial deriavtive for the acceleration in @xmath115 direction becomes : @xmath116 .
\end{split}\ ] ] the derivative with respect to @xmath113 for the radial acceleration becomes : @xmath117.\ ] ] the derivative with respect to @xmath113 for the acceleration in @xmath113 direction becomes : @xmath118 .
\end{split}\ ] ] the derivative with respect to @xmath113 for the acceleration in @xmath115 direction becomes : @xmath119 .
\end{split}\ ] ] the derivative with respect to @xmath115 for the acceleration in radial direction becomes : @xmath120.\ ] ] the derivative with respect to @xmath115 for the acceleration in @xmath113 direction becomes : @xmath121.\ ] ] the derivative with respect to @xmath115 for the acceleration in @xmath115 direction becomes : @xmath122.\ ] ] | we follow the sinking of two massive black holes in a spherical stellar system where the black holes become bound under the influence of dynamical friction .
once bound , the binary hardens by three - body encounters with surrounding stars .
we find that the binary wanders inside the core , providing an enhanced supply of reaction partners for the hardening .
the binary evolves into a highly eccentric orbit leading to coalescence well beyond a hubble time .
these are the first results from a hybrid `` self consistent field '' ( scf ) and direct aarseth @xmath0-body integrator ( nbody6 ) , which combines the advantages of the direct force calculation with the efficiency of the field method . the code is designed for use on parallel architectures and is therefore applicable to collisional @xmath0-body integrations with extraordinarily large particle numbers ( @xmath1 ) .
this creates the possibility of simulating the dynamics of both globular clusters with realistic collisional relaxation and stellar systems surrounding supermassive black holes in galactic nuclei . |
` the world to seoul , seoul to the world . '
this was the slogan of the 1988 seoul olympics games , and is also the slogan of the korean stock market .
the globalization means that foreign traders have an influence on the korean market and its synchronization with world markets .
interdisciplinary study has received much attention , with considerable interest in applying physics to economics and finances @xcite . since
a financial market is a complex system , many researchers have developed network theory to analyze such systems .
the concept of an asset tree constructed by a minimum spanning tree is useful in investigating market properties @xcite .
the minimum spanning tree ( mst ) is derived for a unique sub - network from a fully connected network of the correlation matrix .
the mst of @xmath0 nodes has @xmath1 links ; each node represents a company or a stock and edges with the most important correlations are selected
. then clusters of companies can be identified .
the clusters , a subset of the asset tree , can be extended to portfolio optimization in practice .
the companies of the us stock market are clearly clustered into business sectors or industry categories @xcite . nowadays , many emerging markets experience the globalization that is making rapid progress , and the influence of developed markets is becoming stronger .
most markets synchronize with the us market and globalization is leading to characteristic changes in emerging markets @xcite .
several results have been reported on the necessity to find a model appropriate to emerging markets , because the models for mature markets can not be applied to emerging markets universally @xcite .
the korean market is representative of emerging markets and is subject to synchronization with external markets @xcite .
clustering in the korean market differs from that in the us market and is due to foreign factors @xcite . in this paper
, we explore characteristics of the korean stock market .
we construct the minimum spanning tree ( mst ) shifting a time window of approximately two decades and analyze the time - dependent properties of the clusters in the mst that the market conditions are not stationary .
then we investigate the market with knowledge of the history of the korean market .
the korea stock exchange ( kse ) opened in 1956 . at that time , only 12 companies were listed on the market . as the korean economy has developed , the stock market has undergone many changes under the influence of factors inside and outside the market .
we deal with the daily closure stock prices for companies listed on the kse from 4 january 1980 to 30 may 2003 .
the stock had a total of 6648 price quotes over the period .
we select 228 companies that remained in the market over this period of 23 years . fig .
[ index ] shows the index for those companies .
the representative kse index , kospi , is an index of the value - weighted average of current stock prices . the index of fig .
[ index ] is a price - equally - weighted index , similar to use for the dow jones industrial average ( djia ) .
many previous studies on the stock market assumed a certain number of trading days to constitute a year .
however , it is not easy to apply such an assumption to our data set , because the korean market opening time changed in 2000 . before 20th may 2000 ,
the market opened every day except sunday , and from monday to friday after 21th may 2000 .
most of data set falls into the former period , so we assume 300 trading days for one year .
the x - axis values in fig .
[ index ] were calculated under this assumption .
we use the logarithmic return of stock @xmath2 , which can be written as : @xmath3 where @xmath4 is the price of stock @xmath2 .
the cross - correlation coefficients between stock @xmath2 and @xmath5 are defined as : @xmath6 and form a correlation matrix @xmath7 .
.industry categories of the korea stock exchange in our data set [ cols="<,<,<",options="header " , ] the top panel of fig .
[ correlation ] shows the mean correlation coefficient calculated with only non - diagonal elements of @xmath7 .
the second shows the standard deviation , the third , the skewness and the last , the kurtosis .
it has been reported that when the market crashes , the correlation coefficient is higher @xcite . in the us market ,
the effect of black monday ( 19 october 1987 ) was clearly visible for these four coefficients , with correlations among them also apparent @xcite . however , crash effects on the korean market ( the late 1980s bubble crash and the 1997 asian financial crisis ) are visible , but not clear in comparison with the us market , and the korean market coefficients do not have clear correlations .
we investigate more properties of the market through the mst that is a simple graph with @xmath0 nodes and @xmath1 links .
the most important connection is linked when it is constructed .
it is known that the us market network is centralized to a few nodes @xcite .
the hub of the us market is approximately general electric ( ge ) , and it is possible to make clusters ( subsets of the mst ) of the us market with industry categories or business sectors @xcite .
however , the korean market has no comparable hub for the whole market , and the clusters are constructed with the msci index @xcite .
we regard this result as the effect of globalization and market integration .
thus , we obtained the msts from 1980 to 2003 with time windows of width @xmath8 corresponding to daily data for @xmath8=900 days and @xmath9=20 days . during this period
there is no comparable hub , but we can form clusters with industry categories for some periods .
then we define the parameter _
grouping coefficient_. the grouping coefficient of a specified industry category @xmath10 is defined as : @xmath11 where @xmath12 represents the nodes in category @xmath10 , @xmath13 is the number of links that are connected to node @xmath2 and @xmath14 is the number of links from the node included in category @xmath10 .
table [ category ] shows 25 industry categories of the korean stock market .
in fact , there are rather more categories than 25 . however , the standard for grouping is excessively detailed , and combinations of these categories are mostly used .
the categories in table [ category ] are reconstructed from hankyoreh , a popular newspaper in korea .
fig . [ group]a shows the _ grouping coefficient _ for each category over the whole period .
we observe that categories 8 , 18 , 21 , 22 and 23 form a well - defined cluster .
we focus on the maximum grouping coefficient for each industry category .
for example , there are only four companies in the finance category ( 23 ) and the maximum value of the coefficient is only 0.6 ( = 3/5 ) because of the characteristics of the mst .
we take the maximum value when the nodes are linked linearly .
[ group]b shows the ratio of the grouping coefficient to the maximum value for each category .
categories 18 , 21 , 22 and 23 are almost complete clusters in this plot . and ( b ) shows the ratio of the coefficient to the maximum value of @xmath15.,width=302 ] we previously investigated characteristics of the korean stock market @xcite using a data set from 2001 to 2004 , and found that the market forms clusters when the morgan stanley capital international ( msci ) index is exploited . however , fig .
[ group ] shows that some industry categories can be applied to form the clusters .
we consider the history of the korean market , including the globalization effect .
[ grouptime ] shows the grouping coefficient @xmath16 for the whole market as a function of time .
this coefficient is calculated with all of the nodes , and the ratio of connections between companies in the same category to the total number of links . before the mid-1980s
, the korean market had developed according to a planned economy and had many restrictions on trading of stocks . at that time , the market was unstable because of the poor liquidity .
this is one possible explanation for lower value in the early 1980s in fig .
[ grouptime ] . as the market prospered in the mid-1980s , clusters of industry categories
also extensively formed .
the 1988 seoul olympics games and the 1997 asian financial crisis hastened globalization of the korean market . in particular , globalization of the korean market progressed to synchronization with external markets .
this explains the decreasing coefficient in fig .
[ grouptime ] after 1988 .
the index continues to show a decreasing trend , which means that the formation of clusters in the korean market is related to the msci index .
the msci korea index has been calculated from 1988 , when the grouping coefficient in fig .
[ grouptime ] has almost a maximum value .
the msci korea index is a factor of the korean market s globalization and market integration .
because foreign traders strongly influence the korean market , the msci index is a good reference for their trading @xcite .
we have studied the korean stock market network with the daily closure stock price .
the analysis shows that the grouping coefficient changes with elapsing time . with globalization ,
the market is synchronized to external markets , and the number of clusters of industry categories decreases . finally , the market forms clusters according to the msci index .
we think the tendency of synchronization will be stronger and the clusters of the msci index or foreign factors will be firmer .
our future research will determine other properties of the globalization effect with other statistical analysis in the korean market .
r. n. mantegna , and h. e. stanley , _ an introduction to econophysics : correlations and complexity in finance _
( cambridge university press , cambridge , 1999 ) . w. b. arthur , s. n. durlauf , and d. a. lane , _ the economy as an evolving complex system ii _
( perserus book , massachusetts , 1997 ) . j .-
bouchaud , and m. potters , _ theory of financial risks _ ( cambridge university press , cambridge , 2000 ) .
t. aste , t. di matteo , and s. t. hyde , physica a * 346 * , 20 ( 2005 ) .
t. kaizoji , and m. kaizoji , physica a * 344 * , 240 ( 2005 ) .
r. n. mantegna , eur .
j. b * 11 * , 193 ( 1999 ) .
g. bonanno , g. caldarelli , f. lillo , and r. n. mantegna , phys .
e * 68 * , 046130 ( 2003 ) .
m. mcdonald , o. suleman , s. williams , s. howison , and n. f. johnson , e - print arxiv : cond - mat/0412411 .
onnela , a. chakraborti , k. kaski , j. kertsz , and a. kanto , phys .
e * 68 * , 056110 ( 2003 ) .
f. climent , and v. meneu , int .
finance * 12 * , 111 ( 2003 ) .
k. matia , m. pal , h. salunkay , and h. e. stanley , europhys . lett . * 66 * , 909 ( 2004 ) .
yoon , j. s. choi , k. kim , and y. kim , j. korean phys .
soc . * 46 * , 719 ( 2005 ) .
k. e. lee , and j. w. lee , j. korean phys . soc . *
46 * , 726 ( 2005 ) . s .-
yoon , k. kim , and j. s. choi , j. korean phys .
soc . * 46 * , 1071 ( 2005 ) .
k. e. lee , and j. w. lee , j. korean phys . soc . *
47 * , 185 ( 2005 ) .
j. s. choi , k. kim , s. m. yoon , k. h. chang , and c. c. lee , j. korean phys .
* 47 * , 171 ( 2005 ) .
jung , s. chae , j .- s .
yang , and h .-
moon , to be published in physica a , doi:10.1016/j.physa.2005.06.081 .
s. drozdz , f. grummer , a. z. gorski , f. ruf , and j. speth , physica a * 287 * , 440 ( 2000 ) .
kim , y. lee , b. kahng , and i. kim , j. phys .
* 71 * , 2133 ( 2002 ) . | we study the effect of globalization on the korean market , one of the emerging markets .
some characteristics of the korean market are different from those of the mature market according to the latest market data , and this is due to the influence of foreign markets or investors .
we concentrate on the market network structures over the past two decades with knowledge of the history of the market , and determine the globalization effect and market integration as a function of time . |
rapidly evolving coherent scalar fields have been widely studied in cosmology .
their dynamics is not only important during the reheating epoch after inflation , but they can also support periods of accelerated expansion in both the early universe @xcite or at late times @xcite . concerning the dark matter problem , non - thermal candidates like the axion @xcite or other massive scalar @xcite or pseudoscalar fields @xcite also fall in this class .
these models can be interpreted as bose - einstein condensates , where the scalar particles occupy the lowest quantum state of the potential @xcite . finally , the possibility of ultra - light scalar fields as dark matter candidates has been explored in different works @xcite by tuning appropriately the potential and initial conditions @xcite .
the general analysis of a homogeneous oscillating scalar field in an expanding universe was performed by turner in @xcite . for a power - law potential @xmath0 , the rapid scalar oscillations around the minimum of such a potential behave as a perfect fluid with an effective equation of state @xmath7 .
his results can be recovered by means of a generalization of the virial theorem @xcite .
recently , it has been shown that a fast oscillating abelian vector @xcite , non - abelian vector @xcite or arbitrary spin field @xcite will behave in a very similar way .
the purpose of this work is to analyse the growth of perturbations in these coherent oscillating scalar theories for arbitrary power law potential .
this subject has been mainly studied for harmonic potential models that mimic the standard dark matter case @xcite , as it happens for the axion field @xcite .
it has been proved by using the linear perturbation theory that the axion was equivalent to cdm for high enough masses @xcite . however , gravitational instabilities of oscillations in a harmonic potential are suppressed on small scales @xcite .
this analysis determines the cut - off in the matter power spectrum and its deviations with respect to the cdm phenomenology . on the other hand ,
the dynamical stability ( ignoring metric perturbations ) of general coherent oscillating scalar dark energy models has been analysed in different works @xcite , even by considering nonlinear evolutions @xcite .
they conclude that potentials supporting accelerated expansion are generically unstable with respect to the growth of inhomogeneities .
this work is organized as follows : we will briefly review the standard average approach for the background evolution of a scalar field under a power law potential ( section ii ) , as well as set the equations that rule its perturbations ( section iii ) . after the preliminary discussion , we will analyse the well - known case of a massive scalar by means of an adiabatic expansion approach ( section iv )
. the perturbations evolution of power - law potential models will be studied following the average approach ( section v ) .
firstly , we will compute the effective sound speed , which is in general the quantity that rules the evolution , using the perturbed version of the generalized virial theorem .
this method allows to extend previous results to an arbitrary power - law potential .
also , exploiting this equation , we will be able to derive a general expression for a possible anharmonic correction in a massive scalar theory .
after that ( section vi ) , we will check the validity of the result for the effective sound speed by studying the exact system of equations ( non - averaged ) in both super - hubble and sub - hubble limits . for small wavenumbers
, we will show what is the natural ansatz for @xmath8 ( subsection a ) , whereas in the sub - hubble limit we will study the exact solution of the non expanding equations thanks to floquet s theorem ( subsection b ) .
finally , we will see that at high @xmath4 , a cut - off is always expected ( subsection c ) .
we are interested in studying the cosmological evolution of a homogeneous scalar field which is rapidly oscillating around the potential minimum .
let us then consider a scalar field theory in cuved space - time with lagrangian @xmath9 the equation of motion can be written as @xmath10 where @xmath11 represents the derivative with respect to its argument . by considering a friedmann - lematre - robertson - walker metric in conformal time @xmath12 , the equation of motion takes the form @xmath13 where @xmath14 and @xmath15 .
the corresponding energy - momentum tensor reads @xmath16 following the same approach as in previous works @xcite , we are interested in the cosmological evolution generated by the effective energy - momentum tensor obtained after averaging over the fast scalar oscillations .
thus , we will concentrate on the average einstein equations given by @xmath17 in the particular case of a homogeneous scalar field in flat rw background , they reduce to @xmath18 in order to obtain the average equation of state of the oscillating scalar , we apply a generalization of the virial theorem @xcite .
let us consider that the typical frequency of the @xmath19 oscillations is @xmath20 and let us calculate the average in a certain time interval @xmath21 such that it is large compared to the oscillation period but small compared to the hubble time @xcite , i.e. , @xmath22 .
we start by calculating the average of the total derivative given by @xmath23 .
thus , if the @xmath19 oscillations are bounded @xmath24 thus , comparing with @xmath25 , we see that ( [ bounded ] ) is suppressed by a factor @xmath26 . using ( [ eqphi ] ) and neglecting @xmath27 terms
, we can write @xmath28 the error introduced by neglecting the total derivative can be reduced by taking large @xmath21 , so that the minimum limit is set by @xmath29 by using these equations , we can reach a useful expression for the the average equation of state : @xmath30 considering a power - law potential @xmath31 , the last expression results @xmath32 ( see @xcite for an alternative discussion ) .
therefore , from the conservation equation @xmath33 we can show that the evolution of the average energy density is @xmath34 so that from the friedmann equation ( [ eins00 ] ) we get : @xmath35
let us consider now a perturbation on the homogeneous evolution studied above for @xmath19 : @xmath36 with @xmath6 a small perturbation .
the only sourced metric perturbations are scalars . therefore we can write @xmath37 where we have chosen the longitudinal gauge for the computation . by taking the fourier transformation in the spatial coordinates , the equation of motion for the perturbation results @xmath38 and , following the same approach as for the background ,
the average einstein equations ( [ einsav ] ) read to first order in perturbations , @xmath39 where the perturbed energy - momentum tensor components read @xmath40 and the effective sound speed is : @xmath41 in the following section , we will analyse this system for the simple case of a massive scalar . by making an adiabatic expansion ,
we will obtain the background solution of @xmath19 and compute the averages of eqs .
( [ eins00])-([einsii ] ) explicitly , together with the corresponding effective sound speed .
homogeneous scalars are widely considered in cosmology , particularly fast oscillating massive scalars are specially relevant since its perturbations mimic those of dust perfect fluids .
this fact makes them a good candidate for solving the dark matter problem . assuming a quadratic potential ( @xmath42 ) and redefining the field @xmath43 in ( [ eqphi ] ) , the equation of motion turns into @xmath44 making an adiabatic expansion @xcite with @xmath45 , the solution to the leading adiabatic order results @xmath46 with @xmath47 and @xmath48 integration constants , and @xmath49 setting the origin of time adequately , the background field reads @xmath50\;.\nonumber\\\end{aligned}\ ] ] the average energy - momentum tensor is equivalent to a dust perfect fluid ( see eq .
( [ omega ] ) or @xcite ) @xmath51 the problem of the perturbations of the massive oscillating scalar has been previously studied in @xcite . in our case , in order to solve the system , we will combine equations ( [ peins00 ] ) and ( [ peins0i ] ) , obtaining an equation which together with ( [ peinsij ] ) , form an algebraic system for @xmath52 and @xmath53 as a function of @xmath19 and @xmath8 .
@xmath54 notice that the origin of the error comes from the use of the virial theorem ( [ virial ] ) and ( [ eins00 ] ) to pass the term @xmath55 to the left - hand side on ( [ peins00 ] ) .
this expression gives @xmath53 as a function of the field and its perturbation , thus , using ( [ adphi ] ) we can obtain the solution of ( [ eqdphi ] ) for @xmath56 in an adiabatic expansion . as we have just seen the background equations of motion
have been solved using a wkb approximation ( [ adphi ] ) which can be written as an expansion in the so called adiabatic parameter @xmath57 . in order to solve the perturbation equations we can similarly use this method . in this case
, there is a new scale , @xmath4 , independent of the other two : @xmath58 and @xmath59 .
therefore , we have two different expansion parameters : the same @xmath60 as in the background case and the new parameter @xmath61 . in order to simplify the calculations and work with a single parameter
, we will assume that the new parameter @xmath61 is related to @xmath62 by @xmath63 ) .
thus we will work in different regimes by assigning different values to the exponent @xmath64 .
thus for example , as we are interested in cosmological perturbations , we typically expect @xmath65 , i.e. both @xmath58 and @xmath4 of the same adiabatic order , or in other words , @xmath66 .
however , as we will see below , this is not the only interesting range in @xmath4 .
for example , we will show that when @xmath67 , the behaviour of perturbations changes .
in such a case , @xmath68 .
finally we will also solve the system for scales @xmath69 , i.e. , @xmath70 .
we will proceed as in the background by assuming an adiabatic ansatz for the field perturbation , @xmath72 the amplitudes can be expanded in the adiabatic parameter , @xmath73 where @xmath74 and @xmath75 are of adiabatic order @xmath76 and @xmath62 respectively . in the standard case @xmath77 ( @xmath78 ) , we have access to super - hubble and sub - hubble modes provided @xmath79 . from the leading order of the equation of motion ( [ eqdphi ] ) , we obtain @xmath80 and once we take this constraint into account , the equation implies @xmath81 by solving this system , we can easily compute the perturbations to the leading order , @xmath82 as it can be seen the solution mimics the perturbations of dust perfect fluids , i.e. it has vanishing effective sound speed , constant @xmath52 and @xmath83 at late times .
the case @xmath84 corresponds to super - hubble modes and has the same solutions ( [ deltaphisca])-([deltapsca ] ) just neglecting @xmath4 in comparison with @xmath58 .
let us first consider the @xmath86 ( @xmath87 ) case .
expanding the equations of motion in @xmath62 we obtain the system @xmath88 with solution @xmath89 \;,\end{aligned}\ ] ] where @xmath90 and @xmath91 are integration constants .
the cosmological perturbations read @xmath92 + \mathcal{o}(\epsilon)\;,\label{psiscak } \\
\\ \left\langle\delta \rho_k\right\rangle & = & - \frac{2}{3 } \frac{k^2 \psi_k}{a^2}+ \mathcal{o}(\epsilon)\ ; , \\ \nonumber \\
c_{\text{eff}}^2 & = & \frac{k^2}{4 m^2 a^2 } + \mathcal{o}(\epsilon)\;.\label{ceffscak}\end{aligned}\ ] ] in this case @xmath52 oscillates slowly compared to @xmath5 with a decaying amplitude at early times and is constant at late times . on the other hand @xmath93
also oscillates with decaying amplitude at early times and grows as @xmath94 at late times thus matching the dusty behaviour . for modes with @xmath95 ,
the expressions are equivalent to ( [ psiscak])-([ceffscak ] ) neglecting the factors that are a power of @xmath96 .
finally , for modes with @xmath97 , @xmath8 oscillates with a higher frequency than @xmath19 and all perturbations go to zero in average .
consequently , a cut - off is expected as we consider larger-@xmath4 modes .
in general for a power - law potential , the background field equation is non linear and we can not obtain the explicit adiabatic expansion of the solutions as in the quadratic case . we can still define a typical frequency of the background oscillations @xmath5 and compare it with the other two scales in the problem @xmath58 and @xmath4 .
moreover , in this case , the pressure is not negligible in comparison with the energy density .
this fact makes easy to compute the effective sound speed @xmath98 , which is the quantity that controls the evolution of sub - hubble perturbations . combining ( [ peins00 ] ) and ( [ peinsii ] )
, we get a single equation for the metric perturbation : @xmath99 \psi_k = 0\;. \label{ceffprop}\end{aligned}\ ] ] if @xmath19 dominates the background energy density , ( [ ceffprop ] ) can be rewritten as @xmath100 where here , @xmath101 and @xmath2 correspond to the average background evolution . in order to compute the effective sound speed , let us consider the following average in a period @xmath21 , which verifies @xmath102 : @xmath103 if the field evolution is periodic or bounded , averaging during long enough periods , the left - hand term of the last equation will be negligible in comparison with the average of @xmath104 . in principle , except for a quadratic potential , @xmath6 will have some growing modes .
those modes make @xmath105 not to vanish , but as long as it oscillates around zero , the following discussion holds at leading order . focusing on the first order of perturbations and introducing the equation of motion ( [ eqdphi ] ) : @xmath106 where we have neglected the scale factor derivatives as we are considering time intervals much smaller than the inverse of the expansion rate . by using ( [ pvirial ] ) , we obtain the following expression for the effective speed of sound @xmath107 in the second equation the terms proportional to @xmath53 are averaged out as their fast oscillating part can be expressed as a total derivative or coincide with the background generalization of the virial theorem ( [ virial ] ) . if a power - law potential , @xmath31 , is considered : * when @xmath108 , i.e. the frequency of @xmath19 is almost equal to the perturbation frequency , the behaviour is similar to a perfect fluid with constant equation of state ( see fig .
[ figsubh ] for the case @xmath42 and fig .
[ eceff4 ] for @xmath109 ) , i.e. up to @xmath110 : @xmath111 notice that for @xmath112 , there is an instability in agreement with @xcite . in this case
@xmath113 is not well defined when @xmath114 . * for massive scalar fields @xmath115 , @xmath116 reproducing the result obtained in @xcite : @xmath117 although this expression seems to be valid for all @xmath4 , when @xmath4 is comparable with @xmath59 fails since the averaging time interval has the same order of the @xmath118 oscillation period .
see curve ( c ) in fig .
[ eceff2 ] at early times .
the high @xmath4 limit will be discussed in section viii .
+ the corresponding comoving jeans length @xmath119 , which satisfies @xmath120 reads : @xmath121 * because of the vanishing of the potential terms in the numerator of ( [ cseff ] ) in the harmonic case , we can also consider anharmonic corrections , @xmath122 in a simple way . due to the averaging process the sine mode of @xmath56 gives a negligible contribution : @xmath123 the same situation occurs with the term @xmath124 .
therefore , only the cosine mode contributes and we will be able to extract the common factor @xmath125 both in the numerator and the denominator and simplify the expression
. we will only consider even powers @xmath126 since odd exponents make the sinusoidal resulting function to oscillate around zero and are suppressed by the average .
finally , @xmath127 for example , for the case @xmath128 , we recover the sound speed obtained in @xcite .
in order to check the validity of the results obtained above from the averaged equations , in this section we will compare them with exact results in certain particular limits .
let us then consider the einstein equations with the exact ( non - averaged ) energy - momentum tensor . at the background level we have : @xmath129 whereas to first order in perturbations they read , @xmath130 where the expression for the perturbed energy - momentum tensor components are given in ( [ de00 ] ) , ( [ deii ] ) and ( [ de0i ] ) . notice that in this section all the geometric quantities are not averaged so that in order to compare with the result of the previous section we will explicitly take the corresponding average of the obtained results .
in the super - hubble limit @xmath131 , we can neglect the terms proportional to @xmath4 in ( [ e00 ] ) , and introducing ( [ e0i ] ) in ( [ e00 ] ) , we obtain @xmath132 it is interesting to note that any time the oscillating @xmath133 is zero , the perturbation @xmath134 must be zero too .
therefore , it is natural to assume the ansatz @xmath135 . when it is substituted in the previous equation and by using eq .
( [ eqphi ] ) , we can write the metric perturbation as : @xmath136 if we take into account ( [ e0i ] ) , we can obtain the following equation for @xmath137 : @xmath138 by combining equations ( [ exeins00 ] ) and ( [ exeinsii ] ) , we can integrate the last equation obtaining : @xmath139 where @xmath140 is an integration constant . solving this equation
we obtain @xmath141 where @xmath142 and @xmath143 are integration constants . substituting in ( [ psish ] ) @xmath144 as it can be seen in fig .
[ figsuperh ] for the case of a quadratic potential , @xmath53 does not oscillate in an important way .
on the other hand , computing the energy density contrast : @xmath145 notice that since @xmath146 oscillates around a non - vanishing value , the same behaviour is expected for @xmath147 .
this agrees with the numerical result in fig .
[ figsuperh ] .
finally , it can be seen that averaging the previous expression , we find a perfect agreement with the results in previous section .
thus , substituting ( [ a ] ) in ( [ psiex ] ) and ( [ fex ] ) , and averaging in ( [ deltaex ] ) we find that at late times : @xmath148 where the @xmath110 error comes from the average in @xmath149 .
this result agrees with the standard expression for super - hubble modes in perfect fluid cosmologies with constant equation of state .
notice that ( [ psisuperh ] ) is in good concordance with the numerical solution shown in fig .
[ figsuperh ] . in this section
, we will consider a different approach to check the results of previous section .
this approach is valid for sub - hubble modes of power law potential theories with @xmath150 .
fortunately , as we shown , the @xmath115 case is well understood .
using ( [ eqphi ] ) and ( [ eqdphi ] ) , we can write the perturbation of the energy density and pressure as , @xmath151 where @xmath152 , @xmath153 . multiplying ( [ e00 ] ) by the expected effective sound speed @xmath154 and subtracting it from ( [ eii ] ) : @xmath155 by using ( [ deltarho ] ) and
( [ deltap ] ) , @xmath156 notice that the order of the @xmath19 amplitude can be estimated through ( [ eins00 ] ) , @xmath157 thus , @xmath158 so eq . ( [ eqjo ] ) can be approximated by @xmath159 moreover , in time intervals @xmath160 , we can write approximately @xmath161 the solution of this equation is decomposed in a solution for the homogeneous equation , @xmath162 and a particular solution for the inhomogeneous one . in order to obtain it
let us take the fourier transform of ( [ ecminpsi ] ) , @xmath163 where @xmath164 denotes the convolution operation . in the limit @xmath165
, @xmath166 oscillates with @xmath5 and , thus , @xmath167 by taking the inverse fourier transform , we reach the general solution , @xmath168 where @xmath169 .
as it can be seen , @xmath170 , however for sub - hubble modes , @xmath171 and , consequently , the particular solution can be neglected .
in addition , the source on ( [ exacddpsi ] ) can also be neglected , thus obtaining the same approximated equation ( [ ceffprop ] ) of the effective model .
a more rigorous analysis can be done thanks to floquet s theorem . within time
intervals @xmath172 , ( [ eqdphi ] ) takes the form , @xmath173 with @xmath174 and @xmath175 .
notice that as @xmath19 is periodic , so is @xmath176 and the floquet s theorem gives us the general form of the solution of ( [ ecpertflo ] ) .
following @xcite , let us define @xmath177 and @xmath178 as the normalized solutions corresponding to the initial conditions : @xmath179 , @xmath180 and @xmath181 , @xmath182 .
the corresponding characteristic equation reads @xmath183 \tau + 1 = 0\;,\end{aligned}\ ] ] whose associated roots can be written as @xmath184 and the floquet s theorem implies : 1 . if @xmath185 , ( [ ecpertflo ] ) has two linearly independent solutions @xmath186 where @xmath187 and @xmath188 are @xmath189-periodic functions . 2 . if @xmath190 , ( [ ecpertflo ] ) has a periodic solution with period @xmath189 ( @xmath191 ) or @xmath192 ( @xmath193 ) .
denoting by @xmath194 this one and by @xmath195 another linearly independent solution : @xmath196 where @xmath197 is a constant . as an example
, let us consider the quartic potential case @xmath198 , @xmath199\;;\;z = \sqrt{\frac{\lambda}{2}}\phi_0 \eta + \delta\ ; ; \\ \ddot{\delta\phi}_k & + & \left ( 6\ ; \text{sn}[z , i]^2 + \epsilon_k^2 \right ) \delta\phi_k = 0\;;\;\ ; \epsilon_k \equiv \frac{2 k^2}{\lambda \phi_0 ^ 2}\;,\label{pfloquet4}\end{aligned}\ ] ] with @xmath200 and @xmath201 integration constants and @xmath202 $ ] the corresponding jacobi elliptic function . if @xmath203 , @xmath204 \right)+\delta\phi_{20 } \frac{d}{dz}\left ( z \cdot \text{sn}[z , i ] \right)\;. \label{dphi4k0}\end{aligned}\ ] ] notice that , as jacobi ellliptic functions are periodic with period @xmath205 , @xmath206 making the change @xmath207 , we can see that this solution corresponds to the @xmath208 case of the floquet s theorem : @xmath209 \right )
\nonumber \\ & + & \delta\phi_{20 } \left ( \frac{\pi}{4 k } \frac{d}{dx}\left ( \frac{4 k x}{\pi } \text{sn}\left[\frac{4 k x}{\pi},i \right ] \right)\right . \nonumber \\ & + & \left . 4 k \frac{d}{dx}\left ( \text{sn}\left[\frac{4 k x}{\pi},i\right ] \right ) \right)\;,\end{aligned}\ ] ] by comparing with ( [ floquet2 ] ) .
however , in general , when @xmath210 , @xmath185 and it can be seen that for modes with @xmath211 , @xmath8 is an oscillating function modulated by a periodic function of frequency @xmath212 . in fig .
[ figfloquet ] an explicit numerical calaculation shows how the number of nodes is multiplied by a factor of 10 when @xmath4 is increased from @xmath213 to @xmath214 . for a general power law potential , if @xmath215 , the analogous solution to ( [ dphi4k0 ] ) can be written as @xmath216 which also corresponds to the second case of floquet s theorem . for @xmath217
, @xmath134 behaves analogously to the @xmath218 case and , thus , we will assume that @xmath56 has the form : @xmath219 as @xmath220 , solutions can be divided in @xcite : 1 .
stable solutions with @xmath221 : @xmath222 2 .
unstable solutions with @xmath223 and @xmath224 : @xmath225 notice that @xmath226 and @xmath227 are real @xmath189-periodic functions .
we can synthesise both cases in a single expression : @xmath228 where @xmath229 and @xmath230 are @xmath189-periodic functions .
we consider @xmath231 , and @xmath232 for stable modes , which are those in which we are interested . due to the periodic properties of @xmath19 and @xmath134
, their product can be expanded in fourier series as , @xmath233 from this series , we can obtain a simple expression for @xmath234 , @xmath235 we can obtain the general solution of ( [ ecminpsi ] ) , @xmath236 with @xmath237 notice that in the limit in which we are interested ( @xmath238 ) , @xmath239 .
therefore the particular solution is at leading order equal to @xmath166 and is negligible in comparison with the homogeneous one , in agreement with our preliminary analysis ( [ psiorder ] ) . from the general solution ( [ analyticpsi ] )
, we can obtain an expression for the effective sound speed up to @xmath110 : @xmath240 } { \psi_{0k } \cos \left(\frac{2 k}{\omega_{\text{eff}}}c_{\text{eff}}z+\delta_{0k}\right)}}{1 + \frac{2 \left[\text{re}(b_{0k})\cdot \cos(\alpha_k z ) - \text{im}(b_{0k})\cdot \sin(\alpha_k z)\right ] } { \psi_{0k } \cos \left(\frac{2 k}{\omega_{\text{eff}}}c_{\text{eff}}z+\delta_{0k}\right ) } } \;.\ ; \label{ceffcorr } \end{aligned}\ ] ] where we have considered only the @xmath241 mode in ( [ analyticpsi ] ) since the @xmath242 modes vanish when taking the average .
this expression depends not only on @xmath243 but also on the initial conditions given by @xmath244 and @xmath245 .
however , from ( [ psiorder ] ) we know that @xmath246 . on the other hand we have to determine the size of the parameter @xmath243 . with this purpose
, we will follow the discussion made in @xcite writing @xmath176 and a possible solution @xmath177 as fourier expansions : @xmath247 we introduce the tentative solution @xmath177 in ( [ eqfloq ] ) , resulting the following system : @xmath248 we eliminate @xmath249 , obtaining the determinant @xmath250 and an equation for @xmath243 , @xmath251 .
this equation can be rewritten as @xmath252 we know the solution when @xmath253 , where one of the modes is periodic .
consequently @xmath254 .
notice that we have also checked this behaviour numerically in fig .
[ figfloquet ] . expanding ( [ eqdelta ] ) in @xmath243
, @xmath255 even if we can not compute @xmath256 analytically for @xmath257 , we know that expanding in @xmath258 the first correction of the matrix elements is @xmath259 . thus from equation ( [ eqdelta2 ] ) , we expect that @xmath260 .
therefore , the correction to the effective sound speed given by ( [ ceffcorr ] ) is : @xmath261 which is a generalization of ( [ csk ] ) for @xmath262 .
we have seen that for @xmath263 an oscillating scalar field with power - law potential behaves as a perfect fluid for which @xmath264 .
let us now consider the opposite limit with @xmath265 . for sub - hubble modes @xmath53 is well approximated by @xmath266 in this case the field perturbation oscillates much faster than the background field .
the typical frequency of @xmath134 oscillations would be @xmath4 @xcite , thus @xmath267 we can also reach an equivalent expression using ( [ analyticpsi ] ) . in this limit , we can neglect @xmath113 from eq .
( [ ecpertflo ] ) in comparison with @xmath268 .
it implies @xmath269 .
if we assume that the lowest @xmath270 coefficients are responsible of the main contribution to the fourier expansion @xmath271 and , thus , @xmath272 the gravitational potential , and accordingly the density perturbation , oscillates around zero as shown in fig .
[ figlargek ] .
because of this fact , all the perturbations vanish in average but the effective sound speed is @xmath273 ( see fig . [ figlargek ] ) according to ( [ cseff ] ) , since in both @xmath274 and @xmath275 , the kinetic term dominates .
in this work we have shown that a coherent homogeneous scalar field oscillating in a power - law potential behaves as an adiabatic perfect fluid with constant equation of state both at the background and perturbation levels .
thus , scalar perturbations are shown to propagate , to the leading order in @xmath276 , with a sound speed given by @xmath1 . the first correction to this expression
is shown to be @xmath277 .
the robustness of the result has been shown by studying the exact system in the sub - hubble and super - hubble limits as well as in the numerical computations .
these results extend previous analysis done in the massive case @xmath42 and opens the possibility of using this kind of models as perfect fluid analogues in different cosmological contexts . in the general case
, we have shown that there are departures from the perfect fluid behaviour on small scales with a cut - off around @xmath278 very much as in the harmonic case .
notice that for @xmath112 the negative value of @xmath279 suggests the generation of instabilities as found in previous works in the homogeneous case .
the analysis performed in this paper could be extended to higher - spin oscillating fields .
in particular in the massive vector case , which as shown in @xcite behaves as non - relativistic matter at the background level , this study would allow to determine the growth of structures and its viability as dark matter candidate .
work is in progress in this direction @xcite .
* acknowledgements : * this work has been supported by micinn ( spain ) project numbers fis2011 - 23000 , fpa2011 - 27853 - 01 , fis2014 - 52837-p and consolider - ingenio multidark csd2009 - 00064 .
acknowledge support from complutense university under grant ct4/14 .
we would like to thank alberto diez tejedor for his useful comments .
t. damour and v. f. mukhanov , phys .
lett . * 80 * , 3440 ( 1998 ) ; a. r. liddle and a. mazumdar , phys .
d * 58 * , 083508 ( 1998 ) ; j. w. lee , s. koh , c. park , s. j. sin and c. h. lee , phys . rev .
d * 61 * , 027301 ( 2000 ) ; v. h. cardenas and g. palma , phys .
d * 61 * , 027302 ( 2000 ) ; m. sami , grav .
cosmol .
* 8 * , 309 ( 2003 ) .
a. r. liddle and r. j. scherrer , phys .
d * 59 * ( 1999 ) 023509 ; j. a. gu , arxiv:0711.3606 [ astro - ph ] ; s. d. h. hsu , phys . lett .
b * 567 * , 9 ( 2003 ) ; s. dutta and r. j. scherrer , phys .
d * 78 * , 083512 ( 2008 ) ; m. sami and t. padmanabhan , phys . rev .
d * 67 * , 083509 ( 2003 ) [ erratum - ibid .
d * 67 * , 109901 ( 2003 ) ] [ arxiv : hep - th/0212317 ] . v. sahni and l. m. wang , phys .
d * 62 * , 103517 ( 2000 ) ; e. masso , f. rota and g. zsembinszki , phys .
d * 72 * , 084007 ( 2005 ) ; r.d .
peccei , h.r .
quinn , phys .
* 38 * ( 1977 ) 1440 ; phys . rev .
* d16 * ( 1977 ) 1791 ; s. weinberg , phys .
* 40 * ( 1978 ) 223 ; f. wilczek , phys . rev . lett .
* 40 * ( 1978 ) 279 ; j. preskill , m.b .
wise , f. wilczek , phys .
b120 * ( 1983 ) 127 ; l.f .
abbott , p. sikivie , phys . lett .
* b120 * ( 1983 ) 133 ; m. dine , w. fischler , phys .
* b120 * ( 1983 ) 137 .
turner , , 1243 ( 1983 ) ; j.e .
kim , phys .
150 * , 1 ( 1987 ) . b. de carlos , j. a. casas , f. quevedo and e. roulet , phys .
b * 318 * , 447 ( 1993 ) ; m. gasperini and g. veneziano , phys .
d * 50 * , 2519 ( 1994 ) ; j. a. r. cembranos , phys .
lett . * 102 * , 141301 ( 2009 ) ; j. phys
. conf .
* 315 * , 012004 ( 2011 ) .
a. dobado and a. l. maroto , nucl .
b * 592 * , 203 ( 2001 ) ; j. alcaraz _ et al .
rev.*d67 * , 075010 ( 2003 ) ; j. a. r. cembranos , a. dobado and a. l. maroto , phys . rev . *
d65 * 026005 ( 2002 ) ; phys . rev
. lett .
* 90 * , 241301 ( 2003 ) ; phys .
d * 68 * , 103505 ( 2003 ) ; a. l. maroto , phys . rev .
d * 69 * , 043509 ( 2004 ) ; phys .
d * 69 * , 101304 ( 2004 ) .
sin , , 3650 ( 1994 ) ; s.u .
ji and s.j .
sin , , 3655 ( 1994 ) ; j .- w .
lee and i .-
koh , , 2236 ( 1996 ) ; w. hu , astrophys .
j. * 506 * , 485 ( 1998 ) ; f. ferrer and j.a .
grifols , j. cosmol .
astropart .
0412 * , 012 ( 2004 ) ; c.g . bhmer and t. harko , j. cosmol .
astropart . phys .
* 0706 * , 025 ( 2007 ) ; l.a .
urea - lpez , j. cosmol .
astropart .
phys . * 0901 * , 014 ( 2009 ) ; j .- w .
lee , j. korean phys .
* 54 * , 2622 ( 2009 ) ; j .- w .
lee and s. lim , j. cosmol .
astropart .
1001 * , 007 ( 2010 ) ; a.p .
lundgren , m. bondarescu , r. bondarescu and j. balakrishna , astrophys .
j. * 715 * , l35 ( 2010 ) ; i. rodrguez - montoya , j. magaa , t. matos , and a. prez - lorenzana , astrophys .
j. * 721 * , 1509 ( 2010 ) .
t. harko , , 123515 ( 2011 ) ; t. harko , j. cosmol .
astropart .
phys . * 1105 * , 022 ( 2011 ) ; p .- h .
chavanis , , 043531 ( 2011 ) ; p .- h .
chavanis and l. delfini , , 043532 ( 2011 ) ; v. lora , j. magaa , a. bernal , f.j .
snchez - salcedo , and e.k .
grebel , j. cosmol .
astropart .
1202 * , 011 ( 2012 ) ; p.j.e .
peebles and a. vilenkin , , 103506 ( 1999 ) ; p.j.e .
peebles , astrophys .
j. * 534 * , l127 ( 2000 ) ; t. matos , f.s .
guzmn and l.a .
urea - lpez , class . quant .
* 17 * , 1707 ( 2000 ) ; a. arbey , j. lesgourgues , and p. salati , , 123528 ( 2001 ) ; j. magaa , t. matos , v.h .
robles , and a. surez , arxiv:1201.6107v1 ( 2012 ) .
t. matos and l.a .
urea - lpez , class .
quantum grav .
* 17 * , l75 ( 2000 ) ; t. matos and l.a .
urea - lpez , , 063506 ( 2001 ) ; t. matos , j .-
luvano , i. quiros , l.a .
urea - lpez , and j.a .
vzquez , , 123521 ( 2009 ) ; a. surez and t. matos , mon . not .
r. astron .
416 * , 87 ( 2011 ) ; j. magaa , t. matos , a. surez , f.j .
snchez - salcedo , arxiv:1204.5255v1 ( 2012 ) .
j. a. r. cembranos , c. hallabrin , a. l. maroto and s. j. n. jareno , phys .
d * 86 * , 021301 ( 2012 ) .
j. a. r. cembranos , a. l. maroto and s. j. n. jareno , phys .
d * 87 * , no . 4 , 043523 ( 2013 ) ; arxiv:1309.7447 [ hep - ph ] .
j. a. r. cembranos , a. l. maroto and s. j. n. jareno , jcap * 1403 * ( 2014 ) 042 [ arxiv:1311.1402 [ gr - qc ] ] .
p. sikivie and q. yang , , 111301 ( 2009 ) ; d.j.e .
marsh and p.g .
ferreira , , 103528 ( 2010 ) ; d.j.e .
marsh , e. macaulay , m. trebitsch , and p.g .
ferreira , arxiv:1110.0502v1 ( 2011 ) ; b. kain and h.y .
ling , , 023527 ( 2012 ) ; p.h .
chavanis , astron .
astrophys .
* 537 * , a127 ( 2012 ) ; o. erken , p. sikivie , h. tam , and q. yang , 2012 , , 061304 ( 2012 ) ; o. erken , p. sikivie , h. tam , and q. yang , , 063520 ( 2012 ) . w. hu , r. barkana and a. gruzinov , phys .
lett . * 85 * , 1158 ( 2000 ) ; m. alcubierre , a. de la macorra , a. diez - tejedor and j. m. torres , phys .
d * 92 * ( 2015 ) 6 , 063508 ; j. m. torres , m. alcubierre , a. diez - tejedor and d. nez , phys .
d * 90 * ( 2014 ) 12 , 123002 .
a. taruya , phys .
d * 59 * , 103505 ( 1999 ) ; s. tsujikawa , phys .
d * 61 * , 083516 ( 2000 ) ; s. kasuya , phys .
b * 515 * , 121 ( 2001 ) ; l. a. boyle , r. r. caldwell and m. kamionkowski , phys . lett .
b * 545 * , 17 ( 2002 ) .
s. kasuya , m. kawasaki and f. takahashi , phys .
b * 559 * , 99 ( 2003 ) .
n. d. birrell and p. c. w. davies .
quantum fields in curved space .
cambridge ( 1982 ) . | the fact that fast oscillating homogeneous scalar fields behave as perfect fluids in average and their intrinsic isotropy have made these models very fruitful in cosmology . in this work
we will analyse the perturbations dynamics in these theories assuming general power law potentials @xmath0 . at leading order in the wavenumber expansion , a simple expression for the effective sound speed of perturbations
is obtained @xmath1 with @xmath2 the effective equation of state .
we also obtain the first order correction in @xmath3 , when the wavenumber @xmath4 of the perturbations is much smaller than the background oscillation frequency , @xmath5 .
for the standard massive case we have also analysed general anharmonic contributions to the effective sound speed .
these results are reached through a perturbed version of the generalized virial theorem and also studying the exact system both in the super - hubble limit , deriving the natural ansatz for @xmath6 ; and for sub - hubble modes , exploiting floquet s theorem . |
in this paper we investigate the excitations of nuclear matter with the pion quantum numbers . our analysis is based on studies the in - medium pion dispersion equation .
interactions of pions with baryons in nuclear matter is included in framework of migdal model @xcite .
solutions of the pion dispersion equation in this model were considered on the physical sheet of the complex plane of pion frequency @xmath16 .
we expand this analysis to the unphysical sheets of the riemann surfaces .
we have included first the unphysical sheets to the analysis in our earlier papers @xcite . in @xcite
we studied the branches of solutions responsible for the pion condensation and the long wavelength instability and it was shown that they are the separate branches supplemented to the well - known zero - sound , pion , and isobar ones . in this paper
we continue to study the solutions of zero - sound and pion dispersion equations responsible for the instability of nuclear matter using the retarded pion propagator .
we analyze zero - frequency solutions of the pion dispersion equation depending on the value of @xmath1 and define to what type excitations these zero - frequency solutions refer .
this could permit us to do conclusion about the character of the phase transition related to the considered instability . except the purely theoretical interest
the problem has various applications .
investigation of the pion excitations in nuclear matter , started long ago , continue to play an important role in nuclear physics .
special interest in study of the pionic many - body degrees of freedom was stimulated by prediction and investigation of the pion condensation @xcite . in recent years
the study of the pionic many - body degrees of freedom is related to the investigations of the excited states and of phase transitions in nuclear matter in heavy ion collisions
. detailed knowledge of in - medium pion dynamics is essentially important for description of mesons ( @xcite ) , nucleons ( @xcite ) and @xmath17-isobars @xcite , @xcite .
analysis of the dilepton production from @xmath18 annihilation requires the knowledge of the pion dispersion in hot and dense matter @xcite .
to study of the pion dynamics in heavy ion collisions we need the relativistic description of the pion self - energy at the finite temperatures .
solutions of the relativistic pion dispersion equations were presented in papers @xcite .
a pion self - energy with the correct nonrelativistic limit was obtained in papers @xcite .
when considering the pion dynamics in nuclear matter at high densities and temperatures , it is important to have as a `` point of reference '' a reliable description of the nuclear matter excitations with pion quantum numbers in nonrelativistic limit .
such description have been obtained in the pioneering papers of migdal and his group @xcite , followed by the numerous papers @xcite . in this paper
we study the solutions of the nonrelativistic pion dispersion equations in symmetric nuclear matter at zero temperature following to @xcite .
the aim of this paper is to study the solutions with zero frequency @xmath16=0 depending on the value of the coupling constant @xmath1 .
these solutions characterize the stability boundary ( in this model it is the boundary on the density ) . in this paper
it is shown that at different values of @xmath1 the zero - frequency solutions belong to the different types of excitations .
the value of @xmath1 is considered in the interval @xmath19 .
when we change @xmath1 , the branches of solutions are changed as well : certain solutions go over from the physical to unphysical sheets and vice verse . to identify the solutions on the unphysical sheets
it is important to know the solutions on the physical ones .
the branches of solutions which are obtained in the present paper reproduce the results of @xcite on the physical sheet of the complex plane of @xmath16 . to do comparison with other papers
the simple model with the constant effective quasiparticle interaction is very useful .
however , the pion condensation in this model emerges at too low density and this is not consistent with the results of investigations , @xcite .
it is well known that the solutions of the pion dispersion equations , @xmath2 , describe the different types of excitations in the nuclear matter .
they are : spin - isospin zero - sound wave , @xmath20 , pion wave , @xmath21 , isobar wave , @xmath22 , and others @xcite .
the appearance of the solution with @xmath16=0 means that at the certain values of nuclear density @xmath3 and wave vector @xmath23 the frequency of a definite type of excitations @xmath24 turns to zero : @xmath25=0 .
if @xmath26 one obtains @xmath27 , and the amplified solution takes place .
this signals the instability of nuclear matter .
the change of @xmath1 causes the changes of the values of @xmath28 and @xmath29 .
moreover , at a special situation zero - frequency solution passes to another type of excitations @xmath30 : @xmath31=0 .
it was shown in @xcite that the phase transition is determined by the type of those excitations which become unstable .
thus it is important to know to what type of excitations the solutions with @xmath16=0 belongs . in fig .
1 we show the values of critical densities @xmath32 and wave vectors @xmath33 for which the pion dispersion equation @xmath34 has solutions with @xmath16=0 at any value of @xmath1 .
we are interested in solutions which satisfy additional restriction : @xmath35=0 . in eq .
( 1 ) @xmath36 is the pion self - energy part ( retarded polarization operator ) .
the pion self - energy is formed on the basic of particle - hole and isobar - hole loops renormalized due to the effective interactions of quasiparticles : @xmath37 @xmath38 and @xmath39 .
( the effective constant @xmath1 is regarded as @xmath40 through the paper . ) in fig .
1 the critical density @xmath28 and wave vector @xmath29 are presented for three models of the pion self - energy .
the matter is stable at @xmath41 for the every model .
the physical variables which are related to the pion propagator @xmath42 diverge along the curves in fig . 1 .
this corresponds to the phase transition @xcite in nuclear matter . at @xmath41 , the nuclear matter
is considered as a normal fermi liquid , but at @xmath5 there is phase transitions into new state which may include the standing spin - isospin wave or `` condensates of excitation '' .
let us introduce the useful definitions .
the waves @xmath20 , @xmath21 , @xmath22 , and others we name branches of solutions .
the set of branches of solutions of the same type for the different values of @xmath1 composes a family ( @xmath43 , @xmath44 , @xmath45 , and others ) .
when we say about zero - frequency solutions or @xmath16=0 solutions , we mean that the condition @xmath35=0 is satisfied in this point , as well
. return to fig . 1 .
there are known two reasons for instability in the framework of migdal theory @xcite in the nonrelativistic symmetric cold nuclear matter at not too high densities . in paper
@xcite it was shown that fermi liquid becomes unstable when the attractions between the quasiparticles is too strong . in spin - isospin channel
this corresponds to @xmath46 @xcite . in fig .
1 we see that there are not solutions with @xmath16=0 at @xmath46 .
this is because of the matter is unstable at the any density .
another region of @xmath13 corresponds to the instability in respect to transition into a new ground state containing condensate of pions @xcite . in the present paper the main attention
is paid to study that families of solutions which are responsible for instability along the curves in fig . 1 , from long wavelength instability to pion condensation .
we start with the analysis of the zero - sound dispersion equation for spin - isospin excitations with quantum numbers @xmath0 .
this equation is closely related to the pion dispersion equation .
it was shown in paper @xcite that for the case of strong attraction between the quasiparticles ( @xmath46 ) the zero - sound dispersion equation has a growing with time solution which corresponds to instability in response of small - amplitude , long wavelength density fluctuations .
for @xmath47 the unstable solutions do not appear @xcite .
in this paper the family of solutions responsible for the long wavelength instability is built .
we denote this family as @xmath9 . the index
sd " means spinodal decomposition " .
it is known @xcite , that the thermodynamic stability condition requires the partial derivative of the pressure with respect to the volume to be negative , i.e. , @xmath48 .
the process which takes place in matter , when this condition is broken , is known as a spinodal decomposition @xcite .
it is shown in @xcite , that there is a relation between the effective quasiparticle interaction and the partial derivative @xmath49 @xmath50 is the number of particles , @xmath51 is the effective scalar - isoscalar quasiparticle interaction .
the stability condition is broken at @xmath52 .
for the zero - sound dispersion equation this corresponds to the appearance of the imaginary amplified solutions .
it looks reasonable to use the notation @xmath14 for such solutions .
here we use the same notation for the analogous solutions , which appear in the spin - isospin channel at @xmath46 . moreover , as it is shown below , at @xmath53 all the solutions belong to the same family .
we use the notation @xmath14 for them as well .
the branches @xmath14 are imaginary and can be found for all values @xmath53 .
now we turn to the pion dispersion equation . in the channel of excitations with the pion quantum numbers
the particle - hole interaction can be modified by taking into account the pion - nucleon interactions @xcite . as a result , collective and particle - hole excitations are renormalized due to interaction with pions and , vice versa , the pion excitations are changed due to interaction with particle - hole ones in the nuclear matter .
it was predicted @xcite , that the account of the pion - nucleon interaction is responsible for the phase transition into the ground state with the pion condensation . in fig . 1 it is demonstrated that due to the pion - nucleon interaction we obtain the phase transition not only for @xmath54 but for all ( considered ) values of @xmath1 . for the pion dispersion equation ( [ 1 ] )
we can find the branch of solutions corresponding to @xmath14 .
we denote it as @xmath55 .
the correspondence is established in the following way .
it will be seen from the form of the pion self - energy that @xmath14 are the poles of @xmath56 , and for each value of @xmath57 the frequency @xmath55 tends to @xmath14 as pion - nucleon coupling constant @xmath58 .
we show also that there is a special value of @xmath1 , denoted by @xmath59 ( @xmath60 ) .
when @xmath61 , the solutions with @xmath16=0 are on the branches @xmath55 . when @xmath10 , the type of the family carrying zero - frequency solutions is changed .
we shall refer to @xmath62 as `` transition value '' .
when @xmath10 , zero - frequency solutions pass to the branches of solutions of another family .
it is denoted as @xmath11 ( condensate " ) .
the solutions of ( 1 ) referred to this family , @xmath15 , are identified in the following way . in papers
@xcite it was shown that among the solutions of the pion dispersion equation the imaginary solutions emerge under special conditions .
instability of matter , related to these imaginary solutions is interpreted as a phase transitions into a new ground state containing the pion condensate . the branch @xmath15 reproduces the imaginary solutions presented in @xcite on the physical sheet of @xmath16 .
it is demonstrated that @xmath11 is an independent family of solutions along with @xmath44 , @xmath43 , @xmath45 or @xmath63 and it is responsible for pion condensation . in sect .
2 the expressions for the retarded pion self - energy are presented and the singularities on the complex plane are investigated .
we consider the structure of the riemann surfaces that are defined by the form of the pion self - energy .
the unphysical sheets are described . in sect .
3 the family @xmath9 is considered for the zero - sound and pion dispersion equations .
it is shown how zero - frequency solutions @xmath64=0 at @xmath3 ( fig .
1 ) come these branches . in sect .
4 the different types of solutions of the pion dispersion equation are considered on the physical and unphysical sheets .
it is shown that there is an independent family of solutions @xmath11 that is related to the pion condensation . in sect .
5 we determine the value of @xmath62 and show how the solution with @xmath16=0 pass from the family @xmath55 to @xmath15 when @xmath65 . in this paper
we solve the dispersion equations in the symmetric nuclear matter with zero temperature .
the equilibrium density is @xmath66 , @xmath67=0.268 gev . during the computations ,
the effective mass of quasiparticles is assumed to be @xmath68=0.8@xmath69 , while @xmath69=0.94 gev is the vacuum nucleon mass .
the pion mass is @xmath70=0.14 gev , the isobar mass @xmath71 gev ( the isobar width is neglected ) .
following to the paper @xcite , the effective nucleon quasiparticle interaction can be written as @xmath72 @xmath73 where @xmath74 are the pauli matrices in the isospin and spin spaces .
the factor @xmath75 , where @xmath76 is a density of states for two sorts of nucleon on the fermi surfaces , @xmath77 ; @xmath78 are the landau - migdal dimensionless parameters . in this paper
we use a simple model where parameters are taken to be a constant : @xmath79 .
the landau - migdal parameter @xmath1 found from the analysis of the nuclear experimental data is @xmath80 @xcite .
this parameter describes the short - range correlations between the nucleons with a range @xmath81 : @xmath82 .
the relation @xmath83 sets the correspondence of the landau - migdal parameters to the effective correlative parameters @xmath84 @xcite .
the landau - migdal parameter @xmath85 corresponds to rescattering of @xmath86 quasiparticles .
the parameters @xmath38 and @xmath39 determine the rescattering of @xmath87 and @xmath88 .
the parameters @xmath38 and @xmath39 are not changed in the course of paper and are taken equal to @xmath38=0.4 and @xmath39=1.6 @xcite . in the paper
we take the pion - nucleon coupling constant @xmath89 and the pion - nucleon - isobar coupling constant @xmath90 .
in this section we present the expressions for the pion self - energy in nuclear matter .
these expressions are well known @xcite .
we do an analytical continuation in @xmath16 of the polarization operator @xmath56 to the unphysical sheets of the complex plane .
the search of zero - frequency solutions is related to investigation of the solutions both on the physical and unphysical sheets .
we consider excitations in nuclear matter which are described by the solutions of the pion dispersion equation ( 1 ) .
the causality requirement leads to the known interpretation of the solutions of the dispersion equation in the complex plane of @xmath16 .
the stable excitations are described by the solutions with real frequencies , placed on the real axis . the damping solutions ( resonances )
are placed on the unphysical sheets under the cuts , which the polarization operator has on the real axis .
this is landau damping which is related to overlapping of the frequency of excitations with the frequency of free particle - hole pairs ( which give the imaginary part in the pion self - energy ) .
we can introduce the damping of solutions in another way , with the help of the isobar width ( taken as the experimental constant , @xmath91=115 mev ) or quasiparticle damping constant , for example . in this case the real excitations are displaced to the lower half plane of the physical sheet @xcite .
we do not consider this damping henceforth . at certain conditions ( large density , a strong attraction between particles ) the imaginary solutions arise in the upper half - plane of the physical sheet .
they correspond to the excitations with amplify amplitude .
such excitations correspond to an unstable regime .
we find the corresponding solutions with negative imaginary part in the lower half - plane of the physical sheet , as well .
the amplified solutions in the spin - isospin channel ( as well as all the solutions with @xmath92 ) are interpreted , following @xcite , as the frequencies of @xmath93-type excitations . while the solutions in the lower half - plane ( as well as all the solutions with @xmath94 ) are considered as @xmath95-type excitations .
their frequencies have to be taken with an opposite sign in respect to the solutions . for the physical excitations of @xmath96-type we put into correspondence the solutions with @xmath92 .
in symmetrical nuclear matter with @xmath97 all three types of solutions , i.e. , @xmath98 are identical . in the asymmetric nuclear matter
the solutions carrying different charges are transformed in a different way with the change of the parameter of asymmetry and of the density .
the method used in this paper permits to follow these transformations from symmetric up to neutron nuclear matter .
pion self - energy in nuclear matter is due to @xmath99- and @xmath100-interaction of the pions and nucleons .
the total pion self - energy is thus a sum of a scalar and vector terms : @xmath101 the main results of this paper are obtained for @xmath102 operator .
one can consider some models for @xmath103 and they give the expressions like @xmath104 without @xmath16 and @xmath57 dependence @xcite .
the influence of @xmath105 on the final results has the quantitative character and is discussed in appendix a. following the papers @xcite we write @xmath106 as a sum of the nucleon and isobar parts @xmath107 the operators @xmath108 and @xmath109 can be expressed through the lowest order polarization insertions @xmath110 and @xmath111 which can be presented as the integrals over the particle - hole ( @xmath112 ) and isobar - hole loops ( @xmath113 ) .
the vertices of these loops are described by the pion - nucleon and isobar - nucleon - pion coupling constants @xmath114 , @xmath115 and by the form factors which take the nonzero baryon size into account , @xmath116 , where @xmath117 and @xmath118=0.667 gev , @xmath119=1.0 gev .
polarization operators @xmath110 , @xmath111 can be written as : @xmath120 @xmath121 then for @xmath122 , @xmath123 we have the expressions @xcite : @xmath124 @xmath125 in this equation @xmath126 . in eqs .
( [ 5f ] ) we see that @xmath108 is @xmath110 with one vertex renormalized due to nucleon - hole and isobar - hole interaction .
the denominator @xmath127 has a form @xmath128 and the condition @xmath129 gives a zero - sound dispersion equation .
solutions of this equation describe the excitations in nuclear matter , if we do not take into account the renormalization due to the pion - baryon interaction .
equation ( 1 ) has the solutions corresponding to the solutions of eq .
( [ 7 ] ) .
thus at @xmath130 , eq . ( [ 7 ] ) has the solutions @xmath14 ( which will be shown and discussed below ) , while at @xmath131 there are zero - sound solutions @xmath20 @xcite .
these solutions are the poles of @xmath108 and @xmath109 ( [ 5f ] ) .
the equation ( 1 ) has the corresponding solutions : @xmath55 and @xmath132 .
it can be seen from the structure of eq .
( 1 ) that when @xmath114 and @xmath133 approach zero points , the branches @xmath55 ( @xmath132 ) approach to the poles of polarization operator @xmath56 , i.e. , @xmath20 ( @xmath14 ) .
the causal polarization operator @xmath110 can be presented as a sum of two terms , the first term corresponds to excitation and the second to the annihilation of the particle - hole pairs .
we want to study the excitations in nuclear matter which appear as a response on an external perturbation and we use the retarded operators in eq .
( 1 ) . for the real @xmath16
they are determined as @xcite @xmath134 and @xmath135 is expressed through the integrals @xcite @xmath136 .\ ] ] the function @xmath135 can be written as @xmath137 the first integral in ( [ 9a ] ) corresponds to @xmath138 and the second one to @xmath139 .
equation ( [ 9a ] ) can be used to obtain a condition which relates the values of @xmath135 in the two points of the complex plane @xcite @xmath140 the same equation can be written for @xmath141 and , consequently , for operators in eqs .
( [ 2f])-([6f ] ) . henceforth only the retarded operators are used and index @xmath142 is omitted .
we write the expression for @xmath143 in such a way that it does not contain the overlapping logarithm cuts ( compare with @xcite ) : @xmath144 @xmath145 for @xmath146 .
each of logarithms has its own region of integration over @xmath147 and over the angle between @xmath148 and @xmath149 in ( [ 9a ] ) .
these regions are shown in @xcite . at @xmath150 , @xmath151 is the migdal s function @xcite @xmath152\,\ ] ] where @xmath153 , @xmath154 .
function @xmath155 , where @xmath156 is determined by ( [ 11 ] ) with @xmath157 .
let us consider the cuts of @xmath110 in the complex plane of @xmath16 .
it is clear that , for @xmath158 , there are two cuts ( we denote them by @xmath159 and @xmath160 ) .
the cuts originated by the logarithmic terms of ( [ 10 ] ) are at @xmath161 @xmath162 the cuts of the function @xmath163 lie on the negative semi - axis symmetrically with respect to the cuts in ( [ 12 ] ) .
therefore , @xmath110 has four cuts in the complex plane of @xmath16 , which are shown in fig .
2 . when @xmath164 , the cuts are caused by the singularities of @xmath151 , while at @xmath165 the cuts are related to the logarithms of @xmath163 . the imaginary factors in denominators in ( [ 9a ] ) determine the physical sheet in the complex plane of frequency .
on the physical sheet the first logarithm in ( [ 10 ] ) has the imaginary part on the upper edge of the cut @xmath159 equal to @xmath166 , and @xmath167 on the lower edge of the cut . in the point @xmath168=0 , where the cuts @xmath159 and @xmath160 touch each other , the imaginary part of @xmath151 is continuous but
there is a jump of the derivative .
this is shown in @xcite , fig .
let us now determine the unphysical sheets which we pass on when go under the cuts @xmath159 and @xmath169 ( fig . 2 ) .
consider the cut @xmath159 .
we denote the first logarithm in ( [ 10 ] ) as @xmath170 . while the frequency @xmath16 goes along the cut @xmath159 ( [ 12 ] ) : @xmath171 then @xmath172 is changed in the interval @xmath173 .
the cut @xmath159 in the complex plane of @xmath16 corresponds to the cut along the negative real axis in the complex plane of @xmath172 .
when we go to the cut @xmath159 from above ( the arrow in fig .
2a ) , this corresponds that we go to the cut from below in the complex plane of @xmath172 .
going over under the cut , we pass on to the unphysical sheet neighboring with the physical one on the riemann surface of @xmath174 .
the values of @xmath174 on the neighboring sheet differ by of @xmath175 from the values on the physical one at the same @xmath16 .
as example , the value of @xmath176 on the lower edge of cut @xmath159 is @xmath177 .
but the value at the same point of @xmath16 on the unphysical sheet is @xmath178 .
thus , we have the continuous changing of logarithm along the arrow in fig . 2a .
then we conclude that @xmath56 in eq .
( [ 1 ] ) changes continuously as well .
thus , going over under the cut @xmath159 in fig .
2a we pass on to the unphysical sheet of the complex plane of @xmath16 that is placed under the physical sheet .
let us denote the unphysical sheet with the same letter as a cut : @xmath159 .
the part of the unphysical sheet @xmath159 is shown in fig .
2b by the shading with the right slope .
now we define an unphysical sheet related to the cut @xmath169 ( denote the sheet by @xmath169 ) .
the cut @xmath169 stems from the first logarithm in @xmath139 .
we designate @xmath179 = @xmath180 .
going over under the cut of @xmath181 we add the shift @xmath182 to the value of @xmath181 .
then , the logarithm on the sheet @xmath169 under the lower edge of the cut @xmath169 is equal to @xmath183 .
this expression is used for the analytical continuation to the sheet @xmath169 .
the part of the sheet @xmath169 is shown on the fig .
2b by the shading with the left slope .
in the paper we consider the solutions of ( [ 1 ] ) which are placed on the shaded parts of the unphysical sheets @xmath159 and @xmath169 in fig .
each logarithmic term in ( [ 9 ] ) , ( [ 10 ] ) has its own riemann surface .
an important point here is that moving on riemann surface of one of logarithms we stand on the physical sheet for the other logarithms .
thus , we have built up two unphysical sheets @xmath159 and @xmath169 .
let us show that the values of @xmath110 are real and coincide on the negative imaginary axes of the sheets @xmath159 and @xmath169 . on the imaginary axis
we denote @xmath184 . on the negative imaginary axis of the sheet @xmath159 , the function @xmath112 ( [ 9 ] ) , ( [ 10 ] ) takes a form @xmath185 @xmath186.\ ] ] @xmath187 \ .\ ] ] here only the first and the second terms of @xmath138 and @xmath139 ( [ 10 ] ) are shown . on the other hand , on the negative imaginary axis of the sheets @xmath169 the function @xmath112 ( [ 9 ] )
is @xmath188 @xmath189\ ] ] and after reduction we have the expression ( [ 81 ] ) again .
this is the illustration of the following connection between the values of @xmath110 on @xmath159 and @xmath169 sheets : @xmath190 thus , @xmath110 is defined on the lower semiplanes of the unphysical sheets @xmath159 and @xmath169 ( the shaded regions in fig .
2b ) , and the values of @xmath110 coincide on the negative imaginary axes of the sheets @xmath159 and @xmath169 .
the special solutions of ( 1 ) which we are interested in , are placed on the shaded part of the complex plane of @xmath16 in fig .
2b and on the imaginary axes of the physical and unphysical sheets .
in this part we consider solutions of the pion dispersion equation ( 1 ) at @xmath191 .
the definition of @xmath62 will be given in the sect . 5 , since it is determined by the behavior of branches which are studied in sects .
we start by considering the zero - sound dispersion equation , and investigate its solutions for different values of @xmath40 .
we demonstrate the modification of these solutions when we pass to the pion dispersion equation .
in this subsection we consider the unstable solutions @xmath14 of the zero - sound dispersion equation . for @xmath192 , they are amplified , placed on the physical sheet and reflect the instability of nuclear matter in respect to the small fluctuations of density . for @xmath193 , the branches @xmath14 are damped , and come to the negative imaginary axes ( which belongs to the unphysical sheets @xmath159 and @xmath169 ) .
in this section we do not take into account the excitations of isobar and restrict ourselves by @xmath102 in ( [ 2f ] ) .
the zero - sound dispersion equation stems from ( [ 7 ] ) and has the form @xmath194 the results obtained for this equation are valid both in spin - isospin and in isoscalar , spin , isoscalar channels .
first we obtain the solutions of eq .
( [ 17x ] ) in the kinetic theory of landau and compare them with the known results .
the function @xmath143 ( [ 9],[10 ] ) in the long wavelenths limit is @xmath195 substituting this expression in ( [ 17x ] ) , we obtain the known zero - sound dispersion equation @xcite @xmath196 here we denote @xmath197 . when @xmath46 , this equation has two ( positive and negative ) solutions on the imaginary axis of the physical sheet @xcite .
let us denote @xmath198 , then eq .
( [ 5x ] ) transforms into @xmath199 to obtain solutions for @xmath193 , we take the long wavelength limit of eq .
( [ 81 ] ) and substitute it in ( [ 17x ] ) . on the negative imaginary axis of the unphysical sheet , we get an equation @xmath200 solutions of this equation are negative , i.e. , @xmath201 .
this equation coincides with that presented in paper @xcite .
it was pointed out in @xcite that the solution of eq .
( [ 9x ] ) do not correspond to an actual solution " . in agreement with this statement we see that solutions of ( [ 9x ] ) are located on the negative imaginary axis belonging both @xmath159 and @xmath169 unphysical sheets .
it was explained in @xcite the damping of this mode is the landau damping @xcite . in fig . 3 the solutions @xmath202 to eq . ( [ 17x ] )
are shown at @xmath203 .
they are similar to the solutions at @xmath203 in fig . 1 of @xcite ( except the curve @xmath204 at @xmath193 ) .
the shaded region means that the solution are on the unphysical sheets . in fig . 3 it is demonstrated that when the attraction between the particles is increasing , the damping solution placed on the unphysical sheet goes over to the physical sheet , turning to amplified solution . considering the unphysical sheets we can find two solutions of ( [ 17x ] ) for any value of @xmath203 ( fig . 3 ) .
however , at @xmath193 they are on the different unphysical sheets .
it was shown that the solutions of eq .
( [ 9x ] ) are negative thus being located on the unphysical sheet @xmath159 .
now we look for solutions with @xmath205 ( curve @xmath204 in fig . 3 ) .
the situation is the same as we had with the sheets @xmath159 and @xmath169 .
there are two new unphysical sheets @xmath206 and
@xmath207 which overlay the physical one . in appendix b the construction of these sheets is presented . the dispersion equation ( [ 17x ] ) on the imaginary positive unphysical axis
is the same on the both @xmath206 and @xmath207 sheets and has a form @xmath208 the solutions of this equation are positive .
this equation turns into eq .
( [ 9x ] ) if to change the sign of @xmath202 .
the solutions of this equation are shown by the curve @xmath204 in fig .
3 . they are located on the positive imaginary unphysical axis ( this is marked by the horizontal shading ) .
let us compare fig . 3 and fig . 1 in @xcite . in paper @xcite
only one curve is presented which corresponds to decreasing of @xmath202 with increasing of @xmath1 .
this curve describes the physical processes : the amplifying of the amplitudes of the small fluctuations at @xmath46 and the landau damping of excitations at @xmath193 . according to @xcite in the spin - isospin channel
this curve can be considered as describing the @xmath93-type excitations .
another curve shown in fig .
3 corresponds to the @xmath95-type excitations .
their frequencies are to be taken with the opposite sign in respect to the ones shown in fig .
there are two solutions for @xmath96-type excitations . one of them corresponds to the growing excitations ( at @xmath46 ) and the second should be taken with the opposite sign and becomes to degenerate with the first one in the symmetric and asymmetric matter .
it is well known that the kinetic theory of landau gives the linear dependence of frequencies of collective excitations on the wave vectors , i.e. , @xmath209 . when we pass to rpa using eqs .
( [ 9 ] ) , ( [ 10 ] ) , the shape of the dependence changes and we obtain the branches of solutions @xmath2 . in fig .
4 we show the imaginary branches @xmath14 obtained in rpa .
the shaded fields on the edges of figures indicate that the branches are located on the unphysical imaginary axis .
we see that the branches change continuously with @xmath1 .
when @xmath210 , the branches @xmath14 are located on the unphysical sheet only , but for @xmath211 they go over to the physical sheet in the interval of wave vectors @xmath212 . in analogy with fig .
3 , we can find the second solution of eq .
( [ 1 ] ) for each values of @xmath1 and @xmath57 .
the symmetric " branch has the conjugated values on the physical sheet @xmath213 , but it goes to the unphysical sheet @xmath214 when @xmath215 .
all solutions belong to the same family @xmath9 .
this family exists at @xmath203 .
we consider the influence of the pion - nucleon interaction on the branches @xmath14 .
following @xcite we include only the direct term of the interaction of the pion with a particle - hole pair , and do not take into account the noncentral interaction of the quasiparticles . in this section
we take into account the isobar - hole pairs contribution to the pion self - energy , eq .
( [ 7 ] ) .
first let us investigate how the contribution of isobar - hole excitations changes the stability conditions .
let us assume in eq .
( [ 17x ] ) that @xmath16=0 and @xmath216 . in this limit
this equation reduces to @xmath217 .
this gives the value of @xmath40 when instability occurs in the long wavelength limit .
if we take into account the isobar - hole loops , this condition takes the form @xmath218 at @xmath16=0 and @xmath219 , @xmath220 is proportional to nuclear density : @xmath221 . if @xmath222 , the corrections produced by @xmath223 disappear . as a result
the values of @xmath28 and @xmath29 on the solid lines in fig .
1 approach the dashed ones . in fig .
1 the solid lines are obtained for the minimal densities when solutions with @xmath16=0 emerge for every @xmath40 .
it is shown in fig .
5a , how @xmath55 are modified when @xmath114 is changed from 0 to 1 . at @xmath114=0 the branch of solutions of the zero - sound dispersion equation , eq .
( [ 7 ] ) , @xmath14 is presented .
( it should be noted that the results in fig .
5a are obtained for @xmath109 contribution being included in pion self - energy ( [ 2f ] ) while those in fig .
4 do not include @xmath109 . ) as we see from fig .
5a , the growth of @xmath114 results in emerging of the amplified solutions on the physical sheet at @xmath224 . in fig .
5b the branches @xmath55 are shown for different values of density of nuclear matter .
the curve 3 is calculated for @xmath3 and contains the point @xmath225=0 .
it is demonstrated how the zero value @xmath16=0 is placed on this branch ( the point @xmath226 ) . at the point
@xmath226 the condition @xmath227=0 is satisfied .
the density @xmath32 and wave vector @xmath33 , corresponding to zero solution , are presented in fig . 1 ( denoted by @xmath226 ) .
when density is lower than the critical one , @xmath55 are completely situated on the unphysical sheet ( curve 4 ) and the nuclear matter is stable .
when @xmath5 , the amplified solutions appear on the physical sheet in the certain interval of @xmath57 ( the curves 1 and 2 ) .
figures 5a , b demonstrate the typical behavior of the branches of solutions of ( 1 ) at @xmath228 .
thus , we obtain the instability of nuclear matter related to @xmath55 for all @xmath228 .
instability occurs at @xmath229 .
the corresponded solutions of the zero - sound dispersion equation , @xmath14 , demonstrate instability at @xmath230 only .
we refer @xmath55 to @xmath9 family .
now let us consider in fig .
1 another well studied region @xmath232 .
this region of @xmath40 was investigated in connection with prediction of the pion condensation . in fig .
6 the solutions of eq .
( 1 ) are presented at @xmath233 and @xmath234=1.78 .
such figures were published in papers @xcite . in fig .
6 we see the pion branch of solutions @xmath21 , it starts at the frequency @xmath235 .
it turns into the free pion solution as @xmath58 .
the isobar branch @xmath22 starts at the frequency which is determined by the isobar and nucleon mass difference .
the zero - sound branch @xmath132 starts at @xmath236=0 , and at certain value of @xmath57 it acquires an imaginary part due to landau damping @xcite .
it goes over to unphysical sheet related to the cut @xmath160 , eq . (
[ 12 ] ) , @xcite . besides the numerated branches ,
one more imaginary branch is presented in fig .
let us denote it @xmath15 .
tracing of this branch of solutions stimulated development of the theory of the pion condensation " @xcite . in this section
we investigate the family of solutions @xmath11 and show that this family is responsible for the instability at @xmath231 . in fig . 6 a part of a branch @xmath15
is presented that is located on the physical sheet in the interval @xmath237 .
the rest of this branch is on the unphysical sheet @xmath159 . in fig .
7 we present the same branch @xmath15 as in fig . 6 , but here
we show it from another point of view , i.e. , considering the complex plane of @xmath16 ( the solid line ) .
the branch @xmath15 starts at @xmath57=0 at the point ` 1 ' and has the real value @xmath238 . with the increasing of @xmath57 it moves to the unphysical sheet @xmath159 .
it follows from ( [ 6a ] ) that there is another branch of solutions @xmath239 , which starts at the point @xmath240 and is placed on the unphysical sheet @xmath169 ( the dashed curve ) . below in paper
, the hatching on the edge of figures with the right slope means that the drawing branches are located on the unphysical sheet @xmath159 .
the left slope means that they are on the sheet @xmath169 .
the physical sheet is above the axis of abscissas . in point ` 2 ' @xmath15 and @xmath239 coalesce on the imaginary axis .
then @xmath15 ascends the imaginary axis and @xmath239 descends .
( to determine which of branches go up we can take into account the width of isobar in eq .
( [ 4f ] ) as in @xcite . )
thus , two simple roots of eq .
( 1 ) which move on the unphysical sheets @xmath159 and @xmath169 ( when @xmath57 increase ) coalesce on the imaginary axis at @xmath241 . here
the roots become imaginary ones , and continue to move up and down the imaginary axis which belongs to both @xmath159 and @xmath169 sheets . in fig .
7 the values of @xmath15 and @xmath239 are shifted from the imaginary axis by hand to demonstrate the trajectories of solutions with increasing of @xmath57 ( it is shown by arrows ) .
the wave vector @xmath242 corresponds to the turning point of the trajectories ` 3 ' . at this point , @xmath15 and @xmath239
reach their maximal and minimal values .
when @xmath243 , the branches return along the imaginary axis . in fig .
7 the solutions are presented for @xmath244 . at larger @xmath57 ,
the behavior of branches has a nontrivial character but it is not related to physical processes and is not discussed here . in figs . 6 , 7 it is shown how 1 ) the branch @xmath15 goes from the unphysical to the physical sheet at @xmath245 ; 2 ) it becomes imaginary , with growing value of @xmath246 , on the physical sheet ; 3 ) the branch returns to the unphysical sheet at @xmath247 . in fig .
8 the complex plane of @xmath16 is presented where @xmath15 and @xmath239 are shown for different densities of nuclear matter . in these figures
( except fig .
8a ) , the branches are drawn up to the turning point on the imaginary axis , i.e. , at @xmath248 ( see fig .
7 ) . in fig .
8a the density is small and the branches do not attain the imaginary axis . in fig .
8b the branches encounter on the imaginary axis but completely stay on the unphysical sheets ( @xmath249 ) . in fig .
8c the result for the critical density @xmath28 is presented . on the real axis
the solution with @xmath16=0 appears . here
@xmath250 ( @xmath251 ) and we see that the conditions @xmath252=0 and @xmath253=0 are satisfied . the wave vector and density corresponding to the point @xmath254 in fig .
8c are marked by @xmath254 in fig . 1 .
in fig . 8d it is demonstrated that at @xmath5 the solutions with @xmath246 ( @xmath255 ) exist on the physical sheet .
the critical densities and wave vectors in fig . 1 for @xmath256 correspond to the solutions denoted by @xmath254 in fig .
such solutions @xmath15 can be found not only when @xmath231 , but for @xmath191 also .
it should be mentioned , that an attempt to shift the pion condensation to larger densities means that @xmath15 go on to stay on the unphysical sheet at larger densities .
there are several reasons that can influence our results .
it was shown in @xcite that the important factor shifting the pion condensation to the larger densities is the decreasing of the effective nucleon mass with growth of density .
this decreasing is predicted by the scaling of brown - rho @xcite and qcd sum rules @xcite .
however , at smaller values of the nucleon mass the lowest order expansion in powers of @xmath257 in eq .
( [ 9a ] ) is not valid any more , thus requiring relativistic kinematics for the nucleon . besides
, the formation of isobar fermi surface must be taken into account , as soon as the density becomes so large that the quasiparticle energy on the fermi surface exceeds the difference of the isobar and nucleon mass . in figs . 7 , 8 the branches @xmath15 start at the point @xmath258 .
inclusion of the scalar part of the pion self - energy @xmath105 , eq .
( [ 1f ] ) , displaces the initial value to @xmath259 ( appendix a ) .
however , this does not modify the curves in fig . 1 qualitatively .
in sects . 3 , 4 we have presented zero - frequency solutions of eq .
( 1 ) , which appeared for different values of @xmath1 . in this section
the transitional " value @xmath62 is obtained . in sects .
3,4 it was shown that at @xmath191 the zero - frequency solutions belong to the family @xmath9 and at @xmath231 they refer to @xmath11 . in this section
the transition of solution with @xmath16=0 from one family to another is demonstrated . in the vicinity of @xmath62 , at @xmath260 , the branches belonging to @xmath63 and @xmath11 families are placed close to each other on the same unphysical sheets . in figs . 5 , 7 ,
8 these branches are shown separately . in fig .
9 the branches of the both families are shown simultaneously . in fig .
9a at @xmath261 and @xmath3 , the branches @xmath15 , @xmath239 and @xmath55 are presented simultaneously .
the critical value of the wave vector is @xmath262 .
the branches @xmath15 and @xmath239 behave like in fig .
the branch @xmath55 is shown by the dotted curve and it is presented in the interval @xmath263 . the behavior of @xmath55 is similar to the curve 3 in fig .
5b between @xmath57=0 and the point @xmath226 , which corresponds to @xmath23 . in fig .
9a number ` 1 ' denotes the start of branches at @xmath57=0 ; ` 2 ' means the turning point of @xmath55 : @xmath264=0.077@xmath70 ; ` 3 ' corresponds to @xmath57=0.1@xmath70 .
we see that @xmath55 descends when @xmath57 changes from 0 to @xmath264 and then ascends and turns to zero at @xmath23 : @xmath265=0 ( point @xmath266 in fig . 1 ) .
now let us enlarge a little the coupling constant : @xmath40=-0.0018 .
the following calculations are made at the critical density @xmath3 , corresponding to this value of @xmath40 . in this case
we have a coalescence of @xmath15 and @xmath239 on the imaginary axis .
the wave vector of coalescence is equal to @xmath267=0.0962@xmath70 . in this difficult for drawing case , the presentation of the branches is made in two steps . in fig .
9b the branches are shown at @xmath268 , i.e. , before coalescence and in fig .
9c at @xmath269 .
the position of branches at @xmath57=0.094@xmath70 are marked by ` 2 ' . in fig .
9b it is shown that the position of @xmath55 at point ` 2 ' is much lower on the negative imaginary axis than the @xmath270 and @xmath271 .
further , in fig .
9c we see the coalescence of @xmath15 and @xmath239 at @xmath241 ( point ` 3 ' ) .
then @xmath15 and @xmath239 go in the opposite directions along the imaginary axis . in fig .
9c it is shown that the branch @xmath239 descends and meets the branch @xmath55 in the point ` 4 ' ( at @xmath272=0.109@xmath70 ) . [
the branch @xmath55 has turned upward at @xmath57=0.104@xmath70 . ] after the second coalescence , the branches @xmath239 and @xmath55 go in the opposite sides on the complex plane .
thus , the branch @xmath239 has blocked @xmath55 on the unphysical sheet .
the branch @xmath15 goes over to the real axis and a solution with @xmath16=0 ( point @xmath42 ) belongs to this branch : @xmath252=0 at @xmath29=1.60@xmath70 ( the point @xmath42 in fig . 1 ) .
thus , the value of @xmath62 is determined by the interaction of branches on the unphysical sheets shown in fig .
we can conclude that @xmath62 is a special value of the spin - isospin effective quasiparticle interaction when 1 ) there is coalescence of @xmath15 and @xmath239 on the imaginary axis at the critical density @xmath3 and 2 ) @xmath273 @xmath274 , @xmath275 .
the second condition is satisfied as a rule , since @xmath276 tends to @xmath277 as @xmath278 ( compare with fig .
3 ) . in fig .
9 the results are presented for @xmath279 ( [ 1f ] ) , i.e. , without the excitation of isobars . in this case
we have obtained @xmath62=-0.0019 . if to solve eq .
( 1 ) with @xmath280 , then @xmath62=0.0020 .
the influence of @xmath99-wave pion - nucleon interaction is discussed in appendix a. it is interesting to mention that the figure similar to fig .
9 is presented in @xcite for the solutions of the zero - sound dispersion equation .
their figure demonstrates the behavior of solutions for the different values of @xmath51 and the branches of solutions depend on @xmath281 instead of @xmath57 ( this means the dependence on the value of the quasiparticle effective mass @xmath282 along the branch ) . in paper @xcite @xmath51 and @xmath281
are the zero and the first multipolar components of the scalar - isoscalar effective interaction of the quasiparticles .
* a. * in this paper two families of solutions to the pion dispersion equation are calculated , which determine the stability boundary of the nuclear matter ( the critical density @xmath28 ) depending on the value of spin - isospin coupling constant @xmath1 .
it is shown that there is a special value of @xmath65 such that for @xmath283 the stability boundary are related to the family @xmath63 and for @xmath284 it refers to @xmath11 .
it is interesting to investigate , whether we can obtain the information about the difference of the phase transitions at @xmath8 and @xmath10 using the branches of solutions . to do this
let us consider the rate of growth of excitations in the vicinity of @xmath62 . for @xmath285 , eq .
( 1 ) has solutions with @xmath286 . in figs . 5 , 7 ,
8d such solutions are presented .
we determine @xmath287 , @xmath288 .
we can take several values of @xmath1 close to @xmath62 and calculate the maximal values of @xmath289 ( i.e. , @xmath290 ) depending on @xmath291 for the every @xmath1 . in fig .
10 these results are demonstrated for @xmath292 , where @xmath62=-0.0019 . in this figure
we can not observe the essential difference in the dependence of @xmath290 on @xmath291 for @xmath8 and @xmath10 .
it is possible to conclude that either this test is not sensitive to the type of excitations , or the character of the phase transition changes continuously with @xmath1 from long wavelength instability up to the pion condensation .
* b. * now let us consider the singularity of the integrals containing the pion propagator @xmath42 ( 1 ) : @xmath293 , in the poles of @xmath294 when @xmath3 and @xmath295 .
the function @xmath296 is not related to the poles of @xmath294 . in the sect .
3.1 for the zero - sound dispersion equation ( [ 7 ] ) , it was shown that there are two solutions on the physical sheet . at @xmath193 ,
the amplified solutions go over to the unphysical sheets @xmath159 , @xmath169 and the second branch goes over to the unphysical sheets @xmath214 , @xmath207 ( fig . 3 ) .
similar situation takes place for the pion dispersion equation . up to now
we considered the solutions which are located on the physical and unphysical @xmath159 and @xmath169 sheets , but there are the symmetrical " solutions .
they have the conjugated values on the physical sheet and their unphysical sheets are @xmath206 and @xmath207 .
this means that at @xmath3 and @xmath23 not only @xmath297=0 ( @xmath288 ) but the symmetrical " branch as well . as a result
, the function @xmath298 has a form ( we leave the singular part only ) @xmath299 since we are interested in such zero solutions that @xmath300=0 , then @xmath301 as @xmath295 and the integral diverges along the curves in fig . 1 .
this correspond to situation of the phase transitions @xcite . * c. * in the paper we do not consider the case when @xmath38 , @xmath39 are modified simultaneously with @xmath40 .
but as it is shown above , the existence of @xmath62 and @xmath15 , @xmath14 does not depend on whether we take excitation of the isobar into account or not .
therefore we do not expect the qualitative changes in our results if we change the values of @xmath38 and @xmath39 together with @xmath40 .
in the paper we investigate of the zero - frequency solutions of the pion dispersion equation depending on the values of the spin - isospin coupling constant of quasiparticles @xmath1 . for every @xmath1 ,
the critical density @xmath28 and the wave vector @xmath29 corresponding to the solution with @xmath16=0 were calculated .
then the branches of solutions , containing these zero solutions , are constructed and it is determined to what type of excitations ( to what family ) these solutions referred .
it is shown that the zero - frequency solutions belong to the different types of excitations , depending the value of @xmath1 . in all models used in calculations ,
the transitional " value @xmath62 ( @xmath60 ) is found , such that at @xmath7 the zero solution passes from the branches of the type @xmath9 to the ones of type @xmath11 . at @xmath61 ,
the unstable amplified branches of solutions for the zero - sound dispersion equation ( @xmath14 ) and for the pion dispersion equation ( @xmath55 ) are constructed ( figs . 3 , 4 , 5 ) .
5a we see that if to take the pion - nucleon coupling constant @xmath114 value to be very small , @xmath14 and @xmath55 are very close .
when we increase @xmath114 up to 1 , the behavior of @xmath55 is change : the damping solutions become amplified ones ( fig . 5 ) .
we obtain that the zero - frequency solutions of eq .
( 1 ) at @xmath61 belong to the same family of solutions which is responsible for the long wavelength instability in the kinetic theory of landau .
figure 1 demonstrates that the stability boundary continuously extends to @xmath302 .
but at these @xmath1 the unstable solutions belong to family @xmath11 .
it is shown that this family contents the solutions responsible for the pion condensation .
it is possible to expect that a new state of medium after phase transition is different depending the value of @xmath1 .
we investigated the dependence of the maximal growth rate of perturbations at several @xmath1 in the vicinity of @xmath62 on density . but this dependence does not exhibit the noticeable modifications at @xmath62 .
author thanks s. v. tolokonnikov for the constructive criticism .
also , author thanks v. r. shaginian and e. g. drukarev and especially m. g. ryskin for the numerical fruitful discussions .
the work was supported by the grants rfbr-03 - 02 - 17724 .
in this appendix the scalar part of the pion self - energy , @xmath105 , ( 2 ) and its influence on the behavior of the curves in fig . 1 are considered .
we construct @xmath105 using the relation of gell - mann oakes
renner @xcite , @xcite .
it permits to express the pion mass in medium through the quark condensates in nuclear matter @xmath303 where @xmath304 - masses of @xmath305- and @xmath306-quarks ( @xmath307=11 mev ) , @xmath308 - coupling constant of the pion decay in the medium ( here we assume that @xmath309=92 mev , i.e. , the vacuum and medium values of @xmath310 coincide ) .
the value @xmath311 is a scalar quark condensate in nuclear matter .
it may be present as @xmath312 here @xmath313 is a scalar quark condensate in vacuum , @xmath313=-0.03 gev@xmath314 ; @xmath315 is the matrix element of the scalar quark condensate calculated over the nucleon @xmath316 .
then from ( a1 ) we obtain @xmath317 on the other hand , we can define @xmath318 using eq .
( 1 ) as @xmath319 the form of @xmath105 can be obtained comparing ( a2 ) and ( a3 ) provided that @xmath320 @xmath321 in fig .
1 the dotted lines correspond to the critical density @xmath28 and wave vector @xmath29 obtained for eq .
( 1 ) with the pion self - energy ( [ 1f ] ) . ( recall that for each @xmath1 in fig . 1
we show the minimal @xmath28 when solution @xmath16=0 appears ) . in this model
the transition value of @xmath1 is @xmath62=0.0026 .
here we construct the unphysical sheets @xmath206 and @xmath207 .
the solutions of eq .
( 19 ) that are presented by the curve @xmath204 in fig .
2 are placed on these sheets .
to pass on to the sheet @xmath206 we do the analytical continuation in @xmath16 of @xmath322 from the lower edge of the cut @xmath159 upwards , adding @xmath323 to @xmath324 .
the sheet @xmath206 is situated above the physical sheet .
the value of @xmath325 above the upper edge of the cut @xmath159 is @xmath326 .
this expression is continued on the sheet @xmath206 . in similar way we construct the sheet @xmath207 making the analytical continuation of @xmath327 from the lower edge of cut @xmath169 , adding @xmath323 to logarithm values .
99 a. b. migdal , rev .
mod . phys . * 50 * , 107 ( 1978 ) ; a. b. migdal , d. n. voskresenskii , e. e. saperstein , and m. a. troitskii , phys . rep . * 192 * , 179 ( 1990 ) .
e. g. drukarev , m. g. ryskin , and v. a. sadovnikova , eur .
j. * a4 * , 171 ( 1999 ) .
v. a. sadovnikova and m. g. ryskin , phys . at
. nucl . * 64 * , 440 ( 2001 ) . r. f. sawyer and d. j. scalapino , phys
* d7 * , 953 ( 1973 ) .
g. e. brown and w. weise , phys .
* 27 * , 1 ( 1976 ) .
e. oset , w. weise , and h. toki , phys . rep . * 83 * , 281 ( 1982 ) .
g. mao , l. neise , h. stocker , and w. greiner , phys . rev . *
c59 * , 1674 ( 1999 ) .
j. helgesson and j. randrap , phys .
rev , * c52 * , 427 ( 1995 ) . c. gale and j. kapusta , phys .
* c35 * , 2107 ( 1987 ) .
l. h. xia , c.m .
ko , l. xiong , and j. q. wu , nucl .
a485 * , 721 ( 1988 ) .
l. liu and m. nakano , nucl . phys . * a618 * , 337 ( 1997 ) .
j. diaz alonso and a. perez canyellas , nucl .
a526 * , 623 ( 1991 ) ; l. mornas , nucl - th/0107005 .
l. b. leinson and a. perez , eur .
j. * a20 * , 371 ( 2004 ) .
m. f. m. lutz , phys .
b552 * , 159 ( 2003 ) ; phys.lett . *
b 566 * , 277 ( 2003 ) .
backman , g. e. brown , and j. a. niskanen , phys .
rep . * 124 * , 1 ( 1985 ) . c. garsia - recio , e. oset , and l. l. salsedo , phys
* c37 * , 194 ( 1988 ) .
g. chanfray and m. ericson , phys .
b141 * , 163 ( 1984 ) .
j. delorme and p. a. m. guichon , phys
. lett . * b263 * , 157 ( 1991 ) .
m. colonna and ph .
chomaz , phys . rev . * c49 * , 1908 ( 1994 ) . c. j. pethick and d. g. ravenhall , ann . phys . * 183 * , 131 ( 1988 ) .
t. ericson and w. weise , _ pions and nuclei " _
, clarendon press , 1988 .
l. d. landau and e. m. lifshitz , _ statistical physics " _
( pergamon press , 1989 ) .
pomeranchuk , sov .
jetp * 8 * , 361 ( 1958 ) .
d. pines and p. nozieres , _ the theory of quantum liquids " _ ( w.a.benjamin,inc . , 1966 ) .
h. mueller and b.d .
serot , phys . rev . *
c52 * , 2072 ( 1995 ) .
j. speth and e. werner , w. wild , phys . rep .
* 33 * , 127 ( 1977 ) . a. l. fetter and j. d. walecka,_quantum theory of many - particle systems"_(mc - graw - hill , new york , 1971 ) .
dickhoff , a. faessler , j. meyr - ter - vehn , and h. muthner , phys .
rev . * c23 * , 1154 ( 1981 ) . s .- o . backmann and w. weise ,
`` mesons in nuclei '' _ , v.iii , 1095 , ed . by m.
rho , d. wilkinson , ( north - holland pub .
company , 1976 ) .
g. e. brown and m. rho , phys .
lett . * 66 * , 2720 ( 1991 ) .
r. j. furnstahl , d. k. griegel , and t .d .
cohen , phys . rev . *
c46 * , 1507 ( 1992 ) .
e. g. drukarev , m. g. ryskin , v. a. sadovnikova , th .
gutsche , and a. faessler , phys . rev . * c69 * , 065210 ( 2004 ) .
m. gell - mann , r. oakes , and b. renner , phys . rev .
* 175 * , 2195 ( 1968 ) .
fig . 1 . critical values of nuclear density @xmath28 and wave vector @xmath29 , at which in eq .
( 1 ) solutions @xmath16=0 appear .
results for three models of the pion self - energy are presented : solid lines correspond @xmath328 ( [ 1f ] ) , ( [ 2f ] ) ; dashed curves are for @xmath329 ; dotted curves are for the total @xmath330 ( [ 1f ] ) . fig
. 2 . complex plane of the frequency @xmath16 .
a ) numbers @xmath159 and @xmath160 denote the cuts ( [ 12 ] ) of @xmath138 and @xmath169 and @xmath331 denote the cuts of @xmath139 ; b ) unphysical sheets are shown : sheet @xmath159 ( the shading with the right slope ) and sheet @xmath169 ( the shading with the left slope ) .
3 . solutions of eqs .
( [ 6x ] ) , ( [ 9x ] ) , ( [ 13 ] ) at @xmath203 in the kinetic theory of landau .
the shading with the left slope marks unphysical sheet @xmath159 and the horizontal shading refers to @xmath206 .
results are obtained at @xmath332 .
imaginary solutions @xmath14 of eq .
( [ 7 ] ) .
a ) dependence of @xmath14 on @xmath114 : the curve ( 1 ) @xmath114=0 ; ( 2 ) 0.4 ; ( 3 ) 0.6 ; ( 4 ) 1.0 .
the results are obtained at @xmath332 and @xmath40=-0.4 .
b ) dependence of @xmath14 on the value of density : the curve ( 1 ) @xmath333=1 ; ( 2 ) 0.318 ; ( 3 ) 0.265 ; ( 4 ) 0.221 .
the results are obtained at @xmath114=1 and @xmath40=-0.4 .
point @xmath226 marks solution with @xmath16=0 .
fig . 6 .
different types of solutions of eq .
( 1 ) : spin - isospin zero - sound branch @xmath132 , pion branch @xmath21 , isobar branch @xmath22 and condensate branch @xmath15 .
results are obtained at @xmath234=1.78 and @xmath40=2 .
we computed that @xmath334=1.35@xmath70 , @xmath335=2.35@xmath70 .
fig . 7 .
branches @xmath15 and @xmath239 on the complex plane of frequency . the shading with the right ( left ) slope marks the unphysical sheet @xmath159 ( @xmath169 ) .
the physical sheet is above the abscissas axis .
axes of reference are shown by the dotted lines .
the solid curve correspond to @xmath15 and the dashed one is for @xmath239 .
the number ` 1 ' marks start of branches at @xmath57=0 ; ` 2 ' is the coalescence point at @xmath267=1.28@xmath70 ; ` 3 ' is the turning point at @xmath264= 1.88@xmath70 .
variables @xmath336 , @xmath234 and @xmath40 are the same as in fig .
fig . 8 .
branches @xmath15 and @xmath239 on the complex plane of frequency at different densities @xmath234 . in figures
b , c , d branches are shown at @xmath337 .
point @xmath254 marks solution with @xmath16=0 .
notations of lines and sheets are the same as in figs . 6 , 7 .
transition of a zero solution from family @xmath9 to family @xmath11 .
the solid , dashed and dotted curves denote @xmath15 , @xmath239 and @xmath55 , correspondingly .
a ) @xmath1=-0.0020 .
the number ` 1 ' corresponds to @xmath57=0 ; ` 2 ' means turning point of @xmath55 : @xmath264=0.077@xmath70 ; ` 3 ' @xmath57=0.1@xmath70 .
arrows show movement of branches with @xmath57 .
point @xmath266 corresponds to @xmath338=0 .
b ) @xmath1=-0.0018 .
branches are shown for @xmath268 ( point ` 2 ' ) .
c ) continuation of figure b. branches are shown for @xmath339 .
point ` 3 ' marks the coalescence point of @xmath15 and @xmath239 at @xmath57=0.0962@xmath70 .
point ` 4 ' shows the coalescence point of @xmath239 and @xmath55 at @xmath57=0.109@xmath70 .
point @xmath42 corresponds to @xmath252=0 . | in this paper we consider instability of nuclear matter which takes place when the frequencies of the collective excitations turn to zero . we investigate collective excitations with pion quantum numbers @xmath0 .
we study the dependence of zero - frequency solutions of the pion dispersion equation on the value of the spin - isospin quasiparticle interaction @xmath1 .
the solutions of the pion dispersion equation describe the different types of the excitations in the matter , @xmath2 . at the critical density @xmath3 one of solutions of
the definite type turns to zero : @xmath4=0 . when @xmath5 , the excitations @xmath6 become amplified .
it is shown that there is such a transitional " value of @xmath7 that for @xmath8 the zero - frequency solutions belong to the type @xmath9 while for @xmath10 they pertain to the type @xmath11 .
the solutions of the type @xmath9 correspond to instability to small density fluctuations of the nuclear matter at @xmath12 . on the other hand , @xmath11 is responsible for the pion condensation " at @xmath13 . for the stable nuclear matter
the branches of solutions @xmath14 and @xmath15 are located on the unphysical sheets of the complex plane of frequency . |
since the work of lieb _ et al .
_ @xcite , fourier microscopy has become an increasingly important experimental technique for nano - optics .
it is now commonly used to study quantum emitters @xcite , optical nanostructures @xcite , and the interactions of these two systems @xcite .
for example , fourier microscopy has been used to characterize the orientation of single molecules @xcite and luminescent excitons in layered materials @xcite , the radiation pattern and directivity of optical antennas @xcite , and the multipolar origin of quantum transitions @xcite .
these fourier microscopy studies all share a common goal , namely to measure quantitative information about the angular spectrum radiated by a microscopic sample .
however , a surprisingly wide range of optical systems and setups have been used to achieve this goal , including many different objective classes with varying levels of aberration correction and numerical aperture ( na ) .
for example , researchers have used everything from dry objectives with 0.8 na @xcite to 1.49 na total internal reflection fluorescence ( tirf ) objectives @xcite and even 1.65 na high - index objectives @xcite .
researchers have also used several different configurations to image the back focal plane ( bfp ) .
some configurations place a bertrand lens before the microscope s image plane @xcite ; others place a bertrand lens after the image plane @xcite , while a third set use relay optics to reimage and magnify the bfp @xcite . beyond nano - optics , fourier microscopy is also becoming an important tool for wide - field imaging and structured illumination applications .
recently , techniques have been developed to reconstruct high - resolution , wide - field images from multiple fourier - space measurements @xcite .
in addition to imaging , fourier - space techniques are also being used for optical trapping .
for example , by leveraging the fourier transform properties of an objective , researchers have shown how spatial light modulators can be used to simultaneously trap many particles in arbitrary 3d configurations @xcite . adapting optical microscopes to fourier - space imaging and manipulation can introduce unexpected challenges .
for example , even the simple task of focusing in fourier space can lead to counterintuitive results . whereas real - space alignment can be readily achieved by focusing on fine features in the image ,
such fine features in fourier space are generally the product of aberrations ( e.g. , distortion near the pupil edge ) . in this context
, fourier microscopy raises a number of design choices that are distinct from real - space imaging .
specifically , most commercially available objectives and configurations have been optimized for different real - space applications .
yet , it is unclear which of these corrections are most important for quantitative fourier imaging and also to what extent the optics for fourier imaging will introduce additional aberrations .
the purpose of this paper is to systematically evaluate the optical design choices and experimental parameters inherent in fourier microscopy , including which microscope objectives are best suited for fourier imaging , the ideal collection method and the relative advantages of different bertrand lens configurations .
to quantitatively examine these design choices , we first model complete microscope systems in zemax .
detailed information about commercial microscope objectives and tube lenses are inferred by examination of published patents from nikon , olympus , and zeiss .
( as a potential resource to readers , we provide zemax files for all optical elements as well as the combined systems in ref . @xcite . ) based on ray - tracing analysis , we show that the ideal objective for fourier microscopy is a plan - corrected apochromat with a high numerical aperture and low magnification .
furthermore , we show that placing the bertrand lens into the infinity " space between the objective and tube lens can yield significant imaging improvements .
fourier microscopy typically involves three basic components : a microscope objective , a tube lens , and a bertrand lens .
this paper focuses on the application of fourier imaging to modern microscopes , where an infinity - corrected microscope objective is used in combination with a tube lens to produce an image of the object at the exit port of the microscope .
the addition of a bertrand lens allows one to image the fourier transform of the object by effectively re - imagining the objective s bfp ; in this way , the tube lens and bertrand lens can be seen as a two - lens optical system used to image the objective s bfp . for the purpose of this paper ,
we assume that the bertrand lens will be a standard achromatic doublet designed for visible wavelengths .
the term bertrand lens is generally used to describe a lens that performs a fourier transform without changing the position of the conjugate plane . for simplicity here
though , we refer to any lens used to perform a fourier transform as a bertrand lens . as shown in fig
. [ fourierconfigs ] , there are four commonly used configurations for bfp imaging .
the first two simply place an achromatic doublet ( thorlabs ac254 - 050-a and ac254 - 100-a , respectively ) either before @xcite or after @xcite the microscope s image plane as shown in fig .
[ fourierconfigs](a ) and fig .
[ fourierconfigs](b ) , respectively .
the first configuration typically limits the bertrand lens to have a focal length of @xmath050 mm ( due to finite accessible space before the image plane ) and is therefore limited in its magnification of the bfp .
however , this configuration allows for simple switching between fourier and real - space imaging by inserting or removing the bertrand lens .
the second configuration allows for greater magnification of the bfp image , but can not be used for real - space imaging without additional optics .
the remaining two configurations attempt to overcome these limitations by using relay optics to move the image plane further from the microscope exit port .
figure [ fourierconfigs](c ) shows a modified 4f relay ( thorlabs ac254 - 100-a and ac254 - 050-a ) , where the second of the two lenses can be replaced by one with twice the focal length to obtain a real - space image @xcite . the alternate configuration in fig .
[ fourierconfigs](d ) uses a pair of relaying lenses ( thorlabs ac254 - 200-a and ac254 - 150-a ) followed by an achromatic pair ( thorlabs map104040-a ) that can be removed in order to obtain a real - space image @xcite .
although these two designs offer greater flexibility , the additional elements significantly increase alignment difficulty and tolerancing errors . in the following analysis
, we will weigh the advantages and disadvantages of these common configurations while also identifying a new approach .
before considering the imaging configurations , it is possible to determine the ideal microscope objective . while microscope manufacturers do not tend to supply consumers with detailed models , their patent applications for objective lenses often specify a great deal of information .
these patents often included the radius of curvature , thickness , refractive index , and abbe number for each optical surface in the objective and associated tube lens .
therefore , by searching the patent literature , it is possible to infer likely designs for commonly used objectives .
although these patents do not include details about any surface coatings used to minimize reflections ( and therefore can not be used to model overall system throughput ) , they do specify enough information for modeling optical aberrations . in this section , we provide the rationale by which we examined the patent literature and came to identify likely designs for commercially available components . to guide our search ,
we have sought to identify patent applications that were submitted near the commercial release date of new objectives , e.g. identifying a 2004 patent application from nikon @xcite describing tirf objectives with `` na larger than 1.45 '' that predates the 2005 release of their 1.49 na tirf objectives @xcite . wherever possible
, we have then tried to use physical examination ( i.e. , disassembly of objectives and tube lenses ) to confirm basic design properties such as the number of elements and their relative curvatures . for every objective examined
, we did notice slight discrepancies in specific details , e.g. in the curvatures or thicknesses of some lenses , but discrepancies are to be expected from patent specifications . despite being sold under the same commercial name ,
the designs for objectives and tube lenses are presumably subject to continuous improvements and modifications .
for example , over the last 20 years , the removal of environmentally hazardous materials from optical glasses ( e.g. , arsenic , lead and hcfcs ) has required the redesign of many optical components @xcite . nevertheless , we believe these detailed patents are still a helpful source for design optimization , particularly for understanding the differences between objective classes , even if the final objectives may vary from these exact specifications .
commercial patents generally specify optical materials by their abbe number and refractive index at the d - line ( 587.5 nm ) . with these parameters alone , optical models are only accurate over a relatively small range of wavelengths near the d - line .
to enable more accurate analysis over the full visible spectrum , we performed a glass substitution using zemax s built in glass catalogs .
where possible , we replaced each material with tabulated data from a `` standard '' or `` preferred '' glass with an abbe number and index identical to the patent specification .
if an exact match could not be found for the abbe number , we looked for the closest match that still provided an exact index match .
where no index match was possible , we simply used the original abbe number and refractive index .
these substitutions greatly improved chromatic aberrations and brought them qualitatively in line with the spectrally dependent field curvature plots provided in the patents .
we further validated this process by examining the glass manufacturers . for the zeiss objective , it was possible to use schott glass for all but one surface .
( carl zeiss microscopy and schott are both subsidiaries of the carl zeiss foundation . )
similarly for olympus , all but one surface was ohara glass .
( ohara lists olympus as one of its major customers @xcite . ) finally , nikon objectives appear to use schott and a variety of japanese optical glass manufacturers , including its subsidiary hikari .
the first element we were able to identify in commercial microscopes is the tube lens , because it is a relatively simple component .
since tube lenses are also integral to all infinity - corrected microscope designs , they are often defined in multiple patents . by searching the patent literature
, we found numerous patents from nikon @xcite , olympus @xcite and zeiss @xcite in which the same tube lens was specified for a given manufacturer .
in addition , we were able to obtain and disassemble lenses from all three manufacturers .
we found that the zeiss tube lens ( part number : 452960 ) matched the single element design described in objective patents .
we also found that the nikon tube lens for ti - u inverted microscope ( part number : mea53210 - 1xl ) agrees with the patents in the number of components and their relative curvatures , but component thicknesses differed by @xmath11 mm .
however , the olympus tube lens ( part number : u - tlu-1 - 2 ) had a different number of elements than the patent specification . despite these discrepancies , we are confident in the basic designs as they have been listed in many patents spanning several decades and , therefore , seem to be the tube lenses used by the manufacturers to specify objective performance . in tables [ nikontubelens ] - [ zeisstubelens ] of appendix a , we present the tube lens implementations specified by nikon , olympus , and zeiss patents . a number of different objectives were chosen from nikon , olympus , and zeiss based on their availability in the patent literature and in order to cover a large range of na and aberration corrections . a detailed list of objective manufacturers , magnifications , na , aberration corrections and their associated patents can be found in table [ table : objectives ] .
although we can not be certain that these are exact matches to commercial products , we still believe that the qualitative results obtained from analyzing these patents can help identify the optimal objective class for fourier imaging .
.patent sources used for objective modeling [ cols="^,^,^,<,^ " , ] @xmath2 modeled as 1.49 na to match possible commercial realization [ table : objectives ] to ensure that the zemax models faithfully represent the patent specifications , we first reproduced the d - line distortion and spherical aberration figures in the patents .
as discussed above , we then performed a detailed glass substitution process ; this allowed us to confirm the wavelength dependent spherical aberration plots .
note that the glass substitution procedure does not directly affect our analysis , because we perform all subsequent aberration calculations at the d - line .
nevertheless , we chose to perform the glass substitutions to ensure we correctly modeled the objectives and to create a library of objectives that would be of broader use .
in addition to limited information on the optical glasses that are used , almost none of the patents specify component semi - diameters ( which are relevant to vignetting ) .
it is not possible with these patent specifications to assume that the objectives could achieve their full na over the full field of view , because this would require physically impossible lens surfaces , e.g. lenses that intersect other lenses and/or themselves . in order to determine the semi - diameters
, we allowed these values to float such that the objective collects all the light from an on - axis point source which emits at the specified na . using this process , and comparing the resulting semi - diameters to patents where semi - diameters were specified @xcite , we were able to validate this approach .
specifically , for ref .
@xcite , there was @xmath3 @xmath4 m error on all surfaces . for ref .
@xcite , there was @xmath5 mm error on all surfaces and an exact match on the surface that is the primary source of vignetting .
for the results of the glass substitution and semi - diameter determination , see tables [ nikon1p4_60x ] - [ zeiss1p45_100x ] in appendix b. figure [ objectiveschematic ] shows an example of the resulting modeled surfaces from a nikon 1.4na , 60x objective .
for real - space imaging of structures and specimens , a large variety of microscope objectives have been developed , each with different optimization metrics in mind .
these includes apochromatic ( low chromatic aberrations ) , plan ( flat field and low distortion ) , and tirf ( small depth of field ) objectives .
however , since these were all designed for real - space imaging , their performance for fourier imaging is not well known . here
, we examine the effect of aberration corrections , nominal magnification , and numerical aperture as well as the choice of confocal or full - field collection on the quality of the resulting fourier image .
significant deviations from the ideal @xmath6 apodization factor ( dashed black line ) are seen for tirf ( e - g ) and high na water immersion objectives ( a ) , whereas apo ( h ) and plan - apo ( b - d ) objectives perform well even over a relatively large bl displacement of @xmath7 cm . ] when selecting an objective for fourier - space imaging , it is well known that a great deal of information exists beyond the critical angle for total internal reflection ( e.g. , @xmath8 for a sample - air interface ) .
thus , a common approach is to maximize angular collection by using the largest possible na objective .
however , image distortion increases significantly with na .
this is especially true when the na approaches the refractive index of the immersion medium , @xmath9 .
indeed , a key measure in the design of high na objectives is the ratio @xmath10 .
tirf patents often present their claims in terms of such ratios ; for example , ref .
@xcite specifies a tirf objective relative to a @xmath11 condition . as this number approaches unity and higher angle rays
are collected , both aberrations and tolerancing errors are expected to increase . to examine this effect , we plot in fig .
[ apodization ] the apodization factors for the objectives listed in table [ table : objectives ] as a function of defocusing a bertrand lens ( 50 mm achromatic doublet ) placed before the image plane . for each objective
, distortion data was extracted as a function of bertrand lens position and apodization factors were calculated according to ref .
the apodization factor is a dimensionless scaling term that describes how projected ray density increases with increasing field angle ( @xmath12 ) .
this factor is derived from the distortion introduced by the optical system , but is generally assumed to be @xmath6 for systems obeying the abbe sine condition ( i.e. , systems with a wide field of view such as microscopes ) @xcite .
note that the apodization factor was chosen as our metric of objective quality , because it is typically the only correction made to fourier - space images to take into account the optical setup @xcite .
we specifically examine defocus misalignment as it is often the most difficult tolerancing error to correct , because it tends to couple with distortion in common fourier microscopy configurations . while decentering and tilt of the bertrand lens also lead to undesirable changes to the fourier image , they are relatively easy to observe and correct for even in fourier space .
( for example , by observing fluorescence from an isotropic thin film emitter , tilt and decenter are seen as clear radial asymmetries in the resulting fourier image , which are readily corrected . ) finally , we chose to use the bip fourier imaging configuration shown in fig .
[ fourierconfigs](a ) due to its simple design , which allowed for easy determination of the objective s influence . to analyze the results in fig .
[ apodization ] , consider first the objectives that use a standard @xmath11.515 refractive index immersion oil .
as can be seen from a comparison to the ideal @xmath6 apodization shown as a dashed black line , the plan - apo objectives in fig .
[ apodization](b - d ) perform well over the full numerical aperture , whereas tirf objectives in fig .
[ apodization](e - g ) perform quite poorly at na values beyond 1.3 .
note that the sharp peaks and sudden drop off at high na values are features commonly seen during experimental alignment of the bertrand lens .
interestingly , the 1.65 na objective performs quite well , but this is explained by the fact that it uses a high @xmath11.780 refractive index immersion oil . the high index 1.65 na objective
thus has a relatively low @xmath13 ratio of 0.927 , which is comparable to that of the 1.4 na objectives .
in contrast , the 1.25 na water immersion objective , which has a relatively high @xmath13 ratio of @xmath10.938 , performs quite poorly .
this underscores the point that na alone may not be the best criteria in selecting an objective for fourier microscopy . in particular , one should be careful about the use of tirf and other high @xmath13 objectives , especially in applications where a high na plan - apo objective may be suitable . for completeness , we note that the objective performance trends discussed above for the bip configuration were also observed for the aip configuration in fig .
1(b ) as well as the preferred configuration discussed in section 5 where the bertrand lens is placed before the tube lens . while the choice of imaging configuration does impact the magnitude of tolerancing errors , the plan - apo and high - index apo objectives consistently demonstrated lower tolerancing errors than the tirf and 1.25 na water immersion objectives .
in addition to aberration corrections and na , magnification is an important parameter for image quality and resolution . for real - space imaging ,
the choice is fairly obvious , the highest na and magnification will allow you to resolve the smallest features .
however , for fourier - space imaging , something counterintuitive occurs . for the same na ,
you actually obtain a larger fourier image when using an objective with a lower nominal magnification .
to see why , consider the simple schematic shown in fig .
[ mtfcompare ] , which shows how the fourier image is formed at the bfp of a single lens . as the focal length
is increased ( in order to decrease the magnification for a fixed tube lens ) , the semi - diameter must increase in order to maintain the same na . thus , as can be seen in eq .
[ equation : magnification ] below , the bfp semi - diameter height , @xmath14 , is inversely proportional to the objective s nominal real - space magnification , @xmath15 , specified by the manufacturer : @xmath16 is the half angle determined by the na of the objective .
( b ) comparison of modulation transfer functions for the imaging configuration shown in fig .
[ fourierconfigs](a ) using 60x and 100x nikon 1.4 na plan - apo objectives . ] where @xmath17 is the focal length of the tube lens , @xmath18 is the immersion oil index of refraction , and na is the numerical aperture of the objective .
however , there s still a question as to whether this improves the angular resolution ; the image may be larger , but it may also be more blurred out . to demonstrate that this is not the case , we examine the modulation transfer function ( mtf ) for two comparable systems with differing magnification .
( the mtf is defined as the contrast ratio that would be observed at the image plane for a sine wave object of a given spatial frequency . ) specifically , we plot the mtf for two nikon objectives of the same na but different magnifications ( objectives 3 and 4 in table 1 ) .
note that , not only do these two objectives have the same na , their specifications came from the same patent @xcite making them the ideal comparison . as can be seen in fig .
[ mtfcompare](b ) , both objectives have a very similar mtf ( and therefore comparable angular resolutions ) .
thus , the lower magnification objective maintains bfp image quality while increasing its size .
( however , it is important to note that the smaller bfp images from high magnification objectives may still be desirable when working with low light samples , e.g. weak single emitters . )
although these objectives were all designed for low distortion wide - field imaging , they still suffer from reduced throughput at the edges of the field of view .
for real - space imaging , this simply leads to a darker image at the edges than in the center .
however , for fourier imaging , vignetting decreases the observed intensity ( collection efficiency ) at large angles when imaging the full field of view .
this , in turn , can lead to quantitatively different results when fitting the resulting radiation pattern .
thus , it is important to consider the effect of the collection area ( which is often linked to the excitation spot size ) .
m spot , ( b ) 100 @xmath4 m spot and ( c ) full field of view for the objectives listed in table .
[ table : objectives ] using the imaging configuration shown in fig .
[ fourierconfigs](a ) . ] for collection from a 1 @xmath4 m confocal region at the center of the field of view , vignetting is negligible , and all objectives transmit essentially 100% of rays up to their full na .
however , for wide - field collection , almost all objectives experience serious vignetting far short of their designed na . to illustrate this
, we plot the fraction of transmitted rays in fig . [ vignetting ] for three different collection areas , ranging from a 10 @xmath4 m diameter area in fig .
[ vignetting](a ) to a 100 @xmath4 m diameter area in fig .
[ vignetting](b ) to the specified full - field in fig .
[ vignetting](c ) .
( note that the full - field size depends on the manufacturer and magnification , because fields are generally specified by image , rather than object , size .
nikon and olympus specify the full - field to be a 22 mm image , whereas zeiss specifies a 25 mm image ) .
while the vignetting effects within a 10 @xmath4 m central area are minor , it is clear from fig .
[ vignetting ] that increasing the field beyond 100 @xmath4 m ( or operating farther than 50 @xmath4 m from the center of the field ) will reduce the effective na and can distort the fourier image .
thus , although wide - field excitation of a fluorescent sample can greatly improve the signal - to - noise ratio , it will lead to an undesirable quantitative modification to the fourier image in almost all microscope objectives and should therefore be avoided .
from the analysis in section 4 , we have determined that the best choice for fourier - space imaging is a plan corrected apochromatic objective with a low magnification and a high na .
( although all calculations were done at the d - line and therefore chromatic aberrations were not considered , the specific objectives that performed the best in our modeling were plan - apos .
we would also recommend the use of apochormatic objectives in general , because the majority of fluorescence experiments involve broadband emitters . ) with this in mind , we will now consider the relative performance of the four imaging configurations shown in fig .
[ fourierconfigs ] . to compare these four configurations , we use the 60x 1.4na nikon plan - apo objective , and extract the field curvature and apodization factor at the detector plane .
( as mentioned above , the apodization factor is often the only modification used to account for the optical setup in the literature . ) as can be seen in fig .
[ configurationcompare](a ) , the apodization factors are nearly identical for all four configurations .
however , fig .
[ configurationcompare](b ) shows clear differences in the field curvatures .
the field curvature is plotted in units of mm as the distance between the paraxial image plane and the real image plane as a function of @xmath19 . while the aip and bip configurations have field curvatures on the order of 0.1 mm , the m4f and ac relay configurations exhibit curvature on the order of 1 mm .
therefore , it seems that the simpler aip and bip configurations may be slightly better , but given the small curvature ( i.e. @xmath01.3 mm ) for all systems , it is not clear that this would lead to a significant imaging improvement .
( a - d ) , respectively . ( a ) apodization factor for each configuration shown together with the ideal @xmath6 apodization factor ( dashed black line ) for a system obeying the abbe sine condition . ( b ) sagittal and tangential field curvature as a function of field angle for four configurations shown in fig . [ fourierconfigs ] . ]
the choice between the four fourier imaging configurations in fig .
[ fourierconfigs ] is often limited by practical constraints .
for example , it is often useful to place additional optics ( e.g. , polarizers , beamsplitters , filters , etc . ) in the light path .
making the focal length of the bertrand lens as long as possible can also be desirable , because it makes the bfp image on the detector larger and therefore reduces pixelation .
thus , despite being the simplest modification to a standard microscope , the bip configuration in fig .
[ fourierconfigs ] is overly restrictive .
requiring the detector to be at the exit port image plane puts significant limits on the focal length of the bertrand lens and leaves little room for other optical components .
the aip configuration alleviates both of these issues , but can not be used to obtain real - space images at the same detector plane .
finally , the m4f and ac relay configurations are the most flexible in magnification and allow for the removal or replacement of lenses to obtain real - space images .
however , this increased flexibility comes at the cost of increased field curvature as well as increased complexity and associated tolerancing errors .
( c , f ) . ]
interestingly , there is one other place where we have access to the beam path in an standard inverted microscope , namely in the infinity " space between the objective and tube lens .
although this space is typically occupied by excitation and emission filters , these are easily removed , and fourier imaging optics can be put in their place .
the advantage of this setup is that the bertrand lens , which now effectively acts as a tube lens , can be placed such that it directly focuses on the objective s bfp , as shown in fig .
[ bertrandbbl](a ) .
this configuration combines the simplicity of the one - lens configurations in fig .
[ fourierconfigs](a , b ) and the bfp magnification flexibility of the relay configurations in fig .
[ fourierconfigs](c , d ) without curvature or tolerancing complications . as can be seen by comparison of fig .
[ bertrandbbl](b , c ) to fig .
[ apodization](c , f ) , tolerancing errors are greatly reduced when the bertrand lens is placed before the tube lens .
this is primarily due to the fact that the bertrand lens is now directly imaging the back - focal - plane of the objective , and therefore forms an infinite conjugate pair with the tube lens , whereas in all other configurations they form a finite conjugate pair .
importantly , the configuration shown in fig . [ bertrandbbl](a ) also provides a major advantage for system alignment , because deviations in the position of the bertrand lens cause almost pure defocus .
in the four common configurations shown in fig .
[ fourierconfigs ] , movement of the bertrand lens leads to both defocus and distortion , and it is distortion changes near the pupil edge that often produce the sharpest features during bfp alignment .
however , when positioning the bertrand lens in infinity space , the lack of distortion effects means that sharp features in the resulting fourier image do correspond to good focus and , thus , proper alignment .
in conclusion , we have used patent information to model a range of microscope objectives in order to determine the ideal fourier - space imaging setup . we have shown that , despite the significantly larger angles available to tirf objectives , large deviations from ideal apodization can lead to severely aberrated fourier images , especially for commonly used imaging configurations .
it is therefore best to use a high na , plan - apo objective with a low magnification . also , despite the gains in
signal - to - noise when collecting from the full field of view , vignetting leads to an undesirable decrease in throughput at high angles and should therefore be avoided by using on - axis small - area collection . finally , when using a standard commercially available microscope , the choice of imaging configuration outside the microscope is primarily a practical concern .
however , with simple modifications , placing a lens between the objective and tube lens can significantly improve imaging performance and simplify the alignment process .
we hope that these results , together with the tabulated surface data in the appendices and zemax models in ref .
@xcite , will help advance the use of fourier microscopy in nano - optics and nanophotonics .
we also hope the methods and analysis presented here will be of interest to other fields , such as biological imaging , where complete models of microscope objectives and systems could help optimize high resolution techniques .
listed below are the specifications for the most likely implementations of nikon , olympus , and zeiss tube lenses inferred from the patent literature , where r is the radius of curvature , d is the thickness , h is the semi - diameter height , and mfr is the glass manufacturer .
( semi - diameters are set to the smallest clear aperture measured from actual tube lenses . )
listed below are the specifications for the microscope objectives listed in table [ table : objectives ] , where the notation is the same as described above in appendix a. note that for surfaces where an exact glass match could not be obtained ( and for the immersion oil for every objective ) , the index and abbe number are given at the d - line instead .
a. drezet , a. hohenau , d. koller , a. stepanov , h. ditlbacher , b. steinberger , f. aussenegg , a. leitner , and j. krenn , `` leakage radiation microscopy of surface plasmon polaritons , '' mat .
b * 149 * , 220229 ( 2008 ) . c. huang , a. bouhelier , g. c. d. francs , a. bruyant , a. guenot , e. finot , j .- c .
weeber , and a. dereux , `` gain , detuning , and radiation patterns of nanoparticle optical antennas , '' phys . rev .
b * 78 * , 155407 ( 2008 ) .
h. aouani , o. mahboub , n. bonod , e. devaux , e. popov , h. rigneault , t. w. ebbesen , and j. wenger , `` bright unidirectional fluorescence emission of molecules in a nanoaperture with plasmonic corrugations , '' nano lett .
* 11 * , 637644 ( 2011 ) .
k. lee , x. chen , h. eghlidi , p. kukura , r. lettow , a. renn , v. sandoghdar , and s. gtzinger , `` a planar dielectric antenna for directional single - photon emission and near - unity collection efficiency , '' nature photon . * 5 * , 166169 ( 2011 ) .
s. a. guebrou , c. symonds , e. homeyer , j. plenet , y. n. gartstein , v. m. agranovich , and j. bellessa , `` coherent emission from a disordered organic semiconductor induced by strong coupling with surface plasmons , '' phys .
* 108 * , 066401 ( 2012 ) .
n. hartmann , g. piredda , j. bertelot , g. c. des francs , a. bouhelier , and a. hartschuh , `` launching propagating surface plasmon polaritons by a single carbon nanotube , '' nano lett .
* 12 * , 177181 ( 2012 ) .
n. hartmann , d. piatkowski , r. ciesielski , s. mackowski , , and a. hartschuh .
, `` radiation channels close to a plasmonic nanowire visualized by back focal plane imaging , '' acs nano * 7 * , 1025710262 ( 2013 ) .
y. chen , d. zhang , l. zhu , r. wang , p. wang , h. ming , r. badugu , and j. r. lakowicz , `` tamm plasmon- and surface plasmon - coupled emission from hybrid plasmonic - photonic structures , '' optica * 1 * , 407413 ( 2014 ) .
l. shi , t. hakala , h. rekola , j .-
martikainen , r. moerland , and p. trm , `` spatial coherence properties of organic molecules coupled to plasmonic surface lattice resonances in the weak and strong coupling regimes , '' phys .
lett . * 112 * , 153002 ( 2014 ) .
g. bulgarini , m. e. reimer , m. b. bavinck , k. d. jns , d. dalacu , p. j. poole , e. p. bakkers , and v. zwiller , `` nanowire waveguides launching single photons in a gaussian mode for ideal fiber coupling , '' nano lett .
* 14 * , 41024106 ( 2014 ) .
j. a. kurvits , m. jiang , and r. zia , `` zemax ( optical design ) files of microscope objective , tube lens , and fourier imaging setups , '' figshare ( 2015 ) .
[ retrieved 14 july 2015 ] , http://dx.doi.org/10.6084/m9.figshare.1481270 .
`` nikon instruments wins r&d magazine s top 100 new products of the year , '' http://www.nikoninstruments.com/about-nikon/news-room/us-news/nikon-instruments-wins-r-d-magazine-s-top-100-new-products-of-the-year , [ accessed : 2015 - 7 - 9 ] . | fourier microscopy is becoming an increasingly important tool for the analysis of optical nanostructures and quantum emitters .
however , achieving quantitative fourier space measurements requires a thorough understanding of the impact of aberrations introduced by optical microscopes , which have been optimized for conventional real - space imaging . here , we present a detailed framework for analyzing the performance of microscope objectives for several common fourier imaging configurations . to this end , we model objectives from nikon , olympus , and zeiss using parameters that were inferred from patent literature and confirmed , where possible , by physical disassembly .
we then examine the aberrations most relevant to fourier microscopy , including the alignment tolerances of apodization factors for different objective classes , the effect of magnification on the modulation transfer function , and vignetting - induced reductions of the effective numerical aperture for wide - field measurements .
based on this analysis , we identify an optimal objective class and imaging configuration for fourier microscopy .
in addition , as a resource for future studies , the zemax files for the objectives and setups used in this analysis have been made publicly available . |
under the influence of the high pressure , the elemental calcium undergoes a series of structural phase transitions . in particular , one can distinguish seven phases in the range of pressure ( @xmath12 ) from @xmath13 to @xmath14 gpa @xcite , @xcite , @xcite ( please see figure ( a ) for the details ) .
the two first phases , namely ca - i and ca - ii , have been classified as a fcc and bcc structures , respectively @xcite , @xcite .
the third phase ( ca - iii ) has been primarily linked with the sc structure , however the recent reports suggest other assignments .
on the basis of the theoretical studies , teweldeberhan _ et al .
_ proposed the _
structure @xcite .
also vote in favor of the _ cmmm _ structure @xcite . on the other hand , mao _
_ have predicted the transition from the sc - like structure to the monoclinic phase at @xmath15 k and @xmath16 gpa @xcite .
it needs to be underlined that the stability of the structure sc in the area of the existence of the phase ca - iii is being confirmed by the results achieved by errea _
et al . _ and yao _
_ , at least for the temperature of @xmath17 k @xcite , @xcite .
the existence of the phases ca - iv and ca - v has been experimentally examined in the papers @xcite and @xcite .
et al . _ have proposed for them the following assignment : the structure ca - iv should be characterized by @xmath18 and ca - v by _
space groups , respectively @xcite . in the year
@xmath19 , nakamoto _ et al .
_ discovered the new ca - vi phase with the _ pnma _ structure @xcite .
further , in the year @xmath20 , sakata _
et al . _ have reported the existence of the host - guest phase ca - vii @xcite .
we can notice that the high - pressure phase of the host - guest character had been previously predicted by arapan _
et al . _ and then by ishikawa _ et al . _
@xcite , @xcite .
the first mention of the existence of the pressure - induced superconducting state in calcium was provided by dunn and bundy in 1981 @xcite .
fifteen years later , okada _
_ determined the dependence of the critical temperature ( @xmath11 ) on the pressure up to the 150 gpa @xcite ( please see figure ( b ) for details ) . in the year 2006 , yabuuchi _ et al .
_ have repeated the experimental studies of okada @xcite .
it has been found that the values of the critical temperature increase much faster together with the increase of @xmath12 in comparison to the results achieved by okada .
the last notable experimental results have been obtained by sakata _
_ @xcite . on the basis of figure ( b )
, it can be easily noticed that for @xmath21 gpa , the critical temperature takes the value equal to @xmath22 k ( the highest observed @xmath11 among all elements ) .
however , this result has been challenged by andersson @xcite . in the presented paper , we have determined all relevant thermodynamic parameters of the superconducting state that is induced in calcium under the pressure at @xmath0 gpa .
we draw the readers attention to the fact that the pressure of @xmath0 gpa represents the highest value of @xmath12 considered by yabuuchi _
additionally , the high value of the critical temperature at @xmath23 gpa , which is equal to @xmath24 k , has been recently confirmed by the results obtained by sakata _
_ @xcite .
for the purpose of this paper , we have assumed that the phase ca - vi is being characterized by the @xmath25 crystal structure . to support this assumption
we quote the results presented in : @xcite , @xcite and @xcite .
on the imaginary axis ( @xmath26 ) , the order parameter ( @xmath27 ) and the wave function renormalization factor ( @xmath28 ) can be calculated by using the eliashberg equations @xcite : @xmath29 @xmath30 in equations ( [ r1 ] ) and ( [ r2 ] ) the symbol @xmath31 denotes the @xmath32-th matsubara frequency , where @xmath33 , and @xmath34 is the boltzmann constant . the complicated form of the electron - phonon pairing kernel is represented by the expression : @xmath35 where @xmath36 is the maximum phonon frequency ( @xmath37 mev ) , and @xmath38 indicates the eliashberg function , which models the shape of the electron - phonon interaction in a detailed way . in the presented paper , the form of the @xmath38 function has been taken from yin _ et al
. _ @xcite .
the depairing interaction between electrons is described with the use of the coulomb pseudopotential ( @xmath39 ) .
the symbol @xmath40 denotes the heaviside function and @xmath41 is the phonon cut - off frequency : @xmath42 . in the presented paper
the eliashberg equations have been solved for @xmath43 matsubara frequencies ( @xmath44 ) . in this case , the obtained solutions are stable for the temperatures greater than or equal to @xmath45 k. a detailed discussion of the numerical method has been presented in @xcite-@xcite .
the physical value of the coulomb pseudopotential ( @xmath46 ) can be defined by using the condition : @xmath47_{t = t_{c}}=0 $ ] , where the critical temperature is equal to the experimental value ( @xmath2 k ) @xcite . for selected values of the coulomb pseudopotential ( @xmath48 ) .
( b ) the maximum value of the order parameter as a function of the coulomb pseudopotential . ] in figure ( a ) we have presented the dependence of the order parameter on @xmath49 for selected values of @xmath39 .
one can notice that together with the increase of the coulomb pseudopotential , the largest value of the order parameter ( @xmath50 ) decreases .
additionally , in figure ( b ) we have outlined the complete form of the function @xmath51 . on the basis of the obtained results
, we have found that the physical value of the coulomb pseudopotential is equal to @xmath52 .
the above result means that the depairing electron correlations in calcium are relatively strong ( the classical low - temperature superconductors @xmath46 is about @xmath53 @xcite ) .
it can be noted that similar non - standard value of @xmath46 has been obtained for lithium and @xmath54 @xcite , @xcite , @xcite , @xcite , @xcite .
for example , the properties of the superconducting state in the fcc phase of lithium for the pressure values @xmath55 gpa ( @xmath56 k ) and @xmath57 gpa ( @xmath58 k ) have been specified in the paper @xcite .
it has been shown that the physical value of the coulomb pseudopotential increases with @xmath12 from @xmath59 to @xmath60 . in the case of @xmath54 ,
the parameter @xmath46 is equal to @xmath61 ( @xmath62 gpa and @xmath63 k ) @xcite .
the high values of @xmath46 in ca , li , and @xmath54 are difficult to explain in the framework of the classical morel - anderson ( ma ) model @xcite , where : @xmath64^{-1}$ ] .
the symbol @xmath65 is defined by : @xmath66 , where @xmath67 is the electronic density of states at the fermi level and @xmath68 is the coulomb potential ; @xmath69 and @xmath70 denote the characteristic electron and phonon frequency , respectively .
since @xmath71 , the ma pseudopotential is of the order @xmath53 and @xmath72 . it can be noted that the ma model corresponds to treating the irreducible vertex to the first order in @xmath68 .
recently , bauer , han , and gunnarsson have extended the ma theory to the second order in @xmath68 .
the main result is that the retardation effects lead to the reduction @xmath73 also in the higher order calculation , but not as efficiently as in the first order @xcite .
the model presented in the paper @xcite is probably the most advanced attempt to explain the high values of @xmath46 in ca , li , and @xmath54 .
in particular , bauer _
et al . _ have given the following expression for the physical value of the coulomb pseudopotential : @xmath74+a\mu^{2}\ln\left[\frac{\alpha\omega_{e}}{\omega_{ph}}\right]},\ ] ] where @xmath75 and @xmath76 .
on the basis of the equation ( [ rdodatkowe ] ) one can easily estimate the value of @xmath68 for the real materials . in the paper , we assume the following : @xmath77 ( @xmath78 is the half - band width ) , @xmath79 ( @xmath80 $ ] ) , and @xmath81 ( the constant dos ) . in the literature ,
the values of all the important parameters are provided only for @xmath82 ( @xmath62 gpa ) . in particular : @xmath83 , @xmath84 mev , and @xmath85 mev @xcite , @xcite .
the result is the following : @xmath86 mev .
next , we address an important issue , namely , how big is the error bar of the calculated physical value of the coulomb pseudopotential ( @xmath87 ) .
first of all , we can notice that in the framework of the presented analysis , the value of @xmath46 depends on the shape of the eliashberg function and the accuracy of the experimental value of @xmath11
. for calcium the appropriate eliashberg function , taken from @xcite , has been calculated by using the linear - response method ( full - potential lmtart code @xcite ) . in that paper ,
the eliashberg function error bar has been omitted ( @xmath88_{\alpha^{2}f\left(\omega\right)}=0 $ ] ) .
on the other hand , the value of the critical temperature has been measured with the accuracy about @xmath89 k @xcite .
on the basis of these facts , we have obtained : @xmath88_{t_{c}}=\pm 0.02 $ ] .
the solid line denotes the results obtained with the help of the modified allen - dynes formula .
finally , the dashed and dotted lines have been generated based on the classical allen - dynes formula and the mcmillan expression , respectively . ] in the framework of the presented formalism , the exact value of the critical temperature should be obtained on the basis of eliashberg equations . however , in the case of the data interpretation , it is far more convenient to use the simple formula that explicitly reproduces the results of the advanced numerical calculations . in the branch literature there are known two basic formulas that serve for the determination of the critical temperature s value .
the first one has been introduced by mcmillan @xcite ; the second one is the allen - dynes expression @xcite .
unfortunately , in the case of calcium , both formulas considerably underestimate the critical temperature . according to the above
, the allen - dynes formula has been modified in such a way , that it allows us to reproduce the numerical results correctly .
particularly , in order to achieve the proper values of the fitting parameters , the dependence of the critical temperature on the coulomb pseudopotential has been analyzed on the level of the eliashberg equations ( only @xmath38 has been considered as the physical input parameter ) .
next , the least - squares method was applied .
the obtained result is presented below : @xmath90,\ ] ] where the functions @xmath91 and @xmath92 are expressed by the formulas : @xmath93^{\frac{1}{3}},\ ] ] and @xmath94 the parameters , that depend on the eliashberg function , can be determined on the basis of the expressions : @xmath95 and @xmath96 . for calcium under the pressure at @xmath0 gpa , we have respectively : @xmath97 @xmath98 and @xmath99 . the fitting functions @xmath100 and @xmath101 are presented in the following forms : @xmath102 and @xmath103 in figure we have presented the numerical solutions obtained with the use of the eliashberg equations and the modified allen - dynes formula . additionally , for the comparison purposes
, we have depicted the results based on the classical formulas derived by allen - dynes and mcmillan . on the basis of figure
, one can observe , that the expression ( 4 ) perfectly reproduces the exact eliashberg numerical predictions .
the constants in the equation ( 4 ) deviate notably from the original parameterization .
this situation is connected with the fact that the analysis based on the real - axis eliashberg equations suggests only the semiphenomenological form of the @xmath11-formula : @xmath104 $ ] ( see the detailed discussion in @xcite , p. 1051
- 1052 ) . in the case of calcium under the pressure at @xmath0 gpa ,
the values of the allen - dynes parameters are inappropriate .
thus , the constants ( @xmath105 ) should be fit to the data taken from the exact solutions of the eliashberg equations .
we can notice that the change of the coefficient in the expression ( 4 ) under @xmath106 from @xmath107 to @xmath108 slightly lowers the phonon frequency ; the two remaining parameters ( 1.03 and 0.06 ) , in comparison to the classical parameterization , increase the value of the effective electron - phonon coupling constant .
moreover , the parameterization of the strong - coupling correction function ( @xmath91 ) and the shape correction function ( @xmath92 ) also deviates from the original form .
the achieved result indicates that for the high - pressure superconducting state in calcium the shape function has greater significance than in the classical superconductors .
the value of the critical temperature for @xmath109 gpa has been also calculated in the paper @xcite . by using the allen - dynes formula
the authors have qualitatively reconstructed the experimental value of @xmath11 . however , in the examined case the physical value of the coulomb pseudopotential has been strongly lowered ( @xmath110 ) . in the last step ,
we boldly underline that in the literature exist other calculations of @xmath111 than has been presented in the paper @xcite .
in particular , lei _
et al . _ have suggested a very large value of the electron - phonon coupling constant ( @xmath112 for @xmath113 gpa and the sc structure ) @xcite . on the other hand , aftabuzzaman and islam
have predicted @xmath114 for @xmath23 gpa and the _ pnma _ structure @xcite .
the last result is similar to the result obtained by yin _
the form of the order parameter on the imaginary axis for selected values of the temperature has been presented in figure ( a ) .
it has been shown that together with the increase of @xmath49 the absolute values of @xmath115 are decreasing and are subjected to saturation .
it should be underlined that taking the negative values by the order parameter function is connected with the non - zero value of the coulomb pseudopotential .
when analyzing the temperature s dependence of the order parameter , we found that absolute values of the function @xmath115 decrease together with the temperature s growth .
the above result means that together with the growth of the temperature , the less number of matsubara frequencies give significant contribution to the eliashberg equations .
the full dependence of the maximum value of the order parameter ( @xmath50 ) on the temperature has been plotted in figure ( b ) .
we can observe that the values @xmath116 with a good approximation reproduce the temperature dependence of the energy gap at the fermi level . for selected values of the temperature .
( b ) the maximum value of the order parameter as a function of the temperature . ] in figure ( a ) we have presented the form of the wave function renormalization factor on the imaginary axis .
similarly as for the order parameter , the increase of @xmath49 causes the decrease of the successive values of the function @xmath117 . in the case of the high values of @xmath49 ,
the function @xmath117 is subjected to the saturation and takes the value equal to one .
further , figure ( b ) presents the full dependence of the maximum value of the wave function renormalization factor on the temperature .
it can be noted that the presented function with a good approximation determines the temperature dependence of the electron effective mass .
moreover , from the obtained results , we can conclude that the electron effective mass takes a high value in the entire area of the existence of the superconducting state . for selected values of the temperature .
( b ) the maximum value of the wave function renormalization factor as a function of the temperature . ]
in order to determine the physical value of the order parameter for the chosen temperature , the solutions of the eliashberg equations on the imaginary axis ( @xmath118 ) should be analytically continued on the real axis ( @xmath119 ) . in the presented paper
we have used the method introduced by beach _
the form of the order parameter on the real axis is being reproduced by using the function : @xmath120 where @xmath121 and @xmath122 denote the number coefficients , and @xmath123 .
the dependence of the real and imaginary part of the order parameter on the frequency for selected values of the temperature has been presented in figure .
additionally , the rescaled eliashberg function has been specified .
on the basis of the presented results , one can observe that in the range of the low frequencies ( from @xmath13 to about @xmath124 mev ) , only the real part of the order parameter takes the non - zero values . from the physical point of view
, the obtained result defines the lack of the damping effects . for the higher values of the frequency ( from about @xmath124 mev to about @xmath125 mev ) , the real part of the order parameter takes relatively high values , which are clearly induced by the characteristic peaks in the eliashberg function .
furthermore , we can notice that in the discussed range of the energy , the imaginary part of the order parameter becomes non - zero and strongly increases together with the increase of frequency .
for the higher frequencies ( above @xmath125 mev ) , the real part of the order parameter begins to vanish .
this fact is related to the extinction of the eliashberg function itself .
the physical value of the order parameter for the chosen temperature should be determined on the basis of the expression @xcite , @xcite : @xmath126.\ ] ] in the case of the superconductors the most interesting is the value of the order parameter for the temperature of zero kelvin ( @xmath127 ) .
on the basis of the simple calculations we have made the following estimation : @xmath128 mev .
let us mention that familiarity with the value of @xmath129 and @xmath11 allows to calculate the dimensionless ratio : @xmath3 .
in the case of calcium we have obtained the following : @xmath130 the above result indicates that @xmath131 considerably exceeds the value predicted by the bcs theory : @xmath132_{\rm bcs}=3.53 $ ] @xcite .
the influence of the electron - phonon interaction on the electron effective mass ( @xmath133 ) can be determined on the basis of the expression : @xmath134m_{e}$ ] , where the symbol @xmath135 denotes the value of the wave function renormalization factor on the real axis and @xmath136 is the bare electron mass .
the form of the wave function renormalization factor on the real axis has been calculated with the use of the analytical continuation method : @xmath137 where @xmath138 and @xmath139 are the number coefficients , and @xmath123 . in figure
we have presented the shape of the function re@xmath140 $ ] and im@xmath140 $ ] for the critical temperature . similarly to the situation which took place in the case of the order parameter , for the low frequencies
the non - zero is only the real part of the wave function renormalization factor . in the energy range around @xmath124 mev
we can observe characteristic but not so strong amplification of re@xmath140 $ ] , which is clearly correlated with the peaks of the eliashberg function .
additionally , the function im@xmath140 $ ] is non - zero . in the range of the high frequencies ,
re@xmath140 $ ] decreases together with the increase of @xmath119 . on the temperature .
] next , the dependence of the ratio @xmath141 on the temperature has been determined .
the results have been presented in the inset in figure .
we have found that the electron effective mass is large in the entire range , in which the superconducting state exists , and reaches its maximum equal to @xmath142 for @xmath48 .
we can notice that for @xmath48 the value of the ratio @xmath141 can be calculated with a great accuracy by using the simple formula : @xmath143 .
the consistency between the exact numerical result and the analytical approach is the measure of the presented analysis . from the physical point of view
, the result presented above is particularly important , since it can be verified in a simple way if the measurement of the sommerfeld coefficient is to be made in the future .
the thermodynamic critical field ( @xmath9 ) and the difference between the specific heat in the superconducting and normal state ( @xmath144 ) can be calculated on the basis of the free energy difference ( @xmath145 ) : @xmath146 the dependence of the free energy difference on the temperature has been presented in figure .
we can see that in the whole range of the existence of the superconducting phase , the value of the ratio @xmath147 is negative . from the physical point of view , it means that the superconducting state is thermodynamically stable .
the thermodynamic critical field should be determined on the basis of the expression : @xmath148}.\ ] ] the influence of the temperature on the value of the ratio @xmath149 has been presented in the figure . and @xmath149 as a function of the temperature . ]
the difference of the specific heat has been determined on the basis of the formula : @xmath150 } { d\left(k_{b}t\right)^{2}}.\ ] ] additionally , the values of the specific heat in the normal state have been also determined : @xmath151 where @xmath152 . in figure [ fig09 ] ,
we have plotted the dependence of the specific heat in the superconducting state and the normal state on the temperature .
the characteristic `` jump '' , which appears at the critical temperature , can be easily noticed . on the basis of the specified thermodynamic functions
, we have calculated the values of the dimensionless ratios : @xmath153 and @xmath154 .
we have obtained the following : @xmath155 and @xmath156 taking into account the results above , we can state that the values of the considered ratios significantly diverge from the values predicted by the classical bcs theory . in particular :
@xmath157_{\rm bcs}=1.43 $ ] and @xmath158_{\rm bcs}=0.168 $ ] .
in the paper , we have determined all relevant thermodynamic parameters of the superconducting state in calcium under the pressure at @xmath0 gpa .
we have conducted all numerical calculations in the framework of the eliashberg formalism , where the electron - phonon spectral function @xmath38 has been taken form the paper @xcite . on the basis of the exact numerical results
, we can state that the depairing electron correlations in calcium are relatively strong ( @xmath1 ) . in the next step , the values of the parameters in the allen - dynes formula have been calculated .
it has been shown that the critical temperature is properly determined by the modified allen - dynes expression .
furthermore , we have proven that the thermodynamic properties of the superconducting state significantly diverge from the predictions based on the simple bcs theory .
in particular , the following values of the thermodynamic ratios have been obtained : @xmath159 , @xmath160 , and @xmath161 .
d. szczniak would like to acknowledge the financial support under the `` young scientists '' program , provided by the dean of the faculty of mathematics and science jdu professor z. stpie ( grant no .
dsm / wmp/1/2011/17 ) .
some calculations have been conducted on the czstochowa university of technology cluster , built in the framework of the platon project , no .
poig.02.03.00 - 00 - 028/08 - the service of the campus calculations u3 .
99 h. olijnyk , w.b .
holzapfel , phys .
a * 100 * , 191 ( 1984 ) .
t. yabuuchi , y. nakamoto , k. shimizu , t. kikegawa , j. phys .
jpn . * 74 * , 2391 ( 2005 ) .
m. sakata , y. nakamoto , k. shimizu , t. matsuoka , y. ohishi , phys .
b * 83 * , 220512(r ) ( 2011 ) .
a.m. teweldeberhan , j.l .
dubois , s.a .
bonev , phys .
lett . * 105 * , 235503 ( 2010 ) .
y. nakamoto , m. sakata , k. shimizu , h. fujihisa , t. matsuoka , y. ohishi , t. kikegawa , phys .
b * 81 * , 140106 ( 2010 ) .
mao , l.wang , y. ding , w. yang , w. liu , d.y .
kim , w. luo , r. ahuja , y. meng , s. sinogeikin , j. shu , h.k .
mao , proc .
usa . * 107 * , 9965 ( 2010 ) .
i. errea , b. rousseau , a. bergara , int
. j. high pressure res .
* 28 * , 443 ( 2008 ) .
y. yao , r. martonak , s. patcjkovskii , d.d .
klug , phys .
b * 82 * , 094107 ( 2010 ) .
h. fujihisa , y. nakamoto , k. shimizu , t. yabuuchi , y. gotoh , phys .
lett * 101 * , 095503 ( 2008 ) .
y. nakamoto , m. sakata , k. shimizu , h. fujihisa , t. matsuoka , y. ohishi , t. kikegawa , phys .
b * 81 * , 140106(r ) ( 2010 ) .
s. arapan , h.k .
mao , r. ahuja , proc .
* 105 * , 20627 ( 2008 ) .
t. ishikawa , h. nagara , n. suzuki , t. tsuchiya , j. tsuchiya , phys .
b * 81 * , 092104 ( 2010 ) . k.j .
dunn , f.p .
bundy , phys . rev .
b * 24 * , 1643 ( 1981 ) .
s. okada , k. shimizu , t.c .
kobayashi , k. amaya , s. endo , j. phys .
. jpn . * 65 * , 1924 ( 1996 ) .
t. yabuuchi , t. matsuoka , y. nakamoto , k. shimizu , j. phys .
. jpn . * 75 * , 083703 ( 2006 ) .
m. andersson , phys .
b * 84 * , 216501 ( 2011 ) .
yin , f. gygi , w.e .
pickett , phys .
b * 80 * , 184515 ( 2009 ) .
m. aftabuzzaman , a.k.m.a .
islam , j. phys . : condens .
matter * 23 * , 105701 ( 2011 ) .
for discussion of the eliashberg equations [ originally formulated by g.m .
eliashberg , soviet .
jetp * 11 * , 696 ( 1960 ) ] we refer to : + ( a ) p.b .
allen , b. mitrovi , in : _ solid state physics : advances in research and applications _ , edited by h. ehrenreich ,
f. seitz , d. turnbull , ( academic , new york , 1982 ) , vol 37 , p. 1 ; + ( b ) j.p .
carbotte , f. marsiglio , in : the physics of superconductors , edited by k.h .
bennemann , j.b .
ketterson , ( springer , berlin , 2003 ) , vol 1 , p. 223 .
( a ) r. szczniak , acta phys .
pol . a * 109 * , 179 ( 2006 ) ; + ( b ) r. szczniak , solid state commun . *
138 * , 347 ( 2006 ) .
( a ) r. szczniak , physica status solidi ( b ) . * 244 * , 2538 ( 2007 ) ; + ( b ) r. szczniak , solid state commun . *
144 * , 27 ( 2007 ) .
( a ) r. szczniak , solid state commun . * 145 * , 142 ( 2008 ) ; + ( b ) r. szczniak , m.w . jarosik , solid state commun . * 149 * , 2053 ( 2009 ) ; + ( c ) r. szczniak , phys . lett . a * 373 * , 473 ( 2009 ) . ( a )
jarosik , r. szczniak , d. szczniak , acta phys . pol .
a * 118 * , 1031 ( 2010 ) ; + ( b ) m.w .
jarosik , r. szczniak , acta phys . pol .
a * 119 * , 863 ( 2011 ) .
( a ) r. szczniak , m.w .
jarosik , physica b * 406 * , 2235 ( 2011 ) ; +
( b ) r. szczniak , m.w .
jarosik , physica b * 406 * , 3493 ( 2011 ) .
carbotte , rev .
62 * , 1027 ( 1990 ) .
jishi , m. benkraouda , j. bragin , j. low temp . phys . * 147 * , 549 ( 2007 ) .
g. profeta , c. franchini , n.n .
lathiotakis , a. sanna , m.a.l .
marques , m. luders , s. massidda , e.k.u . gross , a. continenza , phys .
lett . * 96 * , 047003 ( 2006 ) .
m. luders , m.a.l .
marques , a. floris , g. profeta , n.n .
lathiotakis , c. franchini , a. sanna , a. continenza , s. massidda , e.k.u .
gross , psik newsletter * 76 * ( 2006 ) scientific highlight of the month .
r. szczniak , m.w .
jarosik , d. szczniak , physica b * 405 * , 4897 ( 2010 ) .
r. szczniak , d. szczniak , solid state communications * 152 * , 779 ( 2012 ) .
p. morel , p.w .
anderson , phys . rev .
* 125 * , 1263 ( 1962 ) .
j. bauer , j.e .
han , o. gunnarsson , http://arxiv.org/abs/1202.5051 y. xie , a.r .
oganov , y. ma , phys .
lett . * 104 * , 177005 ( 2010 ) .
( a ) s.y .
savrasov , phys .
b * 54 * , 16470 ( 1996 ) ; + ( b ) s.y .
savrasov , d.y .
savrasov , phys .
b * 54 * , 16487 ( 1996 )
mcmillan , phys .
rev . * 167 * , 331 ( 1968 ) .
allen , r.c .
dynes , phys .
b * 12 * , 905 ( 1975 ) .
s. lei , d.a .
papaconstantopoulos , m.j .
mehl , phys .
rev b * 75 * , 024512 ( 2007 ) .
k.s.d . beach , r.j .
gooding , f. marsiglio , phys .
b * 61 * , 5147 ( 2000 ) . ( a )
j. bardeen , l.n .
cooper , j.r .
schrieffer , phys . rev . * 106 * , 162 ( 1957 ) ; + ( b ) j. bardeen , l.n .
cooper , j.r .
schrieffer , phys .
108 * , 1175 ( 1957 ) . | the thermodynamic parameters of the superconducting state in calcium under the pressure at @xmath0 gpa have been calculated within the framework of the eliashberg approach .
it has been shown that the value of the coulomb pseudopotential is high ( @xmath1 ) and the critical temperature ( @xmath2 k ) should be determined from the modified allen - dynes formula .
in addition , it has been found that the basic dimensionless ratios of the thermodynamic parameters significantly diverge from the bcs predictions , and take the following values : ( i ) the zero temperature energy gap to the critical temperature ( @xmath3 ) is equal to 4.01 .
( ii ) the ratio @xmath4 equals @xmath5 , where @xmath6 and @xmath7 denote the specific heats for the superconducting and normal state , respectively . (
iii ) the quantity @xmath8 , where @xmath9 indicates the thermodynamic critical field .
finally , it has been proven that the electron effective mass is large and takes the maximum of @xmath10 at @xmath11 .
# 1([#1 ) ] # 1[#1 ] * keywords : * ca - superconductor , high - pressure effects , thermodynamic properties . |
the large magellanic cloud ( lmc ) is the nearest galaxy to the milky way ( with the exception of a merging dwarf galaxy ) , one in which individual stars can be distinguished and studied fairly easily with instruments of the highest angular resolution such as the hubble space telescope ( hst ) .
the lmc is different from the milky way ( mw ) in that it shows more intense star formation ( sf ) , at least when compared with the solar neighborhood , _
i.e. , _ the mw region within @xmath02 kpc of the sun where the interstellar extinction is reasonably small and individual stars can be easily studied in the visible region of the spectrum .
the difference in sf , coupled with the much smaller size of the lmc compared with the mw , implies that this is a dwarf irregular or a tidally truncated small spiral , in which the sf processes are probably different from those in the mw .
feitzinger ( 1987 ) argued that the lmc is a good example of a galaxy in which the sf proceeds primarily via the stochastic self - propagating sf mechanism ( sspsf : gerola & seiden 1978 ) , whereas the conventional view ( _ e.g. , _ kaufman 1979 ) is that the sf in the mw is driven mainly by spiral density waves .
the different sf mechanism , together with the possibility of studying individual stars , are the reasons the lmc is a very popular target for studies of sf - related phenomena ( _ e.g. , _
geha 1998 , battinelli & demers 1998 ) .
there is good reason to study the population of field stars in the lmc ; these objects represent the unspectacular star formation processes in the lmc and , by inference , in dwarf irregular galaxies in general . if we manage to understand the `` quiet '' mode of star formation , _
i.e. , _ that which is not generated in a starburst , we may better understand the origin of the uv radiation in the universe .
note that in starburst galaxies only @xmath020% of the uv light at 2200 is produced by stars in clusters ( meurer 1995 ) and the majority of the light originates from a general population of uv stars .
this diffuse uv emission might be due to the stars that have been ejected from regions of active sf .
studies of the field star population of the lmc were reported by _
e.g. , _ elson ( 1997 ) , geha ( 1998 ) .
these were performed with hst in optical and near - infrared bands , being rather insensitive to massive / hot stars .
the stellar populations of the lmc were reviewed recently by feast ( 1995 ) , who did not find strong evidence for a starburst triggered by a lmc - smc - milky way interaction .
he argued that the sf increase in the lmc was caused by the collapse of the system to a plane about 4 gyrs ago .
the nature of the fainter ultraviolet ( hereafter uv , covering the approximate wavelength range @xmath51000 - 3000 ) sources has not yet been resolved .
this is in part due to the fact that since the td-1 mission in the late 1960s ( thompson 1978 ) there has not been an all - sky survey in the uv
. there has also not been an instrument capable of providing good imaging in the uv , from which photometric parameters and shape information on extended objects could be derived .
previous deep uv observations have been perfomed by the uit shuttle - borne telescope ( stecher 1992 ) and by the foca balloon - borne telescope ( milliard 1992 ) .
the latter are in a spectral band centered at 2015 and @xmath6 wide ; the width of the band depends essentially on the altitude of the balloon which carried the foca telescope to 42 + km altitude .
because of this , the foca passband has a slight red leak ( @xmath7 ) , which depends on the exact altitude of the instrument during the observation .
the results from the diverse foca flights have not yet been published in a systematic manner , as the ground - based follow - up is still in its early phases .
the few published foca results , supplemented by ground - based follow - up observations from large telescopes ( _ e.g. , _ treyer 1998 ) , are sufficiently intriguing to warrant concerted attempts to observe faint uv sources with good angular resolution .
these results indicate that @xmath050% of the uv sources in a field centered on the coma cluster are background galaxies ranging to [email protected] with strong emission lines .
similar results were announced by the group headed by r. ellis ( treyer 1998 ) .
other uv imaging instruments , such as s201 ( carruthers 1977 ) or faust ( bowyer 1995 ) , had much lower angular resolution than either uit ( @xmath03 arcsec ) or foca ( @xmath010 arcsec ) and did not reach objects as faint as these
. however , these low angular resolution imagers offer unique panoramic views of the lmc in the uv .
previous uv imaging observations of the lmc were reported by page & carruthers ( 1981 ) using the s201 camera carried to the moon s surface during the apollo 16 flight ; by courts ( 1984 ) with the wide - field camera operated on the space shuttle ; by smith ( 1987 ) using a rocket - borne uit prototype ; and by courts ( 1995 ) with the faust instrument .
in addition , observations in the uv near 30 dor with photometers were reported by koornneef ( 1977 ) using the ans satellite with five uv bands and with an angular resolution of 2.5 arcmin ; by nandy ( 1979 ) and morgan ( 1979 ) from the td-1 satellite with four bands and with 1.7 arcmin resolution ; and on stellar assocations in the lmc with the skylab s183 experiment ( vuillemin 1988 ) .
most lmc imaging in the uv was done with very wide field cameras and with low angular resolution , of order 3 - 5 arcmin . despite the low resolution , which allows only for the mapping of the general uv emission pattern , these observations generated quantitative measures of the emission from unresolved sources and diffuse nebulosity in the lmc .
in particular , page & carruthers ( 1981 ) measured the total uv emission from the lmc to be s@xmath9=3.4 10@xmath10 erg sec@xmath11 @xmath12 @xmath11 ster@xmath11 ( at 1400 ) and s@xmath13=3.8 10@xmath10 erg sec@xmath11 @xmath12 @xmath11 ster@xmath11
( at 1300 ) .
the 30 doradus and sn 1987a regions in the lmc were observed with higher angular resolution by the uit and the results were published by cheng ( 1992 ) .
the latter observation is relevant , because it relates to a region of the lmc very close to that studied here .
similar uit observations , of the associations lh 52 and lh 53 , and the snr n49 in the lmc , were reported by hill ( 1995 ) .
the field population of uv stars was studied by hill ( 1994 ) , who found a lower star formation ( sf ) rate there than in the stellar associations of the lmc , as well as a sf less biased toward high - mass objects .
finally , all the uv sources in the direction of the lmc which have been observed by uit have been collected in a catalog by parker ( 1998 ) .
the results from the two uit flights constitute the deepest catalog of uv objects in the lmc published to date ( parker 1998 ) .
the uit observations yielded photometry from images with reasonable resolution : fwhm=3.4 arcsec . a comparison of the uit - derived magnitudes and synthetic photometry for 255 stars in common with the iue observatory show no systematic shift , but a standard deviation of 0.25 mag in the distribution of magnitudes .
it is possible that many uv sources are cosmologically - important objects , as treyer ( 1998 ) have shown .
these sources may account for a large part of the extragalactic diffuse uv background ( duvb ) ; this may not vanish completely in directions with very low hi column densities , indicating that not all the duvb is dust - scattered galactic uv light ( leinert 1998 ) .
this motivated us to study the deepest uv observations obtained in parallel mode by the largest orbiting astronomical telescope , the hubble space telescope .
hst is equipped , in principle , with three uv - imaging instruments : wfpc2 , foc , and stis .
the second instrument has an extremely restricted field of view , while the third has a similarly small field but a much higher photometric accuracy , dynamic range , and throughput than the foc .
however , stis can not be used ( yet ) for parallel uv imaging observations because of the possible damage to the mama detectors by uv sources , and most of its pointings are in the direction of crowded , well - defined fields such as galactic nuclei and globular clusters where the sources are fainter than the damage limit .
it is clear that , at present , only wfpc2 can offer a reasonably large field combined with some uv throughput for the serendipitous study of the uv sky .
a number of papers have been published which analyse data collected with the wfpc2 uv filter set .
most of these filters have significant red leaks and the accuracy of the photometry can not always be fully estimated . for this reason ,
many researchers prefer to use the f160bw filter .
this `` wood s filter '' separates a pure uv band with an effective wavelength @xmath14=1491 and bandpass 446 ( holtzman 1995 : h95b ) .
the f160bw filter has a very low throughput ( approximately one order of magnitude below other uv filters from the wfpc2 complement ) ; however , its bandpass has no red leak , unlike other wfpc2 filters .
it also vignettes @xmath01/8 of the field imaged through each wf ccd ( the pc chip is not vignetted ) .
more details about this filter can be found in watson ( 1994 ) .
observations using f160bw to study hot stars in globular clusters were reported by _
e.g. , _ mould ( 1996 ) and cole ( 1997 ) . hunter ( 1997 ) studied the uv emission from another cluster ( r136 ) with the f170w filter , as did gilmozzi ( 1994 ) for the `` double cluster '' ngc 1850a+b ; the f170w filter has a very strong red leak and the interpretation of the results with it is difficult . in order to realize fully the uv imaging potential of the wfpc2+f160bw combination , a program of parallel imaging
was initiated by jm .
some archived observations obtained in this program are used here to study the field population of uv stars in the lmc .
we use the results to estimate the diffuse uv emission from the lmc and by inference , from other star - bursting dwarf galaxies , and to evaluate the degree of contamination by field uv stars of the pointed observations of globular and open clusters in the lmc .
our observations and reductions are described in section 2 and the photometry is discussed in section 3 .
optical counterparts are identified in section 4 , and their nature is discussed in section 5 .
the detection of a significant number of uv stars allows us to draw some general conclusions about the sf in the lmc .
the hst observed objects in the n132d region of the lmc with the faint object spectrograph from 23 to 26 august 1995 collecting data for proposal 5607 . during these very long pointings
the wfpc2 was commanded to observe in parallel with the f160bw filter .
these observations are ideal to study the population of hot stars in the crowded environment of the lmc . during the lmc session
the wfpc2 acquired 12 images with f160bw with exposure times of 2300 sec to 4900 sec , which we analyze here .
a log of the hst data sets used in this paper is given in table 1 .
the region observed by hst is centered at ( j2000 ) 5@xmath15 25@xmath16 , 69@xmath17 30 ; the foreground extinction in this direction is e(b
v)@xmath180.1 ( lucke 1974 ) .
the area is located near the southeast end of the lmc bar , not on the bar itself but rather on its northern edge , @xmath19 off its major axis , and @xmath20 off the lmc s kinematic center .
the list of uv - bright objects or regions in the lmc detected by s201 ( page & carruthers 1981 ) contains no object in this direction , and neither does the catalog of infrared sources in the lmc and smc ( schwering & israel 1990 ) .
the map of radio continuum produced by xu ( 1992 ) shows that at the location of the hst wfpc2 pointing there are no strong ridges of 6.3 cm continuum emission .
the map of diffuse h@xmath21 emission ( fig . 8 in xu 1992 ) shows that the line emission is also minimal in this location , but that 30 dor , about half a degree to the east , is very bright .
the region is one of those studied by davies ( 1976 ) with uk schmidt plates through a 100 wide h@xmath21 filter to search for nebulosities .
the n132d region is noted in their table ii along with n132h as being faint and having `` knots in envelope '' , a size of 7@xmath224 arcmin , and corresponding to the mc39 11 cm radio source .
mc39 is a radio and xray - detected supernova remnant ( clark 1982 ) , located @xmath010 arcmin south of the wfpc2 region imaged in the uv .
the wide field imaging by the s183 experiment indicates that the emission at 2600 ( 250 passband ) from the location of the hst pointings analyzed here is less than 2.69 10@xmath23 erg sec@xmath11 @xmath12 @xmath11 ( square arcmin)@xmath11 ( vuillemin 1988 , fig .
this figure also shows that the hst pointing was not towards any of the uv - bright stellar associations in the lmc . the image of the n132 field in fig .
3 of parker ( 1998 ) contains the hst field studied here in its lower - left ( north of the center ) part .
no exceptionally high stellar density is apparent , nor is this a very sparse field . a careful comparison between the [email protected] diameter image obtained by smith ( 1987 ) at 1590 and shown as the top figure of their plate 23 and the location of the n132 field in the r - band image shown by parker ( 1998 , fig .
2 ) emphasizes that the location of the hst region studied here is off the lmc bar .
although the diffuse uv emission in the vicinity of n132 seems to be higher than from the rest of the lmc disk , it is clear that the hst observations pointed to an area of unexceptional ( faint ) diffuse uv emission , between shapley constellations ii and v ( see fig . 4 of smith 1987 ) .
as it is not clear how good is the flat fielding of the very wide field image in smith ( 1987 ) , we can not remark on the absolute uv surface brightness from the location of hst observations .
the region observed with the hst is therefore typical of the `` field '' stellar population of the lmc , _
i.e. , _ not containing stellar associations or clusters , hii regions or known supernova remnants . it is , thus , representative of the field stellar population in the nearest dwarf galaxy which shows intense star formation .
the lmc images obtained by hst are shown in the mosaic of figure 1 , where the different frames have been combined using the astrometric parameters of the pc chip from set a as produced by the hst reduction pipeline .
the other images were linked into the mosaic through stars in common with previously linked images .
we indicate the scale of the figure with a 30 arcsec bar and the north and west directions by vectors plotted at the figure s lower right corner . the mosaic is shown here for display purposes only and it was not used in the photometric reduction .
note also that in its creation we made no attempt to remove distortions , which are small , in general , for the wfpc2 .
the images are grouped by coordinates in sets ( a through f ) , which were obtained in the same sky position , with the same guide stars , and at the same hst roll angle . using the task * gcombine * of iraf ( in stsdas.toolbox.imgtools ) , with the ccdcrrej option which is based on rejection of pixels higher by more than five standard deviations from their immediate surroundings , or with the combination of images based on crrej ( in stsdas.hst_calib.wfpc ) where pixels in stacked images are rejected by a similar algorithm , it is possible to eliminate most of the cosmic rays ( crs ) .
these are especially troublesome in very long wfpc2 exposures and it is virtually impossible to see any uv source before combining at least two images and rejecting the cr events . even with the ccdcrrej option enough traces of cr events remained in the combined frame , mainly at locations where one cr track crossed another , to significantly confuse the detection of faint objects .
this is because of the low throughput of the wfpc2+f160bw combination , the very low uv sky background in this spectral region , and the susceptibility of the wfpc2 ccds to cr events .
however , these cases could be rejected by blinking the combined images against individual frames used in the same combination . the case for dataset
f is particularly difficult to analyze , as it consists of a single long exposure with significant but not identical overlap with set e. for this dataset we identified only objects visible in the original frames which had the proper intensity profile of a star ( after some cr - cleaning with the * cosmicray * task in noao.imred.ccdred ) .
we used the combined frames to identify genuine uv sources by blinking combined sets of frames with significant overlap .
the objects recognized this way are listed with a double or triple entry in table 2 .
we accepted only those sources which appeared in a combined image , and which we could recognize on the two ( or three ) original images which made up the combined image and had the `` proper '' shape parameters ( [email protected] arcsec ) expected of genuine stellar images .
we calculated the celestial coordinates of all sources and performed aperture photometry with the * phot * package ( in noao.digiphot.apphot ) with a 0.5 arcsec round aperture .
this is justified , given the psf fwhm of images obtained with f160bw ( up to @xmath00.26 arcsec ; watson 1994 ) .
the * phot * package performs aperture photometry and subtracts the sky background as estimated from a ring around the aperture .
when performing aperture photometry , the undersampling of the wf psf is not important .
the photometry has not been corrected for charge transfer efficiency ( whitmore & heyer 1997 ) , because we found this to be insignificant in comparison with the other uncertainties in the data .
for the same reason we did not correct for psf variability .
we performed aperture corrections by subtracting 0.1 mag from the * phot * results .
we also corrected for the molecular contamination of the wfpc2 ( mackenty 1995 ) .
the photometry is reported in table 2 as _ monochromatic magnitudes _ at @xmath25=1491 , with @xmath26 and @xmath27 , where @xmath28 is the gain ratio and @xmath29 is the zero point constant from holtzman ( 1995 ) .
we did not specifically exclude sources in the areas vignetted by the f160bw filter .
these are included in the table and are marked by a * v * following the dataset and ccd chip identifiers given in column 6 . among the 341 independent measurements , only 28 are of objects in vignetted locations on the wf chips . in four cases
a source with multiple measurements appears vignetted in one image but is not vignetted on others ; this allows a comparison of photometry which helps evaluate the effect of vignetting . with one exception ( source no .
45 ) the two photometry measurements agree to within 0.27 mag or better .
we conclude that the vignetting by the f160bw filter does not appear to significantly influence our photometry .
we present the raw uv photometry results for all the objects identified in the wfpc2 images in table 2 .
each independent measurement is identified by a source number , which is the running number of uv sources and indicates the order in which the sources were identified .
some objects appear in more than one ( combined ) image ; for each measurement we present the derived value and the photometric error , the image set from which the particular result originated ( a , b , c , _ etc . _ ) , and the ccd chip in which the object appeared .
for example , the first object in table 2 was detected in the pc chip of the combined set a , at pixel coordinates ( 242.117 ; 260.523 ) and its raw uv magnitude was [email protected] .
the on - chip pixel coordinates are listed beside the celestial coordinates , to aid further investigations of the sources .
the correlation between different measurements was done through the celestial coordinates ( calculated with the iraf * metric * task , in stsdas.hst_calib.wfpc ) , recognizing two separate observations as belonging to the same object if their image centers were separated by @xmath180.5 arcsec . the photometric error listed in table 2
is that given by the * phot * task in iraf , _
i.e. , _ a formal error depending on the photon count and the local background ; this does not reflect the calibration error of the photometry from f160bw images or residual flat field errors .
note that the f160bw flat field correction can be quite noisy ( biretta & baggett 1998 ) . the calibration error , on the other hand
, is @xmath010% at the center of the ccd chips ( baggett 1998 ) .
we combined the different measurements of the same object in a single entry in table 3 , which has the source coordinates , uv magnitudes and errors .
we identify a source by its first listed source number in table 2 .
figure 2 shows the distribution of remeasurements of the same object ; approximately half of the objects have been measured more than once .
figure 3 shows the distribution of apparent uv magnitudes from table 3 .
objects with more than one measurement are shown in the lower panel . a formal comparison of the two magnitude distributions , _
e.g. , _ via a kolmogorov - smirnov test of the cumulative distributions , shows that they are not drawn from the same parent population .
this is probably the combination of differences introduced at the faint end of the distribution by a few spurious sources and at the bright end by the differnt areal coverage ; while singly - detected sources can appear anywhere on the mosaic in fig .
1 , the multiply - measured objects * must * reside in the region of image overlap , which is smaller than the total area observed by hst .
note also that faint sources may register on an individual image but not on another from a different image combination ; it would contribute a real source measured only once , although it may reside in an overlap region .
3 shows that ( a ) brighter objects are more frequently seen in multiple exposures , and ( b ) despite the caveat in the previous paragraph , the faint end of the distribution is not dominated by the singly - measured objects . in fact , the drop in the number - magnitude plots occurs in both histograms at m@xmath4=19 , indicating a similar degree of completeness for the singly - measured and multiply - measured objects .
it is possible that a few faint sources are spurious , but most represent real objects conservatively chosen while adopting stellar images as candidate objects .
we have likely missed some fainter lmc objects through this methodology , but we are fairly certain that we excluded virtually all spurious sources .
the number - magnitude diagrams indicate that our photometry becomes incomplete at [email protected] . comparing the independent measurements of the same objects , after correcting for wfpc2 contamination , we derive an unbiased estimator of the measurement error , which disregards calibration unknowns but which is more representative of the internal photometry error than the value given by *
phot*. this is the measurement error we adopt and present in table 3 . we show in figure 4 the distribution of the individual fuv measurements for objects which have been measured independently twice ( filled squares ) where the error bars are the formal * phot * errors of these points , and the average value adopted in table 3 ( empty diamonds ) with its error bar estimated from the dispersion of the individual measurements .
it is clear that the dispersion of the measurements is reasonably small for objects with m@xmath3119 , but increases for objects at m@xmath3219 to one or more magnitudes .
this demonstrates once again that the completeness limit of the uv observations reported here is near [email protected] .
we note that none of the @xmath0200 uv objects identified here appear convincingly extended when examined on the hst f160bw images .
we conclude tentatively that most of them are probably stars or starlike objects , and that in the direction of the small lmc area sampled by the observations there are no uv - bright background galaxies . the ultraviolet imaging telescope ( uit ) imaged regions of the lmc in fuv and nuv bands during the astro-1 flight .
results of these observations , pertaining to the stellar associations lh52 and lh53 , were published by hill ( 1995 ) .
the results concerning field stars in the lmc were discussed in the context of the imf in the field and in associations ( hill 1994 ) . despite the shallowness of the uit results , by @xmath03 mag relative to the hst data reported here , it is worthwhile first to consider the relative density of field stars .
we find @xmath0200 stars in 12 arcmin@xmath34 to m@xmath3319 , while hill ( 1994 ) find that about 70% of their 1563 stars , to m@xmath3315 , belong to the field population .
this yields a projected density of 0.9 star arcmin@xmath35 for stars brighter than m@xmath4=16 , from which we expect to find 10 stars brighter than m@xmath4=16 in our data .
instead , as fig .
5 demonstrates , we find almost four times this number of stars in our hst images .
an extensive catalog of uv stars in the two mcs , originating from both astro-1 ( 1990 ) and astro-2 ( 1995 ) observations , was published by parker ( 1998 ) .
the uit observations most relevant to the analysis presented here are toward n132 and were obtained during the astro-2 flight with the b1 filter ( @xmath36=1521 , @xmath37=354 ) .
this bandpass is very close to that of the hst wfpc2 f160bw , with a 30 difference in @xmath36 and the hst bandpass being wider by 92 than that of the present observations , therefore a direct comparison of measured magnitudes is justified .
the lmc area sampled by the hst f160bw mosaic is much smaller than the uit image and lies close to the northern edge of the n132 field .
the list of parker has been examined and the objects in common with our data set have been extracted and are presented in table 4 .
a comparison between the uit b1 magnitudes and those derived here for the same objects indicates that , with the exception of one object , the uit stars are reported brighter by @xmath01 magnitude than the hst photometry .
this could , in principle , be the effect of a consistent calibration error of the f160bw magnitudes relative to the calibration of uit , which is derived from objects in common with iue .
we deem it more likely to be the result of confusion contributions in the moderate angular resolution uit images .
in fact , in many cases we can identify the stars which would have been included as single stars in the uit photometry ; these additional contributors are listed on separate lines in table 4 , below the primary hst candidate counterpart . despite the inclusion of the additional stars , we were not able to reproduce the uit magnitudes by adding up their measured flux . obviously ,
more objects than the stars we detected must be included in order to account for the difference in magnitudes , from which one can assume that there are even more faint uv stars not detected in our wfpc2 images which crowd into the uit psf and are measured as a single star .
note that the uv survey to be conducted by galex ( bianchi & martin 1997 ) will have similar problems to those of uit in observing the lmc , as its angular resolution shall be only 3 - 5 arcsec .
these would only be more acute in moderately crowded fields , as galex is supposed to reach fainter uv magnitudes than did uit .
one basic requirement for discussing the nature of the fainter uv sources is the correspondence with an optical counterpart .
unfortunately , the f160bw observations in the direction of the lmc were not collected together with hst imaging with other filters , thus no images with comparable resolution in a different spectral band exist in the hst archives .
we attempted photometry on the scans of the second stsci digitized sky survey ( dss - ii ) , but the relatively large pixel size used for these scans , compounded with the plate scale of the schmidt plates , resulted in extreme crowding of the stellar images .
no photometry to a reasonable accuracy could be performed on the stars at the location of the hst region studied here using the dss scans .
we searched published catalogs with photometric information and could identify 58 uv sources in table 3 with objects listed in the roe / nrl catalog ( yentis 1992 ; wallin 1994 ) .
this catalog is derived from cosmos [email protected] scans of the short red ukst survey plates in the direction of the lmc , which were scanned with a very small laser beam , resulting in a smaller pixel size than the 15@xmath39 m of the dss - ii .
the astrometry is based on the hst guide star catalog and the positions are accurate to @xmath02 arcsec .
the roe / nrl catalog also lists magnitudes , which are in a band close to johnson r , have a typical accuracy of 0.2 mag , and reach objects as faint as @xmath021.5 mag .
our identification of an optical counterpart for a uv source required a positional coincidence within 2 arcsec ; this is justified , given the astrometric accuracy of the roe / nrl catalog , but may result in some chance coincidences .
the success rate in finding optical counterparts by cross - correlating with the roe / nrl catalog is 78% for objects with m@xmath4016 and drops to @xmath41% for m@xmath4218 .
the dependence of the success rate on the uv brightness , which should in principle be related to the optical brightness of the counterpart , indicates that most optical identifications are probably not chance effects .
however , the fact that we could not find optical counterparts for all the bright uv sources is puzzling .
one possibility , which we could not test , is that some possible counterparts were rejected by the roe automatic processing of the cosmos scans because of extreme crowding , thus they are not listed in the roe / nrl catalog .
we transformed approximately the roe / nrl catalog magnitudes to the v band to determine the nature of the lmc uv sources by assuming that the objects are earlier than mid - f and converting their uv - r color indices to uv - v .
the lmc objects for which we found optical counterparts in the roe / nrl catalog are listed in table 5 , along with their uv and optical data . the error in uv
v is obtained as the harmonic mean of the uv photometric error in table 3 and an assumed 0.2 mag error in v propagated from the `` short - red '' magnitude listed in the roe / nrl catalog and the adopted transformation .
we compare the properties of the lmc stars to stars in the milky way via an observational color - magnitude diagram based on high galactic latitude objects observed with the faust shuttle - borne telescope .
we elected to use observed photometry instead of convolving theoretical model atmosphere spectral energy distributions ( seds ) with the different instrumental responses , or using iue spectra ( _ e.g. , _
fanelli 1992 ) convolved similarly , because the uv photometric results are unbiased and uv - selected , just as the lmc sample is unbiased and uv - selected .
the comparison is made with stars observed by the faust experiment in the direction of the north galactic pole ( ngp ) , coma , and virgo ( brosch 1995 , 1997 , 1998 ) which have parallax measurements in the hipparcos and tycho catalogs .
we consider the lmc stars listed in table 5 to be at a distance modulus of 18.5 for the purpose of determining their absolute magnitudes , and plot them in the same diagram with the faust objects .
the faust observations are in a spectral band centered at @xmath43=1650 and @xmath44300 wide , which is not too different from the wfpc2+f160bw band definition ( @xmath36=1491 , @xmath44446 ) .
the two bands mostly overlap , with the hst one being wider and bluer than that of faust .
as both bands are not defined extremely accurately , we can not derive a color term to transform between the two .
it is possible that some chromatic effects may be present , mainly in the photometry of the high t@xmath45 stars which are very blue .
we note also that the faust observations are biased * against * the detection of hot main sequence stars because of the direction of observation ( high @xmath46b@xmath46 ) and the relative rarity of these stars in the immediate solar neighborhood where hipparcos parallaxes are available . for this reason , the earliest main sequence object we can locate on
the faust color - magnitude diagram is late - b .
the uv color - magnitude diagram shown in fig .
6 indicates that the lmc objects join up smoothly with the galactic uv stars in coma , virgo , and the north galactic pole region .
these have been identified in the publications mentioned above as main sequence stars . by inference
, we deduce that at least the 58 objects with roe / nrl catalog counterparts are also mostly main sequence objects .
note that the apparent width of the upper main sequence , as defined by the hst data from the lmc field , is similar to that of the lower main sequence defined by the faust stars in the mw .
figure 5 shows that the f160bw mini - survey of the lmc field becomes progressively incomplete for m@xmath47 .
considering the distance modulus , this indicates that we are observing lmc objects in the general field with m@xmath48 , but we detect objects as faint as m@xmath3 + 2.5 .
the faintest objects for which we have uv - optical color information may , therefore , be early - f main sequence stars .
we added the uv emission from all the detected stars and found that these account for 2.53 10@xmath49 erg sec@xmath11 @xmath12 @xmath11 for the entire area surveyed here .
this translates into 2.49 10@xmath10 erg sec@xmath11 @xmath12 @xmath11 ster@xmath11 at 1500 from stellar sources not in clusters or associations , and is consistent with the integrated lmc emission at 1400 from page & carruthers ( 1981 ) ; their value is higher by one third , but they include all uv sources , cluster and field , in their estimate .
the average uv brightness measured by td-1 ( morgan 1979 ) in the immediate vicinity of the hst - imaged region of the lmc is 6.8 10@xmath10 erg sec@xmath11 @xmath12 @xmath11 ster@xmath11 at 1550 .
this is higher by a factor of 2.7 than we measure , but the nearest td-1 measurement ( region iii ) lies @xmath020 arcmin south of the region studied here , deeper into the lmc bar . _ the calculation shows that the large part of the uv radiation in the lmc , at least for @xmath501500 , is produced by field stars and not by objects in clusters or associations .
_ this is consistent with the estimate by parker ( 1998 ) that @xmath060% of the uv light in the n132 image originates in stars fainter than 16 mag or from diffuse emission .
it is also consistent with the measurement of meurer ( 1995 ) that in starbursts , clusters of young stars provide at most 20% of the uv emission .
our integrated value for the 1500 emission , combined with the upper limit for the diffuse emission at this location and at 2600 from vuillemin ( 1988 ) , indicates a uv color index [ 1500][2600]@xmath510.26 for the integrated stellar population detected in the uv .
photometry of field uv sources with the faust experiment was done , as mentioned above , in a spectral band very similar to that defined by wfpc2 and f160bw . for comparison purposes ,
we use here the results for the ngp ( brosch 1995 ) and the virgo regions ( brosch 1997 ) , from which we selected only the stellar sources .
data from the coma field ( brosch 1998 ) is not used because this faust field is not representative of the `` general field '' of the high - latitude mw , as it contains the mel 111 open cluster . a comparison with the lmc stars detected here with the hst is valid , because with the sensitivity limit of faust it observed mostly galactic sources . at the ngp and virgo high ( northern ) galactic latitude of the faust exposures referred to here
we should have a fair indication of ( half ) the projected density of uv stars in the milky way ( mw ) . in these regions
faust detected 172 stars to a uv magnitude limit of @xmath014 and a completeness limit of @xmath012.5 .
the faust images considered here cover a solid angle of @xmath00.05 ster , which projects to @xmath01.7 10@xmath2 pc@xmath34 when viewed from outside the mw and adopting a 1 kpc scale height for the uv stars .
the coverage of the lmc f160bw mosaic is @xmath012 square arcmin , which corresponds to @xmath04400 pc@xmath34 at the distance of the lmc , thus about one quarter the projected area in the milky way sampled by faust .
therefore , the projected number density of uv stars in the hst observation of the lmc field region is [email protected] stars pc@xmath35 , whereas faust counted [email protected] stars pc@xmath35 in the mw . the latter value should be doubled to account for the full width of the galactic disk when viewed from outside .
it is possible to compare directly the faust results with those of the present hst observations .
while faust sampled the ( thick ) galactic disk and the halo , hst observed the edge of the bar and the disk of the lmc to a similar depth .
the reason is that the objects deteced by faust reach into the thick disk and the halo of the mw to @xmath01 kpc , while the hst observed objects at @xmath060 kpc ; the gain in sensitivity with hst makes up for the increased distance to yield approximately similar depths of observation .
thus , if the projected density of uv - bright stars in the lmc and in the disk and halo of the mw would have been the same , similar projected number densities of stars should have been observed , assuming that the lmc thickness is probably not much different than that of the mw .
however , note that the value of n@xmath54 for the lmc field is higher than that in the mw , as reflected by the faust data .
the spatial density of hot stars in the lmc field regime is apparently higher by a factor of 2 - 3 than that in the mw .
_ we emphasize that our observations indicate a significant contamination by young field stars in all cluster uv observations in the lmc .
_ this is especially important in case one aims to detect blue stragglers , or hot evolved stars which belong to the target cluster ( _ e.g. , _ mould 1996 , where four blue stragglers were identified ) ; if the detected objects have a projected density of @xmath017 arcmin@xmath35 to m@xmath319 they may well be only field interlopers , as already remarked by cole ( 1997 ) .
finally , we can use our data , in combination with the similar magnitude estimates by uit in n132 ( parker 1998 ) to produce a combined ( observational ) luminosity function for uv stars from m@xmath4=8 to 19 . for this purpose
we selected the n132 objects from the lmc catalog of parker and scaled our hst stellar uv magnitude distribution to the total number of stars in the uit sample with 12@xmath18m@xmath3115 .
this is the magnitude range where the uit star counts should be essentially complete and where no hst objects were probably missed .
there are 2271 uit stars in the n132 region , whereas there are only 17 hst stars in the same magnitude range , yielding a scale factor for the number count of 133.59 .
this is close to the ratio of sampled areas between the uit field and that of the hst mosaic ( @xmath0105 ) , providing some confidence for this scaling approach .
note that the possible @xmath0one magnitude discrepancy between the uit and hst magnitudes would not affect our conclusions significantly , because the slope of the uv luminosity function is defined mainly by its faint end , _
i.e. , _ by the hst measurements .
we show in figure 7 the two luminosity functions plotted on the same scale .
it is evident that the scaling procedure continues the trend of higher star counts to m@xmath318 , after which faint stars are progressively lost from the hst images .
figure 8 shows the logarithm of the combined star counts , where for m@xmath3113 we adoped the uit star counts and for the fainter magnitudes those from the ( scaled ) hst data set .
we compare the slope of this observed luminosity function with the models presented in parker ( 1998 ) for the uit star counts .
the models allow a comparison with an lmc metallicity population formed continuously over a period @xmath55 t with a salpeter imf , using fig .
11 from parker and comparing the predicted slope of the star count distribution between m@xmath4=10 and m@xmath4=15 with the same parameter from the actual , combined star count .
the closest fit is with a stellar population formed over @xmath55t@xmath241 gyr , because the younger star formation processes produce a distribution of the fainter uv stars which is much shallower than observed .
this is consistent with the finding of geha ( 1998 ) that the sf rate in the lmc did not change appreciably in the last 2 gyr .
it is similarly possible to compare the slope of predicted luminosity functions for a continuous star formation over 1 gyr , with different imf slopes ( fig .
14 of parker ) , with that actually observed .
the comparison indicates that the parker ( 1998 ) model with the slope most similar to that measured here is that with an initial mass function slope @xmath56 of 1.8 , close to the salpeter value and found also to best fit the uit data .
although the data do not really warrant it , we note that the measured slope is smaller than that predicted by the models with @xmath56=1.8 but higher that that for @xmath56=1.0 .
thus , a salpeter slope with @xmath56=1.4 would probably be better than the 1.8 value found to best fit the bright end of the uv luminosity function by parker
we analyzed a set of very deep hst images obtained in a uv spectral band which is not contaminated by visible light leaks .
our findings can be summarized as follows : 1 .
we identified 198 uv sources with stars in the `` general field '' of the lmc , at the edge of the bar and not associated with any known star cluster or association .
2 . a comparison with galactic objects for the 30% of the sources for which we found optical catalog counterparts indicates that the objects are main sequence stars .
the uv emission from the stellar sources we detected , scaled to the entire solid angle covered by the lmc , accounts for 2/3 of the uv emission from the cloud .
the observed uv luminosity function joins up smoothly with that defined for brighter objects , but on a wider field , from the uit observation of the n 132 region .
the joint luminosity function of uv stars , which covers the uv magnitude range 8 to 19 , confirms the claim by parker ( 1998 ) that the lmc `` field '' domain has been forming stars continuously over ( at least ) 1 gyr , with an imf close to the salpeter law .
uv research at tel aviv university is supported by grants from the ministry of science and arts through the israel space agency , from the austrian friends of tel aviv university , and from a center of excellence award from the israel science foundation .
nb acknowledges support from a us - israel binational award to study uv sources measured by the faust experiment , and partial support from the space telescope science institute during a sabbatical visit .
we thank stefano casertano for discussions on wfpc2 calibrations , and an anonymous referee for extensive remarks which improved significantly the presentation .
baggett , s. , caserano , s. & the wfpc2 group 1998 , stsci isr wfpc2 98 - 01 .
battineli , p. & demers , s. 1998 , aj , 115 , 1472 .
bianchi , l. & martin , c. 1997 in `` the ultraviolet astrophysics beyond the iue final archive '' ( r. gonzales - riestra , w. wamsteker & r.a .
harris , eds . ) , esa sp-413 .
biretta , j. & baggett , s. 1998 , stsci stan / wfpc2-no.33 , august .
bowyer , s. , sasseen , t.p . , wu , x. & lampton , m. 1995 , apjs , 96 , 461 .
brosch , n. , almoznino , e. , leibowitz , e.m . ,
netzer , h. , sasseen , t. , bowyer , s. , lampton , m. & wu , x. 1995 , apj 450 , 137 .
brosch , n. , formiggini , l. , almoznino , e. , sasseen , t. , lampton , m. & bowyer , s. 1997 , apjs , 111 , 143 .
brosch , n. , ofek , e.o . ,
almoznino , e. , sasseen , t. , lampton , m. & bowyer , s. 1998 , mnras , 295 , 959 .
carruthers , g.r .
& page , t. 1977 , apj , 211 , 728 .
cheng , k .-
, michalitsianos , a.g . , hintzen , p. bohlin , r.c . , oconnell , r.w . ,
cornett , r.h .
, morton , r.s . ,
smith , a.m. , smith , e.p .
& stecher , t.p . 1992 , apj , 395 , l29 .
clark , d.h .
, tuohy , i.r . ,
dopita , m.a . ,
mathewson , d.s .
, long , k.s . , szymkowiak , a.e .
& culhane , j.l .
1982 , apj , 225 , 440 .
cole , a.a .
, mould , j.r . , gallagher , j.s .
iii , clarke , j.t . ,
trauger , j.t . ,
ballester , g.e . ,
burrows , c.j . , casertano , s. , griffiths , r. , hester , j.j . ,
hoessel , j.g .
holtzman , j.a .
, scowen , p.a . ,
stapelfeldt , k.r .
& westphal , j.a .
1997 , aj , 114 , 1945 .
courts , g. , viton , m. , sivan , j .-
, decher , r. & gary , a. 1984 , science , 225 , 179 .
courts , g. , viton , m. , bowyer , s. , lampton , m. , sasseen , t.p . &
wu , x. 1995 , a&a , 297 , 338 .
davies , r.d . ,
elliott , k.h .
& meaburn , j. 1976 , mem .
, 81 , 89 .
elson , r.a.w . ,
gilmore , g.f . & santiago , b.x .
1997 , mnras , 289 , 157 .
fanelli , m.n .
, oconnell , r.w . , burstein , d. & wu , c .- c .
1992 , apjs , 82 , 197 . feast , m.w .
1995 in `` stellar populations '' ( p.c .
van der kruit & g. gilmore , eds . ) dordrecht ; kluwer academic publishers , pp . 153 - 163 .
feitzinger , j.v . ,
haynes , r.f . , klein , u. , wielebinski , r. & perschke , m. 1987 , vistas in astr . , 30 , 243 .
geha , m.c . , holtzman , j.a .
, mould , j.r . , gallagher , j.s.iii , watson , a.m. , cole , a.a . ,
grillmair , c.j . ,
staplefeld , k.r . ,
ballester , g.e . ,
burrows , c.j . , clarke , j.t .
, crisp , d. , evans , r.w .
, griffiths , r.e . , hester , j.j . ,
scowen , p.a . ,
trauger , j.t . & westphal , j.a .
1998 , aj , 115 , 1045 .
gerola , h. & seiden , p.e .
1978 , apj , 223 , 129 .
gilmozzi , r. , kinney , e.k . , ewald , s.p . ,
panagia , n. & romaniello , m. 1994 , apj , 435 , l43 .
hill , r.s . ,
isensee , j.e . , cornett , r.h . ,
bohlin , r.c .
, oconnell , r.w . ,
roberts , m.s . ,
smith , a.m. & stecher , t.p 1994 , apj , 425 , 122 .
hill , r.s . ,
cheng , k. -p . ; bohlin , r.c .
, oconnell , r.w . ,
roberts , m.s . ,
smith , a.m. & stecher , t.p . , 1995 , apj , 446 , 622 .
holtzman , j.a .
, burrows , c.j . , casertano , s. , hester , j.j . ,
trauger , j.t . , watson , a.m. & worthey , g. 1995 , pasp , 107 , 156 ( h95a ) .
holtzman , j.a . ,
burrows , c.j . , casertano , s. , hester , j.j . ,
trauger , j.t . ,
watson , a.m. & worthey , g. 1995 , pasp , 107 , 1065 ( h95b ) .
hunter , d.e . ,
vacca , w.d . ,
massey , p. , lynds , r. & oneil , e.j .
1997 , aj , 113 , 1691 .
kaufman , m. 1979 , apj , 232 , 707 .
koornneeff , j. 1977 , a&as , 29 , 117 .
leinert , ch . ,
bowyer , s. , haikala , l.k .
, hanner , m.s . ,
hauser , m.g .
, levasseur - regourd , a .- ch . ,
mann , i. , mattila , k. , reach , w.t . ,
schlosser , w. , staude , h.j . ,
toller , g.n . ,
weiland , j.l . ,
weinberg , j.l .
& witt , a.n .
1998 , a&as , 127 , 1 .
lucke , p.b .
1974 , apjs , 28 , 73 .
lucke , p.b . & hodge , p.w .
1971 , aj , 75 , 171 .
mackenty , j. w. , baggett , s.m . ,
biretta , j. , hinds , m. , ritchie , c.e .
, feinberg , l.d .
& trauger , j.t .
1995 , proc .
spie , 2478 , 160 .
meurer , g.r . ,
heckman , t.m .
, leitherer , c. , kinney , a. , robert , c. & garnett , d.r .
1995 , aj , 110 , 2665 .
milliard , b. donas , j. & laget , m. 1991 , adv .
space res . , 11 , 135 .
morgan , d.h . , nandi , k. & carnochan , d.j .
1979 , mnras , 188 , 131 .
mould , j.r . ,
watson , a.m. , gallagher , j.s .
iii , ballester , g.e . , burrows , c.j . , casertano , s. , clarke , j.t .
, crisp , d. , griffiths , r. , hester , j.j .
, hoessel , j.g .
holtzman , j.a .
, scowen , p.a . ,
stapelfeldt , k.r . ,
trauger , j.t . & westphal , j.a .
1996 , apj , 461 , 762 .
nandy , k. , morgan , d.h , & carnochan , d.j .
1979 , mnras , 186 , 421 .
page , t. & carruthers , g.r .
1981 , `` revised listing - s201 far - ultraviolet atlas of the large magellanic cloud '' , nrl memorandum report 4660 .
parker , j.wm . , hill , j.k . ,
cornett , r.h .
, hollis , j. , zamkoff , e. , bohlin , r.c . , oconnell , r.w . , neff , s.g .
, roberts , m.s . , smith , a.m. & stecher , t.p
1998 , aj , 116 , 180 .
schwering , p.b.w . &
israel , f.p .
1990 `` atlas and catalogue of infrared sources in the magellanic clouds '' , dordrecht : kluwer academic publishers .
smith , a.m. , cornett , r.h . & hill , r.s .
1987 , apj , 320 , 609 .
stecher , t.p . , baker , g.r .
1992 , apj , 395 , l1 .
treyer , m. , ellis , r.s . , milliard , b. , donas , j. & bridges , t.j .
1998 , mnras , in press ( astro - ph/9806056 ) .
thompson , g.i . ,
nandy , k. , jamar , c. , monfils , a. , houziaux , l. , carnochan , d.j . & wilson , r. 1978 `` catalog of stellar ultraviolet fluxes '' , src .
vuillemin , a. 1988 , a&as , 72 , 249 .
watson , a.m. , mould , j.r . , gallagher , j.s . ,
ballester , g.e . ,
burrows , c.j . , casertano , s. , clarke , j.t . , crisp , d. , griffiths , r.e . ,
hester , j. j.f , hoessel , j.g . ,
holtzman , j.a .
, scowen , p.a . ,
saplefeldt , k.r . ,
trauger , j.t . , westphal , j.a.199 , apj , 435 , 55 .
whitmore , b. & heyer , i. 1997 , instrument science report wfpc2 97 - 08 .
xu , c. , klein , u. , meinert , d. , wielebinski , r. & haynes , r.f .
1992 , a&a , 257 , 47 .
yentis , d.j . ,
cruddace , r.g . ,
gursky , h. , stuart , b.v . , wallin , j.f . ,
macgillivray , h.t . &
collins , c.a .
1992 , in `` digitized optical sky surveys '' ( h.t . macgillivray & e.b .
thomson , eds . ) , dordrecht : kluwer , p. 67 .
wallin , j.f .
, yentis , d.j . ,
macgillivray , h.t . ,
bauer , s.b . &
wong , c.s .
1994 , aas bull . , 184 , 2704
cccccc a & u2ou1001 t & 5:25:17.6 & 69:30:26 & 2500 & 230895 a & u2ou1101 t & 5:25:17.6 & 69:30:26 & 4400 & 230895 a & u2ou1201 t & 5:25:17.6 & 69:30:26 & 4400 & 230895 f & u2ou1301 t & 5:25:31.4 & 69:30:52 & 4400 & 230895 b & u2ou1501 t & 5:25:22.7 & 69:30:09 & 2300 & 260895 b & u2ou1502 t & 5:25:22.7 & 69:30:09 & 2300 & 260895 c & u2ou1601 t & 5:25:36.8 & 69:30:30 & 4900 & 260895 c & u2ou1701 t & 5:25:36.8 & 69:30:30 & 4400 & 260895 d & u2ou1e01 t & 5:25:17.8 & 69:30:30 & 4600 & 240895 d & u2ou1f01 t & 5:25:17.8 & 69:30:30 & 4600 & 240895 e & u2ou1g01 t & 5:25:31.8 & 69:30:51 & 4600 & 240895 e & u2ou1h01 t & 5:25:31.8 & 69:30:51 & 4600 & 240895 ccccccc 1 & 242.117 & 260.523 & 5:25:12.886 & 69:30:37.13 & a - pc & 19.190 ( 0.041 ) 1 & 140.889 & 329.237 & 5:25:12.881 & 69:30:37.22 & d - pc & 18.847 ( 0.035 ) 2 & 308.600 & 288.237 & 5:25:12.468 & 69:30:39.53 & a - pc & 18.112 ( 0.020 ) 2 & 205.689 & 361.042 & 5:25:12.461 & 69:30:39.65 & d - pc & 18.051 ( 0.021 ) 3 & 393.000 & 795.000 & 5:25:08.099 & 69:30:35.52 & a - pc & 19.072 ( 0.034 ) 4 & 442.279 & 556.773 & 5:25:09.893 & 69:30:41.22 & a - pc & 13.752 ( 0.002 ) 4 & 322.418 & 637.817 & 5:25:09.888 & 69:30:41.32 & d - pc & 13.444 ( 0.002 ) 5 & 531.523 & 665.313 & 5:25:08.753 & 69:30:43.39 & a - pc & 18.934 ( 0.033 ) 5 & 405.079 & 751.607 & 5:25:08.749 & 69:30:43.54 & d - pc & 18.779 ( 0.032 ) 6 & 597.824 & 177.715 & 5:25:12.537 & 69:30:53.58 & a - pc & 17.590 ( 0.010 ) 6 & 500.647 & 268.901 & 5:25:12.533 & 69:30:53.72 & d - pc & 17.279 ( 0.013 ) 7 & 373.000 & 066.000 & 5:25:14.509 & 69:30:50.67 & d - pc & 16.583 ( 0.009 ) 8 & 758.794 & 236.243 & 5:25:12.185 & 69:31:05.32 & d - pc & 19.942 ( 0.081 ) 10 & 095.365 & 263.299 & 5:25:15.811 & 69:30:07.98 & a - wf2 & 18.220 ( 0.020 ) 10 & 106.265 & 306.270 & 5:25:15.812 & 69:30:08.10 & d - wf2 & 18.086 ( 0.019 ) 11 & 152.362 & 529.463 & 5:25:16.430 & 69:29:41.14 & a - wf2 & 16.467 ( 0.008 ) 11 & 146.375 & 575.363 & 5:25:16.436 & 69:29:41.27 & d - wf2 & 16.221 ( 0.007 ) 12 & 182.786 & 067.505 & 5:25:15.895 & 69:29:40.17 & a - wf2 & 19.190 ( 0.036 ) 12 & 205.649 & 116.813 & 5:25:13.054 & 69:30:23.56 & d - wf2 & 18.972 ( 0.034 ) 13 & 201.310 & 428.790 & 5:25:14.938 & 69:29:49.07 & a - wf2 & 15.717 ( 0.006 ) 13 & 201.505 & 477.811 & 5:25:14.953 & 69:29:49.10 & d - wf2 & 15.484 ( 0.005 ) 14 & 261.667 & 457.343 & 5:25:14.043 & 69:29:44.40 & a - wf2 & 15.525 ( 0.005 ) 14 & 260.222 & 509.980 & 5:25:14.040 & 69:29:44.47 & d - wf2 & 15.318 ( 0.005 ) 15 & 292.361 & 174.331 & 5:25:11.744 & 69:30:10.03 & a - wf2 & 19.755 ( 0.052 ) 15 & 309.946 & 229.913 & 5:25:11.726 & 69:30:10.14 & d - wf2 & 18.927 ( 0.032 ) 16 & 317.258 & 626.340 & 5:25:14.077 & 69:29:26.70 & a - wf2 & 19.116 ( 0.041 ) 16 & 305.174 & 682.373 & 5:25:14.078 & 69:29:26.78 & d - wf2 & 18.845 ( 0.042 ) 17 & 330.646 & 558.050 & 5:25:13.428 & 69:29:32.66 & a - wf2 & 17.274 ( 0.013 ) 17 & 322.711 & 614.787 & 5:25:13.431 & 69:29:32.86 & d - wf2 & 17.088 ( 0.012 ) 18 & 336.000 & 789.620 & 5:25:14.726 & 69:29:10.86 & a - wf2 & 17.309 ( 0.016 ) 19 & 346.644 & 172.694 & 5:25:10.765 & 69:30:08.47 & a - wf2 & 19.033 ( 0.032 ) 19 & 362.435 & 231.432 & 5:25:10.769 & 69:30:08.54 & d - wf2 & 18.731 ( 0.029 ) 20 & 531.974 & 152.208 & 5:25:07.318 & 69:30:04.32 & a - wf2 & 18.099 ( 0.019 ) 20 & 548.410 & 222.324 & 5:25:07.333 & 69:30:04.46 & d - wf2 & 17.855 ( 0.017 ) 21 & 714.051 & 479.000 & 5:25:06.085 & 69:29:27.74 & a - wf2 & 17.795 ( 0.021 ) 21 & 710.531 & 559.679 & 5:25:06.095 & 69:29:27.93 & d - wf2 v & 17.718 ( 0.030 ) 22 & 719.335 & 417.689 & 5:25:05.614 & 69:29:33.39 & a - wf2 & 21.144 ( 0.282 ) 22 & 729.462 & 514.000 & 5:25:05.523 & 69:29:31.72 & d - wf2 & 17.370 ( 0.021 ) 23 & 750.094 & 434.000 & 5:25:05.170 & 69:29:30.81 & a - wf2 & 19.497 ( 0.084 ) 23 & 716.000 & 499.988 & 5:25:05.687 & 69:29:33.40 & d - wf2 & 19.820 ( 0.134 ) 24 & 089.643 & 459.087 & 5:25:17.123 & 69:29:49.75 & a - wf2 & 19.532 ( 0.045 ) 24 & 088.011 & 501.180 & 5:25:17.119 & 69:29:49.89 & d - wf2 & 19.260 ( 0.039 ) 25 & 645.560 & 028.878 & 5:25:04.619 & 69:30:20.24 & d - wf2 & 15.771 ( 0.006 ) 26 & 771.340 & 076.698 & 5:25:02.585 & 69:30:12.39 & d - wf2 & 17.360 ( 0.015 ) 31 & 054.017 & 397.253 & 5:25:21.550 & 69:30:41.15 & a - wf3 & 17.643 ( 0.015 ) 31 & 072.656 & 372.330 & 5:25:21.545 & 69:30:41.39 & d - wf3 & 17.381 ( 0.013 ) 31 & 178.305 & 274.530 & 5:25:21.556 & 69:30:40.94 & b - wf4 & 17.330 ( 0.015 ) 32 & 067.471 & 563.700 & 5:25:24.590 & 69:30:45.32 & a - wf3 & 15.895 ( 0.006 ) 32 & 075.996 & 539.289 & 5:25:24.602 & 69:30:45.46 & d - wf3 & 15.672 ( 0.006 ) 32 & 344.651 & 269.488 & 5:25:24.597 & 69:30:45.08 & b - wf4 & 15.584 ( 0.006 ) 33 & 091.434 & 478.313 & 5:25:23.224 & 69:30:40.35 & a - wf3 & 18.351 ( 0.022 ) 33 & 104.822 & 455.424 & 5:25:23.223 & 69:30:40.48 & d - wf3 & 18.103 ( 0.020 ) 33 & 260.864 & 241.980 & 5:25:23.220 & 69:30:40.07 & b - wf4 & 17.907 ( 0.020 ) 34 & 275.671 & 633.669 & 5:25:27.136 & 69:30:28.18 & a - wf3 & 18.748 ( 0.033 ) 34 & 279.313 & 621.811 & 5:25:27.154 & 69:30:28.35 & d - wf3 & 18.488 ( 0.030 ) 34 & 425.554 & 065.323 & 5:25:27.152 & 69:30:27.96 & b - wf4 & 18.497 ( 0.032 ) 35 & 303.764 & 447.109 & 5:25:23.991 & 69:30:19.50 & a - wf3 & 18.142 ( 0.020 ) 35 & 318.739 & 437.390 & 5:25:23.987 & 69:30:19.67 & d - wf3 & 17.937 ( 0.019 ) 120 & 043.898 & 242.704 & 5:25:23.966 & 69:30:19.70 & b - wf3 & 17.988 ( 0.024 ) 36 & 361.684 & 320.639 & 5:25:22.082 & 69:30:09.89 & a - wf3 & 18.525 ( 0.025 ) 36 & 384.462 & 314.714 & 5:25:22.079 & 69:30:10.12 & d - wf3 & 18.351 ( 0.024 ) 121 & 109.647 & 119.213 & 5:25:22.071 & 69:30:10.18 & b - wf3 & 18.411 ( 0.029 ) 37a & 453.913 & 587.461 & 5:25:27.433 & 69:30:09.89 & a - wf3 & 19.250 ( 0.054 ) 37a & 459.992 & 586.638 & 5:25:27.429 & 69:30:10.07 & d - wf3 & 18.830 ( 0.044 ) 122 & 187.228 & 392.523 & 5:25:27.417 & 69:30:10.17 & b - wf3 & 19.176 ( 0.054 ) 37b & 457.481 & 572.207 & 5:25:27.185 & 69:30:09.12 & a - wf3 & 21.788 ( 0.498 ) 37b & 463.361 & 566.679 & 5:25:27.081 & 69:30:09.24 & d - wf3 & 20.348 ( 0.153 ) 38a & 466.876 & 556.314 & 5:25:26.955 & 69:30:07.75 & a - wf3 & 19.474 ( 0.064 ) 38a & 474.728 & 557.458 & 5:25:26.974 & 69:30:07.95 & d - wf3 & 19.555 ( 0.077 ) 38b & 471.438 & 549.874 & 5:25:26.862 & 69:30:07.05 & a - wf3 & 20.002 ( 0.098 ) 38b & 483.000 & 551.000 & 5:25:26.911 & 69:30:06.92 & d - wf3 & 21.066 ( 0.297 ) 39 & 535.855 & 227.626 & 5:25:21.510 & 69:29:50.51 & a - wf3 & 19.375 ( 0.044 ) 39 & 563.767 & 232.487 & 5:25:21.514 & 69:29:50.69 & d - wf3 & 19.160 ( 0.044 ) 40 & 551.226 & 293.726 & 5:25:22.788 & 69:29:51.16 & a - wf3 & 19.865 ( 0.066 ) 40 & 574.895 & 299.707 & 5:25:22.813 & 69:29:51.36 & d - wf3 & 19.649 ( 0.066 ) 126 & 301.585 & 104.793 & 5:25:22.770 & 69:29:51.48 & b - wf3 & 19.584 ( 0.069 ) 41 & 591.000 & 588.000 & 5:25:28.300 & 69:29:57.07 & a - wf3 & 18.666 ( 0.042 ) 41 & 594.612 & 595.292 & 5:25:28.275 & 69:29:57.50 & d - wf3 & 20.071 ( 0.159 ) 42 & 612.000 & 447.605 & 5:25:25.918 & 69:29:50.48 & a - wf3 & 17.370 ( 0.015 ) 42 & 622.056 & 472.000 & 5:25:26.184 & 69:29:51.50 & d - wf3 & 18.517 ( 0.036 ) 43 & 612.000 & 379.000 & 5:25:24.704 & 69:29:48.25 & a - wf3 & 17.819 ( 0.018 ) 43 & 639.653 & 383.644 & 5:25:24.652 & 69:29:47.48 & d - wf3 & 18.228 ( 0.028 ) 44 & 607.375 & 335.463 & 5:25:23.887 & 69:29:47.28 & a - wf3 & 17.379 ( 0.014 ) 44 & 628.585 & 344.541 & 5:25:23.886 & 69:29:47.48 & d - wf3 & 17.239 ( 0.014 ) 127 & 355.420 & 150.181 & 5:25:23.879 & 69:29:47.56 & b - wf3 & 17.319 ( 0.014 ) 45 & 708.733 & 534.000 & 5:25:28.060 & 69:29:44.36 & a - wf3 & 18.843 ( 0.067 ) 45 & 713.612 & 546.306 & 5:25:27.988 & 69:29:44.84 & d - wf3 v & 21.214 ( 0.713 ) 46 & 793.000 & 300.134 & 5:25:24.429 & 69:29:28.83 & a - wf3 & 18.695 ( 0.042 ) 51 & 047.224 & 456.328 & 5:25:12.589 & 69:31:10.06 & a - wf4 & 18.857 ( 0.064 ) 52 & 084.929 & 118.391 & 5:25:15.438 & 69:30:39.84 & a - wf4 & 19.605 ( 0.098 ) 52 & 053.441 & 080.329 & 5:25:15.422 & 69:30:39.98 & d - wf4 & 18.972 ( 0.041 ) 53 & 094.265 & 149.601 & 5:25:15.396 & 69:30:43.01 & a - wf4 & 14.354 ( 0.008 ) 53 & 060.674 & 112.426 & 5:25:15.375 & 69:30:43.18 & d - wf4 & 14.151 ( 0.003 ) 104 & 792.612 & 429.584 & 5:25:15.426 & 69:30:43.56 & b - pc & 14.212 ( 0.003 ) 54 & 098.477 & 754.720 & 5:25:11.608 & 69:31:39.53 & a - wf4 & 19.469 ( 0.097 ) 55 & 109.924 & 713.486 & 5:25:12.074 & 69:31:36.14 & a - wf4 & 18.812 ( 0.067 ) 56 & 135.128 & 552.157 & 5:25:13.531 & 69:31:21.99 & a - wf4 & 16.911 ( 0.026 ) 56 & 077.397 & 516.988 & 5:25:13.519 & 69:31:22.15 & d - wf4 & 16.939 ( 0.011 ) 57 & 143.718 & 188.409 & 5:25:16.011 & 69:30:48.25 & a - wf4 & 17.467 ( 0.034 ) 57 & 108.329 & 153.795 & 5:25:16.019 & 69:30:48.35 & d - wf4 & 17.254 ( 0.013 ) 58 & 206.444 & 734.497 & 5:25:13.637 & 69:31:41.34 & a - wf4 & 19.138 ( 0.080 ) 58 & 159.454 & 706.861 & 5:25:14.009 & 69:31:42.41 & d - wf4 & 20.012 ( 0.093 ) 59 & 206.333 & 553.424 & 5:25:14.783 & 69:31:24.48 & a - wf4 & 19.166 ( 0.078 ) 59 & 148.403 & 522.668 & 5:25:14.773 & 69:31:24.61 & d - wf4 & 18.894 ( 0.032 ) 60 & 210.734 & 485.336 & 5:25:15.289 & 69:31:18.25 & a - wf4 & 18.633 ( 0.059 ) 60 & 157.354 & 454.904 & 5:25:15.296 & 69:31:18.37 & d - wf4 & 18.395 ( 0.024 ) 61 & 227.305 & 219.682 & 5:25:17.303 & 69:30:53.94 & a - wf4 & 14.985 ( 0.011 ) 61 & 189.720 & 190.583 & 5:25:17.285 & 69:30:54.08 & d - wf4 & 14.857 ( 0.004 ) 62 & 253.328 & 175.909 & 5:25:18.043 & 69:30:50.80 & a - wf4 & 17.814 ( 0.040 ) 62 & 218.465 & 148.465 & 5:25:18.035 & 69:30:50.90 & d - wf4 & 17.585 ( 0.015 ) 63 & 257.563 & 532.684 & 5:25:15.824 & 69:31:24.24 & a - wf4 & 18.909 ( 0.068 ) 63 & 200.553 & 503.562 & 5:25:15.815 & 69:31:24.25 & d - wf4 & 19.600 ( 0.054 ) 64 & 342.984 & 796.393 & 5:25:15.676 & 69:31:51.71 & a - wf4 & 16.687 ( 0.024 ) 64 & 269.822 & 773.505 & 5:25:15.667 & 69:31:51.87 & d - wf4 & 16.149 ( 0.008 ) 65 & 367.672 & 614.385 & 5:25:17.258 & 69:31:35.62 & a - wf4 & 18.945 ( 0.072 ) 65 & 305.810 & 593.494 & 5:25:17.244 & 69:31:35.77 & d - wf4 & 18.705 ( 0.032 ) 66 & 332.477 & 179.500 & 5:25:19.430 & 69:30:53.73 & a - wf4 & 19.378 ( 0.086 ) 66 & 297.485 & 156.963 & 5:25:19.421 & 69:30:53.93 & d - wf4 & 19.060 ( 0.036 ) 67 & 380.004 & 751.588 & 5:25:16.617 & 69:31:48.79 & a - wf4 & 18.907 ( 0.074 ) 67 & 309.666 & 731.261 & 5:25:16.591 & 69:31:48.99 & d - wf4 & 18.739 ( 0.039 ) 68 & 458.448 & 160.663 & 5:25:21.800 & 69:30:56.21 & a - wf4 & 20.619 ( 0.169 ) 68 & 424.491 & 145.929 & 5:25:21.791 & 69:30:56.33 & d - wf4 & 20.102 ( 0.080 ) 69 & 473.572 & 554.724 & 5:25:19.533 & 69:31:33.58 & a - wf4 & 19.108 ( 0.078 ) 69 & 413.697 & 538.809 & 5:25:19.500 & 69:31:33.55 & d - wf4 & 19.334 ( 0.050 ) 70 & 499.100 & 538.825 & 5:25:20.100 & 69:31:32.96 & a - wf4 & 20.228 ( 0.151 ) 70 & 441.542 & 526.181 & 5:25:20.074 & 69:31:33.18 & d - wf4 & 19.673 ( 0.068 ) 71 & 483.147 & 229.535 & 5:25:21.800 & 69:31:03.49 & a - wf4 & 18.764 ( 0.064 ) 71 & 444.793 & 216.463 & 5:25:21.779 & 69:31:03.65 & d - wf4 & 18.507 ( 0.026 ) 72 & 575.000 & 691.462 & 5:25:20.468 & 69:31:49.74 & a - wf4 & 18.744 ( 0.076 ) 72 & 507.595 & 682.304 & 5:25:20.444 & 69:31:49.85 & d - wf4 & 18.388 ( 0.034 ) 73 & 637.564 & 553.584 & 5:25:22.458 & 69:31:38.97 & a - wf4 & 16.397 ( 0.021 ) 73 & 578.677 & 549.316 & 5:25:22.442 & 69:31:39.18 & d - wf4 & 16.211 ( 0.008 ) 74 & 661.381 & 711.638 & 5:25:21.864 & 69:31:54.43 & a - wf4 & 23.046 ( 5.329 ) 75 & 690.004 & 709.468 & 5:25:22.390 & 69:31:55.20 & a - wf4 v & 16.453 ( 0.027 ) 75 & 629.224 & 724.146 & 5:25:22.429 & 69:31:57.16 & d - wf4 v & 16.275 ( 0.012 ) 76 & 735.370 & 686.709 & 5:25:23.333 & 69:31:54.54 & a - wf4 v & 15.700 ( 0.019 ) 76 & 668.214 & 687.938 & 5:25:23.332 & 69:31:54.73 & d - wf4 v & 15.748 ( 0.008 ) 77 & 798.535 & 543.642 & 5:25:25.366 & 69:31:43.37 & a - wf4 v & indef indef 77 & 739.637 & 549.094 & 5:25:25.352 & 69:31:43.60 & d - wf4 v & 14.723 ( 0.004 ) 78 & 655.113 & 651.461 & 5:25:22.145 & 69:31:48.68 & a - wf4 & 16.992 ( 0.029 ) 78 & 590.229 & 647.508 & 5:25:22.136 & 69:31:48.82 & d - wf4 & 16.882 ( 0.013 ) 79 & 759.209 & 568.999 & 5:25:25.604 & 69:31:46.04 & d - wf4 v & 16.451 ( 0.014 ) 80 & 760.697 & 349.908 & 5:25:26.789 & 69:31:25.25 & d - wf4 & 15.582 ( 0.006 ) 81 & 773.092 & 177.880 & 5:25:27.921 & 69:31:09.27 & d - wf4 & 19.485 ( 0.080 ) 82 & 673.152 & 602.838 & 5:25:22.780 & 69:31:44.73 & a - wf4 & 19.070 ( 0.094 ) 82 & 608.571 & 600.550 & 5:25:22.713 & 69:31:44.86 & d - wf4 & 18.372 ( 0.036 ) 83 & 627.839 & 723.154 & 5:25:21.202 & 69:31:54.46 & a - wf4 v & 19.087 ( 0.134 ) 83 & 556.981 & 718.108 & 5:25:21.159 & 69:31:54.63 & d - wf4 & 19.264 ( 0.094 ) 100 & 168.000 & 446.000 & 5:25:16.766 & 69:30:16.13 & b - pc & 18.408 ( 0.028 ) 101 & 215.000 & 513.000 & 5:25:16.097 & 69:30:17.35 & b - pc & 19.122 ( 0.051 ) 102 & 635.401 & 631.425 & 5:25:14.121 & 69:30:34.21 & b - pc & 18.717 ( 0.037 ) 103 & 647.000 & 632.000 & 5:25:14.085 & 69:30:34.72 & b - pc & 18.640 ( 0.036 ) 110 & 299.000 & 701.661 & 5:25:19.213 & 69:29:04.33 & b - wf2 & 17.863 ( 0.021 ) 111 & 300.088 & 387.000 & 5:25:17.617 & 69:29:34.38 & b - wf2 & 18.471 ( 0.028 ) 112 & 297.306 & 361.398 & 5:25:17.540 & 69:29:36.95 & b - wf2 & 19.023 ( 0.041 ) 113 & 347.844 & 719.000 & 5:25:18.430 & 69:29:01.34 & b - wf2 & 18.084 ( 0.025 ) 114 & 406.051 & 641.000 & 5:25:16.966 & 69:29:07.22 & b - wf2 & 19.030 ( 0.050 ) 115 & 483.812 & 398.539 & 5:25:14.333 & 69:29:28.46 & b - wf2 & 18.852 ( 0.036 ) 116 & 622.497 & 323.459 & 5:25:11.424 & 69:29:31.95 & b - wf2 & 18.583 ( 0.033 ) 117 & 745.000 & 439.029 & 5:25:09.786 & 69:29:17.60 & b - wf2 & 19.280 ( 0.077 ) 123 & 225.288 & 772.000 & 5:25:34.508 & 69:30:16.64 & b - wf3 & 17.930 ( 0.035 ) 124 & 241.545 & 725.462 & 5:25:33.743 & 69:30:13.89 & b - wf3 & 18.377 ( 0.048 ) 125 & 303.973 & 496.000 & 5:25:29.912 & 69:30:01.78 & b - wf3 & 19.302 ( 0.065 ) 128 & 378.000 & 650.865 & 5:25:33.092 & 69:29:58.83 & b - wf3 & 18.508 ( 0.052 ) 129 & 412.622 & 530.883 & 5:25:31.087 & 69:29:52.35 & b - wf3 & 18.764 ( 0.050 ) 130 & 529.365 & 075.459 & 5:25:23.422 & 69:29:28.92 & b - wf3 & 19.852 ( 0.090 ) 131 & 584.646 & 510.429 & 5:25:31.605 & 69:29:35.35 & b - wf3 & 17.379 ( 0.022 ) 132 & 653.131 & 348.662 & 5:25:29.013 & 69:29:24.40 & b - wf3 & 18.302 ( 0.041 ) 133 & 707.435 & 485.697 & 5:25:31.777 & 69:29:22.96 & b - wf3 v & 17.089 ( 0.021 ) 134 & 771.515 & 521.146 & 5:25:32.752 & 69:29:17.87 & b - wf3 v & 19.138 ( 0.256 ) 135 & 786.484 & 300.654 & 5:25:28.825 & 69:29:10.50 & b - wf3 & 17.318 ( 0.025 ) 142 & 488.014 & 783.042 & 5:25:24.477 & 69:31:38.04 & b - wf4 & 18.736 ( 0.078 ) 143 & 568.832 & 625.478 & 5:25:26.787 & 69:31:25.34 & b - wf4 & 15.517 ( 0.006 ) 144 & 601.438 & 272.557 & 5:25:29.249 & 69:30:52.58 & b - wf4 & 19.118 ( 0.055 ) 145 & 612.297 & 329.781 & 5:25:29.146 & 69:30:58.32 & b - wf4 & 18.940 ( 0.057 ) 146 & 629.694 & 328.539 & 5:25:29.459 & 69:30:58.70 & b - wf4 & 17.674 ( 0.022 ) 147 & 704.360 & 415.911 & 5:25:30.346 & 69:31:09.18 & b - wf4 & 19.057 ( 0.081 ) 148 & 729.542 & 595.259 & 5:25:29.839 & 69:31:26.88 & b - wf4 v & 15.263 ( 0.007 ) 149 & 749.617 & 198.932 & 5:25:32.299 & 69:30:49.79 & b - wf4 & 17.924 ( 0.032 ) 160 & 068.322 & 145.245 & 5:25:34.631 & 69:30:25.36 & c - wf2 & 18.601 ( 0.039 ) 161 & 088.595 & 469.467 & 5:25:35.909 & 69:29:53.87 & c - wf2 & 17.623 ( 0.018 ) 162 & 093.028 & 072.716 & 5:25:33.807 & 69:30:31.64 & c - wf2 & 19.435 ( 0.073 ) 163 & 144.503 & 243.762 & 5:25:33.751 & 69:30:14.05 & c - wf2 & 17.623 ( 0.018 ) 164 & 142.619 & 086.538 & 5:25:32.990 & 69:30:29.06 & c - wf2 & 19.019 ( 0.052 ) 165 & 356.613 & 587.368 & 5:25:31.621 & 69:29:35.47 & c - wf2 & 16.552 ( 0.011 ) 166 & 380.868 & 710.416 & 5:25:31.780 & 69:29:23.07 & c - wf2 & 16.193 ( 0.010 ) 167 & 476.880 & 190.383 & 5:25:27.413 & 69:30:10.35 & c - wf2 & 18.190 ( 0.028 ) 168 & 499.370 & 213.699 & 5:25:27.127 & 69:30:07.56 & c - wf2 & 18.937 ( 0.049 ) 169 & 506.226 & 205.546 & 5:25:26.956 & 69:30:08.14 & c - wf2 & 18.945 ( 0.049 ) 170 & 475.997 & 458.809 & 5:25:28.786 & 69:29:44.57 & c - wf2 & 18.148 ( 0.031 ) 171 & 517.928 & 656.790 & 5:25:29.015 & 69:29:24.61 & c - wf2 & 17.389 ( 0.027 ) 172 & 628.318 & 047.514 & 5:25:23.963 & 69:30:19.93 & c - wf2 & 17.052 ( 0.014 ) 173 & 718.884 & 359.675 & 5:25:23.866 & 69:29:47.74 & c - wf2 & 16.350 ( 0.011 ) 174 & 751.491 & 113.581 & 5:25:22.071 & 69:30:10.36 & c - wf2 & 17.248 ( 0.019 ) 180 & 051.431 & 142.692 & 5:25:36.224 & 69:30:38.05 & c - wf3 & 18.946 ( 0.060 ) 181 & 065.528 & 505.268 & 5:25:42.880 & 69:30:46.50 & c - wf3 & 17.114 ( 0.014 ) 182 & 072.679 & 141.546 & 5:25:36.310 & 69:30:36.03 & c - wf3 & 18.104 ( 0.028 ) 183 & 114.264 & 105.729 & 5:25:35.871 & 69:30:31.07 & c - wf3 & 18.040 ( 0.027 ) 184 & 148.323 & 181.513 & 5:25:37.413 & 69:30:29.90 & c - wf3 & 18.058 ( 0.028 ) 185 & 202.542 & 443.744 & 5:25:42.451 & 69:30:31.82 & c - wf3 & 17.172 ( 0.015 ) 186 & 257.723 & 099.560 & 5:25:36.480 & 69:30:17.29 & c - wf3 & 19.778 ( 0.109 ) 187 & 281.897 & 420.586 & 5:25:42.442 & 69:30:23.55 & c - wf3 & 17.904 ( 0.028 ) 188 & 296.206 & 425.715 & 5:25:42.605 & 69:30:22.35 & c - wf3 & 16.062 ( 0.008 ) 189 & 353.159 & 510.794 & 5:25:44.447 & 69:30:19.17 & c - wf3 & 13.448 ( 0.002 ) 200 & 374.164 & 500.066 & 5:25:44.372 & 69:30:16.89 & c - wf3 & 17.764 ( 0.026 ) 201 & 397.615 & 345.481 & 5:25:41.664 & 69:30:10.52 & c - wf3 & 17.503 ( 0.019 ) 202 & 424.650 & 161.750 & 5:25:38.456 & 69:30:02.98 & c - wf3 & 18.740 ( 0.044 ) 203 & 535.846 & 358.931 & 5:25:42.610 & 69:29:57.64 & c - wf3 & 13.922 ( 0.003 ) 204 & 570.584 & 293.642 & 5:25:41.606 & 69:29:52.55 & c - wf3 & 17.414 ( 0.019 ) 205 & 570.775 & 210.666 & 5:25:40.098 & 69:29:50.34 & c - wf3 & 17.940 ( 0.027 ) 206 & 584.649 & 525.937 & 5:25:45.901 & 69:29:57.45 & c - wf3 & 16.287 ( 0.011 ) 207 & 668.161 & 219.805 & 5:25:40.767 & 69:29:41.24 & c - wf3 & 18.819 ( 0.063 ) 210 & 083.143 & 175.582 & 5:25:34.401 & 69:30:50.67 & c - wf4 & 17.310 ( 0.016 ) 211 & 089.948 & 147.000 & 5:25:34.660 & 69:30:48.19 & c - wf4 & 18.478 ( 0.037 ) 212 & 117.067 & 153.880 & 5:25:35.128 & 69:30:49.61 & c - wf4 & 19.662 ( 0.095 ) 213 & 130.968 & 461.566 & 5:25:33.725 & 69:31:19.20 & c - wf4 & 15.590 ( 0.006 ) 214 & 132.439 & 438.356 & 5:25:33.882 & 69:31:17.06 & c - wf4 & 17.674 ( 0.022 ) 215 & 173.487 & 145.009 & 5:25:36.188 & 69:30:50.29 & c - wf4 & 15.173 ( 0.005 ) 216 & 235.989 & 625.341 & 5:25:34.780 & 69:31:37.80 & c - wf4 & 18.937 ( 0.059 ) 217 & 279.585 & 274.929 & 5:25:37.422 & 69:31:05.48 & c - wf4 & 17.918 ( 0.025 ) 218 & 319.086 & 620.000 & 5:25:36.312 & 69:31:39.65 & c - wf4 & 18.998 ( 0.064 ) 219 & 397.579 & 230.891 & 5:25:39.800 & 69:31:04.67 & c - wf4 & 16.305 ( 0.009 ) 220 & 423.869 & 126.613 & 5:25:40.826 & 69:30:55.40 & c - wf4 & 14.387 ( 0.003 ) 221 & 470.121 & 563.511 & 5:25:39.361 & 69:31:38.42 & c - wf4 & 13.829 ( 0.002 ) 222 & 508.547 & 240.440 & 5:25:41.768 & 69:31:08.63 & c - wf4 & 16.744 ( 0.011 ) 223 & 581.176 & 315.656 & 5:25:42.697 & 69:31:17.83 & c - wf4 & 18.833 ( 0.055 ) 224 & 600.979 & 310.116 & 5:25:43.086 & 69:31:17.91 & c - wf4 & 19.132 ( 0.073 ) 225 & 668.361 & 305.426 & 5:25:44.324 & 69:31:19.30 & c - wf4 & 18.497 ( 0.047 ) 226 & 732.990 & 129.566 & 5:25:46.412 & 69:31:04.39 & c - wf4 & 15.352 ( 0.006 ) 227 & 747.563 & 627.673 & 5:25:44.034 & 69:31:52.10 & c - wf4 v & 16.911 ( 0.051 ) 228 & 758.000 & 500.755 & 5:25:44.912 & 69:31:40.36 & c - wf4 v & 18.298 ( 0.086 ) 229 & 731.764 & 296.000 & 5:25:45.507 & 69:31:20.20 & c - wf4 & 15.889 ( 0.008 ) 240 & 072.732 & 070.708 & 5:25:29.258 & 69:30:53.16 & e - wf2 & 18.284 ( 0.031 ) 241 & 156.000 & 326.598 & 5:25:29.033 & 69:30:26.59 & e - wf2 & 18.956 ( 0.048 ) 242 & 139.914 & 688.479 & 5:25:31.124 & 69:29:52.35 & e - wf2 & 18.108 ( 0.031 ) 243 & 237.294 & 301.000 & 5:25:27.431 & 69:30:26.86 & e - wf2 & 17.470 ( 0.017 ) 244 & 247.279 & 281.896 & 5:25:27.153 & 69:30:28.42 & e - wf2 & 17.685 ( 0.020 ) 245 & 279.606 & 732.290 & 5:25:28.819 & 69:29:44.47 & e - wf2 & 18.230 ( 0.046 ) 246 & 281.311 & 462.645 & 5:25:27.439 & 69:30:10.25 & e - wf2 & 18.078 ( 0.026 ) 247 & 300.837 & 466.093 & 5:25:27.112 & 69:30:09.36 & e - wf2 & 19.797 ( 0.106 ) 248 & 270.900 & 207.564 & 5:25:26.336 & 69:30:34.98 & e - wf2 & 19.185 ( 0.060 ) 249 & 331.142 & 079.421 & 5:25:24.598 & 69:30:45.61 & e - wf2 & 14.776 ( 0.004 ) 250 & 379.159 & 474.646 & 5:25:25.710 & 69:30:06.51 & e - wf2 & 20.190 ( 0.149 ) 251 & 392.711 & 475.000 & 5:25:25.459 & 69:30:06.04 & e - wf2 & 19.592 ( 0.086 ) 252 & 431.162 & 322.179 & 5:25:23.994 & 69:30:19.73 & e - wf2 & 17.086 ( 0.013 ) 253 & 461.645 & 303.259 & 5:25:23.350 & 69:30:20.76 & e - wf2 & 19.084 ( 0.053 ) 254 & 414.554 & 108.455 & 5:25:23.230 & 69:30:40.66 & e - wf2 & 17.120 ( 0.013 ) 255 & 498.102 & 076.538 & 5:25:21.541 & 69:30:41.48 & e - wf2 & 16.527 ( 0.009 ) 256 & 522.289 & 632.635 & 5:25:23.896 & 69:29:47.61 & e - wf2 & 16.285 ( 0.011 ) 257 & 553.549 & 388.252 & 5:25:22.100 & 69:30:10.17 & e - wf2 & 17.394 ( 0.018 ) 258 & 635.320 & 568.437 & 5:25:21.518 & 69:29:50.78 & e - wf2 & 18.085 ( 0.049 ) 259 & 662.106 & 702.000 & 5:25:21.709 & 69:29:37.33 & e - wf2 v & 16.176 ( 0.036 ) 260 & 668.000 & 759.000 & 5:25:21.890 & 69:29:31.79 & e - wf2 v & 16.881 ( 0.192 ) 270 & 054.313 & 503.182 & 5:25:37.873 & 69:31:08.27 & e - wf3 & 17.795 ( 0.023 ) 271 & 071.041 & 473.684 & 5:25:37.414 & 69:31:05.86 & e - wf3 & 18.086 ( 0.028 ) 272 & 069.799 & 675.279 & 5:25:41.076 & 69:31:11.26 & e - wf3 & 18.917 ( 0.067 ) 273 & 079.882 & 699.812 & 5:25:41.579 & 69:31:10.96 & e - wf3 & 17.604 ( 0.025 ) 274 & 102.244 & 704.530 & 5:25:41.765 & 69:31:08.99 & e - wf3 & 16.773 ( 0.013 ) 275 & 113.652 & 592.643 & 5:25:39.793 & 69:31:05.02 & e - wf3 & 16.423 ( 0.009 ) 276 & 149.598 & 165.465 & 5:25:32.211 & 69:30:50.16 & e - wf3 & 17.467 ( 0.018 ) 277 & 172.319 & 279.436 & 5:25:34.395 & 69:30:51.05 & e - wf3 & 17.335 ( 0.016 ) 278 & 202.425 & 369.526 & 5:25:36.187 & 69:30:50.61 & e - wf3 & 15.284 ( 0.005 ) 279 & 217.944 & 620.025 & 5:25:40.840 & 69:30:55.78 & e - wf3 & 14.398 ( 0.003 ) 280 & 251.000 & 728.724 & 5:25:42.966 & 69:30:55.47 & e - wf3 & 17.932 ( 0.041 ) 281 & 338.737 & 699.430 & 5:25:42.888 & 69:30:46.41 & e - wf3 & 16.822 ( 0.016 ) 282 & 325.910 & 472.064 & 5:25:38.700 & 69:30:41.51 & e - wf3 & 18.868 ( 0.054 ) 283 & 308.327 & 426.849 & 5:25:37.768 & 69:30:42.00 & e - wf3 & 16.365 ( 0.009 ) 284 & 326.544 & 338.481 & 5:25:36.255 & 69:30:37.91 & e - wf3 & 18.975 ( 0.054 ) 285 & 347.444 & 336.793 & 5:25:36.326 & 69:30:35.85 & e - wf3 & 17.933 ( 0.025 ) 286 & 311.361 & 210.011 & 5:25:33.848 & 69:30:35.91 & e - wf3 & 19.317 ( 0.074 ) 287 & 388.582 & 301.690 & 5:25:35.898 & 69:30:30.98 & e - wf3 & 17.887 ( 0.025 ) 288 & 421.748 & 376.416 & 5:25:37.435 & 69:30:29.83 & e - wf3 & 17.912 ( 0.024 ) 289 & 426.167 & 217.360 & 5:25:34.564 & 69:30:25.08 & e - wf3 & 18.697 ( 0.044 ) 290 & 475.195 & 638.307 & 5:25:42.482 & 69:30:31.68 & e - wf3 & 16.914 ( 0.019 ) 291 & 523.612 & 141.245 & 5:25:33.684 & 69:30:13.77 & e - wf3 & 17.822 ( 0.023 ) 292 & 568.756 & 620.401 & 5:25:42.627 & 69:30:22.31 & e - wf3 & 15.916 ( 0.010 ) 293 & 626.353 & 706.461 & 5:25:44.472 & 69:30:19.09 & e - wf3 v & 13.132 ( 0.002 ) 294 & 670.698 & 539.728 & 5:25:41.672 & 69:30:10.40 & e - wf3 v & 17.446 ( 0.038 ) 295 & 682.269 & 146.000 & 5:25:34.596 & 69:29:58.77 & e - wf3 & 18.327 ( 0.041 ) 296 & 750.581 & 194.599 & 5:25:35.816 & 69:29:53.59 & e - wf3 & 17.674 ( 0.028 ) 297 & 206.881 & 139.141 & 5:25:32.027 & 69:30:44.03 & e - wf3 & 19.052 ( 0.058 ) 298 & 236.359 & 150.349 & 5:25:32.378 & 69:30:41.46 & e - wf3 & 19.903 ( 0.118 ) 300 & 117.117 & 139.786 & 5:25:30.292 & 69:31:08.99 & e - wf4 & 18.500 ( 0.037 ) 301 & 102.388 & 661.826 & 5:25:27.260 & 69:31:58.27 & e - wf4 & 18.468 ( 0.040 ) 302 & 151.012 & 732.273 & 5:25:27.777 & 69:32:06.37 & e - wf4 & 17.760 ( 0.028 ) 303 & 141.534 & 319.154 & 5:25:29.764 & 69:31:26.75 & e - wf4 & 14.974 ( 0.004 ) 304 & 207.554 & 416.185 & 5:25:30.445 & 69:31:37.84 & e - wf4 & 16.768 ( 0.011 ) 305 & 202.375 & 632.783 & 5:25:29.216 & 69:31:58.32 & e - wf4 & 17.329 ( 0.017 ) 306 & 234.848 & 587.563 & 5:25:30.031 & 69:31:54.93 & e - wf4 & 19.327 ( 0.076 ) 307 & 241.896 & 607.643 & 5:25:30.072 & 69:31:57.06 & e - wf4 & 17.275 ( 0.016 ) 308 & 221.568 & 338.525 & 5:25:31.113 & 69:31:30.78 & e - wf4 & 14.791 ( 0.004 ) 309 & 251.783 & 170.844 & 5:25:32.572 & 69:31:15.62 & e - wf4 & 19.170 ( 0.064 ) 310 & 302.518 & 549.339 & 5:25:31.467 & 69:31:53.20 & e - wf4 & 18.555 ( 0.041 ) 311 & 322.960 & 533.000 & 5:25:31.934 & 69:31:52.26 & e - wf4 & 18.695 ( 0.043 ) 312 & 322.468 & 186.645 & 5:25:33.759 & 69:31:19.08 & e - wf4 & 15.599 ( 0.006 ) 313 & 323.569 & 163.505 & 5:25:33.368 & 69:31:26.45 & e - wf4 & 17.680 ( 0.020 ) 314 & 382.990 & 243.649 & 5:25:34.566 & 69:31:26.21 & e - wf4 & 19.120 ( 0.059 ) 315 & 472.313 & 492.403 & 5:25:34.864 & 69:31:52.49 & e - wf4 & 19.702 ( 0.115 ) 316 & 535.841 & 242.504 & 5:25:37.354 & 69:31:30.38 & e - wf4 & 19.307 ( 0.073 ) 317 & 536.663 & 461.382 & 5:25:36.194 & 69:31:51.30 & e - wf4 & 18.316 ( 0.040 ) 318 & 560.390 & 277.688 & 5:25:37.605 & 69:31:34.39 & e - wf4 & 18.847 ( 0.052 ) 319 & 574.503 & 574.484 & 5:25:36.283 & 69:32:03.13 & e - wf4 & 19.150 ( 0.116 ) 320 & 587.925 & 473.000 & 5:25:37.057 & 69:31:53.86 & e - wf4 & 18.696 ( 0.061 ) 321 & 596.447 & 264.982 & 5:25:38.327 & 69:31:34.14 & e - wf4 & 18.586 ( 0.043 ) 322 & 661.675 & 288.691 & 5:25:39.395 & 69:31:38.29 & e - wf4 & 13.836 ( 0.002 ) 323 & 724.334 & 265.986 & 5:25:40.628 & 69:31:37.92 & e - wf4 & 18.924 ( 0.083 ) 324 & 731.899 & 529.597 & 5:25:39.381 & 69:32:03.18 & e - wf4 v & 18.139 ( 0.070 ) 325 & 749.995 & 672.309 & 5:25:38.940 & 69:32:17.20 & e - wf4 v & 14.710 ( 0.012 ) 326 & 783.278 & 591.757 & 5:25:39.966 & 69:32:10.44 & e - wf4 v & 16.052 ( 0.026 ) 340 & 306.655 & 070.736 & 5:25:24.597 & 69:30:45.41 & f - wf2 & 15.515 ( 0.005 ) 341 & 473.100 & 057.589 & 5:25:21.543 & 69:30:41.28 & f - wf2 & 17.273 ( 0.013 ) 342 & 531.532 & 611.436 & 5:25:23.891 & 69:29:47.36 & f - wf2 & 17.103 ( 0.014 ) 350 & 130.376 & 721.599 & 5:25:41.565 & 69:31:10.63 & f - wf3 & 18.135 ( 0.028 ) 351 & 152.138 & 726.227 & 5:25:41.783 & 69:31:08.75 & f - wf3 & 17.530 ( 0.018 ) 352 & 156.710 & 613.613 & 5:25:39.810 & 69:31:04.74 & f - wf3 & 17.203 ( 0.013 ) 353 & 231.375 & 385.683 & 5:25:36.202 & 69:30:50.37 & f - wf3 & 16.023 ( 0.007 ) 354 & 262.401 & 634.561 & 5:25:40.843 & 69:30:55.54 & f - wf3 & 15.121 ( 0.005 ) 355 & 612.542 & 613.529 & 5:25:42.638 & 69:30:22.01 & f - wf3 & 16.714 ( 0.013 ) 356 & 675.392 & 695.695 & 5:25:44.477 & 69:30:18.82 & f - wf3 v & 13.843 ( 0.003 ) 357 & 766.953 & 176.323 & 5:25:35.808 & 69:29:53.41 & f - wf3 & 16.662 ( 0.010 ) 358 & 222.567 & 754.311 & 5:25:42.721 & 69:31:03.09 & f - wf3 & 16.104 ( 0.008 ) 359 & 520.155 & 636.918 & 5:25:42.486 & 69:30:31.43 & f - wf3 & 17.808 ( 0.026 ) 360 & 387.792 & 706.562 & 5:25:42.904 & 69:30:46.11 & f - wf3 & 17.656 ( 0.022 ) 361 & 397.182 & 712.011 & 5:25:43.058 & 69:30:45.40 & f - wf3 & 17.932 ( 0.027 ) 362 & 340.460 & 436.454 & 5:25:37.790 & 69:30:41.78 & f - wf3 & 17.162 ( 0.012 ) 370 & 190.129 & 372.325 & 5:25:29.419 & 69:31:33.22 & f - wf4 & 13.918 ( 0.003 ) 371 & 185.623 & 302.985 & 5:25:29.785 & 69:31:26.58 & f - wf4 & 15.736 ( 0.006 ) 372 & 257.507 & 395.704 & 5:25:30.469 & 69:31:37.67 & f - wf4 & 17.499 ( 0.015 ) 373 & 266.626 & 317.549 & 5:25:31.134 & 69:31:30.65 & f - wf4 & 15.502 ( 0.005 ) 374 & 357.737 & 159.598 & 5:25:33.772 & 69:31:18.95 & f - wf4 & 16.368 ( 0.008 ) 375 & 359.546 & 139.348 & 5:25:33.934 & 69:31:17.12 & f - wf4 & 16.820 ( 0.010 ) 376 & 703.313 & 240.469 & 5:25:39.399 & 69:31:38.14 & f - wf4 & 14.585 ( 0.003 ) 377 & 203.469 & 731.000 & 5:25:27.377 & 69:32:07.17 & f - wf4 & 16.609 ( 0.010 ) 16 & 2 & 5:25:14.08 & 69:29:26.70 & 18.46 ( 0.14 ) 103 & 1 & 5:25:14.09 & 69:30:34.72 & 18.23 ( 0.04 ) 102 & 1 & 5:25:14.12 & 69:30:34.21 & 18.30 ( 0.04 ) 115 & 1 & 5:25:14.33 & 69:29:28.46 & 18.28 ( 0.04 ) 7 & 1 & 5:25:14.51 & 69:30:50.67 & 16.19 ( 0.01 ) 18 & 1 & 5:25:14.73 & 69:29:10.86 & 16.79 ( 0.02 ) 59 & 2 & 5:25:14.77 & 69:31:24.61 & 18.49 ( 0.14 ) 251 & 1 & 5:25:25.46 & 69:30 : 6.04 & 19.06 ( 0.09 ) 79 & 1 & 5:25:25.60 & 69:31:46.04 & 15.92 ( 0.01 ) 250 & 1 & 5:25:25.71 & 69:30 : 6.51 & 19.66 ( 0.15 ) 42 & 2 & 5:25:25.92 & 69:29:50.48 & 17.42 ( 0.57 ) 148 & 3 & 5:25:29.84 & 69:31:26.88 & 14.74 ( 0.38 ) 125 & 1 & 5:25:29.91 & 69:30 : 1.78 & 18.73 ( 0.06 ) 306 & 1 & 5:25:30.03 & 69:31:54.93 & 18.79 ( 0.08 ) 307 & 1 & 5:25:30.07 & 69:31:57.06 & 16.74 ( 0.02 ) 149 & 3 & 5:25:32.30 & 69:30:49.79 & 17.05 ( 0.31 ) 298 & 1 & 5:25:32.38 & 69:30:41.46 & 19.37 ( 0.12 ) 309 & 1 & 5:25:32.57 & 69:31:15.62 & 18.64 ( 0.06 ) 134 & 1 & 5:25:32.75 & 69:29:17.87 & 18.57 ( 0.26 ) 164 & 1 & 5:25:32.99 & 69:30:29.06 & 18.45 ( 0.05 ) 128 & 1 & 5:25:33.09 & 69:29:58.83 & 17.94 ( 0.05 ) 313 & 1 & 5:25:33.37 & 69:31:26.45 & 17.15 ( 0.02 ) 314 & 1 & 5:25:34.57 & 69:31:26.21 & 18.59 ( 0.06 ) 295 & 1 & 5:25:34.60 & 69:29:58.77 & 17.79 ( 0.04 ) 160 & 2 & 5:25:34.63 & 69:30:25.36 & 18.09 ( 0.07 ) 211 & 1 & 5:25:34.66 & 69:30:48.19 & 17.91 ( 0.04 ) 216 & 1 & 5:25:34.78 & 69:31:37.80 & 18.37 ( 0.06 ) 315 & 1 & 5:25:34.86 & 69:31:52.49 & 19.17 ( 0.12 ) 212 & 1 & 5:25:35.13 & 69:30:49.61 & 19.09 ( 0.09 ) 296 & 2 & 5:25:35.82 & 69:29:53.59 & 16.53 ( 0.50 ) 183 & 2 & 5:25:35.87 & 69:30:31.07 & 17.41 ( 0.06 ) 186 & 1 & 5:25:36.48 & 69:30:17.29 & 19.21 ( 0.11 ) 320 & 1 & 5:25:37.06 & 69:31:53.86 & 18.16 ( 0.06 ) 316 & 1 & 5:25:37.35 & 69:31:30.38 & 18.77 ( 0.07 ) 184 & 2 & 5:25:37.41 & 69:30:29.90 & 17.43 ( 0.06 ) 270 & 1 & 5:25:37.87 & 69:31 : 8.27 & 17.26 ( 0.02 ) 321 & 1 & 5:25:38.33 & 69:31:34.14 & 18.05 ( 0.04 ) 202 & 1 & 5:25:38.46 & 69:30 : 2.98 & 18.17 ( 0.04 ) 282 & 1 & 5:25:38.70 & 69:30:41.51 & 18.33 ( 0.05 ) 325 & 1 & 5:25:38.94 & 69:32:17.20 & 14.17 ( 0.01 ) 221 & 3 & 5:25:39.36 & 69:31:38.42 & 13.48 ( 0.43 ) 324 & 1 & 5:25:39.38 & 69:32 : 3.18 & 17.60 ( 0.07 ) 326 & 1 & 5:25:39.97 & 69:32:10.44 & 15.52 ( 0.03 ) 205 & 1 & 5:25:40.10 & 69:29:50.34 & 17.37 ( 0.03 ) 323 & 1 & 5:25:40.63 & 69:31:37.92 & 18.39 ( 0.08 ) 207 & 1 & 5:25:40.77 & 69:29:41.24 & 18.25 ( 0.06 ) 220 & 3 & 5:25:40.83 & 69:30:55.40 & 14.04 ( 0.42 ) 280 & 1 & 5:25:42.97 & 69:30:55.47 & 17.40 ( 0.04 ) 223 & 1 & 5:25:42.70 & 69:31:17.83 & 18.26 ( 0.05 ) 361 & 1 & 5:25:43.06 & 69:30:45.40 & 17.41 ( 0.03 ) 224 & 1 & 5:25:43.09 & 69:31:17.91 & 18.56 ( 0.07 ) 227 & 1 & 5:25:44.03 & 69:31:52.10 & 16.34 ( 0.05 ) 225 & 1 & 5:25:44.32 & 69:31:19.30 & 17.93 ( 0.05 ) 200 & 1 & 5:25:44.37 & 69:30:16.89 & 17.19 ( 0.03 ) 189 & 3 & 5:25:44.45 & 69:30:19.17 & 12.89 ( 0.36 ) 228 & 1 & 5:25:44.91 & 69:31:40.36 & 17.73 ( 0.09 ) 229 & 1 & 5:25:45.51 & 69:31:20.20 & 15.32 ( 0.01 ) 206 & 1 & 5:25:45.90 & 69:29:57.45 & 15.72 ( 0.01 ) 226 & 1 & 5:25:46.41 & 69:31 : 4.39 & 14.78 ( 0.01 ) rrrr 15414 & 14.06 ( 0.04 ) & 25 & 15.24 ( 0.01 ) 15511 & 11.47 ( 0.08 ) & 4 & 13.20 ( 0.16 ) 15543 & 14.59 ( 0.06 ) & 6 & 17.04 ( 0.16 ) 15556 & 15.05 ( 0.05 ) & 17 & 16.65 ( 0.10 ) 15557 & 14.90 ( 0.03 ) & 56 & 16.40 ( 0.01 ) 15565 & 13.35 ( 0.04 ) & 14 & 14.89 ( 0.11 ) 15584 & 13.58 ( 0.04 ) & 13 & 15.07 ( 0.13 ) & & 24 & 18.86 ( 0.14 ) 15591 & 12.20 ( 0.03 ) & 53 & 13.75 ( 0.11 ) 15593 & 14.31 ( 0.04 ) & 64 & 15.86 ( 0.28 ) & & 67 & 18.29 ( 0.09 ) 15598 & 14.25 ( 0.03 ) & 11 & 15.81 ( 0.13 ) & & 12 & 18.67 ( 0.04 ) 15602 & 12.79 ( 0.03 ) & 114 & 18.46 ( 0.05 ) 15616 & 12.83 ( 0.02 ) & 61 & 14.39 ( 0.07 ) & & 62 & 17.17 ( 0.12 ) & & 66 & 18.65 ( 0.16 ) 15681 & 15.35 ( 0.04 ) & 31 & 16.62 ( 0.42 ) & & 33 & 17.22 ( 0.53 ) 15697 & 14.30 ( 0.04 ) & 73 & 15.77 ( 0.10 ) 15709 & 13.86 ( 0.05 ) & 76 & 15.20 ( 0.02 ) & & 74 & 22.53 ( 5.33 ) & & 75 & 15.84 ( 0.11 ) 15721 & 15.32 ( 0.06 ) & 44 & 16.30 ( 0.38 ) & & 43 & 17.50 ( 0.20 ) 15731 & 13.64 ( 0.04 ) & 32 & 14.85 ( 0.49 ) 15744 & 13.06 ( 0.03 ) & 77 & 14.94 ( 0.00 ) & & 79 & 15.92 ( 0.01 ) 15777 & 13.79 ( 0.04 ) & 80 & 14.96 ( 0.34 ) 15810 & 15.25 ( 0.05 ) & 135 & 16.75 ( 0.03 ) 15821 & 15.54 ( 0.05 ) & 146 & 16.70 ( 0.35 ) & & 145 & 18.07 ( 0.26 ) & & 144 & 18.15 ( 0.42 ) 15835 & 13.74 ( 0.04 ) & 148 & 14.74 ( 0.38 ) 15853 & 15.71 ( 0.09 ) & 304 & 16.54 ( 0.37 ) 15865 & 13.65 ( 0.04 ) & 308 & 14.56 ( 0.36 ) 15873 & 15.11 ( 0.05 ) & 133 & 15.98 ( 0.45 ) 15906 & 16.53 ( 0.16 ) & 134 & 18.57 ( 0.26 ) 15922 & 14.25 ( 0.03 ) & 213 & 15.25 ( 0.45 ) & & 214 & 16.63 ( 0.40 ) 15933 & 16.08 ( 0.12 ) & 210 & 16.77 ( 0.03 ) & & 211 & 17.91 ( 0.04 ) & & 212 & 19.09 ( 0.09 ) 15953 & 14.04 ( 0.06 ) & 215 & 14.89 ( 0.46 ) 15978 & 15.02 ( 0.04 ) & 283 & 16.16 ( 0.41 ) & & 282 & 18.33 ( 0.05 ) 15988 & 14.16 ( 0.04 ) & 325 & 14.17 ( 0.01 ) 15995 & 12.82 ( 0.03 ) & 221 & 13.48 ( 0.43 ) & & 323 & 18.39 ( 0.08 ) 16004 & 14.99 ( 0.05 ) & 219 & 16.03 ( 0.49 ) 16005 & 15.39 ( 0.06 ) & 326 & 15.52 ( 0.03 ) 16019 & 13.23 ( 0.03 ) & 220 & 14.04 ( 0.42 ) & & 280 & 17.40 ( 0.04 ) 16038 & 14.98 ( 0.05 ) & 222 & 16.41 ( 0.45 ) & & 272 & 18.38 ( 0.07 ) & & 273 & 17.31 ( 0.27 ) 16049 & 14.38 ( 0.05 ) & 188 & 15.63 ( 0.42 ) & & 187 & 17.33 ( 0.03 ) 16052 & 12.76 ( 0.03 ) & 203 & 13.35 ( 0.00 ) 16053 & 15.37 ( 0.05 ) & 181 & 16.60 ( 0.42 ) & & 361 & 17.41 ( 0.03 ) 16076 & 14.15 ( 0.41 ) & 189 & 12.89 ( 0.36 ) & & 225 & 17.93 ( 0.05 ) 16087 & 15.48 ( 0.40 ) & 200 & 17.19 ( 0.03 ) 16096 & 14.74 ( 0.04 ) & 206 & 15.72 ( 0.01 ) crr 26 & [email protected] & [email protected] 25 & [email protected] & [email protected] 20 & [email protected] & [email protected] 4 & [email protected] & [email protected] 6 & [email protected] & [email protected] 12 & [email protected] & [email protected] 56 & [email protected] & [email protected] 53 & [email protected] & [email protected] 64 & [email protected] & [email protected] 11 & [email protected] & [email protected] 114 & [email protected] & [email protected] 61 & [email protected] & [email protected] 69 & [email protected] & [email protected] 68 & [email protected] & [email protected] 36 & [email protected] & [email protected] 78 & [email protected] & [email protected] 73 & [email protected] & [email protected] 331 & [email protected] & [email protected] 76 & [email protected] & [email protected] 46 & [email protected] & [email protected] 32 & [email protected] & [email protected] 77 & [email protected] & [email protected] 80 & [email protected] & [email protected] 3702 & [email protected] & [email protected] 168 & [email protected] & [email protected] 302 & [email protected] & [email protected] 81 & [email protected] & [email protected] 45 & [email protected] & [email protected] 45 & [email protected] & [email protected] 170 & [email protected] & [email protected] 305 & [email protected] & [email protected] 146 & [email protected] & [email protected] 148 & [email protected] & [email protected] 304 & [email protected] & [email protected] 308 & [email protected] & [email protected] 131 & [email protected] & [email protected] 133 & [email protected] & [email protected] 134 & [email protected] & [email protected] 313 & [email protected] & [email protected] 213 & [email protected] & [email protected] 212 & [email protected] & [email protected] 215 & [email protected] & [email protected] 319 & [email protected] & [email protected] 182 & [email protected] & [email protected] 186 & [email protected] & [email protected] 316 & [email protected] & [email protected] 221 & [email protected] & [email protected] 326 & [email protected] & [email protected] 205 & [email protected] & [email protected] 220 & [email protected] & [email protected] 273 & [email protected] & [email protected] 201 & [email protected] & [email protected] 188 & [email protected] & [email protected] 203 & [email protected] & [email protected] 223 & [email protected] & [email protected] 200 & [email protected] & [email protected] 189 & [email protected] & [email protected] 226 & [email protected] & [email protected] | we present an analysis of the deepest pure - uv observations with the highest angular resolution ever performed , a set of 12 exposures with the hst wfpc2 and f160bw filter obtained in parallel observing mode , which cover @xmath012 square arcminutes in the lmc , north of the bar and in the `` general field '' regime of the lmc .
the 341 independent measurements of 198 objects represent an accumulated exposure of @xmath12 10@xmath2 sec and reveal stars as faint as m@xmath322 mag .
the observations show that @xmath02/3 of the uv emission from the lmc is emitted by our hst - detected uv stars in the field , _
i.e. , not _ in clusters or associations .
we identified optical counterparts in the roe / nrl photometric catalog for @xmath0 1/3 of the objects .
the results are used to discuss the nature of these uv sources , to estimate the diffuse uv emission from the lmc as a prototype of dwarf galaxies , and to evaluate the contamination by field stars of uv observations of globular and open clusters in the lmc .
we find that the projected density of uv stars in the general field of the lmc is a few times higher than in the galactic disk close to the sun .
combining our data with observations by uit allows us to define the stellar uv luminosity function from m@xmath4=8 to 18 mag , and to confirm that the field regions in the lmc have been forming stars at a steady rate during the last 1 gyr , with an imf close to the salpeter law . #
1#2#3#4#5#6#7 to#2 ' '' '' |
the recent experimental realization of synthetic gauge field in ultracold atomics gases has greatly extended the horizon of quantum simulation in these systems @xcite .
a particularly important case is the implementation of synthetic spin - orbit coupling ( soc ) , a non - abelian gauge field , in these systems , where the internal degrees of freedom of the atoms are coupled to the atomic center - of - mass motional degrees of freedom @xcite . in condensed - matter materials
, soc plays a key role in many interesting phenomena , such as the quantum spin hall effects , topological insulators , and topological superconductors @xcite . although the form of the synthetic soc currently realized in cold atoms differs crucially from those in condensed - matter systems , there exist various theoretical proposals on realizing synthetic soc which can induce topologically nontrivial phases @xcite .
thus , the hope of simulating the various topological phases , the topological superfluid state in particular , in the highly controllable environment of an ultracold atomic gas stimulated intensive theoretical studies on spin - orbit coupled fermi gases @xcite .
furthermore , recent studies suggest that other exotic superfluid phases and novel phenomena can be engineered with carefully designed configurations @xcite . as such , soc has a great potential of becoming a powerful tool of quantum control in ultracold atomic gases . in this review ,
we focus on the zero - temperature pairing physics in a spin - orbit coupled ultracold fermi gas .
we will discuss the exotic superfluid phases in systems with different spatial dimensions and with different forms of soc .
a fundamentally important effect of soc is the modification of the single - particle dispersion spectra @xcite .
we will start from there and show how this effect leads to interesting pairing phases such as the topological superfluid state , the various gapless superfluid states , the soc - induced fulde - ferrell ( ff ) state , and the topological ff state .
we will also touch upon the topic of exotic few - body states in spin - orbit coupled fermi systems whose stability also benefits from the soc - modified single - particle dispersion .
the paper is organized as follows : in sec .
[ sec_implementation ] , we briefly introduce the implementation scheme of soc in current cold atom experiments . in sec .
[ sec_singlespec ] , we present the single - particle dispersion spectra under various forms of soc . then in sec . [ sec_pairing ] , we analyze the general pairing mechanism in these systems based on the modified single - particle dispersion , and present the exotic superfluid phases and the rich phase diagrams under different configurations .
we further discuss the possibilities of engineering novel pairing states and novel few - body states in sec .
[ sec_engineer ] . finally , we summarize in sec .
[ sec_fin ] .
the principle behind most of the proposals for an artificial gauge potential is based on the adiabatic theorem and the associated geometrical phase @xcite . in general , by engineering the atom - laser interaction , the atoms experience an adiabatic potential when moving through space .
the resulting geometrical phase appearing in the effective hamiltonian gives rise to the artificial gauge potential . to see this , we start from the full hamiltonian @xmath0,\ ] ] where @xmath1 is the kinetic energy and @xmath2 $ ] describes the atom - laser coupling , whose spatial dependence is related to the atomic motion .
formally , let us expand the wave function at any given time @xmath3 onto the eigen basis @xmath4 of @xmath5 @xmath6 where @xmath7 s are the time - dependent expansion coefficients . substituting the expansion above into the time - dependent schrdinger s equation and projecting it into the subspace of the @xmath8-th eigen state
, we have @xmath9 where @xmath10 satisfies @xmath11 . assuming the adiabatic condition , under which the slow center - of - mass motion of an atom adiabatically follows the fast internal dynamics governed by @xmath5
, we may retain only @xmath12 in eq .
( [ eqnadiabaticfull ] ) to get @xmath13 which effectively describes the motion of an atom in the adiabatic potential @xmath10 . to make the geometrical phase stand out , we further examine the term involving the kinetic energy @xmath14\right\ } \nonumber\\ & = & \frac{1}{2m}\left(\mathbf{p}-\mathbf{a}\right)^2c_{\alpha}+wc_{\alpha}.\end{aligned}\ ] ] here , the geometrical vector potential @xmath15 , and the geometrical scalar potential @xmath16 .
hence , apart from an energy shift due to the scalar potential @xmath17 , the effective hamiltonian for an atom in the adiabatic potential @xmath10 can be written as @xmath18 is the effective rabi frequency of the raman process , @xmath19 is the two - photon detuning .
the pseudo - spin states ( @xmath20 ) are hyperfine states within the ground state hyperfine manifold.,width=340 ] the physical implication is just what we have stated at the beginning , for a particle moving in an adiabatic potential , its external motion adiabatically follows the internal dynamics at each spatial location . as a result , the internal states of the particle may change as the particle is moving through space .
when the change in internal states involves only a phase factor , the gauge potential associated with the geometrical phase is abelian , as is the case for synthetic electromagnetic fields .
when the change in internal states involves a general rotation in the hilbert space spanned by the internal states , the gauge potential associated with the geometrical phase can be non - abelian , as is the case for synthetic soc .
experimentally , the adiabatic potential is typically generated by coupling the internal states of an atom with lasers , and it was spielman s group at nist that had first realized a uniform vector gauge potential in a bec of @xmath21rb atoms @xcite .
the scheme utilizes raman lasers to couple hyperfine states in the ground state manifold of @xmath21rb , such that when an atom jumps from one internal state to another via the raman process , its center - of - mass momentum also changes . applying a similar scheme , vortices
were later observed in a bec of @xmath21rb with synthetic magnetic field , followed by the implementation of synthetic electric field @xcite . in 2010 , via a slightly modified raman scheme ( see fig . [ fig_levelscheme ] ) , spielman s group was able to generate synthetic soc , a non - abelian version of the synthetic gauge field , in a bec of @xmath21rb atoms @xcite .
this was soon followed by the realization of synthetic soc in ultracold fermi gases by the shanxi group and the mit group in 2012 @xcite . in the following
, we will address the currently realized synthetic soc in cold atomic gases as the nist soc . in the case of the nist soc
, the effective single - particle hamiltonian can be written as @xmath22 where we have rotated the spin basis in the second line so that the effective hamiltonian takes the form of an equal mixture of rashba @xmath23 and dresselhaus @xmath24 soc . at the moment , only the effective one - dimensional soc in the form of eq .
( [ eqn_nistsoc ] ) has been implemented experimentally .
note that in eq .
( [ eqn_nistsoc ] ) , the effective rabi - frequency @xmath25 and the two - photon detuing @xmath19 can be seen as zeeman fields perpendicular and parallel to the direction of soc , respectively .
a fundamental effect of soc in ultracold fermi gases is the modification of the single - particle dispersion . under soc ,
both the symmetry and the low - energy density - of - states are different , leading to unconventional bec in bose systems , and exotic pairing states in fermi systems with attractive interactions . in this section
, we will focus on the single - particle dispersion under typical forms of soc , and discuss their potential impact on many - body systems .
the single - particle hamiltonian under the rashba soc can be written as @xmath26,\ ] ] where @xmath27 , @xmath28 ( @xmath29 ) is the creation ( annihilation ) operator for the pseudo - spin @xmath30 , @xmath31 is the soc strength , and @xmath32 stands for hermitian conjugate .
the pseudo - spin here is related to the adiabatic potential , and its exact relation with the hyperfine states is scheme dependent .
the hamiltonian can be diagonalized by introducing the annihilation operators in the so - called helicity basis @xmath33 with @xmath34 .
the diagonalized hamiltonian @xmath35 where @xmath36 are the resulting single - particle eigen energies . ) ; the solid curves are the spectra for the helicity branches ( @xmath37 ) .
( b ) the spectra of the helicity branches under soc and an out - of - plane zeeman field @xmath38 .
( c ) the single - particle spectrum of the lower helicity branch under the rashba soc , three - dimensional view .
( d ) the single - particle spectrum of the lower helicity branch under the nist soc , three - dimensional view.,width=340 ] the single - particle spectra under rashba soc is illustrated in fig .
[ fig_specsoc ] , where it is clear that soc breaks the inversion symmetry and splits the spin spectra into two helicity branches . due to the symmetry of the rashba soc , points of the lowest energy in the lower helicity branch form a ring in momentum space , and the ground state
is infinitely degenerate in this case [ see fig . [ fig_specsoc](c ) ] .
correspondingly , the density - of - states for a rashba spin - orbit coupled spectra in three dimensions is a constant at the lowest energy .
this peculiar single - particle spectra and density of states naturally lead to interesting many - body phenomena @xcite . for noninteracting bosons , as the ground - state degeneracy is infinite
, bec is no longer possible at zero temperature even in three dimensions .
however , a weak interaction can induce a spontaneous symmetry breaking , and the bosons will condense to either one point or two opposite points on the degenerate ring in momentum space , depending on the interaction parameters .
this leads to the so - called plane - wave phase and the stripe phase in a uniform interacting bose gas with rashba soc @xcite .
for a noninteracting fermi gas at zero temperature , atoms occupy all the low - energy states up to the fermi energy . as the atom number density increases , the topology of the fermi surface changes and the system undergoes a lifshitz transition .
furthermore , it has been shown that for atoms with attractive @xmath39-wave interaction between the spin species and with rashba soc , the two - body bound state energy is enhanced due to the increased density of states at low energies @xcite .
in fact , a two - body bound state exists in three dimensions even in the weak - coupling limit , similar to the case of a two - dimensional problem without soc , suggesting an effective reduction of dimensions , which is consistent with the soc - modified density of states .
the single - particle hamiltonian under the nist soc can be written as @xmath40 where @xmath28 ( @xmath29 ) is the creation ( annihilation ) operator for different hyperfine states .
compared with eq .
( [ eqn_nistsoc ] ) , the soc parameter @xmath31 is related to the momentum transfer of the raman process , and the effective zeeman field @xmath41 is proportional to the effective rabi - frequency .
the helicity basis that diagonalizes the hamiltonian becomes @xmath42 with @xmath43^{1/2}/\sqrt{2 } [ h^2+\alpha^2k_x^2]^{1/4}$ ] .
the single - particle dispersion of the helicity branches @xmath44 . when @xmath45 , the degenerate ground state manifold consists of two points in momentum space , in contrast to a ring under the rashba soc [ see fig .
[ fig_specsoc](d ) ] .
the spin components of the helicity branches are momentum dependent under the nist soc , which is also different from that of the rashba soc .
although these differences can lead to drastically different many - body effects , as we will show later , cold atomic gases under the nist soc still preserve many of the interesting properties of those under the rashba soc .
for example , in a bec under the nist soc , the stripe phase and the plane - wave phase can still be identified in the finite - temperature phase diagram @xcite . while in a noninteracting degenerate fermi gas ,
topological changes of the fermi surface have been observed experimentally ( see fig . [ zjsocexp ] ) . besides the rashba and the nist soc ,
the isotropic soc of the form @xmath46 , or the three - dimensional soc , has also been discussed recently by various authors @xcite . for bosons ,
spin textures of the ground state under an isotropic soc may acquire interesting topology . for fermions
, it is expected that the system under appropriate parameters can behave like three - dimensional topological insulators or weyl semimetals .
additionally , the application of an effective zeeman field can deform the single - particle dispersion spectra along the direction of the field . as we will see later
, this would lead to novel pairing states under attractive inter - particle interactions .
other more exotic forms of synthetic soc exist and have been discussed in the literature @xcite .
a common feature of these socs is the absence of counterparts in naturally occurring condensed - matter systems , which makes the quantum simulation of systems with exotic forms of soc in ultracold atomic gases more appealing .
with the understanding of single - particle dispersions under soc , we now examine the pairing physics in an attractively interacting fermi gas with soc .
we start with the conventional bcs pairing mechanism in the weak - coupling limit , which should help us to appreciate the fundamentally new pairing mechanisms in a spin - orbit coupled fermi gas .
it is well known that a two - component fermi sea without spin imbalance becomes unstable in the presence of a small attractive interaction .
the resulting cooper instability leads to a bcs ground state , which is the basis for understanding conventional superconductivity in most metals at low enough temperatures . in ultracold fermi gases ,
pairing superfluidity has been experimentally observed thanks to the feshbach resonance technique @xcite .
the effective bcs mean - field hamiltonian of such a system can be written as @xmath47 where @xmath48 is the bare interaction rate , @xmath49 is the pairing order parameter , and @xmath50 is the chemical potential , which , in the weak - coupling limit , is essentially the fermi energy of the system . under the hamiltonian ( [ eqn_bcspairing ] ) , the bcs pairing mechanism in the weak - coupling limit can be schematically illustrated as fig .
[ fig_bcspairing](a ) .
the dispersion of the spin - up particles crosses that of the spin - down holes at the fermi surface , where the pairing mean field @xmath51 couples the two branches , leading to an avoided crossing and the quaiparticle ( quasihole ) dispersions . from the positions of the avoided crossings
, we see that the pairing state in this case has zero center - of - mass momentum . in the presence of a spin imbalance , or an effective zeeman field ,
the fermi surface mismatch between the two spin species can lead to competition between the bcs pairing mechanism and the fulde - ferrell - larkin - ovchinnikov pairing mechanism @xcite . in the latter case ,
the dispersion of the minority spin species is shifted so that pairing can take place near the fermi surface [ see fig .
[ fig_bcspairing](b ) ] , thus leading to a finite center - of - mass momentum pairing state . in both of these cases ,
an important feature is that pairing involves both the particle and the hole branches of different spin species .
we address this kind of pairing mechanism as interbranch pairing .
note in fig .
[ fig_bcspairing](b ) and in the rest of the review , we consider the simple case where pairing occurs with a single - valued center - of - mass momentum , the so - called fulde - ferrell ( ff ) state . under soc , with different single - particle dispersions , a new pairing mechanism emerges , which becomes more transparent and also physically more interesting in the presence of effective zeeman fields . extending the bcs mean - field theory ,
we may write down the effective mean - field hamiltonian of a two - component fermi gas under the rashba soc and an effective out - of - plane zeeman field @xmath52 where @xmath41 is the effective zeeman field along the @xmath53 direction .
the hamiltonian above is written the pseudo - spin basis @xmath54 . under soc , the mean - field ground state of the system has both @xmath39- and @xmath55-wave pairing components , as both @xmath56 and @xmath57 ( @xmath58 ) are finite @xcite .
starting from the hamiltonian above , we now examine the pairing mechanism in the weak - coupling limit . in the absence of any zeeman fields , as illustrated in fig .
[ fig_specsoc](a ) , the two helicity branches cross at the origin in momentum space . under an effective out - of - plane zeeman field , i.e. ,
when @xmath41 is finite , the helicity branches are coupled and a gap opens up at the origin [ see fig . [ fig_bcspairing](c ) ] .
when the chemical potential lies in this gap , as is the case in fig .
[ fig_bcspairing](c ) , pairing can occur within the lowest helicity branch , leading to the opening of pairing gaps at the crossings of the particle and hole dispersions of the lower branch .
this is possible since soc mixes different spin species so that @xmath39-wave pairing between spin - up and spin - down atoms can happen within the same helicity branch .
we call this pairing mechanism intrabranch pairing .
as we will demonstrate in the next section , for a rashba soc , this is the pairing scenario that results in the exotic topological superfluid state . when the chemical potential lies above the gap , or when the interaction becomes stronger , interbranch pairing becomes important and competes with intrabranch pairing . from the location of the avoided crossings
, we see that pairing states in this case have zero center - of - mass momentum . when an in - plane zeeman field is added , the single - particle dispersion spectra become deformed . as shown in fig .
[ fig_bcspairing](d ) , in this case , intrabranch pairing naturally leads to pairing states with finite center - of - mass momenta .
apparently , these soc - induced ff pairing states are the result of interplay between soc and fermi surface asymmetry , and are different in mechanism from the conventional ff ( cff ) states discussed previously . with these qualitative analysis in the weak - coupling limit
, we have already seen the various possibilities of exotic pairing states under soc . in the following ,
let us see some more concrete examples .
perhaps the most important exotic pairing state in a spin - orbit coupled fermi system is the topological superfluid state in two dimensions , originally investigated in the context of semiconductor / superconductor heterstructures with rashba soc , @xmath39-wave pairing order and an out - of - plane zeeman field @xcite . as has been shown in ref .
@xcite , soc breaks the inversion symmetry and mixes up different spin components in the helicity branches .
the introduction of the external zeeman field breaks the time - reversal symmetry and opens up a gap between different helicity branches . in the weak - coupling limit ,
when the fermi surface of the system lies in this gap , the ground state of the system becomes a topological superfluid state with majorana zero modes on the boundaries or at the cores of vortex excitations .
importantly , these non - abelian majorana zero modes are protected by a bulk gap and are therefore useful for fault - tolerant topological quantum computation @xcite .
the stabilization of the topological superfluid state can be understood as the direct result of intrabranch pairing , which leads to a mixture of @xmath39- and @xmath55-wave symmetries in the pairing order parameter @xcite . in parameter regions where the @xmath55-wave pairing dominates
, the @xmath59 symmetry of the pairing order parameter inherited from the rashba soc allows one to map the system onto a @xmath55-wave superfluid with the corresponding symmetry @xcite .
-@xmath50 plane for a two - dimensional fermi gas under rashba soc and an out - of - plane zeeman field @xmath41 .
the first - order phase transition is shown in red solid curve while the second - order phase transitions are shown in dash - dotted black curves . the thin dashed curve in the topological superfluid ( tsf ) region marks the @xmath60 threshold .
the effective zeeman field @xmath41 is taken to be the energy unit , while the unit of momentum @xmath61 is defined through @xmath62 . here , vac represents the vacuum .
see ref .
@xcite.,width=302 ] in the context of ultracold atomic gases , a natural question is whether the topological superfluid state and the associated majorana zero modes can be prepared and probed .
although the experimental realizaiton of the rashba soc has not been achieved yet , there have been various theoretical proposals on its implementation in cold atoms , which gives rise to even more theoretical characterizations of the topological superfluid phase and the related phase transitions in the context of ultracold fermi gases . in two dimensions
, it is found that a topological superfluid phase which supports majorana zero modes can be stabilized @xcite . in a uniform system
, these different phases will give rise to phase - separated states with various first - order boundaries ( see fig .
[ fig_eb ] ) . in a trapping potential , which is more relevant to the experimental conditions , these different phases naturally separate in space , leading to shell structures similar to those in a polarized fermi gas @xcite .
another intriguing finding on the mean - field phase diagram fig .
[ fig_eb ] is the absence of a normal phase , even in the limit of large magnetic field @xmath63 .
instead of being zero in the case without soc , here , the order parameter decreases exponentially with increasing @xmath41 , but remains finite within the entire sf and tsf regime .
this somewhat counterintuitive result is an artifact of the mean - field approach , which completely neglects quantum and thermal fluctuations . a more careful analysis incorporating particle - hole fluctuations in the large - zeeman - field limit shows that there exits a so - called polaron - molecule transition , suggesting that the true ground state is a normal phase , at least in the weak - coupling regime @xcite .
when the rashba soc is replaced by the less symmetric nist soc , the topological superfluid phase would be replaced by various nodal superfluid phases ( nsf1 and nsf2 ) , with either two or four discrete nodal points along the direction with no soc .
the mean - field zero - temperature phase diagram in two dimensions is shown in fig . [ fig_socchapffphase](a ) . note that the cff phase is due to the interbranch pairing mechanism and not the soc - induced ff states that we have discussed previously . in three dimensions , neither the rashba soc nor
the nist soc would lead to a topological superfluid , as the pairing superfluid phases therein are either fully gapped trivial superfluid ( sf ) state or nodal superfluid phase .
interestingly , for gapless superfluid states in a three dimensional fermi gas with the nist soc , the nodal points typically form closed surfaces in momentum space @xcite .
an important consequence of the intrabranch pairing is the dramatically enhanced ff pairing states under soc and fermi surface asymmetry . as illustrated in fig .
[ fig_bcspairing](c ) , under an additional in - plane zeeman field , the fermi surface is deformed , such that it no longer has inversion symmetry along the axis of the transverse field . in the weakly interacting limit ,
it is clear that a simple bcs pairing state with zero center - of - mass momentum becomes energetically unfavorable .
this opens up the possibility of an exotic fflo - like pairing state with finite center - of - mass momentum @xcite . from the general argument
above , we may further infer that such a pairing state is a natural result of the co - existence of soc and fermi surface asymmetry , and should generally exist in such systems , regardless of the exact type of soc in the system . indeed , these exotic pairing states have been reported to exist in various systems of different dimensions and with different forms of soc @xcite . to illustrate this
, we first consider an experimentally relevant system , where the nist soc is imposed on a two - dimensional fermi gas with effective axial and transverse zeeman fields .
similar to the rashba case , a mean - field description of the pairing states here can be seen as a natural extension of the standard bcs theory @xmath64 where @xmath65 , the order parameter @xmath66 .
the hamiltonian ( [ eqnheffsoc ] ) has been written under the hyperfine - spin basis @xmath67 .
again , the soc parameter @xmath31 is related to the momentum transfer of the raman process in the nist scheme , and the effective zeeman field @xmath41 and @xmath68 are proportional to the effecitve rabi - frequency and the two - photon detuning of the raman lasers , respectively .
the zero - temperature thermodynamic potential can be obtained by diagonalizing the effective hamiltonian @xmath69 where the quasiparticle ( quasihole ) dispersion @xmath70 ( @xmath71 , @xmath72 ) are the eigenvalues of the matrix in hamiltonian ( [ eqnheffsoc ] ) , and @xmath73 is the heaviside step function . from the previous general analysis in the weak - coupling limit
, the bcs pairing states with zero center - of - mass momentum would become unstable against an ff pairing state under the fermi surface asymmetry .
this implies an instability of the bcs state with finite @xmath68 .
this point can be demonstrated by performing a small @xmath74 expansion around the local minimum in the thermodynamic potential landscape that corresponds to the bcs pairing state @xmath75 where we have assumed @xmath76 .
it is then straightforward to demonstrate numerically that for @xmath77 , we have @xmath78 , @xmath79 ; while for @xmath80 , we have @xmath81 , which has an opposite sign to that of @xmath68 .
this is a direct evidence that the bcs pairing state in the presence of fermi surface asymmetry and soc becomes unstable against an ff state , with the center - of - mass momentum @xmath74 opposite to the direction of the transverse field @xmath68 .
a qualitative understanding of these ff states is that the combination of soc and fermi surface asymmetry shifts the local minima that corresponds to bcs pairing states ( @xmath82 ) onto the finite-@xmath74 plane .
this is further reflected by the observation that the magnitude of the center - of - mass momentum @xmath83 decreases as fermi surface asymmetry becomes smaller , i.e. , with increasing chemical potential @xmath50 or soc strength @xmath31
@xcite .
we show the mean - field phase diagram in fig .
[ fig_socchapffphase](b ) . recalling fig .
[ fig_eb ] , it is clear that the topological superfluid state under the rashba soc is now replaced by a nodal ff ( nff ) state , while the trivial superfluid state is replaced by a gapped ff ( gff ) state .
these soc - induced ff states originate from intrabranch pairing , and are different in mechanism from the conventional ff state . among these soc - induced ff states ,
perhaps the most interesting case is the topological ff ( tff ) state , where the pairing state can have nonzero center - of - mass momentum and topologically nontrivial properties simultaneously @xcite .
this exotic pairing phase can be understood as derived from a typical topological superfluid phase in two dimensions , where rashba soc , @xmath39-wave pairing order and an out - of - plane zeeman field coexist . with the addition of another effective zeeman field in the transverse direction ,
the fermi surface becomes asymmetric , and according to the preceding analysis , the ground state of the system necessarily acquires a nonzero center - of - mass momentum .
more importantly , the ground state would inherit all topological properties from the topological superfluid state , provided that the deformation of the fermi surface should not be drastic enough to close the bulk gap .
the phase diagram of a two - dimensional fermi gas under rashba soc and cross zeeman fields is shown in fig .
[ fig_socchapffphase](c ) . apparently , the stability region of the tff state roughly corresponds to the topological superfluid state in fig .
[ fig_eb ] .
an advantage of ultracold atomic gases is their highly tunable parameters .
while soc provides the possibility of engineering single - particle dispersion , in principle , one can also change the interaction strength via the well - established feshbach resonance technique .
in fact , when combining soc with other available tools in ultracold atomic gases , we can engineer highly nontrivial states . .
the impurity atoms interact only with the spin - up atoms in a spin - orbit coupled fermi gas .
the fractions of spin components in both helicity branches are momentum dependent , as characterized by @xmath84 and @xmath85 , where @xmath86 is defined in sec .
[ sec_singlespec ] .
see ref .
@xcite.,width=264 ] as an example , we consider a three - component fermi gas , where one fermion species ( ` impurity ' ) is tuned close to a wide feshbach resonance with one of the spin - species in a two - component fermi gas under the nist soc @xcite .
the pairing mechanism here is illustrated in fig .
[ fig_schematic ] . under the nist soc ,
the spin components in both helicity branches are momentum dependent .
the pairing states naturally acquire a nonzero center - of - mass momentum , which is dependent on the position of the fermi surface as well as the interaction strength .
for instance , in the weak - coupling limit , when the fermi surface lies in the lower branch , the center - of - mass momentum is negative , since the pairing is lower - branch dominated ; while if the fermi surface lies in the upper branch , the center - of - mass momentum can be positive , as the ground state is the result of the competition between lower - branch - dominated pairing and the upper - branch - dominated pairing .
this can lead to the competition between pairing states with different center - of - mass momentum , as illustrated in fig .
[ fig_compete ] . apparently , the ff pairing states in this system is different from the soc - induced ff states discussed previously .
qualitatively , we are replacing the fermi surface asymmetry in the previous section with the asymmetry in the spin degrees of freedom of the helicity basis .
relative to the fermi energy @xmath87 as functions of the center - of - mass momentum @xmath88 , with @xmath89 ( solid ) , @xmath90 ( dashed ) , and @xmath91 ( dash - dotted ) . the unit of energy is the fermi energy @xmath92 of a two - component , noninteracting fermi gas in the absence of soc and with the same total number density .
( b ) typical phase diagram in the ( @xmath93 ) plane in a li - k - k system .
the red ( blue ) curve shows the transition boundary between the normal state ( n ) and ff states with different center - of - mass momentum ( ff1 and ff2 ) .
see ref .
@xcite.,width=340 ] a notable feature of the current system is that the two - body bound state is suppressed by soc . in the absence of soc ,
the spin - down atoms are decoupled , and the two - body bound state between the ` impurity ' and the spin - up atom emerges at resonance with @xmath94 , where @xmath95 is the scattering length between the spin - up atom and the ` impurity ' atom . under soc , this two - body bound state threshold is pushed toward the bec limit with @xmath96 ( see fig . [ fig_twobodythresh ] ) .
roughly , as soc establishes ` correlation ' between the spin - up and the spin - down atoms , it is more difficult for the ` impurity ' and the spin - up atoms to combine and form a bound state . as we will see later , a more rigorous description of this peculiar ` correlation ' is the symmetry in the soc - modified single - particle dispersion .
this suppression of two - body bound state naturally leads to the question of the stability of three - body bound states , i.e. , whether a three - body bound state can be more stable than the two - body bound state in certain parameter region ? a short answer to the question above is : not with the nist soc @xcite .
however , it is possible to stabilize three - body bound states , or trimers , if we consider other forms of soc , especially those with a high symmetry .
indeed , if we replace the nist soc with the rashba soc in the previously discussed three - component fermi system , trimers can be stabilized even in the absence of any two - body bound states @xcite .
the stabilization of this so - called borromean state can be understood from the special symmetry of the rashba - modified single - particle dispersion spectra ( see fig . [ fig_borromean ] ) . as illustrated in fig .
[ fig_specsoc](c ) , under the rashba soc , the degenerate subspace of the single - particle ground state forms a ring in momentum space .
with such a spectral symmetry , the scattering within the lowest - energy subspace is blocked due to the total momentum conservation ( see fig . [ fig_borromean ] ) , which effectively suppresses the formation of a two - body bound state .
in contrast , the three - body scattering is not blocked . under the nist soc ,
the single - particle dispersion spectra have less symmetry . hence , although the two - body bound states are also suppressed in those systems , they remain to be energetically more favorable than the trimers . and a fixed @xmath97 . the mass ratio @xmath98 , where @xmath99 is the mass of the impurity atom , @xmath100 is the mass of the atoms under soc . here ,
@xmath101 ( solid ) , @xmath102 ( dashed ) , and @xmath103 ( dash - dotted ) . the unit of energy @xmath104 , and the unit of momentum @xmath105 .
see ref .
@xcite.,width=264 ] . in this @xmath106 system ,
@xmath107 atoms are subject to the rashba soc , and @xmath108 serves as the ` impurity ' .
( upper left ) the two - body scattering within the lowest - energy subspace of the @xmath109 system can not proceed due to the conservation of total momentum .
( upper right ) the three - body scattering within the lowest energy subspace of the @xmath106 system is allowed ( green dashed arrows ) .
( lower panel ) phase diagram for the system in sec .
[ sec52 ] on the @xmath110 plane , where @xmath111 .
the lower and upper solid curves respectively represent the threshold of borromean ( ` b ' ) state and the boundary at which the ordinary trimer ( ` t ' ) merges into the atom - dimer continuum ( ` ad ' ) .
the blue dashed curve is the dimer threshold , which is also the boundary between ` b ' and ` t ' for @xmath112 .
see ref .
@xcite , title="fig:",width=264 ] . in this @xmath106 system ,
@xmath107 atoms are subject to the rashba soc , and @xmath108 serves as the ` impurity ' .
( upper left ) the two - body scattering within the lowest - energy subspace of the @xmath109 system can not proceed due to the conservation of total momentum .
( upper right ) the three - body scattering within the lowest energy subspace of the @xmath106 system is allowed ( green dashed arrows ) .
( lower panel ) phase diagram for the system in sec .
[ sec52 ] on the @xmath110 plane , where @xmath111 .
the lower and upper solid curves respectively represent the threshold of borromean ( ` b ' ) state and the boundary at which the ordinary trimer ( ` t ' ) merges into the atom - dimer continuum ( ` ad ' ) .
the blue dashed curve is the dimer threshold , which is also the boundary between ` b ' and ` t ' for @xmath112 .
see ref .
@xcite , title="fig:",width=264 ] borromean states have been found before in different physical systems . however , previously studied borromean states are nonuniversal , with properties dependent on the short - range details of the two - body interaction potential @xcite . here , the borromean states are universal , with properties only dependent on the two - body @xmath39-wave scattering length and the soc coupling strength . furthermore , near the trimer threshold of such a system , three - body resonances should give rise to interesting many - body states with exotic few - body correlations . the universal borromean state in the spin - orbit coupled fermi system
is thus highly nontrivial .
note that universal trimers have also been identified in a related system , where an atom under isotropic soc interacts with two other atoms @xcite .
both of these cases suggest that exotic few - body states may be stabilized by introducing soc as another dimension of control .
in this brief review , we have focused on the general pairing mechanism in a spin - orbit coupled fermi gas .
as soc changes the single - particle dispersion spectra , pairing superfluidity in the system takes on many new and exotic forms : topological superfluid , soc - induced ff state , topological superfluid so on so forth . besides these exotic pairing states
, soc can also lead to novel universal trimer states , which features exotic few - body correlations upon which even more interesting many - body phases may emerge . in all these cases , by modifying the single - particle dispersion , soc proves to be a powerful tool of quantum control , which , when combined with the outstanding tunability of ultracold atomic gases , is playing an increasingly important role in fulfilling the potential of quantum simulation with ultracold atomic gases .
this work is supported by nfrp ( 2011cb921200 , 2011cba00200 ) , nkbrp ( 2013cb922000 ) , nnsf ( 60921091 ) , nsfc ( 10904172 , 11104158,11374177,11105134 , 1127409,11374283 ) , the fundamental research funds for the central universities ( wk2470000006 ) , the research funds of renmin university of china ( 10xnl016 ) , and the programs of chinese academy of sciences .
w. ketterle and m. w. zwierlein , _ making , probing and understanding ultracold fermi gases _ , ultracold fermi gases , proceedings of the international school of physics `` enrico fermi '' , course clxiv , varenna , 20 - 30 june 2006 , edited by m. inguscio , w. ketterle , and c. salomon ( ios press , amsterdam ) ( 2008 ) . | we review some recent progresses on the study of ultracold fermi gases with synthetic spin - orbit coupling . in particular , we focus on the pairing superfluidity in these systems at zero temperature .
recent studies have shown that different forms of spin - orbit coupling in various spatial dimensions can lead to a wealth of novel pairing superfluidity .
a common theme of these variations is the emergence of new pairing mechanisms which are direct results of spin - orbit - coupling - modified single - particle dispersion spectra . as different configurations
can give rise to single - particle dispersion spectra with drastic differences in symmetry , spin dependence and low - energy density of states , spin - orbit coupling is potentially a powerful tool of quantum control , which , when combined with other available control schemes in ultracold atomic gases , will enable us to engineer novel states of matter . |
the last 15 years have seen a significant series of developments in quantum theory , stemming from the realization by bender and co - authors that a class of non - hermitian hamiltonians possess real spectra under the parity - time ( @xmath0 ) symmetry condition , where parity - time means spatial reflection and time reversal , @xmath1 and @xmath2 @xcite .
this mathematical discovery has initiated numerous studies of open systems with balanced gain and loss even though the generality of this construction is under discussion @xcite .
experimental setups have been offered to create @xmath0-symmetric physical systems in optics @xcite , electronic circuits @xcite , as well as in mechanical systems @xcite . in a number of theoretical studies
it has been demonstrated that @xmath0-symmetric systems often demonstrate unusual and counterintuitive properties .
these include , among others , unconventional beam refraction @xcite , bragg scattering @xcite , symmetry - breaking transitions @xcite and associated ghost states @xcite , a loss - induced optical transparency @xcite , conical diffraction @xcite , a new type of fano resonance @xcite , chaos @xcite , nonlocal boundary effects @xcite , optical switches @xcite and diodes @xcite , phase sensitivity of light dynamics @xcite , and the possibility of linear and nonlinear wave amplification and filtering @xcite .
unexpected instabilities were also identified at the level of @xmath0-symmetric lattices and nonlinear modes were identified in few - site oligomers , as well as in full lattice settings both in 1d @xcite and even in 2d @xcite .
extensions of @xmath0-symmetric considerations in the setting of active media ( of unequal gain and loss ) have also recently been proposed @xcite . motivated by the linear oscillator problems associated with ( linear ) electrical @xcite and mechanical @xcite @xmath0-symmetric experiments ,
klein - gordon field - theoretic generalizations with a @xmath0-symmetric defect have been proposed and the collective variable method has been developed to describe kink dynamics in the system @xcite ; see also for a detailed discussion @xcite .
it was also shown that standing kinks in such models are stable if they are centered at the loss side of the defect @xcite and standing breather may exist only if centered exactly at the interface between gain and loss regions @xcite .
a natural question arises what happens with the moving klein - gordon solitary waves when they interact with the spatially localized @xmath0-symmetric defect .
interaction of solitary waves with local inhomogeneities of media has been attracting attention of researchers for the last two decades .
the reflection windows were observed in the kink - impurity interactions by fei _
et al_. in the sine - gordon ( sg ) @xcite and @xmath3 @xcite models
. scattering of sg breather by localized defects has been investigated in the conservative case @xcite .
it has been shown that the breather can split into a kink and antikink pair or can be accelerated by the defect .
this is possible in conservative systems because the translational kinetic energy of the breather can be partly converted into its internal energy and vice versa .
scattering of linear and nonlinear waves ( solitons ) on defects in @xmath0-symmetric optical waveguide arrays was analyzed @xcite .
it was shown that the incident high - amplitude solitons ( or even linear wavepackets @xcite ) can excite a mode localized on the @xmath0-symmetric defect . by exciting the localized mode of a large amplitude
, it is possible to perform phase - sensitive control of soliton scattering and amplification or damping of the localized mode .
the gain - loss pattern in conjunction with the nonlinearity lead to asymmetric propagation of the incoming wavepackets depending on their direction of incidence .
kinks in non - integrable models such as the perturbed sg equation or @xmath3 model can support internal vibrational modes @xcite . in some cases , impurities can also support localized vibrational modes .
kinks of the integrable sg equation do not bear internal modes @xcite .
when a kink hits an impurity in a conservative model , a part of its energy is trapped towards the excitation of the impurity mode @xcite and another fraction leads to the emission of radiation bursts @xcite .
a merger of a colliding kink and antikink into a breather is possible in a non - integrable system when energy loss to radiation and/or excitation of the kink s internal modes is sufficiently large @xcite .
the binding free kink and antikink into a breather has been addressed in @xcite in presence of spatially periodic perturbation .
external d.c . driving force in the absence of damping for sufficiently large magnitude of the force causes the breather to split into a kink - antikink pair while for small driving force the breather excitations lead to stationary modes @xcite .
the breather can dissociate into a kink - antikink pair under external field @xcite .
conversion of an oscillation mode into a kink - antikink pair has been observed via abrupt distortions of the on - site potential in time or in space @xcite . the recent work of @xcite illustrated that such an evolution is also possible when the breather is subject to gain e.g. on the gain side of a @xmath0-symmetric medium .
interaction of moving solitons with @xmath0-symmetric defects in the realm of the klein - gordon field , to the best of our knowledge , has not been studied previously , in part , arguably , since @xmath0-symmetric field theories is a very recent theme of research . in this paper
, we aim to reveal the principal physical effects observed during the interaction of sg kinks and breathers with a @xmath0-symmetric defect with balanced gain and loss .
in particular , a kink approaching the @xmath0-symmetric defect from the gain side is always transmitted , while from the loss side it may be reflected or transmitted depending on its energy .
this suggests an asymmetric effective dynamics which is identified by means of an explicitly solvable collective coordinate approach .
we also illustrate how this effective collective dynamics can be modified by the presence of the repulsion from an another kink or of the attraction by an antikink . on the other hand , for the breather the dynamics
is sensitively dependent on both the characteristics of the breather and those of the defect , potentially exhibiting either gain or loss of energy for the coherent structure ( the former possibly even featuring the breakup of the breather into a kink and an anti - kink waveforms ) .
the structure of the paper is as follows . in sec .
[ sec : ii ] , following the work @xcite , we introduce the perturbed sg equation and the well - known kink and breather solutions to the integrable sg equation . in sec .
[ sec : collectvar ] , a collective variable method is applied and analytically solved to reveal some features of the kink dynamics in the considered system .
we then report on the numerical results for scattering of kinks in sec .
[ sec : kink ] , pairs of kinks in sec .
[ sec : kinkkink ] , breathers in sec .
[ sec : breather ] and kink - antikink pairs in sec .
[ sec : kinkantikink ] .
our conclusions and some future directions are presented in sec .
[ sec : v ] .
we consider a perturbed sine - gordon equation of the form @xcite : @xmath4 where @xmath5 is the unknown scalar field and lower indices denote partial differentiation .
the perturbation term is in the right - hand side of the equation .
the parameter @xmath6 controls the perturbation amplitude . in order to study the effects of a spatially localized @xmath0-symmetric defect on traveling kinks and breathers , for the function @xmath7 we take @xmath8-\exp[-\beta(x-\delta)^2]\},\ ] ] which has the symmetry @xmath9 .
this ensures that eq .
( [ sge_perturbation ] ) is @xmath0-symmetric physically implying that while eq .
( [ sge_collisions ] ) describes an open system with gain and loss , the gain balances the loss . the gain - loss spatial profile determined by eq .
( [ sge_perturbation ] ) represents a superposition of two bell - shaped functions with the separation between them controlled by the parameter @xmath10 .
the parameter @xmath11 is related to the hump inverse width . for @xmath12 and/or @xmath13 one
has @xmath14 . for definiteness
, here we consider the case of @xmath15 with the gain ( loss ) region @xmath16 ( @xmath17 ) .
in the present study the simulations are carried out for different values of the perturbation amplitude @xmath6 and fixed @xmath18 and @xmath19 .
the choice of @xmath18 makes the hump width comparable to the kink width . for @xmath14
, we have the integrable sg equation with the following kink solution @xmath20\},\ ] ] and the breather solution @xmath21\}}{\omega\cosh[\delta_b\eta(x - x_0-v_bt)]},\ ] ] where @xmath22 is kink velocity , @xmath23 , @xmath24 are the breather velocity and frequency , @xmath25 is the soliton initial position and @xmath26 the energy of the kink and the breather are , respectively @xmath27 far from the defect , solitons move with constant velocities @xmath22 and @xmath23 feeling no perturbation . in the vicinity of the defect , the soliton parameters change and , as it will be shown ,
it is important from which side the soliton hits the defect . to study numerically the effect of the perturbation on the dynamics of the sg solitons , we introduce the mesh @xmath28 , where @xmath29 is the lattice spacing , @xmath30 and propose the following discrete version of the model @xmath31 in which @xmath32 and @xmath33 .
it can be seen that the term @xmath34 in eq .
( [ sge_collisions ] ) is discretized with the accuracy @xmath35 which has already been used by other authors @xcite .
this is done to minimize the effect of discreteness introduced by the mesh .
equations of motion ( [ fk_collisions ] ) were integrated with respect to the temporal variable using an explicit scheme with the accuracy of @xmath36 and the time step @xmath37 .
the simulations reported below in section iv were conducted for @xmath38 and @xmath39 .
a collective variable approach has been developed @xcite to describe the kink dynamics in the model eq .
( [ sge_collisions ] ) .
the kink is effectively described by the one degree of freedom particle of mass @xmath40 , which is the mass of standing kink .
the kink coordinate @xmath41 as a function of time @xmath42 can be found from the following equation of motion @xmath43 with @xmath44 ^ 2\gamma(x)dx,\end{aligned}\ ] ] where the overdot means differentiation with respect to time and the prime denotes differentiation with respect to @xmath45 . substituting the kink solution eq .
( [ kink ] ) into eq .
( [ coll_var1a ] ) one obtains @xmath46}.\end{aligned}\ ] ] the equation of motion ( [ coll_var1 ] ) was integrated numerically for the initial conditions @xmath47 , @xmath48 using the simplest scheme @xmath49 where @xmath50 denotes the time step number , @xmath51 , @xmath52 , and @xmath39 is the time step . the collective variable equation ( [ coll_var1 ] ) can also be solved explicitly with its solution given in the form of a quadrature .
the first integral reads @xmath53 where @xmath54 and @xmath55 is the integration constant .
the second integration gives @xmath56 with the integration constant @xmath57 .
equation ( [ denis6 ] ) gives the solution to eq .
( [ coll_var1 ] ) in an implicit form @xmath58 .
the kink approaching the defect from the gain ( loss ) side is first accelerated ( decelerated ) and then decelerated ( accelerated ) when it enters the lossy ( gain ) side . as a result , the kink experiences a phase shift . to calculate the phase shift we substitute eq .
( [ coll_var2 ] ) into eq .
( [ denis4 ] ) : @xmath59},\end{aligned}\ ] ] where @xmath60 . the function @xmath61}\end{aligned}\ ] ] is odd and hence the function @xmath62 is even .
note that @xmath63 decays exponentially when @xmath64 . from the last statement
it follows the existence and the equality of the following limits @xmath65 where @xmath66 is the value of the limits .
coming back to eq .
( [ denis6 ] ) , we note that the integrand can be presented as the sum @xmath67 substitution of the last equation into eq .
( [ denis6 ] ) gives @xmath68 the integral in eq .
( [ denis13 ] ) is bounded uniformly in @xmath45 since the integrand decays exponentially at infinity .
the right hand side in ( [ denis13 ] ) is a linear function in @xmath42 .
hence , as @xmath69 , each solution to equation ( [ coll_var1 ] ) should behave as @xmath70 where @xmath71 indicates terms bounded as @xmath69 , and @xmath22 is in fact the kink velocity given by the formula @xmath72 the last equation expresses the kink velocity in terms of the model parameters .
below we assume that @xmath73 , and the case of @xmath74 can be treated in a similar way . the function @xmath75 eq .
( [ denis21 ] ) grows at infinity linearly and hence the integral in the left hand side in ( [ denis13 ] ) tends to a constant as @xmath69 .
thus , we can specify behavior of eq .
( [ denis21 ] ) as follows , @xmath76 where now the symbol @xmath77 stands for the terms vanishing as @xmath69 . the quantity @xmath78 is in fact the kink s phase shift due to the defect , which we now calculate .
in order to do it , we substitute eq .
( [ denis14 ] ) into eq .
( [ denis13 ] ) , taking into consideration eq .
( [ denis11 ] ) : @xmath79 in the limit @xmath69 , for positive @xmath22 one has @xmath80 , and eq .
( [ denis15 ] ) becomes @xmath81 subtracting one identity from the other one , finds @xmath82 the integration constant @xmath55 can be found from eq .
( [ denis11 ] ) that allows us to rewrite eq .
( [ denis17 ] ) as @xmath83 if @xmath74 , the similar formula reads as @xmath84 for the kink solution ( [ kink ] ) the function @xmath85 can be cast into the particular form @xmath86}.\ ] ] after changing the order of integration and integrating over @xmath87 one obtains @xmath88\}dx.\ ] ] the kink s phase shift can be now found from eqs .
( [ denis19 ] ) , ( [ denis22 ] ) and eq . ( [ denis23 ] ) by evaluating the integrals numerically .
if the kink approaches the defect from the loss side , it must have sufficient momentum not to be trapped .
the critical kink initial velocity @xmath89 can be found with the help of the collective variable method .
one can present eq .
( [ coll_var1 ] ) for @xmath90 in the form @xmath91}. \nonumber \\\end{aligned}\ ] ] a kink having critical velocity must have @xmath92 at @xmath93 , i.e. , the kink stops when it reaches the center of the defect . setting in eq .
( [ coll_var4 ] ) @xmath94 and @xmath92 after integrating over the the collective variable @xmath45 we have @xmath95- { \tanh}[\delta_k(x - x)]\}dx . \nonumber \\\end{aligned}\ ] ] the value of the integral in eq .
( [ coll_var5 ] ) can be found numerically for the initial condition @xmath96 and recalling that the final stopping point is @xmath97 . for @xmath18 and @xmath19
used in our study one finds @xmath98 . for small kink velocity @xmath99 and @xmath40
so that @xmath100
first we start with the case of the kink - defect interaction , which is simpler . in fig .
[ fig1 ] the results for the case when the kink approaches the defect with @xmath101 from the gain side are presented . in ( a )
the kink position as the function of time is shown by the solid lines for the two values of the initial kink velocity , @xmath102 and @xmath103 , as indicated for each curve .
dashed lines give the results obtained with the help of the theoretical collective variable method eq .
( [ coll_var3 ] ) .
one can see that the collective variable approach gives a very accurate prediction of the actual kink dynamics . in ( b )
time evolution of the kink energy @xmath104 is plotted . from fig .
[ fig1 ] it is clearly seen that the kink moving toward the defect from the gain side is first accelerated and after passing the gain side of the defect it is decelerated by the loss side .
after the kink passes the defect and moves far from it , it restores its initial velocity and energy .
the only effect of the kink - defect interaction in this case is a phase shift .
the maximal kink energy increases with increase in the kink initial velocity @xmath22 for fixed defect amplitude @xmath6 because of the nature of the defect , whose effect is stronger for larger @xmath105 . and
@xmath103 , for the case when the kink approaches the defect from the gain side .
the defect center is located at @xmath106 .
solid lines show the results of the numerical solution for the continuous system and dashed lines show the results obtained with the help of the collective variable method .
horizontal dotted line shows location of the defect center .
( b ) time evolution of the kink total energy with the initial velocities @xmath102 and @xmath103 during the interaction with the defect .
the perturbation amplitude is @xmath101 in both cases.,width=359 ] next , suppose a kink comes from the lossy side . in this case ,
two different scenarios for the kink interaction with the defect are possible depending on its initial velocity @xmath22 ( or perturbation strength @xmath6 ) as presented in fig .
if @xmath22 is large enough ( or @xmath6 is small enough ) , the kink passes through the lossy part of the defect with the velocity smaller than @xmath22 and enters the gain part where it is accelerated up to the initial velocity and then goes on to infinity . in the opposite case ( where @xmath22 is not large enough or @xmath6 is not small enough ) , the kink does not possess sufficient momentum to pass through the lossy part of the defect and it is trapped there . in fig .
[ fig2 ] ( a ) the kink position as a function of time is shown for @xmath107 and @xmath101 with @xmath108 in both cases .
figure [ fig2 ] ( b ) shows the kink total energy as a function of time for these two cases .
as one can see , for the case of @xmath107 kink passes through the defect and restores its initial velocity , while for @xmath101 the kink is trapped by the lossy side of the defect . and
the results are given for @xmath107 and @xmath101 .
the results for the full system described by the partial differential equation ( pde ) of eq .
( [ sge_collisions ] ) ( solid lines ) and the ordinary differential equation ( ode ) of the collective variable approach ( dashed lines ) are compared .
the dotted line shows the location of the defect center .
( b ) time evolution of the kink total energy for the same two cases.,width=359 ] the kink approaching the defect from the lossy side is always trapped there . above the line it passes through the defect and
restores its initial velocity .
solid line is for the continuum system , while dashed line is for collective variable method.,width=359 ] we now further expand on our comparison of the kink dynamics observed in the continuous pde system of eq .
( [ sge_collisions ] ) with that in the single degree of freedom ode model of eq .
( [ coll_var1 ] ) . in fig .
[ fig3 ] the plane of the parameters @xmath6 and @xmath22 is shown with the line which separates the two possible scenarios of the kink - defect interaction when the kink approaches the defect from the lossy side . above the line the kink has sufficient initial momentum to pass through the defect and to restore its initial velocity .
contrary to this , below the line the kink is always trapped in the lossy region of the defect and eventually stops .
the collective variable result eq .
( [ vc ] ) is shown by the dashed line .
the result obtained for the continuum system ( shown by the solid line ) is in a perfect agreement with the collective variable method for small kink velocity and the deviation increases for larger kink velocities .
this is natural to expect as the collective coordinate derivation of @xcite was obtained away from the relativistic regime of large speeds @xmath22 .
nevertheless , we observe that for speeds even nearly half the maximal speed of propagation in the medium the relevant collective coordinate prediction remains very accurate . in fig .
[ fig4 ] , the kink s phase shift due to interaction with the @xmath0-symmetric defect is presented as a function of its initial velocity .
solid lines show the results of the numerical solution for the continuous system and dashed lines show the results obtained with the help of the collective variable method of eqs .
( [ denis19 ] ) , ( [ denis22 ] ) , and ( [ denis23 ] ) . in ( a ) the kink moves toward the defect with strength @xmath107 from the gain side . in ( b )
the kink moves from the opposite side and @xmath109 .
the vertical dotted line shows the critical value of the initial kink velocity for this case .
it can be seen that the accuracy of the collective variable method is very high especially for small @xmath22 .
the plots show smaller phase shift for higher kink initial velocity .
this comes from the fact that higher velocity kink is more accelerated by the perturbation considered here .
-symmetric defect as a function of its initial velocity for the kink moving ( a ) toward the gain side of the defect with strength @xmath107 and ( b ) toward the lossy side of the defect with strength @xmath109 .
solid lines show the results of numerical solution for the continuous pde system and dashed lines show the results obtained with the help of the collective variable ode method eqs .
( [ denis19 ] ) , ( [ denis22 ] ) and ( [ denis23 ] ) .
the vertical dotted line in ( b ) shows the threshold kink velocity @xmath110.,width=340 ] here we demonstrate that the kink k@xmath111 trapped at the lossy side of the defect can be pushed through the defect by the second kink k@xmath112 even if the second kink has velocity smaller than the threshold value . to do
so we consider two well separated kinks moving with the same velocity below the threshold value toward the lossy side of the defect .
the first kink is trapped and the second one pushes it , through their well - known mutual repulsion @xcite , through the defect being either reflected back [ see fig .
[ fig5 ] ( a ) for the case of @xmath113 or trapped itself [ as in fig . [ fig5 ]
( b ) for the case of @xmath114 . note that the threshold kink velocity is @xmath115 for @xmath107 , used for this simulation .
is trapped by the defect and then it is pushed through the defect by k@xmath112 , through their mutual repulsion . in ( a ) k@xmath112
is reflected , while in ( b ) it is trapped by the lossy region of the defect .
horizontal dashed lines show the location of the center of the defect . here
( a ) @xmath116 , ( b ) @xmath117 and @xmath107 in both cases .
the threshold kink velocity is @xmath115 .
the kink initial positions are @xmath118 for k@xmath111 and @xmath119 for k@xmath112.,width=340 ] it was found that the result of the breather - defect interaction importantly depends on the initial breather position @xmath25 , because this parameter controls the breather oscillation phase at which it hits the defect . a moving breather in one oscillation
travels the distance @xmath120 .
this means that it is sufficient to consider the range of the initial breather positions from @xmath25 to @xmath121 . in some cases
the breather can split into kink and antikink after passing the defect .
in this situation it is convenient to present the result of the breather - defect interaction by the total energies of the subkinks constituting the breather under the assumption that the subkinks , when merged into a resulting breather after the interaction , share the breather energy equally . in the cases when the breather splits into a kink - antikink pair , the energies of the subkinks are different and they are calculated after they become well separated .
-symmetric defect in the perturbed sge model eq .
( [ sge_collisions ] ) for the breather approaching ( a , b ) from the gain side of the defect and ( c , d ) from the lossy side of the defect . shown
are the total energies of the subkinks constituting the breather under the assumption that they share the breather energy equally .
when the breather splits into a kink - antikink pair , the subkinks have different energies and the lines split into two .
the defect amplitude is @xmath109 in ( a , c ) and @xmath122 in ( b , d ) .
the breather has initial velocity @xmath123 and frequency @xmath124 .
, width=453 ] in fig .
[ fig6 ] we plot the total energy of the subkinks after the breather collides with the defect as a function of its initial position @xmath25 .
horizontal dashed lines show the initial energy of the subkinks . in ( a ,
b ) we show the case when the breather approaches the defect from the gain side and in ( c , d ) it moves toward the defect from the opposite direction . in ( a , c ) the defect amplitude is a=0.1 , while in ( b , d ) a=0.4 .
the breather has frequency @xmath124 and initial velocity of @xmath123 , so that in all cases @xmath125 .
the plots include the whole period of the breather initial position .
one can see that in ( a - c ) there exist the domains of @xmath25 where the breather ( b ) splits into a kink - antikink ( @xmath126 ) pair . in
( d ) the breather does not gain enough energy from the defect to split .
it can be concluded that the breather can split regardless of the direction it approached the defect .
however , the maximal energy gain is larger when the breather moves toward the gain side of the defect .
interestingly also , although there are ( naturally expected ) cases where the splinters bear a lower energy sum than that of the original breather , there are also ones where their sum exceeds the energy of the original breather .
again , this can happen on either side of approach , although it is again more pronounced when approaching the defect from the gain side . to explain the dependence of the breather - defect interaction on the breather phase we note that the kinetic energy of the moving breather is a periodic function of time and space with the sharp maxima at the points where the subkinks collide .
the perturbation term in eq .
( [ sge_collisions ] ) , as it was already mentioned , acts more prominently for large @xmath105 .
thus , the location of the subkink collision points with respect to the maximum and minimum of @xmath7 is very important . in fig .
[ fig7](a ) the function @xmath7 is shown . in fig .
[ fig7](b ) the breather kinetic energy @xmath127 as the function of its spatial coordinate @xmath128 is given for the two cases , @xmath129 ( solid line ) and @xmath130 ( dash line ) , for @xmath131 and @xmath124 , which corresponds to fig .
[ fig6](c ) .
the perturbation strength is @xmath109 . in both cases
the breather moves from the right to the left and approaches the defect from the loss side .
the solid line shows the case when the maxima of the kinetic energy almost do not catch the lossy region of the defect but one of the maxima takes place near the maximum of the gain region . as a result , the breather gains more energy than it loses and it splits into a kink - antikink pair , so that the kinetic energy does not oscillate after the breather passes the defect .
the dashed line shows the case where one maximum of the breather s kinetic energy nearly fits to the maximal loss and the next maximum nearly fits to the maximal gain . in this case
the breather passes through the defect almost unchanged .
hence , clearly the interplay of the kinetic energy oscillation with the spatial distribution of the gain - loss profile is critical in determining the observed breather - defect interaction phenomenology . for large perturbations ( i.e. , stronger defects )
the maxima of the kinetic energy always catch the lossy region of the defect .
consequently , the breather do not gain more energy than it losses and it never splits into a kink - antikink pair for any initial position of the breather .
( see fig .
[ fig6 ] ( d ) ) . as the function of @xmath132 .
( b ) the kinetic energy of the breather as a function of its position .
the breather comes from the lossy ( positive @xmath128 , i.e. , right end ) side of the defect towards the gain ( negative @xmath128 , i.e. , left end ) side with initial positions @xmath129 ( solid line ) and @xmath130 ( dashed line ) [ see fig . [
fig6](c ) ] . in both cases
@xmath131 , @xmath124 and @xmath109.,width=340 ] examples of the breather interaction with the defect are presented in fig .
[ fig8 ] for different initial breather positions . in ( a - c )
the breather approached the defect from the gain side and in ( d - f ) from the loss side .
the breather parameters are @xmath133 and @xmath124 and the perturbation amplitude
is @xmath109 in all cases . in ( a),(d ) and
( e ) the breather breaks up into subkinks . the breaking up takes place only for breathers with sufficiently small frequencies . in ( b ) and
( f ) after the interaction of the defect breather frequency decreases which means that the breather total energy increases . in ( c )
the breather frequency increases ( total energy decreases ) .
this again corroborates the fact that the breather may either lose or gain energy upon its interaction with the defect ( contrary to what we saw , e.g. , in the case of the kink ) . .
in ( a - c ) the breather moves toward the defect from the gain side and in ( d - f ) from the loss side .
initial breather parameters are @xmath124 , @xmath123 and initial positions are ( a ) @xmath134 , ( b ) @xmath135 , ( c ) @xmath136 , ( d ) @xmath137 , ( e ) @xmath138 and ( f ) @xmath139 .
horizontal dashed lines show the position of the defect center and wavy lines represent the breather s two subkinks.,width=377 ] in fig .
[ fig9 ] we present the results obtained for the case when a well separated antikink and kink move toward the defect with the velocity @xmath22 from the lossy side .
the defect strength is @xmath107 and thus , the kink critical velocity is @xmath115 .
we take ( a ) @xmath116 ( b ) @xmath140 ( c ) @xmath141 and ( d ) @xmath142 , so that in the first three cases @xmath143 and the antikink is trapped at the lossy region .
then the kink approaches the antikink and they create a breather .
interestingly , the breather easily enters the gain side of the defect and it is amplified . in ( a ) the breather splits into a kink - antikink pair with one subkink trapped by the loss region and another one passing through the defect . in ( b )
both subkinks pass through the defect . in ( c )
the breather does not split and it moves away from the defect as a single entity ( i.e. , the antikink and kink remain bound ) . in ( d ) @xmath144 and the antikink is not trapped by the defect and both subkinks pass through the defect effectively without interaction with each other . . in all cases , the kink and antikink move toward the defect from the loss side with equal velocity .
horizontal dashed lines show the location of the center of the defect . in ( a - c )
the velocity of the kinks is smaller than the threshold value @xmath115 : ( a ) @xmath116 ( b ) @xmath140 ( c ) @xmath141 , while in ( d ) @xmath142 , i.e. , both have velocities above the threshold value and overcome the defect , effectively without interacting .
the kink initial positions are @xmath118 for @xmath145 and @xmath119 for @xmath146 .
the insets show the details of the dynamics close to the defect center.,width=453 ] note that in fig .
[ fig9](b ) both kink and antikink have @xmath143 nevertheless , they _ both _
pass through the defect .
two reasons can be given to explain this effect and they both are related to the fact that the kink and antikink form a breather to pass through the defect . in sec .
[ sec : breather ] it was shown that the breather can gain energy from the @xmath0-symmetric defect , depending on the phase , and this is the first reason .
the second reason is that the breather translational degree of freedom is only weakly affected by the perturbation considered in this study .
this is demonstrated in fig .
[ fig10 ] where we contrast the dynamics of breathers and kinks with initial velocities equal to 0.1 , 0.2 and 0.3 in the case of homogeneous loss @xmath147 and @xmath148 .
the breather initial frequency is @xmath124 .
it can be seen that the kink trajectories ( smooth lines ) show that kink propagation velocity gradually decreases , while breather trajectories ( wavy lines ) demonstrate almost constant propagation velocities .
hence , this suggests that while breathers are topologically robust , breathers are more efficient in weathering lossy media and in overcoming barriers imposed by dissipative perturbations . this is a feature that is especially useful in the realm of @xmath0-symmetric perturbations / defects . and @xmath148
. the breather initial frequency is @xmath124 .
the kink trajectories ( smooth lines ) show that kink propagation velocity gradually decreases , while breather trajectories ( wavy lines ) demonstrate almost constant propagation velocities , i.e. , minimal impact in the breather translation by the presence of the dissipative perturbation.,width=377 ]
interaction of sg kinks ( and multi - kinks ) as well as breathers with a @xmath0-symmetric defect bearing balanced regions of positive and negative dissipation of energy was investigated analytically ( wherever possible ) and numerically in the present work .
it was demonstrated that a kink coming from the gain side always passes through the defect and restores its initial velocity ( see fig .
[ fig1 ] ) .
the only effect of the interaction with the defect is a phase shift associated with the kink position .
however , for the kink approaching the defect from the opposite side , there exist two different scenarios , depending on the kink initial velocity @xmath22 . for @xmath143 , where @xmath89 is a threshold value of the velocity
, the kink does not have enough energy to pass through the defect and it is trapped by the lossy side of the defect ( see fig . [ fig2 ] ) , while for @xmath144 it is able to overcome the relevant barrier .
if two well - separated kinks approach the defect from the lossy side with the velocities less than @xmath89 , then one of them can pass through the defect while another one will be either trapped by the lossy region or reflected back ( see fig .
[ fig5 ] ) , i.e. , their pairwise repulsion may modify the collisional outcome with the defect .
the breather - defect interaction is more interesting since the breather can split into subkinks depending on its parameters and also on the amplitude of the defect .
depending on the breather initial phase , its total energy can be increased or decreased after the interaction with the defect ( see fig . [ fig6 ] and fig . [ fig8 ] ) .
this can be explained by the fact that the kinetic energy of the moving breather is a periodic function of time and space with sharp maxima at the points where the subkinks collide .
the type of perturbation considered in the present work is more prominent for large @xmath105 .
change in the breather phase changes the location of the subkink collision points with respect to the maxima of the gain and loss regions thus affecting the overall result of the interaction between the breather and the defect ( see fig .
[ fig7 ] ) .
a well - separated kink and antikink pair moving toward the lossy side of the defect with @xmath143 , may enable _ both _ coherent structures to potentially pass through the defect [ see fig .
[ fig9](b ) ] .
this happens because the kink and antikink form a breather that can gain energy from the defect and whose propagation velocity is less affected by the dissipative term than the propagation velocity of the constituent kink or antikink ( see fig .
[ fig10 ] ) .
we conclude that the @xmath0-symmetric defects give new opportunities in the manipulation with the soliton dynamics in the sine - gordon equation and related field theories .
numerous future directions open up as a result of the present considerations .
one such is to consider other klein - gordon field theories in the presence of @xmath0-symmetric defects , such as , e.g. , the @xmath3 model .
the latter is especially interesting due to the presence of internal modes in the kink dynamics which may have a nontrivial impact on the observed phenomenology .
another relevant consideration is that of higher dimensionality .
examining radial kinks as well as breathers in the higher dimensional versions of the sine - gordon model is a theme that has attracted recent interest @xcite , including the formation of breathers as a result of the interaction of the kinks with a radial domain boundary . developing @xmath0-symmetric variants of the 2d sine - gordon and examining the corresponding dynamics is still an open problem .
finally , comparison of the present features with corresponding bright and dark soliton interactions with @xmath0-symmetric defects within the realm of the focusing and defocusing respectively @xmath0-symmetric nonlinear schrdinger equation would also be a theme of relevance to future studies , especially since the latter is the principal field of optical applications of @xmath0-symmetric models .
such studies are presently under consideration and will be reported in future publications .
d.s . thanks the hospitality of the bashkir state pedagogical university and the institute for metals superplasticity problems , ufa , russia .
thanks financial support provided by the government program 5 - 100 - 2020 and by the russian science foundation grant 14 - 13 - 00982 .
d.i.b . was partially supported by a grant of russian foundation for basic research , the grant of the president of russian federation for young scientists - doctors of science ( project no .
md-183.2014.1 ) and by the fellowship of dynasty foundation for young russian mathematicians .
p.g.k . acknowledges support from the us national science foundation under grants cmmi-1000337 , dms-1312856 , from fp7-people under grant irses-606096 from the binational ( us - israel ) science foundation through grant 2010239 , and from the us - afosr under grant fa9550 - 12 - 10332 . | the scattering of kinks and low - frequency breathers of the nonlinear sine - gordon ( sg ) equation on a spatially localized @xmath0-symmetric perturbation ( defect ) with a balanced gain and loss is investigated numerically .
it is demonstrated that if a kink passes the defect , it always restores its initial momentum and energy and the only effect of the interaction with the defect is a phase shift of the kink .
a kink approaching the defect from the gain side always passes , while in the opposite case it must have sufficiently large initial momentum to pass through the defect instead of being trapped in the loss region . the kink phase shift and critical velocity
are calculated with the use of the collective variable method .
kink - kink ( kink - antikink ) collisions at the defect are also briefly considered , showing how their pairwise repulsive ( respectively , attractive ) interaction can modify the collisional outcome of a single kink within the pair with the defect . for the breather
, the result of its interaction with the defect strongly depends on the breather parameters ( velocity , frequency and initial phase ) and on the defect parameters .
the breather can gain some energy from the defect and as a result potentially even split into a kink - antikink pair or it can lose a part of its energy .
interestingly , the breather translational mode is very weakly affected by the dissipative perturbation , so that a breather penetrates more easily through the defect when it comes from the lossy side , than a kink . in all studied soliton - defect interactions the energy loss to radiation of small - amplitude extended waves is negligible . |
understanding sea quark effects in the light hadron spectrum is an important issue , sharpened by the recent finding of a systematic deviation of the quenched spectrum from experiment@xcite . to this end , we have been pursuing @xmath0 qcd simulations using an rg - improved gauge action and a tadpole - improved clover quark action @xcite , to be called * rc * simulations in this article .
the parameters of these simulations are listed in table [ tab : param ] .
the statistics at @xmath3 have been increased since lattice98 , and the runs at @xmath4 are new .
in addition we have carried out quenched simulations with the same improved action , referred to as * qrc * , for a direct comparison of the full and quenched spectrum .
the @xmath5 values of these runs , given in table [ tab : param ] , are chosen so that the lattice spacing fixed by the string tension matches that of full qcd for each value of sea quark mass at @xmath6 and 2.1 .
quenched hadron masses are calculated for valence quark masses such that @xmath7 0.80.5 , which is similar to those in the * rc * runs . in this report we present updated results of the full qcd spectrum and light quark masses .
we also discuss sea quark effects by comparing the * rc * and * qrc * results . for reference
we use quenched results with the plaquette gauge and wilson quark action @xcite as well , which we denote as * qpw*. [ tab : param ] lllll + lattice & @xmath8 & # traj . &
@xmath9 & @xmath10 [ fm ] + @xmath11 & 0.1409 & 6250 & 0.806(1 ) & 0.289(3 ) + @xmath12 & 0.1430 & 5000 & 0.753(1 ) & 0.152(2 ) + @xmath13 & 0.1445 & 7000 & 0.696(2 ) & 0.269(3 ) + @xmath14 fm & 0.1464 & 5250 & 0.548(4 ) & 0.248(2 ) + @xmath15 & 0.1375 & 7000 & 0.805(1 ) & 0.204(1 ) + @xmath16 & 0.1390 & 7000 & 0.751(1 ) & 0.193(2 ) + @xmath17 & 0.1400 & 7000 & 0.688(1 ) & 0.181(1 ) + @xmath18 fm & 0.1410 & 7000 & 0.586(3 ) & 0.170(1 ) + @xmath19 & 0.1357 & 2000 & 0.806(2 ) & 0.1342(6 ) + @xmath20 & 0.1367 & 2000 & 0.757(2 ) & 0.1259(5 ) + @xmath21 & 0.1374 & 2000 & 0.690(3 ) & 0.1201(5 ) + @xmath22 fm & 0.1382 & 2000 & 0.575(6 ) & 0.1128(3 ) + @xmath19 & 0.1351 & 2000 & 0.800(2 ) & 0.1049(2 ) + @xmath23 & 0.1358 & 2000 & 0.754(2 ) & 0.1012(3 ) + @xmath24 & 0.1363 & 2000 & 0.704(3 ) & 0.0977(3 ) + @xmath25 fm & 0.1368 & 2000 & 0.629(5 ) & 0.0947(2 ) + lllll + & & + @xmath5 & @xmath10 [ fm ] & & @xmath5 & @xmath10 [ fm ] + 2.187 & 0.2079(15 ) & & 2.416 & 0.1359(7 ) + 2.214 & 0.1977(13 ) & & 2.456 & 0.1266(13 ) + 2.247 & 0.1853(9 ) & & 2.487 & 0.1206(9 ) + 2.281 & 0.1727(10 ) & & 2.528 & 0.1130(9 ) + 2.334 & 0.1577(9 ) & & 2.575 & 0.1065(7 ) +
the analysis procedure of our full qcd spectrum data follows that in ref .
@xcite : @xmath26 and @xmath27 are used to set the scale and determine the up and down quark mass @xmath28 , while the strange quark mass @xmath29 is fixed from either @xmath30 or @xmath31 .
we tested several fitting forms for the continuum extrapolation , and found that the fit is stable ; e.g. , for the meson masses , linear extrapolations in @xmath32 and in @xmath33 are consistent with each other and a quadratic fit in @xmath32 is also consistent within 2 standard deviations . here , we present results from the linear extrapolation in @xmath32 .
= 7.5 cm = 7.5 cm fig .
[ fig : spectrum ] shows an update of results for vector meson and octet baryon masses in comparison to those from the * qpw * simulation . with increased statistics at @xmath34 and new points at @xmath35
, we find our conclusion to remain unchanged since lattice98 , _
i.e. , _ meson masses in full qcd extrapolate significantly closer to experiment than in quenched qcd . for baryons ,
the statistical errors are still too large to draw definitive conclusions .
in order to obtain a deeper understanding of the sea quark effect in meson masses , we investigate how their values depend on the sea quark mass . in this test
, the valence strange quark mass is fixed by a phenomenological value of the ratio @xmath36 . to avoid uncertainties that may arise from chiral extrapolations ,
the light dynamical quark mass is set to one of the values corresponding to @xmath37 or 0.5 .
the values of the masses `` @xmath38 '' and `` @xmath27 '' of fictitious mesons for such quark masses can then be determined by interpolations or short extrapolations of hadron mass results . in fig .
[ fig : massratio ] , we plot `` @xmath39 '' as a function of the lattice spacing normalized by `` @xmath27 '' for different sea quark masses . making linear extrapolations in @xmath32 , we observe that the continuum limits of the two quenched simulations * qrc * and * qpw * are consistent . on the other hand , the full qcd result from * rc * exhibits an increasingly clearer deviation from the quenched value toward lighter sea quark masses .
we consider that this result provides a clear demonstration of the sea quark effect on vector meson masses .
= 7.5 cm = 7.5 cm = 7.5 cm we plot our results for light quark masses in the @xmath40 scheme at @xmath412 gev in fig .
[ fig : mq ] , together with the quenched results of ref .
continuum extrapolations are made linearly in @xmath32 with the constraint that the three definitions ( using axial vector ward identity(awi ) or vector ward identity(vwi ) with either @xmath42 from sea quarks or partially quenched @xmath42 ) yield the same value .
we confirm our previous finding@xcite that i ) quark masses in full qcd are much smaller than those in quenched qcd , and ii ) the large discrepancy in the strange quark mass determined from @xmath30 or @xmath31 , observed in quenched qcd , is much reduced .
our current estimate for quark masses in @xmath43 qcd are @xmath44 mev , @xmath45 mev ( @xmath46-input ) and @xmath47 mev ( @xmath48-input ) .
the quoted errors include our estimate of the systematic errors due to the choice of functional form of continuum extrapolations and the definition of the @xmath40 coupling used in the one - loop tadpole improved renormalization factor .
our results for quark masses are smaller than the values often used in phenomenology@xcite , though the ratio @xmath49 26(3 ) is consistent with the result 24.4(1.5)@xcite from chiral perturbation theory .
the small values are quite interesting , especially for the strange quark mass ; a smaller strange quark mass raises the prediction of the standard model for the direct cp violation parameter @xmath50 , as strongly favored by the experimental results from the ktev @xcite and na31 collaborations @xcite .
this work is supported in part by grants - in - aid of the ministry of education ( nos . 09304029 , 10640246 , 10640248 , 11640250 , 11640294 , 10740107 , 11740162 ) .
se and kn are jsps research fellows .
aak , tm and hps are supported by the research for the future program of jsps , and hps also by the leverhulme foundation . | we present updated results of the cp - pacs calculation of the light hadron spectrum in @xmath0 full qcd .
simulations are made with an rg - improved gauge action and a tadpole - improved clover quark action for sea quark masses corresponding to @xmath10.6 and the lattice spacing @xmath20.09 fm .
a comparison of the @xmath0 qcd spectrum with new quenched results , obtained with the same improved action , shows clearly the existence of sea quark effects in vector meson masses .
results for light quark masses are also presented . |
symmetries are ubiquitous in discrete mathematics and geometry .
they are often responsible for the tractability of algorithmic problems and for the beauty of both the investigated structures and the developed methods .
it is common knowledge , however , that the presence of symmetries in integer programs may severely harm the ability to solve them .
the reasons for this are twofold .
first , the use of branch - and - bound methods usually leads to an unnecessarily large search tree , because equivalent solutions are found again and again .
second , the quality of lp relaxations of such programs typically is extremely poor .
a classical approach to `` break '' such symmetries is to add constraints that cut off equivalent copies of solutions , in hope to resolve these problems .
there are numerous examples of this in the literature ; we will give a few references for the special case of graph coloring below .
another approach was developed by margot @xcite .
he studies a branch - and - cut method that ensures to investigate only one representative of each class of equivalent solutions by employing methods from computational group theory .
furthermore , the symmetries are also used to devise cutting planes .
methods for symmetry breaking in the context of constraint programming have been developed , for instance , by fahle , schamberger , and sellmann @xcite and puget @xcite .
the main goal of this paper is to start an investigation of the polytopes that are associated with certain symmetry breaking inequalities . in order to clarify the background , we first discuss the example of a well - known integer programming ( ip ) formulation for the graph coloring problem .
let @xmath2 be a loopless undirected graph without isolated nodes .
a _ ( vertex ) coloring _ of @xmath3 using at most @xmath4 colors is an assignment of colors @xmath5 to the nodes such that no two adjacent nodes receive the same color .
the _ graph coloring _ problem is to find a vertex coloring with as few colors as possible .
this is one of the classical np - hard problems @xcite .
it is widely believed to be among the hardest problems in combinatorial optimization . in the following classical ip formulation ,
@xmath6 are the nodes of @xmath3 and @xmath4 is some upper bound on the number of colors needed .
@xmath7 & x_{ij } + x_{kj } & \leq y_j & \{i , k\ } \in e,\ ; j \in \{1,\dots , c\ } & \text{(i)}\\ & { \displaystyle \sum}_{j=1}^{c } x_{ij } & = 1 & i \in v & \text{(ii)}\\ & \multicolumn{2}{r}{x_{ij } \in \{0,1\ } } & i \in v,\ ; j \in \{1,\dots , c\ } & \text{(iii)}\\ & \multicolumn{2}{r}{y_j \in \{0,1\ } } & j \in \{1 , \dots , c\ } & \text{(iv ) } \end{array}\ ] ] in this model , variable @xmath8 is @xmath1 if and only if color @xmath9 is assigned to node @xmath10 and variable @xmath11 is @xmath1 if color @xmath9 is used
. constraints ( i ) ensure that color @xmath9 is assigned to at most one of the two adjacent nodes @xmath10 and @xmath12 ; it also enforces that @xmath11 is @xmath1 if color @xmath9 is used , because there are no isolated nodes .
constraints ( ii ) guarantee that each node receives exactly one color .
it is well known that this formulation exhibits symmetry : given a solution @xmath13 , any permutation of the colors , i.e. , the columns of @xmath14 ( viewed as an @xmath15-matrix ) and the components of @xmath16 , results in a valid solution with the same objective function value . viewed abstractly , the symmetric group of order @xmath4 acts on the solutions @xmath17 ( by permuting the columns of @xmath14 and the components of @xmath16 ) in such a way that the objective function is constant along every orbit of the group action .
each orbit corresponds to a symmetry class of feasible colorings of the graph . note that `` symmetry '' here always refers to the symmetry of permuting colors , not to symmetries of the graph .
the weakness of the lp - bound mentioned above is due to the fact that the point @xmath18 with @xmath19 and @xmath20 is feasible for the lp relaxation with objective function value @xmath21 .
the symmetry is responsible for the feasibility of @xmath18 , since @xmath22 is the barycenter of the orbit of an arbitrary @xmath23 satisfying ( ii ) in .
it turned out that the symmetries make the above ip - formulation for the graph coloring problem difficult to solve .
one solution is to develop different formulations for the graph coloring problem .
this line has been pursued , e.g. , by mehrotra and trick @xcite , who devised a column generation approach .
see figueiredo , barbosa , maculan , and de souza @xcite and cornaz @xcite for alternative models . another solution is to enhance the ip - model by additional inequalities that cut off as large parts of the orbits as possible , keeping at least one element of each orbit in the feasible region .
mndez - daz and zabala @xcite showed that a branch - and - cut algorithm using this kind of symmetry breaking inequalities performs well in practice .
the polytope corresponding to was investigated by camplo , corra , and frota @xcite and coll , marenco , mndez - daz , and zabala @xcite .
ramani , aloul , markov , and sakallah @xcite studied symmetry breaking in connection with sat - solving techniques to solve the graph coloring problem . the strongest symmetry breaking constraints that mndez - daz and zabala @xcite introduced are the inequalities @xmath24 from each orbit , they cut off all points except for one representative that is the maximal point in the orbit with respect to a lexicographic ordering .
a solution @xmath17 of the above ip - model is such a representative if and only if the columns of @xmath14 are in decreasing lexicographic order .
we introduce a generalization and strengthening of inequalities in section [ sec : ipformulations ] .
breaking symmetries by adding inequalities like does not depend on the special structure of the graph coloring problem .
these inequalities single out the lexicographic maximal representative from each orbit ( with respect to the symmetric group acting on the columns ) of the whole set of all 0/1-matrices with exactly one @xmath1-entry per row .
the goal of this paper is to investigate the structure of general `` symmetry breaking polytopes '' like the convex hull of these representatives .
we call these polytopes _ orbitopes_. the idea is that general knowledge on orbitopes ( i.e. , valid inequalities ) can be utilized for different symmetric ips in order to address both the difficulties arising from the many equivalent solutions and from the poor lp - bounds . in particular with respect to the second goal , for concrete applications it will be desirable to combine the general knowledge on orbitopes with concrete polyhedral knowledge on the problem under investigation in oder to derive strengthened inequalities . for the example of graph coloring , we indicate that ( and how ) this can be done in section [ sec : closingremarks ] . figure [ fig : genericillustration ] illustrates the geometric situation . the case of a symmetric group acting on the columns
is quite important .
it does not only appear in ip - formulations for the graph coloring problem , but also in many other contexts like , e.g. , block partitioning of matrices @xcite , @xmath12-partitioning in the context of frequency assignment @xcite , or line - planning in public transport @xcite .
however , other groups are interesting as well . for instance , in the context of timetabling in public transport systems @xcite , cyclic groups play an important role .
we thus propose to study different types of orbitopes , depending on the group acting on the columns of the variable - matrix and on further restrictions like the number of @xmath1-entries per row being exactly one ( _ partitioning _ ) , at most one ( _ packing _ ) , at least one ( _ covering _ ) , or arbitrary ( _ full _ ) . the main results of this paper are complete and irredundant linear descriptions of packing and partitioning orbitopes for both the symmetric group and for the cyclic group acting on the columns of the variable - matrix .
we also provide ( linear time ) separation algorithms for the corresponding sets of inequalities .
while this work lays the theoretical foundations on orbitopes , a thorough computational investigation of the practical usefulness of the results will be the subject of further studies ( see also the remarks in section [ sec : closingremarks ] ) .
the outline of the paper is as follows . in section [ sec : basicorbitopes ] , we introduce some basic notations and define orbitopes . in section [ sec : optimizingoverorbitopes ] we show that optimization over packing and partitioning orbitopes for symmetric and cyclic groups can be done in polynomial time . in section [ sec : packingandpartitioningorbitopesforcyclicgroups ] we give complete ( totally unimodular ) linear descriptions of packing and partitioning orbitopes for cyclic groups .
section [ packingandpartitioningorbitopesforsymmetricgroups ] deals with packing and partitioning orbitopes for symmetric groups , which turn out to be more complicated than their counterparts for cyclic groups . here , besides ( strengthenings of ) inequalities
, one needs exponentially many additional inequalities , the `` shifted column inequalities '' , which are introduced in section [ sec : shiftedcolumninequalities ] .
we show that the corresponding separation problem can be solved in linear time , see section [ sec : separationalgorithmforscis ] .
section [ sec : completeinequalitydescriptions ] gives a complete linear description , and section [ sec : facets ] investigates the facets of the polytopes .
we summarize the results for symmetric groups in section [ summaryofresultssymmetricgroup ] for easier reference .
finally , we close with some remarks in section [ sec : closingremarks ] .
we first introduce some basic notation . for a positive integer @xmath25 ,
we define @xmath26 } : = \{1,2,\dots , n\}$ ] . we denote by @xmath27 the 0-matrix or 0-vector of appropriate sizes . throughout the paper let @xmath28 and @xmath29 be positive integers . for @xmath30 } \times { [ { q}]}}$ ] and @xmath31 } \times { [ { q}]}$ ] , we write @xmath32 for convenience , we use @xmath33 for @xmath34 and @xmath35 for @xmath36 , where @xmath37 } \times { [ { q}]}$ ] and @xmath38 } \times { [ { q}]}$ ] . if @xmath28 and @xmath29 are clear from the context , then @xmath39 are the entries of the @xmath10th row . let @xmath40 } \times { [ { q}]}}$ ] be the set of @xmath0-matrices of size @xmath41 .
we define @xmath42 @xmath43 @xmath44 .
let @xmath45 be the lexicographic ordering of @xmath46 with respect to the ordering @xmath47 of matrix positions , i.e. , @xmath48 with @xmath49 if and only if @xmath50 , where @xmath51 is the first position ( with respect to the ordering above ) where @xmath52 and @xmath53 differ .
let @xmath54 be the group of all permutations of @xmath26}$ ] ( _ symmetric group _ ) and let @xmath3 be a subgroup of @xmath55 , acting on @xmath46 by permuting columns .
let @xmath56 be the set of matrices of @xmath46 that are @xmath45-maximal within their orbits under the group action @xmath3 .
we can now define the basic objects of this paper .
the _ full orbitope _ associated with the group @xmath3 is @xmath57 we associate with the group @xmath3 the following restricted orbitopes : @xmath58 by definition , @xmath59 is a face of both @xmath60 and @xmath61 . in this paper
, we will be only concerned with the cases of @xmath3 being the _ cyclic group _
@xmath62 containing all @xmath29 cyclic permutations of @xmath63}$ ] ( section [ sec : packingandpartitioningorbitopesforcyclicgroups ] ) or the symmetric group @xmath55 ( section [ packingandpartitioningorbitopesforsymmetricgroups ] ) .
furthermore , we will restrict attention to packing and partitioning orbitopes .
for these , we have the following convenient characterizations of vertices : [ obs : charvert ] a matrix of @xmath46 is contained in @xmath64 if and only if its columns are in non - increasing lexicographic order ( with respect to the order @xmath45 defined above ) .
[ obs : charvert : sym ] a matrix of @xmath65 is contained in @xmath66 if and only if its first column is lexicographically not smaller than the remaining ones ( with respect to the order @xmath45 ) .
[ obs : charvert : cycpack ] in particular , a matrix of @xmath67 is contained in @xmath66 if and only if it has a @xmath1-entry at position @xmath68 .
[ obs : charvert : cycpart ] the main aim of this paper is to provide complete descriptions of @xmath69 , @xmath70 , @xmath71 , and @xmath72 by systems of linear equations and linear inequalities .
if these orbitopes admit `` useful '' linear descriptions then the corresponding linear optimization problems should be solvable efficiently , due to the equivalence of optimization and separation , see grtschel , lovsz , and schrijver @xcite .
we start with the cyclic group operation , since the optimization problem is particularly easy in this case .
[ thm : cyclpartopt ] both the linear optimization problem over @xmath73 and over @xmath74 can be solved in time @xmath75 .
we first give the proof for the packing case .
for a vector @xmath76}\times{[{q}]}}$ ] , we consider the linear objective function @xmath77 the goal is to find a matrix @xmath78 such that @xmath79 is maximal .
let @xmath80 be such a @xmath81-maximal matrix , and let @xmath82 be its first column .
if @xmath83 , then @xmath84 by part of observation [ obs : charvert ] . by the same observation it follows that if @xmath85 and @xmath86}$ ] is the minimum row - index @xmath10 with @xmath87 , then @xmath80 has only zero entries in its first @xmath88 rows , except for the @xmath1-entry at position @xmath89 ( there is at most one @xmath1-entry in each row ) .
furthermore , each row @xmath90 of @xmath80 either has no @xmath1-entry or it has its ( unique ) @xmath1-entry at some position where @xmath81 is maximal in row @xmath10 .
thus , we can compute an optimal solution as follows : ( 1 ) for each @xmath91}$ ] determine a vector @xmath92 that is the zero vector if @xmath81 does not have any positive entries in row @xmath10 and otherwise is the @xmath9-th standard unit vector , where @xmath93}$ ] is chosen such that @xmath94}}\}}$ ] ; set @xmath95 in the first case and @xmath96 in the second .
( 2 ) compute the values @xmath97 and @xmath98 for all @xmath99 .
( 3 ) determine @xmath88 such that @xmath100 is maximal among @xmath101}}\}}$ ] . ( 4 )
if @xmath102 , then @xmath27 is an optimal solution .
otherwise , the matrix whose @xmath10-th row equals @xmath103 for @xmath104 and which is all - zero in the first @xmath88 rows , except for a @xmath1-entry at position @xmath89 , is optimal . from the description of the algorithm it is easy to see that its running time is bounded by @xmath75 ( in the unit - cost model ) .
the partitioning case is then straightforward and even becomes easier due to part of observation [ obs : charvert ] .
[ thm : sympartopt ] both the linear optimization problem over @xmath105 and over @xmath106 can be solved in time @xmath107 .
we give the proof for the partitioning case , indicating the necessary modifications for the packing case at the relevant points . as in the proof of theorem [ thm : cyclpartopt ] , we maximize the linear objective function given by @xmath108 for @xmath109}\times{[{q}]}}$ ] .
we describe a two - step approach . in the first step , for @xmath110}$ ] with @xmath111 and @xmath112}$ ] , we let @xmath113 be @xmath81-maximal among the matrices in @xmath114}}$ ] with exactly ( in the packing case : at most ) one @xmath1-entry in every row .
denote by @xmath115 the @xmath81-value of @xmath113 , i.e. , @xmath116 the values @xmath115 can be computed in time @xmath107 as follows .
first , we compute all numbers @xmath117}}\}}$ ] ( in the packing case : @xmath118}}\}})$ ] ) for all @xmath119}$ ] and @xmath93}$ ] .
this can clearly be done in @xmath75 steps by using the recursions @xmath120 for @xmath121 .
then , after initializing @xmath122 for all @xmath119}$ ] and @xmath112}$ ] , one computes @xmath123 for all @xmath93}$ ] , @xmath124 , and @xmath125 ; see figure [ fig : sympartopt ] .
in the second step , for @xmath119}$ ] and @xmath93}$ ] , let @xmath126 be @xmath81-maximal among the matrices in @xmath127}}$ ] with exactly ( in the packing case : at most ) one @xmath1-entry in every row and with columns @xmath128 being in non - increasing lexicographic order .
thus , by part of observation [ obs : charvert ] , @xmath129 is an optimal solution to our linear optimization problem .
denote by @xmath130 the @xmath81-value of @xmath126 , i.e. , @xmath131 .
_ left : _ computation of @xmath115 .
_ right : _ computation of @xmath130 via the dynamic programming relation
. indicated are the matrix @xmath132 and corresponding term @xmath133 and matrix @xmath134 with corresponding term @xmath135.,title="fig : " ] . _ left : _ computation of @xmath115 .
_ right : _ computation of @xmath130 via the dynamic programming relation . indicated are the matrix @xmath132 and corresponding term @xmath133 and matrix @xmath134 with corresponding term @xmath135.,title="fig : " ] let @xmath136 be the index of the first row , where @xmath126 has a @xmath1-entry in column @xmath9 ( with @xmath137 if there is no such @xmath1-entry ) ; see figure [ fig : sympartopt ] .
then @xmath126 has a @xmath81-maximal matrix @xmath138 in rows @xmath139 with exactly ( in the packing case : at most ) one @xmath1-entry per row and lexicographically sorted columns @xmath140 ( contributing @xmath135 ) . in row
@xmath12 , there is a single @xmath1-entry at position @xmath141 ( contributing @xmath142 ) . and
in rows @xmath143 , we have a @xmath81-maximal matrix @xmath144 with exactly ( in the packing case : at most ) one @xmath1-entry per row in the first @xmath145 columns ( contributing @xmath133 ) and zeroes in the remaining columns .
therefore , we obtain @xmath146 hence , considering all possibilities for @xmath12 , we have @xmath147 for all @xmath119}$ ] and @xmath93}$ ] . for convenience
we define @xmath148 for @xmath149}$ ] with @xmath150 and @xmath151 for all @xmath152}$ ] and @xmath153 .
furthermore , we set @xmath154 for all @xmath155}$ ] .
finally , we define @xmath156 for all @xmath157}$ ] and @xmath158}$ ] .
thus , by dynamic programming , we can compute the table @xmath130 via equation in the order @xmath159 , @xmath160 .
for each pair @xmath161 the evaluation of requires no more than @xmath162 steps , yielding a total running time bound of @xmath107 .
furthermore , if during these computations for each @xmath161 we store a maximizer @xmath163 for @xmath12 in , then we can easily reconstruct the optimal solution @xmath129 from the @xmath12-table without increasing the running time asymptotically : for @xmath119}$ ] , @xmath164}$ ] the matrix @xmath126 is composed of @xmath165 ( if @xmath166 and @xmath167 ) , @xmath168 ( if @xmath169 and @xmath170 ) , and having @xmath171-entries everywhere else , except for a @xmath1-entry at position @xmath172 ( if @xmath173 ) . each single matrix @xmath113 can be computed in @xmath174 steps .
furthermore , for the matrices @xmath113 needed during the recursive reconstruction of @xmath129 , the sets @xmath175}$ ] are pairwise disjoint ( see figure [ fig : sympartopt ] ) .
thus , these matrices all together can be computed in time @xmath75 . at the end
there might be a single @xmath176 to be constructed , which trivially can be done in @xmath75 steps .
thus , with respect to complexity theory there are no `` obstructions '' to finding complete linear descriptions of packing and partitioning orbitopes for both the cyclic and the symmetric group action .
in fact , for cyclic group actions we will provide such a description in theorem [ thm : orbipartcycl ] and theorem [ thm : orbipackcycl ] for the partitioning and packing case , respectively . for symmetric group actions we will provide such a description for partitioning orbitopes in theorems [ thm
: resultspart ] and for packing orbitopes in theorem [ thm : resultspack ] .
the algorithm used in the proof of theorem [ thm : cyclpartopt ] ( for cyclic groups ) is trivial , while the one described in the proof of theorem [ thm : sympartopt ] ( for symmetric groups ) is a bit more complicated .
this is due to the simpler characterization of the cyclic case in observation [ obs : charvert ] and is reflected by the fact that the proofs of theorems [ thm : resultspart ] and [ thm : resultspack ] ( for symmetric groups ) need much more work than the ones of theorems [ thm : orbipartcycl ] and [ thm : orbipackcycl ] ( for cyclic groups ) .
the algorithms described in the above two proofs heavily rely on the fact that we are considering only matrices with at most one @xmath1-entry per row . for cyclic group operations ,
the case of matrices with more ones per row becomes more involved , because we do not have a simple characterization ( like the one given in parts [ obs : charvert : cycpack ] and [ obs : charvert : cycpart ] of observation [ obs : charvert ] ) of the matrices in @xmath66 anymore .
for the action of the symmetric group , though we still have the characterization provided by part of observation [ obs : charvert ] , the dynamic programming approach used in the proof of theorem [ thm : sympartopt ] can not be adapted straight - forwardly without resulting in an exponentially large dynamic programming table ( unless @xmath29 is fixed ) .
these difficulties apparently are reflected in the structures of the corresponding orbitopes ( see the remarks in section [ sec : closingremarks ] ) .
from the characterization of the vertices in parts and of observation [ obs : charvert ] one can easily derive ip - formulations of both the partitioning orbitope @xmath71 and the packing orbitope @xmath72 for the cyclic group @xmath62 .
in fact , it turns out that these formulations do already provide linear descriptions of the two polytopes , i.e. , they are totally unimodular .
we refer the reader to schrijver ( * ? ? ?
19 ) for more information on total unimodularity .
it is easy to see that for the descriptions given in theorems [ thm : orbipartcycl ] and [ thm : orbipackcycl ] below , the separation problem can be solved in time @xmath75 .
[ thm : orbipartcycl ] the partitioning orbitope @xmath71 for the cyclic group @xmath62 equals the set of all @xmath30 } \times { [ { q}]}}$ ] that satisfy the following linear constraints : the equations @xmath177 and @xmath178 for all @xmath179 , the nonnegativity constraints @xmath180 for all @xmath181 and @xmath93}$ ] , the row - sum equations @xmath182 for all @xmath181 .
this system of constraints is non - redundant .
the constraints @xmath182 for @xmath119}$ ] and @xmath183 for @xmath184},j\in{[{q}]}$ ] define an integral polyhedron , since they describe a transshipment problem ( and thus , the coefficient matrix is totally unimodular ) .
hence , the constraint system given in the statement of the theorem describes an integer polyhedron , because it defines a face of the corresponding transshipment polytope . by part of observation
[ obs : charvert ] , the set of integer points satisfying this constraint system is @xmath185 . hence the given constraints completely describe @xmath71 .
the non - redundancy follows from the fact that dropping any of the constraints enlarges the set of feasible integer solutions . ''
stands for a @xmath186 , `` @xmath187 '' for a @xmath188.,title="fig:"][sfl : packcyclcoline ] [ thm : orbipackcycl ] the packing orbitope @xmath72 for the cyclic group @xmath62 equals the set of all @xmath30 } \times { [ { q}]}}$ ] that satisfy the following linear constraints : the constraints @xmath189 and @xmath178 for all @xmath190 , the nonnegativity constraints @xmath180 for all @xmath191 and @xmath93}$ ] , the row - sum inequalities @xmath192 for all @xmath193 , the inequalities @xmath194 for all @xmath181 ( see figure [ fig : orbipackcycl ] for an example ) .
this system of constraints is non - redundant .
.,title="fig:"][sfl : networkmatrix ] from part of observation [ obs : charvert ] it follows that an integer point is contained in @xmath72 if and only if it satisfies the constraints described in the statement , where inequalities ensure that the first column of @xmath14 is lexicographically not smaller than the other ones ( note that we have at most one @xmath1-entry in each row of @xmath14 ) . dropping any of the constraints
enlarges the set of integer solutions , which proves the statement on non - redundancy .
thus , as in the proof of the previous theorem , it remains to show that the polyhedron defined by the constraints is integral .
we prove this by showing that the coefficient matrix @xmath52 of the row - sum inequalities @xmath192 ( for @xmath195 ) and inequalities ( for all @xmath195 ) is a network matrix ( and thus , totally unimodular ) . adding the nonnegativity constraints
amounts to adding an identity matrix and preserves total unimodularity , which also holds for the inclusion of @xmath196 into the system . in order to establish the claim on the network structure of @xmath52
, we will identify a directed tree @xmath138 , whose arcs are in bijection with @xmath197 } \times { [ { q}]}$ ] ( the set of indices of the columns of @xmath52 ) , such that there are pairs of nodes @xmath198 of @xmath138 in bijection with the row indices @xmath199}$ ] of @xmath52 with the following property .
the matrix @xmath52 has a @xmath200-entry in row @xmath201 and column @xmath161 , if the unique path @xmath202 from node @xmath203 to node @xmath204 in the tree @xmath138 uses arc @xmath161 in its direction from @xmath10 to @xmath9 , a @xmath205-entry , if @xmath202 uses @xmath161 in its reverse direction , and a @xmath171-entry , if @xmath202 does not use @xmath161 . for the construction of the tree @xmath138 , we take a directed path @xmath206 of length @xmath28 on nodes @xmath207 with arcs @xmath208 for @xmath119}$ ] ; see figure [ fig : networkmatrix ] . for each @xmath181
, we append a directed path @xmath209 of length @xmath210 to node @xmath211 , where @xmath209 has node set @xmath212 and arcs @xmath213 for @xmath179 . choosing the pair @xmath214 for the @xmath10-th row sum - inequality and the pair @xmath215 for the @xmath10-th inequality , finishes the proof ( using the bijection between the arcs of @xmath138 and the columns of @xmath52 indicated by the notation @xmath216 ) .
for packing orbitopes @xmath217 and partitioning orbitopes @xmath69 with respect to the symmetric group it follows readily from the characterization in part of observation [ obs : charvert ] that the equations @xmath218 are valid .
thus , we may drop all variables corresponding to components in the upper right triangle from the formulation and consider @xmath219 } \times { [ { q}]}}\,:\,{i \geq j}\}}.\ ] ] we also adjust the definition of @xmath220}\ ] ] and define the @xmath9th column for @xmath93}$ ] as @xmath221 furthermore , we restrict ourselves to the case @xmath222 in this context . because of , the case of @xmath223 can be reduced to the case @xmath224 and the case of @xmath225 is of no interest .
the next result shows a very close relationship between packing and partitioning orbitopes for the case of symmetric group actions .
[ prop : projpartpack ] the polytopes @xmath69 and @xmath226 are affinely isomorphic via orthogonal projection of @xmath69 onto the space @xmath227}}\}}\ ] ] @xmath228and the canonical identification of this space with @xmath229 .
the affine subspace @xmath230 of @xmath231 clearly contains @xmath69 .
let @xmath232 be the orthogonal projection mentioned in the statement ( identifying @xmath233 in the canonical way with @xmath234 ) ; note that the first row is removed since it only contains the element @xmath68 .
consider the linear map @xmath235 defined by @xmath236 ( where @xmath237 and @xmath238 ) .
this is the inverse of @xmath239 , showing that @xmath239 is an affine isomorphism .
as we have @xmath240 , this finishes the proof .
it will be convenient to address the elements in @xmath241 via a different `` system of coordinates '' : @xmath242},\ ; 1 \leq \eta \leq p - j+1.\ ] ] thus ( as before ) @xmath10 and @xmath9 denote the row and the columns , respectively , while @xmath243 is the index of the diagonal ( counted from above ) containing the respective element ; see figure [ fig : scispart ] for an example . for @xmath244 and @xmath245 , we write @xmath246 . for @xmath247
we denote by @xmath248 the set of all coordinates ( positions in the matrix ) , where @xmath14 has a @xmath1-entry .
conversely , for @xmath249 , we use @xmath250 for the 0/1-point with @xmath251 if and only if @xmath252 . for @xmath253
, we define the _ column _ @xmath254 and for @xmath255 we write @xmath256 . of course
, we have @xmath257 .
the rest of this section is organized as follows .
first , in section [ sec : ipformulations ] , we deal with basic facts about integer points in packing and partitioning orbitopes for the symmetric group . to derive a linear description of @xmath217 and @xmath69 that only contains integer vertices ,
we need additional inequalities , the _ shifted column inequalities _ , which are introduced in section [ sec : shiftedcolumninequalities ] .
we then show that the corresponding separation problem can be solved in linear time ( section [ sec : separationalgorithmforscis ] ) .
section [ sec : completeinequalitydescriptions ] proves the completeness of the linear description and section [ sec : facets ] investigates the facets of the polytopes .
we first derive a crucial property of the vertices of @xmath217 .
[ lem : xcapcol ] let @xmath14 be a vertex of @xmath217 with @xmath258 @xmath259 .
then we have @xmath260 .
with @xmath261 we have @xmath262 , which implies @xmath263 ( since @xmath14 has at most one @xmath1-entry in row @xmath10 ) .
thus , @xmath264 would yield @xmath265 for all @xmath266 , contradicting the lexicographic order of the columns of @xmath14 ( see part of observation [ obs : charvert ] ) . for @xmath267 and the set @xmath268 , @xmath269 , we call @xmath270 a _ column inequality _
; see figure [ fig : scispart ] for an example with @xmath271 .
the column inequalities are strengthenings of the symmetry breaking inequalities @xmath272 introduced by mndez - daz and zabala @xcite in the context of vertex - coloring ( see in the introduction ) .
[ prop : ipsym ] a point @xmath247 is contained in @xmath217 @xmath273 if and only if @xmath14 satisfies the row - sum constraints @xmath274 @xmath275 for all @xmath184}$ ] and all column inequalities . by lemma [ lem : xcapcol ] ,
inequalities are valid for @xmath217 ( and thus , for its face @xmath69 as well ) . because of the row - sum constraints , all column inequalities are valid as well .
therefore , it suffices to show that a point @xmath276 that satisfies the row - sum constraints @xmath277 and all column inequalities is contained in @xmath64 .
suppose , this was not the case .
then , by part of observation [ obs : charvert ] , there must be some @xmath93}$ ] such that the @xmath278-st column of @xmath14 is lexicographically smaller than the @xmath9th column .
let @xmath10 be minimal with @xmath262 ( note that column @xmath9 can not be all - zero ) .
thus , @xmath265 for all @xmath279 .
this implies @xmath280 , showing that the column inequality @xmath281 is violated by the point @xmath14 for the bar @xmath282 .
proposition [ prop : ipsym ] provides a characterization of the vertices of the packing- and partitioning orbitopes for symmetric groups among the integer points .
different from the situation for cyclic groups ( see theorems [ thm : orbipartcycl ] and [ thm : orbipackcycl ] ) , however , the inequalities in this characterization do not yield complete descriptions of these orbitopes .
in fact , we need to generalize the concept of a column inequality in order to arrive at complete descriptions .
this will yield exponentially many additional facets ( see proposition [ prop : orbipack : facets ] ) . a set @xmath283 with @xmath284 and @xmath285
is called a _ shifted
column_. it is a _ shifting _ of each of the columns @xmath286 as a special case we have column @xmath287 , which is the shifted column @xmath288 for @xmath261 .
by definition , if @xmath289 is a shifted column , then so is @xmath290 for every @xmath291 .
[ lem : xcapsc ] let @xmath14 be a vertex of @xmath217 with @xmath258 @xmath259
. then we have @xmath292 for all shiftings @xmath293 of @xmath294 .
we proceed by induction on @xmath9 .
the case @xmath295 follows from lemma [ lem : xcapcol ] , because the only shifting of @xmath296 is @xmath296 itself .
therefore , let @xmath297 , and let @xmath298 be a shifting of @xmath294 ( hence , @xmath299 ) .
since by assumption @xmath300 , lemma [ lem : xcapcol ] yields that there is some @xmath301 with @xmath302 .
if @xmath303 , then we are done . otherwise , @xmath304 holds .
hence , @xmath305 is a shifting of ( @xmath306 and hence of ) @xmath307 , which , by the inductive hypothesis , must intersect @xmath308 .
[ def : scis ] for @xmath309 , @xmath310 , and a shifting @xmath293 of @xmath294 , we call @xmath311 a _ shifted column inequality _ _ ( sci)_. the set @xmath53 is the _ bar _ of the sci , and @xmath161 is the _ leader _ of ( the bar of ) the sci . the set @xmath293 is the _ shifted column ( sc ) _ of the sci .
see figure [ fig : scispart ] for examples . in particular ,
all column inequalities are shifted column inequalities .
the class of shifted column inequalities , however , is substantially richer : it contains exponentially many inequalities ( in @xmath29 ) .
[ prop : scivalid ] shifted column inequalities are valid both for the packing orbitopes @xmath217 and for the partitioning orbitopes @xmath69 .
as @xmath69 is a face of @xmath217 , it is enough to prove the proposition for packing orbitopes @xmath217 .
therefore , let @xmath309 , with @xmath167 , and let @xmath312 be a shifting of @xmath294 .
denote by @xmath53 the bar of the corresponding sci .
let @xmath247 be a vertex of @xmath217 . if @xmath313 , then clearly @xmath314 holds .
otherwise , there is a unique element @xmath315 . as @xmath316
, we have @xmath301 .
therefore @xmath317 is a shifting of @xmath318 .
thus , by lemma [ lem : xcapsc ] , we have @xmath319 .
this shows @xmath320 , implying @xmath321 . in order to devise an efficient separation algorithm for scis
, we need a method to compute minimal shifted columns with respect to a given weight vector @xmath322 .
the crucial observation is the following .
let @xmath323 with @xmath324 be a shifting of @xmath325 for @xmath326 with @xmath327 . if @xmath328 , then @xmath293 is a shifting of @xmath294 ( _ case 1 _ ) .
if @xmath329 , then @xmath330 is a shifting of @xmath331 ( _ case 2 _ ) ; see figure [ fig : sep ] . .,title="fig:"][sfl
: sep1 ] .,title="fig:"][sfl
: sep2 ] for all @xmath326 , let @xmath332 be the weight of a @xmath333-minimal shifting of @xmath325 .
the table @xmath334 can be computed by dynamic programming as follows ; we also compute a table of values @xmath335 , for each @xmath336 , which are needed later to reconstruct the corresponding shifted columns : for @xmath337 , initialize @xmath338}}\}}$ ] . for @xmath339 , initialize @xmath340 . for @xmath339 , @xmath341 ( with @xmath326
) : compute @xmath342 corresponding to cases 1 and 2 , respectively .
then set @xmath343 thus , the tables @xmath334 and @xmath344 can be computed in time @xmath75 . furthermore ,
for a given @xmath326 , we can compute a @xmath333-minimal shifting @xmath345 of @xmath325 in time @xmath346 from the table @xmath344 : we have @xmath347 for all @xmath93}$ ] , @xmath348 for all @xmath349}$ ] , and @xmath350 for all other @xmath336 .
this proves the following result .
[ thm : sepsci ] let @xmath322 be a given weight vector .
there is an @xmath75 time algorithm that simultaneously computes the weights of @xmath333-minimal shiftings of @xmath325 for all @xmath326 and a data structure that afterwards , for a given @xmath336 , allows to determine a corresponding shifted column in @xmath346 steps .
in particular , we obtain the following : [ cor : sepsci ] the separation problem for shifted column inequalities can be solved in linear time @xmath75 .
let a point @xmath351 be given .
we can compute the @xmath352-values @xmath353 of all bars @xmath354 in linear time in the following way : first , we initialize @xmath355 for all @xmath119}$ ] and @xmath356 .
then , for each @xmath357}$ ] , we calculate the value @xmath358 for @xmath359 .
using theorem [ thm : sepsci ] ( and the notations introduced in the paragraphs preceeding it ) , we compute the table @xmath334 and the mentioned data structure in time @xmath75 . then in time @xmath75 we check whether there exists an @xmath360 with @xmath167 and @xmath361 .
if there exists such an @xmath336 , we compute the corresponding shifted column @xmath362 ( in additional time @xmath363 ) , yielding an sci that is violated by @xmath352 .
otherwise @xmath352 satisfies all scis . of course
, the procedure described in the proof of the corollary can be modified to find a maximally violated sci if @xmath364 does not satisfy all scis . in this section
we prove that nonnegativity constraints , row - sum equations , and scis suffice to describe partitioning and packing orbitopes for symmetric groups
. the proof will be somewhat more involved than in the case of cyclic groups .
in particular , the coefficient matrices are not totally unimodular anymore . in order to see this , consider the three column inequalities @xmath365 the submatrix of the coefficient matrix belonging to these three rows and the columns corresponding to @xmath366 , @xmath367 , and @xmath368 is the matrix @xmath369 whose determinant equals @xmath370 .
note that the above three inequalities define facets both of @xmath217 and @xmath69 for @xmath371 ( see propositions [ prop : orbipack : facets ] and [ prop : orbipart : facets ] , respectively ) .
[ prop : orbipart : descr ] the partitioning orbitope @xmath69 is completely described by the nonnegativity constraints , the row - sum equations , and the shifted column inequalities : @xmath372 let @xmath373 be the polyhedron on the right - hand side of the statement above . from propositions [ prop : ipsym ] and [ prop : scivalid ] we know already that @xmath374 holds .
thus , it suffices to show that @xmath373 is an integral polytope ( as @xmath69 is by definition ) . in the following ,
we first describe the strategy of the proof . for the rest of the proof , fix an arbitrary vertex @xmath22 of @xmath373 .
a _ basis _
@xmath375 of @xmath352 is a cardinality @xmath376 subset of the constraints describing @xmath373 that are satisfied with equality by @xmath352 with the property that the @xmath377-coefficient matrix of the left - hand sides of the constraints in @xmath375 is non - singular .
thus , the equation system obtained from the constraints in @xmath375 has @xmath352 as its unique solution .
we will show that there exists a basis @xmath378 of @xmath352 that does not contain any sci .
thus , @xmath378 contains a subset of the @xmath28 row - sum equations and at least @xmath379 nonnegativity constraints .
this shows that @xmath352 has at most @xmath28 nonzero entries and , since @xmath352 satisfies the row - sum equations , it has a nonzero entry in every row
. therefore , @xmath378 contains all @xmath28 row - sum equations , and all @xmath28 nonzero entries must in fact be @xmath1 . hence , @xmath352 is a 0/1-point .
so the existence of such a basis proves the proposition .
the _ weight _ of a shifted column @xmath380 with @xmath381 ( we will not need shifted columns with @xmath382 here , as they do not appear in scis ) is @xmath383 in particular , if @xmath384 and @xmath385 are two shifted columns with @xmath386 , then we have @xmath387 .
the _ weight _ of an sci is the weight of its shifted column , and the _ weight _ of a basis @xmath388 is the sum of the weights of the scis contained in @xmath388 ( note that a shifted column can appear in several scis ) .
a basis of @xmath352 that contains all row - sum equations and all nonnegativity constraints corresponding to @xmath171-entries of @xmath352 is called _ reduced_. as the coefficient vectors ( of the left - hand sides ) of these constraints are linearly independent , some reduced basis of @xmath352 exists .
hence , there is also a reduced basis @xmath378 of @xmath352 of minimal weight . to prove the proposition
, it thus suffices to establish the following claim .
[ claim : minrednosci ] a reduced basis of @xmath352 of minimal weight does not contain any sci .
the proof of claim [ claim : minrednosci ] consists of three parts : we show that a reduced basis of @xmath352 does not contain any `` trivial scis '' ( claim [ claim : trivialscis ] ) .
we prove that a reduced basis of @xmath352 of minimal weight satisfies three structural conditions on its ( potential ) scis ( claim [ claim : basicstructure ] ) . finally , assuming that a reduced basis of @xmath352 with minimal weight contains at least one sci
, we will derive a contradiction by constructing a different solution @xmath389 of the corresponding equation system .
we are now ready to start with part 1 .
we call an sci with shifted column @xmath293 _ trivial _ if @xmath390 holds or if we have @xmath391 and @xmath392 for all @xmath393 for some @xmath394 ( thus satisfying @xmath395 ) ( see figure [ fig : forbidden ] ) .
[ claim : trivialscis ] a reduced basis @xmath388 of @xmath352 does not contain any trivial scis .
let @xmath293 be the shifted column @xmath293 and @xmath53 be the bar of some sci that is satisfied with equality by @xmath352 .
if @xmath396 , then the coefficient vector of the sci is a linear combination of the coefficient vectors of the inequalities @xmath397 for @xmath398 , which all are contained in @xmath388 ( due to @xmath399 ) .
since the coefficient vectors of the inequalities in @xmath388 form a non - singular matrix , the sci can not be in @xmath388 .
( by `` coefficient vector '' we always mean the vector formed by the coefficients of the left - hand side of a constraint . )
if @xmath293 contains exactly one entry @xmath400 with @xmath401 , then we have @xmath402 .
let @xmath10 be the index of the row that contains the bar @xmath53 .
the nonnegativity constraints @xmath403 for @xmath404 , @xmath405 for @xmath406 , and @xmath407 for @xmath408 are contained in @xmath388 .
since the coefficient vector of the considered sci can linearly be combined from the coefficient vectors of these nonnegativity constraints and of the row - sum equations @xmath409 and @xmath182 , this sci can not be contained in @xmath388 .
[ claim : basicstructure ] a minimal weight reduced basis @xmath388 of @xmath352 satisfies the following three conditions : [ reduced : firstrowentry ] if @xmath51 is contained in the shifted column of some sci in @xmath388 , then there exists some @xmath410 with @xmath411 .
[ reduced : nonzeroleader ] if @xmath161 is the leader of an sci in @xmath388 , then @xmath412 holds .
[ reduced : containleader ] if @xmath161 is the leader of an sci in @xmath388 , then there is no sci in @xmath388 whose shifted column contains @xmath161 .
see figure [ fig : forbidden ] , for an illustration of the three conditions .
assume there exists an sci in @xmath388 with shifted column @xmath293 and bar @xmath53 that contains the first nonzero entry of a row @xmath12 , i.e. , there is @xmath400 with @xmath413 and @xmath414 for all @xmath410 .
let @xmath415 be the entries of @xmath293 above row @xmath12 .
let @xmath416 and @xmath417 .
see figure [ fig : basicstructure ] for an illustration . because @xmath418 is a shifting of @xmath419
, @xmath420 is an sci and hence satisfied by @xmath352 .
since we have @xmath421 ( thus , @xmath422 ) , it suffices to show that replacing the original sci @xmath423 by @xmath424 gives another basis @xmath425 of @xmath352 ( which also is reduced ) , contradicting the minimality of the weight of @xmath388 .
due to @xmath426 , @xmath427 , @xmath428 , and @xmath429 we have @xmath430 therefore , equality must hold throughout this chain .
in particular , this shows @xmath431 .
thus , its suffices to show that the coefficient matrix of the equation system obtained from @xmath425 is non - singular , which can be seen as follows .
since @xmath432 ( see ) , we know that all nonnegativity constraints @xmath403 with @xmath433 are contained in @xmath388 and @xmath425 .
the same holds for @xmath405 with @xmath434 and for @xmath407 with @xmath408 , where row @xmath10 contains bar @xmath53 ( since @xmath435 by ) .
thus , we can linearly combine the coefficient vector of @xmath436 from the coefficient vectors of the constraints @xmath437 , @xmath438 , @xmath439 , and the nonnegativity constraints mentioned above .
since all these constraints are contained in @xmath425 , this shows that the coefficient matrix of @xmath425 has the same row - span as that of @xmath388 , thus proving that it is non - singular as well .
assume that there exists an sci in @xmath388 with leader @xmath161 , bar @xmath53 , and shifted column @xmath293 such that @xmath440 .
if @xmath441 , then we have @xmath442 .
define @xmath443 , @xmath444 , and observe that @xmath445 , @xmath446 , i.e. , @xmath447 and @xmath448 , because a reduced basis does not contain trivial scis by claim [ claim : trivialscis ] ; see figure [ fig : basicstructure ] .
hence , @xmath424 is an sci .
we therefore have : @xmath449 where the first equation holds because @xmath423 is satisfied with equality by @xmath352 and the second equation follows from @xmath450 .
hence , we know that @xmath451 . since we have @xmath421 ( and consequently @xmath452 ) , again it remains to show that the coefficient vector of @xmath423 can be linearly combined from the coefficient vector of @xmath424 and some coefficient vectors of nonnegativity constraints in @xmath388 and @xmath425 .
but this is clear , as we have @xmath450 and @xmath453 , where the latter follows from .
assume that in @xmath388 there exists an sci @xmath454 with leader @xmath255 , bar @xmath455 , and shifted column @xmath456 ( in particular : @xmath457 ) and another sci @xmath458 with bar @xmath459 and shifted column @xmath460 hence , we have @xmath461 .
define @xmath462 ( i.e , the part of @xmath385 lying strictly above row @xmath10 ) and @xmath463 ( i.e , @xmath384 together with the part of @xmath385 strictly below row @xmath10 ) .
clearly , @xmath464 is a shifting of @xmath465 , and @xmath466 is a shifted column as well ( due to @xmath467 ) .
thus , with @xmath468 , we obtain the scis @xmath469 @xmath470 ( see figure [ fig : basicstructure ] ) .
since and are contained in @xmath388 , we have @xmath471 and @xmath472 .
adding these two equations yields @xmath473 because @xmath474 cancels due to @xmath475 .
since @xmath352 satisfies the scis and , equation shows that in fact we have @xmath476 and @xmath477 .
it is not clear , however , that we can simply replace and by and in order to obtain a new basis of @xmath352 .
nevertheless , if @xmath478 , and @xmath479 are the coefficient vectors of , , , and , respectively , we have @xmath480 , which implies @xmath481 let @xmath482 be the subspace of @xmath231 that is spanned by the coefficient vectors of the constraints different from in @xmath388 .
thus , the linear span of @xmath483 is the whole space @xmath231 . due to
, the same holds for @xmath484 ( since @xmath485 ) .
therefore , there is @xmath486 such that @xmath487 spans @xmath231 .
let @xmath488 be the corresponding sci from @xmath489 .
hence , @xmath490 is a ( reduced ) basis of @xmath352 as well .
since we have @xmath491 and @xmath492 ( due to @xmath457 ) , the weight of @xmath425 is smaller than that of @xmath388 , contradicting the minimality of the weight of @xmath388 .
before we finish the proof of the proposition by establishing claim [ claim : minrednosci ] , we need one more structural result on the scis in a reduced basis of @xmath352 .
let @xmath493 be any shifted column with @xmath494 for some @xmath495}$ ] .
we call @xmath496 the _ first nonzero element _ of @xmath293 if @xmath497 holds .
similarly , @xmath496 is called the _ last nonzero element _ of @xmath293 if we have @xmath498 [ claim : firstlast ] let @xmath388 be a reduced basis of @xmath352 , and let @xmath499 be the shifted columns of some scis in @xmath388 ( @xmath500 is allowed ) .
[ claim : firstlast : firstnonzero ] if @xmath161 is the first nonzero element of @xmath384 and @xmath501 , then @xmath161 is also the first nonzero element of @xmath385 .
[ claim : firstlast : lastnonzero ] if @xmath161 is the last nonzero element of @xmath384 with @xmath502 and @xmath501 , then @xmath161 is also the last nonzero element of @xmath385 and @xmath503 .
[ claim : firstlast : both ] if @xmath161 is the last nonzero element of @xmath384 with @xmath502 , then @xmath161 is not the first nonzero element of @xmath385 . . ]
let @xmath504 be two shifted columns of scis with bars @xmath455 and @xmath459 , respectively , in the reduced basis @xmath388 of @xmath352 .
suppose that @xmath505 , i.e. , @xmath506 holds .
define @xmath507 and @xmath508 , see figure [ fig : firstlast ] .
since @xmath509 holds , @xmath510 is a shifted column and @xmath511 is an sci .
thus , we obtain @xmath512 furthermore , since @xmath513 is contained in the basis @xmath388 of @xmath352 , we have @xmath514 subtracting from yields @xmath515 .
we thus conclude @xmath516 ( where the second inequality follows by exchanging the roles of @xmath384 and @xmath385 in the argument ) .
if @xmath161 is the first nonzero element of @xmath384 , then we have @xmath517 .
thus , the first inequality of implies @xmath518 , showing that @xmath161 is the first nonzero element of @xmath385 .
if @xmath161 is the last nonzero element of @xmath384 and @xmath519 holds , then we have @xmath520 . with the second inequality of
we obtain : @xmath521 where the last equation holds because @xmath513 is contained in @xmath388 .
it follows that @xmath503 and @xmath161 is the last nonzero element of @xmath385 .
this follows from the first two parts of the claim , since @xmath388 does not contain any trivial scis by claim [ claim : trivialscis ] .
we will now proceed with the proof of claim [ claim : minrednosci ] .
thus , assume that @xmath378 is a reduced basis of @xmath352 of minimal weight and suppose that @xmath522 contains at least one sci .
we are going to construct a point @xmath523 that satisfies the equation system obtained from @xmath378 , contradicting the fact the @xmath352 is the unique solution to this system of equations . at the beginning , we set @xmath524 , and let @xmath525 be an arbitrary positive number
. then we perform the following four steps ( see figure [ fig : finish ] for illustrations of the first three ) .
[ step : smallestentry ] for every @xmath161 that is the first nonzero element of the shifted column of at least one sci in @xmath378 , we reduce @xmath526 by @xmath527 .
[ step : largestentryvalue1 ] for every @xmath161 that is the last nonzero element of the shifted column @xmath293 of at least one sci in @xmath378 with @xmath528 , we increase @xmath526 by @xmath527 .
[ step : adjustbasis ] for each @xmath119}$ ] and for all @xmath529 ( in this order ) : if @xmath161 is the leader of some sci in @xmath378 , we adjust @xmath526 such that , with @xmath530 , @xmath531 holds .
[ step : correctrowsum ] for each @xmath119}$ ] , adjust @xmath526 in order to achieve @xmath532 , where @xmath533 .
the reason for treating the case @xmath534 separately in step [ step : largestentryvalue1 ] will become evident in the proof of claim [ claim : step4 ] below .
the following four claims will yield that @xmath535 is a solution of the equation system corresponding to @xmath522 .
[ claim : step2 ] after step [ step : largestentryvalue1 ] , for each shifted column @xmath293 of some sci in @xmath378 we have @xmath536 let @xmath293 be the shifted column of some sci in @xmath522 .
it follows from part of claim [ claim : firstlast ] that the first nonzero element @xmath161 of @xmath293 is the only element in @xmath293 whose @xmath535-component is changed ( reduced by @xmath527 ) in step [ step : smallestentry ] .
thus , after step [ step : smallestentry ] we have @xmath537 . if @xmath538 , then , by part of claim [ claim : firstlast ] , @xmath539 is not changed in step [ step : largestentryvalue1 ]
otherwise , @xmath391 , and @xmath540 is increased by @xmath527 in step [ step : largestentryvalue1 ] , where @xmath51 is the last nonzero element of @xmath293 . according to part of claim [ claim : firstlast ] , no other component of @xmath535 belonging to some element in @xmath293 is changed in step [ step : largestentryvalue1 ] .
thus , in both cases the claim holds .
[ claim : step3a ] no component of @xmath535 belonging to the shifted column of some sci in @xmath378 is changed in step [ step : adjustbasis ] .
let @xmath293 be the shifted column of some sci in @xmath522 .
according to part of claim [ claim : basicstructure ] , @xmath293 does not contain the leader of any sci in @xmath522 , since @xmath378 is a reduced basis of minimal weight .
[ claim : step3b ] after step [ step : adjustbasis ] , for each sci in @xmath378 with shifted column @xmath293 and bar @xmath53 we have @xmath541 . for an sci in @xmath522 with shifted column @xmath293 and bar @xmath53 , we have @xmath542 .
thus , from claims [ claim : step2 ] and [ claim : step3a ] it follows that @xmath543 holds after step [ step : adjustbasis ] .
[ claim : step4 ] step [ step : correctrowsum ] does not change any component of @xmath535 that belongs to the shifted column or the bar of some sci in @xmath378 .
let @xmath161 be such that @xmath544 for all @xmath545 and @xmath546 .
by part of claim [ claim : basicstructure ] , @xmath161 is not contained in any shifted column of an sci in @xmath378 .
if @xmath161 is contained in the bar @xmath53 of some sci in @xmath378 , then clearly @xmath435 holds .
thus , after step [ step : adjustbasis ] , we have @xmath547 , which shows that @xmath526 is not changed in step [ step : correctrowsum ] .
we can now finish the proof of the proposition .
claims [ claim : step3b ] and [ claim : step4 ] show that @xmath535 satisfies all scis contained in @xmath378 with equality .
furthermore , in all steps of the procedure only components @xmath526 with @xmath546 are changed ( this is clear for steps [ step : smallestentry ] , [ step : largestentryvalue1 ] , and [ step : correctrowsum ] ; for step [ step : adjustbasis ] it follows from part of claim [ claim : basicstructure ] ) .
since after step [ step : correctrowsum ] , @xmath535 satisfies all row - sum equations , this proves that @xmath535 is a solution to the equation system obtained from @xmath378 .
we assumed that @xmath378 contains at least one sci .
let @xmath293 be the shifted column of one of these .
we know @xmath548 by claim [ claim : trivialscis ] .
thus , let @xmath161 be the first nonzero element of @xmath293 .
hence , after step [ step : smallestentry ] , we have @xmath549 . by part of claim [ claim : firstlast ] , this still holds after step [ step : largestentryvalue1 ] .
as @xmath526 is also not changed in steps [ step : adjustbasis ] and [ step : correctrowsum ] ( see claims [ claim : step3a ] and [ claim : step4 ] ) , we deduce @xmath523 , contradicting the fact that @xmath352 is the unique solution to the equation system belonging to @xmath378 .
this concludes the proof of proposition [ prop : orbipart : descr ] .
we hope that reading this proof was somewhat enjoyable . anyway , at least it also gives us a linear description of the packing orbitopes for symmetric groups almost for free .
[ prop : orbipack : descr ] the packing orbitope @xmath217 is completely described by the nonnegativity constraints , the row - sum inequalities , and the shifted column inequalities : @xmath550 let @xmath551 be the polyhedron on the right - hand side of the statement .
we define @xmath552}}\}}$ ] .
the proof of proposition [ prop : orbipart : descr ] in fact shows that its statement remains true if we drop all scis with shifted column @xmath293 and @xmath553 from the linear description .
this follows from the fact that , due to @xmath554 and claim [ claim : trivialscis ] , no such sci can be contained in any reduced basis of @xmath352 ( using the notations from the proof of proposition [ prop : orbipart : descr ] ) .
thus we obtain @xmath555 with @xmath556 where the last inequalities are equivalent ( with respect to @xmath557 ) to the nonnegativity constraints associated with the elements of @xmath558 by addition of row - sum equations .
define @xmath559}}\}}$ ] , and denote by @xmath560 the orthogonal projection .
since none of the inequalities defining @xmath561 has a nonzero coefficient in @xmath558 , we have @xmath562 , hence @xmath563 .
this yields @xmath564 , which , due to @xmath565 , implies @xmath566 .
thus , we obtain @xmath567 where the first equation is due to proposition [ prop : projpartpack ] , the second equation follows from , and the final arises from identifying @xmath233 with @xmath231 . in this section ,
we investigate which of the constraints from the linear descriptions of @xmath69 and @xmath217 given in propositions [ prop : orbipart : descr ] and [ prop : orbipack : descr ] , respectively , define facets .
this will also yield non - redundant descriptions .
it seems to be more convenient to settle the packing case first and then to carry over the results to the partitioning case .
recall that we assume @xmath568 .
[ prop : orbipack : facets ] [ prop : orbipack : facets : fulldimensional ] the packing orbitope @xmath569 is full dimensional : @xmath570 a nonnegativity constraint @xmath180 , @xmath571 , defines a facet of @xmath217 , unless @xmath572 holds .
the faces defined by @xmath573 with @xmath574 are contained in the facet defined by @xmath575 .
[ prop : orbipack : facets : nonneg ] every row - sum constraint @xmath192 for @xmath119}$ ] defines a facet of @xmath217 .
[ prop : orbipack : facets : rowsum ] [ prop : orbipack : facets : sci ] a shifted column inequality @xmath436 with bar @xmath53 and shifted column @xmath576 defines a facet of @xmath217 , unless @xmath577 and @xmath578 ( exception i ) or @xmath579 and @xmath580 ( exception ii ) hold . in case of exception i , the corresponding face is contained in the facet defined by the sci with bar @xmath53 and shifted column @xmath581 . in case of exception ii ,
the face is contained in the facet defined by the sci @xmath582 . for all @xmath583 , we define @xmath584 by @xmath585 that is
, @xmath586 has @xmath1-entries at position @xmath51 and on the main diagonal up to column @xmath587 , except that @xmath588 unless @xmath589 ; see figure [ fig : orbipack : facets ] .
the columns of each @xmath586 are in non - increasing lexicographic order .
hence , by part of observation [ obs : charvert ] , each @xmath586 is a vertex of @xmath217 . in order to show that these vectors are linearly independent
, we fix an arbitrary ordering of the @xmath586 that starts with @xmath590 . for each @xmath583 ,
all points @xmath591 preceding @xmath586 have a @xmath171-entry at position @xmath51 , while @xmath592 .
this shows that these @xmath376 vertices of @xmath217 are linearly independent .
together with @xmath27 this gives @xmath593 affinely independent points contained in @xmath217 , proving that @xmath217 is full dimensional .
the calculations in the statement are straightforward . for @xmath594 all points @xmath586 with @xmath595
are contained in the face defined by @xmath180 .
since this is also true for @xmath27 , the face defined by @xmath180 contains @xmath376 affinely independent points ( see the proof of part ) , i.e. , it is a facet of @xmath217 . for every vertex @xmath596
contained in the face defined by @xmath573 for some @xmath574 , we have @xmath597 for all @xmath598 ( because otherwise the columns of @xmath352 would not be in non - increasing lexicographic order ) .
this shows that @xmath22 is contained in the facet defined by @xmath575 . in order to show that @xmath192 defines a facet of @xmath217 for @xmath119}$ ] , we construct points @xmath599 ( depending on @xmath10 ) from the points @xmath586 defined in part by adding a @xmath1 at position @xmath600 if @xmath601 ( see figure [ fig : orbipack : facets ] ) .
the @xmath602 points @xmath599 for all @xmath603 , and the unit vector @xmath604 ( with a single @xmath1 in position @xmath600 ) satisfy @xmath182 .
furthermore , they are affinely independent , since subtracting @xmath604 from all vectors @xmath599 yields vectors @xmath605 , which can be shown to be linearly independent similarly to part ; here , we need @xmath606 .
let @xmath423 be an sci with bar @xmath53 , leader @xmath607 , and shifted column @xmath608 .
if @xmath577 and @xmath578 hold ( exception i ) , then the sci is the sum of the sci @xmath609 and the sci with bar @xmath53 and shifted column @xmath610 ; see figure [ fig : orbipack : facets ]
. repeating this argument @xmath611 times proves the second statement of part for exception i. if @xmath579 and @xmath612 with @xmath613 hold ( exception ii ) , then the sci is the sum of the scis @xmath582 , , @xmath614 .
this proves the second statement of part for exception ii . otherwise , let @xmath615 be the set of vertices of @xmath217 that satisfy the sci with equality , and let @xmath616 be the linear span of @xmath615 and the unit vector @xmath617
. we will show that @xmath618 , which proves @xmath619 ( since @xmath620 ) .
hence , the sci defines a facet of @xmath217 .
[ r][r ] .,title="fig:"][sfl : packsci2 ] .,title="fig:"][sfl : packsci3 ] to show that @xmath618 , we prove that @xmath621 for all @xmath622 .
we partition the set @xmath623 into three parts ( see figure [ fig : orbipack : facets : sci ] ): @xmath624 for @xmath625 , denote by @xmath626 the diagonal starting at @xmath627 and ending at @xmath628 .
similarly , denote by @xmath629 the diagonal starting at @xmath630 and ending in @xmath631 or in @xmath632 .
[ claim : facets : ac ] for all @xmath633 we have @xmath634 .
denote the incidence vector of @xmath635 by @xmath636 ( see figure [ fig : orbipack : facets : sci ] ) . both @xmath637 and @xmath638 are vertices of @xmath217 .
we have @xmath639 for @xmath640 .
furthermore @xmath641 for @xmath642 .
hence , these two vertices satisfy the sci with equality and we obtain @xmath643 . [
claim : facets : d ] for all @xmath644 we have @xmath634 .
define the set @xmath645 see figure [ fig : orbipack : facets : sci ] .
let @xmath646 . by construction ,
the three points @xmath647 , @xmath648 , and @xmath649 are vertices of @xmath217 . if @xmath650 , we have @xmath651 and @xmath652 , where we need @xmath653 in case of @xmath654 ( notice that in case of @xmath579 we have @xmath655 ) . due to @xmath656 , both @xmath647 and @xmath648
satisfy the sci with equality .
this yields @xmath657 .
if @xmath658 , then @xmath659 does not hold in all cases ( e.g. , if @xmath660 , we have @xmath661 ) .
however , since @xmath658 , @xmath662 is well - defined and @xmath663 hold .
hence the vertices @xmath664 and @xmath665 satisfy the sci with equality , giving @xmath666 .
[ claim : facets : s ] for all @xmath667 we have @xmath634 .
define the set @xmath668 see figure [ fig : orbipack : facets : sci ] .
the incidence vector @xmath669 is a vertex of @xmath217 , which , due to @xmath670 and @xmath671 satisfies the sci with equality .
thus , from @xmath672 we conclude @xmath621 , since @xmath673 by definition of @xmath233 , and @xmath674 for all @xmath675 by claims [ claim : facets : ac ] and [ claim : facets : d ] .
[ claim : facets : b ] for all @xmath676 we have @xmath634 .
the vector @xmath677 is a vertex of @xmath217 that satisfies the sci with equality .
furthermore , we have @xmath678 where @xmath679 .
thus , we conclude @xmath680 , since @xmath681 for all @xmath682 by claims [ claim : facets : ac ] , [ claim : facets : d ] , and [ claim : facets : s ] . claims [ claim : facets : ac ] to [ claim : facets : b ] show @xmath683 for all @xmath622 .
this proves that the sci defines a facet of @xmath217 ( unless exception i or ii hold ) .
finally , we carry the results of proposition [ prop : orbipack : facets ] over to partitioning orbitopes .
[ prop : orbipart : facets ] the partitioning orbitope @xmath684 has dimension @xmath685 the constraints @xmath182 form a complete and non - redundant linear description of @xmath686 . [
prop : orbipart : facets : nonnegative ] a nonnegativity constraint @xmath180 , @xmath571 , defines a facet of @xmath69 , unless @xmath572 holds .
the faces defined by @xmath573 with @xmath574 are contained in the facet defined by @xmath575 .
[ prop : orbipart : facets : sci ] a shifted column inequality @xmath423 with bar @xmath53 and shifted column @xmath576 defines a facet of @xmath69 , unless @xmath687 ( exception i ) or @xmath577 and @xmath578 ( exception ii ) or @xmath579 and @xmath580 ( exception iii ) . in case of exception i , the corresponding face is contained in the facet defined by @xmath688 , where @xmath10 is the index of the row containing @xmath53 . in case of exception ii ,
the face is contained in the facet defined by the sci with bar @xmath53 and shifted column @xmath689 . in case of exception iii ,
the face is contained in the facet defined by the sci @xmath582 . according to proposition [ prop : projpartpack ] ,
@xmath226 is isomorphic to @xmath69 via the orthogonal projection of the latter polytope to the space @xmath690}}\}}\ ] ] ( and via the canonical identification of @xmath233 and @xmath234 ) .
this shows the statement on the dimension of @xmath69 ; the calculations and the claim on the non - redundancy of the equation system are straightforward .
furthermore , this projection ( which is one - to - one on @xmath686 ) maps every face of @xmath69 that is defined by some inequality @xmath691 with @xmath692 , @xmath693 , and @xmath694 for all @xmath119}$ ] to a face of @xmath226 of the same dimension defined by @xmath695 conversely , if @xmath696 defines a face of @xmath226 for @xmath697 and @xmath698 , then the inequality @xmath699 defines a face of @xmath69 of the same dimension .
due to parts and of proposition [ prop : orbipack : facets ] , this proves part of the proposition , where we use the fact that the inequalities @xmath700 are equivalent to @xmath701 with respect to @xmath69 .
furthermore , due to part of proposition [ prop : orbipack : facets ] , the above arguments also imply the statements of part for @xmath702 ( including exception ii and iii ) .
finally , we consider the case @xmath687 ( exception i ) .
since we have @xmath703 for all @xmath704 , the equation @xmath705 implies @xmath706 and hence @xmath707 ( using the row - sum equation for row @xmath10 containing @xmath53 ) .
this concludes the proof .
we collect the results on the packing- and partitioning orbitopes for symmetric groups .
[ thm : resultspart ] the partitioning orbitope @xmath69 @xmath228for @xmath708 with respect to the symmetric group @xmath55 equals the set of all @xmath709 that satisfy the following linear constraints : the row - sum equations @xmath182 for all @xmath119}$ ] , the nonnegativity constraints @xmath180 for all @xmath710 , the shifted column inequalities @xmath423 for all bars @xmath711 with @xmath309 , @xmath167 , and shifted columns @xmath712 where in case of @xmath579 the last condition reduces to @xmath713 and we additionally require @xmath714 .
this system of constraints is non - redundant .
the corresponding separation problem can be solved in time @xmath75 . for the result on the completeness of the description ,
see proposition [ prop : orbipart : descr ] , for the question of redundancy see proposition [ prop : orbipart : facets ] , and for the separation algorithm see corollary [ cor : sepsci ] .
note that the sci with shifted column @xmath715 and bar @xmath716 defines the same facet of @xmath69 as the nonnegativity constraint @xmath717 .
[ thm : resultspack ] the packing orbitope @xmath70 @xmath228for @xmath718 with respect to the symmetric group @xmath55 equals the set of all @xmath709 that satisfy the following linear constraints : the row - sum inequalities @xmath192 for all @xmath719}$ ] , the nonnegativity constraints @xmath180 for all @xmath720 , the shifted column inequalities @xmath721 for all bars @xmath711 with @xmath309 , @xmath167 , and shifted columns @xmath722 where in case of @xmath579
we additionally require @xmath714 .
this system of constraints is non - redundant .
the corresponding separation problem can be solved in time @xmath75 . for the result on the completeness of the description ,
see proposition [ prop : orbipack : descr ] , for the question of redundancy see proposition [ prop : orbipack : facets ] , and for the separation algorithm see corollary [ cor : sepsci ] .
we close with some remarks on the technique used in the proof of proposition [ prop : orbipart : descr ] , on the combination of scis and clique - inequalities for the graph - coloring problem , and on full and covering orbitopes .
our technique to prove proposition [ prop : orbipart : descr ] can be summarized as follows .
assume a polytope @xmath723 is described by some ( finite ) system @xmath724 of linear equations and inequalities .
suppose that @xmath725 is a subsystem of @xmath724 for which it is known that @xmath725 defines an integral polytope @xmath726 .
one can prove that @xmath727 is integral by showing that every vertex @xmath352 of @xmath727 is a vertex of @xmath728 in the following way .
here we call a basis ( with respect to @xmath724 ) of @xmath352 _ reduced _ if it contains as many constraints from @xmath725 as possible : [ item : remarks : weight ] starting from an arbitrary reduced basis @xmath375 of @xmath352 , construct iteratively a reduced basis @xmath729 of @xmath352 that satisfies some properties that are useful for the second step .
such a proof is conceivable for every 0/1-polytope @xmath727 by choosing @xmath732^n$ ] as the whole @xmath0-cube and @xmath725 as the set of the @xmath733 trivial inequalities @xmath734 , for @xmath735 ( if necessary , modifying @xmath724 in order to contain them all ) .
we do not know whether this kind of integrality proof has been used in the literature
. it may well be that one can interpret some of the classical integrality proofs in this setting .
anyway , it seems to us that the technique might be useful for other polytopes as well . as mentioned in the introduction , for concrete applications like the graph coloring problem one can ( and probably has to ) combine the polyhedral knowledge on orbitopes with the knowledge on problem specific polyhedra .
we illustrate this by the example of clique inequalities for the graph coloring model described in the introduction .
fix a color index @xmath736}$ ] .
if @xmath737 is a clique in the graph @xmath738 , then clearly the inequality @xmath739 is valid .
in fact , the strengthened inequalities @xmath740 are known to be facet - defining for the convex hull of the solutions to , see @xcite .
suppose that @xmath741 is a shifted column and that we have @xmath742 for all @xmath261 with @xmath743 .
then the inequality @xmath744 is valid for all solutions to the model obtained from by adding inequalities ( which are all `` column inequalities '' in terms of orbitopes ) , see figure [ fig : cliquesci ] .
the details and a computational study will be the subject of a follow - up paper . with respect to cyclic group actions , we loose the simplicity of the characterizations in observation [ obs : charvert ] .
the reason is that the matrices under investigation may have several equal nonzero columns .
in particular , the lexicographically maximal column may not be unique . with respect to the action of the symmetric group
, we still have the characterization of the representatives as the matrices whose columns are in non - increasing lexicographic order ( see part [ obs : charvert : sym ] of observation [ obs : charvert ] ) .
the structures of the respective full and covering orbitopes , however , become much more complicated . in particular , we know from computer experiments that several powers of two arise as coefficients in the facet - defining inequalities .
this increase in complexity is reflected by the fact that optimization of linear functionals over these orbitopes seems to be more difficult than over packing and partitioning orbitopes ( see the remarks at the end of section [ sec : optimizingoverorbitopes ] ) .
let us close with a comment on our choice of the set of representatives as the maximal elements with respect to a lexicographic ordering ( referring to the row - wise ordering of the components of the matrices )
. it might be that the difficulties for full and covering orbitopes mentioned in the previous paragraph can be overcome by the choice of a different system of representatives .
the choice of representatives considered in this paper , however , seems to be appropriate for the packing and partitioning cases .
whether the results presented in this paper are useful in practice will turn out in the future . in any case , we hope that the reader shares our view that orbitopes are neat mathematical objects .
it seems that symmetry strikes back by its own beauty , even when mathematicians start to fight it .
we thank the referees for their work .
in particular , we are indepted to one of them for several insightful and constructive remarks , including the proof of theorem [ thm : orbipackcycl ] that we have in this final version .
we furthermore thank yuri faenza , andreas loos , and matthias peinhardt for helpful comments .
t. fahle , s. schamberger , and m. sellmann , _ symmetry breaking _ , in principles and practice of constraint programming
cp 2001 : 7th international conference , t. walsh , ed .
, lncs 2239 , springer - verlag , berlin / heidelberg , 2001 , pp . | we introduce _ orbitopes _ as the convex hulls of @xmath0-matrices that are lexicographically maximal subject to a group acting on the columns .
special cases are packing and partitioning orbitopes , which arise from restrictions to matrices with at most or exactly one @xmath1-entry in each row , respectively .
the goal of investigating these polytopes is to gain insight into ways of breaking certain symmetries in integer programs by adding constraints , e.g. , for a well - known formulation of the graph coloring problem .
we provide a thorough polyhedral investigation of packing and partitioning orbitopes for the cases in which the group acting on the columns is the cyclic group or the symmetric group .
our main results are complete linear inequality descriptions of these polytopes by facet - defining inequalities . for the cyclic group case ,
the descriptions turn out to be totally unimodular , while for the symmetric group case , both the description and the proof are more involved .
the associated separation problems can be solved in linear time . |
the use of quantum systems for information processing was first introduced by benioff @xcite . in 1985
deutsch described quantum computers which exploit the superposition of multi particle states , thereby achieving massive parallelism @xcite .
researchers have also studied the possibility of solving certain types of problems more efficiently than can be done on conventional computers @xcite .
these theoretical possibilities have generated significant interest for experimental realization of quantum computers @xcite .
several techniques are being exploited for quantum computing and quantum information processing , including nuclear magnetic resonance ( nmr ) @xcite .
nmr has played a leading role for the practical demonstration of quantum gates and algorithms .
most of the nmr quantum information processing ( qip ) experiments have utilized systems having indirect spin - spin couplings ( scalar j couplings ) @xcite .
recently nmr qip has been demonstrated in quadrupolar and dipolar coupled systems , obtained by orienting the molecules in liquid crystal media @xcite . in case of homo nuclear dipolar
coupled systems , spins are often strongly coupled and hence can not be addressed individually @xcite
. however the @xmath0 eigen states are collectively treated as an n - qubit system @xcite .
similarly for quadrupolar systems ( spin @xmath1 1/2 ) , individual spins are treated as a multi - qubit system @xcite .
resolved resonance lines provide access to the full hilbert space of @xmath0 dimension @xcite . a quadrupolar nucleus of spin i has ( 2i+1 ) energy levels , which are equi - spaced in the high magnetic field and
the 2i single quantum transitions are degenerate . in the presence of first order quadrupolar interaction
this degeneracy is lifted and gives rise to 2i equi - spaced transitions @xcite . for molecules partially oriented in anisotropic media , such as liquid crystals , such a situation
is often obtained @xcite .
if @xmath2 , such a system can be treated as an n - qubit system @xcite .
the advantage of such systems and the dipolar coupled systems , over the j - coupled systems is that the coupling values are one to two orders of magnitude larger , allowing shorter gate times or the use of shorter transition selective pulses .
so far quadrupolar systems have been used for , quantum simulation , preparation of pseudo pure states , implementation of quantum gates and search algorithms @xcite .
recently das et al have implemented cleve version of 2-qubit dj algorithm on a spin 7/2 nucleus @xcite . in all these cases
the controlled gates are implemented by inverting populations between various levels , by using transition selective @xmath3 pulses .
recently it has been demonstrated that non - adiabatic geometric phases can be used for implementing quantum gates @xcite .
here we use non - adiabatic geometric phases to implement controlled phase shift gates and collins version of dj algorithm on a 3-qubit system obtained by eight eigen states of a spin-7/2 quadrupolar nucleus .
the hamiltonian of a quadrupolar nucleus partially oriented in liquid crystal matrix , in the presence of large magnetic field @xmath4 , having a first order quadrupolar coupling , is given by , @xmath5}s(3i_z^2-i^2 ) = \omega_o i_z+2\pi \wedge ( 3i_z^2-i^2 ) , \label{6eqn1 } \end{aligned}\ ] ] where @xmath6 is the resonance frequency , s is the order parameter at the site of the nucleus , and @xmath7 is the quadrupolar coupling .
if the order parameter s is small , the effective quadrupolar coupling @xmath8 can be of the order of a few khz in spite of @xmath7 being of the order of several mhz .
thus , it is possible to observe the satellite transitions due to first order quadrupolar coupling . a 50 - 50 mixture of cesium - pentadecafluoro - octanate and @xmath9 forms a lyotropic liquid crystal at room temperature @xcite . the schematic energy level diagram of oriented spin-7/2 nucleus is shown in fig .
( [ equi]a ) .
the 8 energy levels are labeled as the basis states of a three qubit system . in the high field approximation , effective quadrupolar coupling ( @xmath8 )
can be considered as a small perturbation to zeeman field .
thus for population purposes the equilibrium density matrix can be considered to be proportional to @xmath10 ( fig . [ equi]a ) .
partially oriented @xmath11 nucleus ( i=7/2 ) gives rise to a well resolved 7 transitions spectrum at room temperatures ranging from 290 k to 315 k fig ( [ equi]b ) . the effective quadrupolar coupling ( @xmath8 ) changes with temperature , since the order parameter is a function of temperature .
all the experiments have been carried out here at temperature 307 k , which gives @xmath12 hz . the equilibrium spectrum is obtained by applying a @xmath13 pulse , is shown in fig.([equi]b ) .
the integrated intensities are in the expected ratio 7:12:15:16:15:12:7 , as determined by the transition matrix elements of @xmath14 ( appendix ) @xcite .
non - adiabatic geometric phase in nmr was first verified by suter et.al @xcite .
non adiabatic geometric phases in nmr , were used to implement controlled phase shift gates , deutsch - jozsa ( dj ) and grover algorithms in weakly j - coupled and strongly dipolar coupled systems @xcite .
a two level subspace ( r , s ) forms a fictitious spin-1/2 subspace , hence the states @xmath15 and @xmath16 can be represented on a bloch sphere ( fig . [ quad - bloch ] ) @xcite .
the two @xmath3 pulses applied on a transition ( r , s ) , cyclically rotates each of the states @xmath15 and @xmath16 @xcite .
the cyclic path followed by each of the two states , makes an equal solid angle ( @xmath17 ) at the center of the bloch sphere @xcite . the geometric phase acquired by the states @xmath15 and @xmath16 is given by @xmath18 and @xmath19 respectively @xcite , where the phase difference of @xmath20 between the states indicate that the two states are traversed in the opposite direction @xcite
it is possible to obtain various solid angles , hence various phases , by shifting the phases of the two @xmath3 pulses with respect to each other ( fig .
[ quad - bloch ] ) @xcite .
it is shown that two @xmath3 pulses with phases @xmath21 and @xmath22 respectively , applied on a transition ( r , s ) , subtends a solid angle @xmath23 , thus the phase acquired by the states @xmath15 and @xmath16 is @xmath24 and @xmath25 respectively .
however the internal hamiltonian evolves during the application of a transition selective @xmath3 pulse , and gives rise to an additional phase known as dynamical phase .
in order to observe only the geometric phase , one has to refocus the evolution of internal hamiltonian . as shown in eq .
( [ 6eqn1 ] ) , the hamiltonian consists of larmor frequency term and quadrupolar coupling term . the larmor frequency is identical to the frequency of central transition d. thus the receiver frequency is set at frequency of transition d ( fig .
[ equi]a ) , hence there is no evolution due to the larmor frequency term .
the duration of the selective @xmath3 pulse is chosen in such a way that the phase acquired by the quadrupolar term makes a complete @xmath26 rotation , in other words @xmath27 , where @xmath28 is duration of the selective @xmath3 pulse which is equal to 1.425 ms .
the phase information can be encoded in the coherences , hence in order to observe the phases the initial state should contain the coherences , which is obtained by applying a non - selective @xmath13 pulse on equilibrium state .
figure ( [ gates]a ) shows the implementation of controlled phase shift gate , represented by the diagonal matrix @xmath29 $ ] , obtained by applying two selective @xmath3 pulses on transition ( 3,4 ) , with phases y and @xmath30 .
the two states @xmath31 and @xmath32 acquire @xmath3 phase shift so the transitions b and d are inverted , confirming the phase gate . since the phases of two @xmath3 pulses are same , it is possible to combine the two @xmath3 pulses in to a single @xmath26 pulse .
such pulses are used in section ( 3.1 ) . in some quantum computing protocols ( for example grover search algorithm and quantum fourier transform )
one requires a phase gate such as @xmath33 ( table .
1 ) , which creates the relative phase @xmath24 at @xmath34 state , for example @xmath35=@xmath36 $ ] .
such phase gates can be implemented by sandwiching various phase gates @xmath37 , as shown in table ( 1 ) .
@xmath37 indicates the geometric phase shift gate obtained by applying a pair of selective @xmath3 pulses on a transition ( i , j ) with phases y and @xmath38 , thus the diagonal elements corresponding to i and j states , acquire a phase shift @xmath24 and @xmath25 respectively .
for example , @xmath39=diag[@xmath40 , @xmath41 , 1 , 1 , 1 , 1 , 1 , 1 ] , which can be obtained by applying two selective @xmath3 pulse on transition ( 1,2 ) with phases y and @xmath42 .
the method of constructing controlled phase shift gates ( table 1 ) , is explained here .
for example , consider a system consisting of n energy levels and ( n-1 ) single quantum transitions between the levels ( 1,2 ) , ( 2,3), ..... (n-1,n ) , as shown in fig.([nlevels ] ) . in this system , the controlled phase shift gate @xmath43 $ ] , can be implemented by sandwiching ( n-1 ) phase shift gates , as , @xmath44 in eq .
( [ eqn5 ] ) , by putting n=8 one can obtain any of the @xmath33 shown in table ( 1 ) . in table
( 1 ) , each of the operators , @xmath33 , requires seven pairs of selective @xmath3 pulses , where each pair of pulses will act on a single transition .
however the @xmath3 pulses on unconnected transitions can be applied simultaneously .
hence one can simultaneously implement @xmath45 , @xmath46 , @xmath47 , @xmath48 by using a pair of multi - frequency ( mf ) pulses , similarly @xmath49 , @xmath50 , @xmath51 can be implemented by using another pair of multi - frequency pulses .
for example @xmath52 can be implemented by using seven geometric phase shift gates ( table 1 ) , obtained by using two pairs of multi - frequency @xmath3 pulse , as described below .
@xmath53 the unitary operators of phase shift gates correspond to diagonal matrices which commute with each other , hence @xmath52 of eq .
( [ 6eqn13 ] ) can be rewritten as , @xmath54.[u_{(3,4)}^{5\phi/8}].[u_{(5,6)}^{3\phi/8}].[u_{(7,8)}^{\phi/8}]\}.\{[u_{(2,3)}^{3\phi/4 } ] .
[ u_{(4,5)}^{\phi/2}].[u_{(6,7)}^{\phi/4}]\ } \nonumber \\ & = & \ { [ ( \pi)_y^a(\pi)_{y+\pi+7\phi/8}^a].[(\pi)_y^c(\pi)_{y+\pi+5\phi/8}^c].[(\pi)_y^e(\pi)_{y+\pi+3\phi/8}^e ] .
[ ( \pi)_y^g(\pi)_{y+\pi+\phi/8}^g]\}. \nonumber \\ & & \ { [ ( \pi)_y^b(\pi)_{y+\pi+3\phi/4}^b].[(\pi)_y^d(\pi)_{y+\pi+\phi/2}^d].[(\pi)_y^f(\pi)_{y+\pi+\phi/4}^f]\ } \nonumber \\ & = & \ { [ ( \pi)^{a , c , e , g}_y.(\pi)^{a , c , e , g}_{(y+\pi+7\phi/8),(y+\pi+5\phi/8),(y+\pi+3\phi/8),(y+\pi+\phi/8 ) } ] \}.\nonumber \\ & & \ { [ ( \pi)^{b , d , f}_y.(\pi)^{b , d , f}_{(y+\pi+3\phi/4),(y+\pi+\phi/2),(y+\pi+\phi/4 ) } ] \ } \label{6eqn5}\end{aligned}\ ] ] the @xmath3 pulses acting on unconnected transitions , a , c , e and g ( eq . [ 6eqn5 ] ) , can be combined ( added ) in to a single pulse , since the operators of the subspaces corresponding to these transitions , @xmath55 , @xmath56 , @xmath57 and @xmath58 @xcite , commute with each other .
similarly one can combine the @xmath59 pulses acting unconnected transitions b , d and f ( eq . [ 6eqn5 ] ) .
figure ( [ gates]b ) shows the spectrum obtained by a @xmath13 pulse followed by the implementation of @xmath60 using four mf-@xmath3 pulses ( eq .
[ 6eqn5 ] ) , the state @xmath61 hence the transition a acquires a @xmath62 phase shift , confirms @xmath63 . table ( 1 ) .
unitary operators of controlled-@xmath64 phase shifted gates + @xmath65=u_{(1,2)}^{7\phi/8}.u_{(2,3)}^{3\phi/4}.u_{(3,4)}^{5\phi/8}.u_{(4,5)}^{\phi/2}. u_{(5,6)}^{3\phi/8}.u_{(6,7)}^{\phi/4}.u_{(7,8)}^{\phi/8}$ ] + @xmath66=u_{(1,2)}^{-\phi/8}.u_{(2,3)}^{3\phi/4}.u_{(3,4)}^{5\phi/8}.u_{(4,5)}^{\phi/2}.u_{(5,6)}^ { 3\phi/8}.u_{(6,7)}^{\phi/4}.u_{(7,8)}^{\phi/8}$ ] + @xmath67=u_{(1,2)}^{-\phi/8}.u_{(2,3)}^{-\phi/4}.u_{(3,4)}^{5\phi/8}.u_{(4,5)}^{\phi/2}.u_{(5,6)}^ { 3\phi/8}.u_{(6,7)}^{\phi/4}.u_{(7,8)}^{\phi/8}$ ] + @xmath68=u_{(1,2)}^{-\phi/8}.u_{(2,3)}^{-\phi/4}.u_{(3,4)}^{-3\phi/8}.u_{(4,5)}^{\phi/2}.u_{(5,6)}^ { 3\phi/8}.u_{(6,7)}^{\phi/4}.u_{(7,8)}^{\phi/8}$ ] + @xmath69=u_{(1,2)}^{-\phi/8}.u_{(2,3)}^{-\phi/4}.u_{(3,4)}^{-3\phi/8}.u_{(4,5)}^{\phi/2}.u_{(5,6)}^ { 3\phi/8}.u_{(6,7)}^{\phi/4}.u_{(7,8)}^{\phi/8}$ ] + @xmath70=u_{(1,2)}^{-\phi/8}.u_{(2,3)}^{-\phi/4}.u_{(3,4)}^{-3\phi/8}.u_{(4,5)}^{-\phi/4}.u_{(5,6)}^ { -5\phi/8}.u_{(6,7)}^{\phi/4}.u_{(7,8)}^{\phi/8}$ ] + @xmath71=u_{(1,2)}^{-\phi/8}.u_{(2,3)}^{-\phi/4}.u_{(3,4)}^{-3\phi/8}.u_{(4,5)}^{-\phi/2}.u_{(5,6)}^ { -5\phi/8}. u_{(6,7)}^{-6\phi/8}.u_{(7,8)}^{\phi/8}$ ] + @xmath72=u_{(1,2)}^{-\phi/8}.u_{(2,3)}^{-\phi/4}.u_{(3,4)}^{-3\phi/8}.u_{(4,5)}^{-\phi/2}.u_{(5,6)}^ { -5\phi/8}.u_{(6,7)}^{-3\phi/4}.u_{(7,8)}^{-7\phi/8}$ ] +
deutsch - jozsa ( dj ) algorithm determines in a single query , whether a given function is constant or balanced @xcite .
a function is `` constant '' if it gives the same output for all inputs , and `` balanced '' if it gives one output for half the number of inputs and another for the remaining half @xcite .
classically for an n bit binary function , at least ( @xmath73 ) queries are needed to determine whether the function is constant or balanced , whereas the dj algorithm requires only a single query @xcite .
the cleve version of the dj algorithm requires an extra qubit ( ancilla qubit ) and uses controlled - not gates @xcite , whereas collins version does not require the ancilla qubit but needs controlled phase shift gates @xcite . for a three qubit system
there are 2 constant and 70 ( @xmath74=@xmath75 ) balanced functions @xcite .
the quantum circuit of the collins version of the dj algorithm is shown in fig .
( [ djckt ] ) .
the algorithm starts with a pure state ( pseudo pure state in nmr ) @xmath61 which is converted to a superposition state of eight basis states @xmath61 , @xmath76 , ..... , @xmath77 , by applying a pseudo hadamard gate on all the three qubits .
thereafter a unitary operator @xmath78 ( oracle ) is applied , followed by detection . theoretically , after @xmath78 one has to apply the hadamard gate . in nmr ,
a hadamard gate is replaced by a pseudo hadamard gate , which is implemented by a @xmath79 pulse , and for detection another @xmath80 pulse is needed .
these two pulses cancel each other and the result of the algorithm is available immediately after @xmath78 .
the unitary operators ( u ) of the oracle , are eight dimensional diagonal matrices .
for two constant functions , u s are given by , @xmath81=(-1).i .
\end{aligned}\ ] ] for balanced functions , we define the unitary operator as @xmath82 , which means that each of the four diagonal elements 1,k , l , and m are equal to -1 , whereas the remaining four diagonal elements are 1 .
for example @xmath83 $ ] .
the unitary operators of 35 balanced functions can be written as , @xmath84.i , \nonumber \\ where \hspace{0.2 cm } k < l < m , \hspace{0.2 cm } k=2,3 .. ,6 , \hspace{0.2 cm } l \ge k+1 , \hspace{0.2 cm } and \hspace{0.2 cm } m \ge l+1 \label{6eqn6}\end{aligned}\ ] ] the unitary operators of the remaining 35 balanced functions are identical to one of the operators of eq .
( [ 6eqn6 ] ) , up to an overall phase factor @xmath20 @xcite . here
we implement the unitary operators corresponding to 11 of the 35 balanced functions ( eq . [ 6eqn6 ] ) and one constant function ( @xmath85 ) . *
( i)preparation of pseudo pure state ( pps ) @xmath86 : * a multi frequency pulse with six harmonics with appropriate amplitudes was applied .
the six harmonics are the frequencies of the six leftmost transitions b , c , ... ,
( [ equi]b ) . under the influence of this pulse ,
the populations of all the states except @xmath87 , start a collective population transfer among them as a linear chain of coupled oscillators @xcite .
choosing the correct amplitudes of various harmonics and duration of the pulse , these six transitions can be simultaneously saturated with average population of these seven levels .
the relative amplitudes of the six harmonics b , c , ... , and g are respectively given by 0.84 , 0.93 , 1 , 1.03 , 1.04 and 1.07 @xcite .
the duration of the pulse is 2.05 ms .
a gradient pulse was applied subsequently to kill ( dephase ) the coherences created during the process .
the final populations were measured using a non - selective small - angle ( 5@xmath88 ) pulse .
the small - angle pulse , converts the population differences between adjacent energy levels into single quantum transitions , within linear approximation .
the spectrum of fig ( [ pps]b ) confirms the preparation of @xmath87 pseudopure state . *
( iii ) coherent superposition : * after the creation of @xmath86 pps , a pseudo hadamard gate ( hard @xmath79 pulse ) creates a superposition of 8 basis states .
it may be noted that unlike in weakly coupled spins , the state created here is not in equal superposition of eigenstates , since the coefficients of various eigenstates are different .
however , as is shown here , the created coherent superposition can be utilized for quantum parallelism , to distinguish different classes of functions .
the state of the system after a @xmath79 pulse on @xmath86 pps , is given by , @xmath89 \nonumber \nonumber \\ & = & \frac{1}{8\sqrt{2}}[\vert
1 \rangle + \sqrt{7 } \vert 2 \rangle + \sqrt{21 } \vert 3 \rangle + \sqrt{35 } \vert 4 \rangle + \sqrt{35 } \vert 5 \rangle + \sqrt{21 } \vert 6 \rangle + \sqrt{7 } \vert 7 \rangle + \vert 8 \rangle ] .
\label{6eqn7}\end{aligned}\ ] ] the corresponding density matrix can be written as , @xmath90 where the elements within the boxes represent single quantum transitions a , b , .. , g. figure ( [ pps]c ) shows the spectrum of @xmath91 , where the intensities of transitions are obtained by modulus of the product of single quantum elements of eq .
[ 6eqn8 ] , with corresponding matrix elements of @xmath92 . *
( iv ) implementation of u : * for constant function @xmath93 ( unit matrix ) , which requires no pulse , thus the final state @xmath94 is given by , @xmath95 as mentioned in section ( 2 ) , the controlled phase shift gate @xmath96 can be achieved by applying two @xmath59 pulses with same phase on transition ( i , j ) , which can be be combined in to a single @xmath97 pulse . for balance functions , u = @xmath82 ( eq .
[ 6eqn6 ] ) , can be decomposed in to , @xmath98[u_{(l , l+1)}^{\pi}.u_{(l+1,l+2)}^{\pi} ...
u_{(m-1,m)}^{\pi } ] \label{6eqn11}\end{aligned}\ ] ] in eq .
( [ 6eqn11 ] ) , @xmath99 and @xmath100 are implemented by applying two @xmath101 pulses on transitions ( 1,k ) and ( l , m ) respectively , if ( 1,k ) and ( l , m ) does not correspond to single quantum transitions , then they can be decomposed in to a series of single quantum transitions , as shown in the second equality of eq .
( [ 6eqn11 ] ) , for example , @xmath102.[u_{(5,6)}^{\pi}. u_{(6,7)}^{\pi}.u_{(7,8)}^{\pi } ] \nonumber \\ & = & [ ( 2\pi)^a.(2\pi)^b.(2\pi)^c].[(2\pi)^e.(2\pi)^f.(2\pi)^g ] \nonumber \\ & = & ( 2\pi)^{a , c , e , g}.(2\pi)^{b , f } , \label{6eqn12}\end{aligned}\ ] ] where in the last equality of eq .
( [ 6eqn12 ] ) , the pulses acting on unconnected transitions ( fig .
[ equi ] ) are combined in to a single mf pulse .
the required pulses for various @xmath82 of eq .
( [ 6eqn6 ] ) are given in table ( 2 ) , the @xmath26 pulses acting on unconnected transitions , are combined in to a single multi - frequency ( mf ) @xmath26 pulse .
the duration of each transition selective pulse , is set such that the evolution due to quadrupolar coupling , makes a complete @xmath26 rotation .
it is to be noted that , in eq .
( [ 6eqn12 ] ) and table ( 2 ) , one can use any phase for ( @xmath26 ) pulses , we have used y phase for all the pulses .
the final state @xmath103 is given by @xmath104 it can be seen that @xmath103 is same as @xmath91 , except that the basis states @xmath105 , @xmath106 , @xmath107 and @xmath108 acquire a phase factor @xmath20 ( = -1 ) .
table ( 2 ) .
unitary operators of balanced functions ( eq . [ 6eqn6 ] ) and corresponding pulse sequences , consisting of @xmath26 pulses , where all the pulses are applied with phase y. sequence & u(1,k , l , m ) & pulse sequence + u(1,2,3,4 ) & @xmath109 & u(1,2,3,5 ) & @xmath109-@xmath110 & u(1,2,3,6 ) & @xmath111-@xmath110 + u(1,2,3,7 ) & @xmath111- @xmath112 & u(1,2,3,8 ) & @xmath113-@xmath112 & u(1,2,4,5 ) & @xmath114 + u(1,2,4,6 ) & @xmath114-@xmath115 & u(1,2,4,7 ) & @xmath116-@xmath115 & u(1,2,4,8 ) & @xmath116-@xmath117 + u(1,2,5,6 ) & @xmath118 & u(1,2,5,7 ) & @xmath118-@xmath119 & u(1,2,5,8 ) & @xmath120-@xmath119 + u(1,2,6,7 ) & @xmath121 & u(1,2,6,8 ) & @xmath121-@xmath122 & u(1,2,7,8 ) & @xmath123 + u(1,3,4,5 ) & @xmath114-@xmath124 & u(1,3,4,6 ) & @xmath114-@xmath125 & u(1,3,4,7 ) & @xmath116-@xmath125 + u(1,3,4,8 ) & @xmath116-@xmath126 & u(1,3,5,6 ) & @xmath118-@xmath124 & u(1,3,5,7 ) & @xmath118-@xmath127 + u(1,3,5,8 ) & @xmath120-@xmath127 & u(1,3,6,7 ) & @xmath121-@xmath124 & u(1,3,6,8 ) & @xmath121-@xmath128 + u(1,3,7,8 ) & @xmath123-@xmath124 & u(1,4,5,6 ) & @xmath111-@xmath124 & u(1,4,5,7 ) & @xmath111-@xmath127 + u(1,4,5,8 ) & @xmath113-@xmath127 & u(1,4,6,7 ) & @xmath129-@xmath124 & u(1,4,6,8 ) & @xmath129- @xmath128 + u(1,4,7,8 ) & @xmath130-@xmath124 & u(1,5,6,7 ) & @xmath129-@xmath131 & u(1,5,6,8 ) & @xmath129- @xmath132 + u(1,5,7,8 ) & @xmath130-@xmath131 & u(1,6,7,8 ) & @xmath113-@xmath131 & & + * ( v ) detection : * the single quantum transitions ( a , b , ..
g ) of the final states ( eq . [ 6eqn9],[6eqn10 ] ) are detected .
figure ( [ dj1 ] ) shows the spectrum of @xmath133 and various @xmath103 .
from the shape of the spectrum one can conclude , whether the final state represents a constant ( or ) a balanced function . for constant function ( @xmath133 ) none of the peaks are inverted , whereas for balanced functions ( @xmath134 ) atleast one of the peaks is inverted .
furthermore the phases of transitions confirm the final state @xmath103 .
for example in the spectrum of @xmath135 , transition d is negative since the phase difference between the states @xmath136 and @xmath137 is @xmath20 , while all others have zero phase difference , similarly one can confirm the other states .
demonstration of quantum computing protocols on various nmr systems , is a promising research area for increasing number of qubits . in this work
controlled phase shift gates are implemented on a oriented spin-7/2 nucleus , using non - adiabatic geometric phases obtained by using selective pulses on single quantum transitions .
it is also demonstrated here that in an n level system , one can implement a controlled phase shift gate by sandwiching various geometric phase shift gates , this method can also be applied to weakly as well as strongly coupled spin-1/2 systems .
the number of selective pulses are reduced by using multi frequency ( mf ) pulses .
collins version of 3-qubit dj algorithm is implemented , where the eight basis states are collectively treated as a three qubit system .
the required controlled phase shift gates of the algorithm , are implemented by using mf-@xmath97 pulses .
the angular momentum operators of spin-7/2 nucleus , are given by @xcite , p. benioff , _ quantum mechanical models of turing machines that dissipate no energy _
lett . _ * 48 * , 1581 ( 1982 ) .
d. deutsch , _ proc .
london , ser .
a _ * 400 * , 97 ( 1985 ) .
d. deutsch and r. jozsa , _ rapid solution of problems by quantum computation _ , _ proc .
a _ * 439 * , 553 ( 1992 ) .
p. w. shor , _ polynominal - time algorithms for prime factorization and discrete algorithms on quantum computer _ _ siam rev_. * 41 * , 303 - 332 ( 1999 ) .
grover , _ quantum mechanics helps in searching for a needle in haystack _ _ phys .
lett . _ * 79 * , 325 ( 1997 ) .
nielsen and i.l .
chuang , `` quantum computation and quantum information '' .
cambridge university press , cambridge , u.k . 2000 .
m. scholz , t. aichele , s. ramelow , and o. benson , _ deutsch - jozsa algorithm using triggered single photons from a single quantum dot _ _ phys .
lett . _ * 96 * , 180501 ( 2006 ) .
cory , a. f. fahmy , and t. f. havel , _ ensemble quantum computing by nmr spectroscopy _ , _ proc .
usa _ * 94 * , 1634 ( 1997 ) .
n. gershenfield and i. l. chuang , _ bulk spin - resonance quantum computation _ , _ science _ * 275 * , 350 ( 1997 ) .
i. l. chuang , l. m. k. vanderspyen , x. zhou , d. w. leung , and s. llyod , _ nature ( london ) _ , * 393 * , 1443 ( 1998 ) . j.a .
jones and m. mosca , _ j. chem .
phys . _ * 109 * , 1648 ( 1998 ) .
i.l . chuang , n. gershenfeld , m. kubinec , phys .
* 80 * , 3408 ( 1998 ) .
t. s. mahesh , kavita dorai , arvind , anil kumar , _ j. mag .
_ * 148 * , 95 ( 2001 ) .
b. m. fung , _ use of pairs of pseudopure states for nmr quantum computing _
a * 63 * , 022304 ( 2001 ) .
a. k. khitrin and b. m. fung , _ nuclear magnetic resonance quantum logic gates using quadrupolar nuclei _ , j. chem . phys . * 112 * , 6963 ( 2000 ) . t. s. mahesh , neeraj sinha , k. v. ramanathan , and anil kumar , _ ensemble quantum - information processing by nmr : implementation of gates and the creation of pseudopure states using dipolar coupled spins as qubits _ , phys .
a * 65 * , 022312 ( 2002 ) .
neeraj sinha , t. s. mahesh , k.v .
ramanathan , and anil kumar , _ toward quantum information processing by nuclear magnetic resonance : pseudopure states and logical operations using selective pulses on an oriented spin 3/2 nucleus _ , j. chem .
phys . * 114 * , 4415 ( 2001 ) .
a. khitrin , h. sun , and b. m. fung , _ method of multifrequency excitation for creating pseudopure states for nmr quantum computing _ , phys
a * 63 * , 020301 ( 2001 ) . v. l. ermakov and b. m. fung , _ experimental realization of a continuous version of the grover algorithm _ , phys .
a * 66 * , 042310 ( 2002 ) .
a. k. khitrin and b. m. fung , _ nmr simulation of an eight - state quantum system _ , phys .
a * 64 * , 032306 ( 2001 ) .
murali , neeraj sinha , t.s .
mahesh , malcom levitt , k.v .
ramanathan , and anil kumar , _ quantum - information processing by nuclear magnetic resonance : experimental implementation of half - adder and subtractor operations using an oriented spin-7/2 system _ , phys .
a * 66 * , 022313 ( 2002 ) .
anil kumar , k.v .
ramanathan , t.s .
mahesh , neeraj sinha and k.v.r.m .
murali , _ developments in quantum information processing by nmt : use of quadrupolar and dipolar couplings _ , pramana - journal of physics , * 59 ( 2 ) * 243 - 254 ( 2002 ) . ranabir das and anil kumar , _ use of quadrupolar nuclei for quantum - information processing by nuclear magnetic resonance : implementation of a quantum algorithm _ ,
a * 68 * , 032304 ( 2003 ) .
sarthour , e.r .
deazevedo , f.a .
bonk , e. l. g. vidoto , t. j. bonagamba , a. p. guimaraes , j.c.c .
freitas , i.s .
oliveira , _ relaxation of coherent states in a two - qubit nmr quadrupole system _ , phys .
a * 68 * , 022311 ( 2003 ) .
bonk , r.s .
sarthour , e.r .
deazevedo , j.d .
bulnes , g. l. mantovani , j.c.c .
freitas , t.j .
bonagamba , a.p .
guimaraes , i.s .
oliveira , _ quantum - state tomography for quadrupole nuclei and its application on a two - qubit system _ phys .
a * 69 * , 042322 ( 2004 ) .
bonk , e.r .
deazevedo , r.s .
sarthour , j.d .
bulnes , j.c.c .
freitas , a.p .
guimaraes , i.s .
oliveira , t.j .
bonagamba , _ quantum logical operations for spin 3/2 quadrupolar nuclei monitored by quantum state tomography _ , j. magn
* 175 * , 226 ( 2005 ) .
ranabir das and anil kumar , _ experimental implementation of a quantum algorithm in a multiqubit nmr system formed by an oriented spin-7/2 system _ , appl .
89 * , 024107 ( 2006 ) .
h. kampermann and w. s. veeman , _ characterization of quantum algorithms by quantum process tomography using quadrupolar spins in solid - state nuclear magnetic resonance _ , j. chem
. phys . * 122 * , 214108 ( 2005 ) .
r. r. ernst , g. bodenhausen , and a. wokaun , _ principles of nuclear magnetic resonance in one and two dimensions , oxford university press _ 1987 .
p. dhiel and c. l. khetrapal _ nmr - basic principles and progress ( springer - verlag ) , new york , 1969 , vol .
1_. y. aharnov and j. anandhan , _ phase change during a cyclic quantum evolution _ , phys
* 58 * , 1593 ( 1987 ) .
d. suter , k. t. mueller and a. pines , _ aharnov - anandhan quantum phase by nmr interferometry _ , phys.rev.lett . * 60 * , 1218 ( 1988 ) .
ranabir das , s. k. karthick kumar , and anil kumar , _ use of non - adiabatic geometric phase for quantum computing by nuclear magnetic resonance _ , j. magn .
reson . * 177 * , 318 ( 2005 ) .
t. gopinath and anil kumar , _ geometric quantum computation using fictitious spin-(1/2 ) subspaces of strongly dipolar coupled nuclear spins _ , phys .
a * 73 * , 022326 ( 2006 ) .
r. cleve , a. ekert , c. macchiavello , and m. mosca , _ quantum algorithms revisited _ , proc .
a * 454 * , 339 ( 1998 ) .
david collins , k. w. kim , and w. c. holton , _ deutsch - jozsa algorithm as a test of quantum computation _ , phys .
a * 58 * , 1633 ( 1998 ) .
oliver mangold , andreas heidebrecht , and michael mehring , _ nmr tomography of the three - qubit deutsch - jozsa algorithm _ , phys .
a * 70 * , 042307 ( 2004 ) . | in this work controlled phase shift gates are implemented on a qaudrupolar system , by using non - adiabatic geometric phases .
a general procedure is given , for implementing controlled phase shift gates in an n level system .
the utility of such controlled phase shift gates , is demonstrated here by implementing 3-qubit deutsch - jozsa algorithm on a 7/2 quadrupolar nucleus oriented in a liquid crystal matrix . . |
till the end of august 1999 , x - ray , optical , and radio afterglows have been observed from about 16 , 11 , and 5 gamma - ray bursts ( grbs ) respectively ( costa et al .
1997 ; bloom et al .
1998 ; groot et al .
1998 ; kulkarni et al . 1998 , 1999 ; harrison et al . 1999 ; stanek et al . 1999 ; fruchter et al .
1999 ; galama et al .
. the so called fireball model ( goodman 1986 ; paczyski 1986 ; mszros & rees 1992 ; rees & mszros 1992 , 1994 ; katz 1994 ; sari , narayan , & piran 1996 ; dermer & mitman 1999 ; dermer , bttcher , & chiang 1999 , 2000 ; dermer 2000 ) is strongly favored , which is found successful at explaining the major features of grb afterglows ( mszros & rees 1997 ; vietri 1997 ; tavani 1997 ; waxman 1997a ; wijers , rees , & mszros 1997 ; sari 1997a ; huang et al .
1998a , b , 1999a , b ; dai & lu 1998a , b , c ; dermer , chiang , & bttcher 1999 ; dermer , chiang , & mitman 2000 )
. however , we are still far from resolving the puzzle of grbs , because their `` inner engines '' are well hidden from direct afterglow observations . to unveil the nature of the `` inner engine '' , we first need to know the energetics involved in a typical burst , which itself depends on two factors : ( i ) the distance scale of grbs , this has been settled since the bepposax discovery of grb 970228 ; ( ii ) the opening angle of grb ejecta , i.e. , whether gamma - rays are radiated isotropically or not , this question is still largely uncertain . most grbs localized by bepposax have indicated isotropic energies of @xmath5
@xmath6 ergs , well within the energy output from compact stellar objects of solar - mass . however , grb 971214 , 980703 , 990123 , and 990510 have implied isotropic gamma - ray releases of @xmath7 ergs ( 0.17 @xmath8 , kulkarni et al .
1998 ) , @xmath9 ergs ( 0.06 @xmath8 , bloom et al . 1998 ) , @xmath10 ergs ( 1.9 @xmath8 , kulkarni et al . 1999 ; andersen et al . 1999 ) , and @xmath11 ergs ( 0.16 @xmath8 , harrison et al . 1999 ) respectively . moreover , if really located at a redshift of @xmath12 as suggested by reichart et al .
( 1999 ) , grb 980329 would imply an isotropic gamma - ray energy of @xmath13 ergs ( 2.79 @xmath8 ) .
such enormous energetics has forced some theorists to deduce that grb radiation must be highly collimated in these cases , with half opening angle @xmath14 , so that the intrinsic gamma - ray energy could be reduced by a factor of @xmath15 @xmath16 , and could still come from compact stellar objects ( pugliese , falcke , & biermann 1999 ) . obviously , whether grbs are beamed or not is of fundamental importance to our understanding of their nature .
how can we tell a jet from an isotropic fireball ?
gruzinov ( 1999 ) has argued that optical afterglows from a jet can be strongly polarized , in principle up to tens of percents , if co - moving magnetic fields parallel and perpendicular to the jet have different strengths and if we observe at the right time from the right viewing angle .
more direct clues may come from afterglow light curves .
rhoads ( 1997 , 1999a , b ) has shown that the lateral expansion ( at sound speed ) of a relativistic jet ( with a lorentz factor @xmath17 ) will cause the blastwave to decelerate more quickly , leading to a sharp break in the afterglow light curve .
the power law decay indices of afterglows from grb 980326 and 980519 are anomalously large , @xmath18 ( groot et al .
1998 ; owens et al . 1998 ) , and optical light curves of grb 990123 and 990510 even show obvious steepening at observing time @xmath19 2 d ( kulkarni et al . 1999 ; harrison et al . 1999 ; castro - tirado et al . 1999 ) .
recently grb 970228 was also reported to have a large index of @xmath20 ( galama et al .
these phenomena have been widely regarded as the evidence of the presence of relativistic jets ( sari , piran , & halpern 1999 ; castro - tirado et al .
1999 ) .
however , numerical studies of some other authors ( panaitescu & mszros 1998 ; moderski , sikora , & bulik 2000 ) have shown that due to the increased swept - up matter and the time delay of the large angle emission , the sideway expansion of the jet does not lead to an obvious dimming of the afterglow .
thus there are two opposite conclusions about the jet effect : the analytical solution predicts a sharp break , while the numerical calculation shows no such sharp breaks .
it is very interesting to note that a recent analytic treatment by wei & lu ( 2000 ) shows that the sharp break predicted by rhoads is usually not exist unless the beaming angle is very small .
this analysis seems to have given some supports to the numerical studies .
we find that previous studies on jet effects need to be improved urgently in the following three aspects : ( i ) afterglows during the mildly relativistic ( @xmath21 ) and non - relativistic ( @xmath22 ) phases of the expansion are of great importance to us , since they may correspond to observing time of @xmath23 2 5 d ( huang et al .
1998a , b ) . however , conventional dynamical model can not transit correctly from the ultra - relativistic phase ( @xmath24 ) to the non - relativistic phase for adiabatic shocks .
this has been stressed by huang , dai , & lu ( 1999 , 2000 ) .
therefore previous numerical studies were based on improper dynamical equations .
they could describe an ultra - relativistic jet , but they gave spurious results in the mildly relativistic and non - relativistic phases ( huang , dai , & lu 1999a , b ) ; ( ii ) it is reasonable to assume that the lateral expansion speed of the beamed ejecta is just the co - moving sound speed @xmath25 .
usually we take @xmath26 , where @xmath27 is the speed of light ( rhoads 1997 , 1999a ) .
however in realistic case we expect @xmath25 to vary with time , and especially it will by no means be @xmath28 when @xmath29 .
this is another reason that previous studies are not proper for mildly relativistic and non - relativistic jets ; ( iii ) in previous studies , the expansion of the beamed ejecta is supposed to be solely adiabatic . in fact the blastwave should be highly radiative at first .
it evolves slowly into an adiabatic blastwave at later stages ( dai , huang , & lu 1999 ) .
this paper is aiming to overcome the problems mentioned just above .
we present our refined dynamical model in section 2 .
it can describe the overall evolution of a realistic , jetted grb ejecta .
synchrotron radiation from shock accelerated electrons is formulated in section 3 .
our detailed numerical results are presented in section 4 .
section 5 is a brief discussion .
let @xmath30 be the radial coordinate in the burster frame ; @xmath31 , and @xmath32 be the time from the event measured in the burster frame , co - moving ejecta frame , and terrestrial observer s frame respectively ; @xmath33 and @xmath34 be the initial lorentz factor and ejecta mass and @xmath35 the half opening angle of the ejecta .
the burst energy is @xmath36 .
it is useful to write down the following simple relations at first : @xmath37 @xmath38 where @xmath39 .
the evolution of radius ( @xmath30 ) and swept - up mass ( @xmath40 ) is described by @xmath41 @xmath42 where @xmath43 is number density of surrounding interstellar medium ( ism ) and @xmath44 is mass of proton .
the opening angle @xmath35 is usually evaluated as @xmath45 , where @xmath46 is co - moving lateral radius of the ejecta ( rhoads 1999a ; moderski , sikora , & bulik 2000 ) .
this is not ideal in numerical evaluations since @xmath35 may be as large as 0.5 0.8 at later times .
here we use the following differential equation : @xmath47 as for the evolution of the lorentz factor @xmath1 , the following equation has been widely used for both isotropic fireballs and beamed ejecta : @xmath48 where @xmath49 is the total mass in the co - moving frame , including internal energy @xmath50 , @xmath51 dm$ ] ( chiang & dermer 1999 ; piran 1999 ; moderski , sikora , & bulik et al .
2000 ) . here
@xmath52 is the radiative efficiency defined as the fraction of the shock - generated thermal energy ( in the co - moving frame ) that is radiated .
@xmath53 corresponds to highly radiative case and @xmath54 corresponds to adiabatic expansion .
however huang , dai , & lu ( 1999a , b ) have demonstrated clearly that for adiabatic expansion equation ( [ dgdmold6 ] ) does not agree with the sedov solution in the non - relativistic limit : equation ( [ dgdmold6 ] ) gives @xmath55 while the sedov solution requires @xmath56 ( sedov 1969 ) .
they have proposed a new equation .
now we briefly repeat their derivation . in the observer s frame ,
since the total kinetic energy of the fireball is @xmath57 ( panaitescu , mszros , & rees 1998 ) , and the radiated thermal energy is @xmath58 ( blandford & mckee 1976 ) , we have @xmath59 = - \epsilon \gamma ( \gamma - 1 ) dm c^2.\ ] ] for the item @xmath50 , it is usually assumed that @xmath60 ( panaitescu , mszros , & rees 1998 ) .
equation ( [ dgdmold6 ] ) has been derived just in this way .
huang , dai , & lu ( 1999a , b ) suggested that @xmath61 is a better approximation that makes equation ( [ dgu7 ] ) suitable for both ultra - relativistic and newtonian shocks . using this expression ,
it is easy to obtain ( huang , dai , & lu 1999a , b ) @xmath62 huang , dai , & lu ( 1999a , b ) have shown that this equation is acceptable for both radiative and adiabatic fireballs , and in both ultra - relativistic and non - relativistic phases . here
we use this equation to follow the overall evolution of beamed grb ejecta .
equations ( [ drdt3 ] ) , ( [ dmdr4 ] ) , ( [ dthetadt5 ] ) , and ( [ dgdmass8 ] ) present a thorough description of jet evolution . but before evaluating them numerically , we should give the expression for @xmath25 and @xmath52 .
the lateral expansion is determined by the co - moving sound speed .
the simple assumption of @xmath26 is not proper for our purpose in this paper .
we must derive @xmath25 from @xmath63 , where @xmath64 and @xmath65 are co - moving pressure and energy density respectively .
kirk & duffy ( 1999 ) have derived @xmath66 c^2 , \ ] ] where @xmath67 is co - moving mass density , and @xmath68 is the adiabatic index .
dai , huang , & lu ( 1999 ) gave a simple and useful expression for @xmath68 , @xmath69 .
since @xmath70 and @xmath71 , it is easy to get @xmath72 in the ultra - relativistic limit ( @xmath73 ) , equation ( [ cssquare10 ] ) gives @xmath74 ; and in the non - relativistic limit ( @xmath75 ) , we simply get @xmath76 .
so equation ( [ cssquare10 ] ) is a reasonable expression and will be used in our model . as usual
we assume that the magnetic energy density in the co - moving frame is a fraction @xmath77 of the total thermal energy density ( dai , huang , & lu 1999 ) @xmath78 and that the shock accelerated electrons carry a fraction @xmath79 of the proton energy . here
@xmath79 is a parameter characterizing the efficiency of energy transport from protons to electrons . in realistic case , it is probable that @xmath79 may vary with time , but little is known about the detailed mechanism .
so we will take @xmath79 as a constant throughout this article .
this implies that the minimum lorentz factor of the random motion of electrons in the co - moving frame is @xmath80 where @xmath81 is electron mass and @xmath82 is the index characterizing the power law energy distribution of electrons .
we here consider only synchrotron emission from these electrons , and neglect the contribution of inverse compton emission because the latter is unimportant particularly at late times ( waxman 1997a ) .
the energy of a typical electron is lost due to both synchrotron radiation and expansion of the ejecta , thus the radiative efficiency of this single electron is given by @xmath83 ( dai , huang , & lu 1999 ) , where @xmath84 is the synchrotron cooling time , @xmath85 , with @xmath86 the thompson cross section , and @xmath87 is the co - moving frame expansion time , @xmath88 .
since all the shock accelerated electrons carry only a fraction @xmath79 of the internal energy , the radiative efficiency of the total ejecta can be given by ( dai , huang , & lu 1999 ) @xmath89 for the highly radiative expansion , @xmath90 and @xmath91 , we have @xmath92
. the early evolution of the ejecta is likely in this regime .
for an adiabatic expansion , @xmath93 or @xmath94 , we get @xmath95 .
the late evolution is believed to be in this regime .
so one expects that in realistic case the radiative efficiency evolves from about 1 to 0 . in this paper
we call the jet whose radiative efficiency evolves according to equation ( [ radeps13 ] ) a `` realistic '' one ( dai , huang , & lu 1999 ) .
in the absence of radiation loss , the distribution of the shock accelerated electrons behind the blastwave is usually assumed to be a power law function of electron energy , @xmath96 where @xmath97 is the maximum lorentz factor , @xmath98 ( dai , huang , & lu 1999 ) , and @xmath82 usually varies between 2 and 3 .
however , radiation loss may play an important role in the process .
electrons with different lorentz factors have different radiation efficiencies .
sari , piran , & narayan ( 1998 ) have derived an equation for the critical electron lorentz factor , @xmath99 , above which synchrotron radiation is significant , @xmath100 electrons with lorentz factors below @xmath99 are adiabatic ones , and electrons above @xmath99 are highly radiative . in the presence of steady injection of electrons accelerated by the shock ,
the distribution of radiative electrons becomes another power law function with an index of @xmath101 ( rybicki & lightman 1979 ) , but the distribution of adiabatic electrons is unchanged .
then the actual distribution should be given according to the following cases ( dai , huang , & lu 1999 ) : \(i ) for @xmath102 , @xmath103 @xmath104 where @xmath105 is the total number of radiating electrons involved .
\(ii ) for @xmath106 , @xmath107 where @xmath108 @xmath109^{-1}n_{\rm ele } .\ ] ] \(iii ) if @xmath110 , then @xmath111 where @xmath112 in the co - moving frame , synchrotron radiation power at frequency @xmath113 from electrons is given by ( rybicki & lightman 1979 ) @xmath114 where @xmath115 is electron charge , @xmath116 , and @xmath117 with @xmath118 being the bessel function .
we assume that this power is radiated isotropically , @xmath119 let @xmath120 be the angle between the velocity of emitting material and the line of sight and define @xmath121 , we can derive the angular distribution of power in the observer s frame ( rybicki & lightman 1979 ) , @xmath122 @xmath123 then the observed flux density at frequency @xmath124 is @xmath125 where @xmath126 is the area of our detector and @xmath127 is the luminosity distance .
photons received by the detector at a particular time @xmath32 are not emitted simultaneously in the burster frame .
this effect has been emphasized by a number of authors ( waxman 1997b ; sari 1997b ; panaitescu & mszros 1998 ) and may be of great importance to jet radiation ( moderski , sikora , & bulik 2000 ; panaitescu & mszros 1999 ) . in order to calculate observed flux densities , we should integrate over the equal arrival time surface determined by @xmath128 within the jet boundaries .
for convenience , let us define the following initial values or parameters as a set of `` standard '' parameters : initial energy per solid angle @xmath129 ergs/@xmath130 , @xmath131 ( i.e. , initial ejecta mass per solid angle @xmath132 ) , @xmath133 @xmath134 , @xmath135 , @xmath136 , @xmath137 kpc , @xmath138 , @xmath139 . for simplicity , we first assume that the expansion is completely adiabatic all the time ( i.e. @xmath140 , we call it an `` ideal '' jet , distinguishing it from the `` realistic '' jet defined in section 2.3 ) . figure 1 shows the evolution of the lorentz factor .
we see that the ultra - relativistic phase lasts only for @xmath141 s , this is the period during which equation ( [ dgdmold6 ] ) can be safely applied .
the mildly relativistic phase lasts from @xmath141 s to @xmath142 s. in short , the ejecta will cease to be highly relativistic at time @xmath143 @xmath144 s. this clearly demonstrates the necessity to replace equation ( [ dgdmold6 ] ) with our improved expression of equation ( [ dgdmass8 ] ) .
figure 1 also indicates that the transition from the relativistic phase to the non - relativistic phase occurs roughly at @xmath145 @xmath146 s. figure 2 illustrates the time dependence of shock radius . in figure 3 we present the evolution of the jet opening angle @xmath35 . during the ultra - relativistic phase ( @xmath147 s )
, @xmath35 increases only slightly .
but at the newtonian stage ( @xmath148 s ) , the increase of @xmath35 is very quick .
again this figure indicates that the critical point is roughly at @xmath145 s. the usual approximation of @xmath149 is not a good approach and our equation ( [ dthetadt5 ] ) is obviously more reasonable .
figure 4 illustrates jet evolution on the @xmath150
@xmath151 plane schematically , where @xmath151-axis is just the symmetry axis of the jet and the lateral expansion is _ approximately _ at @xmath150 direction .
note that the jet is non - relativistic when @xmath152 cm .
again we see the quick lateral expansion at the newtonian stage . in figure 5
we show some examples of equal arrival time surfaces on the @xmath150
@xmath151 plane . here
we assume that the angle between jet symmetry axis and the line of sight is @xmath153 .
figures 6 and 7 illustrate the time dependence of radiative efficiency ( @xmath52 ) and total kinetic energy ( @xmath154 ) for `` realistic '' jets respectively .
although the highly radiative phase usually lasts for very short period ( @xmath155 @xmath156 s ) , it dissipates the total kinetic energy substantially ( up to 90% in less than @xmath16 s ) .
this effect may account for the confusing phenomena observed in some grbs : @xmath1-ray energy released in the initial grb phase ( @xmath157 ) is comparable to or even higher than the total kinetic energy inferred from afterglow observations ( wijers & galama 1999 ) .
figure 8 shows the optical light curves , computed for r band observation . in this figure
, the thick solid line corresponds to an `` ideal '' jet with `` standard '' parameters , and viewing angle @xmath153 . for comparison ,
other lines are drawn with only one parameter altered or only one condition changed .
please note that at end point of each curve , the average electron lorentz factor is already as small as @xmath158 , corresponding to bulk lorentz factor of @xmath159 , thus completely in the newtonian regime .
> from figure 8 , we see that in no case could we observe the theoretically predicted light curve steepening during the _ relativistic _ stage itself ( i.e. , when @xmath160 s ) , consistent with previous numerical studies ( panaitescu & mszros 1999 ; moderski , sikora , & bulik 2000 ) . this has been attributed to the effects of equal arrival time surfaces and more and more swept - up material . here
we would like to propose another more reasonable explanation .
rhoads is correct in saying that the lateral expansion begins to take effect when @xmath161 . in our calculations
this occurs at @xmath162 s. however , the blastwave is already in its mildly relativistic phase at that moment and it will become non - relativistic soon after that ( i.e. , when @xmath163 s , see figure 1 ) .
so it is not surprising that we could not see any obvious breaks during the relativistic phase , they just do not have time to emerge .
fortunately we have pointed out that the lateral expansion is even more notable in the non - relativistic phase , so we expect that a break should occur in the light curve between the relativistic stage and the non - relativistic stage .
the dash - dotted line in figure 8 proves our deduction .
this line corresponds to an `` ideal '' jet with @xmath164 and with other parameters equal to those in the `` standard '' set . here
the decay in the relativistic phase is @xmath165 , in good agreement with theoretical results for isotropic fireballs ( wijers , rees , & mszros 1997 ; sari et al .
1998 ) , and the decay in the non - relativistic phase is @xmath166 , also consistent with rhoads prediction for lateral expansion effect ( but made only for relativistic jets ) .
previous numerical studies are based on dynamical equations improper for non - relativistic blastwaves .
they could not correctly reveal the break between relativistic and newtonian stages ( panaitescu & mszras 1999 ; moderski , sikora , & bulik 2000 ) .
one may ask why other lines in figure 8 do not show any breaks .
the reason is they all correspond to @xmath167 and their peaks appear at relatively late stages .
for example , the thick solid line peaks at about @xmath156 s. then we can not see the initial power law decay ( i.e. , @xmath168 ) in the relativistic phase and can only see the fast decay of @xmath166 at lage stages .
however , we should note that although they do not show any obvious breaks , they all have a common characteristic : a large timing index of @xmath4 2.1 .
we suggest that the sharp decline of afterglows from some grbs ( with @xmath169 2.0 ) itself may just be the evidence of the presence of a jet ( huang , dai , & lu 2000a , b , c ) .
dai & lu ( 1999 , 2000 ) have discussed the optical afterglows from isotropic fireballs during the non - relativistic phase extensively .
the most obvious difference between their results and ours is that our timing index at non - relativistic stage is reasonably larger .
in this paper we investigate the detailed dynamical evolution of jets and their afterglows . our model is simple in form and is easy for numerical evaluations . comparing with previous studies , the model is refined in the following aspects : \(i )
equation ( [ dgdmold6 ] ) has been widely used in previous studies , however , it is not correct for non - relativistic ejecta ( huang , dai , & lu 1999a , b ) .
the dynamics here ( i.e. , mainly our equation ( [ dgdmass8 ] ) ) is applicable to both ultra - relativistic and newtonian jets , so we could follow the overall evolution ( from @xmath24 to @xmath29 ) of a jet by using a single set of differential equations .
numerical results indicate that the ejecta will cease to be ultra - relativistic @xmath156 @xmath144 s after the main grb .
we should consider this refinement seriously .
we find that during the ultra - relativistic phase , the opening angle @xmath35 increases only slightly , but at the newtonian stage it increases quickly to @xmath171 1 .
the highly radiative regime ( when @xmath172 ) lasts usually for very short period , however a substantial fraction of the initial kinetic energy could be dissipated in less than @xmath15 @xmath16 s. as for the light curves of optical afterglows , we find no theoretically predicted breaks at the critical points where the ejecta transits from @xmath173 to @xmath174 , consistent with previous numerical studies ( panaitescu & mszros 1999 ; moderski , sikora , & bulik 2000 ) .
the reason is that at this critical point the lateral expansion just begins to take effect , but its action does not prevail over @xmath2 completely , and also the relativistic phase is too short for the break to appear .
however we have shown that the lateral expansion is even more striking during the non - relativistic phase , so we expect an obvious break between the relativistic stage and the non - relativistic stage .
this has been proved by our calculations made with @xmath164 .
other calcuations made with @xmath167 do not show any obvious breaks , but they are all characterized by a large timing index .
we conclude that the most obvious characteristic of beamed eject is the quick decline of afterglows at late stages ( in fact corresponding to the non - relativistic phase ) , with @xmath4 2.1 , and in some cases , we could even see a sharp break in the light curve between the relativistic stage and the non - relativistic stage .
we suggest that the quick decay ( with @xmath175 1.7 2.0 ) of optical afterglows from some grbs , such as grb 970228 ( @xmath176 d ) , 980326 , and 980519 , and the breaks in the optical light curves of grb 990123 and 990510 , are probably due to beaming effect ( huang , dai , & lu 2000a , b , c ) .
but we should also keep in mind that other factors might lead to a large @xmath177 as well .
for example , the ism density may be a decline function of radius ( usually @xmath178 ) , then an isotropic fireball can well explain the quick decay of optical afterglows from grb 970228 , 980326 , and 990519 . additionally , if an isotropic grb remnant sweeps through a uniform ism ( @xmath179 ) and an uneven ism ( i.e. , @xmath178 ) in succession , then an obvious break is likely to present in the optical light curve .
grb 990123 and 990510 may also correspond to this case .
the suggestion by dai & lu ( 1999 , 2000 ) that the quick decay of afterglows from grb 980519 and 990123 is due to an isotropic newtonian blastwave in a dense ism ( with @xmath180 @xmath134 ) is also possible .
so , the degree of beaming is still very difficult to determine now . with the progress in observing technique ,
when much more grbs are localized , maybe we could also infer some hints from statistical researches .
however the final solution may come from systematic deep optical surveys , which are expected to find many faint decaying optical sources if grbs are highly collimated .
they are afterglows from jetted grbs whose gamma - ray emission deviates the line of sight slightly .
this work was partly supported by the national natural science foundation of china , grants 19773007 and 19825109 , the national climbing project on fundamental researches , and the national project of fundamental researches ( 973 project ) .
andersen , m. , et al .
1999 , science , 283 , 2075 blandford , r.d .
, & mckee , c.f .
1976 , phys .
fluids , 19 , 1130 bloom , j.s . , et al .
1998 , , 508 , l21 castro - tirado , a. , et al . 1999 ,
science , 283 , 2069 chiang , j. , & dermer , c.d .
1999 , , 512 , 699 costa , e. , et al .
1997 , , 387 , 783 dai , z.g . , huang , y.f .
, & lu , t. 1999 , , 520 , 634 dai , z.g . , & lu , t. 1998a , , 298 , 87 dai , z.g .
, & lu , t. 1998b , , 333 , l87 dai , z.g . , & lu , t. 1998c , , 81 , 4301 dai , z.g . , & lu , t. 1999 , , 519 , l155 dai , z.g . , & lu , t. 2000 , , in press , astro - ph/9906109 dermer , c.d . 2000 , submitted , astro - ph/0002306 dermer , c.d . ,
bttcher , m. , & chiang , j. 1999 , , 515 , l49 dermer , c.d . ,
bttcher , m. , & chiang , j. 2000 , , 537 , in press , astro - ph/9910472 dermer , c.d . , & mitman , k.e .
1999 , , 513 , l5 dermer , c.d . ,
chiang , j. , & bttcher , m. 1999 , , 513 , 656 dermer , c.d . , chiang , j. , & mitman , k.e .
2000 , submitted , astro - ph/9910240 fruchter , a.s . , et al . 1999 , , 519 , l13 galama , t.j .
, et al . 1999a , , 398 , 394 galama , t.j . , et al .
1999b , submitted , astro - ph/9907264 goodman , j. 1986 , , 308 , l47 groot , p.j .
, et al . 1998 , , 502 , l123 gruzinov , a. 1999 , , 525 , l29 harrison , f.a . , et al . 1999 , , 523 , l121 huang , y.f .
, dai , z.g . , & lu , t. 1998a , , 336 , l69 huang , y.f . ,
dai , z.g . ,
wei , d.m . , & lu , t. 1998b , , 298 , 459 huang , y.f . ,
dai , z.g . , & lu , t. 1999a , chin . phys . lett . , 16 , 775 ( astro - ph/9906404 ) huang , y.f . ,
dai , z.g . , & lu , t. 1999b , , 309 , 513 ( astro - ph/9906370 ) huang , y.f . , dai , z.g . , & lu , t. 2000a , , 355 , l43 ( astro - ph/0002433 ) huang , y.f . ,
dai , z.g .
, & lu , t. 2000b , , in press huang , y.f . ,
dai , z.g . , & lu , t. 2000c , chin .
, in press katz , j. 1994 , , 422 , 248 kirk , j.g . , & duffy , p. 1999
, journal of physics g , 25 , r163 kulkarni , s.r .
, et al . 1998 , , 393 , 35 kulkarni , s.r . , et al . 1999 , , 398 , 389 mszros , p. , & rees , m.j . 1992 , , 257 , 29p mszros , p. , & rees , m.j . 1997 , , 476 , 232 moderski , r. , sikora , m. , & bulik , t. 2000 , , 529 , 151 owens , a. , et al . 1998 , , 339 , l37 paczyski , b. 1986 , , 308 , l43 panaitescu , a. , & mszros , p. 1998
, , 493 , l31 panaitescu , a. , & mszros , p. 1999
, , 526 , 707 panaitescu , a. , mszros , p. , & rees , m.j .
1998 , , 503 , 315 piran , t. 1999 , , 314 , 575 pugliese , g. , falcke , h. , & biermann , p.l .
1999 , , 344 , l37 rees , m.j . , & mszros , p. 1992
, , 258 , 41p rees , m.j . , & mszros , p. 1994
, , 430 , l93 reichart , d.e . , lamb , d.q . ,
metzger , m.r . ,
quashnock , j.m . ,
cole , d.m . ,
castander , f.j . ,
cooray , a.r .
, & vanden berk , d.e . 1999 , , 517 , 692 rhoads , j. 1997 , , 487 , l1 rhoads , j. 1999a , , 525 , 737 rhoads , j. 1999b , a&as , 138 , 539 rybicki , g.b . ,
& lightman , a.p .
1979 , radiative processes in astrophysics , wiley , new york sari , r. 1997a , , 489 , l37 sari , r. 1997b , , 494 , l49 sari , r. , narayan , r. , & piran , t. 1996 , , 473 , 204 sari , r. , piran , t. , & halpern , j.p .
1999 , , 519 , l17 sari , r. , piran , t. , & narayan , r. 1998 , , 497 , l17 sedov , l. 1969 , similarity and dimensional methods in mechanics , academic , new york , ch .
iv stanek , k.z . , garnavich , p.m. , kaluzny , j. , pych , w. , & thompson , i. 1999 , , 522 , l39 tavani , m. 1997 , , 483 , l87 vietri , m. 1997 , , 488 , l105 waxman , e. 1997a , , 485 , l5 waxman , e. 1997b , , 491 , l19 wei , d.m . , & lu , t. 2000 , , in press wijers , r.a.m.j . ,
& galama , t.j . 1999 , , 523 , 177 wijers , r.a.m.j .
, rees , m.j . , & mszros , p. 1997 , , 288 , l51 | whether gamma - ray bursts are highly beamed or not is a very difficult but important problem that we are confronted with .
some theorists suggest that beaming effect usually leads to a sharp break in the afterglow light curve during the ultra - relativistic phase , with the breaking point determined by @xmath0 , where @xmath1 is the lorentz factor of the blastwave and @xmath2 is the initial half opening angle of the ejecta , but numerical studies tend to reject the suggestion .
we note that previous studies are uniformly based on dynamics that is not proper for non - relativistic blastwaves . here
we investigate the problem in more detail , paying special attention to the transition from the ultra - relativistic phase to the non - relativistic phase .
due to some crucial refinements in the dynamics , we can follow the overall evolution of a realistic jet till its velocity is as small as @xmath3 .
we find no obvious break in the optical light curve during the relativistic phase itself .
however , an obvious break does appear at the transition from the relativistic phase to the newtonian phase if the physical parameters involved are properly assumed . generally speaking ,
the newtonian phase is characterized by a sharp decay of optical afterglows , with the power law timing index @xmath4 2.1 .
this is due to the quick lateral expansion at this stage .
the quick decay of optical afterglows from grb 970228 , 980326 , and 980519 , and the breaks in the optical light curves of grb 990123 and 990510 may indicate the presence of highly collimated @xmath1-ray burst ejecta . |
solitons play an important rle in many areas of physics . as classical solutions of non - linear field theories
, they are localized structures with finite energy , which are globally regular . in general
, one can distinguish between topological and non - topological solitons .
while topological solitons @xcite possess a conserved quantity , the topological charge , that stems ( in most cases ) from the spontaneous symmetry breaking of the theory , non - topological solitons @xcite have a conserved noether charge that results from a symmetry of the lagrangian .
the standard example of non - topological solitons are @xmath1-balls @xcite , which are solutions of theories with self - interacting complex scalar fields .
these objects are stationary with an explicitly time - dependent phase . the conserved noether charge @xmath1 is then related to the global phase invariance of the theory and is directly proportional to the frequency .
@xmath1 can e.g. be interpreted as particle number @xcite .
while in standard scalar field theories , it was shown that a non - renormalizable @xmath2-potential is necessary @xcite , supersymmetric extensions of the standard model ( sm ) also possess @xmath1-ball solutions @xcite . in the latter case ,
several scalar fields interact via complicated potentials .
it was shown that cubic interaction terms that result from yukawa couplings in the superpotential and supersymmetry ( susy ) breaking terms lead to the existence of @xmath1-balls with non - vanishing baryon or lepton number or electric charge .
these supersymmetric @xmath1-balls have been considered as possible candidates for baryonic dark matter @xcite and their astrophysical implications have been discussed @xcite . in @xcite , these objects have been constructed numerically using the exact form of a scalar potential that results from gauge - mediated susy breaking . however , this potential is non - differentiable at the susy breaking scale . in @xcite a differentiable approximation of this potential
was suggested and the properties of the corresponding @xmath1-balls have been investigated in @xmath3 dimensions .
this was extended to @xmath4 dimensions in @xcite .
@xmath1-ball solutions with a @xmath2-potential in @xmath3 dimensions have been studied in detail in @xcite .
it was realized that next to non - spinning @xmath1-balls , which are spherically symmetric , spinning solutions exist .
these are axially symmetric with energy density of toroidal shape and angular momentum @xmath5 , where @xmath1 is the noether charge of the solution and @xmath6 corresponds to the winding around the @xmath7-axis . approximated solutions of the non - linear partial differential equations were constructed in @xcite by means of a truncated series in the spherical harmonics to describe the angular part of the solutions .
the full partial differential equation was solved numerically in @xcite .
it was also realized in @xcite that in each @xmath8-sector , parity - even ( @xmath9 ) and parity - odd ( @xmath10 ) solutions exist .
parity - even and parity - odd refers to the fact that the solution is symmetric and anti - symmetric , respectively with respect to a reflection through the @xmath11-@xmath12-plane , i.e. under @xmath13 .
complex scalar field models coupled to gravity possess so - called `` boson star '' solutions @xcite . in @xcite boson stars have been considered that have flat space - time limits in the form of @xmath1-balls .
these boson stars are hence self - gravitating @xmath1-balls . in @xcite
the gravitating generalizations of the supersymmetric @xmath1-balls studied in @xcite have been discussed in @xmath4 dimensions .
it was found that the behaviour of the mass and charge at the critical value of the frequency depends crucially on the number of dimensions @xmath14 . while in most models considered , the scalar field function is exponentially decaying and hence different notions of a boson star radius exist ,
this is different in models with a v - shaped potential @xcite . in this case ,
compact boson stars with a well - defined outer radius ( very similar to those of `` standard stars '' ) can be given @xcite .
in this paper we are interested in the possibility to detect boson stars through the motion of massive and massless test particles in their space - time .
in particular , we will be interested in the difference between test particle motion in a boson star space - time and a schwarzschild space - time .
since boson stars are very compact , they have been considered as alternatives to supermassive black holes @xcite residing e.g. in the center of galaxies and geodesic motion of massive test particles describing extreme - mass - ratio inspirals ( emris ) has been discussed in some particular cases in @xcite .
while objects with a well - defined surface as alternatives to the supermassive black hole at the center of our own galaxy , the milky way have been ruled out @xcite , the boson stars studied in this paper have a scalar field falling of exponentially at infinity and hence strictly speaking do not have a well - defined outer surface outside which the energy density and pressure , respectively , vanishes .
we hence study _ non - compact _ boson stars in this paper and we will make a detailed analysis of the motion of massive and massless test particles in the space - time of such a boson star . our paper is organised as follows : in section ii , we give the field theoretical model , the ansatz and the equations of motion to describe the space - time of a non - spinning , non - compact boson star . in section iii
we discuss the geodesic equation and give our numerical results in section iv .
we conclude in section v.
in the following we will discuss the field theoretical model to describe the space - time of a non - spinning , non - compact boson star in which the test particles will move on geodesics .
the action @xmath15 of the field theoretical model reads : @xmath16 where @xmath17 is the ricci scalar , @xmath18 denotes newton s constant and the matter lagrangian is given by @xmath19 where @xmath20 denotes a complex scalar field and we choose as signature of the metric @xmath21 .
@xmath22 is the scalar field potential that arises in gauge - mediated supersymmetric breaking in the minimal supersymmetric extension of the standard model ( mssm ) @xcite : @xmath23 where @xmath24 corresponds to the scale below which supersymmetry is broken and is roughly @xmath25 tev , while @xmath8 denotes the scalar boson mass . for @xmath26 we can give a polynomial expansion of this potential which reads @xmath27 neglecting terms of @xmath28 and higher
this potential has the form of the scalar potential frequently used in the study of @xmath1-balls and boson stars @xcite and reads @xmath29 where @xmath30 , @xmath31 , @xmath32 and @xmath33 are constants such that @xmath34 corresponds to the scalar boson star mass @xmath8 . in previous studies ,
the constants have been fixed to particular values which fulfill the requirements to find non - trivial solutions to the equations of motion .
the matter lagrangian @xmath35 ( [ lag ] ) is invariant under the global u(1 ) transformation @xmath36 as such the locally conserved noether current @xmath37 , @xmath38 , associated to this symmetry is given by @xmath39 the globally conserved noether charge @xmath1 of the system then reads @xmath40 finally , the energy - momentum tensor is given by @xmath41 + \partial_{\mu } \phi^ * \partial_{\nu }
\phi + \partial_{\nu}\phi^ * \partial_{\mu } \phi\ .\end{aligned}\ ] ] the coupled system of ordinary differential equations is given by the einstein equations @xmath42 with @xmath43 given by ( [ em ] ) and the klein - gordon equation @xmath44 in the following , we want to study non - spinning boson stars . for the metric we use
the following ansatz in isotropic coordinates @xmath45 \ .\ ] ] for the complex scalar field , we use a stationary ansatz that contains a periodic dependence of the time - coordinate @xmath46 : @xmath47 where @xmath48 is a constant and denotes the frequency . in order to be able to use dimensionless quantities
we introduce the following rescalings @xmath49 and find that the equations depend only on the dimensionless coupling constant @xmath50 where @xmath51 is the planck mass .
note that @xmath52 denotes the ratio between the susy breaking scale and the planck mass which we expect to be small @xmath53 if the susy breaking scale is on the order of @xmath25 tev . in this paper
, we will however also study larger values of @xmath52 in order to understand the qualitative behaviour of the solutions in curved space - time .
note that with these rescalings the scalar boson mass @xmath8 becomes equal to unity . using the rescalings
the einstein equations read @xmath54\ ] ] and @xmath55 \ , \ ] ] while the klein - gordon equation for the scalar field function is @xmath56 \ .\ ] ] the prime here and in the following denotes the derivative with respect to @xmath57 .
the equations ( [ einstein1])-([kg ] ) have to be solved subject to appropriate boundary conditions .
these are given by the requirement of regularity at the origin @xmath58 and by the requirement of finite energy , asymptotically flat solutions @xmath59 the set of coupled , nonlinear ordinary differential equations ( [ einstein1])-([kg ] ) has been studied in detail in @xcite , however in schwarzschild - like coordinates .
we will quote the relevant results in the following when necessary , but refer the reader to that paper for details on the field theoretical solutions .
the behaviour of the metric function @xmath60 determines the mass @xmath61 of the solution .
this is given by @xcite @xmath62 while the explicit expression for the noether charge reads @xmath63 note that the mass @xmath61 is measured in units of @xmath64 .
at spatial infinity @xmath65 the scalar field function decays exponentially @xmath66 hence @xmath67 for @xmath68 .
as such the boson star does not possess a well - defined surface and different notions of the radius of the boson star are possible ( see discussion below ) .
while boson stars with well - defined surface do exist @xcite ( so - called `` compact boson stars '' ) , which hence rather resemble astrophysical objects , boson stars without proper surface could play an important rle as alternatives to supermassive black holes in the center of galaxies .
we consider the geodesic equation @xmath69 where @xmath70 denotes the christoffel symbol given by @xmath71 and @xmath72 is an affine parameter such that for time like geodesics @xmath73 corresponds to proper time .
the geodesic lagrangian @xmath74 for a point particle in the space time ( [ metric ] ) reads @xmath75 \ , \end{aligned}\ ] ] where @xmath76 for massless particles and @xmath77 for massive particles , respectively .
the constants of motion are the energy @xmath78 and the angular momentum @xmath79 of the particle ( absolute value and direction ) .
to fix the direction , we choose @xmath80
. hence the motion will occur in the equatorial plane .
the conserved quantities then are @xmath81 using these constants of motion we can rewrite ( [ lagrangian_geo ] ) as follows @xmath82 where @xmath83 is the effective potential @xmath84 we then find @xmath85 the solution to this will give @xmath86 for a test particle .
the proper time elapsing along the trajectory is given by the integral @xcite @xmath87 in the next sections the evolution of the proper time along a geodesic is shown by points .
since @xmath60 and @xmath88 are only given numerically , @xmath89 and @xmath86 can only be determined numerically .
we have solved the equations of motion ( [ einstein1])-([kg ] ) numerically using a newton - raphson method with an adaptive grid scheme @xcite .
the relative errors of the numerical integration are on the order of @xmath90 .
the geodesic equation is integrated numerically in fortran using the fortran subroutines for mathematical applications
imsl math / library .
the accuracy of the integration as estimated from application of the method for the integration of geodesics in the schwarzschild space - time is on the order of @xmath91 . in order to be able to compare the space - time of a boson star and that of a schwarzschild black hole
, we have first investigated the difference between the metric functions @xmath60 and @xmath88 .
for that , we have to remind ourselves of how the schwarzschild metric looks like in isotropic coordinates which we use in this paper .
this is given by ( [ metric ] ) with @xmath92 where @xmath61 is the adm mass of the solution and the isotropic coordinate @xmath57 is related to the schwarzschild like coordinate @xmath93 by @xmath94 . in fig.[compare_boson_s ] we show the metric functions @xmath95 and @xmath96 for a boson star with given mass @xmath61 and in comparison to that the metric functions for a schwarzschild solution with the same @xmath61 .
( left ) and @xmath96 ( right ) for a boson star with given mass @xmath97 for @xmath98 .
also given are the metric functions @xmath95 and @xmath96 for a schwarzschild solution with the same value of @xmath61 .
note that the schwarzschild solution has @xmath99 at the event horizon @xmath100 . ] clearly , the metric functions agree at large @xmath57 , while for small @xmath57 they deviate strongly . in particular the metric functions
@xmath95 and @xmath96 of the schwarzschild solution tend to zero at the event horizon @xmath101 , while for boson stars @xmath95 and @xmath96 stay perfectly regular for all @xmath102 .
this fact will play an important rle in the behaviour of test particles at small @xmath57 , while our results already indicate here that the behaviour at large @xmath57 will be similar when comparing boson star space - times with the schwarzschild space - time .
note that we have chosen a non - realistically high value of the parameter @xmath52 here in order to show the qualitative features of the changes .
in reality @xmath52 would be many orders of magnitude smaller .
it has been observed before that boson stars exist only in a limited range of the frequency @xmath48 . at the maximal value of @xmath48 both the mass and the charge of the boson stars
tend to zero , while at the minimal value of @xmath48 a second branch of solutions starts to form and finally ends at the critical value of @xmath48 after a number of spirals . in the following ,
we want to discuss the motion of test particles in boson star space - times . here
we concentrate on the boson stars at some particular values of the frequency @xmath48 .
these are indicated in fig.[om_m ] , where we plot the mass @xmath61 as function of @xmath48 for @xmath103 and @xmath98 , respectively .
the exact numerical values of the mass @xmath61 and charge @xmath1 are given in table [ omega_values ] .
is shown as function of @xmath48 for two different values of @xmath52 .
the black dots on the curves indicate the values of @xmath48 that we have investigated in more detail and that are given in table [ omega_values ] . ]
.the values of @xmath48 as indicated by the black dots in fig.[om_m ] together with the corresponding values of the charge @xmath1 , mass @xmath61 and the value @xmath104 .
the ( 1 ) and ( 2 ) denote the solutions on the 1 . and 2 .
branch , respectively .
see also @xcite for further details using the potential ( [ phi6 ] ) . [
cols="^,^,^,^,^,^",options="header " , ] in table [ light_deflection ] we give the light deflection in the boson star space - times indicated in table [ omega_values ] for massless test particles with @xmath105 and @xmath106 .
we observe that the light deflection increases when moving along the branch for @xmath103 up to roughly the minimal possible value of @xmath107 . on the second branch ,
the light deflection decreases again .
similar to what was stated above for massive test particles , there seems to be no direct correlation between the mass of the boson star and the value of light deflection and hence the presence of the scalar field seems to have a non - trivial effect . for @xmath98
we find that the light deflection decreases when moving along the branches .
this again indicates that a semi - analytical treatment of the space - time of boson stars does not lead to the proper results and that numerical techniques are vital to make predictions about observables in such kind of space - times .
in this paper , we have studied the geodesic motion of test particles in boson star space - times and have discussed possibilities to differ them from the geodesics around a schwarzschild black hole .
our work has many interesting applications . as a first example
let us mention the accretion of matter onto massive compact objects such as neutron stars .
neutron stars are very dense objects that are typically modeled using an appropriate equation of state .
boson stars could act as a toy model for these objects . in the following
, one could use our techniques to model accretion flow onto a boson star in a simple ballistic approach .
this was done before in black hole space - times where analytic solutions to the geodesic equation exist @xcite .
this technique could be applied as well using the numerical data of a boson star space - time .
the accretion flow would also be interesting from the point of view of interpreting the boson star as an alternative to a supermassive black hole residing in the center of our milky way .
the accretion of matter leads to the emission of radio waves and these can be used to make predictions about the properties of the supermassive object at the center of our galaxy .
gravitational wave signals would be another possibility .
these have already been discussed in @xcite
. it would also be interesting to see how test particles move in radially excited boson star space - times as well as in the space - time of a rotating boson star .
for once this is interesting since it is believed that the flat space - time counterparts of boson stars , so - called @xmath1-balls could originally form in an excited state @xcite .
furthermore , due to the interaction with surrounding matter , e.g. accretion , boson stars might be excited .
if boson stars should act as toy model for neutron stars it would be surely also of interest to study the rotating counterparts since basically all observed neutron stars seem to rotate , in fact quite quickly as the example of a radio pulsar ( assumed to be a neutron star ) spinning at 716 hz shows @xcite .
it would then be interesting to study the motion of test particles in these space - times . finally , in order to model accretion flow onto dense astrophysical objects more realistically one should consider charged test particles ( that would make up the plasma ) .
hence the equation of motion for the particle should be modified to include the charge of the particle .
studies of charged test particles in some black hole space - times have been done previously @xcite . in order to have interactions of the test particles with the boson star the latter
should be charged as well .
this can be done by promoting the global u(1 ) symmetry of our model to a local u(1 ) and has been considered for a number of boson stars @xcite .
the corresponding equation of motion for the test particle would then have to be solved numerically .
+ + + * acknowledgment * we gratefully acknowledge the deutsche forschungsgemeinschaft ( dfg ) for financial support within the framework of the dfg research training group 1620 _ models of gravity_. we also acknowledge discussions with parinya sirimachan at the initial stages of this paper . 99 n.s .
manton and p.m. sutcliffe , _ topological solitons _
, cambridge university press , 2004 .
r. friedberg , t. d. lee and a. sirlin , phys .
d * 13 * ( 1976 ) , 2739 .
t. d. lee and y. pang , phys . rep .
* 221 * ( 1992 ) , 251 .
s. r. coleman , nucl .
b * 262 * ( 1985 ) , 263 .
volkov and e. whnert , phys .
d * 66 * ( 2002 ) , 085003 .
a. kusenko , phys .
b * 404 * ( 1997 ) , 285 ; phys .
b * 405 * ( 1997 ) , 108 .
a. kusenko , hep - ph/0009089 .
k. enqvist and j. mcdonald , phys .
b * 425 * ( 1998 ) , 309 ; s. kasuya and m. kawasaki , phys . rev .
d * 61 * ( 2000 ) , 041301 ; a. kusenko and p. j. steinhardt , phys .
lett . * 87 * ( 2001 ) , 141301 ; t. multamaki and i. vilja , phys .
b * 535 * ( 2002 ) , 170 ; m. fujii and k. hamaguchi , phys . lett .
b * 525 * ( 2002 ) , 143 ; m. postma , phys . rev .
d * 65 * ( 2002 ) , 085035 ; k. enqvist , _ et al .
_ , phys .
b * 526 * ( 2002 ) , 9 ; m. kawasaki , f. takahashi and m. yamaguchi , phys . rev .
d * 66 * ( 2002 ) , 043516 ; a. kusenko , l. loveridge and m. shaposhnikov , phys .
d * 72 * ( 2005 ) , 025015 ; y. takenaga _ et al . _ [ super - kamiokande collaboration ] , phys .
b * 647 * ( 2007 ) , 18 ; s. kasuya and f. takahashi , jcap * 11 * ( 2007 ) , 019 .
l. campanelli and m. ruggieri , phys .
d * 77 * ( 2008 ) , 043504 ; l. campanelli and m. ruggieri , phys .
d * 80 * ( 2009 ) 036006 .
e. copeland and m. tsumagari , phys.rev .
d * 80 * 025016 ( 2009 ) .
b. hartmann and j. riedel , phys .
d * 87 * ( 2013 ) 044003 .
b. kleihaus , j. kunz and m. list , phys .
d * 72 * ( 2005 ) , 064002 .
b. kleihaus , j. kunz , m. list and i. schaffer , phys .
d * 77 * ( 2008 ) , 064025 .
y. brihaye and b. hartmann , nonlinearity * 21 * ( 2008 ) , 1937 .
d. j. kaup , phys . rev .
* 172 * ( 1968 ) , 1331 .
e. mielke and f. e. schunck , proc .
8th marcel grossmann meeting , jerusalem , israel , 22 - 27 jun 1997 , world scientific ( 1999 ) , 1607 .
r. friedberg , t. d. lee and y. pang , phys .
d * 35 * ( 1987 ) , 3658 .
p. jetzer , phys .
* 220 * ( 1992 ) , 163 .
f. e. schunck and e. mielke , class .
* 20 * ( 2003 ) r31 .
f. e. schunck and e. mielke , phys .
lett . a * 249 * ( 1998 ) , 389 .
a. r. liddle , m. s. madsen , int .
j. mod .
* d1 * ( 1992 ) 101 .
h. arodz and j. lis , phys .
d * 77 * , 107702 ( 2008 ) .
h. arodz and j. lis , phys .
d * 79 * , 045002 ( 2009 ) .
b. kleihaus , j. kunz , c. lammerzahl and m. list , phys .
b * 675 * , 102 ( 2009 ) .
b. kleihaus , j. kunz , c. lammerzahl and m. list , phys .
d * 82 * , 104050 ( 2010 ) .
b. hartmann , b. kleihaus , j. kunz and i. schaffer , phys .
b * 714 * ( 2012 ) 120 [ arxiv:1205.0899 [ gr - qc ] ] . e. tejeda , p. a. taylor and j. c. miller , _ an analytic toy model for relativistic accretion in kerr spacetime _
, arxiv:1207.0733 [ astro-ph.he ] . j. w. t. hessels , s. m. ransom , i. h. stairs , p. c. c. freire , v. m. kaspi and f. camilo , science * 311 * 1901 ( 2006 ) .
s. chandrasekhar , _ the mathematical theory of black holes _ ( oxford university press , oxford 1983 ) ; v. f. gackstatter , ann .
der phys .
* 40 * ( 1983 ) 352 ; j. bik , z. stuchlk , v. balek , bull .
. czechosl .
* 40 * ( 1989 ) 65 ; s. grunau and v. kagramanova , phys .
d * 83 * ( 2011 ) 044009 . | we study the geodesic motion of test particles in the space - time of non - compact boson stars .
these objects are made of a self - interacting scalar field and depending on the scalar field s mass can be as dense as neutron stars or even black holes .
in contrast to the former these objects do not contain a well - defined surface , while in contrast to the latter the space - time of boson stars is globally regular , can however only be given numerically .
hence , the geodesic equation also has to be studied numerically .
we discuss the possible orbits for massive and massless test particles and classify them according to the particle s energy and angular momentum .
the space - time of a boson star approaches the schwarzschild space - time asymptotically , however deviates strongly from it close to the center of the star . as a consequence ,
we find additional bound orbits of massive test particles close to the center of the star that are not present in the schwarzschild case .
our results can be used to make predictions about extreme - mass - ratio inspirals ( emris ) and we hence compare our results to recent observational data of the stars orbiting _
sagittarius a@xmath0 _ - the radiosource at the center of our own galaxy . |
as a complement to detailed simulations of wolf - rayet ( wr ) winds ( e.g. , * ? ? ?
* ; * ? ? ?
* ) , we have developed a set of simple analytic models using diffusive cak - type line driving with frequency redistribution @xcite . in this paper
, we apply these models to a preliminary study of the @xmath0 relation for wr stars , which is important for the study of wr stars as progenitors of long - duration gamma - ray bursts @xcite .
the details of the basic model are discussed in @xcite .
we compare our results with those from @xcite ( hereafter vdk ) using their wn star parameters ( see figure [ vdfig ] ) .
the metal abundances were provided by a. heger ( private communication ) from a 25 @xmath5 evolution model based on solar abundances from @xcite . while the @xmath2 from these abundances is smaller than the traditional @xmath6 , uncertainties in other model parameters
are expected to dominate over uncertainties in abundance .
figure [ vdfig ] shows the metallicity dependence of our wnl model , as compared to the model in vdk .
a least - squares fit to our result over @xmath3 gives @xmath7 , where @xmath8 , very close to the vdk result of @xmath9 .
it appears that our @xmath1 flattens more quickly than those of vdk , perhaps because fe saturates more quickly with our more complete line list .
calculations to higher @xmath2 need to be done to confirm that there is indeed a flattening , however .
log @xmath1 vs. log @xmath10 for a typical wnl star ( @xmath11 ) .
plusses and lines are from @xcite , figure 2 .
crosses are from this paper . ]
we have applied analytic wr wind models to the study of @xmath1 as a function of @xmath2 . our preliminary results for a wnl - type star show a power - law dependence of @xmath1 with @xmath2 over @xmath12 , with index @xmath8 , similar to the the findings vdk .
however , our @xmath1 seems to flatten more quickly .
we plan to cover a wider range in @xmath2 to confirm this and to perform the same analysis for wc stars .
a. onifer would like to thank alexander heger for helpful discussions and valuable data .
portions of this work were performed under the auspices of the u.s .
department of energy by los alamos national laboratory under contract no . | to better understand wolf - rayet stars as progenitors of gamma - ray bursts , an understanding of the effect metallicity has on wolf - rayet mass loss is needed . using simple analytic models , we study the @xmath0 relation of a wn star and compare the results to similar models .
we find that @xmath1 roughly follows a power law in @xmath2 with index 0.88 from @xmath3 and appears to flatten by @xmath4 . |
let @xmath3 be a space endowed with a probability measure @xmath4 , and let @xmath5 and @xmath6 be @xmath7 independent random variables with values in @xmath3 , distributed according to @xmath4 ; from the statistical point of view , @xmath8 is the set of given data .
let @xmath9 be a loss function which associates a real number @xmath10 to any real - valued measurable function @xmath11 defined on @xmath3 and any point @xmath12 .
denote by @xmath13 the loss function @xmath14 associated with @xmath11 and set @xmath15 to be the associated risk .
the risk of any statistic @xmath16 is defined by @xmath17 $ ] .
let @xmath18 be a class ( usually called the _ model _ ) of real - valued measurable functions defined on @xmath3 .
in learning theory , one wants to assume as little as possible on the class @xmath18 , or on the measure @xmath4 .
the aim is to use the data to construct learning algorithms whose risk is as close as possible to @xmath19 ( and when this infimum is attained by a function @xmath20 in @xmath18 , this element is called an _ oracle _ ) .
hence , one would like to construct procedures @xmath21 such that , for some @xmath22 , with high probability , @xmath23 the role of the _ residual term _ ( or _ rate _ ) @xmath24 is to capture the `` complexity '' of the problem , and the hope is to make it as small as possible .
when @xmath24 tends to zero as @xmath25 tends to infinity , inequality ( [ eqoracle - inequality ] ) is called an _
oracle inequality_. when @xmath26 , we say that @xmath27 satisfies an _ exact oracle inequality _
( the term _ sharp _ oracle inequality has been also used ) and when @xmath28 it satisfies a _ nonexact oracle inequality_. note that the terminology `` risk bounds '' has been also used for ( [ eqoracle - inequality ] ) in the literature .
a natural algorithm in this setup is the _ empirical risk minimization procedure _ ( erm ) ( terminology due to @xcite ) , in which the _ empirical risk _ functional @xmath29 is minimized and produces @xmath30 .
note that when @xmath31 does not achieve its infimum over @xmath18 or if the minimizer is not unique , we define @xmath32 to be an element in @xmath18 for which @xmath33 .
this algorithm has been extensively studied , and we will compare our first result to the one of .
one motivation in obtaining nonexact oracle inequalities [ equation ( [ eqoracle - inequality ] ) for @xmath34 is the observation that in many situations , one can obtain such an inequality for the erm procedure with a residual term @xmath24 of the order of @xmath1 , while the best residual term achievable by erm in an exact oracle inequality [ equation ( [ eqoracle - inequality ] ) for @xmath35 will only be of the order of @xmath36 for the same problem .
for example , consider the simple case of a finite model @xmath18 of cardinality @xmath37 and the bounded regression model with the quadratic loss function [ i.e. , @xmath38 with @xmath39 for some absolute constant @xmath40 and @xmath41 .
it can be verified that for every @xmath42 , with probability greater than @xmath43 , @xmath32 satisfies a nonexact oracle inequality with a residual term proportional to @xmath44 . on the other hand ,
it is known @xcite that in the same setup , there are finite models for which , with probability greater than a positive constant , @xmath45 can not satisfy an exact oracle inequality with a residual term better than @xmath46 .
thus , it is possible to establish two optimal oracle inequalities [ i.e. , oracle inequalities with a nonimprovable residual term @xmath24 up to some multiplying constant ] for the same procedure with two very different residual terms : one being the square of the other one
. we will see below that the same phenomenon occurs in the classification framework for vc classes .
thus our main goal here is to present a general framework for nonexact oracle inequalities for erm and rerm ( regularized erm ) , and show that they lead to fast rates in cases when the best known exact oracle inequalities have slow rates .
although the improved rates are significant , it is clear that exact inequalities are more `` valuable '' from the statistical point of view .
for example , consider the regression model with the quadratic loss .
it follows from an exact oracle inequality on the prediction risk [ equation ( [ eqoracle - inequality ] ) for @xmath26 ] , another exact oracle inequality , but for the estimation risk @xmath48 where @xmath49 is the regression function of @xmath50 given @xmath51 , and @xmath52 is the @xmath53-norm with respect to the marginal distribution of @xmath51 . in other words ,
exact oracle inequalities for the prediction risk @xmath54 provide both prediction and estimation results ( prediction of the output @xmath50 and estimation of the regression function @xmath49 ) whereas nonexact oracle inequalities provide only prediction results . of course , nonexact inequalities are very useful when it suffices to compare the risk @xmath55 with @xmath56 ; and the aim of this note is to show that the residual term can be dramatically improved in such cases .
the first result of this note is a nonexact oracle inequality for the erm procedure .
to state this result , we need the following notation .
let @xmath57 be a class of real - valued functions defined on @xmath3 .
an important part of our analysis relies on the behavior of the supremum of the empirical process indexed by @xmath57 @xmath58 where for every @xmath59 , we set @xmath60 and @xmath61 .
recall that for every @xmath62 , the @xmath63 norm of @xmath64 is @xmath65 we will control the supremum ( [ eqsupremum - empirical - process ] ) using the quantities @xmath66 note that for a bounded class @xmath57 , one has @xmath67 and in the sub - exponential case , @xmath68 ( this follows from pisier s inequality ) ; cf .
lemma 2.2.2 in @xcite . throughout this note
we will also use the notation @xmath69 and for any pseudo - norm @xmath70 on @xmath71 , we will denote by @xmath72 the diameter of @xmath57 with respect to this norm . observe that the desired bound depends on the @xmath73 behavior of the _ envelope function _ of the class , @xmath74 .
although this extension seems minor at first , the examples we will present show that the assumption is not very restrictive and allows one to deal with lasso - type situations , in which the indexing class is very small something which is impossible under the @xmath75 assumption . on the other hand
, it should be emphasized that this is not a step towards an unbounded learning theory . for such results
, the analogous assumption should be that the class has a bounded diameter in @xmath73 , which is , of course , a much weaker assumption than a @xmath73 envelope function and requires different methods ; see , for example , @xcite . to obtain
the required bound , we will study empirical processes indexed by sets associated with @xmath57 , namely , the star - shaped hull of @xmath57 around zero and the localized subsets for different levels @xmath76 , defined by @xmath77 given a model @xmath18 and a loss function @xmath9 , consider the loss class and the excess loss class @xmath78 and the excess loss class .
we will assume that an oracle @xmath20 exists in @xmath18 , and from here on set @xmath79 .
[ theoa ] there exists an absolute constant @xmath80 for which the following holds .
let @xmath18 be a class of functions and assume that there exists @xmath81 such that for every @xmath82 , @xmath83 .
let @xmath84 , set @xmath85 for which @xmath86 and put @xmath87 an increasing function satisfying that for every @xmath42 , @xmath88 then , for every @xmath42 , with probability greater than @xmath43 , @xmath89 although the formulation of theorem [ theoa ] requires that for every @xmath90 , @xmath91 , we will show that if @xmath9 is nonnegative , this condition is trivially satisfied for @xmath92 .
unfortunately , this type of condition is far from being trivially satisfied for the excess loss class @xmath93 , which is one of the major differences between exact and nonexact oracle inequalities . indeed ,
bernstein condition _
, that for every @xmath94 , @xmath95 ( see @xcite or section [ seccomments ] below ) , used in to obtain exact oracle inequalities with fast rates ( rates of the order of @xmath1 ) , depends on the geometry of the problem @xcite and may not be true in general .
theorem [ theoa ] is similar in nature to corollary 2.9 of @xcite and a detailed comparison between the two results can be found in section [ seccomments ] .
theorem [ theoa ] is similar in nature to theorem 2 in @xcite .
let @xmath96 be a nondecreasing , continuous function , for which @xmath97 and @xmath98 is nonincreasing .
set @xmath18 to be a class of functions where there is some @xmath99 such that @xmath100 and @xmath101 .
if @xmath102 for any @xmath103 satisfying @xmath104 , and @xmath105 is the unique solution of the equation @xmath106 , then for every @xmath107 , with probability greater than @xmath108 , @xmath109 one of the applications of the above theorem in learning theory is for the loss function @xmath110 .
it leads to an exact oracle inequality for the erm procedure , preformed in a class @xmath18 of vc dimension ( see @xcite for more details ) , and with a residual term of the order of @xmath111 . in comparison , in the same situation , for every @xmath82 , @xmath112
therefore , it follows from theorem [ theoa ] , the argument used to obtain equation ( 29 ) in @xcite ( or example 3 in @xcite ) and the peeling argument which will be presented in ( [ eqpeeling - shahar ] ) below , that for every @xmath107 , with probability greater than @xmath43 , @xmath113 the residual term @xmath114 obtained in @xcite is optimal , but since it heavily depends on the parameter @xmath115 , it ranges between @xmath116 and @xmath117 ( up to a logarithmic factor ) .
in particular , it can be as bad as the _ square root _ of the residual term of the nonexact oracle inequality ( [ eqmassart - neoi ] ) in the same situation .
the main difference between the two results is that the condition @xmath118 for every @xmath94 is always satisfied whereas the condition that for every @xmath82 @xmath119 depends on the relative position of @xmath50 and @xmath18 , and thus on geometry of the system @xmath120 .
it is interesting to note that the residual term in ( [ eqmassart - neoi ] ) always yields fast rate even for hard classification problem such that @xmath121=1/2 $ ] .
this means that while the prediction problem in classification is completely blind to the geometry of the model , the estimation problem is influenced in a very strong way by the geometry of @xmath120 .
thus , estimating the regression function ( or the bayes rule ) is in general much harder than predicting the output @xmath50 .
another related result is the one in @xcite where ( among other results ) an exact oracle inequality is proved for the erm with a residual term @xmath122 .
the residual term is controlled using the empirical oscillation @xmath123 indexed by @xmath124 , and by the @xmath53 diameter @xmath125 @xmath126 note that all the quantities @xmath127 , @xmath128 from @xcite , @xmath122 from @xcite , @xmath129 from @xcite or theorem [ theoexact - oracle - ineq - erm ] below , define the residual terms of the oracle inequalities as a fixed point of some equation .
those appear naturally either from _ iterative localization _ of the excess risk , converging to @xmath122 @xcite , or from an `` isomorphic '' argument @xcite identifying the `` level '' @xmath130 at which the actual and the empirical structures are equivalent .
we refer the reader to those articles for more details .
results in @xcite were obtained under the boundedness assumption @xmath131 because the necessary tools from empirical processes theory , like contraction inequalities @xcite , only hold under such an assumption . in particular , these results do not apply even to the gaussian regression model .
the approach developed in this work provides a slight improvement , since risk bounds hold if the envelope function @xmath132 is sub - exponential ( which is the case for the gaussian regression model with respect to the square loss ) .
one should also mention the subtle but significant gap between the _ margin assumption _ and the bernstein condition which we use .
both state that for every @xmath94 , @xmath133 for some constant @xmath134 . however , in the margin condition @xmath49 has the minimal risk _ over all measurable functions _
( for instance , @xmath49 is the regression function in the regression model with respect to the quadratic loss ) , while in a bernstein condition @xmath20 is assumed to minimize the risk _ over @xmath18_. the two conditions are equivalent only when @xmath135 ( and thus @xmath136 ) .
but in general , they are very different . as a simple example , in the bounded regression model [ i.e. , @xmath137 with respect to the quadratic loss , the margin assumption holds with @xmath138 whereas the bernstein condition is not true in general . for more details on the difference between the margin assumption and the bernstein condition
we refer the reader to the discussion in @xcite .
the second type of application we will present deals with nonexact regularized oracle inequalities . usually a model @xmath18 is chosen or constructed according to the belief that an oracle @xmath20 in @xmath18 is close , in some sense , to some minimizer @xmath49 of the risk function in some larger class of functions @xmath139 [ e.g. , in the regression model , @xmath49 can be the regression function and @xmath140 .
hence , by choosing a particular model @xmath141 , it implicitly means that we believe @xmath49 to be close to @xmath18 in some sense .
it is not always possible to construct a class @xmath18 that captures properties @xmath49 is believed to have ( e.g. , a low - dimensional structure or some smoothness properties ) . in such situations , one is not given a single model @xmath18 ( usually the set @xmath139 is too large to be called a model ) , but a functional @xmath142 , called a _
criterion _ , that characterizes each function according to its level of compliance with the desired property and the smaller the criterion , the `` closer '' one is to the property .
for instance , when @xmath139 is an rkhs , one can take @xmath143 to be the norm in the reproducing kernel hilbert space , or when @xmath139 is the set of all linear functionals in @xmath144 , one may chose @xmath145 for some @xmath146 $ ] .
the extreme case here is @xmath147 and @xmath148 is the cardinality of the support of @xmath115 ; thus a small criterion means that @xmath115 belongs to a low - dimensional space . instead of considering the erm over the too large class @xmath139 ,
the goal is to construct a procedure having both good empirical performances and a small criterion .
one idea , that we will not develop here , is to minimize the empirical risk over the set @xmath149 @xcite , and try to find a data - dependent way of choosing the radius @xmath150 .
another popular idea is to regularize the empirical risk : consider a nondecreasing function of the criterion called a _ regularizing function _ and denoted by @xmath151 and construct @xmath152 with the obvious extension if the infimum is not attained .
the procedure ( [ eqrerm ] ) is called _ regularized empirical risk minimization procedure _ ( rerm ) .
rerm procedures were introduced to avoid the `` over - fitting '' effect of large models @xcite , and later used to select functions with additional properties , like smoothness ( e.g. , svm estimators in @xcite ) or an underlying low - dimensional structure ( e.g. , the lasso estimator ) . in this setup , we are interested in constructing estimators @xmath27 realizing the best possible trade - off between the risk and the regularizing function over @xmath139 : there exists some @xmath22 such that with high probability @xmath153 using the same terminology as in ( [ eqoracle - inequality ] ) , inequality ( [ eqoracle - ineq - penalized ] ) is called a _
regularized oracle inequality_. when @xmath26 , ( [ eqoracle - ineq - penalized ] ) is called an _ exact regularized oracle inequality _ , and when @xmath28 , ( [ eqoracle - ineq - penalized ] ) is called a _
nonexact regularized oracle inequality_. following our analysis of the erm algorithm , the next result is a regularized oracle inequality for the rerm . but before stating this result , one has to say a word on the way the regularizing function @xmath154 and the criterion @xmath143 are related .
the choice of @xmath154 is driven by the complexity of the sequence @xmath155 of models @xmath156 for any @xmath157 , the complexity of @xmath158 is measured by @xmath159 defined as above for some fixed @xmath84 by @xmath160 hence , @xmath161 is a `` level '' in @xmath162 above which the empirical and the actual structures are equivalent ; namely , with high probability , on the set @xmath163 , @xmath164 thus , the function @xmath165 captures the `` isomorphic profile '' of the collection @xmath166 .
up to minor technical adjustments , the regularizing function , defined formally in ( [ eqpenalty - tilde ] ) , is @xmath167 .
we will study two separate situations , both motivated by the applications we have in mind . in the first ,
@xmath143 will be uniformly bounded and may only grow with the sample size @xmath25that is , there is a constant @xmath168 satisfying that for every , @xmath169 .
the second case we deal with is when the `` isomorphic profile '' @xmath165 tends to infinity with @xmath150 . for technical reasons
, we also introduce an auxiliary function @xmath170 , defined in the following assumption .
[ assthm - b ] assume that for every @xmath171 , @xmath172 a.s . and that there are nondecreasing functions @xmath173 and @xmath174 such that for every @xmath157 and every @xmath175 , @xmath176 let @xmath84 and consider a function @xmath177 nondecreasing in its first argument and such that , for any @xmath157 and @xmath42 , @xmath178 assume that either : * there exists @xmath179 such that for every @xmath180 and in this case define @xmath181 , for all @xmath84 and @xmath42 , or * the function @xmath165 tends to infinity with @xmath150 and there exists @xmath182 such that @xmath183 , for all @xmath157 and @xmath42 and , in this case , let @xmath184 be any function in @xmath185 and define @xmath170 such that , for every @xmath42 and @xmath84 , @xmath186,\nonumber\end{aligned}\ ] ] where @xmath187 is the generalized inverse function of @xmath127 [ i.e. , @xmath188 , for all @xmath189 and @xmath190 is some absolute constant .
[ theob ] there exist absolute positive constants @xmath191 , @xmath192 @xmath193 and @xmath190 for which the following holds . under assumption [ assthm - b ] , for every @xmath42 and @xmath194 with probability greater than @xmath195 , @xmath196.\end{aligned}\ ] ] fortunately , @xmath170 usually has little impact on the resulting rates . for instance , in the main application we will present here , . like in theorem [ theoa ] ,
the bernstein - type condition @xmath197 holds when @xmath9 is nonnegative and sub - exponential for @xmath198 .
therefore , and contrary to the situation in exact oracle inequalities , the `` geometry '' of the family of classes @xmath155 does not play a crucial role in the resulting nonexact regularized oracle inequalities .
observe that now the choice of the regularizing function in terms of the criterion is now made explicit : @xmath199 the formulation of theorem [ theob ] seems cumbersome , but it is not very difficult to apply it and here we will present one application dealing with high - dimensional vectors of short support .
other applications on matrix completion , convex aggregation and model selection can be found in @xcite .
formally , let @xmath200 be @xmath7 i.i.d .
random variables with values in @xmath201 , and denote by @xmath202 the marginal distribution of @xmath51 .
the dimension @xmath203 can be much larger than @xmath25 but we believe that the output @xmath50 can be well predicted by a sparse linear combination of covariables of @xmath51 ; in other words , @xmath50 can be reasonably approximated by @xmath204 for some @xmath205 of short support ( even though we will not require any assumption of this type to obtain our results ) .
these kind of problems are called `` high - dimensional '' because there are more covariables than observations .
nevertheless , one hopes that under the structural assumption that @xmath50 `` depends '' only on a few number of covariables of @xmath51 , it would still be possible to construct efficient statistical procedures to predict @xmath50 . in this framework , a natural criterion function is the @xmath206 function measuring the size of the support of a vector .
but since this function is far from being convex , using it in practice is hard ; see , for example , @xcite .
therefore , it is natural to consider a convex relaxation of the @xmath206 function as a criterion : the @xmath0 norm @xcite . in what follows
, we will apply theorem [ theob ] to establish nonexact regularized oracle inequalities for @xmath0-based rerm procedures , and with fast error rates a residual term that tends to @xmath207 like @xmath1 up to logarithmic terms .
the regularizing function resulting from theorem [ theob ] for the @xmath2-loss ( @xmath208 ) will be the @xmath209th power of the @xmath0-norm .
in particular , for the quadratic loss , we regularize by @xmath210 , the _ square of the @xmath0-norm _ , @xmath211 while the standard lasso is regularized by the @xmath0 norm itself .
this choice of the exponent is dictated by the complexity of the underlying models : the sequence of balls @xmath212 trough the isomorphic profile function @xmath165 .
observe that since @xmath213 when @xmath214 , a nonexact oracle inequality for the lasso estimator itself follows from theorem [ theob ] , but with a slow rate of @xmath36 .
using the @xmath209th power of the @xmath0-norm as a penalty function for the @xmath2-risk yields a fast @xmath1 rate ( see theorem [ theoc ] ) .
we will perform this study for the @xmath2-loss function , and in which case , for every @xmath215 , @xmath216 the following result is obtained only under the assumption that @xmath50 and @xmath217 belong to @xmath218 .
since there are no `` statistically reasonable '' @xmath219 variables for , it sounds more `` statistically relevant '' to assume that @xmath220 , @xmath221 are almost surely bounded when one wants results for the @xmath2-risk with @xmath222 , or that the functions are in @xmath223 for @xmath224 ( e.g. , linear models with sub - gaussian noise and a sub - gaussian design satisfy this condition ) .
[ theoc ] let @xmath208 .
there exist constants @xmath191 and @xmath192 that depend only on @xmath209 for which the following holds .
assume that there exists @xmath225 such that @xmath226 , @xmath227 .
for @xmath42 and @xmath228 , let @xmath229 and consider the rerm estimator @xmath230 then , with probability greater than @xmath195 , the @xmath2-risk of @xmath231 satisfies @xmath232 where @xmath233 .
procedures based on the @xmath0-norm as a regularizing or constraint function have been studied extensively in the last few years .
we only mention a small fraction of this very extensive body of work @xcite .
in fact , it is almost impossible to make a proper comparison even with the results mentioned in this partial list .
some of these results are close enough in nature to theorem [ theoc ] to allow a comparison .
in particular , in @xcite , the authors prove that with high probability , the lasso satisfies an exact oracle inequality with a residual term @xmath234 up to logarithm factors , under tail assumptions on @xmath50 and @xmath51 . in
@xcite , upper bounds on the risks @xmath235 $ ] and @xmath236 were obtained for a weighted lasso @xmath231 when @xmath237 for @xmath238 with short support .
exact oracle inequalities for rerm using an entropy - based criterion or on an @xmath239 criterion ( with @xmath240 close to @xmath241 ) were obtained in @xcite for any convex and regular loss function and with fast rates .
similar bounds were obtained in @xcite for a rerm using a weighted @xmath0-criterion . in @xcite
it is shown that the lasso and dantzig estimators @xcite satisfy oracle inequalities in the deterministic design setup and under the rec condition .
in fact , in most of these results the authors obtained exact oracle inequalities with an optimal residual term of @xmath242 , which is clearly better than the rate @xmath243 obtained in theorem [ theoc ] for the quadratic loss and in the same context .
however , it is important to note that all these exact oracle inequalities were obtained under an assumption that is similar in nature to the restricted isometry property ( rip ) , whereas in theorem [ theoc ] one does not need that kind of assumption on the design .
although it seems strange that it is possible to obtain fast rates without rip there is nothing magical here .
in fact , the isomorphic argument used to prove theorem [ theob ] ( and thus theorem [ theoc ] ) shows that the random operator @xmath244 satisfies some sort of an rip , which actually coincides with the rip property in the noise - free case @xmath245 for an isotropic design .
this indicates that rip is not the key property in establishing oracle inequalities for the prediction risk , but rather , the `` isomorphic profile '' of the problem at hand , which takes into account the structure of the class of functions . finally , a word about notation .
throughout , we denote absolute constants or constants that depend on other parameters by @xmath246 , @xmath40 , @xmath192 , @xmath247 , etc .
( and , of course , we will specify when a constant is absolute and when it depends on other parameters ) .
the values of these constants may change from line to line .
the notation @xmath248 ( resp . , @xmath249 ) means that there exist absolute constants @xmath250 such that @xmath251 ( resp . , @xmath252 ) .
if @xmath253 is a parameter , then @xmath254 means that @xmath255 for some constant @xmath256 depending only on @xmath257 .
we denote by @xmath258 the space @xmath144 endowed with the @xmath239 norm @xmath259 .
the unit ball there is denoted by @xmath260 and the unit euclidean sphere in @xmath144 is @xmath261 .
in this section we obtain a general bound on @xmath262 for the @xmath2-loss when @xmath208 , and show that a bernstein - type condition is satisfied under weak assumption on the loss function .
the _ isomorphic property _ of a functions class measures the `` level '' at which empirical means and actual means are equivalent .
the notion was introduced in this context in @xcite .
although it is not a necessary feature of this method , if one wishes the isomorphic property to hold with exponential probability , one can use a high probability deviation bound on the supremum of the localized process . a standard way
( though not the only way , or even the optimal way ! ) of obtaining such a result is through of talagrand concentration inequality @xcite applied to localizations of the function class , combined with a good control of the variance in terms of the expectation ( a bernstein - type condition ) . when applied to an excess loss class , this argument leads to exact oracle inequalities ; see , for example , @xcite .
here we are interested in nonexact oracle inequality , and thus , we will study the isomorphic properties of the loss class . to make the presentation simpler , we are not dealing with a fully `` unbounded theory '' like in @xcite , but rather that the class has an envelope function which is bounded in @xmath263 , and we follow the path of @xcite , in which one obtains the desired high probability bounds using talagrand s concentration theorem .
since we would like to avoid the assumption that the class consists of uniformly bounded functions , an important part of our analysis is the following @xmath73 version of talagrand s inequality @xcite .
[ theoadamcjak ] there exists an absolute constant @xmath264 for which the following holds .
let @xmath6 be @xmath25 i.i.d .
random variables with values in a space @xmath3 , and let @xmath57 be a countable class of real - valued measurable functions defined on @xmath3 . for every @xmath42 and @xmath265 , with probability greater than @xmath266 , @xmath267 using the same truncation argument as in @xcite , it follows that for every single function @xmath268 and every @xmath269 , with probability greater than @xmath266 , @xmath270 and , in particular , if there exists some @xmath81 for which @xmath271 , then for every @xmath272 and @xmath42 , with probability greater than @xmath266 , @xmath273 theorem [ theoadamcjak ] can be extended to classes @xmath57 satisfying some separability property like condition ( m ) in @xcite .
we apply theorem [ theoadamcjak ] in this context and it will be implicitly assumed that every time we use theorem [ theoadamcjak ] , this separability condition holds .
in particular , theorem [ theoadamcjak ] will be applied to the localized sets @xmath274 to get nonexact oracle inequalities for the erm algorithm and to the family @xmath275 to get nonexact regularized oracle inequalities for the rerm procedure .
observe that theorem [ theoadamcjak ] requires that the envelope function @xmath276 is sub - exponential , but since @xmath277 it follows that @xmath278 is not much larger than @xmath279 .
however , this condition can be a major drawback .
for instance , if the set @xmath57 consists of linear functions indexed by the euclidean sphere @xmath280 , and @xmath51 is the standard gaussian measure on @xmath144 , the resulting envelope function is bounded in @xmath281 , but its norm is of the order of @xmath282 . in theorem
[ theoc ] , we bypass this obstacle by assuming that @xmath283 . this assumption is far better suited for situations in which the indexing class is small like localized subsets of @xmath284 that appear naturally in lasso type results .
[ corisomorphy ] let @xmath18 be a functions class and assume that there exists @xmath285 such that for every @xmath82 , @xmath286 .
if @xmath84 and @xmath287 satisfy that @xmath288 then for every @xmath42 , with probability larger than @xmath289 , for every @xmath94 @xmath290 where , for @xmath193 the constant appearing in theorem [ theoadamcjak ] , @xmath291 the proof follows the ideas from @xcite .
fix @xmath292 and @xmath42 , and note that by theorem [ theoadamcjak ] , with probability larger than @xmath266 , @xmath293\\[-8pt ] & & { } + k b_n(v(\ell_f)_{\lambda})\frac x n.\nonumber\end{aligned}\ ] ] clearly , we have @xmath294 and @xmath295 moreover , since @xmath296 is star - shaped , @xmath297 is nonincreasing , and since @xmath298 and @xmath299 , then @xmath300 combined with ( [ eqadamcjak ] ) , there exists an event @xmath301 of probability greater than @xmath266 , and on @xmath301 , @xmath302 hence , on @xmath301 , if @xmath303 satisfies that @xmath304 , then @xmath305 .
moreover , if @xmath306 , then @xmath307 ; hence @xmath308 , and so @xmath309 . in theorem [ theoa ] , the desired concentration properties ( and thus the fast rates in theorem [ theoc ] ) rely on a bernstein - type condition , that for every @xmath94 , @xmath310 assumption ( [ eqbernstein - assumption ] ) is trivially satisfied when the loss functions are positive and uniformly bounded : if @xmath311 , then @xmath312 .
it also turns out that ( [ eqbernstein - assumption ] ) does not require any `` global '' structural assumption on @xmath18 and is trivially verified if class members have sub - exponential tails .
[ lembernstein - condition - under - psi1 ] let @xmath51 be a nonnegative subexponential random variable .
then for every @xmath313 , @xmath314 fix @xmath315 , and note that @xmath316\,dt=\theta^2{\mathbb p}[x\geq\theta]+ 2\int_\theta^\infty t{\mathbb p}[x\geq t]\,dt\nonumber\\ & \leq & 2\theta^2\exp(-\theta/\|x\|_{\psi_1})+4\int _
\theta ^\infty t\exp(-t/\|x\|_{\psi_1})\,dt\\ & \leq & ( 2\theta^2 + 4\theta\|x\|_{\psi_1}+4)\exp(-\theta /\|x\|_{\psi_1}).\nonumber\end{aligned}\ ] ] since @xmath317 , it follows from ( [ eqlemma - bern - main ] ) that , for any @xmath318 , @xmath319 the result follows for @xmath320 . in particular ,
if @xmath321 and @xmath322 for some @xmath323 , then for every @xmath324 , @xmath325 let @xmath326 be the loss class associated with @xmath18 for the erm or with a class @xmath158 for some @xmath157 for the rerm .
the next step is to obtain bounds on the fixed point of the localized process , that is , for some @xmath327 , to find a small @xmath328 for which @xmath329 note that the complexity of the star - shaped hull @xmath330 is not far from the one of @xmath326 itself .
actually , a bound on the expectation of the supremum of the empirical process indexed by @xmath331 will follow from one on @xmath332 for different levels @xmath333 .
this follows from the peeling argument of @xcite : that @xmath334
. therefore , setting @xmath335 , for all @xmath336 and @xmath337 , @xmath338 because if @xmath339 , then the sets @xmath340 are empty .
thus , it remains to bound @xmath341 for any @xmath336 .
let us mention that a naive attempt to control these empirical processes using a contraction argument is likely to fail , and will result in slow rates even in very simple cases ( e.g. , a regression model with a bounded design ) .
we refer to @xcite for more details .
the bounds obtained below on @xmath342 are expressed in terms of a random metric complexity of @xmath326 , which is based on the structure of a typical coordinate projection @xmath343 .
these random sets are defined for every sample @xmath344 by @xmath345 the complexity of these random sets will be measured via a metric invariant , called the @xmath346-functional , introduced by talagrand as a part of the generic chaining mechanism .
[ defgamma2 ] let @xmath347 be a semi - metric space .
an admissible sequence of @xmath348 is a sequence @xmath349 of subsets of @xmath348 such that @xmath350 and @xmath351 for any @xmath352 .
we define @xmath353 where the infimum is taken over all admissible sequences @xmath354 of @xmath348 .
we refer the reader to @xcite for an extensive survey on chaining methods and on the @xmath346-functionals
. in particular , one can bound the @xmath346-functional using an entropy integral @xmath355 where @xmath356 is the minimal number of balls of radius @xmath357 with respect to the metric @xmath203 needed to cover @xmath348 , and @xmath358 is the diameter of the metric space @xmath347 .
we will use the @xmath346-functional to state our theoretical bounds because there are examples in which @xmath359 is significantly smaller than the corresponding entropy integral .
however , in all our concrete applications we will use the bound ( [ eqdudley - gamma2 ] ) since the computation of those is much simpler , the gap is at most logarithmic and the purpose of this note is not to obtain the optimal estimates but to show that the residual terms in exact and nonexact oracle inequalities could be very different .
now , we turn to some concrete examples where @xmath326 is the loss functions class in the regression model with respect to the @xmath2-loss .
let @xmath208 and set the @xmath2-loss function of @xmath11 to be @xmath360 . in this case
, the @xmath2-loss functions class localized at some level @xmath361 is @xmath362 .
the following result is a combination of a truncation argument and rudelson s @xmath363 method . to formulate it ,
set @xmath364 , for any , let @xmath365 , and if @xmath366 , put @xmath367 .
[ propmark - learning ] for every @xmath208 , there exists a constant @xmath191 depending only on @xmath209 for which the following holds . if @xmath18 is a class of functions , then for any @xmath336 : if @xmath224 , then @xmath368 $ ] , if @xmath222 , then @xmath369 is upper bounded by @xmath370.\ ] ] let @xmath371 where @xmath372 is a threshold to be fixed later .
for @xmath82 , set @xmath373 , let , and note that @xmath374 .
thus , @xmath375 to upper bound the truncated part of the process , consider the empirical diameter @xmath376 . by the zin
ginn symmetrization theorem @xcite and the upper bound on a rademacher process by a gaussian one , @xmath377 where @xmath378 are @xmath25 independent standard random variables and @xmath379 denotes the expectation with respect to those variables . for a fixed sample @xmath380 ,
let @xmath381 be the gaussian process defined by @xmath382 , @xmath383 . if @xmath384 , then @xmath385 where we have used that @xmath386 for every @xmath387 . by a standard chaining argument it follows that @xmath388 and thus , @xmath389 .
a bound on the diameter follows from ( [ eqgeneric - chaining - truncation-2 ] ) and the contraction principle , @xmath390 implying that @xmath391 and so @xmath392 next , observe that for @xmath224 , the right - hand side in ( [ eqbound - truncation - part ] ) does not depend on the truncation level @xmath393 , and thus one may take @xmath393 arbitrarily large , leading to the desired result . for @xmath394 ,
consider the unbounded part of the process .
since the envelope function of @xmath332 exhibits a subexponential decay , then @xmath395\,dt\\ & = & \theta^q{\mathbb p}\bigl[\sup_{h\in h_\mu}|h|\geq\theta\bigr]+\int _
{ \theta ^q}^\infty{\mathbb p}\bigl[\sup_{h\in h_\mu}|h|^q\geq t\bigr]\,dt\\ & \leq & 2\theta^q\exp(-\theta^q / m)+2m\exp(-\theta ^q / m).\end{aligned}\ ] ] the result follows by taking @xmath396 .
in this section , we will present the proof of theorem [ theoa ] , which follows the same ideas as @xcite for the excess loss .
[ theosmall - lemma - erm ] there exists an absolute constant @xmath80 for which the following holds .
let @xmath18 be a class of functions , and assume that there is some @xmath174 such that for every @xmath82 , @xmath83 .
for @xmath42 and @xmath228 , consider an event @xmath301 on which for every @xmath82 , @xmath397 where @xmath398 is some fixed increasing function .
then , with probability greater than @xmath399 , @xmath400 fix @xmath42 , let @xmath190 be the constant introduced in ( [ eqadamcjak - single - function - under - bernstein ] ) , consider @xmath401 and without loss of generality one assume that the infimum is achieved . by ( [ eqadamcjak - single - function - under - bernstein ] ) [ for @xmath402 , the event @xmath403
on which @xmath404 has probability greater than @xmath266 .
hence , @xmath405 and on @xmath406 , every @xmath11 in @xmath18 satisfies that @xmath407 since @xmath408 , then @xmath409 and the claim now follows from the choice of @xmath49 .
proof of theorem [ theoa ] let @xmath42 , @xmath84 , and put @xmath410 by theorem [ corisomorphy ] , the event @xmath301 , on which every @xmath94 satisfies that @xmath411 has probability greater than @xmath266 .
now , the result follows from lemma [ theosmall - lemma - erm ] . the remark following theorem [ theoa ] , that if @xmath9 is nonnegative , then @xmath412 satisfies a bernstein - type condition with @xmath413 follows from lemma [ lembernstein - condition - under - psi1 ] .
although the proof of theorem [ theob ] seems rather technical , the idea behind it is rather simple .
first , one needs to find a `` trivial '' bound on @xmath414 , giving preliminary information on where one must look for the rerm function ( this is the role played by the function @xmath170 ) .
then , one combines peeling and fixed point arguments to identify the exact location of the rerm .
note that for @xmath415 , we have @xmath416 for all @xmath417 .
therefore , without loss of generality , we can replace the set @xmath139 by @xmath18 in both the definition of the rerm in ( [ eqrerm ] ) and in the nonexact regularized oracle inequality of theorem [ theob ] .
we begin with the following rough estimate on the criterion of the rerm . in the case
where there is a trivial bound @xmath418 , for all @xmath82 then it follows that for any @xmath84 and @xmath42 , @xmath419 . turning to the second case stated in assumption [ assthm - b ] , recall that @xmath165 tends to infinity with @xmath150 and
there exists @xmath182 such that for every @xmath420 , @xmath421 .
hence , for every @xmath42 and @xmath422 , we set @xmath170 to satisfy that @xmath423,\end{aligned}\ ] ] where @xmath184 is any fixed function in @xmath18 ( e.g. , when @xmath424 , one may take @xmath425 ) , and @xmath187 is the generalized inverse function of @xmath127 . in this case , we prove the following high probability bound on @xmath414 .
[ lempen - of - rerm - small]assume that @xmath426 tends to infinity when @xmath150 tends to infinity and that there exists @xmath182 such that for every @xmath420 , @xmath421 . then , under the assumptions of theorem [ theob ] , for every @xmath42 and @xmath422 , with probability greater than @xmath266 , @xmath427 . by the definition of @xmath428 , @xmath429
since @xmath9 is nonnegative , then @xmath430 , and thus @xmath431 since @xmath432 , for all @xmath157 , one of the following two situations occurs : either @xmath433 or , noting that for every @xmath420 , @xmath434 , then @xmath435 and since @xmath87 is monotone in @xmath150 then @xmath436 . hence , in both cases
@xmath437 on the other hand , according to ( [ eqadamcjak - single - function - under - bernstein ] ) , with probability greater than @xmath266 , @xmath438 .
the result follows by plugging the last inequality in ( [ eqlemma1-eq1 ] ) and since @xmath439 is nondecreasing .
the next step is to find an `` isomorphic '' result for @xmath440 .
the idea is to divide the set given by the trivial estimate on @xmath441 into level sets and analyze each piece separately .
[ lemrisk - rerm ] under the assumptions of theorem [ theob ] , for every @xmath42 , with probability greater than @xmath43 , @xmath442 let @xmath301 be the event @xmath443 and we will show that this event has the desired small probability . clearly , @xmath444\leq{\mathbb
p}[\omega_0(x)\cap\{\operatorname{crit}({\hat f^{\mathrm{rerm}}_n})\leq \alpha_n(\epsilon , x)\}]+{\mathbb p}[\operatorname{crit}({\hat f^{\mathrm{rerm}}_n})>\alpha _ n(\epsilon , x)],\ ] ] and by lemma [ lempen - of - rerm - small ] , @xmath445\leq4\exp(-x)$ ] in the second case of assumption [ assthm - b ] or @xmath446=0 $ ] when there is a trivial bound on the criterion .
therefore , in any case , we have @xmath447\leq4\exp(-x)$ ] .
recall that @xmath448 , for all @xmath449 , and since @xmath87 is monotone in @xmath150 , then @xmath450\\ & & \qquad\leq\sum_{i=0}^{\lfloor\alpha_n(\epsilon , x)\rfloor}{\mathbb p}[\omega _ 0(x)\cap\{i\leq\operatorname{crit}({\hat f^{\mathrm{rerm}}_n})\leq i+1\}]\\ & & \qquad\leq\sum_{i=0}^{\lfloor\alpha_n(\epsilon , x)\rfloor}{\mathbb p}\bigl[\exists f\in f_{i+1}\dvtx\\ & & \hspace*{78.5pt } r(f)\geq(1 + 2\epsilon)r_n(f)+\rho_n\bigl(i+1,x+\log\alpha _ n(\epsilon , x)\bigr)\bigr].\end{aligned}\ ] ] by theorem [ corisomorphy ] , for every @xmath451 and @xmath449 , with probability greater than @xmath452 , for every @xmath453 , @xmath454 .
in particular , @xmath455\\ & & \qquad\leq4\exp\bigl(-\bigl(x+\log\alpha_n(\epsilon , x)\bigr)\bigr).\end{aligned}\ ] ] hence , the claim follows , since @xmath456\\ & & \qquad\leq\sum_{i=0}^{\lfloor\alpha_n(\epsilon , x)\rfloor}4\exp \bigl(-\bigl(x+\log \alpha_n(\epsilon , x)\bigr)\bigr)\leq4\exp(-x).\end{aligned}\ ] ] proof of theorem [ theob ] let @xmath42 and @xmath457 .
without loss of generality , we assume that , for the constant @xmath190 defined in ( [ eqadamcjak - single - function - under - bernstein ] ) , there exists @xmath135 minimizing the function @xmath458 let @xmath403 be the event on which @xmath459 since @xmath460 , then @xmath461 , and by ( [ eqadamcjak - single - function - under - bernstein ] ) [ applied with @xmath462 , @xmath463 .
consider the event @xmath301 , on which @xmath464 and observe that by lemma [ lemrisk - rerm ] , @xmath465\geq 1 - 8\exp(-x)$ ] .
therefore , on @xmath406 , we have @xmath466 where the last inequality follows from the definition of @xmath467 .
hence , by the choice of @xmath49 , it follows that on @xmath468 , @xmath469
theorem [ theoc ] follows from a direct application of theorem [ theob ] , by estimating the specific function @xmath87 and the `` bernstein function '' @xmath470 .
consider the family of models @xmath155 associated with the @xmath471-criterion @xmath472 , where @xmath473 is a linear functional on @xmath144 .
[ lemcomplexity - elastic - net ] there exists an absolute constant @xmath191 for which the following holds . for every @xmath361 and @xmath157 , and every @xmath344
, @xmath474 moreover , if @xmath475 , then @xmath476 the proof of the first part of the claim is rather standard and has appeared in one form or another in several places ; for example , see @xcite .
it follows from ( [ eqdudley - gamma2 ] ) and maurey s empirical method ( cf .
the second part is an immediate corollary of the first one .
proof of theorem [ theoc ] observe that for every @xmath477 , @xmath478 hence , by lemma [ lembernstein - condition - under - psi1 ] , one may take @xmath479 .
next , the @xmath73-norm of the envelope of the class @xmath158 satisfies @xmath480 , and by ( [ eqpeeling - shahar ] ) , proposition [ propmark - learning ] and lemma [ lemcomplexity - elastic - net ] , for every @xmath292 , @xmath481 where @xmath482
. set @xmath483 and observe that @xmath484 . since @xmath485 then one can take @xmath486 .
thus @xmath487 is a valid isomorphic function for this problem .
it is also easy to check that for , @xmath488 .
the result now follows by combining these estimates with theorem [ theob ] .
the goal of this section is to describe the difference between the analysis used in @xcite to obtain exact oracle inequalities for the erm , and the one used in this note to establish nonexact oracle inequalities for the erm ( theorem [ theoa ] ) .
our aim is to indicate why one may get faster rates for nonexact inequalities than for exact ones for the same problem .
one should stress that this is not , by any means , a proof that it is impossible to get exact oracle inequalities with fast rates ( there are in fact examples in which the erm satisfies exact oracle inequalities with fast rates : the linear aggregation problem , @xcite ) .
it is not even a proof that the localization method presented here is sharp .
a detailed study of the isomorphic method and oracle inequalities for a general sub - gaussian case ( i.e. , a sub - exponential squared loss ) , in the sense that the class @xmath18 has a bounded diameter in @xmath223 rather than an envelope function , will be presented in @xcite . however , we believe that this explanation will help to shed some light on the differences between the two types of inequalities , and we refer the reader to @xcite for a more detailed and accurate analysis .
our starting point is the following exact oracle inequality for erm , which is a mild modification of a result from @xcite .
the only difference is that it uses adamczak s @xmath73 version of talagrand s concentration inequality for empirical processes , instead of massart s version .
[ theoexact - oracle - ineq - erm ] there exists an absolute constant @xmath80 for which the following holds .
let @xmath18 be a class of functions and assume that there exists @xmath489 such that for every @xmath490 , @xmath491 .
let @xmath492 be such that @xmath493 , and consider an increasing function @xmath87 which satisfies that , for every @xmath42 , @xmath494 .
then , for every @xmath42 , with probability greater than @xmath43 , the risk of the erm satisfies @xmath495 . a bernstein - type condition
, the essence of which is that it allows one to `` translate '' localization with respect to the loss or the excess loss to a localization with respect to a natural metric . in particular
this leads to the necessary control on the @xmath496 diameter of a random coordinate projection of the localized class .
although the two components seem similar for the exact and nonexact cases , they are very different . indeed , for a nonexact oracle inequality , the bernstein type condition is almost trivially satisfied and requires no special properties on the model / output couple @xmath120as long as the functions involved have well behaved tails . as such , it is an individual property of every class member ; see lemma [ lembernstein - condition - under - psi1 ] . on the other hand , the bernstein condition required for the exact oracle inequality is deeply connected to the geometry of the problem ; see , for example , @xcite .
more accurately , when the target @xmath50 is far from the set of multiple minimizers of the risk , @xmath497 , one can show that a bernstein condition holds for a large variety of loss function @xmath498 .
however , when the target @xmath50 gets closer to the set @xmath499 , the bernstein constant @xmath500 degenerates , and leads to rates slower than @xmath36 even if @xmath18 is a two functions class .
hence , the geometry of the problem ( the relative position of @xmath50 and @xmath18 ) is very important when trying to establish exact oracle inequalities , and the bernstein condition is truly a `` global '' property of @xmath18 .
in particular , this explains the gap that we observed in the example preceding the formulation of theorem [ theoa ] . in that case , the class is a finite set of functions and the set @xmath499 is nonempty .
thus , one can find a set @xmath18 and a target @xmath50 in a `` bad '' position , leading to an excess loss class @xmath501 with a trivial bernstein constant ( i.e. , greater than @xmath502 ) . on the other hand , regardless of the choice of @xmath50 , the bernstein constant of @xmath412 is well behaved .
let us mention that when the gap between exact and nonexact oracle inequalities is only due to the bernstein condition , it is likely that both erm and rerm will be suboptimal procedures @xcite . in particular , when slow rates are due to a lack of convexity of @xmath18 ( which is closely related to a bad bernstein constant of @xmath503 ) , one can consider procedures which `` improve the geometry '' of the model ( e.g. , the `` starification '' method of @xcite or the `` pre - selection - convexification '' method in @xcite ) .
the second aspect of the problem is the fixed point of the localized empirical process .
although the complexity of the sets @xmath501 and @xmath504 seems similar from a metric point of view ( @xmath501 is just a shift of @xmath412 ) the localized star - shaped hull @xmath505 and @xmath506 are rather different .
since there are many ways of bounding the empirical process indexed by these localized sets , let us show the difference for one of the methods based on the random geometry of the classes , and for the sake of simplicity , we will only consider the square loss
. using this method of analysis at hand , the dominant term of the bound on @xmath507 ( for the loss class ) which was obtained in proposition [ propmark - learning ] is @xmath508 a similar bound was obtained for @xmath509 in @xcite and @xcite , in which the dominant term is @xmath510 if this bound is sharp ( and it is in many cases ) , and since @xmath511 is in general a nonzero constant , the fixed point @xmath129 of theorem [ theoexact - oracle - ineq - erm ] is of the order of @xmath512 and thus leads to a rate decaying more slowly than @xmath36 .
in contrast , in the nonexact case one has @xmath513 which is of the order of @xmath1 ( up to logarithmic factors ) when the complexity @xmath514 is `` reasonable . ''
the reason for this gap comes from the observation that functions in the star hull of @xmath412 whose expectation is smaller than @xmath515 are only `` scaled down '' versions of functions from @xmath412 .
in fact , the `` complexity '' of the localized sets below the level of @xmath515 can already be seen at the level @xmath515 . hence , the empirical process those sets index ( when scaled properly ) , becomes smaller with @xmath103 .
in contrast , because there are functions @xmath516 that can have an arbitrarily small expectation , the complexity of the localized subsets of the star hull of @xmath503 ( normalized properly , of course ) can even increase as @xmath103 decreases .
this happens in very simple situations ; for example , even in regression relative to @xmath284 , if @xmath517 , the complexity of the localized sets remains almost stable and starts to decrease only at a very `` low '' level @xmath103 .
this is the reason for the phase transition in the error rate ( @xmath518 ) that one encounters in that problem .
the first term is due to the fact that the complexity of the localized sets does not change as @xmath103 decreases up to some critical level , while the second captures what happens when the localized sets begin to `` shrink . ''
a concrete example of this phenomenon is treated in the supplementary material @xcite in the convex aggregation context . | we show that empirical risk minimization procedures and regularized empirical risk minimization procedures satisfy nonexact oracle inequalities in an unbounded framework , under the assumption that the class has a subexponential envelope function .
the main novelty , in addition to the boundedness assumption free setup , is that those inequalities can yield fast rates even in situations in which exact oracle inequalities only hold with slower rates .
we apply these results to show that procedures based on @xmath0 and nuclear norms regularization functions satisfy oracle inequalities with a residual term that decreases like @xmath1 for every @xmath2-loss functions ( ) , while only assuming that the tail behavior of the input and output variables are well behaved .
in particular , no rip type of assumption or `` incoherence condition '' are needed to obtain fast residual terms in those setups .
we also apply these results to the problems of convex aggregation and model selection . . |
sparse polynomials are a fundamental object in computer algebra .
computer algebra programs including maple , mathematica , sage , and singular use a sparse representation by default for multivariate polynomials , and there has been considerable recent work on how to efficiently store and compute with sparse polynomials @xcite .
however , despite the memory advantage of sparse polynomials , the alternative dense representation is still widely used for an obvious reason : speed .
it is now classical @xcite that two degree-@xmath0 dense polynomials can be multiplied in softly linear time : @xmath1 ring operations , and even better in many cases @xcite . by contrast , two size-@xmath2 sparse polynomials require @xmath3 operations , and this excludes the potentially significant cost of exponent arithmetic .
much of the recent work on sparse arithmetic has focused on `` somewhat dense '' or structured cases , where the sparsity of the product is sub - quadratic @xcite . at the same time , sparse interpolation algorithms , which in the fastest case can learn an unknown @xmath2-sparse polynomial from @xmath4 evaluations , have gained renewed interest @xcite .
most closely related to the current work , @xcite recently presented algorithms to discover the coefficients of a sparse polynomial product , provided a list of the exponents and some preprocessing . in the context of pattern matching problems
, @xcite gave a las vegas algorithm to multiply sparse polynomials with nonnegative integer coefficients whose cost is @xmath5 .
a remaining question is whether output - sensitive sparse multiplication is possible in time comparable to that of dense multiplication .
this paper answers that question , with three provisos : first , our complexity is proportional to the `` structural sparsity '' of the output that accounts for exponent collisions but not coefficient cancellations ; second , our algorithms are randomized and may produce incorrect results with controllably low probability ; and third , we ignore logarithmic factors in the size of the input . to explain the first proviso , define for a polynomial @xmath6 its _ support _
@xmath7 to be the set of exponents of nonzero terms in @xmath6 .
the _ sparsity _ of @xmath6 , written @xmath8 , is exactly @xmath9 . for two polyomials @xmath6 and @xmath10 , we have @xmath11 .
but in many cases the set of _ possible exponents _
@xmath12 is much smaller than @xmath13 .
this _ structural sparsity _
@xmath14 , is an upper bound on the actual sparsity @xmath15 of the product .
strict inequality @xmath16 occurs only in the presence of _ coefficient cancellations_. part of our algorithm s cost depends only on the actual sparsity , and part depends on the potentially - larger structural sparsity .
our algorithms have not yet been carefully implemented , and we do not claim that they would be faster than the excellent software of @xcite and others for a wide range of practical problems .
however , this complexity improvement indicates that the barriers between sparse and dense arithmetic may be weaker than we once thought , and we hope our work will lead to practical improvements in the near future .
our main algorithm is summarized in theorem [ thm : main ] . here and throughout
, we rely on a version of `` soft - oh '' notation that also accounts for a bound @xmath17 on the probability of failure : @xmath18 , for any function @xmath19 , where @xmath20 means @xmath21 for some fixed @xmath22 ( * ? ? ?
* see sec .
25.7 ) .
[ thm : main ] given @xmath23 $ ] with degree bound @xmath24 and height bound @xmath25 , and @xmath26 , algorithm [ proc : sparsemulzz ] correctly computes the product @xmath27 with probability exceeding @xmath28 , using worst - case expected @xmath29 bit operations , where @xmath30 and @xmath31 are the actual and structural sparsity of the product , respectively .
our algorithm relies on two subroutines , both of which are based on techniques from sparse interpolation and rely on number - theoretic results on the availability of primes .
the first subroutine [ proc : sumset]@xmath32 computes the _ sumset _ of two sets of integers @xmath33 and @xmath34 , defined as @xmath35 this algorithm , which may be of independent interest , has softly - linear complexity in the size of the output @xmath36 .
the second subroutine [ proc : knownsupp](@xmath37 ) requires a set containing @xmath38 in order to compute @xmath39 in time softly - linear in the input and output sizes .
it is based on an algorithm in @xcite , but is more efficient for large exponents .
the main steps of our multiplication algorithm are : 1 .
use [ proc : sumset ] to compute @xmath40 .
2 . run [ proc : knownsupp ] with @xmath41 but with smaller coefficients , to discover the true @xmath38 .
3 . run [ proc : knownsupp ] again , with the smaller exponent set @xmath38 but with the full coefficients .
steps 1 and 2 work with a size-@xmath2 exponent set but with small coefficients , and both contribute @xmath42 to the overall bit complexity . step 3 uses the size-@xmath43 true support but with the full coefficients , and requires @xmath44 bit operations , for a total of @xmath45 .
section [ sec : prelim ] states our notational conventions and some standard results , and section [ sec : redexp ] contains the technical number theoretic results on which we base our randomizations .
section [ sec : basecase ] revisits and adapts our sparse interpolation algorithm from issac 2014 that will be a subroutine for our sumset algorithm , presented in section [ sec : sumset ] .
our new method to find the coefficients , once the support is known , is presented in section [ sec : knownsupp ] .
this is then used in concert with our sumset algorithm in section [ sec : smul ] to describe fully the algorithm of theorem [ thm : main ] , and also to explain how this can be easily extended to output - sensitive sparse multiplication over @xmath46 $ ] , where @xmath47 is @xmath48 , @xmath49 , or @xmath50 .
we count the cost of our algorithms in terms of bit complexity on a random - access machine .
we state their costs using @xmath51 notation , meaning that our algorithms have a factor @xmath52 in the running time .
we can make @xmath53 by running the entire algorithm with error bound @xmath54 some @xmath55 times , then returning the most frequent result .
our main algorithm depends on an unknown number - theoretic constant , as discussed in section [ sec : redexp ] .
thus we have proven only the _ existence _ of a monte carlo algorithm .
we also discuss how this could be easily handled in practice .
our randomized procedures return either the correct answer ( with controllable probability @xmath28 ) , or an incorrect answer , or the symbol @xmath56 .
whenever a subroutine returns @xmath56 , we assume the calling procedure returns @xmath56 as well .
we let @xmath57 denote a commutative ring with identity . for @xmath58
$ ] we let @xmath59 $ ] denote the ideal generated by @xmath6 . for @xmath60 , @xmath61 ,
we let @xmath62 and @xmath63 denote the integers @xmath64 and @xmath65 respectively , such that @xmath66 . we write
@xmath48 for @xmath67 , typically represented as @xmath68 . unless otherwise stated we assume @xmath69 $ ] is of the form @xmath70 with coefficients @xmath71 and exponents @xmath72 .
often we assume each @xmath73 and the exponents are sorted @xmath74 , but it is sometimes useful to relax these conditions .
we write @xmath75 and @xmath76 for the sparsity and degree bounds .
when @xmath77 we write @xmath78 for the _ height _ of @xmath6 .
we also use the norm @xmath79 . more generally , we consider multivariate laurent polynomials in the case @xmath77 , the _ sparse representation _ of @xmath6 consists of a tuple @xmath80 , followed by a list of @xmath43 tuples @xmath81 , where each @xmath82 is stored using @xmath83 bits and each @xmath84 is stored using @xmath85 bits . when @xmath57 is instead a finite ring , @xmath86 is omitted and each @xmath82
is stored using @xmath87 bits . when multiplying @xmath23 $ ] , we assume shared bounds @xmath86 and
@xmath0 so that total input / output size is @xmath88 given that @xmath89 .
the _ dense representation _ of an @xmath90-variate polynomial @xmath6 is an @xmath90-dimensional array of @xmath91 coefficients , where exponents are implicitly stored as array indices . in the case
that @xmath77 with bound @xmath86 as above , this requires @xmath92 bits .
the terms `` sparse polynomial '' and `` dense polynomial '' refer only to the choice of representation , and not to the relative number of nonzero coefficients .
typically , we assume @xmath6 is sparse and reserve @xmath93 to indicate a dense polynomial .
converting between the sparse and dense representations has softly linear cost in the combined input / output size .
when computing a sumset @xmath94 , we assume that every integer in @xmath95 or @xmath96 is represented using @xmath85 bits , where @xmath97 .
we cite the following results from integer and polynomial arithmetic , which we use throughout .
assume @xmath98 $ ] as in with bounds @xmath0 , @xmath43 , and @xmath86 as described above .
our algorithm performs arithmetic on modular images of @xmath6 . for @xmath99
$ ] , we represent @xmath100/{\langle}x^p-1 { \rangle}$ ] by the remainder from dividing @xmath101 by @xmath102 .
note we treat @xmath103 and @xmath104 as elements of @xmath105 $ ] and @xmath105/{\langle}x^p-1{\rangle}$ ] respectively . to reduce a sparse polynomial @xmath106
, we reduce each exponent modulo @xmath107 , and then add like - degree terms .
by fact [ fact : int_arithmetic ] , we have : [ cor : modxp_m ] given any @xmath99 $ ] , we can compute @xmath108 using @xmath109 bit operations .
we first recall how to choose a random prime number .
[ fact : rosser ] if @xmath110 , then the number of primes in @xmath111 $ ] is at least @xmath112 .
we test if @xmath107 is prime in @xmath113 time via the method in @xcite .
this test and fact [ fact : rosser ] lead to procedure [ proc : getprime ] .
@xmath56 [ lem : getprime_cost ] [ proc : getprime](@xmath114 ) works as stated and has bit complexity @xmath115 the stated cost follows from fast primality testing due to @xcite .
the probability that any chosen @xmath107 is prime is at least @xmath116 , from fact [ fact : rosser ] .
therefore , using the fact that @xmath117 for any nonzero @xmath118 , the probability that none of the chosen @xmath107 are prime is at most @xmath119 as desired .
it is frequently useful to choose a random prime that divides very few of the integers in some unknown set @xmath120 .
if a fraction of @xmath121 integers in @xmath122 do not vanish modulo @xmath107 , then we call @xmath107 a _ @xmath121-vanish - prime _ for @xmath122 .
we call a @xmath123-vanish - prime for @xmath122 a _ good vanish - prime_. procedure [ proc : vanp ] shows how to choose a random @xmath121-vanish - prime .
@xmath124 [ proc : getprime](@xmath125 , @xmath126 ) [ lem : vanp ] procedure [ proc : vanp ] works as stated to produce a @xmath121-vanish - prime @xmath107 satisfying @xmath127 and has bit complexity @xmath128 .
let @xmath122 be any subset of integers with @xmath129 and @xmath130 .
write @xmath131 , and write @xmath132 for the number of `` bad primes '' for which more than @xmath133 elements of @xmath122 vanish modulo @xmath107 .
since each @xmath134 , this means that @xmath135 , and because @xmath136 , @xmath137 is an upper bound on the number of bad primes .
if @xmath138 is very small , we instead use a similar argument to say that the number of primes for which _ any _ element of @xmath122 vanishes is at most @xmath139 .
then fact [ fact : rosser ] guarantees the prevalence of bad primes among all primes in @xmath140 is at most @xmath126 , so the probability of getting a bad prime , or of erroneously returning a composite @xmath107 , is bounded by @xmath17 .
a closely related problem is to choose @xmath107 so that most integers in a set @xmath122 are unique modulo @xmath107 .
we say that @xmath141 _ collides _ modulo @xmath107 if there exists @xmath142 with @xmath143 .
we say @xmath107 is a _ @xmath121-difference - prime for @xmath122 _ if the fraction of integers in @xmath122 which do not collide modulo @xmath107 is at least @xmath121 .
a 1-difference - prime is called a _ good difference - prime for _ .
procedure [ proc : diffp ] shows how to compute difference - primes , conditioned on the _ diameter _ of the unknown set @xmath122 : @xmath144 [ lem : diffp ] procedure [ proc : diffp ] has bit complexity @xmath128 and works as stated to produce a @xmath121-difference - prime @xmath107 satisfying @xmath145 and @xmath146 let @xmath122 be any set as described .
an element @xmath141 collides modulo @xmath107 iff the product of differences @xmath147 vanishes modulo @xmath107 .
if @xmath148 this can never happen . otherwise , as each such product is at most @xmath149 , the result follows from the lemma [ lem : vanp ] , setting the @xmath0 of the lemma to @xmath150 .
@xmath151 our algorithms often perform arithmetic modulo @xmath102 .
similar to the notion of collisions above for a set of integers modulo @xmath107 , we say two distinct terms @xmath152 and @xmath153 of @xmath69 $ ] _ collide _ modulo @xmath102 if @xmath154 . essentially , reduction modulo @xmath102 `` hashes '' exponent @xmath155 to @xmath156 .
if @xmath107 is a good difference - prime for @xmath7 and @xmath157 is a good vanish - prime for the coefficients of @xmath6 , then @xmath158 with coefficients reduced modulo @xmath157 has the same sparsity as @xmath6 itself .
sometimes we implicitly reduce exponents modulo @xmath107 by evaluating at @xmath107th roots of unity . in such cases we need to construct primes @xmath157 such that @xmath159 , and to find @xmath107th roots of unity modulo each @xmath157 . in principle , this procedure is no different than the previous ones , as there is ample practical and theoretical evidence to suggest that the prevalence of primes in arithmetic progressions without common divisors is roughly the same as their prevalence over the integers in general .
however , the closest to fact [ fact : rosser ] that we can get here is as follows , which is a special case of lemma 7 in @xcite .
[ lem : arithprog ] there exists an absolute constant @xmath160 such that , for all @xmath161 , and for all but at most @xmath162 primes @xmath107 in the range @xmath163 $ ] , there are at least @xmath164 primes @xmath157 in the range @xmath165 $ ] such that @xmath159 .
set @xmath166 , which means @xmath167 .
fixing @xmath168 , and for any @xmath169 , lemma 7 in @xcite guarantees the existence of positive constants @xmath170 and @xmath171 such that the following holds : for all @xmath172 , and for all but @xmath173 integers @xmath174 $ ] , there are at least @xmath175 primes @xmath157 in the range @xmath176 $ ] such that @xmath177 , where @xmath178 is the euler totient function . setting @xmath179 , and letting @xmath180 , @xmath181 , and @xmath182 , the statement of our lemma holds because @xmath183 thus @xmath184 @xmath185 @xmath186 @xmath187 @xmath56 lemma [ lem : arithprog ] forms the basis for algorithm [ proc : getpqw ] , where we assume that the constant @xmath160 is given .
since this constant has not actually been computed , a reasonable strategy would be to choose some small `` guess '' for @xmath160 and run the algorithm until it does not report failure .
if the algorithm fails , it could be due to the random prime @xmath107 being an `` exception '' in lemma [ lem : arithprog ] , or due to unlucky guesses for the primitive roots @xmath188 , or due to the guessed constant @xmath160 being too small . because our primality tests are deterministic , failure due to @xmath160 being too small is detectable by the algorithm returning @xmath56 .
we state the running time and correctness , assuming @xmath160 is sufficiently large , as follows .
[ lem : getpqwcost ] procedure [ proc : getpqw ] has worst - case bit complexity with probability at least @xmath28 , it returns a good difference - prime @xmath107 for @xmath6 , primes @xmath189 such that @xmath190 , and @xmath107th primitive roots modulo each @xmath191 , @xmath192 .
the lower bound @xmath193 guarantees that there are sufficiently many even integers in the range @xmath194 $ ] in order for step [ getpqw : choosea ] to be valid , since for any @xmath195 , we have @xmath196 . for the running time , the outer loop does not affect the complexity in our notation because @xmath197 .
observe also that @xmath198 the running time is dominated by the aks primality tests in the inner loop , which are performed @xmath199 times , each at cost @xmath200 , giving the stated worst - case bit complexity .
all of the checks for primality of @xmath107 and @xmath191 s , as well as the test that each @xmath201 is a @xmath107th primitive root of unity modulo @xmath191 , are deterministic .
therefore the only possibility that the algorithm returns an incorrect result other than @xmath56 is the probability that @xmath107 is not a good difference - prime for @xmath7 . according to the proof of lemma [ lem : diffp ] , the condition @xmath202 , and using the union bound over all outer loop iterations , the probability that _ any _ of the chosen @xmath107 s is not a good difference - prime is less than @xmath126 . consider next a single iteration of the outer loop .
this will produce a valid output unless insufficiently many good @xmath191 s and @xmath201 s are found for that choice of @xmath107 .
from fact [ fact : rosser ] and lemma [ lem : arithprog ] , the probability that @xmath107 is an exception to the lemma is at most @xmath203 , which is less than @xmath204 from the bound @xmath205 .
if @xmath107 is not an exception , then lemma [ lem : arithprog ] tells us that the probability of each @xmath157 being prime is at least @xmath206 .
when @xmath157 is prime , since prime @xmath107 divides @xmath207 , the probability that each @xmath208 is a @xmath107-pru in @xmath209 is @xmath210 , easily making the total probability of successfully adding to @xmath211 and @xmath212 at each loop iteration at least @xmath213 .
let @xmath214 be the size of @xmath33 . by hoeffding s inequality ( @xcite , thm .
1 ) , the probability that fewer than @xmath215 integers are added to @xmath211 after all iterations of the inner loop is at most @xmath216 where the last inequality holds because @xmath217 . therefore , with probability at least @xmath218 , and using again @xmath219 , at least @xmath220 integers are added to @xmath211 each time through the inner loop . since each @xmath221
, this means @xmath222 , and the algorithm will return on step [ getpqw : checkret ] .
combining with the probability that @xmath107 is an exception , we conclude that the probability the algorithm does _ not _ return in each iteration of the outer loop is at most @xmath223 .
as this is repeated @xmath224 times , the probability is less than @xmath126 that the algorithm returns @xmath56 . using the union bound with the probability that any @xmath107 is not a good difference - prime
, we have the overall failure probability less than @xmath17 .
let @xmath225 $ ] be sparse polynomials with @xmath226 and @xmath227 . the subroutine [ proc : sumset ] computes @xmath40 by first reducing the degrees and heights of the input polynomials and then multiplying them .
however , it can not perform the multiplication using a recursive call to [ proc : sparsemulzz ] because the degrees are never reduced small enough to allow the use of dense arithmetic in a base case .
instead , we present here a `` base case '' algorithm which , given @xmath6 , @xmath10 , and a bound @xmath228 , computes @xmath39 , in time softly linear in @xmath229 , and @xmath230 . any algorithm with such running time suffices ; we will use our own from @xcite , which is a monte carlo sparse interpolation algorithm for univariate polynomials over finite fields . to adapt @xcite for multiplication over @xmath231 $ ]
, we first choose a `` large prime '' @xmath232 and treat @xmath233 and their product @xmath27 as polynomials over @xmath234 .
this size of @xmath157 ensures that no extension fields are necessary .
the subroutine [ proc : bb ] specifies how the unknown polynomial @xmath235 $ ] will be provided to the algorithm .
it exactly matches the sorts of black - box evaluations that @xcite requires .
the entire procedure is stated as [ proc : sm1 ] .
@xmath236 @xmath237 via dense arithmetic sparse representation of @xmath238 [ lem : bb ] the algorithm [ proc : bb ] works correctly and uses @xmath239 bit operations .
correctness is clear . to compute @xmath240 ,
we replace every term @xmath152 of @xmath6 and @xmath10 with @xmath241 .
this costs @xmath242 by binary powering . reducing modulo @xmath243 costs @xmath244 by corollary [ cor : modxp_m ] .
dense arithmetic costs @xmath245 bit operations by fact [ fact : dense_arithmetic ] . summing these costs
yields @xmath246 .
@xmath247[proc : getprime]@xmath248 call procedure majorityvotesparseinterpolate from @xcite , with coefficient field @xmath209 , sparsity bound @xmath43 , degree bound @xmath249 , error bound @xmath250 , and black box [ proc : bb]@xmath251 .
the algorithm [ proc : sm1 ] correctly returns the product @xmath27 with probability at least @xmath28 , and has bit complexity @xmath252 in order to use [ proc : bb ] in the algorithm from @xcite , we simply replace the straight - line program evaluation on the first line of procedure computeimage with our procedure [ proc : bb ] .
again , note that as the prime @xmath157 was chosen with @xmath253 , the majorityvotesparseinterpolate algorithm does not need to work over any extension fields .
the correctness is guaranteed by the previous lemma , as well as theorem 1.1 in @xcite .
for the bit complexity , in section 7 of @xcite , we see that the cost is dominated by @xmath254 calls to the black box evaluation function , each of which is supplied @xmath255 , and @xmath256 . applying the bit complexity of lemma [ lem : bb ]
gives the stated result .
let @xmath257 be nonempty , @xmath258 , and @xmath259 throughout this section .
we prove as follows : [ thm : supp ] procedure @xmath260 has bit complexity @xmath261 and produces @xmath36 with probability at least @xmath28 .
we compute the sumset @xmath36 as @xmath262 , @xmath263 $ ] , where @xmath264 and @xmath265 . here
@xmath266 has exponents in @xmath267 and @xmath268 .
thus it suffices to construct the exponents of @xmath266 modulo @xmath269 .
moreover , we have @xmath270 , and that @xmath271 we first show how to compute a tighter upper bound on the true value of @xmath272 . to this end ,
let @xmath145 be a good difference - prime for @xmath262 , using the naive bound @xmath273 from , and define the @xmath274 , where @xmath275 $ ] are defined by @xmath276 then @xmath277 and each term @xmath152 of @xmath266 corresponds to either one or two terms in @xmath278 , of degrees @xmath156 and @xmath279 . therefore @xmath280 we will compute an approximation @xmath281 such that @xmath282 , and therefore @xmath283 . to this end
we present a test that , given @xmath284 , always accepts if @xmath285 and probably rejects if @xmath286 .
we do this for @xmath284 initially @xmath287 , doubling whenever the test rejects .
given the current estimate @xmath284 , we next choose a ( 1/2)-difference - prime @xmath157 for the support of any @xmath288-sparse polynomial with degree @xmath289 , and compute @xmath290 .
we work modulo @xmath291 , such that none of the coefficients of @xmath292 vanish modulo @xmath293 .
if @xmath278 is @xmath284-sparse then @xmath292 is as well .
if @xmath278 has @xmath288 terms then , as fewer than @xmath284 terms of @xmath278 are in collisions , @xmath292 is _ not _ @xmath284-sparse . as no terms of @xmath278 vanish modulo @xmath293 , additional terms in @xmath278 can only increase @xmath294 .
we choose @xmath157 so that the test is correct with probability at least @xmath295 . by iterating @xmath296 times , by hoeffding s inequality ,
the probability is at least @xmath297 that the test runs correctly in at least half of the iterations . as @xmath298
, it suffices that the test is correct @xmath299 times .
the [ proc : sumset ] procedure performs this test on lines [ proc : ss : size0][proc : ss : sizen ] . by corollary [ cor : modxp_m ] the respective costs of constructing @xmath300 and @xmath301 are @xmath302 and @xmath303 , and similarly for @xmath304 .
the cost of the dense arithmetic here is @xmath305 .
given that @xmath306 , the total bit - cost of this part is @xmath261 .
armed with the bound @xmath307 , we aim to compute @xmath27 .
our approach is to compute images @xmath308 for an integer @xmath309 .
since the coefficients of @xmath310 are scaled by powers of @xmath311 , a single term @xmath312 in the original polynomial @xmath266 becomes @xmath313 in @xmath278 and @xmath314 in @xmath310 , and if they are uncollided we can discover @xmath315 by computing their quotient .
modulo @xmath316 , this quotient @xmath315 is simply @xmath317 , from which we can obtain the exponent @xmath318 .
this idea is similar to the `` coefficient ratios '' technique suggested by @xcite , but working modulo @xmath316 allows us to avoid costly discrete logarithms .
procedure [ proc : sumset ] contains the complete description .
@xmath319[proc : diffp]@xmath320 @xmath321 @xmath322 [ proc : ss : size0 ] @xmath323 @xmath324 @xmath325[proc : sm1]@xmath326 @xmath327[proc : sm1]@xmath328 @xmath329 empty list of integers @xmath122 [ proc : sumset ] has four steps that are probabilistic : choosing a good difference - prime @xmath107 , estimating the sumset size @xmath259 , and constructing @xmath278 and @xmath310 . as each is set to fail with probability less than @xmath330 , [ proc : sumset ] succeeds with probability at least @xmath28 .
we now analyze the total cost of this algorithm .
[ proc : diffp ] produces @xmath107 of size @xmath331 . constructing @xmath332 at the beginning , and the reduction of @xmath333 modulo @xmath334 at the end , both cost @xmath261 bit operations .
the search for @xmath284 costs @xmath261 from the previous section . finally , as @xmath335 , @xmath336 , and similarly for @xmath337 , @xmath338 , and @xmath339 , the sparse multiplications due to [ proc : sm1 ] also costs @xmath340 bit operations .
these dominate the complexity as stated in theorem [ thm : supp ] .
we turn now to the problem of multiplying sparse @xmath341 $ ] , provided some @xmath342 .
this algorithm is used twice in our overall multiplication algorithm : first with large @xmath343 but small coefficients , then with the actual support @xmath344 but full - size coefficients .
[ thm : knownsupp ] given @xmath345 $ ] and a set @xmath346 such that @xmath347 , the product @xmath39 can be computed in time @xmath348 where @xmath349 and @xmath24 .
this is @xmath51-optimal , as it matches the bit - size of the inputs .
our algorithm requires a small randomly - selected good difference - prime @xmath107 with @xmath350 bits , and a series of pairs @xmath351 , where each @xmath157 is a slightly larger prime with @xmath352 bits , and @xmath353 is an order-@xmath107 element in @xmath209 .
these numbers are provided by [ proc : getpqw ] ( sec .
[ sec : redexp ] ) .
our algorithm works by first reducing the exponents modulo @xmath107 , then repeatedly reducing the coefficients modulo @xmath157 and performing evaluation and interpolation at powers of @xmath353 .
this inner loop follows exactly the algorithm of @xcite and @xcite for applying a transposed vandermonde matrix and its inverse .
since @xmath107 is a good difference - prime for the support of the product , there are no collisions and this gives us each coefficient modulo @xmath157 .
the process is then repeated @xmath354 times in order to recover the full coefficients via chinese remaindering . without affecting the complexity
, we may assume that @xmath122 contains the support of the inputs too , i.e. , @xmath7 and @xmath355 .
we also assume that @xmath356 , such that no @xmath357 is too large to be an exponent of @xmath39 . under these assumptions , and writing @xmath358 , the stated complexity of our algorithm is simply @xmath359 .
the problem of computing the coefficients of a sparse product , once the exponents of the product are given , has been recently and extensively investigated by van der hoeven and lecerf , where they present an algorithm whose bit complexity ( in our notation ) is @xmath360 ( @xcite , corollary 5 ) .
as @xmath361 , the algorithm here saves a factor of at most @xmath362 in comparison , which could be substantial if the exponents are very large .
their algorithm is more efficient if the support superset @xmath122 is _ fixed _ , in which case they can move the most expensive parts into precomputation and compute the result in the same soft - oh time as our approach , @xmath44 .
furthermore , the support bit - length @xmath363 is at most @xmath364 , but could be as small as @xmath365 , for example if the support contains only a single large exponent . in such cases our savings
is only on the order of @xmath366 . applying transposed vandermonde systems , and their inverses , is an important subroutine in sparse interpolation algorithms , and efficient algorithms are discussed in detail by @xcite and @xcite .
we restate the general idea here and refer the reader to those papers for more details .
if @xmath93 is a dense polynomial , it is well known that applying the vandermonde matrix @xmath367 to a vector of coefficients from @xmath93 corresponds to evaluating @xmath93 at the points @xmath368 . applying the inverse vandermonde matrix corresponds to interpolating @xmath93 from its evaluations at those points .
the product tree method can perform both of these using @xmath369 field operations ( @xcite , chapter 10 ) .
if @xmath6 is instead a sparse polynomial @xmath370 , evaluating @xmath6 at consecutive powers of a single high - order element @xmath353 corresponds to multiplication with the transposed vandermonde matrix : @xmath371 the transposition principle tells us it is possible to compute the maps @xmath372 and @xmath373 in essentially the same time as dense evaluation and interpolation . in particular , if the coefficients @xmath82 are in the modular ring @xmath209 , then the transposed vandermonde map and its inverse can be computed using @xmath374 bit operations @xcite .
@xmath375 @xmath376[proc : getpqw]@xmath377 [ proc : ks : choosep ] @xmath378 [ proc : ks : reduce ] @xmath379 list of @xmath43 empty lists of integers @xmath380 procedure [ proc : knownsupp ] works as stated when @xmath342 . in any case
it has bit - complexity @xmath381 we first analyze the probability of failure when @xmath342 .
the randomization is in the choices of @xmath107 , @xmath157 , and @xmath353 ; problems can occur if these lack the required properties .
if @xmath107 is a good difference - prime for @xmath38 , then by definition there will be no collisions in @xmath382 .
furthermore , if @xmath353 is a @xmath107th root of unity modulo @xmath157 , then there are no collisions among the values @xmath383 , so @xmath384 is invertible modulo @xmath157 .
algorithm [ proc : getpqw ] ensures this is the case with high probability , and if so the algorithm here faithfully computes each coefficient @xmath385 modulo @xmath157 .
conversely , if there are no collisions in @xmath382 , and if no zero divisors modulo @xmath157 are encountered in the application of the vandermonde matrix and its inverse , then the algorithm correctly computes then values @xmath386 , even if @xmath107 is not prime or some @xmath387 is not actually a @xmath107th root of unity .
therefore all failures in choosing tuples @xmath388 are either detected by the algorithm or do not affect its correctness . since that is the only randomized step ,
we conclude that the entire algorithm is correct whenever the input exponent set @xmath122 contains the support of the product . for the complexity analysis
first define @xmath389 .
step [ proc : ks : choosep ] costs @xmath390 bit operations by lemma [ lem : getpqwcost ] . reducing each exponent
@xmath391 modulo @xmath107 , on step [ proc : ks : reduce ] , can be done for a total of @xmath392 bit operations .
now we examine the cost of the for loop that begins on step [ proc : ks : eachq ] . as
the exponents are now all less than @xmath107 , computing each @xmath393 on step [ proc : ks : binpow ] requires only @xmath352 operations modulo @xmath157 , for a total of @xmath394 , which is @xmath395 bit operations . from before we know that applying the transposed vandermonde matrix and its inverse takes @xmath396 , or @xmath395 bit operations . because @xmath397 , the loop on step [ proc : ks : eachq ] repeats @xmath354 times , for a total cost of @xmath398 bit operations .
this also bounds the cost of the chinese remaindering in the final loop .
the complete multiplication algorithm over @xmath231 $ ] that was outlined in the introduction is presented as [ proc : sparsemulzz ] .
@xmath399[proc : sumset]@xmath400 [ proc : smz : sumset ] @xmath401 [ proc : smz : height ]
@xmath402[proc : vanp]@xmath403
@xmath404[proc : knownsupp]@xmath405 [ proc : smz : ks1 ] @xmath406 [ proc : knownsupp]@xmath407 unless failure occurs , we have @xmath408 . every coefficient in @xmath266
is a sum of products of coefficients in @xmath6 and @xmath10 , so the value @xmath409 computed on step [ proc : smz : height ] is an upper bound on @xmath410 , and @xmath107 is a good vanish - prime for @xmath266 .
thus @xmath411 , so the final step correctly computes the product @xmath39 . by the union bound , lemma [ lem : vanp ] and theorems [ thm : supp ] and [ thm : knownsupp ] ,
the probability of failure is less than @xmath17 . writing @xmath14 and @xmath15 , we see that @xmath412 , thus steps [ proc : smz : sumset ] and [ proc : smz : ks1 ] contribute @xmath42 to the overall cost , whereas the last step costs @xmath413 bit operations . as @xmath414 ,
the total bit complexity is @xmath45 , as required .
we now consider extensions of this algorithm to positive and negative exponents ( laurent polynomials ) , multiple variables , and other common coefficient rings , using kronecker substitution .
this is stated in the following theorem .
thmextensions let @xmath415 $ ] be sparse laurent polynomials over @xmath47 , where @xmath416 , @xmath48 , @xmath417 , or @xmath49 .
the product @xmath39 can be computed using @xmath418 bit operations , where @xmath31 is the structural sparsity of the product , @xmath419 , and @xmath96 is the largest bit - length of any coefficient in the input or output .
write the output polynomial @xmath27 as @xmath420 where each @xmath71 and each @xmath84 satisfies @xmath421 .
we first apply the kronecker substitution , providing an easily - invertible map between @xmath46 $ ] and @xmath422 $ ] : @xmath423 for @xmath424 .
this increases the degree to @xmath91 , such that the logarithm of this degree @xmath425 , matching the exponent bit - size in the multivariate representation .
the algorithm [ proc : sumset ] already handles negative exponents ( i.e. , laurent polynomials ) explicitly .
the other primary subroutine to [ proc : sparsemulzz ] is [ proc : knownsupp ] , which only uses the exponents in the set @xmath382 , which are reduced modulo @xmath107 and therefore cause no difficulty if they are negative . thus the multiplication algorithms handle univariate laurent polynomials without any changes .
to extend the multiplication algorithm and it subroutines beyond @xmath416 , we use that our algorithm is also softly - linear in the input _
heights_. this allows us to adapt to many coefficient domain that provides a natural mapping to the integers , and to preserve softly - linear time if that mapping provides only a softly - linear increase in size . for a modular ring @xmath426
, we can trivially treat the inputs as actual integers , then reduce modulo @xmath293 after multiplying .
for a finite field @xmath427 , elements are typically represented as polynomials over @xmath428 $ ] modulo a degree-@xmath429 irreducible polynomial , so these coefficients can be converted to integers using a low - degree kronecker substitution . for the rationals @xmath430 , we might choose a prime @xmath157 larger than the product of the largest numerator and denominator in the output , multiply modulo @xmath157 , then use rational reconstruction to recover the actual coefficients .
in all the above cases , there is growth in the bit - length of coefficients , but only in poly - logarithmic terms of input and output size , therefore not affecting the soft - oh complexity .
the only downside is that we are no longer able to split the cost neatly between @xmath431 and @xmath432 because the unreduced integer polynomial product might have nonzero coefficients which are really zeros in @xmath433 .
we thank mark giesbrecht for helpful discussions on the development of this paper ( and much more ) .
matthew t. comer , erich l. kaltofen , and clment pernet .
sparse polynomial interpolation and berlekamp / massey algorithms that correct outlier errors in input values . in _ proc .
issac 12 _ , pages 138145 .
acm , 2012 .
doi : 10.1145/2442829.2442852 .
mickal gastineau and jacques laskar .
highly scalable multiplication for distributed sparse multivariate polynomials on many - core systems . in _ proc .
casc 2013 _ , pages 100115 , new york , 2013 .
springer - verlag .
doi : 10.1007/978 - 3 - 319 - 02297 - 0_8 .
erich kaltofen and lakshman yagati .
improved sparse multivariate polynomial interpolation algorithms . in _ symbolic and algebraic computation _ ,
volume 358 of _ lect .
notes comput .
_ , pages 467474 .
springer berlin / heidelberg , 1989 .
doi : 10.1007/3 - 540 - 51084 - 2_44 . | we present randomized algorithms to compute the sumset ( minkowski sum ) of two integer sets , and to multiply two univariate integer polynomials given by sparse representations .
our algorithm for sumset has cost softly linear in the combined size of the inputs and output .
this is used as part of our sparse multiplication algorithm , whose cost is softly linear in the combined size of the inputs , output , and the sumset of the supports of the inputs . as a subroutine , we present a new method for computing the coefficients of a sparse polynomial , given a set containing its support .
our multiplication algorithm extends to multivariate laurent polynomials over finite fields and rational numbers .
our techniques are based on sparse interpolation algorithms and results from analytic number theory . |
general relativity ( gr ) as a theory that describes the geometrical structure of space - time and its connection with the energy content of the universe has been highly successful in the last one hundred years .
gr is not only able to give explanation of some discrepancies that had been arisen with newtonian theory in the solar system , but also provides a satisfactory description of the cosmological evolution of the space - time .
however , gr is not able to account for the latest cosmological and astrophysical observations with standard matter sources . among these observations
are the anomalous rotational velocities of objects near the edge of the galaxies , the dynamics of galaxies in clusters , formation of large scale structures or gravitational lensing of background objects by galaxy clusters , such as the bullet cluster . for the explanation of these observations ,
more matter than the standard one is required , the so - called dark matter ( dm ) which has an attractive gravitational character that decelerates the expansion of the universe .
although there are many plausible origins for this component @xcite , dm is usually assumed to be in the form of thermal relics that naturally freeze - out with the right abundance in many extensions of the standard model of particles @xcite .
future experiments will be able to discriminate among the large number of candidates and models , such as direct and indirect detection designed explicitly for their search @xcite , or even at high energy colliders , where they could be produced @xcite .
furthermore , the authors of @xcite , recently awarded with the nobel prize in physics , discovered the accelerated expansion of the universe .
this result has been one of the main reasons for the revision of the cosmological evolution as predicted by gr .
neither the properties of usual matter and energy nor the believed properties of dm can account for this acceleration .
one way of circumventing this problem is to postulating a new kind of fluid , the so - called dark energy , which has a non - standard equation of state allowing a positive energy density while having a negative pressure @xcite .
another possibility to generate the present accelerated expansion of the universe is by the modification of gr @xcite .
indeed , this type of modifications could produce an inflationary epoch in the very early universe .
this epoch is postulated to evade the problems with the initially homogeneous and isotropic state of the universe which arise from considering particle horizons in robertson - walker cosmological models . in this work , we will analyze the accelerated expansion of the universe by its description through its effects on a congruence of geodesics without assuming any specific gravitational theory .
this description has the advantage of being coordinate - independent . in an accelerated expansion
it is expected that neighboring geodesics with similar direction will increasingly recede from each other .
this evolution may be studied either through the geodesic deviation equation @xcite or alternatively by the definition of a kinematic quantity , the expansion , which is quite analogous to the spatial divergence of a vector field .
the expansion rate along the congruence will provide us with the appropriate tool to analyze the consequences of the cosmological accelerated expansion of the universe .
this rate is given by the raychaudhuri equation ( re ) , which will be discussed and used throughout this work .
re was first derived by raychaudhuri in @xcite for congruences of timelike geodesics .
later , sachs @xcite introduced the corresponding re for null geodesics .
et al . _
@xcite provided a generalization of hydrodynamics to gr where results analogous to well - known theorems of hydrodynamics using re equation were obtained .
given a congruence of timelike or null geodesics this equation provides the contraction or expansion rate of the congruence according to some kinematic properties of the congruence and the curvature of the space - time where the congruence is embedded .
in gr , the re together with assumed energy conditions guarantee the attractive character of gravity represented by a non - positive contribution to this equation from the space - time geometry . furthermore
, the re has a fundamental role in the demonstration of important theorems due to hawking and penrose .
these theorems assert that singularities of manifolds representing physical space - times are intrinsic to gr and not a special property of particular solutions of einstein s equations provided that some energy conditions hold @xcite . in this investigation
, we shall analyze the evolution of the expansion of a congruence of geodesics .
this is carried out in a robertson - walker cosmological model from a kinematical point of view since no gravitational theory is assumed beyond the equivalence principle .
this work is organized as follows : first , in section [ raychaudhuriequation ] , we present a short review of the derivation of the re for illustrative purposes and stress some aspects usually disregarded in the available literature . then , the kinematics of a congruence of geodesics in a robertson - walker space - time is studied in section [ rwspace - times ] .
we conclude the paper by giving our conclusions in section [ conclusions ] .
the re provides the expansion rate of a congruence of timelike or null geodesics . in the following
, we will summarize the main results for both cases separately according to the notation in @xcite .
for different reviews and approaches to the re , we refer the reader to @xcite . in the following
let us use abstract index notation ( see @xcite for details ) .
thus , an expression like @xmath0 represents the vector itself and not a component in a certain basis . an expression like @xmath1
must be understood as @xmath2 .
when explicit mention to the components of a tensor in a certain basis is made , greek indices will be used , i.e. , @xmath3 .
the principal advantage of this notation is that it distinguishes equations that hold between tensors and equations that only hold for components in a certain basis .
we use a metric signature ( @xmath4 ) and the riemann and ricci tensors and the ricci scalar are defined respectively as @xmath5 , @xmath6 and @xmath7 .
let us consider a congruence of timelike geodesics and its tangent vector field @xmath0 .
it is assumed that the tangent vector field is normalized , i.e. , @xmath8 , which is always feasible by choosing an affine parameterization of the congruence .
let us define the ( 0,2 ) tensor @xmath9 it is easy to verify that @xmath10 the first expression is a consequence of @xmath0 being normalized whereas the second one is the geodesic equation . as usual the projector tensor @xmath11 is defined by @xmath12 this is a spatial metric provided that @xmath0 is hypersurface orthogonal .
this property of the vector field means that there exists a foliation of the manifold by hypersurfaces orthogonal to @xmath0 .
the metric induced by @xmath13 on these hypersurfaces is @xmath11 . for cases where
an orthogonal hypersurface can not be found , the projector tensor @xmath11 is still well - defined as the tensor that projects tensors in a subspace orthogonal to @xmath0 at each point of the manifold .
it is worth noticing that @xmath14 has no components in the direction of @xmath0 according to expression .
consequently , @xmath14 is rather defined in the subspace orthogonal to @xmath0 .
this subspace is the hypersurfaces tangent space foliating the manifold when @xmath0 is hypersurface orthogonal .
finally , let us consider the following quantities @xmath15 @xmath16 @xmath17}\,;\end{aligned}\ ] ] which are the trace ( with respect to the metric @xmath11 , but the trace is the same respect to @xmath13 because of ) , the trace - free symmetric part and the antisymmetric part of @xmath14 respectively .
these quantities are known as the expansion @xmath18 , shear @xmath19 and twist @xmath20 of the congruence .
thus , @xmath14 can be decomposed as @xmath21 which is analogous to the decomposition of the deformation velocity tensor in hydrodynamics .
the meaning of @xmath18 , @xmath19 and @xmath20 is clear from the analogy with hydrodynamics .
the expansion @xmath18 takes into account the divergence or convergence of nearby geodesics .
it may also be interpreted as a relative change in volume of a fluid particle .
the shear @xmath19 provides information about the deformation of the geodesic congruence without change in volume . finally , the twist @xmath20 contains the vorticity of the congruence .
equations for the rate of these kinematic quantities along each geodesic can be derived in the following way .
after some calculus , it can be proved @xmath22 where the leibniz s rule and the geodesic equation satisfied by @xmath0 have been used .
since we are interested in the equation for the expansion @xmath18 , the trace of becomes @xmath23 where @xmath24 is an affine parameter of the geodesics , i.e. , the proper time of an observer moving along a geodesic of the congruence .
equation is known as the re @xcite .
let us recall that the expansion @xmath18 provides a natural coordinate - independent way of measuring the focusing or divergence of free - falling matter following a certain congruence of geodesics in the gravitational field , i.e. , @xmath25 means divergence and @xmath26 means convergence .
the quantities of the r.h.s . of
are scalars , thus any convenient coordinate system can be used to calculate them .
however , it is important to stress that equation provides the expansion rate as seen by an observer moving along a geodesic of the congruence .
the equivalent counterparts for shear @xmath19 and the twist @xmath20 are obtained from equation by taking the trace - free symmetric part and the antisymmetric part respectively .
let us now discuss the character of the contribution of the different terms in expression .
the quantities @xmath27 and @xmath28 are non - negative quadratic scalars .
the fact that these scalars are non - negative is due to the _ spatial _ character of the tensor @xmath14 reflected by .
this can be understood if one considers a framework with the spatial components of @xmath11 , i.e. , the orthogonal components to @xmath0 , taking the values @xmath29 with @xmath30 at a certain point of the manifold , whereas the components in the direction of @xmath0 being zero .
then , taking into account the spatial character of @xmath14 the values of @xmath31 and @xmath32 are the same in this special framework .
therefore , @xmath33 . the same reasoning is valid for @xmath20 .
thus , it is clear that both the shear @xmath19 and the expansion @xmath18 always tend the congruence to focus whereas the twist @xmath20 always tends to make it diverge . for the contribution of space - time geometry ,
i.e. , for the sign of @xmath34 , there appear three possibilities : @xmath35 @xmath36 @xmath37 the implications of the last three inequalities will be the keystone of this investigation .
the effect that the two first possibilities produce in the congruence may be expressed as a tendency to diverge and a tendency to focus respectively .
it can also be understood as accelerated expansion in the case of and decelerated expansion in the case of . on the other hand
, the value of @xmath38 depends both on the point of the manifold where it is calculated and on the direction of the vector field @xmath0 at that point .
thus , a positive contribution to the re may be obtained at some points of the manifold for certain directions whereas may be negative at those points for different directions .
although it is rarely considered , the expression @xmath39 admits a geometrical interpretation @xcite .
let @xmath40 be a unit vector and @xmath41 with @xmath42 an arbitrary set of unit vectors orthogonal to @xmath40 .
the gaussian curvature @xmath43 of the geodesic surface generated by @xmath40 and @xmath41 is @xmath44 where @xmath45 and @xmath46 are the usual eisenhart @xmath47 s which take the values @xmath48 or @xmath49 according to @xmath50 and @xmath51 .
as is widely known , the sign convention for riemann and ricci tensors is arbitrary a priori . however , let us recall that the sign of the gaussian curvature possesses an intrinsic meaning sign - convention independent .
thus , when all these gaussian curvatures @xmath43 are summed one gets @xmath52 where we have considered that @xmath53 is zero and the spectral decomposition of the metric @xmath54 has been used .
manifestly , the sum does not depend on the election of the set of orthogonal unit vectors @xmath41 under consideration .
then , @xmath55 is the sum of all the gaussian curvatures of the geodesics surfaces generated by @xmath0 and any set of orthogonal unit vectors to it .
this has been referred to as the mean curvature in the direction of @xmath0@xcite .
thus , by following our sign convention @xmath56 and the mean curvature @xmath57 in the direction of a timelike vector @xmath0 becomes @xmath58 hence , the contribution of the space - time geometry to the re has a clear geometrical interpretation as the mean curvature @xmath57 in the direction of the tangent vector field @xmath0 .
furthermore , the inequalities , and can be rewritten in terms of the mean curvature @xmath57 in the direction of @xmath0 as follows @xmath59 @xmath60 @xmath61 thus , if the mean curvature for a given timelike vector @xmath0 at one point of the manifold is positive ( negative ) the space - time geometry would tend to make geodesics with tangent vector @xmath0 passing through this point diverge ( focus ) .
this formulation represents a clear advantage over the previous one in terms of @xmath38 since it does not depend upon any sign convention nor upon the metric signature .
let us recall that when energy conditions are assumed in gr , one always obtains and but not .
the fact that in gr @xmath57 is always negative or zero is interpreted as a manifestation of the attractive character of gravity .
let us return to our discussion about the re .
it is clear that the sign of the expansion rate @xmath62 will depend upon all the contributions of the r.h.s of . from their respectively definitions ,
it is obvious that all these terms depend on the congruence of geodesics under consideration .
however , the terms @xmath63 , @xmath27 and @xmath28 depend on the behavior of the congruence in the neighborhood of the point where they are calculated .
let us recall that these terms come from @xmath64 , then the tangent vector field @xmath0 must be known locally in order to calculate its covariant derivative .
nevertheless , the contribution of space - time geometry @xmath57 depends only upon the direction of the congruence at a given point .
thus , all the congruences pointing in the same direction at a given point , will have the same contribution from space - time .
moreover , there are only two terms that may have a positive contribution : the term @xmath28 and the one given by @xmath57 .
specifically , we shall restrict ourselves to congruences with @xmath65 as we shall explain at the end of this section .
therefore , in this type of congruences the only term that may have a positive contribution to is @xmath57 . an equivalent version of the re for null geodesics can be obtained by considering the null tangent vector field @xmath66 to a congruence of null geodesics @xcite . as expected
the result is @xmath67 where the quantities @xmath68 , @xmath69 and @xmath70 are defined with respect to the congruence of null geodesics , the tensor @xmath71 is the tensor analogous to @xmath11 for null geodesics and @xmath72 is an affine parameter in the following sense .
when null geodesics are considered the tangent vector field has always zero norm and consequently an affine parameter defined as the one which normalizes the tangent vector field does not hold .
for this reason , an affine parameter can be defined in a more general way as the one preserving the geodesic equation in its usual form .
this definition remains valid both for timelike and null geodesics .
the factor @xmath73 in instead of @xmath74 appearing in is due to the fact that the involved subspace in the case of null geodesics is two - dimensional rather than three - dimensional .
thus , equation is the re for null geodesics . finally , let us remind that the same type of space - time geometry contributions provided by the inequalitites , and applies replacing @xmath39 by @xmath75 .
however , an analogous interpretation of the term @xmath75 as a mean curvature is not practicable in this case @xcite . at this stage , we estimate important to stress that the raychaudhuri equations and are identities , i.e. , geometrical statements which do not require any underlying gravitational theory to be derived .
they are quite general and apply whenever a riemannian geometry for the space - time is assumed . a more general equation for the evolution of the expansion of a congruence results if the assumption of being a congruence of geodesics is discarded .
moreover , the re can be generalized to spaces with torsion @xcite . for our purposes and are sufficient .
as was briefly mentioned , there are analogous equations to and for the shear @xmath19(@xmath68 ) and twist @xmath20(@xmath69 ) that provide the rate of these quantities along geodesics .
these equations form a coupled system of non - linear first order differential equations .
they also receive the generic name of raychaudhuri equations .
for instance , the equation for the twist @xmath20 is an homogeneous differential equation in @xmath20 and thus if @xmath20 is zero initially it will remain zero along the geodesics .
this result does not prove that the unique solution is @xmath65 because of the non - linearity of the equation .
however , this conclusion can be eventually proved more rigorously ( see @xcite ) .
this is in fact the analogous result to kelvin s theorem in hydrodynamics showing the analogy between theorems of hydrodynamics and theorems concerning congruences of geodesics @xcite .
the initial value problem of raychaudhuri equations is still an open question .
however , in the case of equation if congruences that satisfy @xmath65 are considered , there would exist hypersurfaces orthogonal to the vector field that foliate the space - time .
this is a consequence of frobenius theorem @xcite .
therefore , for a normalized vector field defined orthogonally to an arbitrary hypersurface not necessarily in the whole manifold
its shear and its expansion ( with components only in that hypersurface ) can be determined .
this way , the expansion rate of the congruence of geodesics defined by this vector field can also be determined . therefore , the rate of the kinematic quantities of the congruence ( @xmath18 , @xmath20 and @xmath19 for timelike geodesics and @xmath70 , @xmath69 and @xmath68 for null geodesics ) is influenced by the geometry of space - time . in particular , in the case of expansion @xmath18 ( @xmath70 ) , the term @xmath76 ( @xmath77 ) is the contribution of space - time geometry for timelike geodesics ( null geodesics ) .
let us recall that the contribution of space - time geometry has more general implications than the other contributions since the former depends solely on the direction of the tangent vector field at one point but not on the local behavior of the tangent vector field as @xmath18 ( @xmath70 ) , @xmath19 ( @xmath68 ) and @xmath20 ( @xmath69 ) .
thus , the contribution of space - time at one point of the manifold is the same for every congruence of geodesics with the same value of the tangent vector field at that point .
however , the quantities @xmath18 , @xmath19 and @xmath20 will be different if the congruences are not locally the same . finally , let us remind here that the field equations derived from the assumed gravitational theory determine the geometry of the space - time .
for instance , in the case of gr , the usual einstein s field equations enable to relate the quantities @xmath39 and @xmath78 which appear in and respectively with the energy - momentum tensor @xmath79 in a very simple way .
hence , the expansion rate is straightforwardly related with the energy - momentum content . as a result ,
general properties of the behavior of the geodesic congruence can be established provided that some energy conditions on the energy - momentum tensor are assumed @xcite .
more precisely , if the strong energy condition and the null energy condition are assumed ( although the last one also follows by continuity if the strong energy condition holds ) one reaches _ the geodesic focusing theorem _ , which assures the convergence until zero size of a congruence of geodesics without vorticity which is initially focusing .
this result is used in the demonstration of the singularity theorems by hawking and penrose and it has recently drawn wide attention due to its prominent role in the holographic principle @xcite .
in this section , let us start by considering a spatially flat robertson - walker metric @xcite , @xmath80 that according to the ultimate measurements of cosmological densities of the universe @xcite , represents correctly the universe at sufficiently large scales in agreement with the hypotheses of homogeneity and isotropy stated in the cosmological principle .
the connection between the inequality or equivalently with the accelerated expansion of the universe is not straightforward .
the accelerated expansion of the universe in a robertson - walker cosmological model is governed by the condition @xmath81 where the derivatives are taken with respect to cosmic time .
however , provided that the inequality holds for all the timelike directions @xmath0 in a region of the manifold , the contribution of space - time geometry then favors a tendency to diverge of all the congruence of geodesics in that region .
therefore , the natural question that arises is whether the condition @xmath81 , when satisfied in a region , implies @xmath82 for every timelike direction @xmath0 .
this question will be addressed throughout the present section .
one may thus wonder about the effects of the cosmological acceleration in the kinematical behavior of a congruence of timelike geodesics . in particular
, it is of special interest to study whether a congruence of timelike geodesics tends to focus or to diverge and under which conditions this happens .
the kinematic quantity of a congruence known as expansion @xmath18 introduced in section [ raychaudhuriequation ] provides us with a useful tool to study the focusing of timelike geodesics . in this sense ,
the inequality @xmath83 states the tendency of nearby geodesics to diverge measured by a comoving observer .
the proper time @xmath24 is the time of the comoving observer . since the l.h.s . of the previous expression
can be replaced by the r.h.s . of the re , a vector field tangent to a congruence of timelike geodesics
must be chosen in order to determine the expansion @xmath18 , shear @xmath19 and twist @xmath20 .
since we are dealing with a robertson - walker metric , the most natural choice for a vector field would be @xmath84 which is the preferred direction of a robertson - walker space - time because of the local isotropy in hypersurfaces orthogonal to @xmath85 .
the vector field gives rise to a congruence of timelike geodesics of the form @xmath86 in comoving coordinates with @xmath87 and @xmath88 constants .
these geodesics are followed by the fundamental observers , those seeing homogeneity and isotropy in constant - time hypersurfaces .
thus , a fundamental congruence refers to the congruence of timelike geodesics generated by @xmath85 . as it is easy to verify by straightforward calculation or by using symmetry arguments
, the kinematic quantities of the fundamental congruence are all zero except the expansion @xmath89 thus , by inspection of it follows that the rate of the expansion seen by a fundamental observer will be entirely determined by the value of the expansion @xmath90 and the contribution of space - time geometry @xmath91 hence @xmath92 , \label{dtexpansionrw}\end{aligned}\ ] ] where the proper time of the fundamental observer is @xmath93 , i.e. , the time coordinate in .
a fundamental observer will then see nearby geodesics with a tendency to diverge if and only if @xmath94 . by imposing this condition on
we get @xmath95 which can be expressed in the following way @xmath96 where @xmath97 is the deceleration parameter and @xmath98 is the hubble parameter . hence , a fundamental observer will perceive nearby geodesics with a tendency to diverge if and only if @xmath99 .
in other words , getting a deceleration parameter smaller than zero is not sufficient for fundamental timelike geodesics to be perceived by a fundamental observer as diverging in an increasingly fashion . in the standard cosmological model @xmath100 ,
the value of @xmath97 today can be determined by using the values of the present relative abundances of matter ( including dark matter ) and cosmological constant , @xmath101 and @xmath102 respectively by the expression @xmath103 where flatness in spatial sections was again assumed .
thus , using the approximate values @xmath104 and @xmath105 given in @xcite , the present value of the deceleration parameter @xmath106 results @xmath107 , with an uncertainty of the order of @xmath108
. this value of @xmath106 does not imply a tendency of nearby geodesics to diverge since @xmath109 .
one may argue that we have assumed gr to calculate the deceleration parameter @xmath106 , and therefore the discussion is not independent of a gravitational theory .
however , there are straightforward estimations of the deceleration parameter using supernovae ia @xcite without appealing to a gravitational theory .
the preferred values obtained for @xmath106 in these analyses are always larger than @xmath49 , although with uncertainties of the order @xmath110 . before continuing with the study of the kinematics of a congruence of timelike geodesics ,
let us mention that in general , the determination of other geodesics is not possible in a closed form since an expression for @xmath111 is usually missing .
therefore , it will not be possible in general to obtain an expression for the tangent vector field @xmath0 of a congruence of geodesics
. however , if our interest is focused on calculating the space - time geometry contribution to the re for timelike geodesics , the knowledge of the vector field @xmath112 expression is not necessary .
the reason is that the value of @xmath57 only depends on the direction of the vector field at a given point .
then , given a timelike vector , the value of @xmath57 can be determined by using the expressions for the components of the ricci tensor . moreover , the value of @xmath57 will be constant in each hypersurface orthogonal to @xmath85 because of homogeneity in these hypersurfaces .
however , if we wish to know the total value of @xmath113 from the re , an expression for the tangent vector field is needed .
it means that we must know the vector field locally at each point of the manifold .
nevertheless , we can use the property stated in equations which asserts the _ spatial _ character of @xmath114 .
the procedure consists on defining a smooth vector field on a spacelike hypersurface .
the only conditions on the definition of this vector field are that it be normalized @xmath8 and that it be orthogonal to the hypersurface where it is defined .
this slice of a vector field will give rise to a congruence of timelike geodesics passing by the hypersurface with the direction of the vector field in each point .
this congruence will have @xmath65 because being zero in the initial hypersurface it will remain so .
however , the shear @xmath19 and the expansion @xmath18 do not need to be zero .
then , instead of considering a vector field tangent to a congruence of geodesics defined in all the manifold , a vector field is instead defined at a spacelike hypersurface .
let us remark that this procedure restricts us to consider congruences of timelike geodesics without vorticity . by this way @xmath14 at the initial hypersurface can be determined .
thus , the expansion rate @xmath113 of the congruence evaluated in the hypersurface will be known from a slice of the geodesic tangent vector field .
the congruences under consideration from now on are obtained from a one - parameter dependent slice of a vector field .
first of all , we must choose a spacelike hypersurface where the vector field will be defined .
the simplest one - parameter dependent @xmath115 family of spacelike hypersurfaces defined by a function @xmath116 in terms of the coordinates is @xmath117 where @xmath118 in order the hypersurfaces to be spacelike .
this constraint is easily understood if one considers the normal vector to the hypersurfaces and this vector is then constrained to be timelike .
this family of hypersurfaces depends in fact on three parameters , namely the three components of @xmath119 .
however , thanks to the spatial isotropy , the direction of @xmath119 does not affect the results .
thus , the relevant parameter is @xmath120 .
moreover , because of spatial homogeneity and the arbitrariness in the election of the time origin we have eliminated the constant that may appear in , i.e. , the hypersurfaces pass through the origin of coordinates .
the above restriction on the value of @xmath115 means that the hypersurface is defined only in the manifold region @xmath121 .
otherwise , the hypersurfaces are not spacelike .
let us now define the following quantity @xmath122 the slice of the timelike vector field which is normalized and orthogonal to the hypersurface given by then becomes @xmath123 where @xmath124 the tangent vector field that gives rise to the fundamental congruence is obtained in the specific case @xmath125 or equivalently @xmath126 . in order to guarantee @xmath127 to be timelike
, one imposes @xmath128 .
then , it follows the restriction on the value of @xmath115 commented above . from , @xmath129 can be interpreted as the spatial coordinate velocity of a particle passing through the hypersurfaces defined by .
the dependence of @xmath129 on the scale factor @xmath130 means that this velocity is not the same in the whole hypersurface defined by a certain value of the parameter @xmath115 .
note that in expression we have chosen the @xmath131-axis in the direction of the spatial coordinate velocity @xmath129 without loss of generality thanks to the spatial isotropy .
the @xmath132 factor just ensures the normalization @xmath133 .
one realizes of the similarity of the former derivation with the usual definitions of @xmath129 and @xmath134 in special relativity .
consequently , the slice of the timelike vector field orthogonal to the hypersurfaces in special relativity becomes @xmath135 with @xmath136 in order to guarantee @xmath137 , with @xmath138 the minkowskian metric .
hence , expressions and are completely valid both in special relativity ( setting @xmath139 ) and in robertson - walker space - times .
however , when we are considering a robertson - walker metric , the maximum coordinate velocity @xmath140 that a physical entity can reach , in the sense that it moves along null geodesics , depends on the value of scale factor @xmath111 at each point . in order to guarantee that the vector field as given by is timelike , the required condition reads @xmath141 .
therefore the maximum coordinate velocity is expressed as @xmath142 .
nevertheless , let us recall that the physical spatial velocity which is measured is @xmath143 .
one may interpret a physical velocity as the ratio between a _
spatial _ and _ time _ distances .
these distances hold for the ones induced by the timelike geodesic one follows .
therefore , for a particle with four - velocity @xmath127 , its physical velocity yields @xmath144 where @xmath145 is given by and @xmath146 is the spatial metric induced by the flow of time in the direction of @xmath147 .
alternatively , one may also build a unit vector orthogonal to @xmath147 being the spatial projection of @xmath127 , namely @xmath148 , and then calculate the ratio between the projection of @xmath127 in the direction of @xmath149 and in the direction of @xmath147 , finally obtaining the same result as in .
therefore , after considering the slice of the timelike vector field @xmath127 defined in the hypersurface , the tensor @xmath114 is calculated in order to obtain both the expansion @xmath18 and the shear @xmath19 ( the twist @xmath20 is zero by construction ) . by following this procedure ,
the value of @xmath113 in the hypersurfaces @xmath150 will be determined .
this will provide us with information about the focusing or divergence of the congruence generated by the vector field slice in terms of the scale factor @xmath111 , its derivatives and the spatial coordinate velocity @xmath129 .
however , in the following we shall focus on the mean curvature @xmath57 .
this is due to the fact that although @xmath19 and @xmath18 may not be zero , their contribution to the re is always non - positive .
moreover , as stressed in section [ raychaudhuriequation ] , the implications of the mean curvature @xmath57 are more general than those of the other terms .
furthermore , @xmath57 is the unique term that may have a positive contribution to the re since for the congruence given rise by the vector field slice we have @xmath151 .
hence , let us proceed with the calculation of @xmath152 in order to extract information about the focusing of these one - parameter dependent congruences . in the following we will obtain a condition on @xmath111 and the spatial coordinate velocity @xmath129 that guarantees @xmath153 , i.e. , a positive contribution to the re .
the motivation of choosing a positive contribution lies in the fact that the fundamental congruence satisfies @xmath154 in the present epoch as follows from .
then , by using the ricci tensor components derived from the metric and the expression for the timelike vector field slice given in , we get @xmath155\ , . \ ] ] thus , imposing @xmath156 , yields @xmath157 where @xmath158 ( because of @xmath159 ) have been used .
this last inequality tells us that the condition @xmath81 is not sufficient for a timelike vector of the form to get a positive contribution to the re from the space - time geometry @xmath152 .
consequently , @xmath160 must be larger than a certain positive value . therefore , a positive contribution to the re from space - time geometry for all timelike congruences is not equivalent to @xmath161 .
the inequality can be recast as @xmath162 which depends only on the physical velocity @xmath143 . in figure
[ fig : qbeta ] , the condition has been depicted . in that figure , the values @xmath163 ensuring a positive contribution to the re from the mean curvature are plotted .
moreover , it is worthwhile to stress that the requirement @xmath164 implies lower values of the deceleration parameter @xmath97 for timelike vectors approaching the null cone .
then , in the limit @xmath165 , i.e. , timelike vectors near the null cone , we get @xmath99 .
therefore , provided that @xmath99 all the congruences of timelike geodesics will have a positive contribution to the re from space - time geometry which means @xmath153 for all timelike vectors @xmath127 .
the implication of this condition must not be confused with : the condition implied @xmath94 for the fundamental congruence whereas in this case the condition @xmath99 implies a positive contribution from the space - time geometry , namely @xmath153 , for every congruence of timelike geodesics .
since the present value of the deceleration parameter is not smaller than @xmath49 , it remains possible to get @xmath166 for timelike congruences whose tangent vector field is sufficiently inclined with respect to @xmath167 .
therefore , using , the following possibilities for the physical velocity @xmath143 may be fulfilled @xmath168 @xmath169 @xmath170 to the re from the space - time geometry .
if @xmath126 in we obtain @xmath81 which means that the fundamental congruence will experience a positive contribution from the space - time if @xmath81 and therefore in agreement with . in figure
[ fig : nullcone ] , a future light cone is represented . using the value for the present deceleration parameter ( @xmath107 ) , the limiting value obtained from
, @xmath171 separates the regions of different contribution to the re .
the requirement @xmath172 constitutes a necessary condition to ensure @xmath173 .
however , due to the non - positive character of all the other terms of the re for a congruence without vorticity , @xmath113 might be negative despite a positive contribution @xmath174 .
let us recall that a value of the deceleration parameter @xmath97 equals to @xmath49 is obtained for @xmath175 which is the solution in robertson - walker cosmological model of the einstein s equations with cosmological constant @xmath176 in vacuum .
this solution is usually referred as the de sitter expanding universe .
this means that a value of @xmath97 lower than @xmath49 implies an expansion rate higher than the rate of expansion of a de sitter solution .
therefore , it seems very unlikely to have @xmath177 in any epoch of the evolution of the universe .
let us recall that even in the inflation phase the rate of expansion is expected to be exponential and therefore @xmath178 .
thus , let us conclude this section by remarking that for timelike vectors sufficiently close to the light cone one obtains @xmath173 in any epoch of the evolution of the universe .
in this investigation we have studied the re and its implications on the evolution of geodesics .
we addressed the kinematics of a congruence of timelike geodesics without vorticity in a robertson - walker cosmological model with no underlying gravitational theory assumed a priori . from a general point of view , we have established the difference between the implications of a cosmic acceleration @xmath81 and a positive contribution to the re from the space - time geometry , i.e. @xmath179 .
we have then considered the fundamental congruence generated by the vector field @xmath180 .
we proved that the tendency to focus or diverge for the fundamental congruence depends of the fact that @xmath97 is either bigger or smaller than @xmath49 respectively .
given the fact that an approximate value for the deceleration parameter today @xmath107 can be found either by assuming gr or just by using supernovae ia data in theory - independent calculations , one concludes that although the fundamental congruence is expanding because of @xmath181 , it has a tendency to focus since @xmath182 .
moreover , we have examined the congruence of timelike geodesics generated by a vector field orthogonal to a one - parameter @xmath115 dependent spacelike hypersurface .
the spatial coordinate velocity @xmath183 represented at each point of the hypersurface the inclination of the vector field with respect to the preferred direction of robertson - walker models @xmath147 . in the case of the slice of a vector field to be normalized and orthogonal to the hypersurface , it is possible to determine the condition for the deceleration parameter @xmath97 which ensures the contribution of the space - time geometry to the re , @xmath184 to be positive .
the aforementioned condition depends both upon the spatial coordinate velocity @xmath129 and the scale factor @xmath111 but only through the physical spatial velocity @xmath143 . in congruences without vorticity , if @xmath174 does not hold all the contributions to the re are non - positive since @xmath65 and the remaining terms are non - positive as we discussed .
we found that this condition is not satisfied in the present universe provided that @xmath185 .
this means that congruences of geodesics without vorticity with its tangent vector fulfilling @xmath185 will not tend to diverge . in the robertson - walker preferred framework
this situation occurs for geodesics moving _ near _ the light cones .
these congruences will tend to focus provided that the other terms of the re are negative . besides , a negative contribution from space - time geometry to the re would mean that the attractive character of gravity is restored at high speeds for @xmath186 .
the term high speeds must be understood with respect to the preferred framework of a robertson - walker cosmological model which is usually identified with the cmb because of its high isotropy . in the previous sense
, systems may decouple from the cosmological accelerated expansion .
this result can help to understand the structure formation of system moving fast , that is favored with respect to slow systems . even taking into account the present value of the deceleration parameter @xmath106
, @xmath172 gives a negative contribution to the re from space - time geometry .
this means that in order to get a negative contribution to re from the space - time geometry , it is necessary to move with a velocity bigger than @xmath187 with respect to the preferred framework of robertson - walker space - times .
although this value is quite large , the deceleration parameter @xmath97 was smaller in the past and consequently the velocity respect the cmb necessary to accomplish a negative contribution to the re from space - time geometry was also smaller . to conclude , we would like to highlight the fact that the results obtained in this investigation provide us with a mechanism for a gravitational system to decouple from the accelerated expansion of the universe . this mechanism does not need to appeal to the gravitational interaction of the system itself .
although the decoupling of astrophysical systems from the cosmic acceleration is usually carried out by the stronger effect of their own gravitational forces , our investigation may help to understand how geodesic motion in an expanding universe influences this decoupling .
* acknowledgments .
* we would like to thank prof .
g. f. r. ellis for drawing our attention to reference @xcite . this work has been supported by micinn ( spain ) projects numbers fis2011 - 23000 , fpa2011 - 27853-c02 - 01 and consolider - ingenio multidark csd2009 - 00064 .
fda would like to thank theoretical physics i department , complutense university of madrid for providing all the required academic resources .
adlcd acknowledges financial support from nrf and urc research fellowships ( south africa ) .
jarc acknowledges the kind hospitality of acgc / university of cape town while elaborating the manuscript .
l. covi , j. e. kim and l. roszkowski , phys .
* 82 * , 4180 ( 1999 ) ; j. l. feng , a. rajaraman and f. takayama , phys .
d * 68 * , 085018 ( 2003 ) ; j. l. feng , a. rajaraman and f. takayama , int . j. mod .
d * 13 * , 2355 ( 2004 ) ; j. a. r. cembranos , j. l. feng , a. rajaraman and f. takayama , phys . rev
. lett . * 95 * , 181301 ( 2005 ) ; j. a. r. cembranos , j. l. feng , l. e. strigari , phys .
d * 75 * , 036004 ( 2007 ) ; j. a. r. cembranos , j. l. diaz - cruz and l. prado , phys .
d * 84 * , 083522 ( 2011 ) .
h. goldberg , phys .
lett . * 50 * , 1419 ( 1983 ) ; j. r. ellis _ et al .
_ , nucl . phys .
b * 238 * , 453 ( 1984 ) ; k. griest and m. kamionkowski , phys .
* 333 * , 167 ( 2000 ) ; j. a. r. cembranos , a. dobado and a. l. maroto , phys .
* 90 * , 241301 ( 2003 ) ; phys .
d * 68 * , 103505 ( 2003 ) ; phys .
d * 73 * , 035008 ( 2006 ) ; phys .
d * 73 * , 057303 ( 2006 ) ; a. l. maroto , phys .
d * 69 * , 043509 ( 2004 ) ; phys .
d * 69 * , 101304 ( 2004 ) ; a. dobado and a. l. maroto , nucl .
b * 592 * , 203 ( 2001 ) ; int .
. phys . * d13 * , 2275 ( 2004 ) [ hep - ph/0405165 ] ; j. a. r. cembranos _ et al .
_ , jcap * 0810 * , 039 ( 2008 ) .
j. a. r. cembranos and l. e. strigari , phys . rev .
d * 77 * , 123519 ( 2008 ) ; j. a. r. cembranos , j. l. feng and l. e. strigari , phys .
lett . * 99 * , 191301 ( 2007 ) ; j. a. r. cembranos , a. de la cruz - dombriz , a. dobado , r. lineros and a. l. maroto , phys .
d * 83 * , 083507 ( 2011 ) ; j. a. r. cembranos , a. de la cruz - dombriz , v. gammaldi and a. l. maroto , phys .
d * 85 * , 043505 ( 2012 ) j. a. r. cembranos , v. gammaldi and a. l. maroto , arxiv:1204.0655 [ hep - ph ]
. j. alcaraz _ et al .
rev.*d67 * , 075010 ( 2003 ) ; p. achard _
et al . _ ,
b597 * , 145 ( 2004 ) ; europhys .
lett . * 82 * , 21001 ( 2008 ) ; j. a. r. cembranos , a. dobado and a. l. maroto , phys . rev . * d65 * 026005 ( 2002 ) ; j. phys . a * 40 * , 6631 ( 2007 ) ; phys
d70 * , 096001 ( 2004 ) ; j. a. r. cembranos _ et al . _ , aip conf
proc . * 903 * , 591 ( 2007 ) ; a. g. riess _ et al . _
[ supernova search team collaboration ] , astron .
j. * 116 * , 1009 ( 1998 ) ; s. perlmutter _ et al . _ [ supernova cosmology project collaboration ] , astrophys . j. * 517 * , 565 ( 1999 ) ; j. l. tonry _ et al . _
[ supernova search team collaboration ] , astrophys .
j. * 594 * , 1 ( 2003 ) .
t. biswas
_ et al . _ ,
* 104 * , 021601 ( 2010 ) ; jhep * 1010 * , 048 ( 2010 ) ; phys .
d * 82 * , 085028 ( 2010 ) .
s. weinberg , rev .
* 61 * , 1 - 23 , ( 1989 ) .
s. nojiri and s. d. odintsov , phys .
d * 68 * , 123512 ( 2003 ) ; a. dobado and a. l. maroto phys . rev .
d * 52 * , 1895 ( 1995 ) ; g. dvali , g. gabadadze and m. porrati , phys . lett .
b * 485 * , 208 ( 2000 ) ; j. a. r. cembranos , phys .
lett . * 102 * , 141301 ( 2009 ) ; phys .
d * 73 * , 064029 ( 2006 ) ; j. a. r. cembranos , k. a. olive , m. peloso and j. p. uzan , jcap * 0907 * , 025 ( 2009 ) ; s. nojiri and s. d. odintsov , int . j. geom .
phys . * 4 * 115 , ( 2007 ) ; d. sez - gmez , gen .
grav . * 41 * , 1527 ( 2009 ) ; s. nojiri , s. d. odintsov and d. sez - gmez , phys .
b * 681 * , 74 ( 2009 ) ; j. beltrn and a. l. maroto , phys .
d * 78 * , 063005 ( 2008 ) ; jcap 0903 , 016 ( 2009 ) ; phys .
d * 80 * , 063512 ( 2009 ) ; int . j. mod
d * 18 * , 2243 - 2248 ( 2009 ) ; j. a. r. cembranos , c. hallabrin , a. l. maroto and s. j. n. jareno , arxiv:1203.6221 [ astro-ph.co ] .
t. harko , t. s. koivisto , f. s. n. lobo and g. j. olmo , phys . rev .
d * 85 * 084016 ( 2012 ) .
t. clifton , p. g. ferreira , a. padilla and c. skordis , phys .
* 513 * , 1 ( 2012 ) ; s. m. carroll , v. duvvuri , m. trodden and m. s. turner , phys .
d * 70 * : 043528 ( 2004 ) ; a. de la cruz - dombriz and a. dobado , phys .
d * 74 * 087501 , ( 2006 ) ; a. de la cruz - dombriz , a. dobado and a. l. maroto , phys .
d * 80 * , 124011 ( 2009 ) [ erratum - ibid .
d * 83 * , 029903 ( 2011 ) ] ; shinichi nojiri and sergei d. odintsov , phys .
505 * , 59 - 144 ( 2011 ) ; e. elizalde and d. sez - gmez , phys .
d * 80 * , 044030 ( 2009 ) ; o. bertolami , c. g. boehmer , t. harko and f. s. n. lobo , phys .
d * 75 * , 104016 ( 2007 ) ; j. a. r. cembranos , a. de la cruz - dombriz and b. m. nunez , jcap * 1204 * , 021 ( 2012 ) p.
k. s. dunsby , e. elizalde , r. goswami , s. d. odintsov and d. sez - gmez , phys .
d * 82 * , 023519 ( 2010 ) ; p. k. s. dunsby _ et al .
_ , phys .
d * 80 * , 061301 ( 2009);phys .
d * 83 * , 024030 ( 2011 ) ; phys .
d * 85 * , 083511 ( 2012 ) ; class .
* 29 * , 135012 ( 2012 ) ; a. de la cruz - dombriz and d. sez - gmez , arxiv:1112.4481 [ gr - qc ] ; s. capozziello and d. sez - gmez , annalen phys . *
524 * , 279 ( 2012 ) ; a. de la cruz - dombriz , a. dobado and a. l. maroto , phys .
d * 77 * 123515 ( 2008 ) ; phys .
d * 103 * , 179001 ( 2009 ) ; a. abebe , m. abdelwahab , a. de la cruz - dombriz and p. k. s. dunsby , class .
* 29 * 135011 ( 2012 ) ; j. a. r. cembranos , a. de la cruz - dombriz and p. j. romero , arxiv:1109.4519 [ gr - qc ] .
j. ehlers , gen .
. grav . * 25 * , 1225 ( 1993 ) , english translation of original article by p. jordan , j. ehlers , w. kundt and r. k. sachs , proc . of the mathematical natural sciences section of the mainz academy of sciences and literature ,
nr . * 11 * , 792 ( 1961 ) .
n. dadhich , ( ed . ) , j. krishna rao , ( ed . ) , j. v. narlikar , ( ed . ) and c. v. vishveshwara , ( ed . ) , new york , usa : wiley ( 1985 ) 236p ; f. m. paiva , m. j. reboucas and a. f. f. teixeira , phys .
a * 126 * , 168 ( 1987 ) ; r. capovilla and j. guven , phys .
d * 52 * , 1072 ( 1995 ) ; s. kar , phys .
d * 53 * , 2071 ( 1996 ) ; s. kar , phys .
d * 54 * , 6408 ( 1996 ) ; b. carter , contemp .
* 203 * , 207 ( 1997 ) ; e. zafiris , j. geom .
* 28 * , 271 ( 1998 ) ; e. zafiris , phys .
d * 58 * , 043509 ( 1998 ) ; j. borgman and l. h. ford , int .
j. mod .
a * 20 * , 2364 ( 2005 ) ; n. dadhich , gr - qc/0511123 . j. ehlers , int . j. mod
. phys .
d * 15 * , 1573 ( 2006 ) ; n. ahmadi and m. nouri - zonoz , phys . rev .
d * 74 * , 044034 ( 2006 ) ; g. f. r. ellis , pramana * 69 * , 15 ( 2007 ) ; s. kar and s. sengupta , pramana * 69 * , 49 ( 2007 ) ; c. mukku , s. m. mahajan and b. a. bambah , pramana * 69 * , 137 ( 2007 ) ; n. ahmadi and m. nouri - zonoz , pramana * 69 * , 147 ( 2007 ) ; n. dadhich , pramana * 69 * , 23 ( 2007 ) ; s. kar , resonance j. sci . educ .
* 13 * , 319 ( 2008 ) ; m. i. wanas and m. a. bakry , int .
j. mod .
a * 24 * , 5025 ( 2009 ) ; a. p.
kouretsis and c. g. tsagas , phys .
d * 82 * , 124053 ( 2010 ) ; g. abreu and m. visser , phys .
d * 83 * , 104016 ( 2011 ) ; c. valls , j. math . phys .
* 52 * , 032703 ( 2011 ) .
l. p. eisenhart , _ riemannian geometry _ , ( princeton university press , princeton , new jersey , usa , 1926 ) .
s. capozziello , g. lambiase and c. stornaiolo , annalen phys .
* 10 * , 713 ( 2001 ) .
r. bousso , rev .
phys . * 74 * , 825 ( 2002 ) .
a. friedmann , zeitschrift fr physik a 10 , 377 , 0939 - 7922 ( 1922 ) ; zeitschrift fr physik a 21 , 326 , 0939 - 7922 ( 1924 ) ; g. lemaitre , mon . not .
soc . * 91 * , 483 ( 1931 ) ; mon . not .
soc . * 91 * , 490 ( 1931 ) ; nature * 127 * , 706 ( 1931 ) [ gen .
grav . * 43 * , 2929 ( 2011 ) ] ; gen .
* 29 * , 641 ( 1997 ) [ annales soc .
a * 53 * , 51 ( 1933 ) ] ; h. p. robertson , astrophys .
j. * 82 * , 284 ( 1935 ) ; astrophys .
j. * 83 * , 187 ( 1935 ) ; astrophys .
j. * 83 * , 257 ( 1936 ) ; h. p. robertson and t. w. noonan , ad-678 - 054 ; a. g. walker , proceedings of the london mathematical society 2 , 42 , 90 ( 1937 ) .
n. jarosik , c. l. bennett , j. dunkley , b. gold , m. r. greason , m. halpern , r. s. hill and g. hinshaw _ et al .
_ , astrophys .
j. suppl .
* 192 * , 14 ( 2011 ) ; e. komatsu _ et al . _
[ wmap collaboration ] , astrophys .
j. suppl .
* 192 * , 18 ( 2011 ) .
r. lazkoz , v. salzano and i. sendra , arxiv:1202.4689 [ astro-ph.co ] . | we study the accelerated expansion of the universe through its consequences on a congruence of geodesics .
we make use of the raychaudhuri equation which describes the evolution of the expansion rate for a congruence of timelike or null geodesics . in particular , we focus on the space - time geometry contribution to this equation . by straightforward calculation from the metric of a robertson - walker cosmological model , it follows that in an accelerated expanding universe the space - time contribution to the raychaudhuri equation is positive for the fundamental congruence , favoring a non - focusing of the congruence of geodesics .
however , the accelerated expansion of the present universe does not imply a tendency of the fundamental congruence to diverge .
it is shown that this is in fact the case for certain congruences of timelike geodesics without vorticity .
therefore , the focusing of geodesics remains feasible in an accelerated expanding universe .
furthermore , a negative contribution to the raychaudhuri equation from space - time geometry which is usually interpreted as the manifestation of the attractive character of gravity is restored in an accelerated expanding robertson - walker space - time at high speeds . |
josephson - qubit devices @xcite are based on the charge and phase degrees of freedom .
the charge qubit is achieved in a cooper - pair box @xcite , where two dominant charge states are coupled through coherent cooper - pair tunneling @xcite . using cooper - pair tunneling in josephson charge devices @xcite and via spectroscopic measurements for the josephson phase device @xcite , it has been possible to experimentally observe energy - level splitting and related properties for state superpositions . in addition , using josephson charge devices prepared in a superposition of two charge states @xcite ,
coherent oscillations were observed .
while operating at the degeneracy point , the charge - qubit states are highly coherent @xcite ( @xmath0 ) , with a decoherence time of @xmath1 ns . these important experimental results indicate that the josephson charge and phase devices are potentially useful for solid - state qubits in quantum information processing .
important open problems would now include implementing a _ two - bit coupling _ and then _ scaling up _ the architecture to many qubits . here , we propose a new quantum - computing ( qc ) scheme based on scalable charge - qubit structures .
we focus on the josephson charge qubit realized in a cooper - pair box .
the coulomb interaction between charges on different islands of the charge qubits would seem to provide a natural way of coupling josephson charge qubits ( e.g. , to connect two cooper - pair boxes via a capacitor ) . using this type of capacitive interbit coupling ,
a two - bit operation @xcite similar to the controlled - not gate was derived . however , as pointed out in @xcite , it is difficult in this scheme to switch on and off the coupling . also , it is hard to make the system scalable because only neighboring qubits can interact .
moreover , implementations of quantum algorithms such as the deutsch and bernstein - vazirani algorithms were studied using a system of josephson charge qubits @xcite , where it was proposed that the nearest - neighbor charge qubits would be coupled by tunable dc squids . in the semiconductor literature
, scalability often refers to reducing the size of the device ( packing more components ) . in qc
, scalability refers to increasing the number of qubits coupled with each other .
a suggestion for a scalable coupling of josephson charge qubits was proposed @xcite using oscillator modes in a @xmath2 circuit formed by an inductance and the qubit capacitors . in this proposal ,
the interbit coupling can be switched and any two charge qubits could be coupled .
nevertheless , there is no efficient ( that is , using one two - bit operation ) qc scheme for this proposal @xcite in order to achieve conditional gates e.g . , the controlled - phase - shift and controlled - not gates .
in addition , the calculated interbit coupling terms @xcite only apply to the case when the following two conditions are met : ( i ) the quantum manipulation frequencies , which are fixed experimentally , are required to be much smaller than the eigenfrequency @xmath3 of the @xmath2 circuit .
this condition _ limits _ the allowed number @xmath4 of the qubits in the circuit because @xmath3 scales with @xmath5 . in other words , the circuits in @xcite are not really scalable .
( ii ) the phase conjugate to the total charge on the qubit capacitors fluctuates weakly .
the limitations listed above do not apply to our approach . in our scheme , a common inductance , but no @xmath2 circuit , is used to couple all josephson charge qubits . in our proposal , both dc and
ac supercurrents can flow through the inductance , while in @xcite only ac supercurrents can flow through the inductance and it is the @xmath2-oscillator mode that couples the charge qubits .
these yield different interbit couplings ( e.g. , @xmath6 type @xcite as opposed to @xmath7 in our proposal ) .
we employ two dc squids to connect each cooper - pair box in order to achieve a _
controllable interbit coupling_. our proposed qc architecture is scalable in the sense that _ any _ two charge qubits ( _ not _ necessarily neighbors ) can be effectively coupled by an experimentally accessible inductance .
we also formulate @xcite an efficient qc scheme that requires only one ( instead of two or more ) two - bit operation to implement conditional gates . + this erice summer - school presentation is based on our work in @xcite . additional work on decoherence and noise - related issues appears in , e.g. , @xcite .
also , work more focused on entanglement and readout issues appears in @xcite .
other interesting studies on charge qubits can be found in @xcite for the adiabatic controlled - not gate , in @xcite for geometric phases , and in @xcite for the dynamics of a josephson charge qubit coupled to a quantum resonator .
figure 1 shows a proposed qc circuit consisting of @xmath4 cooper - pair boxes coupled by a common superconducting inductance @xmath8 . for the @xmath9th cooper - pair box , a superconducting island with charge @xmath10
is weakly coupled by two symmetric dc squids and biased , through a gate capacitance @xmath11 , by an applied voltage @xmath12 .
the two symmetric dc squids are assumed to be equal and all josephson junctions in them have josephson coupling energy @xmath13 and capacitance @xmath14 .
the effective coupling energy is given by the squids , each one enclosing a magnetic flux @xmath15 .
each squid provides a tunable coupling @xmath16 , with @xmath17 and @xmath18 is the flux quantum . the effective phase drop @xmath19 , with subscript @xmath20 labelling the squid above ( below ) the island
, equals the average value , @xmath21/2 $ ] , of the phase drops across the two josephson junctions in the dc squid , where the superscript @xmath8 ( @xmath22 ) denotes the left ( right ) josephson junction . above
we have neglected the self - inductance effects of each squid loop because the size of the loop is usually very small ( @xmath23 @xmath24 m ) .
the hamiltonian of the system then becomes @xmath25 with @xmath26 given by @xmath27 here @xmath28 is the charging energy of the superconducting island and @xmath29 is the total persistent current through the superconducting inductance , as contributed by all coupled cooper - pair boxes .
the offset charge @xmath30 is induced by the gate voltage @xmath12 .
the phase drops @xmath31 and @xmath32 are related to the total flux @xmath33 through the inductance @xmath8 by the constraint @xmath34 where @xmath35 is the externally applied magnetic flux threading the inductance @xmath8 . in order to obtain a simpler expression for the interbit coupling , and without loss of generality ,
the magnetic fluxes through the two squid loops of each cooper - pair box are designed to have the _ same _ values but _ opposite _ directions .
if this were not to be the case , the interbit coupling can still be realized , but the hamiltonian of the qubit circuits would just take a more complicated form . because this pair of fluxes cancel each other in any loop enclosing them , then @xmath36 this imposes the constraint @xmath37 for the average phase drops across the josephson junctions in the squids .
the common superconducting inductance @xmath8 provides the coupling among cooper - pair boxes .
the coupling of selected cooper - pair boxes can be implemented by switching on " the squids connected to the chosen cooper - pair boxes . in this case , the persistent currents through the inductance @xmath8 have contributions from all the coupled cooper - pair boxes .
the essential features of our proposal can be best understood via the very simplified diagram shown in fig . 2 .
as seen in fig .
3(a ) , for any given cooper - pair box , say @xmath38 , when @xmath39 for all boxes except @xmath40 , the inductance @xmath8 only connects the @xmath38th cooper - pair box to form a superconducting loop .
the hamiltonian of the system can be reduced to @xcite @xmath41 where @xmath42\ ] ] is controllable via the gate voltage @xmath43 , while the intrabit coupling @xmath44 can be controlled by both the applied external flux @xmath35 through the common inductance , and the local flux @xmath45 through the two squid loops of the @xmath38th cooper - pair box .
retained up to second - order terms in the expansion parameter @xmath46 where @xmath47 we obtain @xmath48 with @xmath49 the intrabit coupling @xmath44 in ( 9 ) is different from that in @xcite because a very different contribution by @xmath8 is considered . to couple
_ any _ two cooper - pair boxes , say @xmath38 and @xmath50 , we choose @xmath51 for all boxes except @xmath40 and @xmath50 . as shown in fig .
3(b ) , the inductance @xmath8 is shared by the cooper - pair boxes @xmath38 and @xmath50 to form superconducting loops .
the reduced hamiltonian of the system is given by @xcite @xmath52 + \pi_{ij}\,\sigma^{(i)}_x\sigma^{(j)}_x.\ ] ] up to second - order terms , @xmath53 with @xmath54 and @xmath55 here the interbit coupling @xmath56 is controlled by both the external flux @xmath35 through the inductance @xmath8 , and the local fluxes , @xmath45 and @xmath57 , through the squid loops . using these two types of circuits , we can derive the required one- and two - bit operations for qc .
specifically , the conditional gates such as the controlled - phase - shift and controlled - not gates can be obtained using one - bit rotations and only one basic two - bit operation . for details , see ref .
a sequence of such conditional gates supplemented with one - bit rotations constitute a universal element for qc @xcite .
usually , a two - bit operation is much slower than a one - bit operation .
our designs for conditional gates @xmath58 and @xmath59 are _ efficient _ since _ only one _ ( instead of two or more ) basic two - bit operation is used .
the one - bit circuit modeled by hamiltonian ( 9 ) has two eigenvalues @xmath60 , with @xmath61^{1/2}.\ ] ] the corresponding eigenstates are @xmath62 where @xmath63 at these two eigenstates , the persistent currents through the inductance @xmath8 are given by @xmath64 up to the linear term in @xmath65 . in the case
when a dc squid magnetometer is inductively coupled to the inductance @xmath8 , these two supercurrents generate different fluxes through the squid loop of the magnetometer and the quantum - state information of the one - bit structure can be obtained from the measurements . in order to perform sensitive measurements with weak dephasing
, one could use the underdamped dc squid magnetometer designed previously for the josephson phase qubit @xcite .
for the two - bit circuit described by eq .
( 15 ) , the hamiltonian has four eigenstates and the supercurrents through inductance @xmath8 take different values for these four states .
the fluxes produced by the supercurrents through @xmath8 can also be detected by the dc squid magnetometer .
for example , when @xmath66 and @xmath67 for @xmath40 and @xmath50 , the four eigenstates of the two - bit circuit are @xmath68 retained up to linear terms in @xmath65 and @xmath69 , the corresponding supercurrents through the inductance @xmath8 are @xmath70 for @xmath71 to 4 , where @xmath72 these supercurrents produce different fluxes threading the squid loop of the magnetometer and can be distinguished by dc squid measurements .
when the two - bit system is prepared at the maximally entangled bell states @xmath73 these two states should be distinguishable by measuring the fluxes , generated by the supercurrents , through the squid loop of the magnetometer .
the typical switching time @xmath74 during a one - bit operation is of the order of @xmath75 . using the experimental value @xmath76 mk , then @xmath77 ns .
the switching time @xmath78 for the two - bit operation is typically of the order @xmath79 choosing @xmath76 mk and @xmath80 ( i.e. , ten times slower than the one - bit rotation ) , we have @xmath81 in our proposal . this number for @xmath8
is experimentally realizable .
a small - size inductance with this value can be made with josephson junctions .
our expansion parameter @xmath82 is of the order @xmath83 our inductance @xmath8 is related with the inductance @xmath84 in @xcite by @xmath85 let us now consider the case when @xmath80 .
for the earlier design @xcite , @xmath86 since @xmath87 , which requires an inductance @xmath88 @xmath24h .
such a large inductance is problematic to fabricate at nanometer scales . in the improved design @xcite , @xmath89 , greatly reducing the inductance to @xmath90 nh
this inductance is about four times larger than the one used in our scheme , making it somewhat more difficult to realize than our proposed @xmath8 .
we propose a _ scalable _ quantum information processor with josephson charge qubits . we use a common inductance to couple all charge qubits and design _ switchable _ interbit couplings using two dc squids to connect the island in each cooper - pair box .
the proposed circuits are scalable in the sense that any two charge qubits can be effectively coupled by an experimentally accessible inductance . in addition
, we formulate @xcite an efficient qc scheme in which only one two - bit operation is used in the conditional transformations , including controlled - phase - shift and controlled - not gates .
we thank yu .
pashkin , b. plourde and xuedong hu for useful discussions .
this work is supported in part by arda , the afosr , and the us national science foundation grant no . eia-0130383 .
99 see , e.g. , y. makhlin , g. schn , and a. shnirman , rev .
phys . * 73 * , 357 ( 2001 ) , and references therein .
y. nakamura , yu . a. pashkin , and j.s .
tsai , nature ( london ) * 398 * , 786 ( 1999 ) .
y. makhlin , g. schn , and a. shnirman , nature ( london ) * 398 * , 305 ( 1999 ) .
y. nakamura , c.d .
chen , and j.s .
tsai , phys .
lett . * 79 * , 2328 ( 1997 ) .
v. bouchiat , d. vion , p. joyez , d. esteve , and m.h .
devoret , phys .
scripta * t76 * , 165 ( 1998 ) . c.h .
van der wal , a.c.j .
ter haar , f.k .
wilhelm , r.n .
schouten , c.j.p.m .
harmans , t.p .
orlando , s. lloyd , and j.e .
mooij , science * 290 * , 773 ( 2000 ) .
friedman , v. patel , w. chen , s.k .
tolpygo , and j.e .
lukens , nature ( london ) * 406 * , 43 ( 2000 ) .
d. vion , a. aassime , a. cottet , p. joyez , h. pothier , c. urbina , d. esteve , and m.h .
devoret , science * 296 * , 886 ( 2002 ) .
f. plastina , r. fazio , and g.m .
palma , phys .
b * 64 * , 113306 ( 2001 ) .
j. siewert and r. fazio , phys .
* 87 * , 257905 ( 2001 ) .
you , j.s .
tsai , and f. nori , phys .
lett . * 89 * , 197902 ( 2002 ) .
e. paladino , l. faoro , g. falci , and r. fazio , phys .
lett . * 88 * , 228304 ( 2002 ) .
you , x. hu , and f. nori , preprint .
you , j.s .
tsai , and f. nori , preprint .
averin , solid state commun . * 105 * , 659 ( 1998 ) .
g. falci , r. fazio , g.m .
palma , j. siewert , and v. vedral , nature ( london ) * 407 * , 355 ( 2000 ) .
o. buisson and f.w.j .
hekking , cond - mat/0008275 ; also in _ macroscopic quantum coherence and quantum computing _
( kluwer academic , dordrecht , 2000 ) , p. 137 .
f. marquardt and c. bruder , phys .
b * 63 * , 054514 ( 2001 ) .
a.d . armour , m.p .
blencowe , and k.c .
schwab , phys .
88 * , 148301 ( 2002 ) .
you and f. nori , preprint .
additional schematic color diagrams for the proposed circuitry appear in http://www-personal.engin.umich.edu/~nori/scalable/scalableqc_figs.pdf s. lloyd , phys .
lett . * 75 * , 346 ( 1995 ) .
d. deutsch , a. barenco , and a. ekert , proc .
r. soc .
london , ser .
a * 449 * , 669 ( 1995 ) . | two of the major obstacles to achieve quantum computing ( qc ) are ( i ) scalability to many qubits and ( ii ) controlled connectivity between any selected qubits . using josephson charge qubits , here we propose an experimentally realizable method to efficiently solve these two central problems . since any two charge qubits can be effectively coupled by an experimentally accessible inductance ,
the proposed qc architecture is _
scalable_. in addition , we formulate an efficient and realizable qc scheme that requires _ only one _ ( instead of two or more ) two - bit operation to implement conditional gates . |
the experimental discovery of persistent currents in mesoscopic rings pierced by a magnetic flux,@xmath5 earlier proposed theoretically,@xmath6 has revealed interesting new effects .
the currents measured in metallic and semiconducting rings , either in a single ring or an array of many rings , generally exhibit an unexpectedly large amplitude , i.e. , larger by at least one order of magnitude , than predicted by theoretical studies of electron models with either disorder or electron - electron interaction treated perturbatively.@xmath7 it has been suggested that the interactions and their interplay with disorder are possibly responsible for the large currents observed , expecting that the effect of the interactions could counteract the disorder effect . however , no consensus has yet been reached on the role of the interactions . in order to gain theoretical insight , it is desirable to perform numerical calculations which allow to consider both interactions and disorder directly in systems with sizes varying from small to large .
analytical calculations usually involve approximations which mainly provide the leading behavior of the properties for large system sizes .
persistent currents in mesoscopic rings strongly depend on the system size , since they emerge from the coherence of the electrons across the entire system .
hence , it is most important to study the size dependence of the current beyond leading order in microscopic models , for a complete understanding of the experimental results .
exact diagonalization was used to calculate persistent currents in systems with very few lattice sites.@xmath8 in this work , we use the density matrix renormalization group ( dmrg ) algorithm,@xmath9 to study a simplified model incorporating interactions and a single impurity , accounting for disorder , in larger system sizes .
we consider a system of interacting spinless electrons on a one - dimensional ring , with a single impurity , and penetrated by a magnetic field .
we study an intermediate range of system sizes , where analytical results obtained by bosonization techniques for large system sizes , do not yet fully apply . without impurity , and at half - filling , the system undergoes a metal - insulator transition from a luttinger liquid ( ll)@xmath10 to a charge density wave ( cdw)@xmath11 groundstate .
the persistent current of the interacting system with an impurity was studied before with the dmrg , in the ll phase.@xmath12 here we study the persistent current , and also the drude weight characterizing the conducting properties of the system , in both the ll and the cdw phase , investigating the interplay between the impurity and the interactions in the two phases . in mesoscopic systems
the separation between metallic and insulating behavior is not always obvious , since the localization length can be of the order or significantly larger than the system size .
hence , a finite drude weight and a current can be observed in the cdw phase of a mesoscopic system .
it is therefore of great interest to characterize the persistent current and the drude weight in both the ll and the cdw phases of mesoscopic systems .
although the simple model that we consider is not the most appropriate to describe the experimental situation , we hope to obtain useful information for the understanding of the more realistic systems . under a jordan - wigner transformation,@xmath13
the system considered is equivalent to a spin-1/2 xxz chain with a weakened exchange coupling .
hence , our results also provide insight into the spin transport in this type of systems .
the hamiltonian describing a system of spinless fermions on a ring pierced by a magnetic flux , with repulsive interactions and a single hopping impurity , or defect , is given by , @xmath14 where @xmath15 is the hopping term , @xmath16 contains the magnetic flux @xmath17 in units of the flux quantum @xmath18 , @xmath2 measures the strength of the defect with values between @xmath19 and @xmath20 , ( @xmath21 corresponding to the defectless case ) , and @xmath22 is the interaction term , with @xmath23 representing the nearest neighbor coulomb repulsion , and @xmath24 , where @xmath25 and @xmath26 are the spinless fermion operators acting on the site @xmath27 of the ring .
we consider a system of @xmath3 sites , with @xmath3 even , and at half - filling , when @xmath28 particles are present .
the lattice constant is set to one and periodic boundary conditions , @xmath29 , are used . via the gauge transformation @xmath30
, the flux can be removed from the hamiltonian , but in the impurity term where the flux is trapped , and the quantum phase @xmath31 is encoded in a twisted boundary condition @xmath32 .
it is then clear that the energy is periodic in @xmath31 with period @xmath33 , i.e. , it is periodic in the flux @xmath17 threading the ring with period @xmath34.@xmath35 after a jordan - wigner transformation , eqs .
( 2 ) and ( 3 ) can be rewritten , respectively , as @xmath36 and @xmath37with @xmath38 and @xmath39 , and the boundary conditions @xmath40 and @xmath41 .
hence , the model ( 1 ) of spinless fermions is equivalent to a spin-1/2 xxz chain with a weakened exchange coupling , and twisted boundary conditions in the transverse direction .
the half - filled case corresponds to total spin projection @xmath42 .
the persistent current generated on a ring pierced by a magnetic flux , at temperature @xmath43 , can be obtained from the ground state energy @xmath44 , by taking the derivative with respect to @xmath45 , @xmath46 for the spinless fermion system , eqs .
( 2 ) and ( 3 ) , @xmath47 corresponds to the ground state value of the charge current operator @xmath48 , while for the xxz chain , eqs .
( 4 ) and ( 5 ) , it corresponds to the ground state value of the spin current operator @xmath49 . as a consequence of the periodicity of the energy ,
the current is also periodic in @xmath31 , with period @xmath33 .
hence , it can be expressed as a fourier series , @xmath50 and the behavior of the current can be analyzed in terms of the coefficients @xmath51.@xmath12 in the noninteracting case ( @xmath52 ) , it has been found that for large system sizes , the current is invariant under the defect transformation @xmath1,@xmath53 i.e. , @xmath54 we shall investigate the existence of this kind of invariance in the interacting case ( @xmath23 ) , both in the ll and the cdw phases .
the drude weight was proposed by kohn as a relevant quantity to distinguish between a metal and an insulator.@xmath55 it is defined as @xmath56 where @xmath57 is the location of the minimum of @xmath58 , which depends on the parity of the number of electrons , i.e. , @xmath59 or @xmath60 for , respectively , an odd or an even number of electrons . for the spinless fermion system @xmath61 represents the charge - stiffness and measures the inverse of the effective mass of the charge carriers.@xmath62 in a metallic conductor @xmath61 tends to a finite value whereas in an insulator @xmath61 vanishes with the system size @xmath3 , when @xmath4 . in the insulating state
the drude weight decays as @xmath63 , where @xmath64 measures the localization length .
for the xxz chain the drude weight represents the spin - stiffness . in a model of free fermions ( @xmath52 ) with no impurity ( @xmath21 ) , it is straightforward to see that the leading behavior of the persistent current @xmath47 in the system size @xmath3 , has a saw - tooth like shape with slope @xmath65 , where @xmath66 is the fermi velocity . thus , the amplitude of the current scales with @xmath67 , vanishing in the limit @xmath68 .
the discontinuity in @xmath47 , that results from a degeneracy of energy levels associated to the translation symmetry,@xmath69 appears at @xmath70 or @xmath60 for , respectively , an even or an odd number of electrons . in the presence of an impurity ( @xmath71 ) , bosonization@xmath72 and conformal field theory@xmath73 calculations predict that the shape of the current @xmath47 is rounded off , and its amplitude decreases with increasing strength of the scatterer potential , still vanishing with the system size as @xmath67 .
the impurity lifts the degeneracy of the energy levels and the current then varies continuously.@xmath69 the model with interactions ( @xmath74 ) and without defect ( @xmath21 ) , is solvable by the bethe ansatz for periodic boundary conditions ( @xmath70)@xmath75 and also for twisted boundary conditions ( @xmath76).@xmath77 at half - filling , the system exhibits a metal - insulator transition , which occurs at @xmath78 . for @xmath79 ,
the system is in a gapless ll phase , while for @xmath80 it is in a gapped cdw state.@xmath81 the ll phase is characterized by a power - law decay of the correlations .
bosonization predicts that in an homogeneous ll , the leading behavior of the persistent current in the system size @xmath3 , has a saw tooth like shape with slope @xmath82 , where @xmath83 is the velocity of current excitations.@xmath84 since translation invariance is preserved in the presence of interactions , the discontinuity in the current still exists for finite @xmath0.@xmath69 a bethe ansatz calculation shows that the drude weight of an homogeneous ll in the thermodynamic limit , has a finite value , which decreases with increasing strength of the interaction @xmath85.@xmath62 the ll state is strongly affected by the presence of an impurity,@xmath86 and bosonization yields that the current then vanishes as @xmath87 , with @xmath88.@xmath72 the study of the ll phase with @xmath71 , performed with the dmrg,@xmath12 has in fact found this kind of behavior .
the cdw phase is characterized by a localization length @xmath64 , which is associated to the energy gap . from the work of baxter,@xmath89 the drude weight in the gapped phase ,
is expected to behave as @xmath63 , vanishing for an infinite system size .
this behavior implies that although the system is insulating in the infinite system size limit , for a finite system , provided @xmath90 is not too large , @xmath61 is still finite and a current can be observed .
the localization length can then be extracted from the size dependence of the drude weight .
we use the dmrg to numerically calculate the groundstate energy of the spinless fermion system as a function of the magnetic flux , @xmath58 , for fixed interaction @xmath0 and impurity strength @xmath2 , in rings up to @xmath91 sites , keeping up to @xmath92 density matrix eigenstates per block.@xmath93 the dmrg is applied to the hamiltonian ( 1 ) after performing the gauge transformation , which removes the flux into a twisted boundary condition.@xmath94 the states of the system are characterized by the quantum numbers associated to the eigenvalues of the local occupation number @xmath95 and the total number of particles @xmath96 operators , which commute with the hamiltonian ( 1).@xmath97 for each set of @xmath3 , @xmath2 and @xmath0 , we obtained the groundstate energy @xmath98 for @xmath99 values of @xmath31 in the periodicity interval @xmath100 , and using chebyshev interpolation@xmath101 we determined the corresponding current ( 6 ) and drude weight ( 9 ) , by numerical differentiation .
we developed a dmrg algorithm for complex hamiltonian matrices , which allowed to calculate the detailed form of the persistent current @xmath102 as a function of @xmath31,@xmath12 and to obtain the drude weight . in a previous approach
the dmrg was used to calculate the so called phase sensitivity @xmath103 , which is the difference of the groundstate energy at flux @xmath70 and @xmath60 , and can be considered a crude measure of the persistent current.@xmath104 although the calculation of @xmath103 requires considerably less computational effort than the calculation of @xmath105 , because then the hamiltonian matrix is real , the phase sensitivity does not provide information on the shape of the current and the value of the drude weight .
we now present the results obtained for the persistent current and the drude weight , where we take @xmath106 and the interaction @xmath0 is in units of @xmath107 .
1 and 2 , exhibit @xmath108 plotted versus @xmath31 , for respectively , @xmath109 and @xmath110 , which correspond , respectively , to the ll and the cdw phase , considering different impurity strengths @xmath2 , on a fixed system size @xmath111 .
we can see that the effect of the impurities , in both phases , is to reduce the intensity of the current , and to round off the shape of @xmath47 .
the amplitude of the current decreases rapidly with increasing values of @xmath112 , and also with increasing strength of the interaction @xmath0 .
3 shows that the invariance of the current with respect to the defect , described in eq .
( 8) , found for large system sizes in the noninteracting case , is also observed for the interacting case , both in the ll and in the cdw phases , the system size @xmath3 required to reach the invariance being larger for larger interaction @xmath0 .
4 displays the drude weights associated to the systems with different interactions @xmath0 and impurities @xmath2 , fixed @xmath111 , of the currents presented in figs . 1 and 2 .
as one would expect the drude weight decreases with increasing @xmath112 , and also increasing @xmath0 . figs . 5 and 6 , present @xmath108 plotted for several system sizes @xmath3 , respectively , for @xmath109 , in the ll phase , and @xmath110 , in the cdw phase , with @xmath2 fixed at @xmath113 .
we observe that the current vanishes faster than @xmath67 in both phases , exhibiting a different behavior in each phase , @xmath47 vanishing much faster with @xmath3 in the cdw than in the ll phase . in order to analyze the behavior of the current in more detail we have numerically evaluated the coefficients @xmath51 ( for @xmath114 ) of the fourier expansion ( 7 ) .
the first ( @xmath115 ) and the second ( @xmath116 ) fourier coefficients of the current , for @xmath109 and @xmath110 , are shown in fig .
one can clearly see that the coefficients @xmath117 and @xmath118 behave similar to each other in both phases .
however , their behavior in the ll and cdw phase is distinct . in the ll phase ,
the fourier coefficients show a power - law decay with @xmath3 , with the second order coefficient decaying faster , i.e. , with a larger exponent , than the first one . in the cdw phase , the fourier coefficients show a dominant exponential decay with @xmath3 , with the second order coefficient also decaying faster , i.e. , with a smaller localization length , than the first one .
we observe that for longer rings , stronger interactions and also stronger impurities , the current is increasingly more precisely described by its first fourier component . in the ll phase
, this in fact corresponds to the asymptotic behavior predicted by bosonization in the large @xmath3 limit , that is @xmath119 .
however , for the system sizes considered here this asymptotic regime is not reached and the current displays a more complex behavior .
the current is composed of a few fourier components with decreasing weight .
8 presents the first fourier coefficient of the current for different values of the interaction , @xmath109 , @xmath120 , @xmath121 , fixed @xmath122 , from which we extract the dependence of @xmath117 on @xmath3 , in the intermediate range of sizes considered .
we observe that for @xmath109 , in the ll phase , the first coefficient of the current varies as @xmath123 @xmath124 , with @xmath125 , while for @xmath126 and @xmath110 , in the cdw phase , it varies as @xmath127 , respectively , with @xmath128 , @xmath129 , and @xmath130 , @xmath131 .
the exponent @xmath132 is given by the slope of the straight line in fig .
8.a , and the length @xmath133 and the exponent @xmath134 were carefully adjusted in order to obtain the best collapse of the data in fig.8.b , on a plot of @xmath135 vs @xmath136 .
the drude weights characterizing the systems with different interactions @xmath0 , fixed @xmath122 , are presented in fig
10 clearly shows that the results obtained for the drude weight confirm the conducting behavior shown by the first coefficient of the currents in fig.8 .
we observe that , for @xmath109 the drude weight varies with the system size as @xmath137 , with @xmath138 , while for @xmath126 and @xmath110 it varies as @xmath139 , respectively , with @xmath140 , @xmath141 and @xmath142 , @xmath143 .
the exponent @xmath144 is given by the slope of the straight line in fig .
10.a , and the localization length @xmath64 and the exponent @xmath145 were carefully adjusted in order to obtain the best collapse of the data in fig.10.b , on a plot of @xmath146 vs @xmath90 .
the exponents and localization lengths characterizing the drude weight are a little different from those characterizing the first fourier component of the current , as one would expect , since the drude weight contains the contribution from the various fourier components .
one sees that the localization length @xmath64 and the exponent @xmath145 decrease with increasing strength of the interaction @xmath0 .
also , concerning the impurity influence , fig .
4 implies that the exponent @xmath144 in the ll phase increases , and the localization length @xmath64 in the cdw phase decreases , with increasing @xmath147 .
as mentioned , in the large @xmath3 limit the current is expected to behave as its first fourier component , which in the ll phase implies that the exponent @xmath132 should be identified with @xmath148 as calculated from bosonization,@xmath72 where @xmath149 is the ll parameter , calculable from the bethe ansatz.@xmath150 for @xmath109 this leads to @xmath151 , which is much larger than our value of @xmath132 .
we should note that the size dependence found for the first fourier component of the current , and the drude weight , characterizes the behavior of an intermediate and limited range of system sizes .
if one would consider a larger range of systems , in the ll phase , one would most probably see the data for the larger @xmath3 bending down , crossing to an asymptotic power - law behaviour with the exponent approaching @xmath152 .
this was observed in ref .
@xmath153 , where the behavior of the first few fourier components of the current in the ll phase was discussed in detail , with data taken for larger values of @xmath3 and stronger interaction and impurity strengths . also , in the cdw phase we consider systems in an intermediate regime where the localization length is larger or near the system size.@xmath154 for larger systems , the power factors that occur in the first fourier component of the current and the drude weight may decline,@xmath155 possibly leaving a pure exponential behavior in the asymptotic regime . from the results obtained , we observe that the system with @xmath109 and @xmath122 , is characterized by an exponent @xmath156 , which is generated by the interplay of the electron interaction with the impurity , and @xmath61 exhibits then a power - law decay with @xmath3 , which implies vanishing in the limit @xmath4 . on the other hand , the systems with @xmath126 and @xmath110 , fixed @xmath122 ,
are characterized by a localization length @xmath64 , which decreases with increasing interaction and impurity strength , and @xmath61 exhibits now an exponential decay with @xmath3 , also vanishing as @xmath4 .
hence , we find that both in the ll and the cdw phases , with an impurity in the system , the effect of the interaction is to decrease the current and the drude weight . as referred before , our results also provide insight into the spin transport in a spin-1/2 xxz chain with a weak link , and a similar behavior to the one above is implied for the spin current and stiffness .
so , in the gapless xy phase , the spin stiffness decays with a power - law , vanishing in the limit @xmath4 , while in the gapped ising phase , it decays exponentially , also vanishing as @xmath4 . comparing our results for the persistent current in the ll phase , with those obtained in ref .
@xmath153 , we have similarly found that the current vanishes faster than @xmath67 .
one observes that the model parameters strongly influence when the last asymptotic regime described by bosonization is reached .
a calculation of the finite - size corrections to the spin stiffness in a pure spin-1/2 xxz chain,@xmath157 has revealed a size dependence in the gapped phase that has a similar form to the one found here .
the result that the drude weight in a ring in the gapless phase with an impurity drops to zero , is in agreement with a previous result obtained for a spin chain,@xmath158 and with renormalization group arguments , which state that the impurity term is relevant leading to a transmission cut.@xmath159 the renormalization group studies find that either a weak barrier or a weak link lead to an insulating state for repulsive interactions , while in the noninteracting case those are marginal perturbations . in turn
, our work shows that there is an invariance of the current under the defect transformation @xmath160 in the interacting system , as for the noninteracting system , and that implies that a strong link will also reduce the current and the drude weight .
the observation that with an impurity in the system , the interaction always leads to an additional decrease of the current and the drude weight is in agreement with previous results by other authors,@xmath8 and can be understood as it is more difficult to move correlated electrons in a scattering potential than independent electrons .
we have studied the behavior of the persistent current and the drude weight on a mesoscopic ring pierced by a magnetic flux . we considered a model of spinless fermions with repulsive interactions and a hopping impurity ,
which is also equivalent to a spin-1/2 xxz chain with a weakened exchange coupling . using a powerful numerical method , the dmrg with complex fields , we have calculated the detailed form of the current as a function of the magnetic flux , which enabled us to investigate the corrections to the large system - size limit , and also allowed to obtain the drude weight .
we show that the system at half - filling , changes from an algebraic to an exponential behavior as the interaction increases , corresponding to a change from a ll to a cdw phase .
we find that the analytical predictions of bosonization for the ll phase , are not yet fully observed in the intermediate range of system sizes considered .
in addition we observe that the invariance of the current under the defect transformation @xmath160 , seen in the noninteracting system , is also verified in the interacting system , in both phases .
hence , an isolated strong link is not only useless for increasing the persistent current ( as might have been expected ) , but it rather destroys coherence and reduces the current .
the behavior determined for the current is consistent with the behaviour determined for the drude weight , the ll phase being characterized by an exponent @xmath156 , which results from the interplay of the interactions with the impurity , while the cdw phase is characterized by a localization length @xmath64 , which decreases with increasing interaction and impurity strength .
we find that , both in the ll and the cdw phase , with a defect in the system the interactions always suppress the current , and the drude weight drops to zero in the limit @xmath4 . away from half - filling
there is no metal - insulator transition in the pure case , and the system is always metallic .
hence , one does not expect to observe then a change in the current and the drude weight from an algebraic to an exponential decay .
nevertheless , one still expects to observe that in the system with an impurity , the current and the drude weight decrease with increasing impurity and interaction strengthes .
therefore , within the model considered , the interactions can not explain the results observed in the experiments .
@xmath81 in the xxz chain , a transition occurrs at @xmath176 , from an xy model , for @xmath177 , to an ising model , for @xmath178 , @xmath176 corresponding to the heisenberg model .
a gap opens up in the spin - excitation spectrum , that is equivalent to the gap in the charge - excitation spectrum of the spinless fermion system .
@xmath183the number @xmath184 of eigenstates required to achieve convergence , of 1 part in @xmath185 , for the ground state energy @xmath186 , increases with the system size , e.g. , @xmath187 for @xmath188 and @xmath189 for @xmath190 .
@xmath202 for different @xmath0 and @xmath122 .
( a ) @xmath203 vs @xmath201 , for @xmath109 , the line represents @xmath204 .
( b ) @xmath205 vs @xmath3 , for @xmath126 and @xmath121 , the lines represent @xmath127 , with @xmath133 and @xmath134 dependent on @xmath0 .
@xmath61 for different @xmath0 and @xmath122 .
( a ) @xmath206 vs @xmath201 , for @xmath109 , the line represents @xmath137 ; ( b ) @xmath207 vs @xmath3 , for @xmath126 and @xmath121 , the lines represent @xmath139 , with @xmath64 and @xmath145 dependent on @xmath0 . | we study the persistent current and the drude weight of a system of spinless fermions , with repulsive interactions and a hopping impurity , on a mesoscopic ring pierced by a magnetic flux , using a density matrix renormalization group algorithm for complex fields .
both the luttinger liquid ( ll ) and the charge density wave ( cdw ) phases of the system are considered . under a jordan - wigner transformation ,
the system is equivalent to a spin-1/2 xxz chain with a weakened exchange coupling .
we find that the persistent current changes from an algebraic to an exponential decay with the system size , as the system crosses from the ll to the cdw phase with increasing interaction @xmath0 .
we also find that in the interacting system the persistent current is invariant under the impurity transformation @xmath1 , for large system sizes , where @xmath2 is the defect strength .
the persistent current exhibits a decay that is in agreement with the behavior obtained for the drude weight .
we find that in the ll phase the drude weight decreases algebraically with the number of lattice sites @xmath3 , due to the interplay of the electron interaction with the impurity , while in the cdw phase it decreases exponentially , defining a localization length which decreases with increasing interaction and impurity strength .
our results show that the impurity and the interactions always decrease the persistent current , and imply that the drude weight vanishes in the limit @xmath4 , in both phases .
pacs : 71.10.pm , 73.23.ra , 73.63.-b |
physical theories are usually created by accumulating some fragments of information which at the beginning do not allow to predict the final structures .
the classical mechanics was formulated by isaac newton in terms of mass , force , acceleration and the three dynamical laws .
it was not immediate to see the lagrangians , hamilton equations and the simplectic geometry behind .
we can not guess the reaction of newton if he were informed that he was just describing the classical phase spaces defined by the simplectic manifolds quite similarly , max planck , niels bohr , louis de broglie , erwin schrdinger and werner heisenberg could not see from the very beginning that the physical facts which they described would be reduced by born s statistical interpretation to the hilbert space geometry ( as it seems , neither hilbert could predict that ) . yet , once accepted that the pure states of a quantum system can be represented by vectors of a complex linear space and the expectation values are just quadratic forms , the hilbert spaces entered irremediably into the quantum theories .
together appeared the `` density matrices '' as the mathematical tools representing either pure or mixed quantum ensembles .
their role is now so commonly accepted that its origin is somehow lost in some petrified parts of our subconsciousness : an obligatory element of knowledge which the best university students ( and the future specialists ) learn by heart . however , is it indeed necessary ?
can indeed the interference pictures of particle beams limit the fundamental quantum concepts to vectors in linear spaces and `` density matrices '' ?
the desire to find some deeper reasons led a group of authors to postulate the existence of an `` intrinsic logic '' of quantum phenomena , called the _ quantum logic _
@xcite . generalizing the classical ideas , it was understood as the collection @xmath0 of all statements ( informations ) about a quantum object , possible to check by elementary `` yes - no measurements '' . following the good traditions ,
@xmath0 should be endowed with _ implication _ ( @xmath1 ) , and _ negation _ @xmath2 .
the implication defines the partial order in @xmath0 ( @xmath3 reinterpreted as @xmath4 ) , suggesting the next axioms about the existence of the lowest upper bound @xmath5 ( `` or '' of the logic ) and the greatest lower bound @xmath6 ( `` and '' of the logic ) for any @xmath7 .
the `` negation '' was assumed to be involutive , @xmath8 , satisfying de morgan law : @xmath9 as well as other axioms granting that @xmath0 is an orthocomplemented lattice @xcite . until now
, the whole structure looked quite traditional . with one exception : in contrast to the classical measurements , the quantum ones do not commute , which traduces itself into breaking the _ distributive law _
@xmath10 obligatory in any classical logic .
the quantum logic was non - boolean !
an intense search for an axiom which would generalize the distributive law , admitting both classical and quantum measurements , in agreement with birkhoff , von neumann , finkelstein @xcite and thanks to the mathematical studies of varadarajan @xcite convinced c. piron to propose the _
weak modularity _ as the unifying law . to some surprise ,
the subsequent theorems @xcite exhibit certain natural completeness : the possible cases of `` quantum logic '' are exhausted by the boolean and hilbertian models , or by combinations of both . as pointed out by many authors
this gives the theoretical physicists some reasonable confidence that the formalism they develop ( with hilbert spaces , density matrices , etc . )
does not overlook something essential , so there will be no longer need to think too much about abstract foundations .
however , is nt this confidence a bit too scholastic
? it can be noticed that the general form of quantum theory , since a long time , is the only element of our knowledge which does not evolve . while the `` quantization problem '' is formulated for the existing ( or hypothetical ) objects of increasing dimension and flexibility ( loops , strings , gauge fields , submanifolds or pseudo - riemannian spaces , non - linear gravitons , etc . ) , the applied quantum structure is always the same rigid hilbertian sphere or density matrix insensitive to the natural geometry of the `` quantized '' systems .
the danger is that ( in spite of all `` spin foams '' ) we shall invest a lot of effort to describe the relativistic space - times in terms of the perfectly symmetric , `` crystalline '' forms of hilbert spaces , like rigid bricks covering a curved highway .
is there any other option ?
the alternatives arise if one decides to describe the statistical theories in terms of geometrical instead of logical concepts .
what is the natural geometry of the statistical theory ?
it should describe the pure or mixed particle _ ensembles _
( also ensembles of multiparticle systems , including the mesoscopic or macroscopic objects ) .
suppose that one is not interested in the total number of the ensemble individuals , but only in their `` average properties '' .
two ensembles with the same statistical averages can not be distinguished by any statistical experiments : we thus say that they define the same _
state_. now consider the set @xmath11 of all _ states _ for certain physical objects . even in absence of any analytic description ,
there must exist in @xmath11 some simple empirical geometry .
given any two states @xmath12 ( corresponding to certain ensembles @xmath13 , @xmath14 ) and two numbers @xmath15 , @xmath16 , consider a new ensemble @xmath17 formed by choosing randomly the objects of @xmath13 and @xmath14 with probabilities @xmath18 and @xmath19 ; its state , denoted @xmath20 is the _ mixture _ of @xmath21 and @xmath22 in proportions @xmath23 .
if in turn both @xmath24 are mixtures of @xmath25 , then some more information is needed to determine the contents of @xmath26 and @xmath27 in @xmath28 .
it can be most simply expressed by representing @xmath11 as a subset of a certain _ affine space _ @xmath29 , where @xmath30 becomes a linear combination .
for any two points @xmath12 all mixtures @xmath30 ( @xmath31 , @xmath16 ) form then the straight line interval between @xmath21 and @xmath22 , contained in @xmath11 .
hence , @xmath11 is a convex set @xcite . to describe the limiting transitions
, @xmath29 must possess a topology and @xmath11 should be closed in @xmath29 .
the information encoded in the convex structure of @xmath11 might seem poor : it tells only which states are mixtures of which other states ( see fig . [ fig:1 ] ) .
yet it turns out that it contains all essential information about both , logical and statistical aspects of quantum theory . and @xmath26 would represent the pure states .
all points in the interior are mixed states and do not allow a unique definition of the pure components .
thus , e.g. , the state @xmath32 could be represented as a mixture of @xmath21 and @xmath22 or @xmath26 and @xmath27 or in any other way .
[ fig:1 ] ]
the boundary of @xmath11 contains some special points @xmath28 , which _ are not _ untrivial combinations @xmath30 with @xmath33 of any two different points @xmath34 of @xmath11 .
these points , called _ extremal _ , represent the physical ensembles which _ are not mixtures _ of different components , and so are called _
pure_. all subensembles of a pure ensemble define the same _ pure state _
@xmath28 , which therefore represents also the quality of each single ensemble individual .
the convex geometry permits to describe as well more general ensemble properties which might be attributed to the single individuals .
note that , in general , the property of ensemble is not shared by the individuals ( e.g. , a human ensemble can contain 50% of men and 50%of women , but each individual , in general , has only one of these qualities ) .
we say that the subset @xmath35 defines a _ property _ of the single objects if : 1 .
it resists mixing , i.e. , @xmath36 , @xmath15 , @xmath16 @xmath1 @xmath37 ( meaning that @xmath38 is a convex subset of @xmath11 ) , 2 . if the property of mixture is shared by every mixture components , i.e. , @xmath39 and @xmath40 , @xmath25 , @xmath41 .
the subset @xmath35 which satisfies 1 . and
is called a _ face _ of @xmath11 . the whole @xmath11 and
the empty set @xmath42 are the improper faces : all other are plane fragments of various dimensionalities on the boundary of @xmath11 ( see fig .
[ fig:2 ] ) .
in particular , each extremal points is a one point face . in what follows
, we shall be most interested in the topologically _ closed _ faces of @xmath11 representing the `` continuous properties '' of the ensemble objects .
further on by _ faces _ we shall mean _ closed faces_. their whole family @xmath43 admits a partial ordering @xmath44 identical with the set theoretical inclusion : the relation @xmath45 means that the property @xmath46 is _ more restrictive _ than @xmath47 , or @xmath46 _ implies _ @xmath47 .
as easily seen , the intersection of any family of faces is again a face of @xmath11 .
hence , for any two faces @xmath48 there exists also their _ smallest upper bound _ , or _
@xmath49 , defined as the intersection of all faces containing both @xmath46 and @xmath47 .
the set @xmath43 with the partial order @xmath44 ( i.e. , implication ) and operations @xmath50 , @xmath51 is thus a complete lattice generalizing the `` quantum logic '' of the orthodox quantum mechanics : it might be called the _ logic of properties_. we shall see that it does not necessarily include _ negation _ , but it admits a natural concept of _ orthogonality _ @xcite
. represent properties of the single ensemble individuals . the picture in perspective permits to see that @xmath46 and @xmath47 are not orthogonal . ]
a natural counterpart of quantum ensembles are the measuring devices and the simplest such devices are _ particle counters_. by a _ counter _ we shall understand any macroscopic body sensitive ( either perfectly or partly ) to the presence of quanta .
our assumption is also , that each counter reacts only to the properties of each single ensemble individual , without depending on the the rest . in mathematical terms
, each counter can be described by a certain functional @xmath52 $ ] . whose values @xmath53 for any @xmath54 mean the fraction of particles in the state @xmath28 detected by the counter @xmath55 .
if @xmath56 , then the counter @xmath55 detects perfectly all @xmath28-particles , if @xmath57 , it overlooks a part , but if @xmath58 , then @xmath55 is completely blind to the @xmath28-particles .
moreover , if @xmath55 reacts only to single ensemble individuals , then for any mixed state @xmath20 it will detect independently both mixture components : @xmath59 , meaning that @xmath55 is a _ linear functional _ on @xmath11 .
we shall assume , that the values of counters permit to distinguish the different points @xmath54 and moreover , they induce a physically meaningful topology , in which they are _ eo ipso _ continuous .
each continuous , linear functional @xmath55 taking on @xmath11 the values @xmath60 will be called _
normal_. mathematically , the counters are , therefore , the _ normal functionals_. to get their geometric image , assume that the surrounding affine space @xmath61 is spanned by @xmath11 .
hence , every linear functional @xmath55 on @xmath11 defines a unique linear functional on @xmath29 which will be denoted by the same symbol @xmath55 . if @xmath62 on @xmath11 , then @xmath62 on @xmath29 .
if not , then the equations @xmath63 , ( @xmath64 ) split @xmath29 into a continuum of parallel hyperplanes on which @xmath55 accepts distinct constant values . due to the linearity , @xmath55 is completely defined by the pair of hyperplanes on which it takes the values 0 and 1 .
if @xmath55 is normal , @xmath11 is contained in the closed belt of space between both planes .
the question arises , how ample is the set of physical counters ?
since no restrictions are evident , we shall assume that each normal functional represents a particle counter which at least in principle can be constructed .
all distinct ways of counting particles can be thus read from the convex geometry of @xmath11 @xcite .
they turn out closely related with the collection of hyperplanes and those are related with _
faces_. indeed , the hyperplanes @xmath65 and @xmath66 of a counter do not cross the interior of @xmath11 , but can `` touch '' its boundary . as one can easily show ,
their common parts with the border @xmath67 are two `` opposite '' faces ( properties ) of @xmath11 , which awake completely different reactions of the counter : while detecting all particles on one of them , it ignores completely the particles on the other .
any two faces @xmath46 , @xmath47 , for which there exists at least one , so sharply discriminating counter , will be called _ excluding _ or _ orthogonal _ @xmath68
. the `` logic of properties '' , therefore , is a lattice with the relations of _ inclusion _ @xmath69 and _ exclusion _
@xmath70 though without a unique ortho-_complement _
( since for any @xmath71 , amongst all elements orthogonal to @xmath38 no greatest one must exist ) .
apart from orthogonality , the next geometry element of @xmath11 describes the selectivity limits of quantum measurements . given a pair of pure states @xmath72 , consider the family of all counters @xmath55 detecting unmistakeably all particles of the state @xmath28 , i.e. , @xmath73 .
can they ignore completely the particles of the state @xmath74 ?
in general , the answer is negative .
the following lower bound over the counters @xmath55 : @xmath75 called the `` detection ratio '' @xcite , if non - vanishing , describes a minimal fraction of @xmath74-particles which must infiltrate any experiment programmed to detect the @xmath28-state .
the geometric character of this quantity is defined just by convex structure of @xmath11 , which determines the support planes ( see fig .
[ fig:3 ] ) . for some hypothetical ensembles .
the parallel support lines @xmath76 and @xmath77 represent a counter detecting all @xmath28-particles , blind to @xmath78-particles , while the lines @xmath65 and @xmath66 correspond to another counter , detecting all @xmath28-particles , but the minimal possible fraction @xmath79 of the @xmath74-particles .
hence , the detection ratio @xmath80 . ] the information contained in might be significantly weaker if the pure state @xmath28 were not _ exposed _
, i.e. , determined completely as the intersection of @xmath11 and at least one support hyperplane .
such cases do not occur in the orthodox qm , but belong to the general convex set geometry ( see ( * ? ? ? *
fig12 ) ) .
in the orthodox theory the pure states are represented by vectors in a complex , linear space ( an inspiration from the observed interference patterns ) and all measured expectation values are real , quadratic forms of the state vectors @xmath81 ( the consequence of born s statistical interpretation ) .
the mixed states are the probability measures on the manifold of pure states ( the projective hilbert sphere ) . however , since the statistical averages are no more than quadratic forms , the ample classes of probability measures ( interpreted as the prescriptions of forming mixtures ) are physically indistinguishable . the faithful representation of the mixed states as the equivalence classes explains the origin of the `` density matrices '' @xcite .
the elements of the convex geometry provide also an alternative interpretation of the unitary invariants @xmath82 called currently the `` transition probabilities '' .
in fact if @xmath11 is the convex set of density matrices in a hilbert space and if two pure states are represented as @xmath83 and @xmath84 ( @xmath85 ) then the elementary lemma shows that @xmath86 i.e. , the commonly used invariant turns out the detection ratio @xcite , revealing an additional sense of the `` transition probabilities '' .
in fact , as once noticed by peter bergman , the deepest picture of a physical theory is obtained not so much by telling what is possible , but rather by `` no go principles '' , defining what is ruled out ( e.g. , the equivalence principle in general relativity , or the uncertainty principle in qm ) .
one such law emerges from the identity . .
once fixed the image , the geometry of @xmath11 determines uniquely the `` transition probabilities '' between any pair of pure states . on the figure : the pair of support planes @xmath65 and @xmath66 illustrates the maximally selective counter which detects all photons in the vertical polarization @xmath87 , but none in the horizontal @xmath88 .
the intermediate values of @xmath55 on the congruence of parallel planes intersecting @xmath11 define the transition probabilities from all other states to the vertical one @xmath87 . ] indeed , @xmath82 not only defines the fraction of the @xmath89-particles accepted by the @xmath81-filter , but also the fundamental impossibility of accepting less ! every physical process which leads to a certain macroscopic effect for _ all _ @xmath81-particles , must lead to the same effect at least for the fraction @xmath82 of @xmath89-particles .
the purely geometric nature of this law , independent of any analytic expression can be best illustrated by the bloch sphere of the photon polarization states ( see fig . [ fig:4 ] ) on which the linear polarizations occupy a great circle ( the `` equator '' ) , circular polarizations are the poles , and the remaining surface points are the elliptic polarizations .
the interior of the sphere collects the mixed polarizations , the center @xmath90 representing the complete polarization chaos .
the pair of tangent planes @xmath65 and @xmath66 represents a maximally selective counter detecting all photons in the vertical polarization and rejecting the orthogonal one .
the geometry of the sphere @xmath11 determines immediately the `` transition probabilities '' between any two pure states without the need of using the analytic @xmath91 representation ( thus , e.g. , the detection ratio between any linear and circular polarization is 1/2 ) . in case of non - classical ensembles ,
the geometry of @xmath11 expresses still more fundamental law about the indistinguishability of quantum mixtures , the phenomenon which appears if @xmath11 is not a simplex . _
given a mixed ensemble of non - classical objects , one can not , in general , retrospect and find out how the mixture has been prepared .
two mixtures composed of different collections of pure states can be physically indistinguishable ( see also fig .
[ fig:1 ] ) . _ in the bloch sphere of polarization states ( fig . [ fig:4 ] )
the effect is exceptionally simple for its center @xmath90 which can be represented equivalently as a mixture of any pair of orthogonal linear polarizations , or two opposite circular polarizations or in any other way : @xmath92 hence , once having the mixed state @xmath90 one can not go back and identify its pure components : a kind of statistical paradox making it quite difficult to check experimentally some semantic curiosities of the existing theory , which are still waiting for a good inspiration !
the structures reported here contain a certain puzzle .
it is basically not strange that the convex geometry is a language of statistical theories .
yet , it was not expected that the structure of an arbitrary convex set @xmath11 contains the equivalents of principal quantum mechanical concepts .
their properties are distorted , but their meaning is similar .
thus , the _ logic of properties _ is an analogue of the _ quantum logic _
@xcite and the _ detection ratios _ are equivalents of the orthodox `` transition probabilities '' . in many aspect
the hilbertian schemes are distinguished by their maximal regularity and almost crystalline symmetry : to each face of @xmath11 , ( read subspace ) , corresponds a unique orthocomplement , etc .
might this resemble the relation between the euclidean and riemannian geometries ?
if so then could it happen that in some circumstances the quantum systems could obey the generalized convex geometry , dissenting from the hilbertian structure ? in the intents of finding a synthesis of the lattice ( `` logical '' ) and probabilistic interpretations ( since j. von neumann @xcite ) the statistical aspects , in general ,
were subordinated to the assumed structure of the orthocomplemented lattice , and the answer of the axiomatic approach was always the same : the quantum mechanics must be exactly as it is .
this belief turned even stronger due to the theorem of gleason @xcite , as well as due to the profound and elegant generalizations of the algebraic approaches of gelfand and naimark @xcite , haag and kastler @xcite , pool @xcite , araki @xcite , haag @xcite , and other authors , who never resigned from the hilbert space representations . curiously , until today , these convictions find also also a strong support in the well known book of g. mackey @xcite in which , however , the axiomatic approach has some self - annihilating aspects : after a laborious presentation of six axioms on quantum logic @xmath93 , the seventh axiom tells flatly that the elements of the logic are closed vector subspaces of a hilbert space , thus making all previous axioms redundant ! ( a short report on this school of axiomatics , see h. primas @xcite ) .
some opposition is not so surprising
the first descriptions of qm based exclusively on the convex geometry belong to g. ludwig @xcite , though the author adopted axioms in fact limiting the story to the orthodox scheme .
the hypothesis about the possibility of quantum mixtures obeying non - hilbertian geometries was formulated by the present author @xcite , also by davies and lewis @xcite .
the hypothetical geometries succeeded to awake both positive and hostile reactions .
roger penrose at some moment hoped that the atypical structures might tell something about the nonlinear graviton @xcite , though later on he complained @xcite that they give a pure statistical interpretation , without any analytical entity behind ( though inversely , the nonlinear graviton of penrose is a pure analytical entity without any statistical interpretation ! ) .
kibble and s. randjbar - daemi followed @xcite describing the _ classical gravity _ in interaction with the generalized quantum structure @xcite .
some other authors in philosophy of physics stay firmly on the ground of the orthodox theory .
nonetheless , they do nt escape objections .
while putnam considers the orthocomplemented structure of hilbert spaces the `` truth of quantum mechanics '' @xcite ( taking the side of mackey ? ) , john bell and bill hallet @xcite adopt the generalized design proposed in @xcite to show the weakness of putnam s argument .
however , the deformed geometries , if real , must occur in some concrete physical circumstances . where should we look for them ? as it seems , the most natural possibility is to look for nonlinear variants of quantum mechanics .
in fact , already some simple nonlinear cases of the schrdinger s equation admit nonquadratic , positive , absolutely conservative quantities which could be used to define the probability densities @xcite . the quantum mechanics with logarithmic nonlinearity permits to define consistently the reduction of the wave packets @xcite . yet , as shown by haag and bannier @xcite , subsequently also in @xcite , the nonlinear wave equations lead to high mobility of quantum states , breaking the quantum impossibility principles .
the most basic difficulty was noticed by n. gisin , who had shown that if the linear evolution law of quantum states were amended by adding some nonlinear operations , then the breaking of the mixture indistinguishability would make possible to read the instantaneous messages between parts of the entangled particle systems @xcite .
the simplest case would occur in a variant of epr experiment for the sequences of photon pairs in the singlet polarization state @xmath94 emitted in two opposite directions . according to the present day theory
the polarization measurements on the left photons can produce at distance ( due to the correlation mechanism ) any desired mixture of the right photons ( or vice versa ) .
as long as mixtures are indistinguishable , this does not transmit information .
however , if the observer of the right photon states could cause their nonlinear evolution , he could distinguish the quantum mixtures , thus reading hidden information and reconstructing without delay the measurements performed by his distant counterpart on the left epr photons .
so , is the nonlinear qm impossible ?
perhaps , we should not overestimate the axiomatic approaches .
what they usually tell is that we can not modify just one element of the theory , while leaving the whole rest intact .
if in the last decade of xix century some excellent axiomaticians tried to formulate reasonable axioms defining the space - time structure , they would prove beyond any doubt that the space - time must be galilean !
yet , it is not .
the deviations ( in our normal conditions ) are very small , but rather important
what can be impossible in qm , is to conserve the orthodox representation of pure states as the `` rays '' in a complex hilbert space , together with the tensor product formalism , and with the unitary background evolution , but to extend it by adding some nonlinear evolution operations and to expect that the instantaneous information transfers will be still blocked .
however , the whole deduction might be already overloaded by too many axioms . if the evolution were extended by some nonlinear operations , then in the first place
, we would loose the hilbert space orthogonality together with the trace rules for probabilities even without worrying about the superluminal messages returning to the spin or polarization qubits , the possibilities of generalizing the hilbertian structures depends not so much on axioms but rather on precise knowledge of probabilities .
if indeed exactly orthodox , then may be , the qubits can only rigidly rotate
the problems of systems traditionally described by multi- or infinitely - dimensional hilbert spaces are more difficult .
the questions of hans primas , perhaps are still waiting for a good answer : _ does quantum mechanics apply to large molecular systems ? _ _ why do so many stationary states not exist ? _
( see ( * ? ? ? * and 12 ) ) .
indeed , even the problem of how to create in practice the one particle states described by arbitrary wave packets deserves systematic studies @xcite .
as recently noticed , the non - linear modifications of quantum dynamics instead of just extending the techniques of the state manipulation might introduce _ constrains _ , with the restricted @xmath11 no longer obeying the hilbertian geometry @xcite ; an option which might be worth exploring .
all the attempts to see more freedom in quantum structures need some empirical criteria , which would permit to detect the new geometries if they exist . in case of classical state structures
such criteria were found by john bell , in form of bell inequalities expressing the boolean geometry of the state mixtures .
their breaking was the sign that the ensembles are non - classical .
the problems of quantum ensembles , e.g. , whether they indeed obey the hilbert space geometry , are significantly more involved .
the initiative of our colleagues @xcite to describe them in terms of `` apophatic '' ( forbidden ) properties continues indeed the effort of john bell on the new theory level .
some interesting cases might be the `` cross sections '' of @xmath11 , resembling the `` constrained qm '' discussed in @xcite , and the projections ( the collapsed @xmath11 caused by deficiency of observables ? ) .
simultaneously , the mathematical research presented in @xcite is an unexpected school of modesty for all of us , who believed to understand so well the property of nice objects called the `` density matrices '' .
now it turns out that we did not even know the properties of the simple qutrit !
needless to say , should any of the `` forbidden properties '' be detected for any statistical ensemble in some physical conditions , this will be the proof that the theory is at the new conceptual level .
interesting , what about all that will think the physicists of xxii century ? | our knowledge and ignorance concerning the geometry of quantum states are discussed . |
rr lyr variables are important tools in the investigation of many fundamental astrophysical problems
. they provide crucial constraints on the physical mechanisms driving radial oscillations and their interplay with stellar evolution @xcite .
furthermore , rr lyr stars offer the opportunity to study the morphology of the horizontal branch and the oosterhoff dichotomy @xcite .
current empirical evidence indicates that the mean period of fundamental mode ( rrab or fu ) rr lyr stars in galactic globular clusters ( ggcs hereafter ) shows a dichotomous distribution at 0.55 ( ooi ) and 0.65 ( ooii ) days @xcite , where the latter group is more metal - poor .
there is also evidence that the dichotomy is the aftermath of the hysteresis mechanism suggested by @xcite , i.e. that the pulsation mode depends on the direction of the evolution inside the instability strip @xcite . @xcite
employed the rr lyr stars as a probe to identify the two main stellar populations in the galaxy , as well as to study the stellar content of the galactic bulge through low - reddening regions @xcite . beyond stellar evolution ,
rr lyr variables have also played a key role in providing estimates of cosmological parameters , and have been instrumental in measuring the distances to a sizable sample of ggcs .
this allowed the estimation of absolute cluster ages , and in turn set the lower limit on the age of the universe @xcite .
moreover , rr lyr variables have been used to estimate the primordial helium content using the a - parameter , i.e. the mass to luminosity relation of low - mass central helium - burning stars @xcite .
rr lyr stars are also the most commonly adopted population ii distance indicator . with a lower mass than classical cepheids
, they have the key advantage to be ubiquitous , and have been identified in both early and late type stellar systems @xcite .
their individual distances can be evaluated using multiple diagnostics , including a visual magnitude - metallicity relation @xcite and a statistical parallax @xcite .
more importantly , they obey to well defined near - infrared ( nir ) period - luminosity ( pl ) relations @xcite .
these pl relations extend to mid - infrared ( mir ) bands where they have the potential of being very accurate distance indicators due to lower extinction and smaller intrinsic scatter @xcite . the use of the i , v - i reddening free period - wesenheit ( pw ) relation to estimate the individual distances of rr lyr stars dates back to @xcite and to @xcite .
a more recent theoretical framework developed by @xcite further supports the use of optical , optical - nir , and nir period - wesenheit - metallicity ( pwz ) relations to determine individual distances of rr lyr stars .
empirical validations to the above pulsation and evolutionary predictions have been provided by @xcite for rr lyr stars in the ggc m4 and by coppola et al .
( 2015 , in preparation ) for rr lyr stars in the carina dwarf spheroidal .
the carnegie rr lyr program ( crrp ) aims to take full advantage of the unique characteristics of these stars in order to reduce the remaining sources of uncertainty in the hubble constant to @xmath1% .
rr lyr mir pl relations will be used as the foundation of an independent population ii cosmological distance scale to calibrate trgb distances for nearby distances , which in turn can be used to calibrate type ia supernova distances . to achieve this goal , we observed over 1,700 rr lyr variables in 31 selected ggcs , as well as @xmath2 rr lyr stars in strategically distributed areas in the galactic halo and the bulge .
in addition , we observed 48 of the nearest , brightest and less reddened rr lyr stars intended to be used as zero point calibrators .
these observations have been conducted during the warm mission of the spitzer space telescope @xcite infrared array camera ( irac , * ? ? ?
* ) , at 3.6 and 4.5 wavelength , with a cadence designed to obtain complete coverage of each rr lyr variable over at least one full period . in this work we focus on the ggc messier 4 ( m4 , ngc 6121 ) .
this cluster is an ideal laboratory for stellar population studies given its proximity to the sun , which allows us to obtain accurate photometric and spectroscopic data for member stars well below the main - sequence turnoff . due to these characteristics
, m4 has been the subject of intensive observational campaigns over a wide range of wavelengths .
it has a well characterized differential extinction of @xmath3 mag , where the uncertainty is the dispersion due to differential reddening @xcite , and mean metallicity of @xmath4 = -1.10 $ ] , ( * ? ? ?
* and references therein ) . in support of the crrp program ,
we have analyzed available multi - epoch optical and nir data , leading to the identification and characterization of 45 rr lyr variables @xcite . from these observations ,
we have derived accurate distance moduli based on optical and nir pl and pw relations @xcite . in this paper
we combine our previous results with the new mir data obtained as part of the crrp campaign .
in section [ sec : obs ] we present our new spitzer photometry .
light curves for all the rr lyr variables in our sample are measured in section [ sec : mags ] . in section [ sec : pl ] we derive mir pl and period - color ( pc ) relationships for the cluster rr lyr variables , while in section [ sec : dm ] we calculate the m4 distance modulus by calibrating our pl zero point using five nearby calibrator rr lyr stars with known parallax , also observed as part of the crrp program .
dependence of the pl zero point from metallicity is also discussed in section [ sec : dm ] , while section [ sec : concl ] summarizes the results of this work .
the results of our ground - based optical and nir monitoring of the stellar population in m4 have been published in @xcite and @xcite . in this paper
we extend our wavelength coverage to the mir , by including multi - epoch 3.6 and 4.5 photometry of a large portion of the cluster .
our irac observations of m4 were executed on 2013 june 2 - 3 as part of the warm spitzer space telescope cycle 9 ( pid 90088 ) .
the cluster was observed at irac 3.6 and 4.5 @xmath5 bands at 12 equally spaced epochs over @xmath6 hours . for each epoch we obtained 30 frames with a medium scale five - point gaussian dither pattern and a @xmath7 mapping , taken using the irac full frame mode with a frame time of 30 seconds to provide a median total exposure of 804 seconds per pixel .
an @xmath8 area at the center of the cluster was covered in both the 3.6 and 4.5 bands . due to the irac focal plane array configuration , in the outer regions of the cluster
only one band is available , with the area just north of the center of the cluster only covered at 3.6 and the region south of the center only covered at 4.5 .
the total area mapped by our observations covers a strip of @xmath9 , roughly cutting north - to - south through the center of the cluster , as shown in figure [ fig : map ] .
this is about one third of the total cluster area that was mapped in our ground - based optical and nir observations , but is still @xmath10 the size of the half - mass radius of m4 ( @xmath11 according to @xcite ) .
the crrp irac data were reduced starting from the basic calibrated data ( bcds ) generated by the spitzer irac pipeline version s19.1.0 .
photometry was performed using the most recent version of the daophot / allstar / allframe suite of programs for crowded - field stellar photometry @xcite . in order to mitigate the effect of blending in the photometry of the m4 crowded field , we used an input source catalog derived by the higher angular resolution optical and nir images described in @xcite .
in addition to higher resolution , this input source catalog is also deeper . for the optical and nir ,
accurate photometry was achieved down to 24 and 22 magnitudes respectively , compared to 17 mag for the mir images .
this catalog was matched to the spitzer data after applying higher order geometric corrections to the individual irac bcds than the default ones provided by the spitzer irac pipeline . with this procedure
we matched 17,595 sources at 3.6 and 16,313 sources at 4.5 , for which we measured an instrumental magnitude with variable point spread function ( psf ) fitting .
the astrometric accuracy of our final source catalog is of the order of @xmath12 . for the optical and nir images , good photometry
was achieved down to 24 and 20 magnitudes respectively the photometry was calibrated to standard irac vega magnitudes using aperture photometry on mosaic images .
the mosaics were generated using the iracproc package @xcite , a perl data language wrapper script for the standard spitzer science center mosaicking software mopex @xcite .
we adopted a mosaic pixel scale of 0.61@xmath13/pix , or half the native irac pixel scale .
the stars we selected for aperture photometry calibration are far from the center of the cluster to reduce blending .
in addition , we removed any possible variable stars , by avoiding sources with a welch & stetson variability index @xcite greater than @xmath14 from the median .
we selected a total of 38 stars at 3.6 and 44 at 4.5 .
we used an aperture of 3 native irac pixels ( @xmath15 ) with an annulus of 3 to 7 pixels ( @xmath16 to @xmath17 ) .
the aperture corrections necessary to convert magnitudes derived with this aperture / annulus combination to the standard irac 10 pixel aperture photometry are 1.1284 and 1.1274 for 3.6 and 4.5 @xmath5 respectively , calculated on isolated stars in the mosaics .
the calibrating stars cover a range of 4 magnitudes and are distributed randomly across the array , outside of the central region . despite the relatively small size of our adopted photometric apertures , the crowding of the field even outside the center of the cluster ( m4 is projected behind the scorpius - ophiuchus cloud complex ) still results in blending contamination of aperture photometry for our calibration sources .
these blended sources are however in many cases resolved by the psf fitting procedure performed in daophot on the higher resolution visible and nir images , from which our input catalog is derived .
we thus used our point source photometric catalog to improve the photometric zero point calibration by subtracting from the calibrator aperture photometry the flux ( measured from the psf - fitting photometry ) of any blended sources found within our chosen aperture . for each blended source we estimated the contaminating flux by measuring the fraction of its flux falling within our 3 pixel aperture , as a function of its centroid distance from the aperture s center .
for this task we used the irac psf described in @xcite , taking advantage of its higher sampling ( @xmath18 of the native irac pixels ) to derive corrections for fractional pixel shifts . at 3.6 17 of our 38 calibrators
were found to be contaminated by an average of 0.03 mag .
the 4.5 region was less crowded , resulting in only 7 out of 44 contaminated calibrators , but contaminated by 0.11 mag on average .
the magnitudes of the de - blended calibrators were then subtracted from their daophot instrumental magnitudes to derive the final calibration zero point .
we found no trend in the calibration zero point with either brightness or array location .
the uncertainty in our final calibration is 0.018 mag at 3.6 and 0.017 mag at 4.5 , based on the standard deviation between the instrumental and measured magnitudes .
this standard deviation of the calibration increases dramatically for the fainter stars in our calibration sample . if we select only calibrating stars brighter than 14 mag , then the error in the calibration is reduced to 0.015 at 3.6 and 0.013 at 4.5 . at the end of this process
, we obtained a photometric catalog in both irac bands for all point sources detected in our field of view . from this catalog we matched 37 of the 45 rr lyr stars with optical photometry derived by @xcite , 26 of them fundamental mode ( fu , rrab ) pulsators , and 11 pulsating in first overtone ( fo , rrc ) .
of the remaining eight rr lyr stars in the cluster , seven are outside the area mapped in at least one of the two irac bands , and one ( v64 ) is a field variable in the background of the cluster , too faint to be detected by irac .
all m4 rr lyr stars are indicated in figure [ fig : map ] by squares , with open red and filled blue symbols for fu and fo rr lyr stars detected in at least one irac band .
the final magnitude of each non - variable star was derived by averaging the value measured in all epochs and in all dithers . for the rr
lyr variables we instead preserved the photometry for each epoch , averaging only on the individual dithers , after correcting the photometry in each dither for the warm spitzer location - dependent photometric corrections available at the spitzer science center web site .
the photometric uncertainty in each epoch is given by the standard deviation of all of the available measurements .
a sample of the individual data points for one star , v1 , is given in table [ tab : phot ] ; all data is available in the online version of this article .
a few rr lyr stars were covered in only one of the five dither positions , and one magnitude measurement was available in each epoch . for these stars
( v1 , v42 , v2 , and v29 ) the photometric uncertainty is instead derived from the _ repeatability _ parameter provided by daophot , which tends to be larger than the photometric error based exclusively on photon and background noise .
our mir sample of m4 rr lyr variables consists of 37 stars , whose location is shown in figure [ fig : map ] . of those , 26 are pulsating in fundamental mode ( fu , rrab ) and 11 are first overtone pulsators ( fo , rrc ) .
not all sources have complete coverage in the irac dataset : of the 37 rr lyr stars in the area mapped by at least one irac band , only 28 have photometry at both 3.6 and 4.5 , with six sources covered only at 3.6 and three covered only at 4.5 .
the periods of our rr lyr stars range between 0.2275 days ( v49 ) to 0.6240 days ( v39 ) , plus a long - period fu rr lyr ( v52 ) with a period of 0.8555 days .
figure [ fig : cmd ] shows representative color - magnitude diagrams ( cmds ) for m4 based on our mir and optical photometry , with the irac 3.6 magnitude vs. the @xmath19 $ ] ( _ top left _ ) and @xmath20 $ ] ( _ bottom left _ ) colors .
note how the high photometric quality of our data allows a well characterized turn - off in all color combinations , and a well populated sub - giant , rgb and agb branch .
the fo and fu rr lyr stars are represented by filled blue and open red squares respectively .
we see a clear separation of the fo and fu rr lyr stars in the horizontal branch of the cmd , with the fo bluer than the fu variables ( _ right panels _ ) .
we derived mir light curves for each rr lyr in our catalog by phasing their multiple - epoch photometry based on the epoch of maximum ( @xmath21 ) and the period derived by @xcite using the lomb - scargle method from the visible data .
sample light curves are shown in figure [ fig : fo - lightcurves ] for fo variables and [ fig : fu - lightcurves ] for fu variables .
the light curves are shown in order of increasing period . following @xcite , for each star we derived a smoothed light curve using a gaussian local estimation ( gloess ) algorithm , a method where a second - order polynomial is fit locally and points are weighted by their gaussian distance from an interpolation point .
a full description of the gloess method is given in @xcite .
the mean magnitude for all stars was then determined by integrating the smoothed light curve flux intensity over one period , then converting the result back into a magnitude . the variability amplitude in magnitude
was also derived from the smoothed light curve , and the uncertainty is calculated using only the data points that lie within the gaussian window of the maximum and minimum of the smoothed curve .
the uncertainty in the mean magnitude is calculated by the sum in quadrature of photometric uncertainty of each observation ( @xmath22 ) and the uncertainty in the fit ( @xmath23 ) where @xmath24 is the number of observations , @xmath25 is the photometric uncertainty , @xmath26 is the peak - to - peak amplitude , and @xmath27 is the number of uniformly spaced points ( following * ? ? ?
the value of @xmath27 ranges from five to twelve ; five
( e.g. v41 ) when our observations cover more than two periods and twelve when our observations cover one complete period ( e.g. v5 ) .
some light curves however were randomly sampled ( e.g. v52 , where the observations do not cover a full period ) , and the uncertainty in the fit is instead @xmath28 where n is the total number of observations .
table [ tab : smean ] gives the 3.6 and 4.5 irac mean magnitudes and amplitudes for the 37 rr lyr stars in the irac field .
figure [ fig : bailey ] shows the luminosity amplitude vs. period ( bailey diagram ) for all rr lyr stars : filled blue squares for fo and open red squares for fu pulsators .
black crosses indicate fu rr lyr stars identified in @xcite as blazhko variables ( the blazhko effect is characterized by a conspicuous modulation of amplitude and phase ; @xcite ) .
recent long - term monitoring of blazhko rr lyr by the kepler satellite has provided evidence of not only period - doubling @xcite but also the possible occurrence of additional periodicities on a time - scale longer than the typical blazhko period @xcite .
as expected the fo and fu rr lyr variables are well separated in both period and amplitude , and the amplitude of the fu pulsators show a clear trend of decreasing with period .
figure [ fig : ap ] shows instead the amplitude ratio between optical bands and the irac 3.6 band .
the amplitude ratio plot can be used as a diagnostic for blends of stars with different color temperature .
two fo rr lyr stars ( v40 and c1 ) show an unusually high optical to mir ratio , possibly indicative of the presence of a faint optically - bright ( hot ) companion or blend .
another fo rr lyr ( v49 ) shows instead an unusually low amplitude ratio , possibly the consequence of having an infrared - bright ( cool ) companion ( or blend ) .
it is also interesting to note that several ( four out of seven at 3.6 and four out of six at 4.5 ) of the fu rr lyr stars that lie below the general amplitude ratio trend are blazhko variables , possibly due to these stars being observed at different phases in their amplitude modulation cycle ( the optical and mir data were not acquired simultaneously ) .
if this is the case , the remaining fu rr lyr stars in the same amplitude ratio group ( v5 , v3 , and v52 ) could also be blazhko stars that were missed by the period studies ( or they could be stars with a faint cool companion or blend ) .
the left panels of figure [ fig : pl - warm ] show the irac pl relation for fo ( filled blue squares ) and fu ( open red squares ) rr lyr stars in m4 .
blazhko variables are indicated by black crosses .
the statistical uncertainties of the mean magnitudes are smaller than the size of the plotted symbols , and are not shown .
residuals from the pl relation were used as a diagnostic for blended variables .
the central regions of m4 are not particularly dense , and no clear cutoff in radial distance from the center was apparent .
instead we employed a threshold in sigma to identify candidate blends .
two stars , v20 and v21 , in the central regions of the cluster fall outside the threshold .
the variable v20 was over @xmath29 above the fitted pl relation , and as shown in figure [ fig : fu - lightcurves ] , has a very poor quality light curve .
the average magnitude of v21 was more than @xmath30 above the fitted relation .
these stars were not included in the final fit of the pl relations , and are shown in figure [ fig : pl - warm ] as open circles .
all average magnitudes are corrected for extinction ; albeit mir extinction corrections are small compared to optical wavelengths , they still can not be ignored given the high photometric accuracy provided by irac .
we used the total - to - selective extinctions of @xmath31}/e(b - v ) = 0.203 $ ] and @xmath32}/e(b - v ) = 0.156 $ ] from @xcite and a color excess of @xmath33 from @xcite .
this results in extinction corrections of @xmath34 and @xmath35 mag at 3.6 and 4.5 , respectively .
we calculated the zero point and slope ( with their uncertainties and the best fit standard deviation ) for the pl relation using an unweighted least squares fit .
an unweighted fit is preferred to avoid biasing by brighter longer period variables , which tend to have smaller photometric uncertainties .
the pl relations take the form @xmath36 where @xmath37 and @xmath38 are representative of the mean period of the fo and fu variables respectively .
the scatter between the fo and fu pl relations is comparable . due to the small number of fo variables
, we also derived the global pl relations for all rr lyr stars ( figure [ fig : pl - warm ] , _ right _ ) , having _ fundamentalized _ the periods of the fo variables using the relation @xmath39 @xcite .
the fundamentalized pl relation takes the form @xmath40 where @xmath41 is representative of the mean fundamentalized period of all rr lyr stars in our sample .
the zero point @xmath42 , slope @xmath43 , their errors , and the standard deviations for all pl relations are given in table [ tab : pl ] .
note how the secondary modulation of the candidate blazhko variables seem to have no measurable effect on their average magnitudes , as they fit well on the pl relation .
figure 4 in @xcite showed that , using available data , the slope of the rr lyr pl relation is a monotonically increasing function of wavelength , and asymptotically approaches a wavelength - independent slope . in figure [ fig : pl - parameters ] , we recreate this plot by adding the optical and near - infrared slopes from @xcite , the irac mir slopes calculated above , and the results from @xcite .
our data is consistent with the values found by @xcite and @xcite for the wise w1 and w2 bands , and confirm with a high level of accuracy that the slope approaches the value of @xmath44 at infrared wavelengths , as predicted by the period - radius relation in @xcite .
figure [ fig : pc ] shows optical and infrared period - color ( pc ) relations for all m4 rr lyr variables observed with irac ( as before , filled blue and open red squares indicate fundamentalized fo and fu ) .
we fit the data with an unweighted least squares procedure ; the best fit parameters are listed in table [ tab : pc ] .
the figure clearly shows how the slope of the pc relation becomes more shallow as the distance between the two passbands decreases , as expected if the color is a temperature effect related to the radius of the pulsator .
the @xmath45 pc relation , in particular , has been singled out for being the least affected by temperature uncertainties when deriving rr lyr distances with the baade - wesselink method @xcite , as opposed to using the standard @xmath46 color .
our figure shows that pc relationships using mir colors ( and the irac 3.6 magnitude in particular ) are even more advantageous , thanks to the steeper slope ( i.e. @xmath47 for the @xmath48 $ ] relation vs. @xmath49 in the @xmath45 fit ) and slightly smaller dispersion around the best fit relation .
the last panel in figure [ fig : pc ] shows the irac @xmath50 - [ 4.5]$ ] pc relation . in this case
the slope is very shallow ( @xmath51 ) , consistent with zero within @xmath52 .
the scatter around the best fit pc relation is also quite large ( 0.017 mag ) , three times larger than the uncertainty in the average magnitudes of individual stars .
the lack of a significant @xmath50 - [ 4.5]$ ] pc relation for rr lyr stars suggests that these stars do not develop variable co molecular absorption in their atmosphere , which is observed in other pulsating stars in the same instability strip ( e.g. classical cepheids ) .
@xcite found a well defined @xmath50 - [ 4.5]$ ] pc relation for large magellanic cloud cepheids with @xmath53 .
the negative slope of the pc relation and blue colors of these stars is explained by the dissociation and recombination of co molecules in the irac 4.5 band , as previously noted by @xcite .
this effect is particularly strong in cepheids with @xmath54 days , and is reflected in a precise phasing of the light and color curve of these stars .
shorter period cepheids are instead characterized by a flat color curve and red colors .
this is further supported by @xcite , which found that co absorption drops off for temperatures greater than 6000 k. given that these rr lyr stars have higher effective temperatures , earlier spectral types and lower metallicity than cepheids , it is not surprising that we do not find strong co effects in the pc relation and color curve .
the @xmath50 - [ 4.5]$ ] pc relation appears to be flat , even when including galactic rr lyr stars with very different metallicities .
therefore , rr lyr observations at 4.5 are not limited by strong metallicity effects as cepheids are , and this band can be successfully used for distance scale measurements with similar accuracy as the 3.6 band .
the weakness of the 4.5 co band in rr lyr stars , however , prevents this feature from being used as metallicity indicator for this class of variables .
the crrp project relies on establishing rr lyr variables as highly accurate indicators for the first rung of the cosmological distance scale , by using their mir pl relation .
the zero point of this relation , however , needs to be calibrated .
efforts in this direction have been made in the recent past by using galactic rr lyr stars observed by wise @xcite , but the accuracy of the calibration has been limited by the uncertainties in the photometry and distance of the calibrators . as mentioned in section [ sec : intro ] , the crrp project addresses this issue with the observations of 43 galactic rr lyr stars whose geometric parallax will be determined better than 2 - 3% with gaia , and five bright rr lyr stars with parallax already determined by @xcite using the hubble space telescope ( hst ) fine guidance sensor ( fgs ) .
ground - based monitoring and spectral observations of the calibrators is ongoing , to ensure the characterization of their pulsation properties and , if required , to allow the calibration of the metallicity dependence of the pl relation .
the observation of a large sample of ggcs ( themselves used as distance indicators ) as part of the crrp program provides a separate avenue to test the calibration of rr lyr stars and probe for metallicity effects .
the distance of m4 , in particular , has been the subject of intense analysis , most recently by @xcite by using a theoretical and empirical calibration of the rr lyr optical and nir pl relation for the cluster .
the distance modulus derived in this work can be used as an alternative calibration of the mir pl zero point , as well as to test the effects of metallicity on the zero point derived from the five hst / fgs rr lyr calibrating stars .
a similar approach will be followed for the remaining 30 gccs in the crrp program ( spanning a wide range in metallicity ) , for which independent distance moduli will be derived in the optical , nir and mir from their population of rr lyr stars . as mentioned above , at present
there are only five rr lyr variables with available geometric parallax , i.e. the aforementioned stars observed with hst / fgs by @xcite : rr lyr , uv oct , su dra xz cyg ( fu pulsators ) and rz cep ( fo ) .
the four fu pulsators are the same stars used by @xcite to calibrate the rr lyr pl relation in the wise bands . even for these stars , however , the uncertainty in their distance modulus is still large ( up to 0.25 mag ) , significantly reducing their effectiveness as calibrators for the distance scale .
all five stars have been observed as part of the crrp program in order to reproduce the @xcite zero point calibration using the irac 3.6 and 4.5 bandpasses , and with better phase sampling and much smaller ( @xmath55 better s / n ratio ) photometric error .
their basic properties ( derived from @xcite ) are listed in table [ tab : gal ] : their periods range from 0.3086 to 0.6642 days , and their [ fe / h ] ranges from @xmath56 to @xmath57 , significantly more metal - poor than m4 itself . a detailed analysis of the calibration of the zero point using these five rr lyr stars at multi wavelengths will be presented in a forthcoming paper .
the mean magnitude of each star was computed as explained in section [ sec : mags ] .
each calibrator was then corrected for extinction , adopting the @xmath58 reddening from @xcite .
the extinction corrections were computed using the same @xmath59 and @xmath60 relations as in section [ sec : mags ] .
the absolute mean magnitudes , as well as the parallax , lkh correction @xcite , distance moduli , extinction corrections , and apparent mean magnitudes are given in table [ tab : gal ] .
@xcite contains two conflicting results about the parallax to rz cep and although the value of 2.12 mas is their preferred solution ( benedict , private communication ) , we chose to adopt the 2.54 mas solution with an @xmath61 of 1 ( assuming a mean spectral type of a5 @xcite and @xmath62 ) which appears to be more consistent with the new photometric data .
the dominating factor in the uncertainty of the absolute mean magnitudes is the uncertainty in their distances .
based on theoretical nir pl - metallicity ( plz ) relations , the slope of the rr lyr pl relation is expected to be only weakly dependent on metallicity @xcite . therefore , for this work we elected to adopt the well constrained slope of the fundamentalized pl relation derived in section [ sec : pl ] , rather than relying on just five galactic calibrators ( having a smaller range in period than the m4 rr lyr stars ) to simultaneously fit both slope and zero point .
figure [ fig : hst - fit ] shows the resulting pl relation , derived after fundamentalizing the period of rz cep .
the calibrated pl relations derived using a one parameter fit are given by @xmath63}= -2.332(\pm0.106)\log(p ) - 1.176(\pm0.080 ) \\
\nonumber \sigma= 0.095 \\ m_{[4.5 ] } = -2.336(\pm0.105)\log(p ) - 1.199(\pm0.080 ) \\
\nonumber \sigma= 0.095 \end{aligned}\ ] ] where the uncertainty in the zero point is calculated by @xmath64 and @xmath25 are the uncertainties of the absolute magnitudes for the five calibrator stars used in the fit .
the largest factor in the uncertainty of the zero point derived above is due to the uncertainty in the distance modulus of the calibrators , but it is a factor of 3 better than the value obtained in @xcite where both the slope and zero point were fit simultaneously and only four stars were used . in both irac bands , the slope we obtain is marginally shallower ( but still consistent , within the error ) than the values published by @xcite and @xcite for the wise w1 and w2 bands .
further improvements in the uncertainty of the zero point will be possible when more accurate parallaxes for all our calibrator rr lyr stars are obtained with the gaia mission .
from the zero point calibration of the mir rr lyr pl relation in the previous section , we can derive the true distance modulus of m4 : @xmath65 } = 11.406 \pm 0.010 \ , ( \textrm{stat } ) \pm 0.080 \ , ( \textrm{syst } ) \\
\nonumber \pm 0.015 \ , ( \textrm{cal } ) \pm 0.020 \ , ( \textrm{ext } ) \\ \mu_{[4.5 ] } = 11.391 \pm 0.010 \ , ( \textrm{stat } ) \pm 0.080\ , ( \textrm{syst } ) \\
\nonumber \pm 0.013 \ , ( \textrm{cal } ) \pm 0.016 \ , ( \textrm{ext } ) \end{aligned}\ ] ] where we have adopted the m4 pl relation using the fundamentalized period of fo pulsators .
the error on the distance modulus comes from four sources .
the _ statistical _ error is the uncertainty in the zero point of the cluster pl relation , and is derived from the least squares fit to the irac data .
systematic _ error is the uncertainty in the calibrated pl zero point , given in the previous section .
the _ calibration _
error is the dispersion in our photometric zero point calibration to standard irac vega magnitudes , described in section [ sec : obs ] .
the _ extinction _
error is the derived from the uncertainty due to differential reddening given in section [ sec : pl ] .
the distance modulus we find is slightly larger than expected , based on previous works ( @xmath66 in @xcite , @xmath67 in @xcite , and @xmath68 in @xcite , this last value derived using rr lyr itself as zero point calibrator ) , but agree within @xmath69 .
if we exclude the systematic error , which will be significantly reduced once high accuracy geometric parallaxes for a large sample of calibrators will become available from gaia , we have an overall uncertainty better than 0.5% in each band .
the average distance modulus of the two wavelengths then is : @xmath70 where the statistical error is reduced by a factor of @xmath71 ( since by using both bands we take advantage of twice the number of available datapoints ) .
note that in this case the overall error is reduced to 1% , small enough to allow the study of higher order parameters in the rr lyr pl relation , first of all the possible dependence on metallicity , once data from more clusters is analyzed .
based on the models presented in @xcite , while the dependence of the pl slope from metallicity is weak , we do expect a larger effect of [ fe / h ] on the pl relation zero point . therefore calibrating the pl relation with galactic rr lyr stars ,
all with different metallicities , may induce a significant error . figure [ fig : met ] addresses this issue , by plotting the residual between the absolute mean magnitude of each calibrator and the value predicted by the pl relation , as a function of their metallicity .
the metallicities used in the plot are listed in table [ tab : gal ] , and are the same listed in @xcite ( see references therein for the original source of the metallicity measurement ) , converted from the original @xcite metallicity scale to the more recent @xcite scale .
for rr lyr and uv oct an extra step was necessary to correct for the different solar abundances used by @xcite when measuring the metallicity of these two stars , with respect to the @xcite .
figure [ fig : met ] also plots the difference between the m4 distance modulus calculated above for m4 , and the value of @xmath68 derived by @xcite from nir pl relations calibrated using rr lyr . due to the large uncertainties in absolute magnitude and metallicity of the galactic rr lyr stars , no trend is apparent ( slope of [email protected] mag / dex ) .
if we adopt the m4 distance modulus value of @xmath73 derived by @xcite from nir data and _ theoretical _ plz relationships for rr lyr stars ( to avoid the large uncertainty in the calibrator distance that is affecting the zero point calibration ) , we obtain the following calibrated rr lyr pl relations for the two irac bands : @xmath63}=-2.332 ( \pm 0.106 ) \log(p ) - 1.054 ( \pm 0.020 ) \\ m_{[4.5]}=-2.336 ( \pm 0.105 ) \log(p ) - 1.091 ( \pm 0.020 ) \end{aligned}\ ] ] an in - depth study of the mir rr lyr pl relations based on theoretical models calculated for the irac bands will be presented in a forthcoming paper .
we have presented new rr lyr mir pl relations for the nearby globular cluster m4 .
accurate irac photometry allows us to reduce the error in the pl slope by a factor of at least two from previous works .
we have also demonstrated that pl relations in the irac bands are consistent with previous calibrations using wise photometry .
we also presented optical and infrared pc relations .
the steeper slope and smaller dispersion of the @xmath48 $ ] relation suggests that it could be preferred over @xmath74 for use in the baade - wesselink method . unlike for cepheid variables ,
the @xmath50 - [ 4.5]$ ] relation is very shallow , indicating that there is no co absorption , and therefore little metallicity effect , in the 4.5 band . as a result
, the 4.5 band can be used as a distance indicator .
we calibrated the zero point of the mir pl relations with five galactic rr lyr stars with known distances .
the uncertainty in their distances is the largest source of error , and we will be able to provide a much more precise calibration using upcoming results from the gaia mission . due to the uncertainty of the available distances to galactic rr lyr stars and the influence of metallicity in the pl relation , we provided a separate pl relation , where the zero point is calibrated using a distance modulus to m4 based on theoretical nir plz relations .
future work will include analysis the remaining gccs in the crrp program in order to provide a comprehensive calibration of the rr lyr pl relation .
this calibration can then be compared to other distance indicators , such as type ii cepheids and delta scuti variables .
we would like to thank the anonymous referee for providing constructive comments to improve the content of this paper .
this work is based [ in part ] on observations made with the spitzer space telescope , which is operated by the jet propulsion laboratory , california institute of technology under a contract with nasa .
this work was partially supported by prin - inaf 2011 tracing the formation and evolution of the galactic halo with vst " ( p.i . :
m. marconi ) and by prin - miur ( 2010ly5n2 t ) chemical and dynamical evolution of the milky way and local group galaxies " ( p.i . : f. matteucci ) .
makovoz , d. , & khan , i. 2005 , in astronomical society of the pacific conference series , vol .
347 , astronomical data analysis software and systems xiv , ed .
p. l. shopbell , m. c. britton , r. ebert ( san francisco : asp ) , 81 cccccl 56446.223 & 11.354 & 0.030 & 56446.227 & 11.267 & 0.079 + 56446.285 & 11.376 & 0.010 & 56446.293 & 11.317 & 0.050 + 56446.336 & 11.394 & 0.003 & 56446.340 & 11.318 & 0.051 + 56446.391 & 11.329 & 0.011 & 56446.398 & 11.261 & 0.074 + 56446.438 & 11.302 & 0.012 & 56446.441 & 11.224 & 0.089 + 56446.492 & 11.312 & 0.021 & 56446.496 & 11.248 & 0.076 + 56446.543 & 11.358 & 0.016 & 56446.547 & 11.285 & 0.069 + 56446.598 & 11.385 & 0.035 & 56446.602 & 11.312 & 0.075 + 56446.652 & 11.344 & 0.021 & 56446.656 & 11.268 & 0.078 + 56446.699 & 11.309 & 0.015 & 56446.703 & 11.240 & 0.063 + 56446.746 & 11.295 & 0.029 & 56446.750 & 11.233 &
0.077 + 56446.797 & 11.340 & 0.023 & 56446.805 & 11.278 & 0.080 + lccccccccl v1 & 16 23 13.67 & 0.28888261 & -0.5392786 & @xmath75 & @xmath76 & @xmath77 & @xmath78 & rrc + & -26 30 53.6 & & & & & & & + v2 & 16 23 16.30 & 0.5356819 & -0.271093 & @xmath79 & @xmath80 & @xmath81 & @xmath82 & rrab * + & -26 34 46.6 & & & & & & & + v3 & 16 23 19.47 & 0.50667787 & -0.2952681 & & @xmath83 & & @xmath84 & rrab + & -26 39 57.0 & & & & & & & + v5 & 16 23 21.03 & 0.62240112 & -0.2059296 & @xmath85 & @xmath86 & @xmath87 & @xmath88 & rrab + & -26 33 05.8 & & & & & & & + v6 & 16 23 25.79 & 0.3205151 & -0.4941515 & @xmath89 & & @xmath90 & & rrc + & -26 26 16.3 & & & & & & & + v7 & 16 23 25.95 & 0.49878722 & -0.3020847 & @xmath91 & @xmath92 & @xmath93 & @xmath94 & rrab + & -26 27 41.9 & & & & & & & + v8 & 16 23 26.16 & 0.50822359 & -0.2939452 & @xmath95 & @xmath96 & @xmath82 & @xmath97 & rrab + & -26 29 41.6 & & & & & & & + v9 & 16 23 26.80 & 0.57189447 & -0.2426841 & @xmath98 & @xmath99 & @xmath100 & @xmath100 & rrab + & -26 29 48.0 & & & & & & & + v10 & 16 23 29.21 & 0.49071753 & -0.3091684 & @xmath101 & @xmath102 & @xmath103 & @xmath104 & rrab + & -26 28 54.3 & & & & & & & + v11 & 16 23 29.98 & 0.49320868 & -0.3069693 & & @xmath105 & & @xmath100 & rrab * + & -26 36 24.5 & & & & & & & + v12 & 16 23 30.82 & 0.4461098 & -0.3505582 & @xmath106 & @xmath107 & @xmath108 & @xmath109 & rrab + & -26 34 58.9 & & & & & & & + v14 & 16 23 31.29 & 0.46353111 & -0.3339211 & @xmath110 & @xmath111 & @xmath112 & @xmath113 & rrab * + & -26 35 34.5 & & & & & & & + v15 & 16 23 31.96 & 0.44366077 & -0.352949 & @xmath114 & & @xmath115 & & rrab * + & -26 24 18.1 & & & & & & & + v16 & 16 23 32.50 & 0.54254824 & -0.2655616 & @xmath116 & @xmath117 & @xmath118 & @xmath119 & rrab + & -26 30 23.0 & & & & & & & + v18 & 16 23 34.70 & 0.47879201 & -0.3198531 & @xmath120 & @xmath121 & @xmath122 & @xmath123 & rrab + & -26 31 04.6 & & & & & & & + v19 & 16 23 35.05 & 0.46781108 & -0.3299295 & @xmath124 & & @xmath125 & & rrab + & -26 25 36.3 & & & & & & & + v20 & 16 23 35.39 & 0.30941948 & -0.5094523 & @xmath126 & @xmath127 & @xmath128 & @xmath129 & rrc + & -26 32 35.9 & & & & & & & + v21 & 16 23 35.93 & 0.47200742 & -0.3260512 & @xmath130 & @xmath131 & @xmath132 & @xmath118 & rrab + & -26 31 33.6 & & & & & & & + v22 & 16 23 36.95 & 0.60306358 & -0.2196369 & @xmath133 & @xmath134 & @xmath135 & @xmath136 & rrab * + & -26 30 13.0 & & & & & & & + v23 & 16 23 37.33 & 0.29861557 & -0.5248876 & @xmath137 & @xmath138 & @xmath139 & @xmath140 & rrc + & -26 31 56.1 & & & & & & & + v24 & 16 23 38.04 & 0.54678333 & -0.2621847 & @xmath141 & @xmath142 & @xmath143 & @xmath144 & rrab * + & -26 30 41.8 & & & & & & & + v25 & 16 23 39.42 & 0.61273479 & -0.2127275 & @xmath145 & @xmath146 & @xmath147 & @xmath82 & rrab + & -26 30 21.3 & & & & & & & + v26 & 16 23 41.65 & 0.54121739 & -0.2666283 & @xmath148 & @xmath149 & @xmath150 & @xmath151 & rrab
+ & -26 32 41.1 & & & & & & & + v27 & 16 23 43.17 & 0.61201829 & -0.2132356 & @xmath152 & & @xmath118 & & rrab + & -26 27 16.3 & & & & & & & + v28 & 16 23 53.60 & 0.52234107 & -0.2820458 & @xmath153 & @xmath154 & @xmath155 & @xmath155 & rrab * + & -26 30 5.30 & & & & & & & + v29 & 16 23 58.25 & 0.52248466 & -0.2819265 & @xmath156 & & @xmath157 & & rrab * + & -26 21 35.1 & & & & & & & + v36 & 16 23 19.45 & 0.54130918 & -0.2665546 & & @xmath158 & & @xmath159 & rrab + & -26 35 49.0 & & & & & & & + v37 & 16 23 31.60 & 0.24734353 & -0.6066994 & @xmath160 & @xmath161 & @xmath162 & @xmath163 & rrc + & -26 31 30.6 & & & & & & & + v38 & 16 23 32.87 & 0.57784635 & -0.2381876 & @xmath164 & @xmath165 & @xmath166 & @xmath167 & rrab * + & -26 33 03.5 & & & & & & & + v39 & 16 23 34.67 & 0.623954 & -0.2048474 & @xmath168 & @xmath169 & @xmath170 & @xmath136 & rrab * + & -26 32 52.1 & & & & & & & + v40 & 16 23 34.59 & 0.38533005 & -0.4141671 & @xmath171 & @xmath172 & @xmath173 & @xmath174 & rrc + & -26 30 44.2 & & & & & & & + v41 & 16 23 39.50 & 0.2517418 & -0.5990447 & @xmath175 & @xmath176 & @xmath177 & @xmath178 & rrc + & -26 33 59.8 & & & & & & & + v42 & 16 24 02.00 & 0.3068549 & -0.5130669 & @xmath179 & & @xmath128 & & rrc + & -26 22 13.0 & & & & & & & + v49 & 16 23 45.25 & 0.22754331 & -0.6429359 & @xmath180 & @xmath181 & @xmath182 & @xmath183 & rrc + & -26 31 28.4 & & & & & & & + v52 & 16 23 24.06 & 0.85549784 & -0.0677811 & @xmath184 & @xmath185 & @xmath186 & @xmath187 & rrab + & -26 30 27.8 & & & & & & & + v61 & 16 23 29.76 & 0.26528645 & -0.5762849 & @xmath188 & @xmath189 & @xmath190 & @xmath191 & rrc + & -26 29 50.3 & & & & & & & + c1 & 16 23 36.43 & 0.2862573 & -0.5432434 & @xmath192 & @xmath193 & @xmath194 & @xmath195 & rrc + & -26 30 42.8 & & & & & & & + lccccccccc & & fo & & & fu & & & fu+fo & + @xmath50 $ ] & 11.207 & -2.658 & 0.079 & 10.841 & -2.370 & 0.040 & 10.929 & -2.332 & 0.056 + & @xmath196 & @xmath197 & & @xmath198 & @xmath199 & & @xmath200 & @xmath201 & + & & & & & & & & & + @xmath202 $ ] & 11.159 & -2.979 & 0.057 & 10.806 & -2.355 & 0.045 & 10.893 & -2.336 & 0.054 + & @xmath203 & @xmath204 & & @xmath200 & @xmath205 & & @xmath200 & @xmath206 & + & & & & & & & & & + lccccc v - j & @xmath207 & @xmath208 & 0.102 + v - k & @xmath209 & @xmath210 & 0.123 + v - [ 3.6 ] & @xmath211 & @xmath212 & 0.117 + v - [ 4.5 ] & @xmath213 & @xmath214 & 0.121 + @xmath50 - [ 4.5]$ ] & @xmath215 & @xmath216 & 0.017 + lccccc period & 0.308645 & 0.466579 & 0.542600263 & 0.566805 & 0.660419 + @xmath217 $ ] & @xmath218 & @xmath219 & @xmath220 & @xmath221 & @xmath222 + @xmath223 & 0.252 & 0.100 & 0.090 & 0.042 & 0.010 + @xmath31}$ ] & 0.051 & 0.020 & 0.018 & 0.009 & 0.002 + @xmath32}$ ] & 0.039 & 0.016 & 0.014 & 0.007 & 0.002 + & & & & & + parallax
( mas ) & @xmath224 & @xmath225 & @xmath226 & @xmath227 & @xmath228 + lkh correction & -0.05 & -0.09 & -0.03 & -0.02 & -0.11 + @xmath229 & @xmath230 & @xmath231 & @xmath232 & @xmath233 & @xmath234 + & & & & & + @xmath50 $ ] & @xmath235 & @xmath236 & @xmath237 & @xmath238 & @xmath239 + @xmath202 $ ] & @xmath240 & @xmath241 & @xmath242 & @xmath243 & @xmath244 + & & & & & + @xmath245 $ ] & @xmath246 & @xmath247 & @xmath248 & @xmath249 & @xmath250 + @xmath251 $ ] & @xmath252 & @xmath253 & @xmath254 & @xmath255 & @xmath256 + | new mid - infrared period - luminosity ( pl ) relations are presented for rr lyr variables in the globular cluster m4 ( ngc 6121 ) .
accurate photometry was obtained for 37 rr lyr variables using observations from the infrared array camera onboard the spitzer space telescope .
the dispersion of m4 s pl relations is 0.056 , and the uncertainty in the slope is 0.11 mag .
additionally , we established calibrated pl relations at 3.6 and 4.5 using published hubble space telescope geometric parallaxes of five galactic rr lyr stars .
the resulting band - averaged distance modulus for m4 is @xmath0 .
the systematic uncertainty will be greatly reduced when parallaxes of more stars become available from the gaia mission .
optical and infrared period - color ( pc ) relations are also presented , and the lack of a mir pc relation suggests that rr lyr stars are not affected by co absorption in the 4.5 band . |
study of light scalar resonances is one of the central problems of nonperturbative qcd , it is important for understanding the chiral symmetry realization way resulting from the confinement physics . in refs .
@xcite we described the high - statistical kloe data on the @xmath8 decay @xcite in the frame of the kaon loop model @xmath9 @xcite simultaneously with the data on the @xmath0 scattering and the @xmath10 reaction .
the description was carried out taking into account the chiral shielding of the @xmath11 meson @xcite and its mixing with the @xmath12 meson , the data yielded evidence in favor of the four - quark nature of the @xmath11 and @xmath12 mesons .
at the same time it was calculated in ref .
@xcite the @xmath0 scattering amplitude in the @xmath2 complex plane , basing on chiral expansion , dispersion relations , and roy equations . in particular , the pole was obtained at @xmath13 , where @xmath14 which was assigned to the @xmath5 resonance . aiming the comparison of the results of refs . @xcite and @xcite , we built up the s - wave @xmath0 scattering amplitude @xmath1 with @xmath15 with correct analytical properties in the complex @xmath2 plane @xcite .
remain that in our model the @xmath16 matrix of the @xmath0 scattering is the product of the `` resonance '' and `` elastic background '' parts : @xmath17 and we introduced the special @xmath18 parametrization to obtain the correct @xmath1 analytical properties ( @xmath19 had correct analytical properties in refs . @xcite already ) . in ref .
@xcite we successfully described the experimental data and the ref .
@xcite results on the real @xmath2 axis using the constructed @xmath0 amplitude , while the @xmath5 pole was located rather far from the ref .
@xcite result .
we assumed that this deviation is caused by approximate character of the roy equations , that take into account only the @xmath0 decay channel .
this question will be discussed below in more details .
in this paper we present the enlarged data analysis .
we dwell on the minimization of the @xmath3 mixing , that leads to the four - quark scenario for light scalars : the @xmath5(600 ) coupling with the @xmath6 channel is suppressed relatively to the coupling with the @xmath0 channel , and the @xmath7(980 ) coupling with the @xmath0 channel is suppressed relatively to the coupling with the @xmath6 channel @xcite .
inelasticity is also crucial for the analysis , here we describe the peculiar behavior of the data up to @xmath20 gev . in refs .
@xcite we used the factor @xmath21 , caused by the elastic @xmath6 background phase , that allows to correct the kaon loop model , suggested in ref .
@xcite , under the @xmath6 threshold .
now we investigate how small this correction may be .
the set of new fits and tables is presented in sec .
[ sdataanalysis ] .
the residues of the @xmath0 scattering amplitude and its resonance part in resonance poles are presented for the first time . as the analytical background @xmath18 is a rather complicated function , in sec .
[ ssimpleback ] we suggest much more simple background parameterization , practically preserving the resonance features , which is comfortable for experimental data analysis , though allows to describe the results of ref .
@xcite only on the real @xmath2 axis .
the conclusion is in sec .
[ sconclusion ] .
note that the @xmath19 parameterization and the complicated background parameterization are the same as in ref .
the modification of the @xmath22 background phase is described in sec .
[ sdataanalysis ] .
the measure of the @xmath3 mixing intensity is the deviation from the ideal picture , when the @xmath0 scattering phase @xmath23 is equal to @xmath24 at the @xmath25 mass @xmath26 , and equal to @xmath27 at the @xmath12 mass @xmath28 .
we require these phases , @xmath29 and @xmath30 , to be close to their `` ideal '' values .
remain that the background phase of the @xmath6 scattering , @xmath31 , changes the modulus of the @xmath32 amplitude under the @xmath6 threshold , at @xmath33 , in the amplitude @xmath34 @xcite . in ref .
@xcite we define @xmath35 in the present paper we investigate the influence of @xmath21 on the @xmath36 amplitude in the @xmath12 region , @xmath37 mev .
we upgrade the parametrization of the @xmath38 , used in refs .
@xcite , now the @xmath39 is parametrized in the following way : @xmath40 @xmath41 note that the @xmath21 also provides pole absence in the analytical continuation of the @xmath36 amplitude under the @xmath0 threshold , see ref .
@xcite .
the experimental data on the inelasticity @xmath42 , see fig .
4 , favor the low value near @xmath43 gev and sharp growth up to @xmath20 gev . below it
is shown that it is possible to reach such a behaviour .
our results for fits 1 - 5 are shown in tables i - vi and figs . 1 - 24 .
fits 1 - 5 show that the allowed range of @xmath25 and @xmath12 parameters is rather wide .
for example , @xmath44 is @xmath45 gev@xmath46 in fit 1 and more than @xmath47 gev@xmath46 in fit 5 .
this result may be important for coordination of the @xmath44 and @xmath48 @xcite .
note that in fit 4 the @xmath25 and @xmath12 are coupled only with the @xmath0 channel and the @xmath6 channel ( @xmath49 ) .
as seen from table i and figs .
9 - 16 , fit 4 is in excellent agreement with the data and the @xcite results .
we introduce 56 parameters , but for restrictions ( expresses 5 parameters through others ) and parameters ( or their combinations ) , that go to the bound of the permitted range ( 9 effective links ) , the effective number of free parameters is reduced to 42 .
but it is significant that fits describe as the experimental data ( 65 points ) , as well as the @xmath0 amplitude from the @xcite in the range @xmath50 gev@xmath46 which is treated along with experimental data . as in @xcite
we show resonance poles of the @xmath1 on some unphysical sheets of its riemannian surface , depending on sheets of the polarization operators @xmath51 . for this choice of sheets
the imaginary parts of pole positions @xmath52 would be connected to the full widths of the resonances ( @xmath53 ) in case of metastable states , decaying to several channels .
note also that we do not show poles for fit 2 .
one can see that the obtained @xmath11 pole positions lie far from eq .
( [ polesigma ] ) . in ref .
@xcite we noted , that one of the possible reasons is the approximate character of the roy equations , that take into account only the @xmath0 channel .
table i. properties of the resonances and main characteristics are shown .
the resonance masses @xmath54 and widths @xmath55 ( which may be called `` breit - wigner '' masses and widths ) are parameters in the resonance propagators , see ref .
they have clear physical meaning in contrast to the resonance poles in the complex plane . [
cols="^,^,^,^,^,^",options="header " , ]
the background function , suggested in ref .
@xcite to reach the correct analytical properties of the @xmath0 scattering amplitude and used above , is rather complicated and costly in computation . in this section
we suggest much more simple background parameterization , practically preserving the resonance features , which is comfortable for experimental data analysis and allows to describe the results @xcite on the real @xmath2 axis .
this background function is an upgrade of the one , used in ref .
@xcite : @xmath56 here @xmath57 .
note that in comparison with ref .
@xcite the function ( [ phb2 ] ) has a left cut .
let us build the @xmath58 function .
it may be devided into 3 parts : @xmath59 where the first one is the usual @xmath58 function for the experimental data , the second one provides the description of the results @xcite , and the third one provides the restrictions . the @xmath60 is constructed with the help of the same data , as in ref .
@xcite , except the @xmath23 data in the region @xmath61 mev , where we use the @xcite results .
note that in table i we show @xmath62 , obtained in the full region @xmath63 mev with the `` old data '' @xcite .
the @xmath64 caused by the real and imaginary parts of the @xmath65 contributions in the region @xmath66 , and the @xmath67 contribution from the region @xmath68 mev@xmath69 .
here for @xmath70 and @xmath67 we used points and errors , kindly sent us by h. leutwyler , and for @xmath71 the approximate curve @xmath72 , obtained using fig . 1 in ref .
@xcite , providing us with central values , and the error is assumed to be @xmath73 .
note that for @xmath71 we used the `` reper '' points @xmath74 mev@xmath69 , @xmath75 mev@xmath69 , @xmath76 mev@xmath69 , @xmath77 mev@xmath69 , @xmath78 mev@xmath69 , @xmath79 mev@xmath69 , @xmath80 mev@xmath69 , @xmath81 mev@xmath69 , the last is the end of the domain of validity of the roy equations , connected with the lehmann - martin ellipse , see @xcite .
we impose the following set of restrictions , contributing to @xmath82 : \1 ) @xmath83 and @xmath84 to provide small @xmath3 mixing , a kind of diagonalization that results in the four - quark model scenario .
\2 ) @xmath85 for @xmath37 mev .
the maximum is found dynamically ( at every calculation of the @xmath58 function ) , the minimum in our situation is at @xmath86 mev .
\3 ) @xmath87 , trying to be not far from the result @xcite .
\4 ) @xmath88 ; @xmath89 gev @xmath90 gev ; @xmath91 gev @xmath92 gev ; @xmath93 mev @xmath94 to provide reasonable form of the @xmath95 . to provide , for example , the condition @xmath96 , we add to @xmath82 the term @xmath97 where @xmath98 is the big number .
so till @xmath96 the contribution @xmath99 is equal to @xmath100 , but when @xmath101 , @xmath99 becomes large , so the minimization procedure can go outside the barrier only on a negligible distance .
our @xmath82 is the sum of contributions like eq .
( [ setbarrier ] ) . using the constructed @xmath58 function ,
we obtain fit 6 .
one can see that this fit perfectly describes the experimental data and the results based on roy equations on the real @xmath2 axis , see table vii and figs .
note that in table vii the @xmath28 and @xmath44 errors are adduced . to illustrate the abilities of the background ( [ phb2 ] ) , we perform fit 7 with the same resonance parameters as for fit 3 .
fit 7 provides practically the same experimental data description as fit 3 .
the theoretical curves for phase @xmath23 are shown in fig .
25 ( a ) , they are practically the same .
it is obvious that both fit 3 and fit 7 provide practically identical mass spectrum in @xmath8 decay also .
the inelasticity is exactly the same .
additionally , fit 7 and fit 3 provide indistinguishable curves for @xmath1 at @xmath102 , see fig . 25 ( b ) .
our investigation shows that the scenario , based on the four - quark model , completely agrees with the current experimental data and theoretical requirements .
it is shown that the requirement of the weak @xmath103 mixing leads to the @xmath104 and @xmath105 suppression , that is predicted by the four - quark model , see table i. the behaviour of the factor @xmath106 , which corrects the kaon loop model , is model dependent .
we show that for large enough @xmath107 constant ( f.e .
, @xmath108 gev@xmath46 ) the current data ( including the ref .
@xcite results ) may be well - described with this factor close to @xmath45 at 850 mev @xmath109 , but for smaller values of this constant ( f.e .
, @xmath45 gev@xmath46 ) the correction increases . new precise data on the @xmath110 reaction and
the inelasticity ( @xmath42 ) of the @xmath0 scattering would give an ability to understand more about this factor and reduce the region of possible values of parameters .
the obtaining of the @xmath5 pole in ref .
@xcite gave the strong argument in favor of the @xmath25 existence , but followed efforts aiming the precise determination of the pole are not productive .
the roy equations are one - channel , that is , are approximate and even slight discrepancy in the amplitude in the physical region may lead to large changes in the complex plane . remind that the riemannian surface of the @xmath0 scattering amplitude has many sheets ( strictly speaking , infinite number of sheets ) , and even for relatively narrow @xmath12 is sometimes a problem to determine on what sheet should we find the pole , see , for example , fit 2 in ref . @xcite . but
even if we obtained the pole precisely , it would give us practically no information on the resonance nature , because it can not be connected to coupling constant in the hermitian ( or quasi - hermitian ) hamiltonian , see also ref .
@xcite , because of large imaginary part .
besides , the residue of the amplitude in the pole is strongly distorted by the background part of the amplitude , see tables v and vi , that gives essential contribution even for relatively narrow @xmath12 .
let us dwell on the results , presented in table vi .
remind that for a stable particle with the mass @xmath111 there is the pole in the amplitude @xmath112 at @xmath113 , the residue of the amplitude @xmath114 is connected to the coupling constant ( @xmath115 ) of the stable particle with the @xmath0 channel .
one can see that the real part of the @xmath1 residue in the @xmath12 pole is positive , so the coupling constant should be practically pure imaginary , what is physically meaningless .
note that the residue of the amplitude resonance part @xmath116 is good .
that is why the best way for understanding the nature of the light scalars is the investigation of their production mechanisms in physical processes .
the simple background parameterization , suggested in sec . [ ssimpleback ] , may be used for experimental data analysis and the description of the ref .
@xcite results for real @xmath2 .
it is shown that the resonance features are practically preserved , moreover , one can see that for even more simple background , used in ref .
@xcite , they changed not so much , though the ref .
@xcite results were not included . in this investigation
we paid more attention to the inelasticity @xmath42 , namely , we tried to reproduce the peculiar behaviour near the threshold , indicated by the experimental data . unfortunately , the current data have large errors , so the precise measurement of the inelasticity @xmath42 near 1 gev in @xmath117 would be very important .
we thank very much h. leutwyler for providing numerical values of the @xmath118 on the real axis , obtained in ref .
@xcite , useful discussions , and kind contacts .
this work was supported in part by rfbr , grant no 10 - 02 - 00016 . in the kaon loop model , @xmath119 @xcite ,
the amplitude of the signal @xmath120 is @xmath121 where @xmath122 is the kaon loop function , @xmath123 and @xmath124 are polarization vectors of the @xmath123 meson and photon , see ref .
@xcite .
b. hyams et al . ,
nucl . phys . *
b64 * , 134 ( 1973 ) .
p. estabrooks and a.d .
martin , nucl . phys . *
b79 * , 301 ( 1974 ) .
martin , e.n .
ozmutlu , e.j .
squires , nucl .
b121 * , 514 ( 1977 ) . v. srinivasan et al .
, phys . rev .
* d12 * , 681 ( 1975 ) .
l. rosselet et al . , phys . rev .
* d15 * , 574 ( 1977 ) . | in the paper phys . rev . * d83 * , 054008 ( 2011 ) we constructed the @xmath0 scattering amplitude @xmath1 with regular analytical properties in the @xmath2 complex plane , describing both experimental data and the results based on chiral expansion and roy equations .
now the results obtained during development of our work are presented .
we dwell on questions dealing with the low @xmath3 mixing , inelasticity description and the kaon loop model for @xmath4 reaction , and show a number of new fits .
in particular , we show that the minimization of the @xmath3 mixing results in the four - quark scenario for light scalars : the @xmath5(600 ) coupling with the @xmath6 channel is suppressed relatively to the coupling with the @xmath0 channel , and the @xmath7(980 ) coupling with the @xmath0 channel is suppressed relatively to the coupling with the @xmath6 channel .
the correct analytical properties of the @xmath0 scattering amplitude are reached with the help of rather complicated background function .
we also suggest much more simple background parameterization , practically preserving the resonance features , which is comfortable for experimental data analysis , but allows to describe the results based on chiral expansion and roy equations only on the real @xmath2 axis . |
the fast x - ray transient source xte j1901 + 014 was discovered [ 4 ] by the all - sky monitor asm on board the rxte observatory during the powerful outburst on april 6 , 2002 lasted from 3 min to 3.15 hours and reached the peak flux @xmath10.9 crab in the 1.5 - 12 kev energy band ( fig.1 , right panel ) .
the source position was determined as ra = 19@xmath2 01@xmath3 [email protected] , dec = + 1 2415.7(j2000 ; 3uncertainty ) .
the analysis of the archival asm data [ 5 ] revealed a previous outburst from the same position on june 21 , 1997 .
this outburst was longer than 6 min and shorter than 8 hr , with a peak flux of @xmath10.4 crab ( fig . 1 , left panel ) .
the obtained information about xte j1901 + 014 was not enough to make any confident conclusions about its nature , but it was noted that the time scale of this flare is similar to those of such events observed from the black hole binary v4641 sgr . in this report
we briefly present results of observations of xtej1901 + 014 with the integral and rxte observatories .
more detail analysis will be presented separately ( see [ 2 ] ) .
during the outburst in july 1997 the source flux in the 1.5 - 3 kev energy band did not exceed the background level whereas in the harder energy bands , 3 - 5 kev and 5 - 12 kev , it reached @xmath10.13 crab and @xmath10.7 crab , respectively . during the outburst in april 2002
the peak fluxes in these three bands were detected at the levels of @xmath10.8 , @xmath11.1 and @xmath11.2 crab , respectively .
thus both observed outbursts were hard .
we analysed rxte / asm archive data from junuary , 1996 to july , 2006 and could not find other such powerful outbursts from the source .
xtej1901 + 014 was detected in the quiescent state ( outside of outbursts ) by both the spectrometer rxte / pca in september , 1998 and april , 2002 , with the full exposure @xmath11650 s and an average 3 - 100 kev flux of @xmath12.8 mcrab ( is was the same in different years ) and the detector integral / isgri in 2003 - 2004 see above with an average flux of @xmath12.7 mcrab in the 17 - 100 kev energy band .
some aperiodic variability of the source flux was detected in all rxte observations .
we found a number of small flares with a duration of 40 - 60 s and a maximal flux of @xmath16 - 8 mcrab .
the origin of such variability is most likely connected with a nonsteady accretion .
analysis of the rosat all - sky survey source catalogue , has shown that the source 1rxs j190141.0 + 012618 is located in the rxte / asm error box ( 3 ) of xte j1901 + 014 . during the pointed rosat / hri observation performed on october 3 , 1994 , the source was also detected , its position was refined and the source was renamed as 1rxh j190140.1 + 012630 [ 7 ] . using the integral / isgri data we improved an accuracy of the xte j1901 + 014 localization to @xmath11.2 . as it clearly seen from fig.2
the rosat source 1rxh j190140.1 + 012630 confidently the distance between positions of xte j1901 + 014 and 1rxh j190140.1 + 012630 is about 0.3 ) falls into the integral / isgri error box for xtej1901 + 014 , that points that xte j1901 + 014 and 1rxh j190140.1 +
012630 are the same source .
we have only very poor information of the source spectral evolution during the outbursts ( see below ) , but can precisely reproduce its spectrum in the quiescent state . to obtain the broadband spectrum of the source in the quiescent state we used rxte / pca data in the 3 - 20 kev energy band and integral / isgri data in the hard energy band ( @xmath520 kev ) analysis .
it is important to note , that the pca / rxte observations were performed in 1998 , 2002 and the isgri / integral ones - in 2003 - 2004 .
thus our spectral reconstruction is correct in the suggestion that the spectrum shape of the source does not change during this time interval .
the broadband ( 3 - 100 kev ) spectrum of xtej1901 + 014 was approximated by a simple power law model with the interstellar absorption which value was fixed at n@xmath6 = @xmath7 atom/@xmath8 that is typical for this direction to the sky ( it was evaluated from the n@xmath6 map ) . the best - fit photon index is @xmath9=2.15 @xmath10 0.03 ( fig .
3 ) . we analysed small short flares registered by rxte / pca from the source ( see above ) and found that the source spectral shape did not changed during the flares .
xtej1901 + 014 is located near the galactic plane ( l = 35.38 deg , b = -1.62 deg ) , thus the galactic ridge emission could strongly affect the result of spectral measurements with rxte / pca [ 3 ] . in this report the spectrum and lightcurves of xtej1901 + 014 were obtained taking into account this contamination . in order to estimate the galactic ridge emission intensity we used the data obtaned during pointing observations of nearby transient sources performed during their `` turned off '' state .
in particular we used pointing data of gs 1843 - 02 ( l @xmath1131 deg , b @xmath11 - 0.5 deg ) observations , that is the nearest transient x - ray pulsar to obtain the galactic ridge spectrum at its position for xtej1901 + 014 .
the analysis of this data allows us to obtain the galactic ridge spectrum near gs 1843 - 02 . due to the nature of the galactic ridge emission
its spectrum has the same form in different regions of the sky with -5 deg @xmath12 b @xmath12 + 5 deg [ 3 ] .
therefore we can just renormalize this spectrum ( using the scan data ) , to get the galactic ridge spectrum at xtej1901 + 014 position .
the importance of accounting the galactic ridge emission is demonstrated by fig.4 , where the total pca / rxte spectrum is shown along with the galactic ridge and source true spectra .
however using two energy bands of rxte / asm ( 3 - 5 and 5 - 12 kev ) it is possible to roughly estimate evolution of the photon index during the outbursts . according to [ 6 ]
the photon index @xmath13 can be expressed as : @xmath14 where r - the relation between count rates in the 3 - 5 kev and 5 - 12 kev energy bands .
note that this equation was obtained for sources with powerlaw spectra . for the outburst in june , 1997
, we found that @xmath13 was not changed significantly ; for the outburst in april , 2002 - it changed from 2.4 + /-
0.1 to 1.4 + /-
* the broadband spectrum of xtej1901 + 014 in the quiescent state was obtained and investigated for the first time .
it can be approximated by a simple power law model with the photon index of @xmath12.15 without any high energy exponential cutoff . *
the powerful outbursts of the source in 1997 and 2002 are not the 1st type x - ray bursts because they become harder with time , resembling to the well - known outburst of v4641sgr or outbursts from saxj1818.6 - 1703 [ 1 ] . *
a number of small short flares were detected from the source during pointed rxte / pca observations . *
the accuracy of the xtej1901 + 014 localization was improved from 3to @xmath11.2that strengthens the association of the source with the rosat soft source 1rxh j190140.1 + 012630 .
summarizing all the above we can suppose that xtej1901 + 014 belongs to the class of fast x - ray transient sources , but for the final answer on its origin , more observations at different wavelengths are necessary .
the authors thank to e.churazov for the developing of the methods of the analysis of the ibis data and software .
authors also thank m.revnivtsev for the help with the analysis of rxte data and a discussion of the results obtained .
this work was supported by the russian foundation for basic research ( projects no.05 - 02 - 17454 and 02 - 04 - 17276 ) , the russian academy of sciences ( the origins and evolution of stars and galaxies program ) and grant of president of rf ( nsh- 1100.2006.2 ) .
al acknowledges financial support from the russian science support foundation . | we present results of spectral and timing analysis of the fast x - ray transient xte j1901 + 014 based on data of the rxte and integral observatories . with the integral / isgri
the source was detected at a significance level of 20@xmath0 with the persistent flux of @xmath12.7 mcrab in a 17 - 100 kev energy band in 2003 - 2004 ( during long observations of the sagittarius arm region ) .
we added the rxte / pca ( 3 - 20 kev ) data obtained in 1998 to the integral / isgri data to build the broadband spectrum of the source in a quiescent state .
it was found that the spectrum can be well approximated by a simple powerlaw with a photon index of @xmath12.15 . from timing analysis
we found short time scale aperiodic variations which can be connected with instabilities in the accretion flow .
[ 2001/04/25 1.1 ( pwd ) ] |
quantum tunneling is one of the most important phenomenon with wide range of applications in modern technology . in spite of its remarkable success in various applications , there still remains the fundamental question , namely , how much time does a particle take to traverse the barrier ?
( so called tunneling time problem ) .
this problem has been approached from many different viewpoints , but still there is a lack of consensus about the existence of simple expression for this time @xcite .
this is due to the fact that there is no hermitian operator associated with this time .
the various time scales proposed in the literature are based on different operational definitions and physical interpretations .
they may represent different complimentary aspect of tunneling process .
some of them include dwell time , larmor time , complex times , buttiker- landauer times , sojourn times @xcite , time scale based on bohm s view @xcite , phase delay time , etc . in our present work
we concentrate mainly on the concept of wigner phase delay time .
this time is usually taken as the difference between the time at which the peak of the transmitted packet leaves the barrier and the time at which the peak of the incident gaussian wave packet arrives at the barrier . within the stationary phase approximation the phase time can be calculated from the energy derivative of the ` phase shift ' in the transmitted or reflected amplitudes .
this phase delay time is also related to various physical quantities such as partial density of states @xcite . in the case of a quantum tunneling
it has been shown that in the opaque barrier limit the phase delay time does not depend on the barrier width .
this phenomenon is called as the hartman effect .
this implies that for sufficiently large barriers the effective velocity of the particle can become arbitrarily large , even larger than the speed of light in the vacuum ( superluminal ) .
this may be further regarded as the evidence of the fact that quantum systems seem to behave as non - local .
experiments have clearly demonstrated @xcite that tunneling photons " travel with superluminal group velocities .
their measured tunneling time is practically obtained by comparing the two peaks of the incident and transmitted wave packets .
it is important to note that these observations do not violate ` einstein causality ' , i.e. , the signal velocity or the information transfer velocity is always bounded by the velocity of light .
theoretically hartman effect has been generalized to different cases including double barriers , various geometric structures and in the presence of aharonov - bohm flux .
mystery surrounding this effect has also been addressed recently by winful .
he argues that the short time delay observed is due to energy storage and release and has nothing to do with propagation .
since the stored energy or the probability density in the evanescent field decreases exponentially with the distance in the barrier , after certain decay distance it does not matter how much more length the barrier has .
for details we refer to @xcite . in our present work
we study the above mentioned effect in a simple quantum system consisting of a circular loop connected to a single wire of semi - infinite length as shown in fig [ system ] .
in addition we impose aharonov - bohm flux through the loop .
we also focus on the following situation .
there is a potential barrier ( or barriers ) inside the loop , while the potential in the connecting lead is set to zero .
the incident energy of the free propagating electron @xmath0 in the semi - infinite wire is less than the barrier potential height @xmath1 .
the impinging electron on the loop in the sub barrier region travels as an evanescent wave in the loop before being reflected .
we analyze the phase time of the reflected wave .
we show that this phase time in the opaque barrier regime becomes independent of the length of the circumference of the ring and the magnitude of the aharonov - bohm flux , thereby generalizing the hartman effect beyond one dimension and in the presence of magnetic flux .
we have also extended this effect by including an additional potential well in the circular ring .
interestingly the saturated delay time becomes independent of the length of the potential well ( in the large length limit ) in the off resonant case .
the potential well can support many resonant states .
this result is regarded as a space collapse or space destroyer " @xcite . even though in the potential well inside the loop electronic wave can travel as a free propagating mode ( and not as a evanescent mode ) , surprisingly the saturated delay time is independent of the length of the well ( as if it does not count ) . in the sections to follow
we first give the description of the theoretical treatment .
for this we make use of the quantum waveguide approach well known for the electronic transport problems .
the classical relativistic helmholtz equation for the electromagnetic wave propagation is formally identical to non relativistic stationary schrdinger equation .
hence conclusions obtained using quantum tunneling theory are equally valid in the case of electromagnetic phenomenon ( photon tunneling ) . in the later sections
we will discuss our results and conclusions .
we approach this scattering problem using the quantum wave guide theory @xcite . in the stationary case the incoming particles
are represented by a plane wave @xmath2 of unit amplitude .
the effective mass of the propagating particle is @xmath3 and the energy is @xmath4 where @xmath5 is the wave vector corresponding to the free particle .
the wave function in different regions ( which are solutions of the schrdinger equation ) in the absence of magnetic flux are given below @xmath6 with @xmath5 being the wavevector of electrons in the lead and in the intermediate free space between two barriers inside the ring .
@xmath7 and @xmath8 are the wavevector respectively for propagating electrons in the barriers of strength @xmath9 and @xmath10 inside the ring .
the origin of the co - ordinates of @xmath11 and @xmath12 is assumed to be at @xmath13 and that for @xmath14 and @xmath15 are at @xmath16 and @xmath17 respectively . at @xmath16 , @xmath18 , at @xmath17 , @xmath19 and at @xmath13 , @xmath20 , where @xmath21 and @xmath22 are the length of the two barriers separated by a well region of length @xmath23 inside the ring
. total circumference of the ring is @xmath24 .
we use griffith s boundary conditions @xcite @xmath25 and @xmath26 at the junction @xmath13 .
all the derivatives are taken either outward or inward from the junction @xcite .
similar boundary conditions are held at @xmath16 and @xmath17 _ i.e. _ @xmath27 we choose a gauge for the vector potential in which the magnetic field appears only in the boundary conditions rather than explicitly in the hamiltonian @xcite .
thus the electrons propagating clockwise and anticlockwise will pick up opposite phases .
the electrons propagating in the clockwise direction from @xmath13 will pick up phases @xmath28 at @xmath16 , @xmath29 at @xmath17 and @xmath30 at @xmath13 after traversing once along the loop .
the total phase around the loop is @xmath31 , where @xmath32 and @xmath33 are the magnetic flux and flux quantum , respectively .
hence from above mentioned boundary conditions we get for tunneling particle @xmath34+b_1\,\exp(\kappa_1\,lb_1)\hskip 2 cm \nn\\ -\,a_2-b_2\,\exp(-i\,\a_2)=0\ , , \\ \kappa_1\,a_1\,\exp(-\kappa_1\,lb_1)\,\exp(i\,\a_1)-\kappa_1\,b_1\,\exp(\kappa_1\,lb_1)\hskip 1 cm \nn\\ + \,i\,k\,a_2-i\,k\,b_2\,\exp(-i\a_2)=0\ , , \\ a_2\,\exp(i\,k\,w)\,\exp(i\,\a_2)\,+\,b_2\,\exp(-i\,k\,w)\hskip 2 cm \nn\\ -\,a_3-b_3\,\exp(-i\a_3)=0\ , , \\
i\,k\,a_2\,\exp(i\,k\,w)-i\,k\,b_2\,\exp(-i\,k\,w)\,\exp(-i\a_2)\,\hskip 1 cm \nn\\ + \,\kappa_3\,a3-\kappa_3\,b_3\,\exp(-i\,\a_3)=0\ , , \end{aligned}\ ] ] with @xmath35 and @xmath36 being the imaginary wave vectors , in presence of rectangular barriers of strength @xmath9 and @xmath10 respectively , inside the ring .
once @xmath37 is known , the ` reflection phase time ' @xmath38 can be calculated from the energy derivative of the phase of the reflection amplitude @xcite as @xmath39}{\partial e}\ , , \label{phtm}\ ] ] where , @xmath40 is the velocity of the free particle .
the concept of ` phase time ' was first introduced by wigner @xcite to estimate how long a quantum mechanical wave packet is delayed by the scattering obstacle . for a ring system with a rectangular barrier of strength @xmath9 along its entire circumference
we obtain an analytical expression for the reflection amplitude as @xmath41 where @xmath42 . in what follows ,
let us set @xmath43 and @xmath44 .
we now proceed to analyze the behavior of @xmath38 as a function of various physical parameters for different ring systems .
we express all the physical quantities in dimensionless units _
i.e. _ all the barrier strengths @xmath45 in units of incident energy @xmath0 ( @xmath46 ) , all the barrier widths @xmath47 in units of inverse wave vector @xmath48 ( @xmath49 ) , where @xmath50 and the reflection phase time @xmath38 in units of inverse of incident energy @xmath0 ( @xmath51 ) .
after straight forward algebra in the large length ( @xmath52 ) limit and in absence of magnetic flux , we obtain an analytical expression for the saturated phase delay time ( using eq .
( [ phtm ] ) and eq .
( [ ref ] ) ) , which is given by , @xmath53 here the rectangular barrier has strength @xmath9 and width @xmath54 .
= 0 ) , for a ring with a barrier of strength @xmath9 throughout its circumference , the phase time @xmath38 is plotted as a function of ring s circumference @xmath52 . the solid , dotted , dashed and dash - dotted curves are for @xmath55 respectively .
incident energy is set to be @xmath56 . in the inset
the saturated value of phase time @xmath57 is plotted as a function of the barrier s strength for same @xmath0 .
, width=302 ] first we take up a ring system with a single barrier along the circumference of the ring . for a tunneling particle having energy less than the barrier s strength
we find out the reflection phase time @xmath38 as a function of barrier s width @xmath52 which in turn is the circumference of the ring .
we see ( fig .
[ fig : tau - l ] ) that in absence of magnetic flux , @xmath38 evolves as a function of @xmath52 and asymptotically saturates to a value @xmath57 which is independent of @xmath52 thus confirming the ` hartman effect ' . from fig .
[ fig : tau - l ] it is clear that the saturation value increases with the decreasing barrier - strength . in the inset of fig .
[ fig : tau - l ] , we have plotted @xmath57 as a function of barrier - strength . from this we can see that for electrons with incident energy close to the barrier - strength the value of @xmath57 is quite large
. lies throughout its circumference , the saturated phase time @xmath57 is plotted as a function of magnetic flux @xmath32 .
the solid , dashed and dotted curves are for @xmath58 respectively . other system parameters are @xmath59 , @xmath56.,width=302 ] to see the effect of magnetic flux on ` hartman effect ' , we consider the same system but in presence of aharonov - bohm ( ab ) flux .
we find out , for the tunneling particle , the reflection phase time as a function of embedded magnetic flux for different lengths @xmath52 of the barrier covering the ring s circumference .
we have chosen the lengths such that in absence of the ` ab - flux ' , for a given system ( _ i.e. _ for known @xmath0 and @xmath1 ) the reflection phase time @xmath38 gets saturated in these lengths . from fig . [ fig : tau - phi ] we see that @xmath38 as a function of @xmath32 shows ab - oscillations with an average value which is the saturation value @xmath57 for the same system in absence of ab - flux . further observe that ( fig .
[ fig : tau - phi ] ) @xmath38 is flux periodic with periodicity @xmath33 .
this is consistent with the fact that all the physical properties in presence of ab - flux across the ring must be periodic function of the flux with a period @xmath33 @xcite .
however , we see that as we increase @xmath52 the magnitude of ab - oscillation in @xmath38 decreases . consequently in the large length limit the visibility vanishes .
this clearly establishes ` hartman effect ' even in presence of ab - flux .
the constant value of @xmath38 thus obtained in the presence of flux is identical to @xmath57 in the absence of flux ( see fig .
( [ fig : tau - phi ] ) ) in the large length regime and its magnitude is given by eq .
( [ satphtm ] ) .
this result clearly indicates that the reflection phase time in the presence of opaque barrier becomes not only independent of length of the circumference but also is independent of the ab - flux thereby observing the ` hartman effect ' in the presence of ab - flux .
= 0 ) , for a ring with two barriers of strength @xmath9 and @xmath10 separated by an intermediate well region , the phase time @xmath38 is plotted as a function of ring s circumference @xmath52 for different width @xmath23 of the well . the dashed - dotted , solid , dotted and dashed curves are for @xmath60 . other system parameters are @xmath61 , @xmath62 and @xmath56.,width=302 ] now we consider the ring system with the ring having two successive barriers separated by an intermediate free space as shown in fig .
[ system ] . in absence of magnetic flux , we see the effect of ` quantum well ' on the reflection phase time @xmath38 . in fig .
[ fig : tau - w ] @xmath38 is plotted as a function of one of the barrier s length ( say @xmath21 ) while other barrier s length is fixed ( @xmath61 ) and for few different values of length of the well . here , the fixed value of the barrier s length @xmath22 is chosen in such a way that in absence of the well region the reflection phase time reaches saturation at this length . from fig .
[ fig : tau - w ] we see that for all parameter values of well s width , the saturation value of reflection phase time @xmath57 is same and it is equal to what we obtained in absence of the well in the ring system .
thus the saturated phase time becomes independent of the width of the well ( in the long length limit ) in the off resonant case .
this is as if the effective velocity of the electron in the well becomes infinite or equivalently length of the well does not count ( space collapse or space destroyer ) while traversing the ring . finally consider a similar system as that shown in fig .
[ system ] .
here we see the effect of resonances , present in the ring system with a well , on the saturated reflection phase time @xmath57 . for the system described above with @xmath63 we have plotted @xmath57 , for the electrons with incident energy @xmath56 , as a function of the well s width for different parameter values @xmath21 in fig .
[ fig : resonance ] .
we see that the resonances which have lorentzian shape become sharper and narrower as the width of the barrier @xmath21 becomes large . for very large
@xmath21 the resonances are very hard to detect . it should be noted that as we increase the length of the well for fixed @xmath0 for particular barrier lengths incident energy @xmath0 coincides with resonances ( or resonant states ) in the well ( which arise due to constructive interference due to multiple scatterings inside the well ) . for these values of lengths
we observe sharp rise in the saturated delay time and its magnitude depends on the length of the well .
it is worth to mention that away form the resonance the value of @xmath57 is independent of the length of the well ( see fig.[fig : tau - w ] ) and depends only on the barrier strength .
= 0 ) , for a ring with two barriers of strength @xmath9 and @xmath10 separated by an intermediate well region , the saturated phase time @xmath57 is plotted as a function of ring s circumference @xmath52 for different width @xmath21 of the barrier .
the solid and dashed curves are for @xmath64 respectively .
other systems parameters are @xmath65 , @xmath62 and @xmath56.,width=302 ]
we have generalized the hartman effect beyond one dimension and in the presence of magnetic flux .
this is done by studying the phase delay time in a circular geometric ring in the reflection mode .
we have obtained an analytical expression for the saturated delay time .
we have also shown that in the presence of successive opaque barriers separated by the potential well the saturated delay time becomes independent of the width of the well ( in the long length limit ) in the off resonant case .
this is as if the effective velocity of the electron between the well becomes infinite or equivalently length of the well does not count ( space collapse or space destroyer ) .
we have further extended this effect in quantum mechanical networks with many arms wherein we have shown additional existence of new quantum non - local effects associated with the phase delay time @xcite .
our reported results are amenable to experimental verifications for photon tunneling in appropriate electromagnetic geometries .
one of the authors ( sb ) thanks debasish chaudhuri and prof .
binayak dutta - roy for several useful discussions .
99 r. landauer and t. martin , rev .
* 66 * ( 1994 ) 217.e .
h. hauge and j. a. stvneng , rev .
mod . phys .
* 61 * ( 1989 ) 917 .
v. s. olkhovsky and e. recami , physics reports * 214 * ( 1992 ) 339.s .
anantha ramakrishna and n. kumar , europhys .
lett . , * 60 * ( 2002 ) 491 ; colin benjamin and a. m. jayannavar , solid state commun . ,
* 121 * ( 2002 ) 591 .
a. m jayannavar , pramana - journal of physics , * 29 * ( 1987 ) l341 .
m. bttiker , pramana - journal of physics , * 58 * ( 2002 ) 241 .
a. m. steinberg , p. g. kwiat , and r. y. chiao , phys .
lett . * 71 * 708 1993 ; r. y. chiao , p. g. kwiat , and a. m. steinberg , scientific american , p. 38
- 46 , august 1993 .
a. enders and g. nimtz , j. phys .
i * 2 * ( 1992 ) 1693 ; * 3 * ( 1993 ) 1089 ; a. enders and g. nimtz , phys . rev .
e , * 48 * , 632 ( 1993 ) ; g. nimtz , a. enders and h. spieker , _ ibid _ * 4 * ( 1994 ) 565 .
p. guerent , e. marclay , and h. meier , solid state commun .
* 68 * , 977 ( 1988 ) ; ch .
spielmann , r. szipocs , a. sting , and f. krausz , phys .
lett . * 73 * , 2308 ( 1994 ) ; th .
hills et al .
a * 58 * , 4784 ( 1998 ) .
h. g. winful , phys .
* 91 * ( 2003 ) 260401 ; h. g. winful , phys
* e68 * ( 2003 ) 016615 ; h. g. winful , opt .
express * 10 * ( 2002 ) 1491 .
v. s. olkhovsky , e. recami and g. salesi , europhys . lett .
* 57 * ( 2002 ) 879 .
j. b. xia , phys . rev .
* b45 * ( 1992 ) 3593 . p. s. deo and a. m. jayannavar , phys
* b50 * ( 1994 ) 11629 .
s. griffith , trans .
faraday soc .
* 49 * , 345 ( 1953 ) 650.y .
gefen , y. imry and m. ya azbel , phys .
lett . * 52 * ( 1984 ) 129.e .
p. wigner , phys .
* 98 * ( 1955 ) 145 .
s. washburn and r. a. webb , adv .
* 35 * ( 1986 ) 375 .
s. datta , electronic transport in mesoscopic systems , cambridge university press , cambridge , uk , 1997 .
swarnali bandopadhyay , raishma krishnan , and a. m. jayannavar , solid state commun .
* 131 * , 447 ( 2004 ) ; swarnali bandopadhyay and a. m. jayannavar , hartman effect and non - locality in quantum networks , quant - ph/0410126 . | in this work we study the behaviour of wigner phase delay time for tunneling in the reflection mode .
our system consists of a circular loop connected to a single wire of semi - infinite length in the presence of aharonov - bohm flux .
we calculate the analytical expression for the saturated delay time .
this saturated delay time is independent of aharonov- bohm flux and the width of the opaque barrier thereby generalizing the hartman effect .
this effect implies superluminal group velocities as a consequence .
we also briefly discuss the concept called space collapse or space destroyer "
. 0.5 cm |
in a geometrical setting , symmetries are connected with isometries associated with killing vectors and , more generally , killing tensors on the configurations space of the system .
an example is the motion of a point particle in a space with isometries @xcite , which is a physicist s way of studying the geodesic structure of a manifold . in @xcite
such studies were extended to spinning space - times described by supersymmetric extensions of the geodesic motion , and in@xcite it was shown that this can give rise to interesting new types of supersymmetry as well .
the geometric duality between a metric @xmath0 and its non - degenerate killing tensor @xmath1 was discussed in @xcite .
the relation was generalized to spinning spaces , but only at the expense of introducing torsion .
the physical interpretation of the dual metrics was not clarified @xcite .
the geometrical interpretation of killing tensors was investigated in@xcite .
recently the structural equations for a killing tensor of order two was investigated and the geometric duality between @xmath2 and a non - degenerate killing tensor @xmath3 was analyzed in@xcite .
an interesting example of an einstein s metric which admits killing - yano tensors is taub - nut metric .
taub - nut metric is involved in many modern studies in physics .
for example the kaluza - klein monopole of gross and perry @xcite and of sorkin @xcite was obtained by embedding the taub - nut gravitational instanton into five - dimensional kaluza - klein theory .
remarkably , the same object has re - emerged in the study of monopole scattering . in the long distance limit ,
neglecting radiation , the relative motion of the bps monopoles is described by the geodesics of this space @xcite@xcite .
the dynamics of well - separated monopoles is completely soluble and has a kepler type symmetry @xcite .
the geodesic motion of pseudo - classical spinning particles in euclidian taub - nut were analyzed in @xcite and the symmetries of extended taub - nut metrics recently were studied in @xcite .
taub - nut metric admits four killing - yano tensors which generate four non - degenerate killing tensors@xcite . on the other hand for a given manifold @xmath2 which admits a non - degenerate killing tensor @xmath3 two types of dual metrics exist @xcite .
an interesting question is to investigate the connection between the symmetries of dual taub - nut metrics and the symmetries of taub - nut metric . for these reasons
the symmetries of the taub - nut dual metrics will be analyzed .
the aim of this paper is to investigate the _ generic _ and _ non - generic _ symmetries corresponding to geodesic motion of pseudo - classical spinning particles on the taub - nut dual metrics.we will investigate the existence of the runge - lenz symmetry for the dual metrics .
the organization of the paper is as follows . in section 2
the geometric duality is presented . in section 3
we investigate the symmetries corresponding to taub nut dual metrics and we construct the spinning space . in section 4
we present our conclusions . in appendix 1
we write down christoffel symbols and the scalar curvature for two interesting dual metrics .
the calculus for all taub - nut dual metrics were done , but due to their huge and complicated expressions we can not write them out in this paper . in appendix 2 we present two plots of scalar curvature of two interesting dual metrics .
the equation of motion of a particle on a geodesic is derived from the action @xmath4 the hamiltonian has the form @xmath5 where the poisson brackets are @xmath6 .
let us suppose that the metric @xmath2 admits a killing tensor field @xmath3 .
a killing tensor is a symmetric tensor which satisfies the following relation : @xmath7 where @xmath8 denote covariant derivatives . from the covariant components @xmath3 of the killing tensor one can construct a constant of motion @xmath9
. it can be easy verified that @xmath10 .
the formal similarity between the constants of motion h and k , and the symmetrical nature of the condition implying the existence of the killing tensor amount to a reciprocal relation between two different models : the model with hamiltonian h and constant of motion k , and a model with constant of motion h and hamiltonian k.the relation between the two models has a geometrical interpretation : it implies that if @xmath3 are the contravariant components of a killing tensor with respect to the metric @xmath2 , then @xmath2 must represent a killing tensor with respect to the metric defined by @xmath3 .
when @xmath3 has an inverse we interpret it as the metric of another space and we can define the associated riemann - christoffel connection @xmath11 as usual through the metric postulate @xmath12 . here
@xmath13 represents the covariant derivative with respect to @xmath3 .
this reciprocal relation between the metric structure of pairs of spaces constitutes a duality relation : performing the operation of mapping a killing tensor to a metric twice leads back to the original theory .
the geometric duality between @xmath2 and a killing tensor @xmath3 was analyzed in @xcite.in this case killing s vectors equations in the dual space have the following form @xcite @xmath14 here @xmath15 are killing vectors in dual spaces .
let us suppose that metric @xmath2 admits a killing - yano tensor @xmath16 .
a killing - yano tensor is an antisymmetric tensor @xcite which satisfies @xmath17 the corresponding killing - yano equations in the dual space has the form@xcite @xmath18 where @xmath19 is a killing - yano tensor on the dual manifold .
d represents the covariant derivative with respect to @xmath2 .
the four - dimensional taub - nut metric depends on a parameter @xmath20 which can be positive or negative , depending on the application ; for @xmath21 it represents a nonsingular solution of the self - dual euclidean equation and as such is interpreted as a gravitational instanton .
the standard form of the line element is @xmath22 the killing vectors for the metric ( [ ade ] ) have the following form : @xmath23 where @xmath24 the metric ( [ ade ] ) admits four killing - yano tensors@xcite .
three of these , denoted by @xmath25 are special because they are covariant constant . in the two - form notation
the explicit expressions are @xmath26 where the @xmath27 are standard expressions in terms of the 3-dimensional spherical co - ordinates @xmath28 .
the fourth killing - yano tensor has the following form @xmath29 now we would like to investigate the symmetries of the dual taub - nut metrics .
we will construct firstly the taub - nut dual metrics using the geometric duality between @xmath2 and @xmath3 @xcite .
we know that when a manifold m admits a killing - yano tensor @xmath16 we can construct a corresponding killing tensor@xcite @xmath3 as @xmath30 using ( [ eky ] ) and ( [ ky ] ) the line element for the dual metric becomes @xmath31 the taub - nut metric admits three more second - rank killing tensors of the form @xcite @xmath32 here @xmath33 and @xmath34 are given by ( [ kyf ] ) and([ky ] ) .
they form a conserved vector of runge - lenz type .
using ( [ ru ] ) the line element for the dual metrics becomes @xmath35 we found that the dual metric ( [ kilten ] ) has the same killing vectors like taub - nut metric ( [ ade ] ) because relation ( [ kiltex ] ) is identically satisfied .
for @xmath36 the corresponding metric from ( [ taubi ] ) admits two killing vectors ( [ vec14 ] ) and ( [ vec15 ] ) but for @xmath37 we found from ( [ kiltex ] ) only one killing vector ( [ vec14 ] ) . because the weyl tensor of ( [ kilten ] ) has non - vanishing components + ( e.g @xmath38 )
, the metric is not conformally flat .
we have obtained the same result for ( [ taubi ] ) but the expressions are too long to be provided here .
now we would like to investigate the killing - yano tensors of order two for the dual metrics ( [ kilten],[taubi ] ) .
we have six independent components of killing - yano tensor @xmath16 and 24 independent equations ( [ anoua ] ) . replacing @xmath39 in ( [ anoua ] ) and using the corresponding expressions of christoffel s symbols for ( [ kilten ] ) we get a set of killing - yano equations .
we found that ( [ anoua ] ) has no solution , therefore the dual metrics ( [ kilten ] ) and ( [ taubi ] ) have no extra symmetry of runge - lenz type . in @xcite using geometric duality between @xmath0 and @xmath1 four dual tab - nut metrics were found .
the inverse matrix of the covariant form from ( [ kilten ] ) give us the dual line element @xmath40 and the dual metrics corresponding to runge - lenz vector have the following form ( for more details see @xcite ) .
@xmath41 next step is to investigate the symmetries of the metrics ( [ elem],[31eq ] ) .
if we make the transformation to a new variable u as @xmath42 the line element ( [ elem ] ) becomes @xmath43 where @xmath44 and @xmath45 are given by @xmath46 .
the metric ( [ alta ] ) is a particular form of extended taub - nut metric presented in @xcite . on the other hand , in the case of geometric duality between @xmath0 and @xmath1 the equations ( [ kiltex ] ) and ( [ anoua ] ) are not valid ( for more details see @xcite ) .
because of this the symmetries of the metrics ( [ elem],[31eq ] ) will be investigated using ( see @xcite ) the killing vectors equations @xmath47 and the killing - yano equations @xmath48 solving ( [ kilnou ] ) we found that the dual metric ( [ elem ] ) admits the same killing vectors as metric ( [ ade]).this result is in agreement with those from@xcite . for @xmath36 , ( [ kilnou ] ) give us for the corresponding metric in ( [ 31eq ] ) two killing vectors ( [ vec14 ] ) and ( [ vec15 ] ) . in the case
@xmath37 from ( [ 31eq ] ) we found only one killing vector ( [ vec14 ] ) for the corresponding metrics .
now we investigate if the dual metrics ( [ elem],[31eq ] ) admits killing - yano tensors of order two .
our strategy is quite straightforward in this case .
we simply write down all the components of the equation ( [ yanotex ] ) explicitly .
the total number of components of @xmath16 is six , while the number of independent equations for eq.([yanotex ] ) is 24 .
@xmath49 solving ( [ tabel ] ) we found no solution for the metrics ( [ elem ] ) and ( [ 31eq ] ) .
then the dual metrics ( [ elem],[31eq ] ) have no killing - yano tensors and runge - lenz type symmetry .
the weyl tensor has non - vanishing components for ( [ elem])(e.g .
@xmath50 ) and then the metric in not conformally flat
. metrics ( [ 31eq ] ) are not conformally flat because they have non - vanishing weyl tensor components , but all expressions are too long to be written here .
an action for the geodesic of spinning space is : @xmath51 the overdot denotes an ordinary proper - time derivative @xmath52 , whilst the covariant derivative of a grassmann variable @xmath53 is defined by @xmath54 . in general , the symmetries of a spinning - particle model can be divided into two classes .
first , there are conserved quantities which exist in any theory and these are called _ generic _ constants of motion .
it was shown that for a spinning particle model defined by the action ( [ spin ] ) there are four generic symmetries @xcite . 1 .
proper -time translations and the corresponding constant of motion are given by the hamiltonian : @xmath55 2 .
supersymmetry generated by the supercharge @xmath56 3 .
chiral symmetry generated by the chiral charge @xmath57 4 .
dual supersymmetry , generated by the dual supercharge @xmath58 here @xmath59 represents the covariant momentum .
the second kind of conserved quantities , called _ non - generic _ , depend on the explicit form of the metric @xmath60 .
_ non - generic _ symmetries are associated with the existence of killing- yano tensors on a given manifold .
the existence of a killing - yano tensor @xmath16 of the bosonic manifold is equivalent to the existence of a supersymmetry for the spinning particle with supercharge @xmath61 satisfies @xmath62 , where @xmath63 . for the metric ( [ kilten ] )
we found the _
generic _ symmetries ( [ prime][ultim ] ) in the following form @xmath64 @xmath65 @xmath66 @xmath67 because the determinants of all dual metrics ( [ taubi],[elem],[31eq ] ) , do not vanish we have similar expressions , in corresponding cases , for the _ generic _ symmetries + ( [ prime][ultim ] ) .
as it was shown above the dual metrics have no killing - yano tensors , and because of this we have no _ non - generic _ symmetries for metrics ( [ kilten ] , [ taubi],[elem],[31eq ] ) .
recently geometric duality was analyzed for a metric which admits a non - degenerate killing tensor of order two @xcite . for a given manifold which admits a non - degenerate killing tensor of order two , geometric duality
give us two types of dual metrics @xcite .
an interesting example arises when the manifold admits killing - yano tensors because they generate killing tensors . in this paper
the symmetries of the dual taub - nut metrics were investigated .
taub - nut metric admits four killing - yano tensors of order two and we have four corresponding dual metrics .
we have obtained that the number of killing vectors of the dual taub - nut metrics depend drastically on their particular form .
we found that metrics ( [ kilten],[elem ] ) have the same killing vectors as the taub - nut metric ( [ ade ] ) , the corresponding dual metrics for @xmath36 in ( [ taubi],[31eq ] ) have two killing vectors ( [ vec14],[vec15 ] ) , and for @xmath37 we get only one killing vector ( [ vec14 ] ) .
we have obtained , by solving killing - yano equations , that all dual taub - nut metrics do not have killing - yano tensors .
this means that all dual metrics do no admit extra symmetries of runge - lenz type .
the scalar curvature of taub nut metric ( [ ade ] ) is zero , but the corresponding dual metrics ( [ kilten],[taubi],[elem],[31eq ] ) have non vanishing scalar curvatures and no runge - lenz vector .
we found that the scalar curvature of dual metric ( [ kilten ] ) is positive and the corresponding scalar curvature for ( [ elem ] ) is negative .
we have obtained that the dual metric ( [ elem ] ) is a special case of extended taub - nut metric given in @xcite .
we obtained also that all dual taub - nut metrics are not conformally flat .
the spinning space was constructed and the _ generic _ and _ non - generic _ symmetries of the dual taub - nut metrics were analyzed .
we found that dual taub - nut metrics have not _ non - generic _ symmetries .
our result differs from those presented in @xcite .
geometric duality , in case of taub - nut metric , reduces the number of symmetries of the dual metrics . finding the number of symmetries of dual metrics for a given manifold which admits a killing tensor is an interesting problem , and it requires further investigation .
rietdjik , j.w .
van holten 1996 nucl
( b472 ) * 472 - 446 .
za ' nski,_9th italian conference on general relativity and gravitational physics _ , pp.2 - 9 , r.cianci et al .. eds .. world scientific , singapore,(1991 ) + d.baleanu,gr-qc/9805053 + d. baleanu , gr - qc/9805052 d.j.gross and m.j.perry , _ nucl.phys . _ * b226 * ( 1983 ) 29 .
r.d.sorkin , _ phys.rev.lett . _
* 51 * ( 1983 ) 87 .
n.s.manton , _ phys.lett . _
* b110 * ( 1982 ) 54 ; i d , * b154 * ( 1985 ) 397 ; i d , ( e ) * b157 * ( 1985 ) 475 .
m.f.atiyah and n.hitchin , _ phys.lett .
_ * a107 * ( 1985 ) 21 .
gibbons and p.j .
ruback , phys . lett .
* b188 * ( 1987 ) 226 ; commun .
115 * ( 1988 ) 267 .
gibbons and n.s .
manton , nucl .
b274 * ( 1986 ) 183 .
feher and p.a .
horvathy , phys . lett . *
b182 * ( 1987 ) 183 ; i d , ( e ) * b188 * ( 1987 ) 512 .
b. cordani , l.gy .
feher and p.a .
horvathy , phys . lett . * b201 * ( 1988 ) 481 .
d.baleanu , _ helv.phys.acta _ * 67 * ( 1994)405 + d.baleanu , _ il nuovo cimento b _ * 109 * ( 1994)845 + d.baleanu , _ il nuovo cimento b _
* 111*(1996)973 + m. viinescu and d. vaman , _ phys.rev . _ * d57*(1998 ) 3790 - 3793 .
d. baleanu , gen .
rel . and grav .
1998 * 30 no.2 * 195 .
d. baleanu , helv .
acta phys .
1998 * 70 * .
d. baleanu , _ non - generic symmetries on extended taub - nut metric _ particles , fields and gravity , lodz
poland,(1998 ) 15 - 19 april j.w.van holten , phys.lett.*b 342*(1995 ) 47 - 52 for the metric ( [ kilten ] ) non - vanishing christoffel coefficients are + @xmath68 @xmath69 @xmath70 @xmath71 @xmath72 @xmath73 @xmath74 @xmath75 @xmath76 @xmath77 @xmath78 .5 cm and the curvature + @xmath79 .5 cm | in this paper we study the symmetries of the dual taub - nut metrics . _
generic _ and _ non - generic _ symmetries of dual taub - nut metrics are investigated .
the existence of the runge - lenz type symmetry is analyzed for dual taub - nut metrics .
we find that in some cases the symmetries of the dual metrics are the same with the symmetries of taub - nut metric .
.3 in .3 in -1.5 cm 0.5 cm _ joint institute for nuclear research _
+ _ bogoliubov laboratory of theoretical physics _ , + _ 141 980 dubna , moscow region , russia _ + 0.8 cm 1 cm |
an intriguing variation on the grand unification theme that nucleons are likely to be unstable is the proposal that neutrons can oscillate into antineutrons .
neutron - antineutron oscillations were first predicted in 1970 by v. a. kuzmin in a model intended as a realization of requirements given earlier by a.d .
sakharov for evolution of the universe to net baryon asymmetry .
subsequently @xmath0oscillations emerged as a predicted reaction in certain grand unification theories @xcite .
more recently it has been shown that @xmath0oscillations can occur in a large class of supersymmetric @xmath5 models @xcite .
in such models the dominant baryon number violating process is a @xmath6 nucleon transition ( e.g. @xmath0oscillations or p + n @xmath7 pions ) rather than a @xmath8 nucleon - antilepton transition ( e.g. p @xmath9 or p @xmath10 ) .
neutron - antineutron oscillations have also been indicated as viable by recent gut models which invoke the existence of extra spacetime dimensions @xcite . if indeed a neutron can evolve into an antineutron , the experimental signatures for the metamorphosis should be distinctive .
the resulting antineutron will annihilate with a baryon of the surrounding environment , producing multiple mesons ( @xmath11 ) whose visible energy and net momentum are approximately those of two nucleon masses having nuclear fermi motion . from the phenomenology of neutron - antineutron oscillations it can be shown @xcite that the oscillation time @xmath12 of a neutron bound within a nucleus of atomic mass @xmath13 is related to the neutron oscillation time in vacuum according to @xmath14 here @xmath15 , which has units of inverse time , is the suppression factor representing the effect of the nuclear environment which substantially prolongs the effective oscillation time .
detailed calculations of the suppression factor @xmath15 for parent nuclei of experimental interest , including deuterium , oxygen , argon , and iron , have been reported in literature .
the calculations utilize phenomenological frameworks provided by nuclear potential theory , and by s - matrix theory . in the analysis of dover , gal , and richard @xcite ,
it is proposed that neutron - antineutron oscillations will occur mostly in outer nuclear shells and near the nuclear surface .
however reservations concerning this picture have been expressed and in a number of calculations the entire nuclear volume contributes to @xmath0oscillations @xcite .
two types of experiments have been used to search for neutron - antineutron oscillations . in one approach ,
slow neutrons from a fission reactor are channeled through a magnetically shielded vacuum pipe towards a target region .
an antineutron produced during the flight will annihilate in the target and the annihilation products are registered by detectors surrounding the target .
experiments of this type have been carried out at pavia @xcite and at grenoble .
the grenoble reactor experiment obtained @xmath16 s at 90% confidence level ( cl ) .
this is the most stringent oscillation time lower limit reported to date using free neutrons .
the alternate approach , used in this experiment , is to continuously monitor neutrons bound in nuclei , usually as part of an ongoing nucleon decay search .
searches of this type have been reported by the underground experiments homestake @xcite , nusex @xcite , kolar @xcite , imb @xcite , kamiokande @xcite , and frjus @xcite .
the searches by kamiokande and frjus obtained the most stringent @xmath0oscillation time lower limits . in a search based upon a 1.11 kiloton - year ( kty ) exposure of the kamiokande - i water cherenkov detector @xcite , no candidate event was observed .
an oscillation time lower limit of @xmath17 4.3 @xmath18 years at 90% cl was set . using the suppression factor @xmath19 s@xmath20 calculated by dover _
et al . _ for oxygen @xcite
, kamiokande obtained an oscillation time limit for free neutrons of @xmath21 seconds at 90% cl .
the frjus collaboration , in a search using a 1.56 fiducial kty exposure of the experiment s planar iron tracking calorimeter , also reported zero @xmath0oscillation candidates . the oscillation time lower limit for @xmath22 in iron thereby obtained was @xmath23 years at 90% cl .
using @xmath24 s@xmath20 as calculated by dover _
et al . _ for iron , frjus determined the free neutron limit to be also @xmath25 @xmath26 seconds at 90% cl @xcite . in the kamiokande analysis an enhanced probability for @xmath0oscillations to occur in the nuclear periphery as postulated by dover _
et al . _ was assumed . in the monte carlo simulations of the experiment ,
the effect of this assumption is to reduce distortion of the final state meson spectrum arising from intranuclear absorption and inelastic scattering processes . as a result experimental detection efficiencies
are enhanced relative to expectations for the case where oscillations may occur throughout the entire volume of parent nuclei .
the search reported here follows the more conservative approach adopted previously by frjus . for our primary simulation on which our detection efficiency is based , we assume @xmath0oscillation to occur throughout the nuclear volume
. however the effects of peripheral predominance are also described .
soudan 2 is a 963 metric ton ( 770 tons fiducial ) iron tracking calorimeter of honeycomb lattice geometry which operates as a slow drift time projection chamber @xcite .
the tracking elements are one - meter - long drift tubes filled with an argon
co@xmath27 gas mixture .
electrons liberated by throughgoing charged particles drift to the tube ends under the action of a voltage gradient applied along the tubes .
the drift charge is registered by vertical anode wires , while horizontal cathode pad strips register the image charges .
the third coordinate is obtained from the drift time .
the amount of charge measures the deposited ionization .
the drift tubes are laid onto corrugated steel sheets , and the sheets are stacked to form 1@[email protected] m , 4.3 ton modules from which the calorimeter is assembled in building - block fashion . surrounding the tracking calorimeter on all sides but mounted on the cavern walls and well separated from calorimeter surfaces is a 1700 m@xmath29 active shield array of two or three layers of proportional tubes @xcite .
the shield facilitates identification of events which are not contained within the calorimeter .
in particular , it provides tagging of background events initiated by cosmic - ray - induced neutrons .
the detector is located at a depth of 2070 meters water
equivalent on the 27th level of the soudan underground mine state park in northern minnesota .
the modular design enabled data taking to commence in april 1989 when the detector was one - quarter of its final size ; routine operation with the fully deployed detector got underway in november 1993 .
the fiducial ( total ) exposure analyzed here , obtained from data taking through december 2000 , is 5.56 ( 6.96 ) kiloton years .
calibration of calorimeter module response was carried out at the rutherford isis spallation neutron facility using test beams of positive and negative pions , electrons , muons , and protons @xcite .
spatial resolutions for track reconstruction and for vertex placement in anode , cathode , and drift time coordinates are of the same scale as the drift tube radii , @xmath30 0.7 cm . in soudan 2
, ionizing particles having non - relativistic as well as relativistic momenta are imaged with @xmath31 sampling in a fine grained honeycomb lattice geometry .
protons can be distinguished from pions and muons via ionization and ranging , energetic muons discriminated from pions via absence of secondary scatters , and prompt e@xmath32 showers distinguished from photon showers on basis of proximity to primary vertices .
these event imaging capabilities offer advantages , in comparison to water cherenkov detectors and to planar iron calorimeters , for analysis of the complicated multiprong topologies that would arise with annihilations initiated by @xmath0oscillations .
we have developed realistic simulations of @xmath0oscillations yielding annihilations as they would occur in the iron nuclei which comprise the bulk of the calorimeter mass .
generation of events is carried out as follows : momenta of the initial state antineutron and nucleon are assigned according to a distribution based upon a fermi - gas model parameterization of quasi - elastic electron nucleus scattering @xcite .
final state particle four - momenta were constructed in accordance with n - body phase space @xcite .
assignment of reactions to generated events is weighted according to cross section data for @xmath33p annihilation at rest .
for the purpose of channel selection , @xmath34n annihilations are assigned the same cross sections as observed for @xmath33p ; @xmath34p is assumed to have the same total rate as @xmath33p , and cross sections were inferred from @xmath33p as allowed by charge conservation .
we restricted our reaction compilation to cross sections exceeding 2% of the total @xmath35 cross section
. consequently production of @xmath36 , @xmath37 , charged and neutral pions and kaons is represented , however final states with @xmath38 and @xmath39 mesons were neglected .
provision was made to include intranuclear rescattering ( ins ) of final state pions in the simulations .
our treatment follows the approach utilized previously in simulations of atmospheric neutrinos and of nucleon decay @xcite .
pions , either directly produced in annihilation or in @xmath36 decays , are propagated through a model nucleus .
the nuclear radius and density are parameters of the model ; the radius is scaled according to @xmath40 .
scattering in the nuclear medium is characterized using a momentum - dependent pion interaction length .
the nuclear parameters were set by requiring the model to reproduce single and multiple pion production rates observed in bubble chamber @xmath41-deuteron ( @xmath13=2 ) and @xmath41-neon ( @xmath13=20 ) interactions @xcite . for pions from vertices placed at random within an iron nucleus , our model predicts @xmath42 to emerge either unscattered or to have undergone only small angle elastic scattering . about 30%
are predicted to undergo total absorption , while the remainder undergo charge exchange ( 9% ) or emerge having undergone inelastic scattering ( 16% ) .
our treatment of nuclear rescattering does not include proton secondaries . while it is expected that protons would occasionally be ejected from parent nuclei as the result of rescattering
, their momenta would almost always fall below the effective threshold ( approx . 450
mev/@xmath43 ) for creation of distinct tracks in the detector . for simulations with intranuclear rescattering included , the average multiplicity per event for mesons emerging from an iron nucleus is 3.8 with rms deviation 1.2 . the average track plus shower multiplicity emitted from an iron nucleus ( prong multiplicity " ) per event is 5.0 with an rms deviation of 2.1 prongs .
( note these multiplicities are for perfect detection " . )
the event generation produces four - momenta of all particles which exit the parent nuclei .
this information is fed , event - by - event , into the soudan 2 monte carlo ( mc ) which provides a realistic detector response .
the simulation includes detector background noise " arising from natural radioactivity and from the electronics .
event records are generated with format identical to that of data events , allowing @xmath0oscillation events to be processed using the same codes and procedures as for data and for events of the atmospheric neutrino monte carlo .
three separate @xmath44n simulation samples were generated .
for each sample , events were generated at random locations throughout the tracking calorimeter as it evolved during the experiment s eleven years of data - taking . in our primary " simulation , annihilations were started at random points throughout their parent nuclei and pion intranuclear rescattering was implemented . in a second simulation which also included intranuclear rescattering ,
annihilations were restricted to the nuclear periphery ( @xmath45 , where @xmath46 is the nuclear radius ) .
a third simulation was carried out in which no pion intranuclear rescattering was included .
each of the three oscillation samples was processed using a sequence of selections and procedures very similar to that routinely used in reduction of data events in soudan 2 .
each sample was subjected ( in software ) to the hardware trigger .
events which passed the hardware trigger requirements were subjected to two different software filter " codes .
the filters impose event containment criteria , e.g. that no track from the event approaches closer than 20 cm to the detector outer surfaces ; they also mitigate against backgrounds arising from cosmic ray muons , natural radioactivity , and detector noise .
events which survived the filters were then subjected to two separate scans by physicists .
scanning was carried out using interactive color graphics workstations .
the scan rules provided refinements to the filter selections and introduced requirements on imaging quality , e.g. an event was rejected if ( _ i _ ) its primary vertex occurred in material interior to the detector but not instrumented , or if ( _ ii _ ) its proximity to inter - module gaps compromised the reconstruction of the event .
events which survived successive application of the hardware trigger , software filters , and physicists scans are tallied in the upper four rows of table [ tbl : survival_rates ] for each of the three simulations .
these entries are input to the calculation of the detection efficiency for neutron - antineutron oscillations in soudan 2 as will be described in sect .
[ sec : deteff ] .
preliminary to kinematic reconstruction , the topology of each event was characterized in terms of track and shower prongs .
a track prong results when an ionizing , non - showering charged particle ( e.g. a pion , muon , or proton ) traverses drift tubes of the calorimeter s honeycomb lattice leaving a continuous trail of tube hits " .
a shower prong on the other hand is created by photon conversion or by a primary electron or positron .
the electromagnetic shower consists of many distinct particles and exhibits a cone - like pattern of hits , interspersed with gaps , which is aligned with the direction of the initiating photon or electron .
the mean conversion length for photons in soudan 2 is approximately 15 cm . as a consequence ,
photon - induced showers generally appear in the vicinity of but moderately displaced from event primary vertices .
images of two @xmath0oscillation events simulated with full detector response are displayed in fig .
[ fig : nnbmcevents.eps ] , where the anode ( @xmath47 ) versus drift time ( @xmath48 ) view of each event is shown . for event scanning , three views are always used , including cathode - time and anode - cathode as well as anode - time projections .
( the anode - cathode view is automatically assembled for all events , mc as well as data , using demultiplexing and hit - matching algorithms . ) as suggested by fig .
[ fig : nnbmcevents.eps ] , @xmath44n events appear to be energetic yet isotropic to a degree uncommon for data events and for events of the atmospheric neutrino monte carlo . the occurrence of relatively high multiplicity per event of track plus shower prongs is a signature feature of annihilation . from simulation of @xmath0oscillation events in the absence of intranuclear rescattering
, we find the prong multiplicity distribution - after triggering , filtering , scanning , and reconstruction - to have a mean of 6.3 prongs per event with an rms width of 2.8 prongs . here
the prong multiplicity is relatively high ( compared to multiplicities as generated , see sect .
2.2 ) , due to depletion of low multiplicities by the processing . with
intranuclear rescattering included , the prong multiplicity distributions of processed samples are shifted lower .
for both simulations with ins included , the multiplicity distributions have mean values of 4.8 prongs with rms widths of 3.2 prongs .
thus a sizable contribution from multiplicities exceeding three - prongs survives in the latter simulations .
events of this kind are relatively uncommon among contained atmospheric neutrino events .
in addition to event topology , the reconstructed kinematic quantities @xmath49 , the visible energy , and @xmath50 , the net momentum , provide discrimination between and background events .
these quantities and their correlation have been used previously in our searches for nucleon decay @xcite .
@xmath49 is calculated as the sum of the relativistic energies for each of the final state tracks ( using the pion mass ) and showers .
there occur a few short , heavily ionizing tracks which satisfy our identification criteria for protons @xcite . for these ,
only the kinetic energy is added into the calculation of @xmath49 .
distributions of @xmath49 versus @xmath50 for events in the absence of and including intranuclear rescattering , are plotted in fig .
[ fig : nnb_evis_p.eps]a and fig .
[ fig : nnb_evis_p.eps]b respectively .
reconstructed events , even in the absence of ins as in fig .
[ fig : nnb_evis_p.eps]a , exhibit large energy losses arising from low energy prongs and secondary hadronic scatters which are unresolved .
consequently the events cluster well below @xmath49 of 1.88 gev .
comparison of fig .
[ fig : nnb_evis_p.eps]a with fig .
[ fig : nnb_evis_p.eps]b shows that a further large degradation arises with ins .
the ins effects substantially increase the kinematic overlap of events with atmospheric neutrino events .
since events generally have high multiplicities we restrict our analysis to events having a multiprong " topology , that is , having two or more produced particles emerging from primary vertices .
quasi - elastic neutrino interactions which produce a charged lepton plus a recoil proton are readily distinguished and removed . as was done for the simulated @xmath0events , we require all events under consideration to be fully contained in an interior volume which is everywhere 20 cm from calorimeter outer surfaces .
three multiprong event samples , which are in addition to the simulated @xmath0samples discussed above , have been isolated for this search : first and foremost are data events for which the cavern - liner active shield array was quiescent during the allowed time window .
these events comprise our _ shield - quiet data
sample_. they originate mostly with atmospheric neutrino interactions but would also include @xmath0oscillation events .
there are 188 multiprong events in the shield - quiet data of this analysis .
there is a second category of data events which comprises a background to neutrino multiprongs as well as to @xmath0oscillations .
events of this category are usually produced by energetic neutrons released in inelastic cosmic ray muon interactions with the cavern rock surrounding the tracking calorimeter .
most of these _ rock events _ are accompanied by charged particles which give coincident hits in the active shield array ; they constitute our _ shield - tagged rock event sample_. however , a few rock events are not accompanied by shield hits , either because of shield inefficiency ( detection efficiency is @xmath52 ) or because no charged particles enter the cavern along with a rock neutron . the latter _ shield - quiet rock events
_ may end up as background in shield - quiet data .
fortunately the shield - tagged rock events provide a control sample from which the amount of residual rock background in shield - quiet data may be inferred .
the predominant background for @xmath0oscillations arises from the atmospheric neutrino reactions .
their contribution is evaluated using the soudan 2 _ atmospheric neutrino monte carlo sample_. our neutrino mc simulation utilizes the atmospheric flux calculation of the bartol group for the soudan site @xcite .
details of the neutrino event generation and comparisons with low energy @xmath53 data are given elsewhere @xcite .
mc atmospheric neutrino events were inserted into the experiment s data stream during data - taking .
the mc events were introduced at a rate 6.06 times higher than the data rate expected ( for null neutrino oscillations ) from the atmospheric neutrino flux .
the events were processed identically to real data events , with their identity revealed only at the final analysis stage .
the atmospheric @xmath51 mc multiprong sample finally extracted contains 1267 events .
properties of our multiprong data , rock , and neutrino mc samples germane to an @xmath0oscillation search are reported below .
further details can be found in previous publications @xcite .
the prong multiplicity distribution of the shield - quiet data sample is summarized using tallies in the grid displayed in table [ tbl : trk_vs_shw ] . in the table ,
the topology of each event is represented by the number of tracks , @xmath54 , and number of showers , @xmath55 , which comprise the visible final state .
the number of data events which have a particular @xmath56 combination are given by the upper entries at the grid location having the appropriate integer coordinates .
the multiprong data topologies having highest rates are seen to be combinations of lowest multiplicity , namely @xmath57 and ( 1,1 ) . for comparison ,
the corresponding prong distribution for the atmospheric neutrino mc event sample is given by the lower entries in each of the track - shower grid locations of table [ tbl : trk_vs_shw ] .
the mc sample ( for null neutrino oscillations ) has been normalized to the same fiducial exposure as for the data ( 1267/6.06 = 209.1 events ) .
the two distributions are seen to match well , e.g. the @xmath58/bin averages to below 1.0 , with the exception of the ( 1,2 ) topology . since most shield - quiet data multiprongs are initiated by atmospheric neutrinos ,
the agreement is evidence that the neutrino mc provides a good general representation of our data .
tracks & + ' '' '' & _ data _ & 0 & 0 & 0 & 2 & 1 & 1 & + & _ mc _ & 0.3 & 0.3 & 0.3 & 0 & 0.2 & 0.5 & + & _ data _ & 3 & 1 & 3 & 0 & 1 & 0 & + & _ mc _ & 2.1 & 0.3 & 0.2 & 1.6 & 0.2 & 0.2 & + & _ data _ & 16 & 7 & 2 & 2 & 0 & 0 & + & _ mc _ & 14.5 & 5.1 & 3.1 & 1.4 & 1.6 & 0.3 & + & _ data _ & 26 & 20 & 15 & 4 & 0 & 2 & + & _ mc _ & 26.8 & 21.1 & 14.6 & 5.9 & 2.7 & 1.3 & + & _ data _ & & 25 & 10 & 5 & 3 & 3 & + & _ mc _ & & 28.3 & 23.6 & 9.3 & 5.0 & 1.6 & + & _ data _ & & & 14 & 13 & 4 & 1 & + & _ mc _ & & & 12.4 & 9.0 & 2.9 & 2.1 & + & & & & & & & & + & & 0 & 1 & 2 & 3 & 4 & 5 & + & & & & & & & & + kinematics for the shield - quiet multiprong data is shown by the @xmath49 versus vector net momentum diplot of fig . [ fig : evis_p_data_numc.eps]a .
the events populate a broad , correlated region extending from threshold to 1.5 gev ( 1.5 gev/@xmath43 ) in visible energy ( net momentum ) .
portions of this data distribution overlap the distribution predicted for @xmath0events shown in fig .
[ fig : nnb_evis_p.eps]b . a direct comparison with kinematics for the atmospheric neutrino mc sample is provided by fig . [ fig : evis_p_data_numc.eps]b .
note that the latter sample has an exposure - equivalent of 33.7 fiducial kton - years .
the distribution of the mc sample , normalized to the exposure , is very similar to that of the data . in fig .
[ fig : evis_p_data_numc.eps ] , events which have prompt protons are denoted by solid circles .
as previously noted , such events are highly improbable as @xmath0oscillations .
however our atmospheric @xmath59 mc indicates that detectable recoil protons are to be expected in 24% of neutrino multiprongs , and this expectation is born out by the frequency of recoil proton events observed in the data of fig .
[ fig : evis_p_data_numc.eps]a . in fig .
[ fig : evis_p_data_numc.eps]a there is a tendency for data events having recoil protons to occur in a region parallel to and below the more populated kinematic band .
this trend is reproduced in the mc distribution of fig .
[ fig : evis_p_data_numc.eps]b .
rock data events which are flagged by hits in the surrounding shield array , have a distribution of vertex depth into the detector which reflects the mean hadronic interaction length of neutrons in the soudan detector medium , approximately 80 cm .
a rock background component in our shield - quiet multiprong data will have a corresponding vertex depth distribution in the detector .
neutrino and @xmath0oscillation events on the other hand , will distribute uniformly throughout the detector volume .
thus by fitting the distribution of vertex depths observed in our multiprong data to the sum of two component distributions representing vertex depths of shield - tagged rock and of neutrino mc events , the amount of shield - quiet rock contamination can be determined @xcite .
on the basis of a maximum likelihood fit , we estimate that a fraction @xmath60 of shield - quiet data multiprongs are produced by rock neutrons .
four selection criteria for neutron oscillation candidates ( table [ tbl : criteria ] ) have been defined and applied to all event samples .
the selections are designed to maximize @xmath0detection in the shield - quiet multiprong data sample while minimizing backgrounds , the principal one arising from inelastic interactions of atmospheric neutrinos . _ ) _ _ prong multiplicity _ : we require the sum of track and shower prongs directly associated with the primary vertices to be greater than or equal to four .
this value is below but near to the mean multiplicities for processed @xmath0samples as discussed in sect .
[ sec : processed_samples ] . lowering our selection to include three - prongs would increase the efficiency for detection by only 3% , however backgrounds arising from atmospheric neutrinos and rock events would increase by 21% and 32% respectively .
_ primary proton tracks _ : events having proton tracks emerging from their primary vertices are removed from consideration as @xmath0oscillation candidates .
the rate of prompt protons is negligible for annihilation events , however @xmath61 of atmospheric neutrino events and @xmath62 of rock events have visible recoil protons .
_ primary muon tracks _ : pions produced in annihilations usually scatter , become absorbed , or range - to - stopping over distances comparable to the calorimeter s hadronic interaction length of 80 cm . in order to mitigate against @xmath63
charged current background , we regard any non - scattering track which has pion / muon ionization and range - to - stopping @xmath1 150 cm ( 240 g / cm@xmath29 ) to be a muon track . any event having such a track is rejected .
this selection eliminates 36% of the atmospheric neutrino sample while cutting less than 10% of @xmath0events of the primary simulation .
_ event kinematics _ : as our final criterion we require the kinematics of @xmath0candidates to be compatible with those of annihilation .
the event net momentum fraction @xmath65 ( hereafter @xmath66 ) has previously been shown to be a useful variable for separating @xmath0oscillation events from atmospheric neutrino reactions @xcite , and its utility is confirmed in our analysis .
figures [ fig : nnb_sims_final.eps]a and [ fig : nnb_sims_final.eps]b show @xmath66 versus @xmath49 for events after topology cuts ( 1 through 3 ) from simulations in which oscillations occur throughout the parent nuclear volumes . for the case of no
intranuclear rescattering ( fig .
[ fig : nnb_sims_final.eps]a ) , approximately 94% of @xmath0oscillation events occur in a region having @xmath66 less than 0.6 , with @xmath49 ranging from one to two nucleon masses . with intranuclear rescattering included , fig .
[ fig : nnb_sims_final.eps]b , the @xmath0event population shifts to lower values of @xmath49 and to higher values of @xmath66 . nevertheless a degree of clustering remains which allows a useful search region in the ( @xmath49 , @xmath66 ) plane to be defined .
our selection for @xmath0kinematics is depicted by the rectangular region ( dotted - line boundary ) shown in the figs [ fig : nnb_sims_final.eps]-[fig : mpdata - final.ps ] .
candidate @xmath0events are required to have @xmath66 @xmath67 0.6 and @xmath68 mev .
these cuts are designed to minimize the atmospheric neutrino background shown in fig .
[ fig : mpmc - final.ps ] while maximizing the acceptance for events .
the reductions in each annihilation sample upon successive @xmath0candidate selections are summarized in the bottom three rows of table [ tbl : survival_rates ] .
it can be seen that intranuclear rescattering significantly lowers the survival rates . for our primary simulation , with ins operative and with annihilations occurring throughout the volume of the parent nucleus , 86 events from an initial 491 event sample
survive all selections , giving a detection efficiency of 0.18 @xmath2 0.02 .
detection efficiencies for the comparison simulations are higher , @xmath69 for the simulation without ins , and @xmath70 for the simulation which includes ins but restricts annihilations to the nuclear periphery .
shield - quiet data events which satisfy @xmath0selections 13 of table [ tbl : criteria ] distribute in the @xmath66 versus plane as shown in fig .
[ fig : mpdata - final.ps ] . of these sixteen events ,
five occur within the kinematically allowed region and are candidate @xmath0oscillation events of this search .
three of the candidates have four - prong topologies , while the remaining two are five - prong events .
two views of one of the candidates consisting of four pion - or - muon tracks emerging cleanly from a reaction vertex , are shown in fig .
[ fig : r68882.ps ] .
this event is also consistent with multipion production by atmospheric neutrinos , e.g. @xmath71 @xcite .
[ fig : mpmc - final.ps ] shows that , among atmospheric @xmath51 mc events which satisfy topology selections 13 of table [ tbl : criteria ] , there are 28 events ( 11 @xmath63 cc , 7 @xmath72 cc , and 10 nc ) which occur in the selected kinematic region .
that is , 28 neutrino - induced background events would occur in a 33.7 fiducial kiloton - year exposure to an atmospheric neutrino flux having null neutrino oscillations .
however strong evidence for atmospheric neutrino oscillations now exists @xcite , @xcite , .
we allow for neutrino oscillations in our background estimate by weighting our mc @xmath63 cc prediction by the atmospheric neutrino flavor ratio - of - ratios : @xmath73_{data}/[({\mbox{$\nu_{\mu}$}}+ \mbox{$\overline{\nu}_{\mu}$})/ ( { \mbox{$\nu_{e}$}}+ \mbox{$\overline{\nu}_{e}$})]_{mc}$ ] .
we have used the value of @xmath74 measured in this experiment , based upon a 5.14 fiducial kton - year exposure : @xmath75 @xcite .
this correction reduces the neutrino background sample to 24.5 events . after scaling to the data exposure and accounting for statistical error plus uncertainties arising from our atmospheric neutrino mc including the neutrino oscillation correction ( see sect .
[ sec : errors_bkgrds ] ) , we obtain @xmath76 events as the estimated neutrino - induced background of our @xmath0oscillation search . as discussed in sect .
[ sec : nuandrockevts ] , neutron - induced rock events which are background for @xmath0oscillations are those which elude tagging by the active shield . as described in sect .
[ sec : rockbkgrdindata ] , we estimate that a fraction @xmath60 of data multiprongs are rock events .
thus of the 188 shield - quiet data multiprongs , we estimate @xmath77 to be rock events .
however , among the 375 shield - tagged rock multiprong events , only fifteen pass our four @xmath0selection criteria . thus , assuming that the selection efficiency of shield - quiet rock events is the same as that of shield - tagged rock events , the background of rock events which pass all @xmath0selections is estimated to be @xmath78 events . in summary
, we estimate the number of background events to be @xmath79 plus @xmath80 events .
having observed five candidate @xmath0oscillation events and having calculated background neutrino and cosmic ray processes to contribute 4.5 events , we set a lower limit for the @xmath0oscillation time @xmath12 in iron nuclei .
we follow previous experiments in using our best estimate of the background in determining the limit .
additional uncertainties in the limit due to the uncertainties in the background calculation are discussed in sect .
[ sec : stat_and_syst_errors ] .
we use the approach to confidence level construction as formulated by feldman and cousins . for five candidate events with 4.5 background events the signal limit at 90% cl is @xmath81 events . the oscillation time lower limit
is then calculated using @xmath82 here , @xmath83 neutrons in a kiloton of the soudan 2 detector , @xmath84 kiloton years is the full detector exposure , and @xmath85 = 0.18 is the efficiency for detection of @xmath0oscillations in the full detector mass , calculated using our primary @xmath0simulation ( see sect . [
sec : deteff ] ) . at 90% cl
we obtain @xmath86 years .
equation ( [ equ : bound_neutron_time ] ) shows that the free @xmath0oscillation time can be inferred from oscillation time for neutrons of nucleus @xmath13 . for the suppression factor of iron
we use the value @xmath87 obtained by dover , gal , and richard upon averaging over calculations using different nuclear optical potentials @xcite .
similar or moderately higher values have been reported from other calculations . at 90% cl
we obtain @xmath4 seconds .
oscillation time lower limits based upon detection efficiencies of each of our three @xmath0simulations are summarized in table [ tbl : s2_search_limits ] .
the range of values which results from the variation in @xmath85 illustrates the important role of the intranuclear rescattering treatment .
the alternative simulations serve as estimators of sensitivity to systematic uncertainties arising from modelling of the nuclear environment .
.neutron - antineutron oscillation time lower limits ( 90% cl ) arising from three different treatments of pion intranuclear scattering in annihilations
. simulations of the second and third rows are regarded to be realistic ; the one with more conservative nuclear modelling ( bottom row ) is the basis for the limits for this search . [ cols="<,^,^ " , ] a degree of caution is warranted when comparing oscillation time lower limits for free neutrons as calculated by underground experiments versus a reactor neutron beam experiment ( last column of table [ tbl : experiments_and_results ] ) .
the limits of soudan 2 , frjus , and kamiokande are based upon values for the neutron suppression factor as calculated by dover _
_ @xcite . however if values for iron and oxygen are considered which are at the high end of ranges obtained by alberico _
@xcite , then limits at or below the grenoble value are implied for the underground experiments .
the fine - grained soudan 2 tracking calorimeter has been used in a search for neutron - to - antineutron oscillations occurring with bound neutrons . based upon a fiducial exposure of 5.56 kiloton - years of the underground detector ,
a new oscillation time lower limit for @xmath0oscillations in iron nuclei has been determined at 90% cl : @xmath88 assuming the suppression factor for iron is @xmath89 @xcite , the corresponding limit at 90% cl for n to @xmath90oscillations of free neutrons is @xmath91 the search reported here is background - limited .
candidate events are observed to occur at a rate similar to that predicted for backgrounds .
since the predominant background arises from kinematic overlap with multiprong reactions initiated by atmospheric neutrinos , it seems unlikely that future underground experiments can avoid also becoming background - limited in larger exposures .
thus reactor neutron beam experiments , rather than underground experiments monitoring bound neutrons , may offer a more promising route for future improvements in sensitivity to ( see @xcite and references therein ) .
this work was supported by the u.s .
department of energy , the u.k . particle physics and astronomy research council , and the state and university of minnesota .
we gratefully acknowledge the minnesota department of natural resources for allowing us to use the facilities of the soudan underground mine state park .
mohapatra and r.e .
marshak , phys .
lett . * 44 * , 1316 ( 1980 ) ; phys . lett . *
94b * , 183 ( 1980 ) ; l .-
n . chang and n .- p .
chang , phys .
lett . * 92b * , 103 ( 1980 ) ; t.k . kuo and s. love ,
* 45 * , 93 ( 1980 ) ; r. cowsik and s. nussinov , phys . lett . *
101b * , 237 ( 1981 ) ; s. rao and r. shrock , phys . lett . *
116b * , 238 ( 1982 ) ; nucl . phys . * b232 * , 143 ( 1984 ) .
w. m. alberico , _ in : _ proceedings of the workshop on future prospects of baryon instability search in p - decay and n - n oscillation experiments , ed .
s.j . ball and y.
a. kamyshkov , ornl-6910 , march 1996 ; pp .
221 - 234 .
g. backenstoss _
et al . _ ,
b * 228 * , 424 ( 1983 ) . a convenient compilation of channel rates based on this work appears in yu.a .
golubkov and m.yu .
khlopov , astro - ph/0005419 ; submitted to yad .
n. horwitz _
et al . _ ,
115 * , 472 ( 1959 ) ; r. armenteros and b. french , _ in : _ high energy physics , edited by e.h.s .
burhop ( academic , new york , 1969 ) , vol . 4 , p. 237 ; c. ghesquiere , _ in : _ proceedings of the symposium on antinucleon - nucleon interactions , liblice - prague , edited by l. montanet , cern 74 - 18 , ( 1974 ) , p. 436 ; p.
pavlopoulous _ et al .
_ , _ in : _ nucleon - nucleon interactions 1977 , proc .
of the second int .
vancouver , edited by h. fearing , d. measday , and a. strathdee ( aip conf .
41 ) , ( aip , new york , 1978 ) , p. 340 .
kamyshkov , nucleon instability and ( @xmath92 ) non - conservation " , _ in : _ proc .
nnn99 workshop , aip conf .
533 , edited by m.v .
diwan and c.k .
jung , stony brook , new york , september 1999 ; p. 84 | we have searched for neutron - antineutron oscillations using the 5.56 fiducial kiloton - year exposure of the soudan 2 iron tracking calorimeter .
we require candidate @xmath0occurrences to have @xmath1 4 prongs ( tracks and showers ) and to have kinematics compatible with annihilation within a nucleus .
we observe five candidate events , with an estimated background from atmospheric neutrino and cosmic ray induced events of 4.5 @xmath2 1.2 events .
previous experiments with smaller exposures observed no candidates , with estimated background rates similar to this experiment .
we set a lifetime lower limit at 90% cl for the @xmath0oscillation time in iron : @xmath3 years .
the corresponding lower limit for oscillation of free neutrons is @xmath4 seconds .
pacs numbers : 11.30.fs , 12.20.fv , 12.60.jv , 14.20.dh 6.5 in 8.25 in 0.0 in 0.0 in |
the existence of @xmath0-wave superconductivity in the one - band two - dimensional hubbard model has been established through a variety of theoretical methods at both weak @xcite and strong coupling @xcite , in other words for one - site interaction @xmath2
either much smaller or much larger than the bandwidth @xmath10 . generalizations of dynamical mean - field theory are particularly suited for the strong coupling limit , but they are also an excellent guide to the physics at weak to intermediate coupling .
@xcite , these calculations suggest that pairing is maximized at intermediate coupling , where the on - site interaction @xmath2 is of order the bandwidth @xmath3 .
some non - perturbative calculations based on weak coupling ideas even agree at intermediate coupling @xcite with strong - coupling based approaches . in all these approaches , spin fluctuations with either an antiferromagnetic or a singlet character @xcite
have been argued to drive the pairing .
these spin fluctuations result from the presence of an on - site coulomb repulsion @xmath2 . at strong coupling ,
the characteristic energy scale of these fluctuations , the exchange interaction @xmath6 , is given by @xmath11 , and the @xmath0-wave gap symmetry adopted by the cooper pairs allows them to avoid the direct effect of the on - site repulsion @xmath2 .
little attention has been paid so far to the effect of the nearest - neighbor coulomb repulsion ( or extended hubbard interaction ) @xmath5 that we expect to be detrimental to @xmath0-wave superconductivity .
roughly speaking , in a simple bcs picture that does not take retardation into account , we expect the effective interaction to be the difference @xmath12 . in ordinary phonon - mediated superconductivity ,
the repulsion @xmath5 is replaced by a smaller pseudopotential @xmath13 to account for the fact that binding occurs at low frequencies through phonons while the coulomb interaction acts over a broad energy scale .
this so - called anderson - morel mechanism @xcite , leads to the following estimate for the coulomb pseudopotential @xmath14 where the debye frequency is @xmath15 and the fermi energy @xmath16 .
one expects that in strongly - correlated superconductivity , the ratio @xmath17 must be replaced by a number closer to unity in which case this mechanism would no - longer be effective and superconductivity should disappear as soon as @xmath18 .
this issue is crucial to understand high - temperature superconductors since that condition , or the weaker condition @xmath19 with @xmath20 the spin gap , is likely to be satisfied in these materials . from the value of the near - neighbor coulomb interaction with a relative dielectric constant of order 10 we estimate @xmath21 mev while @xmath22 mev @xcite .
so far , it has been shown using a variational wave - function approach for the @xmath23 model at strong coupling that superconductivity persists as long as @xcite @xmath24 where @xmath25 is the doping .
a large - n calculation gives superconductivity at least up to @xmath26 , @xcite while density matrix renormalization group ( dmrg ) calculations on hubbard or @xmath27 ladders @xcite suggest that pairing can survive up to @xmath28 .
@xcite at very small coupling ( @xmath29 ) , it has been argued @xcite that pairing is destroyed as soon as @xmath30 .
this weak coupling bound is close to the result of a flex calculation .
@xcite here we show that @xmath0-wave superconductivity in the one - band two - dimensional hubbard model at strong coupling is in fact more robust than expected . even for @xmath7 , as long as the inequality @xmath31 is satisfied , superconductivity persists
this illustrates differences between pairing at weak and at strong coupling , especially in the pseudogap regime .
the resilience of @xmath0-wave superconductivity to @xmath5 can be traced to the increase in the effective @xmath6 caused by @xmath5 at strong coupling , i.e. when at half - filling the system is a mott insulator .
this increase in @xmath6 is visible in the dynamics of both the spin susceptibility and the cooper pair green s function which is enhanced by @xmath5 at low frequency .
the pair - breaking effect of @xmath5 manifests itself at higher frequency .
this leads overall to a sizeable range of values of @xmath5 where the order parameter is essentially independent of @xmath5 in the underdoped regime where a pseudogap appears .
recall that at strong coupling , the pseudogap extends up to optimal doping , a sign that mott physics extends well beyond half - filling @xcite .
we use cellular - dynamical mean - field theory that allows us to study the strong - coupling limit and allows one to take into account both @xmath6 and @xmath5 as well as the effect of retardation .
we first present the method and model and conclude with a discussion after the presentation of the results obtained from large scale numerical calculations .
we start from the one - band extended hubbard model on a square lattice , @xmath32 where @xmath33 is the destruction ( creation ) operator for an electron of spin @xmath34 at site @xmath35 and @xmath36 is the corresponding number operator ( @xmath37 ) .
we assume a band structure close to that of yba@xmath38cu@xmath39o@xmath40 , with nearest - neighbor hopping @xmath41 set to unity , diagonal hopping @xmath42 and third - neighbor hopping @xmath43 , unless otherwise indicated .
this model is solved with a 4 site plaquette cellular dynamical mean field theory ( cdmft ) at @xmath44 using an exact diagonalization solver @xcite that allows us to obtain real - time quantities without analytic continuation .
this approach has been used to reveal the presence of @xmath0-wave superconductivity in the one - band hubbard model,@xcite and to study the pairing dynamics and retardation when @xmath45 .
@xcite the results obtained for @xmath0-wave superconductivity with the plaquette are essentially identical to those obtained with larger clusters .
@xcite as in previous studies of the 2d hubbard model with this approach @xcite , the plaquette is hybridized with a set of 8 bath orbitals , as illustrated in fig .
[ fig : cluster ] , with 6 parameters to be determined through the cdmft self - consistency relation : the hybridizations @xmath46 between the bath and the cluster , the bath energies @xmath47 and the bath @xmath0-wave pairing parameters @xmath48 which , when nonzero , signal the presence of superconductivity and lead to a non - vanishing cluster @xmath0-wave order parameter @xmath49 the order parameter is extracted from the lattice green s function while the doping is measured on the cluster .
the distance function is defined in ref . .
we use cutoff @xmath50 and fictitious inverse temperature @xmath51 . while the extended interaction @xmath5 within the cluster can be treated exactly with an approach like cdmft , the coupling to neighboring clusters requires the hartree approximation .
@xcite more specifically , the model hamiltonian eq .
is modified by replacing the inter - cluster interaction by @xmath52 where @xmath53 denotes nearest - neighbor pairs within the plaquette and @xmath54 nearest - neighbor pairs across plaquettes .
the mean field @xmath55 must be determined self - consistently ; in practice , it is treated like the six bath parameters wihtin the cdmft self - consistency loop and thus both the dynamical mean field ( represented by the bath parameters ) and the static hartree mean field are converged simultaneously . to verify the accuracy of this approach , consider momentarily the half - filled , particle - hole symmetric extended hubbard model with @xmath56 . in that simple case , the hartree field @xmath55
is fixed by particle - hole symmetry .
we find that the dynamical spin susceptibility at the antiferromagnetic wavevector @xmath57 has a dominant peak at a low frequency @xmath58 .
the position of this peak is shown on fig .
[ fig : chi_hf ] as a function of @xmath5 for two values of @xmath2 . in the large @xmath2 limit
, the half - filled extended hubbard model should map to the heisenberg model with a super - exchange parameter @xmath9 . the energy denominator is easily understood by comparing the energy , at @xmath59 , between a configuration where all sites are exactly singly - occupied and another configuration in which one electron has vacated a site in order to doubly occupy a neighboring site .
four @xmath5 bonds are lost , but three are gained . since we expect the frequency @xmath58 to scale like @xmath6 in the large-@xmath2 limit , fig .
[ fig : chi_hf ] confirms that we obtain the correct scaling with @xmath5 .
hence , our approach leads to the correct strong - coupling physics .
to concentrate on the effect of short - range spin fluctuations on superconductivity , we ignore the possibility of long - range antiferromagnetic order predicted by cdmft at low doping , @xcite as well as the possibility of long - range charge density waves in the presence of @xmath5 . at weak coupling charge - density waves
may occur for @xmath60 in two - dimensions .
@xcite we found a similar result at half - filling for strong coupling .
the @xmath0-wave order parameter for ybco hopping parameters is displayed on fig .
[ fig : op ] for hole - doping at @xmath61 , 8 and 16 . in each panel , obtained for a given value of @xmath2 , we display the results for different values of @xmath5 .
the mott transition at half - filling , that separates weak from strong coupling , occurs around @xmath62 .
@xcite as observed before , @xcite in the weak coupling case ( @xmath61 ) superconductivity is strongest at half - filling when competition with antiferromagnetism is prohibited ( fig .
[ fig : op]a ) . at larger coupling ,
the mott transition destroys superconductivity at half - filling ( figs .
[ fig : op]b , c ) .
the value of @xmath5 influences the order parameter in strikingly different ways at weak and at strong coupling . for @xmath61 ,
superconductivity has essentially disappeared at @xmath63 , in agreement with the upper limit @xmath64 found by weak coupling analysis .
@xcite this is consistent with the fact that at weak coupling , @xmath5 always decreases the strength of spin fluctuations .
@xcite by contrast , at strong coupling , @xmath5 can increase the strength of spin fluctuations through @xmath9 , and for the same ratio @xmath65 , the order parameter is still large for @xmath66 on fig .
[ fig : op]b .
in fact , it is barely influenced by @xmath5 in the pseudogap ( underdoped ) regime close to half - filling ( @xmath67 ) .
on the lower panel , for @xmath68 , we have @xmath69 at @xmath70 and superconductivity still persists .
the ratio @xmath71 is maximum at @xmath72 .
we estimate that @xmath73 is the upper bound for the the appearance of @xmath0-wave superconductivity ( ignoring competing orders ) .
the link between superconducting order parameter and spin fluctuations was demonstrated at @xmath45 by the correlation between the peaks in the imaginary part of the anomalous self - energy and the peaks in the spin fluctuation @xcite .
this correlation persists here ( not shown ) . to further investigate this link between spin fluctuations and superconductivity ,
let us thus first focus on the spin dynamics revealed by the the local spin spectral function @xmath74 illustrated on fig .
[ fig : chi ] .
panel ( a ) shows @xmath74 at @xmath66 , @xmath75 for the underdoped ( @xmath67 ) and overdoped ( @xmath76 ) regimes .
the corresponding charge susceptibility is negligible on this scale .
spin fluctuations are much smaller in the overdoped regime and their weight is spread in wavevector contrary to the underdoped case .
panel ( b ) shows that in the underdoped regime ( @xmath67 ) where there is a pseudogap , a moderate extended interaction @xmath5 at strong coupling mainly shifts the spectrum of @xmath77 spin fluctuations to higher frequencies , as expected from the increase of @xmath9 caused by @xmath5 ; it decreases somewhat the overall spectral weight without affecting the spectral shape . to show that low - frequency spin fluctuations are apparent in the pair dynamics
, we compare the characteristic frequency @xmath58 of spin fluctuations , given by the position of the dominant peak of @xmath78 , with a characteristic frequency in the pair dynamics , as was done in ref . .
the pair dynamics can be studied through the integral @xmath79 where @xmath80 is the retarded gorkov function ( or anomalous green s function ) defined in imaginary time by @xmath81 with @xmath35 and @xmath82 nearest - neighbors .
the infinite frequency limit of @xmath83 is equal to @xmath84 which in turn is proportional to the @xmath44 @xmath0-wave order parameter @xmath85 .
@xmath86 is useful to estimate the frequencies relevant for binding .
we call @xmath86 the cumulative order parameter . in bcs theory , @xmath87 is a monotonically increasing function of @xmath88 that reaches its asymptotic value at the bcs cutoff frequency @xmath89 .
@xcite in the eliashberg approach that includes retardation as well as the coulomb pseudopotential , @xcite the function overshoots its asymptotic value at frequencies near the main phonon frequencies before decaying to its final value because of pair breaking effects at higher frequencies .
we define the characteristic frequency @xmath90 as the point where @xmath87 reaches half of its asymptotic value . if one imagines that @xmath91 is made of a single peak , then that peak would be located at @xmath90 .
figure [ fig : ig ] illustrates the cumulative order parameter @xmath87 for various values of @xmath5 .
the top two panels are for two values of doping @xmath92 at @xmath66 , and the bottom one for @xmath68 in the underdoped regime .
the asymptotic value is indicated by a horizontal line . following a sharp rise around @xmath93 ,
the function has a maximum and then decreases towards its asymptotic value , as in eliashberg theory .
the link between the characteristic spin - fluctuation frequency @xmath58 and the characteristic frequency of the pair @xmath90 is summarized on fig .
[ fig : u8 ] , which shows the evolution with doping of : ( a ) the position @xmath58 of the main af susceptibility peak , ( b ) its strength , and finally ( c ) the characteristic frequency @xmath90 , for four values of @xmath5 at @xmath66
. clearly , @xmath58 , @xmath94 and @xmath90 all decrease with doping .
this shows that spin fluctuations and pair dynamics are strongly linked . even for @xmath45
however , the increase in @xmath6 as measured by @xmath95 and @xmath90 in the underdoped regime , @xmath96 , does not correspond to an increase in the order parameter .
as can be seen in fig .
[ fig : op ] , for @xmath67 and @xmath76 the @xmath0-wave order parameter is essentially identical when @xmath45 and the increase in @xmath90 near half - filling manifests itself only in the maximum value of the cumulative order parameter @xmath87 in fig .
[ fig : ig ] , not in the value of the order parameter .
clearly the order parameter does not increase in the underdoped regime despite the increase in the characteristic spin fluctuation frequency @xmath95 because of another effect .
that effect is the pseudogap .
the pseudogap removes many of the states near the fermi energy that would otherwise be paired .
the presence of the pseudogap can be seen by contrasting the single - particle local density of states at two dopings . in the overdoped regime ( @xmath76 )
the local density of states has no pseudogap in the normal state , as seen from the dashed red line on fig .
[ fig : dos]b .
only the superconducting gap is visible .
by contrast , even at @xmath45 there is clearly a pseudogap in the underdoped regime , as can be seen from fig .
[ fig : dos]a .
the effect of superconductivity manifests itself only through the small more symmetric gap near @xmath97 .
@xcite the pseudogap , as measured from the peak to peak distance , increases slightly with @xmath5 in the underdoped regime .
the detrimental effect of the pseudogap on the superconducting order parameter was demonstrated in fig . 7 of ref . where the bubble contribution to the pairing susceptibility decreases as we approach half - filling .
this means that in the absence of interactions between the particles forming the pair ( represented by vertex corrections ) , self - energy effects disfavor superconductivity .
this physics is also present at weak to intermediate coupling when the pseudogap is induced by long wavelength antiferromagnetic fluctuations @xcite ( and not by the incipient short - range mott localisation that appears at strong coupling @xcite ) .
the effect of spin fluctuations and of near - neighbor repulsion @xmath5 on pairing is relatively straightforward in the overdoped regime ( @xmath98 ) .
there , as seen on fig .
[ fig : op]b and figs .
[ fig : u8]a , c , the order parameter increases with the characteristic spin frequency @xmath58 and the characteristic pair frequency @xmath90 . like for the @xmath45 case
, superconductivity disappears at finite @xmath5 when the amplitude @xmath94 of the lowest frequency peak in the spin fluctuations ( fig .
[ fig : u8]b ) vanishes , @xcite as observed in experiment .
@xcite the main effect of @xmath5 is to decrease the spin fluctuations , as happens systematically at weak coupling .
@xcite as we approach half - filling , the decrease of the order parameter for all values of @xmath5 ( figs .
[ fig : op]b , c ) , despite the increase in the amplitude @xmath94 of the lowest peak in the spin fluctuations ( fig .
[ fig : u8 ] ) , is mainly due to the increase of the pseudogap : it removes more and more states from the fermi energy as the doping @xmath92 decreases ( fig .
[ fig : dos ] ) .
note that @xmath5 causes opposite changes in the amplitude @xmath94 and in the characteristic frequencies ( figs [ fig : u8 ] ) , which might explain the near independence of the order parameter with respect to @xmath5 in figs .
[ fig : op]b , c . however , these changes are small on a relative scale and are probably not the main reason for the insensitivity of the order parameter to @xmath5 .
the resilience of @xmath0-wave superconductivity to @xmath5 at strong coupling is best understood from the @xmath68 results for the cumulative order parameter @xmath86 shown in fig .
[ fig : ig]c for @xmath67 .
the largest value of @xmath86 scales roughly like @xmath9 , in other words @xmath5 increases the binding at low frequency , where the retardation is large .
however , @xmath5 also has the expected pair - breaking effect : larger @xmath5 causes larger pair - breaking so that the low frequency increase in @xmath87 is essentially compensated by the time @xmath87 reaches its asymptotic value ( the order parameter ) . to understand more deeply the dual role of @xmath5 , as both pair binding and pair breaking , it is helpful to return to the solution of the eliashberg equations in the electron - phonon case . figs . 7 , and 13 of ref .
show that the imaginary part of the gap function remains positive for all frequencies when the coulomb pseudopotential vanishes .
by contrast , one verifies from fig .
11 of the same paper that the imaginary part of the gap function changes from positive to negative as frequency increases when the coulomb pseudopotential is finite .
a sign change in the imaginary part of the gap function should lead to a similar behavior in the imaginary part of the gorkov function @xmath99 . in this case
its integral @xmath87 , as defined in eq .
( [ f(w ) ] ) , should reach a maximum before decreasing to its asymptotic large frequency value .
by contrast , when the imaginary part of the gap function remains positive at all frequencies , i.e. in the absence of coulomb pseudopotential , @xmath87 should increase monotonically to its asymptotic value , as in bcs theory . since we observe that @xmath87 has a maximum at finite frequency , we are in the case where the coulomb pseudopotential is important .
comparison of figs . 11 and 13 of ref .
also shows that the maximum value of the imaginary part of the gap function ( and that of its frequency integral , as one can check ) increases with the electron - phonon coupling constant . in our case
, the increase in @xmath9 caused by @xmath5 similarly leads to an increase in the maximum value of the cumulative order parameter @xmath87 .
in other words , we can surmise that the increase in @xmath5 leads to larger binding at low frequency , like an increase in the electron - phonon coupling constant , but it also leads to more pair - breaking at large frequency , coming from the coulomb pseudopotential - like effect of @xmath5 , and these two effects nearly cancel each other . in the electron - phonon case
, one also notes that the imaginary part of the gap function vanishes at about two to three times the maximum phonon frequency .
this is where the cumulative order parameter @xmath87 would reach its asymptotic value . in the case of spin fluctuations , @xmath100 is a measure of the width of the spin - fluctuation spectrum . comparing fig .
[ fig : ig]a for @xmath66 and fig .
[ fig : ig]c for @xmath68 , one notices that the frequency for which @xmath87 reaches its asymptotic value is smaller by a factor of roughly two for @xmath68 where the @xmath45 value of @xmath6 is twice as small as for @xmath66 .
the opposing effects discussed above conspire to leave the order parameter rather insensitive to @xmath5 in the presence of a pseudogap .
the pseudogap induced by mott physics @xcite is detrimental to superconductivity , but in its presence superconductivity is effectively protected from near - neighbour repulsion @xmath5 .
indeed , in a doped mott insulator , short - range incipient localisation is strong enough to create a pseudogap , but while @xmath5 causes pair - breaking at high frequency , it also enhances spin fluctuations ( @xmath9 ) at low frequencies , thus compensating the pair - breaking effect .
overall , retardation effects are crucial for the resilience of @xmath0-wave superconductivity to near - neighbor repulsion .
we are indebted to d.j .
scalapino , a. chubukov , j. carbotte and g. kotliar for useful discussions and to giovanni sordi for a careful reading of the manuscript .
this work was supported by nserc ( canada ) , cfi ( canada ) , cifar , and the tier i canada research chair program ( a .- m.s.t . ) .
computational facilities were provided by compute canada and calcul qubec .
d. snchal , cluster dynamical mean field theory , in _ theoretical methods for strongly correlated systems _ , edited by a. avella and f. mancini , volume 171 of _ springer series in solid - state sciences _ , chap .
341370 , springer , 2012 . | many theoretical approaches find @xmath0-wave superconductivity in the prototypical one - band hubbard model for high - temperature superconductors . at strong - coupling ( @xmath1 , where @xmath2 is the on - site repulsion and @xmath3 the bandwidth ) pairing
is controlled by the exchange energy @xmath4 .
one may then surmise , ignoring retardation effects , that near - neighbor coulomb repulsion @xmath5 will destroy superconductivity when it becomes larger than @xmath6 , a condition that is easily satisfied in cuprates for example . using cellular dynamical mean - field theory with an exact diagonalization solver for the extended hubbard model ,
we show that pairing _ at strong coupling _ is preserved , even when @xmath7 , as long as @xmath8 .
while at weak coupling @xmath5 always reduces the spin fluctuations and hence @xmath0-wave pairing , at strong coupling , in the underdoped regime , the increase of @xmath9 caused by @xmath5 increases binding at low frequency while the pair - breaking effect of @xmath5 is pushed to high frequency .
these two effects compensate in the underdoped regime , in the presence of a pseudogap . while the pseudogap competes with superconductivity , the proximity to the mott transition that leads to the pseudogap , and retardation effects , protect @xmath0-wave superconductivity from @xmath5 . |
to model magnetization dynamics , currently two paradigms are commonly used in the field : the micromagnetic model and the heisenberg spin model .
the micromagnetic model @xcite is ideal when simulating systems with linear dimensions of the order of a few nanometers or larger ; since it is a continuous model that is discretized for computational application , its reliability decreases dramatically when simulating magnetic structures exhibiting a large gradient that can not be resolved by the finite size cells . a textbook example for this scenario
is offered by bloch points @xcite ( see fig . [
fig : bloch ] ) , domain walls and spin waves also belong to this category for particular values of the material parameters .
the heisenberg model @xcite is a discrete description , where with every atom in the lattice of the ferromagnet a magnetic moment is associated . since this is a discrete model , its capability to simulate any magnetic structure is not limited by computational artifacts originating from the discretization of a continuum model , which makes it distinct from micromagnetism . on the other hand
, the heisenberg model can not be efficiently used to simulate systems larger than a few nanometers due to the computational time increasing faster than linearly with the number of atoms .
@xcite in the presented approach ( fig .
[ fig : geom ] ) , the entire system is simulated using the micromagnetic model while one or more regions of it containing large gradient structures ( e.g. bloch points ) , are simulated using the discrete heisenberg model .
the main obstacle for the development of a combined multiscale technique consists of devising accurate conditions to make the interface between regions on two different scales magnetically smooth , in order to prevent any interface related artifacts . while in magnetization dynamics , adaptive mesh refinement techniques
@xcite have been used , none of these employed different models for different scales .
one related approach has been proposed , addressing the problem of interfaces between layers of different magnetic materials @xcite . however , the lack of proper interface conditions , in particular the choice of applying a coarse scaled exchange field on the magnetic moments along the interface in the fine scale region , restricts the validity of this approach to the systems with uniform magnetization across the interface . while this shortcoming has been later resolved in refs .
, these approaches were devised to evaluate equilibrium configurations rather than simulating dynamical systems .
one further related approach @xcite employed the finite elements method .
it should be noted however that while in this case the atomic lattice in the heisenberg model can be rendered more accurately , the computational times can not be dramatically reduced as shown for our finite differences approach in @xcite , making this approach considerably slower .
one further multiscale approach @xcite , devised for a different scale combination than the presented one , proposed to use the micromagnetic model as the fine scale model and the maxwell equations as the coarse scale model , this is however restricted to systems with slowly varying magnetization . another work @xcite uses special relativity to evaluate a corrective term to the landau - lifshitz - gilbert equation in the case of domain wall motion . in continuum
mechanics @xcite , multiscale approaches are commonly applied to the investigation of mechanical properties of materials , such as their response to deformations and fractures .
however , so far it is unclear whether one can develop such a multiscale model for magnetization dynamics that allows one to carry out valid simulations of systems that can not be modeled with the currently available approaches . in this paper
we show the details of the multiscale approach , with a particular focus on the interface conditions that we developed to obtain a smooth interaction between regions on different scales .
finally , demonstrations of the validity for the approach are shown , demonstrating the transmission of spin waves across the scale interface without attenuation , and comparing the simulated ground state for structures exhibiting dzyaloshinskii - moriya interaction to the analytical theory . ]
.[fig : geom ] ]
the multiscale approach solves the landau - lifshitz - gilbert equation numerically for two different models : the coarse grained micromagnetic model , which simulates the whole sample ; and the fine scale model , which is used for magnetic structures that can not be accurately described by the micromagnetic model , discretizing the magnetization field at atomic resolution and simulating it in the intrinsically discrete heisenberg spin model .
our software executes in parallel two independent solving routines , one for each model ( it is in principle possible to execute any number of fine scale solving routines ) , performing one full computational step on the coarse scale one and then a short series of steps on the fine scale one centered around the time coordinate of the coarse one ( see fig .
[ fig : time_coord ] ) . , is executed , then corrections to the coarse scale effective field ( generated by the heisenberg fine scale one ) are evaluated .
[ fig : time_coord ] ] the main task towards the development of this technique consisted in modeling the interaction between different regions .
this was achieved by applying , after each coarse scale step , a set of magnetic fields designed to approximate the effect of the non - local terms of the effective magnetic field from one region on the other , see fig . [
fig : boundary ] , namely exchange and stray field .
these magnetic fields are designed as follows : the exchange field , generated by the fine scale magnetic moments closest to the interface ( interfacial moments ) , on their neighboring cells in the coarse scale ( interfacial cells ) is evaluated by averaging all the interfacial moments inside each coarse scale cell .
the average vector is rescaled by the volume @xmath0 of a cell in the atomic lattice , in order to obtain the magnetization @xmath1 , rather than the magnetic moment @xmath2 .
a new finite difference mesh , with coarse scale discretization is created and the cells corresponding to the internal surface of the fine scale region are filled with the difference between the magnetization of the same cell in the original coarse mesh and the new vectors . in this way , the linearity of the exchange field with respect to the magnetization is exploited to evaluate a correction to the field , calculated in the micromagnetic formulation , generated by the original coarse scale cells alone .
the corrected exchange field , exerted by the multiscale cell @xmath3 on the micromagnetic cell @xmath4 is calculated as : @xmath5 here @xmath6 denotes the magnetization in the cell @xmath3 in the purely micromagnetic simulation , while @xmath7 is defined as : @xmath8 where the sum runs over all the magnetic moments @xmath9 located along the interface on the side of cell @xmath3 that is neighboring cell @xmath4 .
this effective field term is evaluated in the micromagnetic model .
likewise , to evaluate the exchange field generated by interfacial cells on interfacial moments , interpolation is employed in order to define a set of new magnetic moments ( ghost moments @xcite ) to act as first neighbors to the interfacial ones . the exchange field generated by the ghost moments
is evaluated in the heisenberg spin model .
a combination of fine scale moments and coarse scale magnetization is used in the interpolation in order to ensure a smooth transition in the magnetic pattern across the interface .
this means that each ghost moment results from the interpolation of atomistic and aptly renormalized micromagnetic vectors .
the interpolation can be linear , bilinear or quadrilinear according to the dimensionality of the coarse scale mesh .
the same techniques , based on the average of interfacial magnetic moments , and the calculation of ghost moments through interpolation across the interface , are employed when evaluating antisymmetric exchange ( dzyaloshinskii - moriya interaction ) across the scale interface . ]
the stray field contains all the long range contributions to the effective fields .
the implementation of this field constitutes one of the main differences between the two models . in both scales
the demagnetization tensor formulation was employed,@xcite as well as the calculation method based on fft for efficient calculation.@xcite while for the coarse cells the demagnetization tensor describes the interaction between two uniformly magnetized solid rectangles , according to the calculations carried on by newell _
et al . _ , @xcite the demagnetization tensor used for magnetic moments in the fine scale ,
is defined as : @xmath10,\ ] ] where @xmath11 and @xmath12 are the position of two magnetic moments , @xmath13 is 3 @xmath14 3 identity matrix , and symbol @xmath15 denotes the tensor product .
similarly to the exchange field , the stray field is linear in the magnetization vector and this property is exploited likewise .
the correction to the stray field generated in the micromagnetic system by fine scale regions is evaluated using the averaged value of magnetic moments in each cell . in order to evaluate the complete demagnetization field acting on the fine scale system , the coarse scale magnetization structure
is copied into a new mesh and the cells corresponding to the fine scale region are filled with zero vectors . the stray field generated by this system
is evaluated .
this technique is employed in order for the field generated by the fine scale region on itself not to be evaluated twice .
since the field has the same discretization as the structure generating it , the result is then interpolated , in order for it to have the discretization of the fine scale mesh .
the type of linear interpolation depends , as for the ghost moments , on the dimensionality of the mesh .
this is the only case for an effective field term evaluated micromagnetically to be applied on the fine scale region .
this approximation is made necessary by the computational complexity of the algorithm calculating the field , increasing with @xmath16 where @xmath17 is the number of cells .
this dependence is due to the method employed for calculating the demagnetization tensor based on fast fourier transform ( fft ) . @xcite
having implemented the approach , we run a series of tests as a demonstration of the validity of our model .
the simulated system was a one - dimensional nanowire , @xmath18 long with a square @xmath19 nm@xmath20 section .
the fine scale domain was @xmath21 long ( fig .
[ fig : wire ] ) , the material parameters for this system are those commonly used for permalloy , namely : @xmath22 a / m , exchange constant @xmath23 j / m , and gilbert damping constant @xmath24 .
@xcite for the purpose of efficiency and due to the constraints of the finite difference method , upon which the original software is based , the crystal in the atomistic region is considered to be ordered in a simple cubic lattice with a lattice constant @xmath25 nm , comparable to the ones of iron and nickel .
is applied to a section of the fine scale region to excite spin waves .
the amplitude of the spin wave is evaluated in the atomistic fine scale cell @xmath3 and the coarse scale cell @xmath4 which is described micromagnetically .
[ fig : wire ] ] spin waves of different frequencies @xmath26 were excited applying an alternating transversal magnetic field , @xmath27 , to a short ( @xmath28 long ) section of the wire .
the magnetization as a function of time was measured on the atomistic moment furthest from the region where @xmath27 is applied ( @xmath29 ) , and on the neighboring micromagnetic cell ( @xmath30 ) , the transversal component of the two arrays was normalized , and then analyzed using fft in order to find @xmath31 and @xmath32 .
peaks with frequency corresponding to the frequency of @xmath27 were easily identifiable .
the height of such peaks increased linearly with the amplitude of @xmath27 .
the peaks , @xmath33 and @xmath34 , were squared and the transmission coefficient @xmath35 across the interface has been evaluated by calculating the ratio between the two : @xmath36 for some values of the frequency , a purely atomistic simulation was performed for comparison and with , the aim of obtaining the relation between frequency and wavelength . using fft in the space domain ,
the corresponding wavenumber @xmath37 was measured for each value of the excitation frequency .
in particular , such fourier transforms were evaluated at different time instants and then averaged .
once again peaks were easily identifiable . by means of linear regression
( see fig .
[ fig : fit ] ) the dependence @xmath38 was measured and the wavelength corresponding to each value of the excitation frequency was calculated as @xmath39 . and the excitation frequency @xmath40 .
[ fig : fit ] ]
three sets of simulations were performed , with different lengths of the micromagnetic cells , corresponding to ten , twenty and thirty times @xmath41 .
the data show ideal transmission for frequency values smaller than a sharply defined cut - off frequency . the same data , as a function of the wavelength , show consistently that the transmission drops to zero at a cut - off wavelength corresponding to a specific value of the coarse cell size .
this universal behavior can be considered as a limitation of computational micromagnetism , which does not allow one to simulate very short wavelength spin waves without refining the mesh , introducing therefore a dramatic increase in the computation time ( fig .
[ fig:1 ] ) .
( a ) and wavelength @xmath42 ( b ) .
a transmission of 1 ( @xmath43 ) for a wide range of wavelengths demonstrates the numerical validity of the model .
[ fig:1],title="fig : " ] ( a ) and wavelength @xmath42 ( b ) .
a transmission of 1 ( @xmath43 ) for a wide range of wavelengths demonstrates the numerical validity of the model .
[ fig:1],title="fig : " ] since we assume that the frequency cut - off is a consequence of the coarse scale not being able to resolve waves with such a high frequency , we simulated a similar system , this time with the excitation being applied on the coarse scale region only . here
the waves propagate into and then out of the fine scale region and the transmission is measured for waves leaving the fine scale region ( fig . [ fig:2 ] ) .
the test was repeated using periodic boundary conditions to make sure that the sharp cut - off was not caused by the waves being reflected at the end of the wire .
both tests were then repeated for different values of the exchange constant . .
the data shown is the result of an average on all the frequencies .
peaks with frequency higher than @xmath44 were not visible in the fourier transform , underlining the fact that the cutoff is a consequence of the waves not being resolved for the chosen cell - size .
the observed transmission of approximately @xmath45 shows the validity of the method with no artificial attenuation at the interface between the regions where different models are used .
[ fig:2 ] ] in order to measure the cut - off frequencies , a linear regression was executed on all the transmission values between @xmath46 and @xmath47 , the intersection of this line with the transmission value of @xmath48 was defined as the cut - off frequency .
we assume the cut - off to be a direct consequence of the exchange interaction not being accurately evaluated in the micromagnetic model when the angle in the magnetization between two neighboring cells is too large .
the dependence of the cut - off frequency on the exchange constant supports this hypothesis ( see fig .
[ fig:3 ] ) . for waves excited in the fine scale region with closed and periodic boundary conditions ( bc ) for different values of the exchange constant @xmath49 . a cut - off frequency depending on the exchange constant demonstrates that this phenomenon is strictly micromagnetic and is not introduced by the multi - scale approach .
[ fig:3 ] ]
to demonstrate the reliability of the method used to evaluate effective fields across the interface by direct comparison to analytical theory , a system exhibiting antisymmetric exchange@xcite was simulated .
a nanowire , similar in shape to the one used to test spin wave transmission , with the parameters @xmath50 a / m , exchange constant @xmath51 j / m .
different values of @xmath52 were used .
the vector @xmath53 scales the energy density of the dzyaloshinskii - moriya interaction ( dmi ) as calculated in ref .
[ ] : @xmath54 the system was relaxed in a coarse scale simulation , then a fine scale region was applied on a section of the wire and the system was relaxed again . the relaxed state ( see fig . [
fig : dmi ] ) showing continuity in the helix structure , typical of systems exhibiting dmi , with a pitch in agreement with the predicted@xcite value of @xmath55 .
the pitch was evaluated from the fourier transform in the space domain for the two components of the helix , using the data points from both scales and taking the peak value from the fourier transform .
the components of @xmath56 evidently have a perfectly sinusoidal shape , see fig .
[ fig : dmi ] ( a ) .
plane , showing continuity and consistency of the period in the coarse and fine scale .
the dashed lines show the position of the fine scale region .
b ) the wavenumber of the helix increases linearly with the dmi constant and is consistent with the expected value . @xcite
[ fig : dmi],title="fig : " ] plane , showing continuity and consistency of the period in the coarse and fine scale .
the dashed lines show the position of the fine scale region .
b ) the wavenumber of the helix increases linearly with the dmi constant and is consistent with the expected value .
@xcite [ fig : dmi],title="fig : " ]
a tracking algorithm was devised in order to keep the fine scale region as small as possible , it scans the fine scale region for the position of the structure of interest ( soi ) , usually the spin structure with large magnetization gradients , and shifts the fine scale region by an integer number of coarse scale cells units , in order to always have the soi close to its center .
when micromagnetic cells previously not part of the fine scale region become included , interpolation is applied in order to fill in the fine scale mesh with magnetic moments that accurately reproduce the coarse scale magnetization and are continuous within and across the scale interface . to show that the fine scale area can be reliably moved , a test was performed .
this test simulated domain wall motion in a nanostrip ( @xmath57 ) induced by a unidirectional magnetic field .
the material parameters of the strip are the same as the nanowire from the previous test with the only exception of gilbert damping @xmath58 .
the domain wall is initially in the center of the fine scale region , when the distance from the starting position becomes larger than a certain threshold ( tracking distance ) , the whole fine scale region is shifted , in order to keep it centered .
the test was repeated for different tracking distances to show that this process does not influence the dynamics of the system ( fig .
[ fig:4 ] ) . ]
we have presented an innovative methodology to perform magnetization dynamics simulations in the systems which can not be accurately simulated otherwise . since some of such systems describe phenomena , including vortex cores switching @xcite and skyrmion nucleation , @xcite
are considered to be important problems in spintronics , we deem this methodology a key step to advance this field . in order to improve the technique and establish multiscale simulations as a valuable tool , its basic features have been described and its limits have been tested .
the transmission data for the spin waves shows that information about magnetic structures in the fine region can cross perfectly the scale interface , thus demonstrating the reliability and numerical validity of our model . a thorough analysis of the cut - off phenomenon found for spin wave transmission shows that in the presence of spin waves with a short wavelength the multiscale approach can be reliably used under the condition that the waves do not leave the fine scale region .
meanwhile , the traditional approach a refinement of the whole mesh would increase the computational time dramatically .
the simulations including the dmi further show that the method employed for evaluating cross scale interactions ensures continuity between the regions of different scales and yield quantitative agreement with the analytical theory .
moreover , the domain wall data indicates the reliability of the tracking algorithm and its effectiveness as a method to keep the size of the fine scale regions at a minimum and not introducing artifacts to the simulated results . as a future direction we propose to analyze the dynamics of magnetic vortex core reversal , a phenomenon that requires a similar approach in order
to be accurately simulated .
@xcite further research will include magnetic structures such as skyrmions which are stabilized by dmi and where the nucleation involves bloch points .
in the long term , there is room for further improvements : generalizing the approach beyond simple cubic lattices in the fine scale region , optimization of the computational routines , extension of this approach to antiferromagnets and nonzero temperatures are some of the examples that will broaden the applicability even further .
a.d.l . is a recipient of a scholarship through the excellence initiative by the graduate school materials science in mainz ( gsc 266 ) , b.k .
is the recipient of the carl zeiss postdoc scholarship multiskalensimulationen fr energiesparende magnetisierungsmanipulation .
the authors acknowledge the support of spinnet ( daad spintronics network , project number 56268455 ) and the dfg ( sfb trr 173 spin+x ) .
o.a.t . acknowledges support by the grants - in - aid for scientific research ( grants no .
25800184 , no .
25247056 and no .
15h01009 ) from mext , japan .
we are thankful to u. nowak and d. hinzke from the university of konstanz for their help with testing the fine scale model . 1 a. aharoni , j. phys . colloq . * 32 * , 966 - 971 ( 1971 ) .
a. thiaville , j.m .
garcia , r. dittrich , j. miltat , t. schrefl , phys .
b * 67 * , 094410 ( 2003 ) .
a. polyakov , phys .
b * 59 * , 79 - 81 ( 1975 ) .
d. hinzke and u. nowak , phys .
sol . ( a ) * 189 * , 475 - 480 ( 2002 ) . c. schieback , m. klui , u. nowak , u. rdiger , p. nielaba , eur .
j. b * 59 * , 429 - 433 ( 2007 ) . c. abert , l. exl , g. selke , a. drews , t. schrefl , ieee trans .
magn . * 48 * , 1105 - 1109 ( 2012 ) .
b. krger , g. selke , a. drews , d. pfannkuche , ieee trans . magn . * 49 * , 4749 - 4755 ( 2013 ) .
garca - cervera and a.m. roma , ieee trans .
magn . * 42 * , 1648 - 1654 ( 2006 ) .
tako , t. schrefl , m.a .
wongsam , r.w .
chantrell , j. appl . phys . * 81 * , 40824083 ( 1997 ) . f. garcia - sanchez , o. chubykalo - fesenko , o. mryasov , r.w .
chantrell , k.y .
guslienko , appl .
* 87 * , 122501 ( 2005 ) .
f. garcia - sanchez , o. chubykalo - fesenko , o. mryasov , r.w .
chantrell , physica b : condens .
matter * 372 * , 328 ( 2006 ) .
f. garcia - sanchez , o. chubykalo - fesenko , o. mryasov , r.w .
chantrell , k.y .
guslienko , j. appl . phys . * 97 * , 10j101 ( 2005 ) . t. jourdan , a. marty , f. lancon , phys .
b * 77 * , 224428 ( 2008 ) . t. jourdan , a. masseboeuf , a. marty , f. lancon , p. bayle - guillemaud , j. appl . phys . * 106 * , 073913 ( 2009 ) . c. andreas , a. kakay , r. hertel , phys . rev .
b * 89 * , 134403 ( 2014 ) .
f. bruckner , m. feischl , t. fhrer , p. goldenits , m. page , d. praetorius , m. ruggeri , d. suess , math .
models methods appl .
* 24 * , 2627 ( 2014 ) .
p. weinberger , e.y .
vedmedenko , r. wieser , r. wiesendanger , philos . mag . * 91 * , 2248 - 2262 ( 2011 ) .
r.e . miller and e.b .
tadmor , model .
* 17 * , 053001 ( 2009 ) . c. hertel , m. schmichen , s. lbig , j. frhlich , j. lang , theor .
fluid . dyn . * 27 * , 817 - 841 ( 2012 ) .
r. hertel , s. gliga , m. fahnle , c.m .
schneider , phys .
98 * , 117201 ( 2007 ) .
elas and a. verga , eur .
j. b * 82 * , 159 ( 2011 ) .
b. krger , doctoral dissertation available at : + ediss.sub.uni-hamburg.de/volltexte/2012/5887/pdf/dissertation.pdf ( 2011 ) .
newell , w. williams , d.j .
dunlop , j. geophys .
98 * , 9551 - 9555 ( 1993 ) .
the micromagnum software is available at : + micromagnum.informatik.uni-hamburg.de/ .
r. hertel and c.m .
schneider , phys .
lett . * 97 * , 177202 ( 2006 ) .
o. a. tretiakov and o. tchernyshyov , phys .
b * 75 * , 012408 ( 2007 ) .
caputo , y. gaididei , f.g .
mertens , d.d .
sheka , phys .
lett . * 98 * , 056604 ( 2007 ) .
guslienko , k.s .
lee , s.k .
kim , phys .
lett . * 100 * , 027203 ( 2008 ) .
j. sampaio , v. cros , s. rohart , a. thiaville , a. fert , nat . nano .
* 8 * , 839844 ( 2013 ) .
x. zhang , m. ezawa , y. zhou , sci .
rep . * 5 * , 9400 ( 2015 ) .
i. dzyaloshinskii , _ a thermodynamic theory of `` weak '' ferromagnetism of antiferromagnetics_. j. phys .
solids * 4 * , 241 - 255 ( 1958 ) t. moriya , _ anisotropic superexchange interaction and weak ferromagnetism_. phys.rev . * 120 * , 91 ( 1960 ) . | simulations of magnetization dynamics in a multiscale environment enable rapid evaluation of the landau - lifshitz - gilbert equation in a mesoscopic sample with nanoscopic accuracy in areas where such accuracy is required .
we have developed a multiscale magnetization dynamics simulation approach that can be applied to large systems with spin structures that vary locally on small length scales . to implement this ,
the conventional micromagnetic simulation framework has been expanded to include a multiscale solving routine .
the software selectively simulates different regions of a ferromagnetic sample according to the spin structures located within in order to employ a suitable discretization and use either a micromagnetic or an atomistic model . to demonstrate the validity of the multiscale approach , we simulate the spin wave transmission across the regions simulated with the two different models and different discretizations .
we find that the interface between the regions is fully transparent for spin waves with frequency lower than a certain threshold set by the coarse scale micromagnetic model with no noticeable attenuation due to the interface between the models . as a comparison to exact analytical theory ,
we show that in a system with dzyaloshinskii - moriya interaction leading to spin spiral , the simulated multiscale result is in good quantitative agreement with the analytical calculation . |
the current - phase relation characterizes the flow of a superfluid / superconductor through a weak link @xcite .
the latter is a constricted flow configuration that can be realized in different ways : i ) apertures in impenetrable walls mostly for helium , ii ) sandwich or bridge structures for superconductors , and iii ) penetrable barriers created by laser beams for ultracold dilute gases .
much information about such systems can be extracted from the current - phase relation , which , given a fluid , depends only on the link properties .
for instance , with @xmath3he , the transition from the usual ac josephson effect to a quantized phase slippage regime @xcite corresponds to the switching from a sine - like current phase relation to a multivalued one @xcite .
a weak link configuration can be modelled very generally upon taking a portion of a superfluid / superconductor to have `` different conduction properties '' with respect to the rest of the system .
two pieces of superconductor joined by a third superconducting region with a smaller coherence length provide one example , whose current - phase relation in one dimension has been studied with the ginzburg - landau equation @xcite . in the context of ultracold dilute gases , raising a repulsive penetrable barrier across the flow yields an equivalent configuration . for instance , with bose - einstein condensates ( bec ) , josephson effect(s ) have been theoretically studied @xcite and experimentally demonstrated using multiple well traps @xcite .
theoretically , the current - phase relation has been studied for a flow through a repulsive square well with fermions across the bcs - bec crossover by means of one dimensional bogoliubov - de gennes equations @xcite , for weak barriers with bosons in a local density approximation @xcite , and for fermions on the bec side of the crossover using a nonlinear schrdinger equation approach @xcite . in this manuscript
, we study the current - phase relation for a bec flowing through a repulsive square well
. the weak link configuration , and in turn the current - phase relation , is then determined by the barrier height with respect to the chemical potential and by the barrier width with respect to the healing length .
though we solve a one - dimensional gross - pitaevskii equation , the results presented in this manuscript are not just relevant for becs , but also include the essential features of current - phase relations of superconducting or superfluid he - based weak links when governed by the ginzburg - landau equation . for any barrier width
, we find that in the limit of zero barrier height , the current phase relation tends to @xmath4 , with @xmath5 being the bulk sound velocity , which corresponds to the phase across a grey soliton at rest with respect to the barrier . on the other hand ,
if the barrier height is above the bulk chemical potential at zero current , the limit of tunneling flow is reached either when the barrier height is much bigger than the bulk chemical potential at zero current or when the barrier width is much larger than the bulk healing length . in this regime , we recover the the usual josephson sinusoidal current - phase relation and obtain an analytical expression for the josephson critical current as a function of the weak link parameters . for barriers wider than the healing lenght inside the barrier region
, we observe two families of multivalued ( often called reentrant ) current - phase relations .
the first , already studied since the early works on superconductivity @xcite , shows a positive slope of the current when the phase difference is close to @xmath6 , thereby reaching a phase difference larger than @xmath6 at least for small currents .
the second family , appearing at a smaller barrier height , has instead a negative slope of the current close to @xmath6 , and in some cases can remain within the @xmath7 interval across the whole range of currents .
these two families can also be distinguished by the maximum number of different positive currents corresponding to the same phase difference : two for the first family , three for the second one . as the first kind of reentrant behavior
was proven to be connected to the onset of phase - slippage in the ac josephson effect @xcite , the second might then be connected to the appearance of new features in the josephson dynamics .
we finally observe that the hysteresis characterizing both families of reentrant current - phase relations is always due to the competition between a hydrodynamic component of the flow and a nonlinear - dispersive component , the latter due to the presence of a soliton inside the barrier region .
the two components can coexist only for barriers wide enough to accomodate a soliton inside . in this spirit
, we develop a simple analytical model which describes very well reentrant regimes of current - phase relations .
we consider a dilute repulsive bose - einstein condensate at zero temperature flowing through a 1d rectangular potential barrier .
we look for stationary solutions of the 1d gpe @xcite : @xmath8 where @xmath9 $ ] is the complex order parameter of the condensate , @xmath10 is the chemical potential , and @xmath11 with @xmath12 the atom mass and @xmath13 the @xmath14-wave scattering length .
the order parameter phase @xmath15 is related to the superfluid velocity via @xmath16 .
the piecewise constant external potential describes the rectangular barrier of width @xmath17 and height @xmath18 : @xmath19 we consider solutions of eq .
( [ eq : gp ] ) which are symmetric with respect to the point @xmath20 , therefore discarding cases in which a reflected wave is present @xcite .
such symmetric solutions in the presence of a barrier exist due to the nonlinearity in the gpe .
we also restrict our analysis to subsonic flows @xmath21 , with @xmath22 being the sound velocity for a uniform condensate of density @xmath23 . as boundary conditions , we fix the condensate density @xmath23 and velocity @xmath24 at @xmath25 , thereby determining the chemical potential @xmath26 . using the relation @xmath27 $ ] , eq .
( [ eq : gp ] ) can be split into a continuity equation , stating the constancy in space of the current @xmath28 , and an equation for the density only @xmath29 where we have used the continuity equation @xmath30 .
its solution @xmath31 is expressed in terms of jacobi elliptic functions @xcite . due to symmetry ,
we need only consider half of the space @xmath32 .
the solution outside the barrier @xmath33 becomes @xmath34\ , \label{eq : solutionout}\ ] ] where @xmath35 , @xmath36 , and @xmath37 is the density at the barrier edge @xmath38 .
the solution inside the barrier @xmath39 is : @xmath40}{{\rm cn}^2[b_1 x ,
k_1]}&,&\ \delta\geq 0\text { and } a_1\geq 0\ , \\ n_2(x)&=n_0+a_2\frac{1-{\rm cn}[b_2x , k_2]}{1+{\rm cn}[b_2 x , k_2]}&,&\ \text{else}\ , \end{aligned } \right .
\label{eq : solutions}\ ] ] where @xmath41 is the density at @xmath20 , @xmath42 with @xmath43 , @xmath44 , and @xmath45 .
finally , the parameters entering the jacobi sines @xmath46 and cosines @xmath47 are @xmath48 , @xmath49 , and @xmath50 , @xmath51 .
given @xmath23 and @xmath24 , we are left with two free parameters : @xmath41 and @xmath37 , which are next determined by matching the density and its derivative at the barrier edge @xmath38 .
first , the derivative matching condition , enforced using the first integral of eq .
( [ eq : denequation ] ) , allows us to write @xmath37 as a function of @xmath41 for any value of @xmath52 : @xmath53\ .
\label{eq : nd}\ ] ] the density matching equation @xmath54 ( @xmath55 ) is then solved by a numerical root finding method , yelding @xmath41 .
two bounded solutions are always found @xcite , an example is given in the upper panel of fig .
[ fig1 ] . in the following
, the solution which tends to a plane wave for @xmath56 will be referred to as `` upper solution '' , while the one tending to a grey soliton will be called `` lower solution '' @xcite . for given barrier parameters
@xmath18 and @xmath57 , and at a fixed density at infinity @xmath23 , the solutions exist up to a critical injected velocity @xmath58 , at which they merge and disappear .
this behavior was found in the case of a repulsive delta potential in @xcite .
similarly , in @xcite , the same kind of merging was reported for a 1d bec flow through a repulsive square well when the width of the latter increases . as pointed out in the introduction , the current - phase relation @xmath0 for a given superfluid
only depends on the properties of the weak link , in our case the barrier height @xmath18 with respect to the chemical potential and the width @xmath17 with respect to the healing length . for a fixed current @xmath59
, the phase difference across the system is calculated using the relation @xmath60 and then renormalized by the phase accumulated by the a plane wave with the same boundary conditions in absence of barrier ( see lower panel in fig . [ fig1 ] ) .
two different values of @xmath61 are found , corresponding to the upper and lower solutions . .
density ( upper panel ) and phase ( lower panel ) as a function of position are shown for both the upper and the lower solution ( see inset ) .
dashed lines in the lower panel correspond to the phase accumulated by a plane wave in absence of barrier . ] in this section , we will use dimensionless quantities , employing the chemical potential at zero current @xmath62 as the unit of energy , the bulk healing length @xmath63 as the unit of length , and @xmath64 as the unit of time . exploiting the symmetry of the system about @xmath20
, the phase difference can be written as @xmath65\ , i=1,2 $ ] , where the third term is the phase difference accumulated by the plane wave .
the limit can be calculated using eq .
( [ eq : solutionout ] ) , yielding @xmath66 .
\label{eq : phase2}\ ] ] the first two terms in eq .
( [ eq : phase2 ] ) correspond to the phase acquired inside the barrier while the third , which we can call the pre - bulk term , gives the phase accumulated outside the barrier , where the density has not yet reached its bulk value @xmath23 .
we have taken the latter to be one . in fig .
[ fig2 ] , the current phase relation is shown for different barrier widths and heights . each curve has a maximum at the point @xmath67 , with @xmath68 being the critical current at which the two stationary solutions merge and disappear .
the upper solutions constitute the part of the current - phase diagram which connects the maximum with the point @xmath69 , while the lower ones belong to the branch connecting the maximum to the point @xmath70 .
indeed , we will now show that , for any @xmath57 and @xmath56 , the upper branch tends to a plane wave , while the lower branch tends to a grey soliton .
in order to have a finite @xmath37 in this limit , the term in square brackets in eq .
( [ eq : nd ] ) must tend to zero , yielding a cubic equation with two coincident solutions @xmath71 and a third @xmath72 , where we have set @xmath73 for simplicity .
for @xmath71 we have @xmath74 , corresponding to the plane wave solution @xmath75 .
for @xmath72 we obtain instead @xmath76 , @xmath77 , corresponding to a grey soliton solution @xmath78 .
therefore , in this limit @xmath79 for any @xmath80 , meaning that the upper branch is actually a vertical line , while for the lower branch we have @xmath81 this curve has a maximum at @xmath82 , corresponding to @xmath83 , that is , the sound velocity @xmath5 in dimensionless units . the arrows in fig .
[ fig2 ] sketch the behavior of the maximum of the current - phase relation @xmath67 as the height @xmath18 is increased at a fixed barrier width @xmath17 . for any width , the current - phase diagram initially takes a cosine - like shape ( eq .
( [ eq : freephase ] ) ) when @xmath84 ( red squares in fig .
[ fig2 ] ) , and tends to a @xmath85 shape for sufficiently large @xmath18 , characterizing the josephson regime of tunneling flow ( blue triangles in fig .
[ fig2 ] ) . between these two limits , the behavior of the maximum
is determined by the barrier width . for thin barriers ( @xmath86 ) , as shown in the left panel of fig .
[ fig2 ] , the point @xmath67 moves down - right , reaching the josephson regime keeping @xmath87 always smaller than @xmath88 . on the other hand , for sufficiently wider barriers , @xmath87 is able to reach values larger than @xmath88 within a finite range of @xmath18 during the down - right displacement of the maximum .
the latter then moves down - left to finally enter the josephson regime , as shown in the right panel of fig .
we note that in this way , while @xmath18 is increased , the phase @xmath87 takes the value @xmath88 twice , but only the second time entering the josephson regime with a sinusoidal current - phase relation . the first time ( orange diamonds in the right panel of fig .
[ fig2 ] ) the flow is not yet in the tunneling regime since @xmath18 is much smaller than the chemical potential , indeed the current - phase relation is symmetrical with respect to @xmath88 , but not sinusoidal .
( left panel ) and @xmath89 ( right panel ) , for different barrier heights @xmath18 ( see inset ) .
the arrows sketch the behavior of the maximum of the curves upon increasing @xmath18 . ]
as described in the previous section , for strong enough barriers the flow enters the tunneling regime , and the current - phase relation takes a sinusoidal form . in the following , we will describe analytically this behavior , deriving a relation between the josephson critical current and the barrier parameters @xmath18 and @xmath57 .
since we are now interested in tunneling flows , we will take @xmath90 and will show that the josephson regime is attained by either increasing the barrier height @xmath18 or its width @xmath17 .
since in this regime the injected velocity of a stationary flow @xmath24 must be much smaller than the sound velocity @xmath5 , the chemical potential can be written as @xmath91 , upon neglecting the kinetic energy term @xmath92 . moreover , the density inside the barrier @xmath41 being exponentially small , we can write @xmath93 , with @xmath94 , where we have neglected @xmath95 . thus ,
both the upper and lower solutions are of the kind @xmath96 , with @xmath97 , @xmath98 , and @xmath99 .
the density in the josephson regime has thus the form @xmath100 in order to write the density matching equation @xmath101 , we approximate @xmath37 by discarding both @xmath95 and @xmath102 in eq .
( [ eq : nd ] ) , obtaining a quadratic equation for @xmath41 . further assuming that @xmath103 , with @xmath104 being the kinetic energy associated with the barrier length scale @xmath57 , the solutions of the above equation are of the form @xmath105 with @xmath106 , and @xmath107 .
the josephson critical current corresponds to the merging of the two solutions at @xmath108 , and reads @xmath109 where we have used @xmath110 , for @xmath111 . the critical velocity for a bosonic and fermionic superfluid flowing through a repulsive square well has been calculated in @xcite within the local density approximation ( for bec case see also @xcite ) .
analytical expressions for the critical current of a bec flow were found for both slowly varying and weak barriers @xcite .
( [ joscrit ] ) thus enriches the above set of analytical results by providing the critical current for strong barriers . finally , we calculate the current - phase relation using eq .
( [ eq : phase2 ] ) .
since @xmath112 is exponentially small , only the first term in eq .
( [ eq : phase2 ] ) contributes , and the integral can be performed analytically to obtain : @xmath113 thus , we recover the sinusoidal current - phase relation characterizing a josephson regime of tunneling flow . .
insets show in more detail the shape of the diagram close to the critical point .
the arrows sketch the behavior of the maximum of the curves upon increasing @xmath18 . ] for the upper and lower solutions . ]
when the barrier width @xmath17 greatly exceeds the healing length , the current - phase relation becomes multivalued as shown in fig .
[ fig3 ] for @xmath114 .
these so - called reentrant current - phase diagrams were first predicted for long superconducting weak links @xcite .
remarkably , they have been experimentally demonstrated with superfluid @xmath115 @xcite . in our system ,
two kinds of multivalued diagrams are found for a fixed barrier width @xmath17 .
we designate the first kind ( green dots ) , appearing at larger barrier heights @xmath18 , as reentrant type i , and the second kind ( orange diamonds in fig . [ fig3 ] ) , occurring for smaller @xmath18 , as reentrant type ii .
they differ in the behavior of the lower branch at small currents @xmath116 .
the phase decreases with increasing @xmath80 in type ii diagrams while it increases in type i , reaching values larger than @xmath6 for @xmath116 .
these two families of diagrams also differ in the number of positive values of the current @xmath80 corresponding to the same phase difference @xmath61 .
indeed , diagrams of type i can have two values of @xmath80 at the same @xmath61 , while diagrams of type ii can allow for three .
the existence of these type ii diagrams , to our knowledge , has not been discussed in the literature . in @xcite , only type
i current - phase relations were observed and connected to the onset of phase - slip dissipation in the system . in this spirit
, a type ii diagram might for instance lead to an instability of a different kind . in the remainder of this section
, we will develop an analytical model that captures the essential features underlying both kinds of reentrant current - phase diagrams .
examination of typical density profiles belonging to the reentrant regime ( see fig
. [ fig4 ] ) suggests to construct the lower solution by starting from the upper one at the same current , then adding a grey soliton inside the barrier region .
since we are dealing with wide barriers , we describe the upper solution in the local density approximation ( lda ) ( its current - phase relation in this approximation is also discussed in @xcite ) . at a fixed current @xmath80 ,
the density inside the barrier @xmath117 is constant and equal to @xmath118 , when @xmath119 , or @xmath120 , when @xmath121 with @xmath122 defined by @xmath123 , and @xmath124 .
the phase difference calculated within lda misses the pre - bulk term in eq .
( [ eq : phase2 ] ) , thus , for the upper branch , it is simply @xmath125 . now , for the density profile of the lower branch @xmath126 , we take a grey soliton ( eq . ( [ eq : solutionout ] ) ) placed inside the barrier at @xmath20 , with a bulk density given by @xmath41 and a bulk velocity @xmath127 while in the region
@xmath128 we keep the density profile of the upper branch , that is , a constant density @xmath23 and velocity @xmath129 .
notice that in this section @xmath41 stands for the density of the upper solution at @xmath20 , as indicated in fig .
the density at @xmath20 for the lower solution corresponds to the center of the dip in the grey soliton density profile . finally , using eq .
( [ eq : phase2 ] ) we obtain the phase difference for the lower branch @xmath130 at a given current @xmath80 , the overall phase difference corresponding to the lower solution has two contributions : 1 ) the `` hydrodynamics phase '' @xmath131 coming from lda and 2 ) the `` nonlinear - dispersive phase '' @xmath132 accumulated across the grey soliton . while @xmath131 is a monotonically increasing function of @xmath80 , @xmath133 is instead monotonically decreasing , starting from @xmath6 at zero current @xcite .
therefore , the hysteresis characterizing a reentrant current - phase relation is due to the competition between the hydrodynamic and the nonlinear - dispersive components of the flow , which can coexist only for barriers wide enough to accomodate a soliton inside . in particular
, we can derive a condition for the appearance of type i reentrant behavior upon expanding eq .
( [ reentrantphase ] ) for small currents , and requiring @xmath134 . using @xmath135 , for @xmath136 , we get @xmath137 , where @xmath138 , and we have taken @xmath139 .
the condition for type i reentrance to appear is thus @xmath140 . for @xmath116
, we have @xmath141 , and since within the lda approximation @xmath142 , the condition @xmath140 takes the simple form @xmath143 with @xmath144 being the healing length inside the barrier region where the density is @xmath41 .
equation ( [ condreentrant ] ) , holding for @xmath142 , has a clear physical meaning : in order to have a type i reentrant current phase diagram , the barrier width @xmath17 must be sufficiently larger than @xmath145 , which is the characteristic size of a soliton placed inside the barrier . in the left panel of fig .
[ fig5 ] , we compare the current - phase relation calculated with the above model ( solid lines ) to the exact results . within the reentrant regime , for both type i and type ii , the agreement is striking with only slight differences close to the the maximum @xmath67 . on the other hand , for thin / strong barriers , lda , and in turn
the above model , is in clear disagreement with the exact results .
( see cases @xmath89 in the left panel of fig .
[ fig5 ] ) .
the difference between type i and type ii diagrams has a physical interpretation within the above model , namely that the hydrodynamic component of the flow dominates for all currents in type i reentrance ( excluding the region @xmath146 @xcite ) , while it is overcome by the nonlinear - dispersive part for sufficiently small current in type ii . in the literature @xcite , multivalued current - phase relations
are typically modelled by describing the weak - link with an equivalent circuit containing a linear inductance in series with a sinusoidal inductance , the latter corresponding to an ideal josephson junction . when compared to our gpe exact results ( see right panel in fig .
[ fig5 ] ) , this model in general fails to describe the curvature of both branches , and misses type ii reentrance , as well as nearly free regimes ( e.g. red squares in fig .
[ fig2],[fig3 ] ) since it does not allow the phase @xmath87 , corresponding to the maximum of the diagram , to be smaller than @xmath88 .
it proves quite accurate only for sufficiently large barrier heights @xmath18 , very close to the josephson regime . in the @xmath3he experiment of @xcite ,
this so called deaver - pierce model @xcite agrees well with the measured current - phase relation .
this might be due to the fact that these experiments are performed with a fixed weak link configuration , moving the system across the transition between josephson and type i reentrant regime , upon changing the @xmath3he healing length with temperature , but always staying sufficiently close to the josephson regime .
we studied the current - phase relation of a bec flowing through a weak link created by a repulsive barrier .
the link is thus modelled by two parameters , the barrier height and width , which fix the current - phase relation . even though we solved a simplified model , we believe that the results obtained will also be relevant for superconducting and superfluid he - based weak links .
we obtained analytical results for the weak barrier limit , for which the current - phase relation has a @xmath147 form , and for the strong barrier limit , for which it takes a @xmath148 form characterizing the josephson regime .
in particular , we derived an expression for the josephson critical current as a function of the link parameters . finally , we found two kinds of multivalued current - phase diagrams which we show , by means of an analytical model , to appear due to the competition between a hydrodynamic and a nonlinear - dispersive part of the flow , which can coexist only for barriers wide enough to accomodate a soliton inside .
the first kind , showing two different positive currents at same phase difference , has recently been experimentally demonstrated with @xmath3he and proven to be connected to the appearance of phase slippage in the ac josephson dynamics .
the second one , new to our knowledge , can instead allow for three different positive currents corresponding to the same phase difference .
we believe that this new kind of hysteresis in the current - phase relation can be associated with new features emerging in the josephson dynamics , which will be studied elsewhere .
the los alamos national laboratory is operated by los alamos national security , llc for the national nuclear security administration of the u.s .
department of energy under contract no .
de - ac52 - 06na25396 .
the study of the stability of the two solutions is beyond the scope of the present paper .
this issue has been addressed in @xcite .
g. mamaladze and o. d. cheshvili , zh .
fiz . * 50 * , 169 ( 1966 ) [ sov .
jetp * 23 * , 112 ( 1966 ) ] ; j. s. langer , and v. ambegaokar , phys . rev . * 164 * , 498 ( 1967 ) ; b. t. seaman , l. d. carr , and m. j. holland , phys . rev .
a * 71 * , 033609 ( 2005 ) ; w. d. li , phys .
a * 74 * , 063612 ( 2006 ) .
by further approximating @xmath139 we get @xmath149 .
since @xmath41 is a monotonically decreasing function of @xmath80 , the argument of @xmath150 increases for increasing current .
therefore , starting from @xmath6 at @xmath151 , @xmath133 decreases monotonically with @xmath80 .
close to the critical point @xmath67 , and for every current - phase relation , the lower branch always has a phase decreasing with increasing current , up to the critical point itself , at which it meets the upper branch .
in particular , for reentrant diagrams , this means that the dispersive part of the flow always dominates over the hydrodynamic part sufficiently close to the critical point . | we study the current - phase relation of a bose - einstein condensate flowing through a repulsive square barrier by solving analytically the one dimensional gross - pitaevskii equation .
the barrier height and width fix the current - phase relation @xmath0 , which tends to @xmath1 for weak barriers and to the josephson sinusoidal relation @xmath2 for strong barriers . between these two limits ,
the current - phase relation depends on the barrier width .
in particular , for wide enough barriers , we observe two families of multivalued current - phase relations .
diagrams belonging to the first family , already known in the literature , can have two different positive values of the current at the same phase difference .
the second family , new to our knowledge , can instead allow for three different positive currents still corresponding to the same phase difference .
finally , we show that the multivalued behavior arises from the competition between hydrodynamic and nonlinear - dispersive components of the flow , the latter due to the presence of a soliton inside the barrier region . |
most of the standard literature in real options is based on one or both of the following unrealistic assumptions : ( 1 ) that the time horizon for the problem at hand is infinite and ( 2 ) that the real asset under consideration is perfectly correlated to a traded financial asset .
the _ infinite maturity _
hypothesis helps to reduce the dimensionality of the problem by removing its dependence on time , therefore concentrating on stationary solutions only , while the _ spanning asset _ hypothesis allows the introduction of useful replication arguments developed for derivative pricing in complete markets .
together they led to the development of a coherent and intuitive approach for investment under uncertainty , well
represented for instance in dixit and pindyck ( 1994 ) , where the decisions to start , abandon , reactive and mothball a given project were reduced to the solution of systems of linear equations . since then , several authors have dropped the artifice of an infinite maturity time , therefore resourcing to numerical methods for dealing with the non - stationary valuation problem .
these include finite difference methods for the associated partial differential equation and lattice methods for discrete time option pricing .
however , even recent books such as smit and trigeorgis ( 2004 ) carry the assumption that real - options valuation is still applicable provided we can find a reliable estimate for the market value of the asset " ( page 102 ) , which is tantamount to saying that markets are sufficiently complete " . in reality , most investment problems where the real options approach is deemed relevant occur in markets which are far from being complete .
for example , almost by definition an r&d investment decision concerns a product which is _ not _ currently commercialized and therefore commands uncertainty that is at best imperfectly correlated with available financial assets .
exceptions to this adherence to a near completion " assumption , but still in the context of an infinite time horizon , are hugonnier and morellec ( 2004 ) and henderson ( 2005 ) . in the first paper ,
a risk averse manager facing an investment decision tries to maximize his expected utility considering the effect that shareholders external control will have on his personal wealth . by assuming that the underlying project is subject to both market risk , which the manager can hedge using a traded financial asset , and idiosyncratic risk , which can not be hedged in the available financial market , the authors reduce this decision to an investment problem in an incomplete market .
by contrast , under a similar model for a project with both systematic and idiosyncratic risks , henderson ( 2005 ) uses an exponential utility framework in order to actually calculate the value for the opportunity to invest as a derivative in an incomplete market , therefore remaining closer in spirit to the real options paradigm . in this paper
, we study a finite horizon version of henderson s model .
we first review the mechanism for pricing derivatives in an incomplete markets using an exponential utility in the context of investment decisions in a simple one period binomial model in section [ one - period - section ] . in section [ multi - period - section ]
we extend the valuation procedure to a multi
period model used an an approximation for continuous time markets , followed by numerical experiments exploring the properties of the option to invest in section [ numerical ] , including comparisons with the corresponding infinite horizon and complete markets limits .
section [ conclusion - section ] then presents conclusions drawn from the model , especially in contrast with alternative ways of dealing with market incompleteness in the context of real options .
consider an investor who needs to decide whether to pay a sunk cost @xmath0 in exchange of a project with current value @xmath1 .
assume that such investment can be made either at time @xmath2 or postponed until time @xmath3 , when the project value is expected to rise or fall according to specified probabilities , so that the opportunity to invest is formally equivalent to discrete - time early - exercise call option with strike price @xmath0 having the project value as the underlying asset .
when the project value is perfectly correlated to the price of a traded financial asset , such option can be priced using standard arbitrage and replication arguments . in the absence of such spanning asset ,
the option to invest becomes analogous to a derivative in an incomplete market .
instead of wishfully pretending that risk neutral and replication arguments can still be used in this case , we argue that the investor s risk preference should be explicitly used for valuing the option to invest . for this , we follow hobson and henderson ( 2002 ) and consider a utility indifference framework based on an exponential utility of the form @xmath4 .
let us assume the existence of a riskless cash account with constant annualized interest rate @xmath5 , which we use as a fixed _ numeraire _ , and denote the discounted project value by @xmath6 and the discounted price of a correlated traded financial asset by @xmath7 .
we then specify their one period dynamics by @xmath8 where @xmath9 and @xmath10 , for positive initial values @xmath11 and historical probabilities @xmath12 . without the opportunity to invest in the project @xmath6 , a rational agent with initial wealth @xmath13
will keep an amount @xmath14 in the cash account and holding @xmath15 units of the traded asset @xmath7 in such a way as to maximize the expected utility of the terminal wealth @xmath16 that is , the investor will try to solve the optimization problem @xmath17\ ] ] suppose next that the investor is offered the opportunity to invest in the project at the end of the period , thereby receiving a discounted payoff @xmath18 . in order to acquire this option
, the investor must spend an amount of money @xmath19 at time @xmath20 .
for instance , @xmath19 might be the price of land that will allow a subsequent real estate development , or the price of a license to explore a natural resource .
the key assumption now is that , after obtaining the option , the investor will try to use the financial market in such a way as to maximize the utility of the total terminal position . in other words , an investor with initial wealth @xmath13 who acquires the option for the price @xmath19 will try to solve the modified optimization problem @xmath21 \label{optimalhedge}\ ] ] following hodges and neuberger ( 1989 ) , we define the _ indifference price _ for the option to invest in the final period as the amount @xmath22 that solves the equation @xmath23 denoting the two possible pay - offs at the terminal time by @xmath24 and @xmath25 , it is then a straightforward calculation to show that , for an exponential utility , such indifference price is given by @xmath26 where , for fixed parameters @xmath27 the function @xmath28 is given by @xmath29 with @xmath30 we will henceforth refer to @xmath22 as the _ continuation value _ for holding the option of investing at a later time . if we now introduce the possibility of investment at time @xmath20 , it is clear that immediate exercise of this option will occur whenever its _ exercise value _ @xmath31 is larger than its continuation value @xmath22 .
that is , from the point of view of this agent , the value at time zero for the opportunity to invest in the project either at @xmath20 or @xmath32 is given by @xmath33
consider now a continuous - time two factor market of the form @xmath34 where @xmath35 are constants and @xmath36 are standard independent brownian motions under the historical probability measure @xmath37 .
as before , @xmath38 corresponds to the discounted price of a traded financial asset which is correlated to the discounted value of the underlying project @xmath39 with a correlation parameter @xmath40 .
an approximation for this market can be obtained by dividing the time interval @xmath41 $ ] into @xmath42 subintervals with equal time steps @xmath43 and taking the one period dynamics for the discrete time processes @xmath44 to be given by .
we then need to choose the dynamic parameters @xmath45 and the one - period probabilities @xmath46 so that , in the limit of small @xmath47 , such dynamics matches the distributional properties of the continuos time processes @xmath38 and @xmath39 . for instance , one can verify that , up to terms of order @xmath47 , the choices @xmath48,\end{aligned}\ ] ] provide the correct variances , means and correlation for @xmath38 and @xmath39 . supplemented by the condition @xmath49 these equations uniquely determine the historical probabilities @xmath46 .
having fixed these parameters , let us choose a sufficiently large integer @xmath50 and denote @xmath51 ranging from @xmath52 to @xmath53 , respectively the highest and lowest achievable discounted project values starting from the middle point @xmath1 with the multiplicative parameter @xmath54 .
in practice , @xmath50 should be chosen so that such highest and lowest values are comfortably beyond the range of project values that can be reached during the time interval @xmath41 $ ] with reasonable probabilities ( say by corresponding to returns which are away from their mean by more than four standard deviations ) .
then each realization for the discrete - time process @xmath55 following the dynamics can then be thought of as a path over a @xmath56 rectangular grid having the values as its repeated columns .
the discounted value of the option to invest on the project can then be determined as a function @xmath57 on this grid , with the index @xmath58 referring to the underlying project value @xmath59 , and the index @xmath60 referring to time @xmath61 .
we start by specifying the following boundary conditions : @xmath62 condition corresponds to the fact that at maturity the option to invest should be exercised whenever the project value exceeds the investment cost , whereas conditions and mean that such option should always be exercised when the project value is at its highest but is worthless when the project is at its lowest .
the values in the interior of the grid are then obtained by backward induction as follows : @xmath63 that is , at each node on the grid , the investor chooses between exercising the investment option , obtaining its immediate exercise value @xmath64 or holding the option one step into the future , retaining its continuation value @xmath65 .
accordingly , at each time @xmath66 , the exercise threshold @xmath67 is defined as the project value for which the exercise value for the option becomes higher than its continuation value . for project values below @xmath67 ,
the investor will prefer to hold the option , while for project values higher than such threshold , preference for immediate exercise will prevail .
we now investigate how the exercise threshold , and consequently the value of the option to invest , varies with the different model parameters . in
what follows , unless explicitly indicated , we fix the investment cost @xmath68 , the risk free interest @xmath69 , the time to
maturity @xmath70 , the dynamic parameters for the traded asset @xmath71 , @xmath72 , @xmath73 and the volatility for the project @xmath74 . given these parameters ,
the capm equilibrium expected rate of return on the project for a given correlation @xmath40 is @xmath75 the difference @xmath76 , known as the below
equilibrium rate of return shortfall , should then be interpreted as the incomplete market analogue of a dividend rate paid by the project , which we fix at @xmath77 . because incompleteness is the main theme of this paper , we starting with the dependence with correlation . in the complete market case ,
corresponding to @xmath78 in our model , the investment threshold can be obtained in closed - form in the limit of an infinite time horizon . for the parameters above
, such formula gives @xmath79 ( see page 153 in dixit and pindyck ( 1994 ) ) .
by contrast , the investment threshold obtained from a simple npv criterion ( that is , invest whenever the net present value for the project is positive ) in this case is equal to @xmath80 .
this constitutes the most widespread result from real option theory : irreversibility and time flexibility lead to investors waiting until much larger thresholds before committing to an investment decision . in the left panel of figure [ corr_gamma ]
we see what reservations need to be made in the presence of market incompleteness .
predictably , the exercise thresholds decrease as we move away from @xmath81 , meaning that the presence of unhedged risks decreases the value of the option to invest .
interestingly however , even at its minimum , corresponding to @xmath82 , the investment threshold is still higher than suggested by npv .
that is , even when the risk in the project is entire idiosyncratic and therefore can not be hedged with financial assets , time flexibility still confers an option value to the opportunity to invest which is higher than its net present value , _ irrespectively of any replication argument_. [ insert figure [ corr_gamma ] here ] moving to the dependence with risk aversion , for an incomplete market and still in the infinite time horizon setup , mcdonald and siegel ( 1986 ) obtain the exercise threshold by assuming that investors require compensation for market risks whilst being risk
neutral towards idiosyncratic risk . as observed in henderson ( 2005 )
, this corresponds to the limit @xmath83 in an exponential utility framework . for the parameters above , with the adjustment for market risks done according to capm , this threshold coincides with the value @xmath84 of dixit and pindyck ( 1994 ) .
we can observe these limiting behavior in the right panel of figure [ corr_gamma ] , together with the fact that the investment threshold for our model expectedly decreases as a function of the parameter @xmath85 .
that is , risk aversion can significantly erode the option value obtained from time flexibility . in the limit @xmath86
, one can explicitly show that expression tends to the subhedge price of the derivative , which is zero for a call option , so that value for the investment opportunity reduces to its net present value , with the corresponding investment threshold @xmath80 . as we observe in the graph
, this erosion of value with risk aversion is faster the lower the correlation between the project and the traded asset . as with classical real options
, we can verify in our model that heightened uncertainty in project values increases the value of time flexibility , which is reflected in the increasing threshold with respect to the volatility @xmath87 observed in the left panel of figure [ vol_delta ] . in the same figure , we observe how the threshold decreases as a function of @xmath88 , implying that the incentives for immediate investment in the project are higher the more its rate of return falls below its equilibrium value , which is the incomplete market analogue of more investment when dividends are higher .
[ insert figure [ vol_delta ] here ] all of these features were already present in the elegant closed - form solution obtained in henderson ( 2005 ) for the infinite time horizon limit and under the somewhat artificial restriction that the investment cost should grow at a rate @xmath89 .
our contribution consists on the one hand in removing this restriction ( our investment cost can grow at any rate , in particular at a rate @xmath90 as above ) and on the other hand in numerically calculating the exercise threshold as a function of time to maturity . in figure
[ maturity ] we see that the exercise threshold can take a long time to converge to its asymptotic value , particularly in the desirable cases of low risk aversion and high correlation , indicating that for typical maturities of only a couple of years the stationary solution provides a poor approximation for its finite horizon counterpart .
[ insert figure [ maturity ] here ] we conclude this section with a graph of option values as a function of the current level of the underlying project .
these should be compared with either figure 5.3 on page 154 of dixit and pindyck ( 1994 ) or figure 2 of henderson ( 2005 ) , which use the same base parameters as ours , except for time to maturity , which we take to be @xmath70 .
we can observe once more that higher correlations lead to higher option values .
moreover , we confirm our previous observation that even for @xmath82 the opportunity to invest is more valuable then its net present value , represented in the graph by the solid line depicting the function @xmath91 .
finally , notice how the smooth pasting and matching conditions , which were not _ a priori _ assumed in our model , are satisfied by the option values , in the sense that the curves in figure [ value ] smoothly match the function @xmath91 at the corresponding exercise thresholds , marked in the graph by the two vertical dotted lines . in all of the numerical experiments above , we used a fixed time step @xmath92 , so that the relative precision for project values on the grid is of the order @xmath93 .
for each point marked in the pictures above , the thresholds and option values were obtained on a typical @xmath94 grid in approximately 15 seconds using a desktop pc at 3mhz .
we have proposed a multi period binomial model for assessing the value of an option to invest on a project in a finite time horizon in the absence of a perfectly spanning financial asset .
the exercise thresholds obtained from our model exhibit the expected dependence with respect to correlation , uncertainty , risk aversion , dividend rates and time to maturity . in particular , for perfectly correlated or perfectly anti correlated assets the threshold approaches that of a complete market , whereas when the aversion to idiosyncratic risks tends to zero the threshold approaches the one obtained under similar assumptions by mcdonald and siegel ( 1986 ) .
moreover , we verify that even in the zero correlation case , whereby none of the risk in the project can be hedge in a financial market , the paradigm of real options can still be applied to value an investment decision , since the opportunity to invest still carries an option value above its net present value . in other words , it is time flexibility itself , more than the possibility of replication , that is the source of the extra value of an investment opportunity .
this value , however , erodes sharply at higher levels of risk aversion , and even more so when the project is uncorrelated to financial markets .
apart from the outright use of risk neutral valuation even when markets are incomplete - under the wishful assumption that they are complete enough for all practical purposes - the most widespread alternative method for dealing with incompleteness in a real options context is through the use of dynamic programming with an exogenous discount rate .
this is the approach indicated for instance in the second half of chapter 5 in dixit and pindyck ( 1994 ) , in which an investor equates the expected capital appreciation from a project to the expected rate of return on the investment opportunity , using a corporate rate of return , which is different from the risk free interest rate and meant to express corporate risk preferences . despite its popularity , such approach has the serious theoretical drawback that the fully nonlinear risk preferences of a corporation can hardly be expressed through a single discount factor .
in fact , the majority of financial economics literature use an expected utility function _ together _ with an exogenous discount factor in order to model risk preferences . at a more practical level , this dynamic programming approach with a corporate discount rate obscures the most important aspects of real options , namely the intuition that can be gain when managerial decisions are treated as options .
for example , under the option paradigm , investment on a multi - stage project is analogous to a portfolio of options , each having its own value and interacting in a complex manner towards the value of the whole project .
precisely because such analogies are completely lost in the dynamic programming approach , authors such as dixit and pindyck dropped it in the remaining of their book in favor of a contingent claim analysis , which then formally relies back on the complete market framework with a spanning asset hypothesis .
in comparison , our proposed method handles incompleteness by explicitly introducing risk preferences in an economically sound utility based framework for the realistic case of a finite time horizon , while retaining the computational complexity of a standard binomial valuation .
the method can then be extended in order to provide incomplete market versions for all the standard managerial decisions treated as real options .
for example , all the thresholds for investment , abandonment , suspension and reactivation of a project in an incomplete market are obtained in detail in the forthcoming paper by grasselli and nitzan ( 2006 ) . as an application in a different direction ,
a modification of the method proposed here was used in grasselli ( 2005 ) for the valuation of employee stock options . in this way
, our valuation method considerably enlarges the domain of applicability of both real options and indifference pricing . | we apply a utility based method to obtain the value of a finite time investment opportunity when the underlying real asset is not perfectly correlated to a traded financial asset . using a discrete time algorithm to calculate the indifference price for this type of real option , we present numerical examples for the corresponding investment thresholds , in particular highlighting their dependence with respect to correlation and risk aversion .
* key words : * real options , incomplete markets , exponential utility , optimal exercise policy . |
the emission from narrow - line region ( nlr ) of active galactic nucleus ( agn ) is an important tool to study the relation between the activity of the central supermassive black hole ( smbh ) and the growth of its host galaxy in which the smbh resides , both because the nlr emission is mainly resulted from the illumination by the central agn and because the nlr kinematics is believed to be mainly dominated by the gravity of the bulge ( see review in wilson & heckman 1985 and references therein , whittle 1992a , b ; nelson & white 1996 ) . the gravity dominated kinematics motivates a number of previous studies to demonstrate that the line width of the agn s strong [ ] @xmath15007 emission line can be used as a proxy for the stellar velocity dispersion of the bugle ( e.g. , nelson & white 1996 ; nelson 2000 ; boroson 2003 ; komossa & xu 2007 ) . basing upon the tight @xmath2 relationship ( e.g. , tremaine et al . 2002 ; ferrarese & merritt 2000 ; magorrian et al .
1998 ; gebhardt et al .
2000 ; haring & rix 2004 ) , the proxy therefore allows one to easily estimate @xmath3 in a large sample of agns ( e.g. , grupe & mathur 2004 ; wang & lu 2001 ; komossa & xu 2007 ) .
it is well known for a long time that the line profiles of the [ ] doublelets show a blue asymmetry with an extended blue wing and a sharp red falloff in a large fraction of agns ( e.g. , heckman et al .
1981 ; whittle 1985 ; wilson & heckman 1985 ; grupe et al 1999 ; tadhunter et al .
2001 ; veron - cetty et al . 2001 ; zamanov et al . 2002 ; komossa & xu 2007 ; xu & komossa 2009 ; greene & ho 2005 ; de roberties & osterbrock 1984 ; storchi - bergmann et al . 1992 ; arribas et al . 1996 ; christopoulou et al . 1997 ) .
the blue asymmetry requires a narrow core gaussian profile ( @xmath4 ) with a blueshifted , broad gaussian component ( @xmath5 ) to reproduce the observed asymmetric profiles for both [ ] @xmath6 emission lines .
the spectroscopic monitor revealed a variability time scale from one to ten years for the blue wings of the [ ] @xmath0 lines in two type i agns ( izw1 : wang et al . 2005 ; ngc5548 : sergeev et al .
1997 ) , which means that the blue wings are likely emitted from the intermediate - line region located between the traditional blr and nlr .
in addition to the blue asymmetry , the redshifts of the [ ] doublelets are often found to be negative compared to the redshifts measured from both stellar absorption features and h@xmath7 emission line ( i.e. , [ ] blueshifts , e.g. , phillips 1976 ; zamanov et al .2002 ; marziani et al .
2003 ; aoki et al .
2005 ; boroson 2005 ; bian et al . 2005 ; komossa et al .
although they are rare cases , the objects with strong [ ] blueshifts larger than 100@xmath8 are called `` blue outliers '' .
the popular explanation of the observed [ ] emission - line profile is that the material outflow from central agn plays important role in reproducing the observed blue asymmetry and blueshift . with
the advent of the high spatial resolution of hubble space telescope ( hst ) , spatially resolved spectroscopic observations of a few nearby seyfert 2 galaxies indicate that the nlrs show complicate kinematics , which could reproduce the observed [ ] line profiles by the radial outflow acceleration ( or deceleration ) and/or jet expansion ( e.g. , crenshaw et al .
2000 ; crenshaw & kraemer 2000 ; ruiz et al . 2001 ; nelson et al .
2000 ; hutchings et al . 1998
; das et al . 2005 , 2006 , 2007 ; kaiser et al . 2000 ; crenshaw et al . 2010 , schlesinger et al . 2009 ; fischer et al . 2010 ; fischer et al .
2011 ) .
recent systematical studies suggested that the blue asymmetry is related with the activity of the central smbh .
veron - cetty et al . (
2001 ) indicated that half of their sample of narrow - line seyfert 1 galaxies ( nls1s ) shows a broad and blueshifted [ ] @xmath15007 component in addition to the unshifted narrow core component .
nelson et al .
( 2004 ) found a correlation between the blue asymmetry and eigenvector - i space by studying the [ ] @xmath15007 line profiles of the pg quasars .
the quasars associated with larger blue asymmetries tend to be stronger emitters presumably having larger eddington ratios ( @xmath9 , where @xmath10 is the eddington luminosity , see also in xu et al . 2007 ; boroson 2005 ; greene & ho 2005 ; mathur & grupe 2005 ) .
similar as the blue asymmetry , the [ ] blueshift is also found to be related with a number of agn properties .
some authors claimed that the [ ] blueshift is directly correlated with @xmath9 ( e.g. , boroson 2005 ; bian et al .
2005 ) , although the correlation might not be the truth ( e.g. , aoki et al .
marziani et al .
( 2003 ) pointed out that all the `` blue outliers '' have small h@xmath7 line widths ( @xmath11 ) and high @xmath9 ( see also in zamanov et al .
2002 ; komossa et al .
2008 ) .
agns are now widely believed to co - evolve with their host galaxies , which is implied by the tight @xmath2 correlation ( see the citations in the first paragraph ) and by the global evolutionary history of the growth of the central smbh that traces the star formation history closely from present to @xmath12 ( e.g. , nandra et al .
2005 ; silverman et al .
2008 ; shankar et al . 2009 ; hasinger et al . 2005 ) .
a number of studies recently provided direct evidence supporting the co - evolutionary scenario in which an agn evolves along the eigenvector - i space from a high @xmath9 state to a low @xmath9 state as the circumnuclear stellar population continually ages ( e.g. , wang et al .
2006 ; wang & wei 2008 , 2010 ; kewley et al . 2006 ; wild et al .
2007 ; davis et al . 2007 ) .
the results of theoretical simulations indicate a possibility that a major merger between two gas - rich disk galaxies plays important role in the co - evolution of agns and their host galaxies ( e.g. , di matteo et al . 2007 ; hopkins et al . 2007 ;
granato et al .
detailed analysis even suggests a delay of @xmath13myr for the detectable agn activity after the onset of the star formation activity ( e.g. , wang & wei 2005 ; schawinski et al .
2009 ; hopkins et al . 2005 ; davis et al . 2007 ; wild et al . 2010 ) .
the delay might be resulted from the feedback from either the central agn ( e.g. , hopkins et al .
2005 ) or the stellar winds ( e.g. , morman & scoville 1988 ) , and/or resulted from the purely dynamical origin ( hopkins 2011 ) .
the outflow origin for both blue asymmetry and blueshift naturally gives us the hint that the [ ] emission - line profile ( and the inferred nlr kinematics ) co - evolves with the stellar population of the host galaxy .
we here report a study that is an effort to examine the evolution of the [ ] @xmath15007 emission - line profile . in principle , both broad- and narrow - line agns are needed to be analyzed to give a complete study . although the [ ] line profiles can be easily measured in the spectra of typical type i agns , the host galaxy properties are hard to be determined because of the strong contamination from the central agn s continuum . since our aim is to study the relationship between the line profile and the host galaxy properties , the narrow emission - line galaxies from sloan digital sky survey ( sdss ) are adopted in our study because their spectra allow us to simultaneously measure both [ ] @xmath15007 line profile and stellar population properties in individual object .
the high - order dimensionless line shape parameters are adopted by us to describe the line profile in details .
the paper is organized as follows .
the sample selection is presented in 2 .
3 describes the data reduction , including the stellar component removal , the line profile measurements , and the stellar population age measurements .
our analysis and results are shown in 4 .
the implications are discussed in the next section . a @xmath14 cold dark matter ( @xmath14cdm )
cosmology with parameters @xmath15 , @xmath16 , and @xmath17 ( spergle et al . 2003 ) is adopted throughout the paper .
we select a sample of narrow emission - line galaxies from the value - added sdss data release 7 ( abazajian et al . 2009 ) mpa / jhu catalog ( heckman et al .
2004 ; kauffmann et al . 2003a ,
b , c ; tremonti et al .
2004 ; see heckman & kauffmann 2006 for a review ) , the catalog provides the measured spectral and physical properties of 927,555 individual narrow emission - line galaxies , including seyfert galaxies , liners , transition galaxies and star - forming galaxies .
the integrated spectra of the narrow emission - line galaxies allow us to study the relationship between the [ ] @xmath15007 emission - line profiles and the circumnuclear stellar populations , because their continuum is dominated by the starlight components . at first
, we select the galaxies whose spectra have median signal - to - noise ratio per pixel of the whole spectrum s / n@xmath1820 from the mpa / jhu sdss dr7 catalog to ensure that the stellar features could be properly separated from each observed spectrum .
secondly , all the emission lines used in the traditional three baldwin - phillips - terlevich ( bpt ) diagnostic diagrams are required to be detected with a significant level at least 3@xmath19 .
thirdly , the spectral profiles of the [ ] @xmath15007 emission lines are only measured for the galaxies whose [ ] lines have s / n larger than 30 .
finally , the galaxies with redshifts within the range from 0.11 to 0.12 are removed in the subsequent spectral analysis .
the removement is used to avoid the possible fake spectral profile caused by the poorly subtracted strong sky emission line [ ] @xmath15577 at the observer frame .
after the selection based on the mpa / jhu catalog , we remove all the galaxies whose [ ] line widths are smaller than @xmath20 from our subsequent statistical comparisons according to our spectral profile measurements ( see section 3 for details ) . the mean intrinsic resolution used in the above criterion
is adopted to be 65@xmath8 according to the sdss pipelines measurements ( e.g. , york et al . 2000 ) .
there are two reasons for this instrumental resolution selection . at first
, the cut on line width removes the low mass starforming galaxies without bugle .
secondly , the cut can alleviate the impact on the measured line shape parameters caused by the instrumental resolution .
for a pure gaussian profile , the intrinsic line width @xmath19 can be obtained by the equation @xmath21 , where @xmath22 and @xmath23 is the observed line width and the instrumental resolution , respectively . however , this relationship is only a first - order approximation for the line profiles that deviate from a pure gaussian profile .
our analytic calculation indicates that the correction of the instrumental resolution depends on the deviation not only for the line width , but also for the other two high - order dimensionless line shape parameters ( see section 4 and appendix a for details ) . after removing the duplicates in the mpa / jhu catalog ,
there are in total 3,339 entries fulfilling the above selection criteria .
the spectra with the highest s / n ratios are kept in our final sample in the duplicate removements .
given the line fluxes reported in the mpa / jhu catalog , these galaxies are subsequently classified into three sub - samples according to the [ ] /h@xmath7 versus [ ] /h@xmath24 diagnostic diagram using the widely accepted classification schemes ( kauffmann et al .
2003a ; kewley et al .
2001 ) . out of the 3,339 galaxies , there are 2,333 seyfert galaxies / liners ( hereafter agns for abbreviation ) , 793 transition galaxies located between the empirical and the theoretical demarcation lines , and 190 star - forming galaxies .
there are 5,856 star - forming galaxies with line width @xmath25 , whose line profiles are believed to be dominated by a symmetric gaussian function .
these galaxies are used as a comparison sample to present the scatters of the measured high - order dimensionless line shape parameters for a symmetric gaussian line profile , because of the limited wavelength sampling and limited signal - to - noise ratio .
these scatters are useful in determining how strong a line profile deviates from a symmetric gaussian profile . the three diagnostic bpt diagrams are shown in figure 1 for the final samples , in which the agns , transition galaxies and star - forming galaxies are symbolized by the red , green and blue solid squares , respectively .
the contours represent the distributions of the star - forming galaxies in the comparison sample .
the spectra of the selected objects are reduced and analyzed by our principal component analysis ( pca ) pipeline ( see wang & wei ( 2008 ) for details ) . at first , each spectrum is corrected for the galactic extinction using the extinction law with @xmath26 ( cardelli et al .
1989 ) , in which the color excess @xmath27 is taken from the schlegel , finkbeiner , and davis galactic reddening map ( schlegel et al .
1998 ) . secondly ,
each extinctions - corrected spectrum is then transformed to the rest frame , along with the flux correction due to the relativity effect , given the redshift provided by the sdss pipelines ( glazebrook et al . 1998
; bromley et al . 1998 ) .
the stellar absorption features are subsequently separated from each rest - frame spectrum by modeling the continuum and absorption features by the sum of the first seven eigenspectra .
the eigenspectra are built from the standard single stellar population spectral library developed by bruzual & charlot ( 2003 ) .
each of the spectra is fitted over the rest - frame wavelength range from 3700 to 7000 by a @xmath28 minimization , except for the regions with strong emission lines ( e.g. , balmer lines , [ ] @xmath294959 , 5007 , [ ] @xmath296548,6583 , [ ] @xmath296716 , 6731 , [ ] @xmath13727 , [ ] @xmath14363 and [ ] @xmath16300 ) .
the rest - frame emission - line isolated spectra are used to parametrize the spectral profiles of their [ ] @xmath15007 emission lines , after the stellar components are removed from the observed spectra .
the emission - line profile can be parametrized by many possible ways .
the widely used ones include the fwhm and the second moment of the line that is defined as @xmath30 where @xmath19 is in units of @xmath31 , @xmath32 and @xmath33 is the line centroid ( i.e. , the first moment ) and the flux density of the continuum - subtracted line flux , respectively . for a pure gaussian profile
, we have a relationship @xmath34 , which means that both fwhm and @xmath19 comparably describe the line broadening if the line profile is a gaussian function .
however , as stated in greene & ho ( 2005 ) , @xmath19 contains more information on the line profile broadening if the profile deviates from a pure gaussian profile .
heckman et al . ( 1981 ) defined an asymmetry index @xmath35 to measure the line asymmetry of [ ] emission lines , where @xmath36 and @xmath37 is the half width to the left and right of the line center defined as the 80% peak intensity level .
basing upon the index @xmath38 , the authors found that the blue asymmetry is shown in about 80% of the 36 seyfert and radio galaxies .
alternatively , the asymmetry index @xmath39/fwhm$ ] is widely used to measure the asymmetry of agn s broad emission lines , where @xmath40 are profile centroids measured at different levels ( see review in sulentic et al .
in addition to the asymmetry , the emission - line shapes of agns vary from extremely peaked to very `` boxy '' .
previous studies used the ratio of line widths at different levels to parametrize such line shapes ( e.g. , marziani et al .
1996 ) . to make the statistical study on [ ] line profile is feasible for a large sample , we adopt the high - order dimensionless line shape parameters to describe the line profile departures from a pure gaussian function ( see binney & merrifield ( 1998 ) for more details ) :
@xmath41 where @xmath42 is the @xmath43-order moment defined as @xmath44 and @xmath19 is the second order moment defined in eq ( 1 ) , respectively .
the first shape parameter @xmath45 is termed the `` skewness '' that measures the deviation from symmetry .
a pure gaussian profile corresponds to @xmath46 .
the emission - line shape with a positive value of @xmath45 shows a red asymmetry , and the shape with a negative value a blue asymmetry .
the second parameter @xmath47 is termed the `` kurtosis '' that measures the symmetric deviation from a pure gaussian profile ( for a pure gaussian profile , we have @xmath48 ) .
the emission line with peaked profile superposed on a broad base has a value of @xmath49 .
a value of @xmath50 corresponds to a `` boxy '' line profile .
we refer the readers to figure 11.5 in binney & merrifield ( 1998 ) for how the values of @xmath45 and @xmath47 vary with the line shapes .
the bulk relative velocity shift of the [ ] emission line is measured relative to the h@xmath7 emission line in each emission - line isolated spectrum , because the h@xmath7 line can be easily detected and because the line shows very small velocity shift relative to the galaxy rest - frame ( e.g. , komossa et al .
the velocity shift is calculated as @xmath51 , where @xmath52 and @xmath53 is the rest - frame wavelength of the [ ] @xmath0 emission line and the velocity of light , respectively .
@xmath54 denotes the wavelength shift of the [ ] line with respect to the h@xmath7 line .
the shift is determined from the measured line centroids ( @xmath55 ) and the rest - frame wavelengths in vacuum of both lines . with the definition of the velocity shift
, a blueshift corresponds to a negative value of @xmath56 , and a redshift to a positive value . by following our previous studies again ,
we use the two lick indices , i.e. , the 4000 break ( @xmath57 ) and the equivalent width of h@xmath58 absorption feature of a - type stars ( h@xmath59 ) ( bruzual 1983 ; balogh et al .
1999 ) . the index h@xmath60 is defined by worthey & ottaviani ( 1997 ) as @xmath61 where @xmath62 is the flux within the @xmath63 feature bandpass , and @xmath64 the flux of the pseudo - continuum within two defined bandpasses : blue @xmath65 and red @xmath66 . ]
, as the indicators of the ages of the stellar populations of the galaxies ( e.g. , heckman et al .
2004 ; kauffmann et al . 2003 ; kauffmann & heckman 2008 ; kewley et al .
2006 ; wang & wei 2008 , 2010 ; wild et al .
both indices are measured in the removed stellar component for each spectrum , and are reliable age indicators until a few gyr after the onset of a burst ( e.g. , kauffmann et al . 2003 , bruzual & charlot 2003 ) .
the sub - sample of partially obscured agns associated with broad h@xmath24 emission lines ( i.e. , seyfert 1.8/1.9 galaxies ) is selected from the parent agn and transition galaxy samples .
although the existence of the broad h@xmath24 emission is direct indicative of the accretion activity of the central smbh , the underlying agn s continuum in these partially obscured agns could potentially decrease the measured two indices .
these partially obscured agns are therefore removed from the subsequent analysis when the two indices are involved . following our previous studies ( wang & wei 2008 ) ,
the broad - line agns are selected by the means of the high - velocity wings of the broad h@xmath24 components on their blue side after the stellar features are removed from the spectra .
the red wing is not adopted in the selection because of the superposition of the strong [ ] @xmath16583 emission line .
the partially obscured agns are at first automatically selected by the criterion @xmath67 , where @xmath68 is the specific flux of the line wing averaged within the wavelength range from 6500 to 6350 in the rest - frame , and @xmath69 is the standard deviation of the continuum flux within the emission - line free region ranging from @xmath70 and @xmath71 .
the automatically selected agns are then inspected one by one by eyes . in total
, there are 174 broad - line seyfert galaxies / liners and 55 broad - line transition galaxies .
the mpa / jhu catalog contains many duplicates because of the repeat observations .
the duplicates allow us to roughly estimate the uncertainties for the line shape parameters and the stellar population age indicators .
the duplicates are reduced and measured by the same method as described above . for the three main sub - samples
( i.e. , agns , transition galaxies and star - forming galaxies ) , the uncertainties are estimated to be @xmath72 and @xmath73 for the shape parameters @xmath45 and @xmath47 , respectively .
the two lick indices have uncertainties : @xmath74 , and @xmath75h@xmath76 . for the comparison sample ,
the duplicates provide the uncertainties of @xmath77 and @xmath78 for the parameters @xmath45 and @xmath47 , respectively .
after obtaining all the required parameters , the relation between the [ ] line profiles and the stellar population properties is examined in this section .
the [ ] line profile is compared between the agns and the transition galaxies in this section . to avoid the possible systematics caused by the host galaxy properties , we compare the line profiles in a sample of matched galaxy pairs ( e.g. , kauffmann et al .
the matched pairs are created according to the sdss dr4 mpa / jhu agn catalog .
the catalog provides the measurements of galaxy properties for the agns that are classified according to the demarcation line given in kauffmann et al .
we select the pairs that have @xmath79 , @xmath80 , @xmath81 , @xmath82 , and @xmath83 , where @xmath84 is the stellar mass in unit of @xmath85 , @xmath86 is the effective stellar mass - density ( @xmath87 , where @xmath88 is the half - light radius in the @xmath89-band ) in unit of @xmath90 , @xmath91 is the concentration index defined as @xmath92 the ratio of the radius enclosing 90% of the total flux to that enclosing 50% of the flux in the @xmath93-band , @xmath94 is the stellar velocity dispersion and @xmath89 is the redshift . because there are much more agns than the transition galaxies in our sample , a large fraction of the transition galaxies have a few of agn pairs . in these cases ,
one agn is assigned to a pair by a random selection , and the final distributions are built from 100 monte - carlo iterations . finally , there are totally 163 distinct galaxy pairs , in which no galaxy is assigned to a pair more than once .
the left panel in figure 2 compares the cumulative distributions of the [ ] line widths in terms of the calculated velocity dispersions .
the agns and transition galaxies are shown by the red - solid and green - dotted lines , respectively .
the error - bars are resulted from our iterations .
one can see clearly that the line profiles of the agns are systematically wider than that of the transition galaxies .
a two - side kolmogorov smirnov test yields a maximum discrepancy of 0.282 with a corresponding probability that the two samples match of @xmath95 .
in fact , the comparison of the parameter @xmath47 indicates that the difference in the distributions is likely due to the fact that the transition galaxies have systematically smaller values of @xmath47 than the agns ( see section 4.3 for more details ) .
that means the transition galaxies are associated with weaker [ ] broad components compared to the agns , which implies that the feedback caused by the material outflow is less stronger in the transition galaxies than in the agns .
the distributions of the [ ] @xmath0 relative velocity shifts defined in section 3.2.2 are compared in the right panel of figure 2 between the agns and transition galaxies .
the symbols are the same as the left panel .
a two - side kolmogorov smirnov test shows that the two samples are matched at a probability of 77% , although a careful examination suggests that the agns marginally tend to be associated with larger negative velocity shifts compared with the transition galaxies .
the left - bottom panel in figure 3 shows the @xmath47 versus @xmath45 diagram .
the agns , transition galaxies and star - forming galaxies are symbolized by the red , green and blue points , respectively .
the over - plotted contours present the distribution of the star - forming galaxies in the comparison sample .
the two crosses at the left - bottom corner show the typical uncertainties for both parameters as roughly estimated from the duplicates . the red - solid one indicates the uncertainties for the large line - width objects , and the black - dashed one for the comparison sample .
the left - upper panel and the right - bottom panel shows the distributions of the parameters @xmath45 and @xmath47 for all the sub - samples ( including the comparison sample ) , respectively .
the distributions are color - codded for the sub - samples as same as the left - bottom panel .
as expected from an emission - line profile dominated by the instrumental resolution , the distribution of the comparison sample is strongly clustered around the point ( i.e. , @xmath46 and @xmath48 ) associated with a pure gaussian function .
the last column in table 1 lists the average and median values of @xmath45 ( @xmath96 and @xmath97 ) for the comparison sample .
the same values are listed in the last column in table 2 but for the parameter @xmath47 ( @xmath98 and @xmath99 ) .
the dispersions is calculated to be 0.14 and 0.43 for the parameters @xmath45 and @xmath47 , respectively .
the distribution therefore validates our spectral measurements because the distribution of the comparison sample is highly consistent with the prediction of a pure gaussian function .
similar as the comparison sample , the statistics are tabulated in table 2 and 3 for both agns and transition galaxies as well .
one can find from the tables that the line profiles of these galaxies systematically deviate from that of the control sample ( i.e. , deviate from a pure gaussian profile ) by not only smaller @xmath45 ( i.e. , a stronger blue asymmetry ) , but also larger @xmath47 ( i.e. , a stronger broad base ) .
the main panel in figure 3 shows that both agns and transition galaxies form a sequence starting from the pure gaussian region to the upper - left corner .
in fact , this phenomenon is qualitatively in agreement with the fact that two gaussian components , one representing the narrow line core and the other representing the blueshifted broad wing , are usually required to model the observed [ ] line profiles .
a minor fraction of the objects deviate from the sequence by their positive values of @xmath45 . by inspecting the spectra of these objects one by one by eyes , we find that the deviations are mainly resulted from the contamination at the [ ] red wing caused by the weak low - ionization emission line @xmath15016 ( veron et al .
2002 ) .
it is interesting to see that the star - forming galaxies with @xmath100 follow the sequence as well , although most of them have gaussian line profiles . in order to validate the deviations in these galaxies , the spectra with @xmath101
are then inspected one by one by eyes again . to understand the nature of these star - forming galaxies with blue asymmetric [ ] emission lines , we need to carefully model all the strong emission lines in individual spectrum .
the issue is out of the topic of this paper , and will be examined in our subsequent studies . the results from the two - sides kolmogorov - smirnov tests are tabulated in table 3 and 4 for the parameters @xmath45 and @xmath47 , respectively .
each entry contains the maximum absolute discrepancy between the used two sub - samples , and the corresponding probability that the two sub - samples match .
although the agns and the transition galaxies show similar distributions for @xmath45 , the agns significantly differ from the transition galaxies in the distribution of parameter @xmath47 .
the fraction of objects with strong broad [ ] wings is larger in the agns than in the transition galaxies previous studies frequently identified a strong correlation between the [ ] line widths and the [ ] blueshifts ( e.g. , bian et al .
2005 ; komossa et al .
2008 ; zamanov et al .
the correlation is commonly explained by the material outflows .
the similar correlations are also identified in iron coronal lines ( erkens et al .
1997 ) and optical complex ( hu et al . 2008 ) in typical type i agns .
the left panel in figure 4 plots the velocity shift as a function of the line width for the agns ( red points ) and the transition galaxies ( green points ) .
spearman rank - order tests show a marginal anti - correlation between the velocity shifts and the line widths for the agns ( with a spearman correlation coefficient @xmath102 at a significance level @xmath103 , where @xmath104 is the probability that there is no correlation between the two variables . ) , and , however , a stronger correlation for the transition galaxies ( with @xmath105 and @xmath103 ) . as an additional examination
, the middle panel in the figure plots the velocity shift against the parameter @xmath45 .
strong correlations between the two variables can be identified for both agns ( @xmath106 and @xmath103 ) and transition galaxies ( @xmath107 and @xmath103 ) , which means that larger the blueshift , stronger the blue asymmetry .
one can see from the correlations that the objects with the largest velocity shifts deviate from the correlations .
these objects instead show symmetrical line profiles . in order to further describe their line profiles ,
the right panel in the figure plots the velocity shift against the parameter @xmath47 .
we find that these [ ] lines have `` boxy '' line profiles rather than peaked ones .
in fact , a `` boxy '' line profile could be reproduced by the sum of two or more ( distinct ) peaks with comparable fluxes and line widths ( see examples of spectra in the figure 1 in greene & ho 2005 ) .
the evolution effect on the [ ] line profile is examined in this section by using the two indices , @xmath57 and h@xmath108 , as the indicators of the ages of the circumnuclear stellar populations . as described in section 3.3 ,
the partially obscured agns are excluded from the analysis throughout this section .
the two indices are plotted against the parameter @xmath45 ( the left panels ) and against the relative velocity shift ( the right panels ) for the agns in figure 5 . at the first glance of the four panels ,
we fail to directly identify any significant correlation between the line profile parameters and the stellar population ages . to examine the evolution effect in more details ,
we divide the agn sample into two groups according to their @xmath45 values : one group for the galaxies with @xmath109 and the another one for the galaxies with @xmath101 .
the distributions of the @xmath57 values are compared between the two groups in the upper - left panel of figure 7 .
the comparison indicates a significant difference between the two groups . ] .
the agns with larger amount of blue asymmetry are systematically associated with younger stellar populations .
a two - sides kolmogorov smirnov test indicates that the difference between the two distributions is at a significance level @xmath110 with a maximum absolute discrepancy of 0.22 .
similar as the parameter @xmath45 , the agn sample is instead separated into two groups by the relative velocity shift at @xmath111 .
we identify a marginal trend that larger the blue velocity shifts , younger the associated stellar populations ( see the bottom - left panel in figure 7 ) . the same statistical test yields a probability that the two groups match of @xmath112 ( with a maximum absolute discrepancy of 0.11 ) .
figure 6 is the same as figure 5 but for the transition galaxies .
the same methods as the agn sample are adopted to divide these galaxies into two groups .
the right panels in figure 7 compare the distributions of @xmath57 between the two groups .
the two - sides kolmogorov - smirnov tests suggest that the two distributions match at the probabilities @xmath113 for the @xmath45 separation and @xmath114 for the @xmath56 separation , which is much less significant than the agn sample .
in fact , these results are easily understood because the transition galaxies are usually systematically younger than the agns ( see also in kewley et al .
2006 ; schawinski et al .
2007 ; wang & wei 2008 ) . in summary ,
our results indicate that the agns associated with young stellar populations show a wide range in their [ ] line profiles that vary from a blue asymmetrical shape to a gaussian function .
in contrast , the agns associated with old stellar populations always show symmetrical line profiles .
the evolution of [ ] @xmath15007 emission - line profile is studied by using a large sample of narrow emission - line galaxies selected from the mpa / jhu sdss dr7 catalog .
the line - profile parameterizing allows us to reveal a trend that agns with more significant blue asymmetries tend to be associated with younger stellar populations . at the beginning of discussion
, we argue that the trend is not driven only by the orientation effect ( antonucci 1993 ; elitzur 2007 ) .
it is widely accepted that the commonly observed blue asymmetrical and blue - shifted [ ] line profiles are caused by the wind - driven nlr outflows .
the orientation - driven scenario therefore results in an odd corollary that a fraction of the agns with young stellar populations have outflow directions closer to the line - of - sight of observers than the agns with old stellar populations , while the starlight components from the host galaxies are believed to be isotropic .
the revealed trend can be alternatively driven by the activity of the central smbh .
in fact , there is many observational evidence supporting that the [ ] line blue asymmetries and blueshifts are correlated with @xmath9 or eigenvector - i space ( e.g. , boroson & green 1992 ; boroson 2002 ) . ] in typical type i agns ( see the description and citations in section 1 ) .
the relationship between the [ ] line profile and @xmath9 is examined in the current narrow - line agn sample to present a complete understanding .
similar as the previous studies of narrow - line agns , the parameter @xmath115}}/\sigma_*^4 $ ] ( e.g. , heckman et al .
2004 ) is used as a proxy of @xmath116 , where @xmath115}}$ ] and @xmath94 is the [ ] @xmath0 line luminosity and the bulge velocity dispersion , respectively .
the transition galaxies are excluded from the subsequent calculations because of the possible contamination caused by regions .
the bolometric luminosity @xmath117 is transformed from @xmath115}}$ ] through the bolometric correction @xmath118}}\approx3500 $ ] ( see heckman et al .
( 2004 ) for details ) . by assuming the balmer decrement for the standard case b recombination and the galactic extinction curve with @xmath26 , the [ ] line luminosity
is corrected for the local extinction that is inferred from the narrow - line ratio @xmath119 .
the blackhole mass is estimated from the @xmath2 calibration : @xmath120 ( tremaine et al . 2002 ) , in which the velocity dispersion is measured for each spectrum through our pca fittings .
the galaxies with @xmath121 are removed from the analysis because the sdss instrumental resolution is @xmath122 .
the results are shown in figure 8 .
the left panel presents an anti - correlation between @xmath115}}/\sigma_*^4 $ ] and @xmath57 .
larger the eddington ratio , and younger the stellar population , which is consistent with the previous studies ( e.g. , kewley et al .
2006 ; kauffmann et al .
the middle panel plots @xmath115}}/\sigma_*^4 $ ] as a function of the parameter @xmath45 . when compared with figure 5
, one can find a similar trend in the @xmath115}}/\sigma_*^4 $ ] vs. @xmath45 plot as that in the @xmath57 vs. @xmath45 plane .
agns with large amount of blue asymmetry tend to have large @xmath9 .
a much stronger trend can be identified in the right panel between @xmath115}}/\sigma_*^4 $ ] and the parameter @xmath47 .
the trend indicates a fact that agns with larger eddington ratios tend to be associated with stronger [ ] broad components .
however , we fail to find a significant trend between @xmath115}}/\sigma_*^4 $ ] and the bulk relative velocity shifts @xmath56 .
spearman rank - order tests are performed to show either the eddington ratio or the stellar population is intrinsically related with the line asymmetry . the resulted correlation coefficient matrix is listed in table 5 .
all the entries in the table have a probability of null correlation @xmath103 .
although both @xmath123 and @xmath57 are correlated with @xmath45 at comparable significance levels , @xmath47 is more strongly correlated with @xmath123 than with @xmath57 .
the stronger correlation therefore indicates that the trend between the line asymmetry and @xmath57 is likely driven physically by the evolution of the eddington ratio ( see below ) .
the trend that agns with more significant blue asymmetry tend to be associated with younger stellar populations provides a piece of direct evidence supporting the co - evolution of the nlr kinematics and the growth of the host galaxy .
agns with stronger outflows tend to be at their earlier evolutionary stage as inferred from the stellar population ages .
the analysis in the above section further suggests that the trend is likely driven by @xmath9 .
the important evolutionary role of @xmath9 has been frequently proposed in recent studies ( e.g. , wang & wei 2008 , 2010 ; kewley et al .
2006 ; heckman et al .
2004 ) . putting these pieces together
yields an improved co - evolution scenario in which agn likely evolve from a high-@xmath124 state with strong outflow to a low-@xmath124 state with weak outflow as the circumnuclear stellar population continually ages .
this evolutionary scenario is consistent with the current understandings of agns .
lieghly et al .
( 1997 ) reported evidence of relativity outflows in three nls1s . in comparison with typical broad - line seyfert galaxies , nls1s are `` young '' agns ( mathur 2000 ) typical of smaller @xmath3 , higher @xmath124 , larger [ ] emission - line asymmetries , younger circumnuclear stellar populations possibly at post - starburst phase , and enhanced star formations as recently revealed by the observation of _
spitzer ( e.g. , boroson 2005 ; zamanov et al .
2002 ; boroson & green 1992 ; wang et al .
1996 ; boller et al .
1996 ; sulentic et al .
2000 ; boroson 2002 ; xu et al . 2007 ; zhou et al .
2005 ; wang & wei 2006 ) .
in addition to the observational ground , the feedback from central agn is frequently involved in the modern numerical and sim - analytical galaxy evolution models to quench the star formation and to blow the gas or dust away through the powerful agn s wind ( e.g. , springel et al .
2005 ; di matteo et al . 2005 ; kauffmann & heckman 2008 ; fabian 1999 ; hopkins et al .
2005 , 2008a , b ; croton et al . 2006 ; somerville et al . 2008 ; khalatyan et al .
2008 ) . the numerical simulation done by hopkins et al .
( 2005 ) suggests a scenario that the powerful agn s wind is required to occur in the young agn phase in which the central smbh is heavily obscured by the starforming gas and dust .
_ transition galaxies are the narrow emission - line galaxies located between the theoretical and empirical demarcation lines in the [ ] /h@xmath7 versus [ ] /h@xmath24 diagnostic diagram ( kauffmann et al . 2003 ; kewley et al .
their emission - line properties are explained by the mixture of the contributions from both star formations and typical agns ( e.g. , kewley et al .
2006 ; ho et al .
1993 , see recent review in ho 2008 ) .
our spectral profile analysis indicates that the transition galaxies differs from the seyfert galaxies in their [ ] emission - line profiles .
the transition galaxies systematically show weaker [ ] broad wings , and narrower [ ] line widths than the seyfert galaxies , which implies that the two kinds of objects are different from each other in their mass outflow kinematics .
the line profile comparison allows us to suspect that the transition galaxies `` bridge '' the agn - starburst co - evolution from early starburst - dominated phase to late agn - dominated phase ( e.g. , yuan et al .
2009 ; wang & wei 2006 ; schawinski et al .
2007 , 2009 ) .
the evolutionary role of the transition galaxies has been reported in the recent studies by many authors . with the large sdss spectroscopic database ,
transition galaxies are found to be systematically associated with younger stellar populations than seyfert galaxies ( e.g. , kewley et al .
2006 ; wild et al . 2007 ; and also see figure 6 in this paper ) .
westoby et al .
( 2007 ) found that the distribution of the h@xmath24 equivalent widths of the transition galaxies peaks at the valley between the red and blue sequences . by analyzing the mm - wavelength observation taken by the 30 m iram , schawinski et al .
( 2009 ) suggested that molecular gas reservoir is destructed by the agn feedback at early agn+starforming transition phase .
we systematically examined the evolutionary issue of the [ ] @xmath0 emission - line profile by using a large sample of narrow emission - line galaxies selected from the mpa / jhu sdss dr7 value - added catalog .
the sample is separated into three sub - samples ( i.e. , star - forming galaxies , transition galaxies and seyfert galaxies / liners ) basing upon the line ratios given in the catalog .
two shape parameters , skewness ( @xmath45 ) and kurtosis ( @xmath47 ) , are additionally used to quantify the profile deviation from a pure gaussian .
our analysis indicates that a ) the transition galaxies are systematically associated with narrower line widths and with weaker [ ] broad wings than the agns ; b ) the agns with stronger blue asymmetries tend to be associated with younger stellar populations .
the evolutionary significance of the [ ] line profile suggests a co - evolution of the outflow feedback and the agn s host galaxy .
we would like to thank the anonymous referees for his / her constructive comments in improving the paper .
this study was supported by the national science foundation of china ( under grant 10803008 ) , and by the national basic research program of china ( grant 2009cb824800 ) .
this study uses the sdss archive data that was created and distributed by the alfred p. sloan foundation .
a toy model of emission - line profile is constructed here to illustrate the effect of the instrumental resolution on the measured line shape parameters .
the model is constructed by assuming an observed [ ] emission line profile that is prescribed by the sum of two pure gaussian profiles : @xmath125 , where @xmath126 is a normalized gaussian distribution with the peak at zero and the variance of @xmath127 , @xmath128 is the same distribution but with the peak shifted at @xmath129 and the variance of @xmath130 , and @xmath24 is a parameter much less than unit . with this two gaussian components model , @xmath126 represents the narrow core of the emission , and @xmath131 the shifted , low - contrast broad wing ( we require @xmath132 for avoiding double peaked profile ) .
the normalizations of the two distributions yield @xmath133 . the first moment ( i.e. , the line centroid )
is written as @xmath134 , which is not affected by the instrumental resolution . with the first moment ,
the high - order moments are written as @xmath135 the second moment of the total line profile is therefore inferred to be approximately the linear combination of the square of the second moments of the two distributions : @xmath136 . in the context of the toy model , the linear combination means that the instrumental resolution @xmath23 can be removed from the measured values according to the equation @xmath137 , which means that the removal depends on the relative strength of the low - contrast gaussian profile , i.e. , the parameter @xmath24 .
the third moment of the total profile can be trivially obtained through the integration @xmath138 . by ignoring the terms with high - orders of @xmath24 ,
the third moment approximately equals to @xmath139 .
the result suggests that the measured third moment depends not only on the relative shift of the low - contrast gaussian profile , but also on the relative strength denoted by the parameter @xmath24 .
although the second term in the above result means that the third moment does not depend on the instrumental resolution in our toy model at the first order approximation : @xmath140 , the definition of the dimensionless shape parameter @xmath45 introduces a non - linear dependence on the instrumental resolution in the denominator ( see eq . 2 in the main text ) .
by adopting the same approach , the fourth moment is calculated to be @xmath141 , where one sees that the second and third terms introduce again a non - linear dependence on the instrumental resolution . with this toy model , the galaxies
whose [ ] @xmath15007 lines are narrower than 2@xmath23 are then dropped out from our statistical analysis in order to avoid the effect of the instrumental resolution on the measured line shape parameters .
abazajian , k. n. , et al .
2009 , , 182 , 543 aoki , k. , kawaguchi , t. , & ohta , k. 2005 , , 618 , 601 antonucci , r. r. j. 1993 , , 31 , 473 arribas , s. , mediavilla , e. , & garcia - lorenzo , b. 1996 , , 463 , 509 baldwin , j. a. , phillips , m. m. , & terlevich , r. 1981 , pasp , 93 , 5 balogh , m. l. , et al . , 1999 , , 527 , 54 bian , w. h. , yuan , q. r. , & zhao , y. h. 2005 , , 364 , 187 binney , j. , & merrifield , m. 1998 , galactic astronomy , princeton , new jersey boller , t. , brandt , w. n. , & fink , h. 1996 , , 305 , 53 boroson , t. a. 2002 , , 565 , 78 boroson , t. a. 2003 , , 585 , 647 boroson , t. a. 2005 , , 130 , 381 boroson , t. a. , & green , r. f. 1992 , , 80 , 109 bromley , b. c. , et al .
1998 , , 505 , 25 bruzual , a. g. 1983 , , 273 , 105 bruzual , g. , & charlot , s. 2003 , , 344 , 1000 cardelli , j. a. , clayton , g. c. , & mathis , j. s. , 1989 , , 345 , 245 christopoulou , p. e. , holloway , a. j. , steffen , w. , mundell , c. g. , thean , a. h. c. , goudis , c. d. , meaburn , j. , & pedlar , a. 1997 , , 284 , 385 crenshaw , d. m. , & kraemer , s. 2000 , , 532 , 101 crenshaw , d. m. , et al .
2000 , , 120 , 1731 crenshaw , d. m. , schmitt , h. r. , kraemer , s. b. , mushotzky , r. f. , & dunn , j. p. 2010
, , 708 , 419 croton , d. j. , et al .
2006 , , 365 , 11 das , v. , et al .
2005 , , 130 , 945 das , v. , crenshaw , d. m. , kraemer , s. b. , & deo , r. p. 2006
, , 132 , 620 das , v. , crenshaw , d. m. , & kraemer , s. b. 2007 , , 656 , 699 davies , r. , mueller sanchez , f. , genzel , r. , tacconi , l. , hicks , e. , friedrich , s. , & sternberg , a. 2007 , , 671 , 1388 de robertis , m. m. , & osterbrock , d. e. 1984 , , 286 , 171 di matteo , t. , springel , v. , & hernquist , l. 2005 , nature , 433 , 604 elitzur , m. 2007 , in asp conf .
373 , the central engine of active galactic nuclei , ed .
l. c. ho & j. -m .
wang ( san francisco , ca : asp ) , 415 erkens , u. , appenzeller , i. , & wagner , s. 1997 , , 323 , 707 fabian , a. c. 1999 , , 308 , l39 ferrarese , l. , & merritt , d. 2000 , , 539 , 9 fischer , t. c. , crenshaw , d. m. , kraemer , s. b. , schmitt , h. r. , mushotsky , r. f. , & dunn , j. p. 2011
, , 727 , 71 fischer , t. c. , crenshaw , d. m. , kraemer , s. b. , schmitt , h. r. , & trippe , m. l. 2010 , , 140 , 577 gebhardt , k. et al .
2000 , , 539 , 13 glazebrook , k. , offer , a. , r. , & deeley , k. 1998 , , 492 , 98 granato , g. l. , de zotti , g. , silva , l. , bressan , a. , & danese , l. 2004 , , 600 , 580 greene , j. e. , & ho , l. c. 2005 , , 627 , 721 grupe , d. , beuermann , k. , mannheim , k. , & thomas , h , -c .
1999 , , 350 , 805 grupe , d. , & mathur , s. 2004 , , 606 , 41 haring , n. , & rix , h. w. 2004 , , 604 , 89 hasinger , g. , miyaji , t. , & schmidt , m. 2005 , , 441 , 417 heckman , t. m. 1981 , , 250 , 59 heckman , t. m. , & kauffmann , g. 2006 , newar , 50 , 677 heckman , t. m. , kauffmann , g. , brinchmann , j. , et al . 2004 . , 613 , 109 ho , l. c. 2008 , , 46 , 475 ho , l. c. , filippenko , a. v. , & sargent , w. l. w. 1993 , , 417 , 63 hopkins , p. f. 2011 , astro - ph / arxiv:1101.4230 hopkins , p. f. , hernquist , l. , martini , p. , cox , t. j. , robertson , b. , di matteo , t. , & springel , v. 2005 , , 625 , 71 hopkins , p. f. , bundy , k. , hernquist , l. , & ellis , r. s. 2007 , , 659 , 976 hopkins , p. f. , cox , t. j. , keres , d. , & hernquist , l. 2008a , , 175 , 390 hopkins , p. f. , hernquist , l. , cox , t. j. , & keres , d. 2008b , , 175 , 356 hu , c. , wang , j. m. , ho , l. c. , chen , y. m. , zhang , h. t. , bian , w. h. , & xue , s. j. 2008 , , 687 , 78 hutchings , j. b. , et al .
1998 , , 492 , 115 kaiser , m. e. , et al .
2000 , , 528 , 260 kauffmann , g. , & heckman , t. m. 2009 , , 2009 , 397 , 135 kauffmann , g. , heckman , t. m. , tremonti , c. , et al .
2003a , , 346 , 1055 kauffmann , g. , heckman , t. m. , white , s. d. m. , et al .
2003b , , 341 , 54 kauffmann , g. , heckman , t. m. , white , s. d. m. , et al .
2003c , , 341 , 33 kauffmann , g. , heckman , t. m. , de lucia , g. , brinchmann , j. charlot , s. , tremonti , c. , white , s , d. m. , & brinkmann , j. 2006 , , 367 , 1394 kewley , l. j. , dopita , m. a. , sutherland , r. s. , heisler , c. a. , & trevena , j. 2001 , , 556 , 121 kewley , l. j. , groves , b. , kauffmann , g. , & heckman , t. 2006 , , 372 , 961 khalatyan , a. , cattaneo , a. , schramm , m. , gottlber , s. , steinmetz , m. , & wisotzki , l. 2008 , , 387 , 13 komossa , s. , & xu , d. 2007 , , 677 , 33 komossa , s. , xu , d. , zhou , h. , storchi - bergmann , t. , & binette , l. 2008 , , 680 , 926 leighly , k. m. , mushotzky , r. f. , nandra , k. , & forster , k. 1997 , , 489 , 25 magorrian , j. , et al .
1998 , , 115 , 2285 mathur , s. 2000 , , 314 , l17 mathur , s. , & grupe , d. 2005 , , 633 , 688 marziani , p. , sulentic , j. w. , dultzin - hacyan , d. , calvani , m. , & moles , m. 1996 , , 104 , 37 marziani , p. , sulentic , j. w. , zamanov , r. , calvani , m. , dultzin - hacyan , d. , bachev , r. , & zwitter , t. 2003 , , 145 , 199 nandra , k. , laird , e. s. , & steidel , c. c. 2005 , , 360 , l39 nelson , c. h. 2000 , , 544 , 91 nelson , c. , plasek , a. , thompson , a. , gelderman , r. , & monroe , t. 2004 , aspc , 311 , 83 nelson , c. h. , weistrop , d. , hutchings , j. b. , crenshaw , d. m. , gull , t. r. , kaiser , m. e. , kraemer , s. b. , & lindler , d. 2000 , , 531 , 257 nelson , c. h. , & whittle , m. 1996 , , 465 , 96 osterbrock , d. e. astrophysics of gaseous nebulae and active galactic nuclei , ( mill valley : univ .
science books ) phillips , m. m. 1976 , , 208 , 37 ruiz , j. r. , crenshaw , d. m. , kraemer , s. b. , bower , g. a. , gull , t. r. , hutchings , j. b. , kaiser , m. e. , & weistrop , d. 2001 , , 122 , 2961 schawinski , k. , thomas , d. , sarzi , m. , maraston , c. , kaviraj , s. , joo , s. , yi , s. k. , & silk , j. 2007 , , 382 , 1415 schawinski , k. , virani , s. , simmons , b. , urry , c. m. , treister , e. , kaviraj , s. , & kushkuley , b. 2009 , , 692 , 19 schlegel , d. , finkbeiner , d. p. , davis , m. 1998 , , 500 , 525 schlesinger , k. , pogge , r. w. , martini , p. , shields , j. c. , & fields , d. 2009 , , 699 , 857 sergeev , s. g. , pronik , v. i. , malkov , yu . f. , & chuvaev , k. k. 1997 , , 320 , 405 shankar , f. , bernardi , m. , & haiman , z. 2009 , , 694 , 867 silverman , j. d. , et al .
2008 , , 679 , 118 somerville , r. s. , hopkins , p. f. , cox , t. j. , robertson , b. e. , & hernquist , l. 2008 , , 391 , 481 storchi - bergmann , t. , wilson , a. s. , & baldwin , j. a. 1992 , , 396 , 45 spergel , d. n. , et al .
2003 , , 148 , 175 springel , v. , white , s. d. m. , jenkins , a. , et al .
2005 , nature , 435 , 629 sulentic , j. w. , marziani , p. , & dultzin - hacyan , d. 2000 , , 38 , 521 tadhunter , c. , wills , k. , morganti , r. , oosterloo , t. , & dickson , r. 2001 , , 327 , 227 tremaine , s. , gebhardt , k. , bender , r. , et al .
2002 , , 574 , 740 tremonti , c. a. , et al .
2004 , , 613 , 898 veilleux , s. , & osterbrock , d. e. 1987 , , 63 , 295 veilleux , s. , cecil , g. , & bland - hawthorn , j. 2005 , , 43 , 769 veron , p. , goncalves , a. c. , & veron - cetty , m. -p .
2002 , , 384 , 826 wang , j. , & wei , j. y. 2006 , , 648 , 158 wang , j. , & wei , j. y. 2008 , , 679 , 86 wang , j. , & wei , j. y. 2009 , , 696 , 741 wang , j. , & wei , j. y. 2010 , , 719 , 1157 wang , j. , wei , j. y. , & he , x. t. 2005 , newa , 10 , 353 wang , j. , wei , j. y. , & he , x. t. 2006 , , 638 , 106 wang , t. , brinkmann , w. , & bergeron , j. 1996 , , 309 , 81 wang , t. , & lu , y. 2001 , , 377 , 52 westoby , p. b. , mundell , c. g. , & baldry , i. k. 2007 , , 382 , 1514 whittle , m. 1985 , , 213 , 1 whittle , m. 1992a , , 387 , 109 whittle , m. 1992b , , 387 , 121 wilson , a. s. , & heckman , t. m. 1985 , in astrophysics of active galaxies and quasi - stellar objects , ed .
j. s. miller ( mill valley : university science books ) wild , v. , et al .
2007 , , 381 , 543 wild , v. , heckman , t. m. , & charlot , s. 2010 , , 405 , 933 worthey , g. , & ottaviani , d. l. 1997 , , 111 , 377 xu , d. w. , & komossa , s. 2009 , , 705 , 20 xu , d. w , komossa , s. , zhou , h. y. , wang , t. g. , & wei , j. y. 2007 , , 670 , 60 york , d. g. , et al .
2000 , , 120 , 1579 yu , q. j. , & tremaine , s. 2002 , , 335 , 965 yuan , t. -t . , kewley , l. j. , & sanders , d. b. 2010 , , 709 , 884 zamanov , r. , marziani , p. , sulentic , j. w. , calvani , m. , dultzin - hacyan , d. , & bachev , r. 2002 , , 576 , 9 zhou , h. y. , wang , t. g. , dong , x. b. , wang , j. , & lu , h. 2005 , mem .
soc . astron .
italiana , 76 , 93 ccccc property & agns & transitions & sf galaxies/@xmath142 & sf galaxies/@xmath143 + ( 1 ) & ( 2 ) & ( 3 ) & ( 4 ) & ( 5 ) + numbers & 2333 & 793 & 190 & 5843 + @xmath144 & -0.167 & -0.170 & -0.09 & 0.01 + @xmath145 & -0.165 & -0.167 & -0.08 & 0.01 + @xmath146 & 0.31 & 0.28 & 0.32 & 0.14 + ccccc property & agns & transitions & sf galaxies/@xmath142 & sf galaxies/@xmath143 + ( 1 ) & ( 2 ) & ( 3 ) & ( 4 ) & ( 5 ) + numbers & 2333 & 793 & 190 & 5843 + @xmath147 & 3.00 & 2.79 & 2.84 & 2.61 + @xmath148 & 2.83 & 2.64 & 2.56 & 2.58 + @xmath149 & 0.70 & 0.56 & 1.01 & 0.43 + cccc sub - sample & agns & transitions & sf galaxies/@xmath142 + & ( 1 ) & ( 2 ) & ( 3 ) + agns & - & 0.034(0.48 ) & 0.159(@xmath150 ) + transitions & - & - & 0.164(@xmath151 ) + sf galaxies/@xmath142 & - & - & - + cccc sub - sample & agns & transitions & sf galaxies/@xmath142 + & ( 1 ) & ( 2 ) & ( 3 ) + agns & - & 0.154(@xmath152 ) & 0.214(@xmath153 ) + transitions & - & - & 0.133(@xmath154 ) + sf galaxies/@xmath142 & - & - & - + | the agn - host co - evolution issue is investigated here by focusing on the evolution of the [ ] @xmath0 emission - line profile . in order to simultaneously measure both [ ] line profile and circumnuclear stellar population in individual spectrum ,
a large sample of narrow emission - line galaxies is selected from the mpa / jhu sdss dr7 catalog . by requiring
that 1 ) the [ ] line signal - to - noise ratio is larger than 30 , 2 ) the [ ] line width is larger than the instrumental resolution by a factor of 2 , our sample finally contains 2,333 seyfert galaxies / liners ( agns ) , 793 transition galaxies , and 190 starforming galaxies . in additional to the commonly used profile parameters ( i.e. , line centroid , relative velocity shift and velocity dispersion ) , two dimensionless shape parameters , skewness and kurtosis , are used to quantify the line shape deviation from a pure gaussian function .
we show that the transition galaxies are systematically associated with narrower line widths and weaker [ ] broad wings than the agns , which implies that the kinematics of the emission - line gas is different in the two kinds of objects . by combining the measured host properties and line shape parameters , we find that the agns with stronger blue asymmetries tend to be associated with younger stellar populations
. however , the similar trend is not identified in the transition galaxies .
the failure is likely resulted from a selection effect in which the transition galaxies are systematically associated with younger stellar populations than the agns .
the evolutionary significance revealed here suggests that both nlr kinematics and outflow feedback in agns co - evolve with their host galaxies . |
using the language of non - relativistic quantum field theory , the csl model is formulated in terms of a stochastic differential equation in the fock space associated with the system @xcite .
given different types of particles , where the type @xmath2 has mass @xmath3 , the mass - proportional csl model @xcite is defined by @xmath4 { \mathrm{d}}w_t({\bf y } ) \right .
\nonumber \\ & & \left.- \frac{\gamma}{2 m^2_0 } \int { \mathrm{d}}{\bf y } [ \hat{m}({\bf y})-\langle m({\bf y } ) \rangle_t ] ^2 { \mathrm{d}}t \right ] { \vert \varphi_t \rangle},\end{aligned}\ ] ] where @xmath5 is the standard quantum hamiltonian , @xmath6 , @xmath7 is a reference mass ( usually the mass of a nucleon ) and @xmath8 is a smeared mass density operator : @xmath9 here , @xmath10 and @xmath11 are , respectively , the creation and the annihilation operator of a particle of type @xmath2 in the point @xmath12 , while @xmath13 is an ensemble of independent wiener processes , one for each point in space .
the model is characterized by two parameters : @xmath14 , which sets the strength of the collapse process , and @xmath15 , which determines the threshold above which spatial superpositions are suppressed .
the choice of the numerical values for these parameters is of course ultimately dictated by the agreement with the experimental data ; the originally proposed values are @xcite @xmath16 and @xmath17 .
-3cm*(a ) * + the mass density operators @xmath18 in eq.([eq : sdecsl ] ) induce a collapse of the wavefunction @xmath19 around the common eigenvectors of the position operators of the particles composing the system @xcite .
hence , the asymptotic wavefunction is sharply localized around defined positions , excluding possible spatial superpositions .
the collapse rate for a microscopic system is given by @xmath20 .
such a small value guarantees that the spatial localization due to the noise field can be safely neglected if a microscopic system is taken into account .
now instead , consider a macroscopic rigid body in a superposition of two states distant more than @xmath15 .
its center of mass collapses with an effective rate @xcite @xmath21 where @xmath22 is the number of constituents of the body contained in a volume @xmath23 and @xmath24 denotes how many such volumes are held in the macroscopic body .
this relation clearly shows the amplification mechanism , which is at the basis of every collapse model .
the localization induced by the noise field grows with the size of the system , so that the center of mass of any macroscopic object behaves , for all practical purposes , according to classical mechanics .
the peculiar property of the csl model is the quadratic dependence of the rate @xmath25 on the number of constituents , which is a direct consequence of the action of the noise field on identical particles @xcite .
the main features of the csl model are summarized in fig.[fig:1 ] , where we represent the time evolution of the position probability distribution of one particle , which is initially in a superposition of two gaussian states .
the wavefunction is subjected continuously to the action of the noise , which suppresses the superposition between the two gaussians , leading to a gaussian state localized around one of the two initial peaks , in a time scale fixed by the collapse rate , see fig.[fig:1 ] * ( a ) * and * ( b)*. the diffusive nature of the dynamics in the csl model is clearly illustrated by the time - evolution of the position variance , see fig.[fig:1 ] * ( c)*. a relevant drawback of the original csl model , as well as of most collapse models , is that the average kinetic energy of the quantum system diverges on the long time scale .
the model predicts that the energy of a particle with mass @xmath26 increases linearly in time with a rate @xmath27 as will become clear by the following analysis , the reason for such an energy increase is precisely due to the absence of dissipation within the model .
the noise acts like an infinite temperature background , steadily increasing the energy of the system .
now that we have clarified the problem of the csl model we want to work out , as well as the features that must be preserved , we are in the position to formulate a new , dissipative csl model . as for the original model , the most compact way to do so , is to define a proper stochastic differential equation .
specifically , we consider the following non - linear stochastic differential equation : @xmath28 { \mathrm{d}}w_t({\bf y } ) \right . \\ & & \left.- \frac{\gamma}{2 m^2_0 } \int { \mathrm{d}}{\bf y } [ [ \hat{\mathbb{l}}^{\dag}({\bf y})\hat{\mathbb{l}}({\bf y})+r^2_t({\bf y})-2r_t({\bf y})\hat{\mathbb{l}}({\bf y } ) ] { \mathrm{d}}t \right ] { \vert \varphi_t \rangle } , \nonumber\end{aligned}\ ] ] with @xmath29 and @xmath30 where @xmath31 the inclusion of dissipation calls for the introduction of a new parameter , @xmath32 , with the dimension of a velocity . crucially , this parameter is related to the temperature of the noise field , as it will be shown later ( see eq.([eq : t ] ) ) , where the numerical choice of @xmath32 will be also discussed .
the structure of the stochastic differential equation ( [ eq : sdecsld ] ) generalizes that of eq.([eq : sdecsl ] ) to the case of non self - adjoint operators @xcite . indeed , for @xmath33 , so that @xmath34 , one recovers the original csl model .
the physical meaning of the operator @xmath35 in eq.([eq : ly ] ) is better understood by taking into account also its momentum representation .
one has @xmath36 where @xmath37 and @xmath38 are , respectively , the creation and annihilation operator of a particle of the type @xmath2 with momentum @xmath39 . by eqs.([eq : ly ] ) and ( [ eq : lymom ] ) , we see that the action of the collapse noise can be compared to that of an external potential which depends not only on the position , but also on momentum of the system , thus inducing dissipation . in particular , since the exchanged momentum @xmath40 in the spatial direction @xmath41 has a gaussian distribution peaked around @xmath42 , the action of the noise will suppress high momenta , so that the mean kinetic energy of the system , as well as the mean momentum , is subject to relaxation .
this is explicitly shown in sec .
[ sec : era ] .
indeed , contrary to any external field , the collapse noise induces an anti - hermitian coupling with matter , which is necessary in order to induce localization .
in addition , the introduction of dissipation also leads to an hermitian contribution to the coupling , see sec . [
sec : haa ] . in order to study the solutions and properties of eq.([eq : sdecsld ] ) , it is often convenient to deal with an equivalent linear stochastic differential equation @xcite . here , we briefly sketch the standard procedure which provides the linear stochastic differential equation associated with the non - linear one in the form given by eq.([eq : sdecsld ] ) . for a complete treatment ,
the reader is referred to @xcite .
consider a non - linear stochastic differential equation as eq.([eq : sdecsld ] ) . recall that @xmath43 denotes an ensemble of independent wiener processes defined on a common probability space @xmath44
let @xmath45 be the ensemble of processes given by @xmath46 where @xmath47 has been defined after eq.([eq : sdecsld ] ) .
now , by means of the girsanov theorem one can define a new probability @xmath48 on @xmath49 such that @xmath50 is an ensemble of wiener processes under @xmath48 @xcite .
in addition , one can define a random vector @xmath51 such that @xmath52 where @xmath19 satisfies eq.([eq : sdecsld ] ) , while @xmath53 satisfies @xcite @xmath54 { \vert { \psi}_t \rangle}.\end{aligned}\ ] ] this is the linear stochastic differential equation associated with the dissipative csl model . in this paragraph , we show that eq.([eq : sdecsld ] ) can be written in a way such that the coupling between the classical collapse noise and quantum matter is made explicit .
in particular , the introduction of dissipation within the csl model leads to an hermitian contribution to the coupling of the collapse noise with quantum matter , in addition to a contribution in the usual ( for collapse models ) anti - hermitian form @xcite .
it is here convenient to use the stratonovich formalism @xcite and to consider the decomposition of @xmath35 into its hermitian and anti - hermitian part , @xmath55 , with @xmath56 and @xmath57 self - adjoint operators .
thus , the non - linear stochastic equation given by eq.([eq : sdecsld ] ) leads to @xmath58 { \vert \varphi_t \rangle},\label{eq : eq}\end{aligned}\ ] ] where @xmath59 is the white - noise field , which can be formally written as @xmath60 and satisfies the relations @xmath61 = 0 \qquad \mathbb{e}[w({\bf x},t ) w({\bf y},t ' ) ] = \delta({\bf x } - { \bf y})\delta(t - t'),\ ] ] @xmath62 being the stochastic average under the reference probability @xmath63 .
moreover , @xmath64 is an hermitian contribution coming from the passage to the stratonovich formalism and reads @xmath65 on the other hand , @xmath66 includes the non - linear contributions preserving the norm of the state vector and is given by @xmath67 - \frac{\sqrt{\gamma}}{m_0 } \int { \mathrm{d}}{\bf y } \langle\hat{\mathbb{l}}^{(a)}({\bf y})\rangle_t w({\bf y},t ) ; \end{aligned}\ ] ] compare with eq.(7.43 ) in @xcite .
equation ( [ eq : eq ] ) describes the coupling between the classical field @xmath59 and quantum matter .
now , since @xmath35 is not a self - adjoint operator such a coupling has an hermitian , as well as an anti - hermitian contribution .
note that in the original csl model the collapse noise is coupled with matter only via an anti - hermitian term @xcite . to be explicit , eq.([eq : lymom ] ) implies @xmath68 and thus @xmath69 and @xmath70 of course , @xmath71 for @xmath72 .
in this section , we investigate the energy behavior of a system subjected to the collapse noise in our extended model .
we deal with the master equation implied by the stochastic differential equation given by eq.([eq : sdecsld ] ) . after presenting the equation in a second - quantization formalism ,
we describe the corresponding operators in the case of a fixed number of particles . in particular , by focusing on the one - particle case ,
we show explicitly the exponential relaxation of the energy to a finite value .
the non - linear , as well as the linear , stochastic differential equation fully fixes the collapse model we are defining here .
however , one is often interested in the predictions of the model related with the statistical mean of some physical quantity , @xmath73,\ ] ] where , as usual , @xmath19 is the stochastic state of the system satisfying eq.([eq : sdecsld ] ) .
for this reason , it can be convenient to deal directly with the evolution of the average state @xmath74,\ ] ] so that one recovers the usual relation @xmath75 by using the it calculus , it is easy to see that eq.([eq : sdecsld ] ) implies the following master equation : @xmath76 + \frac{\gamma}{m^2_0 } \int { \mathrm{d}}{\bf y } \left[\hat{\mathbb{l}}({\bf y})\hat{\rho}(t)\hat{\mathbb{l}}^{\dag}({\bf y } ) \right . \nonumber\\ & & \left .
- \frac{1}{2}\left\{\hat{\mathbb{l}}^{\dag}({\bf y } ) \hat{\mathbb{l}}({\bf y } ) , \label{eq : mecsldiss } \hat{\rho}(t)\right\ } \right].\end{aligned}\ ] ] this is a lindblad master equation @xcite , indicating that we are in the presence of a time - homogeneous markovian dynamics .
the lindblad operators are the same operators as those appearing in the stochastic differential equation defining the model , see eq.([eq : ly ] ) or eq.([eq : lymom ] ) .
the expressions of the lindblad operators @xmath35 restricted to a sector of the fock space with a fixed number of particles is easily obtained as follows .
let us assume for simplicity that we have @xmath77 particles of the same type and mass @xmath26 .
the corresponding restriction of @xmath35 reads @xmath78 where @xmath79 and @xmath80 are , respectively , the position and momentum operator of the @xmath81-th particle and @xmath82 is the constant given by eq.([eq : kj ] ) with @xmath83 .
indeed , consider the hilbert space @xmath84 and the corresponding fock space @xmath85 , where @xmath86 denotes the symmetric or antisymmetric part of the tensor product @xmath87 , @xmath77 times .
now consider the operator on @xmath88 given by @xcite @xmath89 where @xmath90 is a single - particle operator on @xmath84 , with @xmath91 and @xmath92 , respectively , position and momentum operators on @xmath84 .
hence , the restriction of @xmath93 on the @xmath77-particle sector of the fock space reads @xmath94 @xmath95 and @xmath96 being the position and momentum operator of the @xmath81-th particle .
the relation between eq.([eq : c ] ) and eq.([eq : un ] ) is indeed the same as that between eq.([eq : lymom ] ) and eq.([eq : lybb ] ) .
if we further restrict to the case of a single free particle with mass @xmath26 , we end up with the following master equation for the one - particle average state @xmath97 @xmath98 + \frac{\gamma m^2}{(2 \pi \hbar)^3 m^2_0 } \nonumber\\ & & \left ( \int { \mathrm{d}}{\bf { q } } \ , e^{\frac{i}{\hbar } { \bf { q } } \cdot \hat{{\bf x } } } l({\bf q } , \hat{\bf p } ) \hat{\rho}^{(1)}(t ) l({\bf q } , \hat{\bf p})e^{-\frac{i}{\hbar } { \bf { q } } \cdot \hat{{\bf x}}}\right . \nonumber\\ & & \left .
-\frac{1}{2}\left\{l^2({\bf q } , \hat{\bf p } ) , \hat{\rho}^{(1)}(t)\right\}\right),\label{eq : mecsldiss2}\end{aligned}\ ] ] with @xmath99 let us note that the inclusion of dissipation in the csl model preserves the invariance under translations of the system s evolution , but it breaks the invariance under boosts , as directly seen by the master equation ( [ eq : mecsldiss2 ] ) @xcite . nevertheless , the characterization of the overall dynamics by means of a proper first - principle underlying theory , which involves both the sources of the collapse noise and the quantum systems affected by it , should allow to recover a fully covariant description ; see also the discussion in the next paragraph .
the master equation for the system s statistical operator provides us with the evolution equation of the mean kinetic energy of the one - particle system , @xmath100 exploiting the translation covariance of eq.([eq : mecsldiss2 ] ) @xcite one gets @xmath101 this equation is solved by @xmath102 with relaxation rate @xmath103 and asymptotic kinetic energy @xmath104 as expected , now we do have dissipation . the mean energy of the system can decrease as a consequence of the action of the noise .
moreover , even if the energy grows , there is an upper threshold value above which it can not increase .
the long - time energy divergence is now avoided @xcite .
this is precisely the result we wanted and the most natural way to interpret it is to say that the collapse noise has a finite temperature toward which the system thermalizes @xcite .
explicitly , @xmath105 corresponds to a noise temperature @xmath106 where we used eq.([eq : kj ] ) and @xmath107 is the boltzmann constant . the original csl model is recovered in the limit @xmath108 : the noise acts like an infinite temperature background , which explains the energy divergence . the temperature of the noise in eq.([eq : t ] )
does not depend on the mass of the system , which is a very important feature of our model .
in addition , the state of the system actually equilibrates to the canonical gibbs state , see appendix [ sec : adc ] .
these hallmarks of the evolution induced by eq.([eq : sdecsld ] ) depend substantially on the choice of the operators @xmath109 in eq.([eq : ly ] ) .
it is an open question to identify the entire class of operators satisfying these natural requests . in appendix
[ sec : adc ] , we take into account a physically motivated alternative to the choice made in eq.([eq : ly ] ) , showing how the relaxation dynamics of the resulting collapse model is very similar to that presented here and , in particular , the noise temperature is still given by eq.([eq : t ] ) .
the exponential relaxation of the energy @xmath110 in eq.([eq : ht ] ) is the same as that in the dissipative ghirardi - rimini - weber ( grw ) model recently introduced in @xcite .
this is not surprising , since , as for the case without dissipation , the extended grw and csl models share the same one - particle master equation .
if we think that the collapse model fixed by eq.([eq : sdecsld ] ) describes in a phenomenological way the action of a real physical field filling space , it is now clear how the principle of energy conservation can be reestablished .
the energy gained or lost by the system can be ascribed to an energy exchange with the noise field , as the latter can be influenced back by the presence of the system .
an explicit characterization of this process requires an underlying theory , which has to guarantee the classical nature of the noise field , with its own ( non - quantum ) equations of motion , in order to provide a proper objective collapse of the wavefunction @xcite .
in addition , one can already say that a collapse noise with typical cosmological features would correspond to a low - temperature noise @xcite , at most of the order of few kelvins . by eq.([eq : t ] ) , we see that the noise temperature @xmath111 is in one - to - one correspondence with the new parameter @xmath32 . for instance , @xmath112 ( i.e. @xmath113 for a nucleon ) gives @xmath114 .
hence , more than the specific value of the noise temperature , the important thing is that even in the presence of a low - temperature noise the resulting collapse model can be introduced in a consistent way ; see also the discussion in the next section .
it is worth noting that it is not always possible to properly modify a given collapse model to include dissipation via the action of a low - temperature noise . in @xcite
we discuss how such a modification is not feasible for the disi - penrose model @xcite , in which gravity is involved to provide a phenomenological description of the wavefunction collapse . in our model , every system with a temperature higher than about @xmath115
is cooled by the action of the collapse noise .
thus , we are led to reject the bounds on the collapse rate @xmath1 relying on a balance between the system s heating due to the action of the noise in the original csl model and the cooling due to , e.g. , the universe expansion or the energy radiation .
this is the case for the heating of the protons constituting the intergalactic medium or for the energy accumulation in interstellar dust grains @xcite .
note how , in particular , the heating of the igm provides the second strongest bound to date on the localization rate @xmath1 @xcite .
even more , we expect that cosmological data will put strong bounds on the dissipation parameter @xmath82 ( equivalently , on @xmath32 ) .
the modified long - time behavior of the energy predicted by our model will have to be compared with the constraints coming from such data .
some preliminary results have been obtained for the non - dissipative csl model @xcite .
dissipative effects are expected to play an important role also in the experimental investigation of collapse models via optomechanical systems @xcite , where proper signatures could be visible in the density noise spectrum of the mechanical oscillator , or via the spontaneous photon emission from electrically charged particles @xcite , as the latter is registered over a period of several years . in both situations , we expect that dedicated experiments should allow to restrict the possible values of @xmath82 ; of course , also in relation with the other parameters of the model .
as recalled in sec . [ sec : tcm ] , any proper collapse model is characterized by the amplification mechanism . the localizing action of the collapse noise has to increase with size of the system , which guarantees a consistent description of microscopic and macroscopic systems within a unique theoretical framework . here
, we show that the amplification mechanism holds in our extended model , at least as long as one deals with a macroscopic rigid body .
the description of more complex systems , where the internal dynamics becomes involved , calls for a more detailed specification of the system s evolution @xcite .
we stress that the following considerations are valid also in the case of a low temperature noise .
as anticipated , even for a noise temperature @xmath114 we have effective localization and amplification mechanisms , so that the noise actually induces a classical behavior of the center of mass of macroscopic objects .
finally , the different role of the momentum - dependent localization operators of our model in , respectively , the energy relaxation and the wave - function localization is clarified by comparing the time scales of the two phenomena .
consider a macroscopic object made up of @xmath77 particles of equal mass @xmath26 .
one can neglect the contributions due to the electrons and consider the mass of the proton @xmath116 equal to the mass of the neutron @xmath117 , i.e. @xmath118 .
in addition , we deal with a rigid body , which allows us to decouple the evolution of the center of mass from that of the relative coordinates @xcite .
let @xmath119 be the position operator of the center of mass , while the relative coordinates @xmath120 , @xmath121 , are fixed by @xmath122 , for a suitable matrix with elements @xmath123 .
we neglect the possible rotations of the system : this greatly simplifies the description , without affecting the physical meaning of the results @xcite . by virtue of the rigid - body approximation ,
the relative coordinates are fixed and the center - of - mass momentum @xmath124 is simply proportional to each individual momentum @xmath125 , according to @xmath126 , for each @xmath2 .
moreover , consider a total hamiltonian @xmath127 , given by the sum of two terms associated with , respectively , the center of mass and the relative degrees of freedom .
thus , the state of center of mass @xmath128 satisfies a stochastic differential equation with the same form as eq.([eq : sdecsld ] ) , where @xmath5 is replaced by @xmath129 and @xmath35 is replaced by [ compare with eq.([eq : lybb ] ) ] @xmath130 @xmath124 being the center - of - mass momentum operator and we introduced the function @xmath131 where @xmath132 is the fixed @xmath2-th relative coordinate of the rigid body .
the factor @xmath133 conveys the influence of the internal structure on the evolution of the center of mass and it is due to the indistinguishability of particles : it is also present in the original csl model @xcite , but not in the grw models @xcite .
this factor determines the specific features of the amplification mechanism within the csl model .
hence , we are interested in the action of the operator @xmath134 on a generic state @xmath135 of the center of mass and , in particular , we focus on the changes ( if any ) in the localization process due to dissipation . using the notation@xmath136 and @xmath137
, we have the ( non - normalized ) wave function @xmath138 now , we use the continuum limit , so that @xmath139 @xmath140 being the macroscopic density of particles , and we obtain @xmath141 since @xmath142 one gets @xmath143 we now assume that this macroscopic density @xmath144 does not vary significantly on the length - scale fixed by @xmath145 , so that the exponential in eq.([eq : expv ] ) varies as a function of @xmath146 much faster than the other terms within the integral @xcite .
thus , we can make the substitution @xmath147 in eq.([eq : expv ] ) , getting @xmath148 this clearly shows how the effects on the localization process due to the presence of the momentum operator in @xmath134 can be safely neglected , thus guarantying the convergence toward well localized states .
the localization of the wavefunction , as , e.g. , represented in fig.[fig:1 ] , is basically not modified by the introduction of dissipation in the model .
one can draw the same conclusion by dealing with the master equation of the average state , and , in addition , one recovers eq.([eq : g ] ) as an effective characterization of the scaling of the localization strength with the size of the system .
the lindblad master equation for the state of the center of mass is given by [ compare with eq.([eq : mecsldiss ] ) ] @xmath149 \nonumber\\ & & + \frac{\gamma}{m^2_0 } \int { \mathrm{d}}{\bf y } \left[\hat{\mathbb{l}}^{(\text{cm})}({\bf y})\hat{\rho}^{(\text{cm})}(t)\hat{\mathbb{l}}^{(\text{cm})\dag}({\bf y})\right . \nonumber\\ & & \left .
- \frac{1}{2}\left\{\hat{\mathbb{l}}^{(\text{cm } ) \dag}({\bf y } ) \hat{\mathbb{l}}^{(\text{cm})}({\bf y } ) , \hat{\rho}^{(\text{cm})}\right\ } \right ] .
\label{eq : mecsldisscm}\end{aligned}\ ] ] by using eq.([eq : agia ] ) and neglecting the free hamiltonian contribution , we end up with the equation in the position representation @xmath150 where @xmath151.\ ] ] the same expression was obtained for the original csl model in @xcite , under the so - called sharp - scanning approximation .
in particular , if we consider a rigid body with constant density @xmath152 , we get @xcite @xmath153 with @xmath154 the number of particles of the body when its center of mass is in the position @xmath155 that are outside the volume occupied by the body when its center of mass is in @xmath156 . indeed ,
if @xmath154 is equal to the total number of particles ( i.e. there is no overlap between the volumes occupied by the macroscopic rigid body when its center of mass is in , respectively , @xmath155 and @xmath156 ) , one recovers eq.([eq : g ] ) , up to an irrelevant constant factor @xmath157 .
the localization rate , which is vey small for microscopic systems increases with the size of the system proportionally to the square of the number of particles , which is a direct signature of the action of the noise on indistinguishable particles .
the comparison between the dissipation rate @xmath158 and the localization rate @xmath25 , see eq.([eq : g ] ) , shows how the two phenomena occur on different time scales : while the center of mass of a macroscopic system will be quickly localized by the action of the noise , dissipation can possibly play a role on the system s evolution only on the long time scale .
we take into account the evolution of the center - of - mass energy of a macroscopic rigid body with @xmath77 nucleons , @xmath159 , where @xmath160 is the total mass .
if we repeat the calculations performed in sects.[sec : eef ] , the master equation ( [ eq : mecsldisscm ] ) , at first order in @xmath82 , leads to an exponential relaxation of the energy with rate @xmath161 where we considered a spherical object with radius radius @xmath64 and constant density and we made the following approximation @xcite @xmath162 the dissipation rate @xmath158 in eq.([eq : chi ] ) is much smaller than the corresponding localization rate , which , according to eq.([eq : g ] ) , is given by @xmath163 the ratio between the two rates in the case @xmath164 is @xmath165 if we consider a reference density @xmath166 , one has @xmath167)^3 $ ] , since @xmath168 of matter contains approximately an avogadro s number of nucleons .
now , set a radius @xmath169 , so that @xmath170 . in this case
, the localization rate is @xmath171 , while the dissipation rate is @xmath172 : the noise localizes the center of mass of the macroscopic body on very short time scales , while the influence of dissipation can be safely neglected during the whole evolution of the macroscopic system .
similarly , if @xmath173 , implying @xmath174 , we get @xmath175 , so that dissipation can still be neglected , while in this case @xmath176 .
moreover , one could wonder how this analysis changes if we choose a different one - particle localization rate @xmath1 .
for the value proposed by adler @xcite , @xmath177 , we have that dissipation can still be neglected for @xmath169 , where @xmath178 ( and @xmath179 ) . instead ,
for @xmath173 , we end up with @xmath180 , so that dissipation can play a role on the secular evolution of the system . however , also in this case the effect of dissipation on the localization of the wavefunction is completely negligible .
localization occurs on a much shorter time scale than dissipation , @xmath181 , and then the influence of the dissipative terms in eq.([eq : expv ] ) can be neglected to study localization , even if it can subsequently play a role in the long - time behavior of the system .
the main purpose of collapse models is to provide a unified framework for the description of microscopic and macroscopic systems , thus avoiding an ad - hoc dividing line within the theory , as well as yielding a dynamical explanation for the collapse of the wavefunction .
the results of this paper point out that this program can be followed taking into account basic physically - motivated demands . we have included dissipation in the csl model , which is up to now the most refined collapse model
. this allowed us to remove the divergence of the energy on the long time scale affecting the original csl model .
this divergence traces back to an infinite temperature of the collapse noise , which is of course an unrealistic feature of the model .
the inclusion of dissipation brings along a new parameter , which is strictly related with the finite temperature of the noise .
significantly , even in the presence of a low - temperature noise the localization and the amplification mechanism are effective , so that the unified description of microscopic and macroscopic systems is still guaranteed .
a realistic description of the wavefunction collapse can be further developed , for example by also including a non - white noise @xcite within the model .
nevertheless , one should keep in mind that the specific features of the collapse noise can be fixed only through a first - principle underlying theory , which can clarify the physical origin of the noise @xcite .
the development of such an underlying theory is one of the main goals of the research on collapse models and , more in general , on the theories going beyond quantum mechanics .
the authors acknowledge financial support by the eu project nanoquestfit , infn , fra-2014 by the university of trieste and cost ( mp1006 ) .
they also thank m. bahrami , l. disi , s. donadi , l. ferialdi , and b. vacchini for many useful discussions .
in this appendix , we discuss an alternative and physically motivated choice for the operators @xmath35 , which differs from that made in eq.([eq : ly ] ) ( or , equivalently , in eq.([eq : lymom ] ) ) .
we will see how this different choice does not affect significantly the features of the relaxation dynamics we are interested in .
we also argue that in both cases the system s state equilibrates to the stationary solution in the gibbs form .
let us consider the following operators : @xmath182 where @xmath183 denotes the modulus of @xmath184 .
indeed , such operators still reduce to those of the original csl model @xcite in the limit @xmath34 , i.e. @xmath185 . the difference with respect to the operators in eq.([eq : lymom ] ) is that now the action of the noise no longer damps the momentum of the system in an isotropic way , but it tends to suppress the momentum of a particle in the state @xmath186 mainly in the direction @xmath187
. a significant motivation for collapse operators as in eq.([eq : lymom2 ] ) would be their correspondence with the operators used to describe the collisional decoherence in the presence of dissipation .
in particular , let us consider for simplicity the one - particle case .
then , the stochastic differential equation ( [ eq : sdecsld ] ) , along with the operators in eq.([eq : lymom2 ] ) , provides us with a master equation for the average one - particle state @xmath97 given by eq.([eq : mecsldiss2 ] ) , where @xmath188 is now @xmath189 this master equation is the same as that describing the collisional dynamics of a test particle interacting with a low - density gas in the weak coupling regime @xcite , once we make a proper identification of the coefficients in the two models .
in particular , the new parameter @xmath32 corresponds to the most probable velocity @xmath190 of the gas particles , which have mass @xmath191 and a maxwell - boltzmann momentum distribution , see @xcite for more details . by eqs.([eq :
mecsldiss2 ] ) and ( [ eq : lqp2 ] ) , we directly have that the mean kinetic energy of the system evolves according to @xmath192 which involves the forth order momentum of the momentum operator and hence does not yield a closed equation for the mean kinetic energy .
however , at first order in @xmath82 this equation is exactly the same as eq.([eq : hht ] ) , so that one recovers an exponential relaxation to a finite asymptotic energy which corresponds to the noise temperature in eq.([eq : t ] ) .
even more , this expression of the noise temperature is proved exactly by studying the relaxation to the equilibrium of the system s state . by direct substitution
, one can easily see that the canonical gibbs distribution @xmath193 is a stationary solution of eq.([eq : mecsldiss2 ] ) , with @xmath188 as in eq.([eq : lqp2 ] ) , if @xmath194 is the inverse temperature corresponding to eq.([eq : t ] ) .
relying on the theorems by spohn @xcite , we conclude that this stationary solution is unique and , for any initial condition , the state @xmath195 converges to it .
hence , the temperature in eq.([eq : t ] ) is the temperature toward which the system thermalizes and is thus identified as the noise temperature .
the theorems by spohn apply when the set of lindblad operators @xmath196 of a given lindblad generator are such that only the multiples of the identity commute with all the elements of @xmath197 and that if @xmath198 is contained in @xmath197 , then also @xmath199 is so .
indeed , these two conditions are satisfied by set of lindblad operators in eqs .
( [ eq : mecsldiss2 ] ) and ( [ eq : lqp2 ] ) . to be precise ,
the theorems by sphon hold in the finite dimensional case , so that , to put the above discussion on a firm mathematical ground , one should confine the whole system in a finite volume ( which would also allow to take @xmath200 as a proper quantum state ) and include a cut - off on the momenta .
in addition , once the uniqueness of the stationary state is guaranteed , the convergence to it for any initial condition can be alternatively proved by exploiting the contractivity of the relative entropy under completely positive and trace preserving maps @xcite .
the same results can be applied to our dissipative csl model , i.e. , if @xmath188 is given by eq.([eq : lqp ] ) . beside providing a further proof of the relation in eq.([eq : t ] ) , this shows that our model predicts a convergence of the average one - particle state of the system to the stationary solution in the canonical form . by replacing @xmath201 with @xmath202 in the argument of the wave function @xmath203 ,
one obtains a shift of the order of @xmath204 , for a macroscopic particle made up of @xmath205 nucleons and a noise temperature of the order of @xmath115 . | collapse models explain the absence of quantum superpositions at the macroscopic scale , while giving practically the same predictions as quantum mechanics for microscopic systems . the continuous spontaneous localization ( csl ) model is the most refined and studied among collapse models .
a well - known problem of this model , and of similar ones , is the steady and unlimited increase of the energy induced by the collapse noise . here
we present the dissipative version of the csl model , which guarantees a finite energy during the entire system s evolution , thus making a crucial step toward a realistic energy - conserving collapse model .
this is achieved by introducing a non - linear stochastic modification of the schrdinger equation , which represents the action of a dissipative finite - temperature collapse noise .
the possibility to introduce dissipation within collapse models in a consistent way will have relevant impact on the experimental investigations of the csl model , and therefore also on the testability of the quantum superposition principle .
the superposition principle lies at the core of quantum mechanics .
the last years have experienced a huge progress in the theoretical and experimental investigation aimed at preparing and observing quantum superpositions of large systems @xcite .
such a progress promises a crucial insight into a question which was born with quantum mechanics itself @xcite .
can quantum mechanics be applied at all scales , including the macroscopic ones , or is there an intrinsic limit , above which its description of reality is not appropriate ?
collapse models @xcite show explicitly how the second point of view can be assumed without the need to introduce an ad - hoc separation between the microscopic and the macroscopic world within the theory @xcite . through a non - linear stochastic modification of the schrdinger equation ,
collapse models predict a behavior of microscopic systems which follows almost strictly that of standard quantum mechanics , while preventing macroscopic systems from being in a superposition of macroscopically distinct positions . the continuous spontaneous localization ( csl ) model @xcite is the most refined collapse model , as it also applies to identical particles . the mass density of a quantum system is coupled with a white - noise field , which can be interpreted as a classical random field filling space @xcite .
different speculations on the origin of the noise field have been developed , tracing it back , e.g. , to gravity @xcite or to cosmological particles @xcite .
however , the full characterization of such a noise calls for a new fundamental theory , which departs from quantum mechanics and can explain the classical nature of the noise , as well as its non - hermitian and non - linear coupling with matter @xcite . in this respect , the csl model , and every collapse model ,
should be seen as a phenomenological model expressing the influence of the noise field in an effective way .
the localization of the wavefunction of macroscopic objects , along with the resulting destruction of quantum superpositions , is not the only distinctive feature of the csl model with respect to the usual schrdinger evolution .
the action of the noise induces a steady increase of the mean kinetic energy , which diverges on the long time scale @xcite , thus manifestly leading to a violation of the principle of energy conservation . despite the smallness of the increase rate , the comparison of the predictions on the secular energy with cosmological data provides some of the strongest experimental bounds on the two intrinsic parameters of the model @xcite .
in particular , the spontaneous heating of the intergalactic medium which would be induced by the stochastic noise sets @xmath0 as an upper bound to the localization rate @xmath1 .
this value coincides with the proposal by adler based on latent image formation in photography @xcite .
a significant and long - time debated @xcite issue is whether the divergence of the energy in collapse models can be avoided , thus pointing to a reestablishment of the energy conservation principle , while preserving the specific features any collapse model must have . in this work
, we prove that this can be attained by properly extending the csl model .
we modify the defining stochastic differential equation via the introduction of new operators , which depend on the momentum of the system .
this determines the occurrence of dissipation @xcite , thus leading to the relaxation of the energy to a finite asymptotic value .
the latter can be associated with a finite temperature of the noise field .
remarkably , such a temperature can take on small values ( of the order of 1 k ) while the effectiveness of the model is maintained .
contrary to a common misconception , the steady increase of the energy is not an unavoidable trait of the collapse models inducing localization in space : in our dissipative model there is a continuous localization of the wavefunction , while the mean energy of the system will typically decrease .
the paper is organized as follows . in sec .
[ sec : tcm ] , we briefly recall the main features of the original csl model put forward in @xcite . in sec . [
sec : deo ] , we introduce the dissipative extension of the csl model via a proper stochastic differential equation , which requires the definition of a new parameter within the model .
the defining operators are interpreted by moving to the momentum representation , in analogy with an external momentum - dependent potential . in sec .
[ sec : era ] , after deriving the master equation for the statistical operator associated with the model , we prove explicitly that the mean kinetic energy relaxes to a finite value , whose natural interpretation is given in terms of a temperature associated with the noise field .
crucially , even in the presence of a low temperature noise one can have an effective collapse model , which leads to reconsider the bounds on the csl parameters due to the long - time energy behavior . in sec .
[ sec : mol ] , we show that our extended model still provides a unified description of microscopic and macroscopic systems by virtue of the so - called amplification mechanism ; the relation between the time scales of energy relaxation and position localization is also discussed .
final remarks and possible future perspectives are given in sec .
[ sec : c ] . |
photoacoustic imaging is a recent medical imaging technique combining the large contrast between healthy and unhealthy tissues of their optical parameters with the high spatial resolution of acoustic ( ultrasonic ) waves .
electromagnetic radiation , sent through a domain of interest , generates some heating and a resulting thermal expansion of the underlying tissues .
the mechanical displacement of the tissues generates acoustic waves , which then propagate through the medium and are recorded by an array of detectors ( ultrasound transducers ) .
the photoacoustic effect is now being actively investigated for its promising applications in medical imaging .
we refer the reader to e.g. @xcite for recent reviews on the practical and theoretical aspects of the method . in an idealized setting revisited below
, the electromagnetic source is a very short pulse that propagates through the domain at a scale faster than that of the acoustic waves .
the measured acoustic signals may then be seen as being emitted by unknown initial conditions . a first step in the inversion
thus consists in reconstructing this initial condition by solving an inverse source problem for a wave equation .
this inversion is relatively simple when the sound speed is constant and full measurements are available .
it becomes much more challenging when only partial measurements are available and the sound speed is not constant ; see e.g. @xcite .
a second step consists of analyzing the initial condition reconstructed in the first step and extracting information about the optical coefficients of the domain of interest .
the second step is much less studied .
the energy deposited by the radiation is given by the product of @xmath1 , the attenuation in the tissue and of @xmath2 , the radiation intensity .
the question is therefore what information on the medium may be extracted from @xmath3 .
the product can be plotted as a proxy for @xmath4 when @xmath5 is more or less uniform .
this , however , generates image distortions as has been reported e.g. in @xcite .
the extraction of e.g. @xmath4 from @xmath3 remains an essentially unsolved problem .
two different regimes of radiation propagation should then be considered . in thermoacoustic tomography ( tat ) ,
low frequency ( radio - frequency ) waves with wavelengths much larger than the domain of interest , are being used .
we do not consider this modality here .
rather , we assume that high frequency radiation is generated in the near infra red ( nir ) spectrum .
nir photons have the advantage that they propagate over fairly large distances before being absorbed .
moreover , their absorption properties have a very large contrast between healthy and cancerous tissues . in this regime
, @xmath2 may be interpreted as a spatial density of photons propagating in the domain of interest .
the density of photons is then modeled by a transport equation that accounts for photon propagation , absorption , and scattering ; see below .
this paper concerns the reconstruction of the optical parameters in a steady - state transport equation from knowledge of @xmath6 .
the derivation of the transport equation given in below is addressed in section [ sec : main ] .
we are concerned here with the setting of measurements of @xmath7 for different radiation patterns .
our most general measurement operator @xmath8 is then the operator which to arbitrary radiation patterns at the domain s boundary maps the deposited energy @xmath7 .
we analyze the reconstruction of the absorption and scattering properties of the photons from knowledge of @xmath8 . the main tool used in the analysis
is the decomposition of @xmath8 into singular components , in a spirit very similar to what was done e.g. in @xcite ( see also @xcite ) in the presence of boundary measurements rather than internal measurements .
the most singular component is related to the ballistic photons .
we show that the analysis of that component allows us to reconstruct the attenuation coefficient @xmath4 and the spatial component @xmath9 of the scattering coefficient .
the anisotropic behavior of scattering is partially determined by the second most singular term in @xmath8 , which accounts for photons having scattered only once in the domain .
although the full phase function of the scattering coefficient can not be reconstructed with the techniques described in this paper , we show that the anisotropy coefficient @xmath0 that appears in the classical henyey - greenstein phase function is uniquely determined by the measurements in spatial dimensions @xmath10 and @xmath11 .
moreover , all the parameters that can be reconstructed are obtained with hlder - type stability .
we present the stability results in detail .
when scattering is very large , then radiation is best modeled by a diffusion equation characterized by two unknown coefficients , the diffusion coefficient @xmath12 and the attenuation coefficient @xmath1 .
this regime is briefly mentioned in section [ sec : diff ] .
the rest of the paper is structured as follows .
section [ sec : main ] is devoted to the derivation of the stationary inverse transport problem starting from the transient equation for the short electromagnetic pulse . the main uniqueness and stability results
are also presented in detail in this section .
the derivation of the uniqueness and stability results is postponed to the technical sections [ sec : forward ] and [ sec : inverse ] .
the former section is devoted to the decomposition of the albedo operator into singular components .
useful results on the transport equation are also recalled .
the latter section presents in detail the proofs of the stability results given in section [ sec : main ] .
the propagation of radiation is modeled by the following radiative transfer equation @xmath13 we assume here that @xmath14 is compactly supported in @xmath15 and in @xmath16 outside of a bounded domain of interest @xmath17 we wish to probe .
the domain @xmath17 is assumed to be an open subset of @xmath18 with @xmath19 boundary . for the sake of simplicity , @xmath17 is also assumed to be convex although all uniqueness and stability results in this paper remain valid when @xmath17 is not convex . here , @xmath20 is light speed , @xmath21 is the unit sphere in @xmath22 and @xmath23 is the transport operator defined as @xmath24 where @xmath25 is the total attenuation coefficient and @xmath26 is the scattering coefficient .
both coefficients are assumed to be non - negative and bounded ( by a constant @xmath27 ) throughout the paper .
we define @xmath28 we also refer to @xmath9 as a scattering coefficient and define @xmath4 as the intrinsic attenuation coefficient .
we assume that for ( almost ) all @xmath29 , we have @xmath30 .
we also assume that @xmath31 for ( almost ) all @xmath32 .
the optical coefficients @xmath26 and @xmath25 are the unknown coefficients inside @xmath17 that we would like to reconstruct by probing the domain @xmath17 by radiation modeled by @xmath14 . in photo - acoustics ,
the emitted radiation generates some heating inside the domain @xmath17 .
heating then causes some dilation , which mechanically induces acoustic waves .
such acoustic waves are measured at the boundary of the domain @xmath17 .
after time reversion , the latter measurements allow us to infer the intensity of the source of heating .
this gives us internal measurements of the solution @xmath33 of the transport equation .
the objective of this paper is to understand which parts of the optical parameters may be reconstructed from such information and with which stability . before doing so , we need an accurate description of the propagation of the acoustic waves generated by the radiative heating .
the proper model for the acoustic pressure is given by the following wave equation ( see e.g. @xcite ) @xmath34 where @xmath35 is the dalembertian defined as @xmath36 with @xmath37 the sound speed , where @xmath38 is a coupling coefficient assumed to be constant and known , and where @xmath39 is the thermal energy deposited by the radiation given by @xmath40 not surprisingly , the amount of heating generated by radiation is proportional to the amount of radiation @xmath33 and to the rate of ( intrinsic ) absorption @xmath4 .
as it stands , the problem of the reconstruction of the source term @xmath39 inside @xmath17 from measurements of @xmath41 on the boundary @xmath42 is ill - posed , because @xmath43 is @xmath44dimensional whereas information on @xmath45 is @xmath46dimensional . what allows us to simplify the inverse acoustic problem is the difference of time scales between the sound speed @xmath37 and the light speed @xmath20 .
to simplify the analysis , we assume that @xmath37 is constant and rescale time so that @xmath47
. then @xmath20 in is replaced by @xmath48 , which in water is of order @xmath49 .
the transport scale is therefore considerably faster than the acoustic scale . as a consequence ,
when the radiation source term @xmath14 is supported on a scale much faster than the acoustic scale , then @xmath33 is also supported on a scale much faster than the acoustic scale and as a result @xmath39 can be approximated by a source term supported at @xmath50 .
more specifically , let us assume that the source of radiation is defined at the scale @xmath51 so that @xmath52 is replaced by @xmath53 where @xmath54 is a function compactly supported in @xmath55 such that @xmath56 .
the transport solution then solves @xmath57 with @xmath47 , we verify that @xmath58 is given by @xmath59 where @xmath33 solves with @xmath60 . because @xmath61
, we verify that @xmath33 decays exponentially in time .
this shows that @xmath58 lives at the time scale @xmath62 so that @xmath63 is also primarily supported in the vicinity of @xmath50 .
let us formally derive the equation satisfied by @xmath64 when @xmath65 .
let @xmath66 be a test function and define @xmath67 as the standard inner product on @xmath68 .
then we find that @xmath69 where @xmath70 with @xmath33 defined in and where we have used the change of variables @xmath71 .
the latter term is therefore equal to @xmath72 up to a small term for @xmath73 sufficiently smooth .
we thus find that @xmath64 converges weakly as @xmath65 to the solution @xmath74 of the following wave equation @xmath75 the inverse problem for the wave equation is now well - posed .
the objective is to reconstruct @xmath76 for @xmath77 from measurements of @xmath41 for @xmath15 and @xmath78 .
such an inverse problem has been extensively studied in the literature .
we refer the reader to e.g. @xcite for a recent inversion with ( known ) variable sound speed . in this paper , we assume that @xmath76 has been reconstructed accurately as a functional of the radiation source @xmath79 .
our objective is to understand which parts of the optical parameters @xmath25 and @xmath26 can be reconstructed from knowledge of @xmath76 for a given set of radiations @xmath79 . note that we will allow ourselves to generate many such @xmath79 and thus consider a multi - measurement setting .
the time average @xmath80 satisfies a closed - form steady state transport equation given by @xmath81 as can be seen by averaging in time since @xmath82 . since @xmath79 is assumed to be supported outside of the domain @xmath17 and scattering @xmath83 outside of @xmath17 ,
the above transport equation may be replaced by a boundary value problem of the form @xmath84 where the sets of outgoing and incoming boundary radiations are given by @xmath85 where @xmath86 is the outward normal to @xmath17 at @xmath78 and @xmath87 is an appropriate set of incoming radiation conditions obtained by solving @xmath88 outside of @xmath17 assuming that @xmath89 is known outside of @xmath17 .
we are now ready to state the inverse transport problem of interest in this paper .
it is well known that admits a unique solution in @xmath90 when @xmath91 , where @xmath92 with @xmath93 the surface measure on @xmath42 .
we thus define the albedo operator as @xmath94 the inverse transport problem with angularly averaged internal measurements thus consists of understanding what can be reconstructed from the optical parameters @xmath25 and @xmath26 from complete or partial knowledge of the albedo operator @xmath8 .
we also wish to understand the stability of such reconstructions .
the inverse transport problem and its stability properties are solved by looking at a decomposition of the albedo operator into singular components .
let @xmath95 be the schwartz kernel of the albedo operator @xmath8 , i.e. , the distribution such that @xmath96 the kernel @xmath95 corresponds to measurements of @xmath7 at @xmath77 for a radiation condition concentrated at @xmath97 and propagating with direction @xmath98 .
such a kernel can thus be obtained as a limit of physical experiments with sources concentrated in the vicinity of @xmath99 and detectors concentrated in the vicinity of @xmath16 .
the kernel @xmath95 accounts for radiation propagation inside @xmath17 , including all orders of scattering of the radiation with the underlying structure .
it turns out that we can extract from @xmath95 singular components that are not affected by multiple scattering .
such singular components provide useful information on the optical coefficients .
let us define the ballistic part of transport as the solution of @xmath100 then for @xmath101 , we define iteratively @xmath102 this allows us to decompose the albedo operator as @xmath103 where @xmath104 for @xmath105 are defined as @xmath8 in with @xmath33 replaced by @xmath106 and where @xmath107 is defined as @xmath108 .
thus , @xmath109 is the contribution in @xmath8 of particles that have not scattered at all with the underlying structure while @xmath110 is the contribution of particles that have scattered exactly once .
let @xmath111 for @xmath105 be the schwartz kernel of @xmath112 and @xmath113 the schwartz kernel of @xmath107 using the same convention as in .
we define @xmath114 for @xmath77 and @xmath115 as @xmath116 .
thus , @xmath114 indicates the time of escape from @xmath17 of a particle at @xmath16 moving in direction @xmath117 . on @xmath42
, we also define @xmath118 as the distribution such that @xmath119 for any continuous function @xmath120 on @xmath42 . finally , we define the following terms that quantify attenuation .
we define the function @xmath121 on @xmath122 as @xmath123 we still denote by @xmath124 the function defined above for @xmath125 in @xmath17 . then by induction on @xmath126 , we define @xmath127 the latter term measures the attenuation along the broken path @xmath128 $ ]
. then we have the following result .
[ thm : decomp ] let @xmath129 , @xmath130 , and @xmath113 be the schwartz kernels defined as above .
then we have : @xmath131 moreover , we have the bound @xmath132 this theorem will be proved in section [ sec : forward ] .
the results show that @xmath129 is more singular than @xmath130 and @xmath113 .
it turns out that @xmath130 is also more singular than @xmath113 in the sense that it is asymptotically much larger than @xmath113 in the vicinity of the support of the ballistic part @xmath129 . that this is the case is the object of the following result ( see also lemma [ lem : norm_scat ] and for `` @xmath133 '' ) . for any topological space @xmath134
, we denote by @xmath135 the set of the bounded continuous functions from @xmath134 to @xmath136 .
[ thm : asympalpha1 ] let us assume that @xmath137 and that @xmath138 . for @xmath139 , let @xmath140 for some @xmath141 .
let @xmath142 be such that @xmath143 .
then we have the following asymptotic expansion : @xmath144 for @xmath10 and @xmath145 , respectively , where we have defined the functions @xmath146 and @xmath147 .
theorem [ thm : asympalpha1 ] will be proved in section [ sec : forward ] .
we thus observe that @xmath130 blows up a priori faster than @xmath113 as @xmath65 , i.e. , as the observation point @xmath16 becomes closer to the segment where the ballistic term @xmath129 is supported .
this singularity allows us to obtain information on the optical coefficients that is not contained in the ballistic part @xmath129 .
moreover , because of the singular behavior of @xmath130 , such information can be reconstructed in a stable manner . as mentioned above
, the singular behaviors of @xmath129 and @xmath130 allow us to extract them from the full measurements @xmath148 .
moreover , such an extraction can be carried out in a stable fashion , in the sense that small errors in the measurement of the albedo operator translates into small errors in the extraction of the terms characterizing @xmath129 and @xmath130 .
more precisely , let @xmath8 be the albedo operator corresponding to optical parameters @xmath149 and @xmath150 the operator corresponding to the optical parameters @xmath151 .
from now on , a term superimposed with the @xmath152 sign means a term calculated using the optical parameters @xmath151 instead of @xmath149 .
for instance @xmath153 is the equivalent of @xmath154 defined in with @xmath149 replaced by @xmath151 .
we first derive the stability of useful functionals of the optical parameters in terms of errors made on the measurements .
let us assume that @xmath8 is the `` real '' albedo operator and that @xmath150 is the `` measured '' operator .
we want to obtain error estimates on the useful functionals of the optical parameters in terms of appropriate metrics for @xmath155 .
we obtain the following two results .
the first result pertains to the stability of the ballistic term in the albedo operator : [ thm : ballistic ] let @xmath8 and @xmath150 be two albedo operators and @xmath139 .
then we obtain that @xmath156 theorem [ thm : ballistic ] is proved in section [ sec : ballistic ] .
the stability result obtained from the single scattering component is based on a singular behavior obtained in the vicinity of the ballistic component .
such a behavior can not be captured by the @xmath157 norm used above .
instead , we define @xmath158 as the schwartz kernel of the albedo operator where the ballistic part has been removed , i.e. , for measurements that are performed away from the support of the ballistic part .
our stability results are obtained in terms of errors on @xmath159 rather than on @xmath8 .
we can then show the following stability result .
[ thm : stabscat ] let us assume that @xmath160 and that @xmath161 .
let @xmath162 and define @xmath163 .
let @xmath142 such that @xmath143 . in dimension @xmath10 , we have @xmath164 where @xmath165 .
when @xmath145 , we have @xmath166 where @xmath167 and where we use the same notation as in theorem [ thm : asympalpha1 ] .
theorem [ thm : stabscat ] is proved in section [ sec : single ] .
such results do not grant uniqueness of the reconstruction of the optical parameters in the most general setting .
however , they do provide stable , unique , reconstructions in several settings of interest .
let us first assume that @xmath168 so that @xmath169
. then knowledge of the albedo operator uniquely determines @xmath89 for all @xmath77 and @xmath115 .
indeed , we deduce from theorem [ thm : ballistic ] that @xmath170 is uniquely determined and hence @xmath171 since the latter equals @xmath172 when @xmath50 .
taking the derivative of the negative of the logarithm of the latter expression gives us @xmath173 for all @xmath139 and @xmath174 and hence @xmath25 for all @xmath175 .
moreover , we have the following stability result .
[ thm : stabscatfree ] recalling that @xmath89 is bounded from above and below by positive constants , we find that when @xmath168 , @xmath176 here , @xmath177 is a constant that depends on the uniform bound @xmath178 .
the above theorem is proved in section [ sec : ballistic ] .
note that the above result is local in @xmath179 and @xmath180 .
in other words , @xmath181 is uniquely determined by @xmath182 , that is by the experiment that consists of sending a beam of radiation in direction @xmath180 passing through the point @xmath16 ( at least asymptotically since such a transport solution is not an element in @xmath183 ) .
we now assume that @xmath184 .
then the ballistic component of @xmath8 and the estimate in theorem [ thm : ballistic ] allow us to uniquely reconstruct both @xmath89 and @xmath25 under the assumption that @xmath185 we recall that @xmath89 is bounded from below by @xmath186 .
for technical reasons , we also assume that @xmath25 is known in the @xmath187vicinity of @xmath42 , i.e. , for all @xmath188 such that @xmath189 for some @xmath190 .
such a hypothesis is not very restrictive from a practical viewpoint .
we denote by @xmath191 the banach space of the continuous linear functionals on the banach space @xmath192 ( where the gradient @xmath193 is understood in the distributional sense ) . under the above assumptions
, we have the following result .
[ thm : stabsym ] the coefficients @xmath89 and @xmath25 are uniquely determined by the albedo operator @xmath8
. moreover , we have the following stability estimate @xmath194 the constant @xmath177 depends on the parameter @xmath195 as well as the uniform bounds @xmath196 and @xmath178 .
this theorem is proved in section [ sec : ballistic ] .
the above result is also local in @xmath179 and @xmath180 . in other words , @xmath181 and @xmath197
are uniquely determined by @xmath182 , i.e. , by the experiment that consists of sending a beam of radiation in direction @xmath180 passing through the point @xmath16 .
stability estimates may be obtained in stronger norms for @xmath198 provided that a priori regularity assumptions be imposed .
we show the [ cor : regsigma ] let us assume that @xmath198 and @xmath199 are bounded in @xmath200 by @xmath201 for @xmath202 and @xmath203 .
then for all @xmath204 , we have @xmath205 where the constant @xmath177 depends on @xmath201 and on the uniform bounds @xmath196 and @xmath178 .
the corollary is proved in section [ sec : ballistic ] .
let us assume that @xmath137 and that @xmath138 and let us assume again that @xmath206 is known in the @xmath187vicinity of @xmath42 , i.e. , for all @xmath207 such that @xmath189 for some @xmath190 .
the stability estimate allows one to uniquely reconstruct @xmath4 and @xmath198 under the symmetry hypothesis , which is quite general physically . indeed , even when attenuation is anisotropic , there is no reason to observe different attenuations in direction @xmath208 and direction @xmath209 .
the stability estimate in theorem [ thm : stabscat ] provides additional information on the optical coefficients , but not enough to fully reconstruct the scattering kernel @xmath26 . in dimension @xmath10 , we gain information only on @xmath210 . in dimension @xmath145 , we garner information about @xmath211 for all @xmath212 orthogonal to @xmath180 .
the integration in @xmath213 means that one dimension of information is lost in the measurements .
thus , @xmath214 dimensions of information are available on the @xmath215dimensional object @xmath26 .
let us consider the case of an isotropic absorption @xmath216 and isotropic scattering in the sense that @xmath217 .
in such a setting , @xmath218 becomes @xmath44dimensional .
yet , available data in dimension @xmath219 give us information on @xmath220 this is different information from the normalization in @xmath221 as a consequence , if @xmath222 is of the form @xmath223 , where @xmath224 is parameterized by one function @xmath0 , then we have a chance of reconstructing @xmath0 from knowledge of @xmath225 and @xmath226 provided @xmath227 ( where @xmath228 ) .
this occurs for the classical henyey - greenstein ( hg ) phase function in dimensions @xmath10 and @xmath11 , where @xmath229 where @xmath230 and @xmath231 for a.e .
@xmath77 . note that @xmath232 for @xmath77 where the function @xmath233 is given by @xmath234 [ thm : hg ] in the hg phase function in dimension @xmath235 , the parameter @xmath0 is uniquely determined by the data provided @xmath227 for a.e . @xmath77 .
theorem [ thm : hg ] follows from theorems [ thm : stabsym ] , [ thm : stabscat ] , and from , , and the following lemma .
[ hgn=3 ] the function @xmath236 is strictly increasing on @xmath237 , @xmath238 , @xmath239 , @xmath240 , where @xmath241 and @xmath242 and @xmath243 . considering , lemma [ hgn=3 ] in dimension @xmath10 is trivial .
the proof of lemma [ hgn=3 ] in dimension @xmath11 is given in section [ sec : single ] .
moreover we have the following stability estimates .
[ thm : stabhg ] in dimension @xmath219 , we have @xmath244 where @xmath245 is defined in theorem [ thm : stabscat ] , and where the constant @xmath177 depends on @xmath195 , @xmath196 and @xmath178 . in addition , for the hg phase function in dimension @xmath235 , we have the following stability estimate : assume @xmath246 and @xmath247 , then @xmath248 where the constant @xmath177 depends on @xmath195 , @xmath196 , @xmath249 , @xmath178 and @xmath250 .
theorem [ thm : stabhg ] is proved in section [ sec : single ] .
the sensitivity of the reconstruction of @xmath0 degrades as @xmath251 converges to @xmath252 in the sense that @xmath253 as @xmath254 , @xmath255 , where @xmath177 is the constant that appears on the right - hand side of when we replace @xmath256 by @xmath257 on the left - hand side of . on the other hand , @xmath258 when @xmath259 , @xmath260 , so that reconstructions of @xmath0 are very accurate for @xmath0 close to @xmath172 , i.e. , in the case of very anisotropic media .
more precisely , using the properties of the function @xmath236 , one can replace the left - hand side of by @xmath261 ( resp .
@xmath262 , resp .
@xmath263 ) provided that @xmath264 ( resp . @xmath265 and @xmath264 , resp .
@xmath266 ) for some constant @xmath267 ( resp .
@xmath268 , resp .
@xmath269 ) and the constant @xmath177 on the right hand side of then depends also on @xmath267 ( resp .
@xmath268 , resp .
@xmath269 ) .
when scattering is large so that the mean free path @xmath270 is small and intrinsic attenuation @xmath1 is small , then radiation inside the domain @xmath17 is best modeled by a diffusion equation @xmath271 where @xmath272 is the spatial density of photons and @xmath12 is the diffusion coefficient .
we refer the reader to e.g. @xcite for references on the diffusion approximation . when scattering is e.g. isotropic , i.e. , when @xmath273 , then we find that @xmath274 , where @xmath9 is introduced in and @xmath275 is the spatial dimension .
when @xmath12 is known , then the reconstruction of @xmath1 may be easily obtained by using only one measurement .
indeed , the measurement @xmath276 so that @xmath2 may be obtained by solving . once @xmath2 is known , it will be positive in @xmath17 provided that @xmath277 is non trivial and non - negative .
then @xmath1 is obtained by dividing @xmath43 by @xmath5 .
when @xmath277 is bounded from below by a positive constant , then we see that the reconstruction of @xmath4 is unique and clearly stable . when @xmath278 are both unknown , then multiple ( at least two ) measurements are necessary
. this problem will be analyzed elsewhere .
this concludes the section on the derivation and the display of the main results .
the mathematical proofs are presented in the following two sections .
in this section , we prove several results on the decomposition of the albedo operator ( lemmas [ lem : dec ] , [ lem : norm_scat ] and [ lem : dec_noyau_stab ] ) and prove theorems [ thm : decomp ] and [ thm : asympalpha1 ] .
we first recall the well - posedness of the boundary value problem ( here @xmath279 , @xmath280 and @xmath281 ) and give a decomposition of the albedo operator .
the boundary value problem is equivalent to the integral equation @xmath282 for @xmath283 and @xmath284 , where @xmath285 is the bounded operator in @xmath183 defined by @xmath286 for a.e .
@xmath287 and for @xmath288 , and @xmath289 is the bounded operator from @xmath290 to @xmath90 defined by @xmath291 for a.e . @xmath287 .
since @xmath292 , it turns out that @xmath293 is invertible in @xmath294 ( * ? ? ? * lemma 2.4 ) , so that the albedo operator @xmath295 is well - defined by and so that the solution @xmath33 of with boundary condition @xmath284 satisfies @xmath296 it follows that @xmath297 where @xmath298 , @xmath299 , is the distributional kernel of @xmath300 defined by @xmath301 for a.e . @xmath77 and @xmath302 , and where @xmath303 , @xmath299 , is the distributional kernel of @xmath304 .
we give the explicit expression of the distributional kernels @xmath303 , @xmath305 , and @xmath298 , @xmath306 in lemma [ lem : dec ] and study the boundedness of @xmath303 in lemma [ lem : norm_scat ] and of @xmath307 in lemma [ lem : dec_noyau_stab ] .
we then prove theorems [ thm : decomp ] and [ thm : asympalpha1 ] and lemmas [ lem : norm_scat ] and [ lem : dec_noyau_stab ] .
the proof of lemma [ lem : dec ] is given in the appendix .
for @xmath308 , @xmath309 , we set @xmath310 .
[ lem : dec ] for @xmath305 and a.e .
@xmath311 , we have @xmath312_{v_i=\widehat{z_i - z_{i+1}},\ i=0\ldots m-2,\atop v_{m-1}=\widehat{z_{m-1}-z_m - t'z_m}}dt'dz_1\ldots dz_{m-1}.\nonumber\end{aligned}\ ] ] for @xmath306 and a.e .
@xmath313 , we have @xmath314_{v_i=\widehat{z_i - z_{i+1}},\ i=0\ldots m-1}dz_1\ldots dz_{m-1}.\nonumber\end{aligned}\ ] ] [ lem : norm_scat ] for @xmath315 , @xmath316 for @xmath317 , we have @xmath318 @xmath319 we do not use and for @xmath133 in order to prove
. however we will use them in the proof of the stability estimates given in theorem [ thm : stabscat ] .
[ lem : dec_noyau_stab ] for @xmath219 , we have @xmath320 in addition , we have @xmath321 for a.e . @xmath77 and for @xmath284 , where @xmath322 [ [ proof - of - theorem - thmdecomp . ] ] proof of theorem [ thm : decomp ] .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + the equality for @xmath129 follows from the definition of the operator @xmath289 . from , and it follows that @xmath323 for a.e . @xmath77 and @xmath284 . performing the change of variables @xmath324 ( @xmath325 , @xmath326 ) on the right - hand side of , we obtain @xmath327 for a.e . @xmath77 and @xmath284 . performing the change of variables
@xmath328 ( @xmath329 , @xmath174 , @xmath330 ) on the right hand side of , we obtain @xmath331 for a.e . @xmath77 and @xmath284 , which yields for @xmath130
now set @xmath332 when @xmath219 .
taking account of lemma [ lem : norm_scat ] and lemma [ lem : dec_noyau_stab ] , we obtain .@xmath333 [ [ proof - of - theorem - thmasympalpha1 . ] ] proof of theorem [ thm : asympalpha1 ] .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + we assume that @xmath334 and @xmath335 .
let @xmath336 and let @xmath337 and let @xmath338 be such that @xmath339 .
set @xmath140 . from
, it follows that @xmath340 @xmath341 @xmath342 for @xmath343 , where @xmath344 for @xmath345 .
we prove for @xmath10 .
consider the function @xmath346 defined by @xmath347 , for @xmath348 .
then performing the change of variables @xmath349 on the right hand side of and performing the change of variables @xmath350 on the right hand side of , we obtain @xmath351 for @xmath343 , where @xmath352 for @xmath353 ( we recall that @xmath354 , @xmath348 ) . note that
using the definition of @xmath355 and @xmath356 , we obtain @xmath357 for @xmath353 . therefore using we
obtain @xmath358 for @xmath353 and @xmath359 .
note also that from it follows that @xmath360 for @xmath353 ( we used the limit @xmath361 as @xmath362 which holds for any positive real numbers @xmath363 and @xmath364 ) .
using , , and continuity and boundedness of @xmath198 and @xmath365 and @xmath4 , and using lebesgue dominated convergence theorem , we obtain @xmath366 finally note that @xmath367 combining , , and , we obtain for @xmath10 .
we prove for @xmath317 . performing the change of variables @xmath368 on the right - hand side of
, we obtain @xmath369 for @xmath343 , where @xmath370 for @xmath371 .
note that @xmath372 therefore , using the lebesgue dominated convergence theorem and continuity and boundedness of @xmath373 , we obtain @xmath374 as @xmath362 . similarly performing the change of variables @xmath368 on the right hand side of , and using lebesgue dominated convergence theorem and continuity and boundedness of @xmath373
, we obtain @xmath375 as @xmath362 .
note that performing the change of variables @xmath376 , @xmath377 , ( @xmath378 , @xmath379 ) we have @xmath380 adding and and using and , we obtain for @xmath317.@xmath333 [ [ proof - of - lemma - lemnorm_scat . ] ] proof of lemma [ lem : norm_scat ] .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + first note that from and , it follows that @xmath381 for a.e .
@xmath382 and for @xmath383 , @xmath384 , where @xmath385 for @xmath386 and @xmath384 .
we prove and .
let @xmath10 .
let @xmath382 be such that @xmath387 for any @xmath388 .
set @xmath389 for any @xmath308 . using and using the equality @xmath390
, we obtain @xmath391 where we used that @xmath392 .
estimate follows from and .
let @xmath393 be such that @xmath394 for any @xmath388 .
using ( with `` @xmath395 '' ) and , we obtain @xmath396 where @xmath397 for @xmath398 .
we prove @xmath399 then estimate follows from , and .
let @xmath400 .
note that by using , we obtain @xmath401 where @xmath402 and @xmath403 denotes the diameter of @xmath17 .
consider @xmath404 a unit vector orthogonal to @xmath180 and perform the change of variable @xmath405 , then we obtain @xmath406 ^ 2}{d\lambda_1d\lambda_2\over ( ( \lambda_{1,x}-\lambda_1)^2+(\lambda_{2,x}-\lambda_2)^2)^{1\over 2}|\lambda_2|^{1\over 4}},\label{p11}\ ] ] where @xmath407 .
then note that @xmath408 combining and , we obtain @xmath409 where @xmath410 , @xmath411 and @xmath412 .
finally note that @xmath413 @xmath414 combining we obtain .
the statements follow from and the following statements @xmath415 for @xmath416 , @xmath317 , @xmath417 and where @xmath418 is defined by and . we prove , which will complete the proof of lemma [ lem : norm_scat ] .
we proceed by induction on @xmath126 .
we prove for @xmath133 .
let @xmath317 .
note that formula still holds .
note also that @xmath419 for @xmath420 , @xmath421 .
therefore @xmath422 thus for @xmath133 follows from and .
let @xmath383 , @xmath384 be such that hold for any @xmath317 .
we prove that
hold for any @xmath317 and for `` @xmath126''@xmath423 .
let @xmath424 be such that @xmath425 for any @xmath426 . from and
it follows that @xmath427 assume that @xmath428 .
then from and for @xmath429 it follows that there exists a constant @xmath177 ( which does not depend on @xmath424 ) such that @xmath430 performing the change of variables @xmath431 , @xmath432 , we obtain @xmath433 combining and ( with `` @xmath434 '' ) , we obtain @xmath435 @xmath436 and @xmath437 and @xmath438 denotes the euclidean ball of @xmath439 of center @xmath252 and radius @xmath403 . using spherical coordinates @xmath440 , @xmath441 and @xmath442 ( @xmath443 ) we obtain @xmath444 ( by convention @xmath445 ) .
note that by performing the change of variables `` @xmath446 '' and using the estimate @xmath447 , we obtain @xmath448 @xmath449 for @xmath450 .
assume @xmath451 . then combining and
, we obtain @xmath452 therefore using also we obtain that holds for `` @xmath126''@xmath453 .
assume @xmath454 .
then note that @xmath455 for @xmath450 . combining , and
, we obtain @xmath456 where @xmath457 and @xmath458 . therefore using also
, we obtain that holds for `` @xmath126''@xmath459 .
assume that @xmath460 . from and for @xmath461
it follows that there exists a constant @xmath177 ( which does not depend on @xmath424 ) such that @xmath462 performing the change of variables @xmath431 , @xmath463 , we obtain @xmath464 combining and ( with `` @xmath434 '' ) , we obtain @xmath465 where @xmath466 and @xmath467 , @xmath437 , and where @xmath438 denotes the euclidean ball of @xmath439 of center @xmath252 and radius @xmath403 . using spherical coordinates @xmath440 , @xmath441 and @xmath442 ( @xmath443 ) , and using the estimate @xmath468 , we obtain @xmath469 note that @xmath470 for @xmath450 . combining , and , we obtain @xmath471 therefore holds for `` @xmath126''@xmath472 . @xmath35 [ [ proof - of - lemma - lemdec_noyau_stab . ] ] proof of lemma [ lem : dec_noyau_stab ] .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + we first prove the estimates and given below @xmath473 for @xmath474 and for @xmath383 such that @xmath475 and @xmath476 , where the positive constants @xmath477 , @xmath478 do not depend on @xmath479 .
let @xmath474 and let @xmath383 be such that @xmath475 and @xmath480 . performing the change of variables @xmath481 , @xmath482 ( where @xmath403 denotes the diameter of @xmath17 ) , we obtain @xmath483 where @xmath484 ( we also used a rotation that maps @xmath485 to the vector @xmath486 ) and @xmath487 performing the change of variables @xmath488 , @xmath489 ) , on the right hand side of we obtain @xmath490 where @xmath491 ( by convention @xmath492 ) .
consider the case @xmath10 and @xmath133 .
using and the estimate @xmath493 , we obtain @xmath494 @xmath495 @xmath496 @xmath497 where @xmath498 .
estimate follows from , and .
consider the case @xmath317 and @xmath476 .
note that @xmath499 for @xmath500 . combining and
, we obtain @xmath501 ( we perform the change of variables @xmath502 , @xmath503 )
. finally follows from and and the estimate @xmath504 .
we are now ready to prove .
let @xmath219 . from
, it follows that @xmath505 for a.e .
@xmath506 , where @xmath507 for @xmath508 .
assume @xmath10 . combining and we obtain @xmath509 performing the change of variables @xmath510 , on the right hand side of
, we obtain @xmath511 assume @xmath317 . using and
, we obtain @xmath512 where @xmath177 does not depend on @xmath513 . performing the change of variables @xmath514 ,
@xmath515 , @xmath516 , we obtain @xmath517 where @xmath484 . performing the change of variables @xmath518 ,
@xmath519 , we obtain @xmath520 for @xmath308 , @xmath521 , where @xmath491 and @xmath522 is defined by . therefore from , and
, it follows that @xmath523 where the positive constants @xmath177 , @xmath524 do not depend on @xmath513 .
finally from and , it follows that @xmath525 for @xmath219 and for @xmath526 , @xmath527 , where the positive constants @xmath177 , @xmath524 do not depend on @xmath513 .
let @xmath219 and let @xmath526 , @xmath527 . from and the estimate @xmath528 it follows that @xmath529 where @xmath530 .
statement follows from and . we prove .
we first obtain @xmath531 for a.e .
@xmath532 and for @xmath284 . therefore using we
obtain @xmath533 for @xmath284 , where @xmath534 .
therefore there exists a ( unique ) function @xmath535 such that @xmath536 set @xmath537 for a.e . @xmath393 and recall the definition of @xmath538 . then follows from .@xmath35
[ [ proof - of - theorem - thmballistic . ] ] proof of theorem [ thm : ballistic ] .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + let @xmath539 and @xmath540 , @xmath541 .
we have : @xmath542(x)dx\right| \leq\lvert a- \widetilde { a}\rvert _ { \mathcal{l}\left(l^1(\gamma_-,d\xi),l^1(x)\right)}.\label{a1}\ ] ] using and the decomposition of the albedo operator ( see theorem [ thm : decomp ] ) we obtain @xmath543 where @xmath544 and where @xmath545 ( and @xmath546 ) . note that
performing the change of variables @xmath547 on the right - hand side of , we obtain @xmath548 where @xmath549 we use the following result : for any function @xmath550 and for a.e .
@xmath551 there exists a sequence of functions @xmath552 ( which does not depend on the function @xmath553 ) , @xmath554 , @xmath555 and @xmath556 such that the following limit holds @xmath557 ( we refer the reader to ( * ? ? ?
* corollary 4.2 ) for the proof of this statement ) .
in particular , the limits holds for some sequence @xmath558 when @xmath559 belongs to the lebesgue set of @xmath553 denoted by @xmath560 and the complement of @xmath560 is a negligible subset of @xmath561 ( see @xcite ) .
let @xmath562 where @xmath563 is the measurable function on @xmath561 defined by @xmath564 for @xmath565 .
note that the complement of @xmath566 is still negligible .
let @xmath567 and let @xmath568 where @xmath569 denotes the set of continuous and compactly supported functions on @xmath570 .
consider the sequence @xmath571 defined by @xmath572 for @xmath77 and @xmath383 where @xmath573 , @xmath574 , where @xmath575 when @xmath576 and @xmath577 otherwise .
the support of @xmath578 concentrates around the line which passes through @xmath579 with direction @xmath580 . from , , ( and the stone - weierstrass theorem )
, we obtain that @xmath581 for the single - scattering part , using and and for @xmath133 we obtain : @xmath582 where @xmath583 and @xmath584 is the function from @xmath561 to @xmath136 defined by @xmath585 ( where @xmath245 is defined in theorem [ thm : stabscat ] ) . from the definition of @xmath586
, it follows that @xmath587 is a bounded and continuous function on @xmath561 . therefore using lebesgue
dominated convergence theorem , we obtain @xmath588 then using , and lebesgue convergence theorem , we obtain @xmath589 . therefore taking account of and
, we obtain @xmath590 combining ( with @xmath591 ) , , , , we obtain @xmath592 for @xmath593 and @xmath594 . from the density of @xmath595 in @xmath596
, it follows that also holds for @xmath597 . applying on @xmath598 for a.e .
@xmath599 ( where @xmath600 when @xmath601 and @xmath602 otherwise ) , we obtain .
@xmath333 [ [ proof - of - theorem - thmstabscatfree . ] ] proof of theorem [ thm : stabscatfree ] .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + let us define @xmath603 note that @xmath604 .
then , @xmath605 thanks to . the point - wise ( in @xmath606 ) control on @xmath607 and
the estimate for @xmath608 show by application of the triangle inequality and the fact that @xmath607 is bounded from below by the positive constant @xmath609 that @xmath610 here , the constant @xmath177 depends on @xmath178 and is of order @xmath611 .
not surprisingly , reconstructions deteriorate when the optical depth @xmath612 of the domain increases .
this shows that @xmath613 this concludes the proof of the theorem .
@xmath35 [ [ proof - of - theorem - thmstabsym . ] ] proof of theorem [ thm : stabsym ] .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + let @xmath614 and let @xmath615 be defined by @xmath616 for @xmath617 .
the function @xmath618 is defined similarly .
we first prove a stability estimate on @xmath619 @xmath620 .
let @xmath621 , @xmath622 .
set @xmath623 and @xmath624 . due to the symmetry of @xmath4 , @xmath198 with respect to the speed variable @xmath208 ( @xmath4
, @xmath625 bounded from below ) , we obtain : @xmath626 where @xmath364 is defined by .
note that from , it follows that @xmath627 where @xmath403 denotes the diameter of @xmath17 .
a similar estimate is valid for @xmath620 and @xmath628 . therefore using the fact that
@xmath629 for @xmath630 , we obtain @xmath631 note that from it follows that @xmath632 for @xmath633 and @xmath634 , where @xmath403 is the diameter of @xmath17 .
a similar estimate is valid for @xmath635 . therefore using the equality @xmath636 ( for @xmath637 , @xmath638 )
, we obtain @xmath639 using , integrating in the @xmath606 variable ( @xmath624 ) and using , we obtain @xmath640 combining and we obtain @xmath641 for a.e .
@xmath621 , where the constant @xmath177 depends only on @xmath403 , @xmath642 , @xmath643 , @xmath644 and @xmath645 . we now prove .
estimate is , in particular , a consequence of the identities @xmath646 for a.e . @xmath621 and @xmath622 .
consider the operator @xmath647 associated with @xmath648 : @xmath649 for @xmath115 .
using the change of variable @xmath650 ( @xmath651 , @xmath652 , @xmath653 ) we have @xmath654 and using integration by part in the inner integral we have : @xmath655 @xmath656 for a.e .
@xmath115 ( we used ) . identifying @xmath657 and @xmath658 , we obtain @xmath659 we prove the following estimate @xmath660 where @xmath177 depends on the uniform bounds @xmath196 and @xmath178 . combining and ,
we obtain , which completes the proof of theorem [ thm : stabsym ] .
the estimate on @xmath181 comes from an estimate on @xmath661 by the triangle inequality .
the latter is given by @xmath662 we just obtained control of @xmath663 in the @xmath157 sense in the @xmath664variable .
it thus remains to control the constant term @xmath665 .
note that the latter term is nothing but the x - ray transform ( radon transform when @xmath10 ) of @xmath198 along the line of direction @xmath180 passing through @xmath179 . in the setting of measurements that are supposed to be accurate in @xmath666 ,
the line integral is not directly captured as it corresponds to a measurement performed at a point @xmath667 .
this is the reason why we assume that @xmath198 is known in the @xmath668vicinity of @xmath669 .
knowledge of @xmath198 and @xmath199 in the @xmath195-vicinity of @xmath42 allows one to control @xmath670 by @xmath671 . when the x - ray transform of @xmath198 is well captured by available measurements , as for instance in the presence of boundary measurements @xcite , then @xmath195 can be set to @xmath252 .
more precisely , we find that @xmath672 where @xmath673 is equal to @xmath172 for @xmath674 . integrating by parts , this shows that @xmath675 thanks to estimate . by lipschitz regularity of the exponential
, we thus have that @xmath676 the stability result on @xmath181 follows from and the triangle inequality as in the proof of theorem [ thm : stabscatfree ] .
@xmath35 [ [ proof - of - corollary - corregsigma . ] ] proof of corollary [ cor : regsigma ] .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + let @xmath202 .
we first derive from the following estimate @xmath677 assuming that @xmath678 is bounded by @xmath201 in @xmath679 , then using and using the complex interpolation result ( * ? ? ?
* theorem 6.4.5 pp .
153 ) , we obtain @xmath680 for @xmath681 , where @xmath682 and @xmath683 . this proves .
we prove . from
, it follows that @xmath684 , where @xmath403 is the diameter of @xmath17 . hence we obtain @xmath685 for @xmath139 and @xmath599 . using and hlder inequality , and using , we obtain @xmath686 for a.e .
@xmath115 and for @xmath687 where @xmath688 .
estimate proves .@xmath35 [ [ proof - of - theorem - thmstabscat . ] ] proof of theorem [ thm : stabscat ] .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + let @xmath689 .
set @xmath690 and let @xmath404 be such that @xmath339 .
let @xmath691 , then @xmath140 .
first assume @xmath10 .
using the equality @xmath692 , and using , , we obtain @xmath693 for @xmath694 . therefore using as @xmath362 and , we obtain .
assume @xmath317 . using the equality @xmath692 , and using , ( for `` @xmath133 '' )
, we obtain @xmath695 for @xmath696 . therefore using as @xmath362 and
, we obtain .
[ [ proof - of - lemma - hgn3 . ] ] proof of lemma [ hgn=3 ] .
+ + + + + + + + + + + + + + + + + + + + + + + first consider the case @xmath10 .
the statement in lemma [ hgn=3 ] is a straightforward consequence of and we have the following inversion formula : @xmath697 let now @xmath11 . using and the identity @xmath698 , @xmath699 , we obtain @xmath700 for @xmath701 .
performing the change of variables @xmath702 ( @xmath703 ) , we obtain @xmath704 performing the change of variables @xmath705 , we obtain @xmath706 for @xmath701 .
note that from the above , it follows that @xmath707 as @xmath708 ( where we used the integral value @xmath709_0^{+\infty}=1 $ ] ) .
differentiating with respect to @xmath251 , we obtain @xmath710 for @xmath701 , where @xmath711 , @xmath712 , @xmath713 , and @xmath714 integrating by parts , we obtain @xmath715 for @xmath701 , where we use that a primitive of the function @xmath716 is given by the function @xmath717 and where we used that the derivative of the function @xmath718 is given by @xmath719 . performing the change of variables `` @xmath208''@xmath720 on the right hand side of
, we obtain @xmath721 for @xmath701 . combining and ,
we obtain @xmath722,\label{h8}\end{aligned}\ ] ] for @xmath701 . using the estimate @xmath723 for @xmath724 and @xmath725
, we obtain that the second integral on the right hand side of is less than the first integral on the right - hand side of for @xmath725 . therefore using also that the second integral
is multiplied by @xmath726 , we obtain @xmath727 for @xmath725 , which proves that @xmath236 is strictly increasing on @xmath728 . @xmath35 [ [ proof - of - theorem - thmstabhg ] ] proof of theorem [ thm : stabhg ] + + + + + + + + + + + + + + + + + + + + + + + + + + + + + we prove . from , , and
, it follows that @xmath729 for @xmath730 . using and the estimate @xmath731 ( see ) where @xmath403 is the diameter of @xmath17
, we obtain @xmath732 for @xmath730 ( we used the identity @xmath733 for @xmath734 and @xmath735 ) .
let @xmath336 . integrating over the line which passes through @xmath179 with direction @xmath180 and using the stability estimate and , we obtain @xmath736 where @xmath737 and @xmath177 is the constant on the right - hand side of .
integrating over @xmath738 with measure @xmath739 , we obtain .
we now prove .
assume @xmath740 . using and @xmath741
, we obtain @xmath742 ( we used the identity @xmath743 for @xmath744 and @xmath745 ) . using the identity @xmath228 , we have @xmath746 ( we used the fact that @xmath747 at the vicinity of the boundary @xmath748 ) . finally combining , , and ,
we obtain .@xmath35 [ [ section ] ] for @xmath305 , let @xmath749 denotes the distributional kernel of the operator @xmath750 where @xmath285 is defined by .
we first give the explicit expression of @xmath751 , @xmath752 .
then by induction we give the explicit expression of the kernel @xmath749 .
finally we prove lemma [ lem : dec ] .
+ from it follows that @xmath753 for a.e .
@xmath287 and for @xmath754 . performing the change of variables
@xmath755 ( @xmath756 ) on the right hand side of , we obtain @xmath757 for a.e .
@xmath287 and for @xmath754 .
therefore @xmath758 for a.e .
@xmath759 . from and it follows that @xmath760 @xmath761 for a.e .
@xmath287 and for @xmath754 .
therefore performing the change of variables @xmath762 ( @xmath763 ) on the right hand side of we obtain @xmath764 for a.e . @xmath759 .
then by induction we have @xmath765 for a.e . @xmath766 .
we prove for @xmath767 . from , and it follows that @xmath768 for a.e .
@xmath77 and for @xmath540 . then performing the change of variable @xmath324 ( @xmath769 , @xmath770 and @xmath771 ) on the right hand side of
, we obtain @xmath772 for a.e .
@xmath77 and for @xmath540 . performing the change of variables
@xmath773 ( @xmath774 , @xmath775 , @xmath776 ) on the right hand side of , we obtain @xmath777 for a.e .
@xmath77 and for @xmath540 , which proves for @xmath767
. then we prove before proving for @xmath306 .
let @xmath306 . from and the definition of the operator @xmath778
, it follows that @xmath779 for a.e .
@xmath780 and for @xmath754 .
therefore performing the change of variables @xmath781 ( @xmath782 , @xmath783 and @xmath784 ) on the right hand side of we obtain @xmath785 @xmath786 , for a.e .
@xmath780 and for @xmath787 , which proves .
let @xmath306 . from , and
, it follows that @xmath788_{v_i=\widehat{z_i - z_{i+1}},\ i=0\ldots m-1 } dz_1\ldots dz_{m-1}dz_mdv_m,\nonumber\end{aligned}\ ] ] for a.e .
@xmath780 and for @xmath540 . then performing the change of variables @xmath789 `` @xmath790 '' @xmath791 ( `` @xmath790''@xmath792 , @xmath775 , @xmath793``@xmath790''@xmath794``@xmath790''@xmath795 ) , we obtain @xmath796 for a.e . @xmath780 and for @xmath540 , which proves .@xmath35
this work was supported in part by nsf grants dms-0554097 and dms-0804696 . | we consider the reconstruction of optical parameters in a domain of interest from photoacoustic data .
photoacoustic tomography ( pat ) radiates high frequency electromagnetic waves into the domain and measures acoustic signals emitted by the resulting thermal expansion .
acoustic signals are then used to construct the deposited thermal energy map .
the latter depends on the constitutive optical parameters in a nontrivial manner . in this paper , we develop and use an inverse transport theory with internal measurements to extract information on the optical coefficients from knowledge of the deposited thermal energy map .
we consider the multi - measurement setting in which many electromagnetic radiation patterns are used to probe the domain of interest . by developing an expansion of the measurement operator into singular components ,
we show that the spatial variations of the intrinsic attenuation and the scattering coefficients may be reconstructed .
we also reconstruct coefficients describing anisotropic scattering of photons , such as the anisotropy coefficient @xmath0 in a henyey - greenstein phase function model . finally , we derive stability estimates for the reconstructions
. * keywords . * photoacoustics , optoacoustics , transport equation , inverse problems , stability estimates , internal measurements . |
detection of terrestrial - mass extra - solar planets ( exoplanets ) requires extremely stable and precise instruments . in the strive to reach extreme levels of doppler precision
, a number of previously uncharacterized instrumental effects will begin to dominate instrument error budgets .
dedicated doppler radial velocity ( rv ) instruments must be able to measure shifts in stellar spectra at precisions approaching 10 cm s@xmath0 in order to detect true earth twins orbiting nearby stars . in the pursuit of reaching this bold precision goal ,
a number of previously uncharacterized instrument noise sources must be studied and mitigated at unprecedented levels . in the past decade , much attention has been paid to maximizing detection sensitivity by using optical fibers to deliver light from the telescope to the spectrograph .
typical multi - mode fibers ( mmfs ) used in current generation seeing - limited doppler instruments have the convenient ability to scramble light , producing an output illumination that is significantly , though not perfectly , decoupled from the incident flux distribution @xcite .
however , these fibers do suffer from mode interference effects , commonly referred to as modal - noise , which can impose a fundamental limit on achievable measurement precision if not addressed properly @xcite .
this has been shown to severely limit both achievable signal - to - noise on stellar targets @xcite , and ability to realize the full potential of coherent frequency calibration sources @xcite .
it is important to note that insufficient image scrambling and fiber modal noise can both limit measurement precision , but are fundamentally different phenomena , as described in @xcite .
as precision goals approach 10 cm s@xmath0 , a number of subtle and largely unexplored instrumental systematics will begin to dominate overall performance error budgets .
more recently , the use of smfs for compact doppler spectrometers has been suggested as a method of minimizing instrument size while overcoming many of these systematics @xcite .
these fibers support only a single spatial propagation mode and therefore do not suffer from the classical modal interference effects of mmfs .
the output intensity distribution of a smf is entirely decoupled from input illumination variations , yielding a fundamentally stable instrument point - spread function ( psf ) with perfect radial and azimuthal scrambling .
these qualities make smfs an attractive choice when designing compact , stabilized doppler velocimeters for instruments on small telescopes @xcite , where mode - matching is relatively efficient , and for larger telescopes with high performance adaptive optics ( ao ) systems working near the diffraction limit @xcite .
while these fibers can produce a stable and perfectly scrambled output illumination , typical smfs do support two fundamental polarization modes . similarly to mmfs , imperfections and internal stresses in the fiber can lead to variable coupling between these two polarization modes @xcite . in smfs , this leads to an effective polarization change of propagating light .
in fact , a single fiber cable can very efficiently rotate or circularize any arbitrary input polarization to an arbitrary output polarization . as such
, stressed single - mode fiber cables are frequently used as in - line wave - plates for polarization rotation of monochromatic sources @xcite .
10% ) polarized supercontinuum source .
top : effective polarization rotation induced by variable twisting of the fiber cable .
bottom : bend - induced polarization rotation measurement . in both cases , the effective polarization angle exiting the fiber varied by 10 s of degrees with minimal stress applied.,width=326 ] here we study the impact of polarization rotation in single - mode fibers on radial velocity measurements by examining the polarization sensitivity of the primary disperser ( in this case , an echelle reflection grating ) used in most doppler spectrometers .
the purpose of this study is to emphasize the scope and magnitude of this effect , and show specifically how this affects high precision rv measurements .
this is of particular importance since there are several instruments currently being designed or built that utilize smf delivery systems , including minerva - red @xcite , ilocator @xcite , and tests of an extreme precision doppler spectrometer for the subaru telescope @xcite .
with any fiber - fed instrument , the internal stresses within the fiber will change as the fiber is perturbed ( e.g. due to telescope tracking , temperature variations , etc . )
this variable stress can change the intrinsic fiber birefringence , which alters the polarization distribution exiting the fiber .
the consequences of this variable birefringence have been studied for interferometric applications ( e.g. @xcite ) , as birefringent effects in standard smfs can degrade fringe contrast @xcite , but they have yet to be thoroughly studied in the context of precision doppler spectroscopy .
the goal of this study is to estimate how these birefringence effects propagate to spectroscopic velocity errors . as such
, we do not present a rigorous mathematical model of fiber birefringence in typical smfs , as this has been abundantly documented and studied in the literature over the past several decades ( e.g. @xcite ) .
instead , we present a brief summary of the scope of the problem and the relevance to astronomical spectroscopy . in smfs ,
any stresses induced on the fiber core , e.g. due to variations in applied pressure on the cable @xcite , bends @xcite , twists @xcite , thermal fluctuations @xcite , or even variations in external magnetic fields @xcite , will lead to variable polarization mode coupling in the fiber core .
this variable coupling will , in effect , rotate the polarization of light propagating through the fiber .
since most smfs typically support two polarization modes , this variable mode coupling can lead to a manifestation of fiber modal noise . for astronomical instruments utilizing dedicated calibration fibers , any variations in the spectral profile due to variable polarization
will not be traced with a calibration source , as the polarization state is specific to the source and fiber pair .
it is important to note that any polarization rotation will also have a wavelength dependence , which depends strongly on the polarization beat length of the fiber in question @xcite and the internal and external stresses applied to the fiber core @xcite .
polarization - maintaining ( pm ) fibers have the ability to maintain a single polarization with a high polarization extinction ratio , but rely on stable and precise alignment of the incident polarization with the fast axis of the fiber for efficient mode propagation .
this generally requires the incident illumination to be strongly linearly polarized , which is usually an impractical choice for most astronomical applications due to the associated efficiency losses .
while ample amounts of literature exists that studies the effects of stress - induced polarization rotation in smfs ( see previous references ) , we perform a series of simple laboratory tests to show the magnitude of possible polarization fluctuations expected from typical stresses associated with astronomical telescope fiber feeds ( namely twists and bends effects ) illuminated by a weakly polarized continuum source , meant to emulate a possible astrophysical target . a 2 meter commercial smf cable ( corning smf-28 ) is illuminated by a supercontinuum source that is intrinsically weakly polarized ( degree of polarization is approximately 10% , as measured by our polarimeter ) .
the output polarization of the fiber cable is then recorded using a commercial pmi - nir meadowlark polarimeter .
an h - band interference filter ( 1.5 - 1.65 @xmath1 m ) is inserted at the fiber exit to ensure the illumination bandpass is restricted to wavelengths where the fiber supports a single spatial mode ( 1.26 - 1.65 @xmath1 m ) .
the fiber is coiled onto a linear translation stage , which allows for a variable stress to be induced in the cable , and terminated with standard telecommunication fc connectors on either side .
a schematic of the experimental setup is shown in figure [ fig : setup ] .
the cable is then perturbed by both translating the linear stage ( inducing a mild bend ) and twisting the fiber by hand . for twist measurements ,
the cable was periodically twisted by hand between 0 and 90 degrees across a 0.5 meter section of the cable . for bend measurements ,
the cable was bent between two fixed mounting points , separated roughly 0.3 meters apart , using a mechanical stage .
this varied the overall fiber bend radius by several inches .
we find that , even for low - levels of added stress , the effective output polarization angle of light exiting the fiber is highly variable .
measurements of the fiber output polarization are shown in figure [ fig : lab_data ] .
mild twists and bends of the fiber cable can induce effective polarization rotations of over 10 degrees .
to estimate the direct effects of polarization variations on measured radial velocities , we focus on the polarization sensitivity of the primary dispersive element in the spectrograph : in this case , an r4 echelle grating .
we use measured @xmath2 and @xmath3 polarization curves for two newport rgl gratings , manufactured with two different grating masters : mr 160 , used on harps @xcite , and mr 257 , used on hpf @xcite and carmenes @xcite instruments , to modulate simulated stellar spectra ( see figures [ fig : grating_response_vis ] & [ fig : grating_response_nir ] ) .
the modulation is a linear combination of different ratios of the measured grating blaze functions for the two orthogonal incidence polarizations , for a variety of spectral orders in the visible and near - infrared ( nir ) .
model spectra are assumed to be noiseless , and are generated at an effective instrument resolution of @xmath4=100,000 . for the mr160 grating , we use a synthetic g2 spectrum ( t@xmath5 = 5700 k ) for radial velocity calculations for four spectral orders between 350 and 700 nm . for the mr257 grating , where our available data is centered in the nir
, we use a synthetic m - dwarf spectrum as a stellar template ( in this case a 3000 k bt - settl model ) .
the variable blaze modulation due to changing polarization can induce apparent velocity offsets . on the smallest scale
, it can introduce a variable gradient across individual spectral features , although this is likely a negligible change . across an order , it changes the relative chromatic weighting of lines , and shifts the center of the cross - correlation peak in that wavelength range .
most importantly , it changes the shape of the aggregate cross - correlation function , and the final determination of radial velocity from a single observation .
these effects are similar to what would be observed due to variable atmospheric refraction effects ( e.g. due to airmass changes , variations in atmospheric dispersion correction , etc . ) .
relative velocity offsets are calculated using weighted mask cross - correlation method ( similar to the methods described in @xcite and @xcite ) with a custom stellar template mask built specifically for this purpose for each spectral class considered ( approximately g2 and m5 ) . for the latter , telluric contamination is neglected since a synthetic spectrum is used , but this could be easily adapted for use with an observed spectrum with a slightly more rigid selection procedure for mask lines .
velocity offsets at all effective polarization combinations are then compared to spectra modulated by an unpolarized blaze function ( average of @xmath2 & @xmath3 ) .
spectra for each order were purposely not continuum normalized when calculating velocity offsets , as the blaze functions contain important information relating to the weights of various spectral features used for the cross - correlation . measured
velocity offsets between polarization angles are shown in figure [ fig : rv_offset ] .
we note that effective shifts of several m s@xmath0 are seen between orthogonal polarizations for nearly all measured echelle orders . even for weakly polarized light
, low - level polarization rotation due to variable mode coupling can still introduce errors at the 10 s of cm s@xmath0 level .
for strongly polarized , coherent sources , such as laser frequency combs , this effect will be even more pronounced .
if not dealt with , this systematic error could conceivably compromise the exquisite stability capabilities of dedicated smf doppler velocimeters .
periodic mechanical agitation of the fiber cable will change the polarization mode coupling of incident radiation .
this method is frequently used to temporally vary the phase distribution of modes in mmfs , which can significantly reduce classical mode interference effects @xcite . in smfs , this agitation could conceivably scramble the effective mode coupling . this could reduce the effective degree of polarization over typical astronomical integration times , though may result in an efficiency drop due to increased bend losses depending on the amplitude and frequency of agitation .
it has been shown that using long fiber cables can effectively act as depolarizers for strongly polarized sources @xcite , though likely not at the level required to completely mitigate this issue for precise rv measurements .
in addition , it may be possible to mitigate this effect post factum through continuum normalization or `` deblazing '' of the spectrum before calculating rvs .
for example , using the continuum normalization routines included in the _ reduce _ package @xcite , we are able to mitigate the rv disparity by about an order of magnitude . depending on the precision level desired
, this might be the most convenient solution , but will remain sensitive to the normalization procedure , and might produce rv offsets as the fits to the varying order response functions change .
additionally , important spectral weighting information , such as absolute chromatic flux level across a given spectral order , is omitted by dividing by a estimated blaze function .
while we have focused entirely on standard smfs in this study , it is import to note that the polarization effects discussed here could conceivably affect multi - mode fiber delivery systems at the cm s@xmath0 level .
this is particularly true for instruments using small diameter , asymmetric core geometry fibers .
these fibers can significantly improve scrambling performance over standard circular - core fibers , but may also be prone to maintaining measurable levels of residual polarization @xcite .
harps has observed subtle , though measurable , variations in measured blaze functions at the @xmath610@xmath7 level @xcite . while these variations could be attributed to a number of error sources ( such as seeing variations , telescope focus drifts , airmass variations , etc . ) , residual polarization in the instrument delivery fiber could also be a contributing factor , and should be considered along with more commonly examined sources of variability .
we present a cautionary note that variable coupling of polarization modes in single - mode fibers can cause systematic shifts in measured radial velocities for spectrographs using standard echelle reflection gratings as the primary disperser .
this effect is comparable to classical modal interference in multi - mode fibers , though it manifests in a more subtle manner .
we perform a simple laboratory test to explore the effects of external stresses on polarization mode coupling by measuring the output polarization of a commercial smf illuminated by a weakly polarized continuum source under external stress . to quantify the magnitude of this effect on rv measurements , we calculate the effective doppler shift of a template spectrum modulated by measured echelle grating response functions for several spectral orders for two different manufactured gratings .
we use a linear combination of these response functions to span a wide range of effective polarizations incident on the grating .
we show that , even for low - level rotation of weakly polarized light , rv offsets of 10 s cm s@xmath0 are entirely possible for instruments using smfs .
this subtle effect is not traced with a simultaneous calibration source , and can therefore lead to insidious systematic shifts in stellar radial velocity measurements . in the era of extreme precision doppler spectrometers striving towards 10 cm
s@xmath0 precision , this effect could conceivably dominate measurement error for smf - based instruments . , m. , tozzi , a. , sanna , n. , zangrilli , l. , & oliva , e. 2014 , in society of photo - optical instrumentation engineers ( spie ) conference series , vol .
9147 , society of photo - optical instrumentation engineers ( spie ) conference series , 66 , n. , guyon , o. , martinache , f. , schwab , c. , & cvetojevic , n. 2014 , in society of photo - optical instrumentation engineers ( spie ) conference series , vol .
9147 , society of photo - optical instrumentation engineers ( spie ) conference series , 7 , s. , ramsey , l. w. , terrien , r. , et al .
2014 , in society of photo - optical instrumentation engineers ( spie ) conference series , vol . 9147 , society of photo - optical instrumentation engineers ( spie ) conference series , 1 , k. s. , ramsey , l. , mahadevan , s. , halverson , s. , & redman , s. l. 2012 , in society of photo - optical instrumentation engineers ( spie ) conference series , vol .
8446 , society of photo - optical instrumentation engineers ( spie ) conference series , 8 , a. , amado , p. j. , caballero , j. a. , et al .
2014 , in society of photo - optical instrumentation engineers ( spie ) conference series , vol . 9147 , society of photo - optical instrumentation engineers ( spie ) conference series , 1 , j. , stahl , o. , schwab , c. , et al .
2014 , in society of photo - optical instrumentation engineers ( spie ) conference series , vol .
9151 , society of photo - optical instrumentation engineers ( spie ) conference series , 52 | exploring the use of single - mode fibers ( smfs ) in high precision doppler spectrometers has become increasingly attractive since the advent of diffraction - limited adaptive optics systems on large - aperture telescopes .
spectrometers fed with these fibers can be made significantly smaller than typical seeing - limited instruments , greatly reducing cost and overall complexity .
importantly , classical mode interference and speckle issues associated with multi - mode fibers , also known as modal noise , are mitigated when using smfs , which also provide perfect radial and azimuthal image scrambling .
however , these fibers do support multiple polarization modes , an issue that is generally ignored for larger - core fibers given the large number of propagation modes . since diffraction gratings used in most high resolution astronomical instruments have dispersive properties that are sensitive to incident polarization changes , any birefringence variations in the fiber can cause variations in the efficiency profile , degrading illumination stability . here
we present a cautionary note outlining how the polarization properties of smfs can affect the radial velocity measurement precision of high resolution spectrographs .
this work is immediately relevant to the rapidly expanding field of diffraction - limited , extreme precision rv spectrographs that are currently being designed and built by a number of groups . |
entanglement based cryptography @xcite , equivalent formally to the bb84 scheme @xcite , is the one that uses the power of quantum entanglement monogamy obeyed by a maximally entangled pure quantum state . if the state is noisy then in some cases it is possible to run an entanglement distillation process @xcite which may be interpreted as quantum privacy amplification @xcite .
since the final output is maximally entangled , it may be used directly for secret key generation .
the efficiency of this procedure is quantified with distillable entanglement @xmath2 , which defines how many singlet states can be obtained in the asymptotic regime per one input . still it was known that certain states which can not be prepared by local operations and classical communication ( locc ) are not distillable , exhibiting the phenomenon of bound entanglement @xcite . for a long time
bound entanglement was believed to be useless for cryptography , but several years ago it was shown @xcite , that at least some bound entangled states may be useful in quantum cryptography .
this is one extreme instance of the general fact that the amount of distillable secure key @xmath3 may exceed the amount of distillable singlets @xmath4 .
the latter effect has been verified in a recent experiment @xcite .
the key ingredient in the complete theory of distilling a secret key from quantum states @xcite is the notion of a _ private bit _ , p - bit or more generally a private dit , p - dit which is a delocalized maximally entangled state that still retains some entanglement monogamy result .
a quantum p - dit is composed from a @xmath5 key part @xmath6 , and the shield part @xmath7 , shared between alice ( subsystems @xmath8 ) and bob ( subsystems @xmath9 ) in such a way that the local von neumann measurements on the key part in a _
particular _ basis will make its results completely statistically uncorrelated from the results of any measurement of an eavesdropper eve on her subsystem @xmath10 , which is a part of the purification @xmath11 of the p - dit state @xmath12 .
there is a nice explanation how the shield part protects the statistics of the measurement on @xmath13 and @xmath14 to be correlated to eve : it just makes it impossible to distinguish the results of the measurement by an external observer .
an obvious way to determine privacy properties is to reconstruct tomographically the complete p - dit state @xmath12 .
this however is a very time consuming process , especially if the system under investigation is high - dimensional .
the aim of the present paper is to give bounds on the distillable secure key based just on few observables .
this advances further the study presented in ref .
@xcite , where it was proposed to carry out a tomography of the so - called privacy - squeezed state of the state of merit .
we demonstrate that a single observable suffices to provide a nontrivial bound .
we also provide more accurate estimates based on two observables .
these results provide tools for application - specific detection of entanglement , refining the fundamental concept of the entanglement witness proposed in @xcite and @xcite that can be also subjected to optimization with respect to local measurements @xcite and used to quantify the amount of entanglement @xcite .
the present results can be viewed as an outcome of a more general research program : _ experimentally friendly detection / estimation of task - dedicated quantum entanglement and/or correlations_. in fact it is quite usual that we are interested in that aspect of entanglement which is useful for specific quantum information task .
the quantity characterizing this aspect may be a monotone but we believe that it need not to be in general .
for instance it is known that there are cases when specific bell inequalities which are important for device independent cryptography are better violated by nonmaximally entangled states . in this context we believe that the present paradigm will lead to systematic development of experimentally friendly detection / estimation of resources for quantum information tasks .
this paper is organized as follows . in sec .
[ sec : keybounds ] we elaborate on lower bounds on distillable entanglement and distillable key . in sec .
[ sec : single ] we present a lower bound on distillable key in terms of single parameter , i.e. single privacy witness
. an approximate version of this bound is presented in appendix . in sec .
[ ref : double ] we discuss how to infer privacy of a noisy state from the expectation values of two observables .
finally , sec .
[ sec : conclusions ] concludes the paper .
let us start by reviewing how an individual observable can be used to estimate distillable entanglement @xmath2 .
the most natural observable in this context is a projector @xmath15 onto a maximally entangled state @xmath16 of two @xmath17-level systems , where @xmath18 and @xmath19 are any two orthonormal bases .
following the idea of ref .
@xcite dealing with the case @xmath20 , there is a protocol for an arbitrary @xmath17 such that if @xmath21 satisfies @xmath22 then the state @xmath23 is distillable @xcite .
the corresponding rate of the distillation protocol can be easily estimated from below by the hashing protocol @xcite which gives an estimate for the distillable entanglement as @xmath24 where @xmath25 denotes the von neumann entropy .
since an application of the so - called @xmath26 twirling @xcite can only decrease distillable entanglement @xmath4 we may twirl the state @xmath27 in order to bring it to a highly symmetric form and then apply the hashing inequality ( [ hashing ] ) which eventually gives : @xmath28 where @xmath29 and @xmath30 is the shannon entropy .
the above formula is valid for any bipartite quantum state @xmath27 .
there are more sophisticated twirling protocols .
for instance for two qubits there is a protocol @xcite utilizing selected random pauli operations that brings the state to a form diagonal in the bell basis : @xmath31 where @xmath32 are bell states . applying the hashing protocol to
@xmath33 leads to : @xmath34 eqs .
( [ fidelity ] ) and ( [ eq : ed>=bell ] ) provide bounds on the key rate for @xmath27 , as distilled singlet states can be used for the standard ekert protocol . in general
however , this may be not the most efficient way to generate the key . as pointed out in the introduction , a straighforward way to verify that we have a p - bit or a state close to p - dit enough that the secret key may be distilled .
it is based on the so - called quantum state tomography which allows to calculate specific entropic functions that can be used to estimate the amount of the secret key from below .
a useful quantity is the devetak - winter function @xmath35 which quantifies the secret key distillable through a specific one - way secret key distillation protocol : @xmath36 the devetak - winter function is given explicitly by the difference of two holevo quantities @xmath37 and @xmath38 which characterize the amount of information that can be recovered respectively by bob and eve from subsystems @xmath14 and @xmath10 after alice carried out a von neumann measurement of the subsystem @xmath13 in the secure basis @xmath39 : k^_dw ( _ abab)= _ ab(_e_abe^ ) - _ ae ( _ b_abe^ ) .
the state @xmath40 is given by : @xmath41 let us now restrict our attention to a situation when the key part is composed of two qubits @xmath13 and @xmath14 .
the complete density matrix @xmath42 can be written as a @xmath43 array of blocks @xmath44 : @xmath45 it can be transformed by locc ( with respect to the partition @xmath46 ) to the form : @xmath47 which can be , , unwisted to bell diagonal matrix which is directly related to privacy squeezed state ( see @xcite ) : @xmath48 where the norm @xmath49 stands for the trace norm .
it will be convenient to write @xmath50 in the form : @xmath51 it is easy to see that @xmath50 is diagonal in the bell basis and the parameters @xmath52 are occupation probabilities of the corresponding bell states
. there is a useful bound on the secret key which is @xcite @xmath53 here @xmath54 is the classical mutual information for the outcomes of von neumann measurements carried out by alice and bob in the secure basis @xmath55 . as the joint probability distribution for these outcomes is @xmath56 , we have @xmath57 , where @xmath58 denotes the binary entropy . note that the bound given in eq .
( [ eq : kd<=p1 ... p4 ] ) is weaker than that on distillable entanglement in eq .
( [ eq : ed>=bell ] ) .
this is because the state @xmath50 does not actually describe the physical system at any stage of the protocol , but is rather a formal construct characterizing privacy properties of the state @xmath59 of the complete system @xmath60 .
the class of secrecy witnesses we shall consider here is defined formally as @xmath61 where @xmath62 is any hermitian matrix satisfying @xmath63 acting on the shield subsystems @xmath64 and @xmath65 .
we will use the expectation value of the secrecy witness @xmath66 to approximate the value of @xmath67 .
in fact we have : w : = | _ |= [ ( _ 00,11 + _ 11,00 ) ] + satisfying @xmath63 one has @xmath68 .
indeed , @xmath63 implies @xmath69 and consequently @xmath70 . sometimes the parameter @xmath71 may give exactly the value of @xmath72 .
we will give an instance of that at the end of this section .
let us also stress that the witness itself ( [ witness ] ) is an observable which must be measured on the original state ( [ initial - state ] ) .
quantitative estimation of the distillable key will be based on the following inequality @xmath73 applied to eq .
( [ eq : kd<=p1 ... p4 ] ) , which yields : @xmath74 as we are interested in the most pessimistic scenario , we need to minimize the right hand side over pairs @xmath75 that satisfy all the constraints .
this gives the _ central formula _
: @xmath76 \label{centralbound}\end{gathered}\ ] ] we found numerically the boundary value @xmath77 at which the above bound becomes strictly positive , i.e. the witness condition @xmath78 , to be equal to @xmath79 .
we can simplify the optimization in eq .
( [ centralbound ] ) by introducing a new pair of variables @xmath80 and @xmath81 .
a straightforward calculation shows that the bound for the key expressed in the new variables takes the form : @xmath82 because the lower limit for @xmath83 is greater or equal to @xmath84 , optimization over @xmath83 for a fixed value of @xmath85 yields @xmath86 .
consequently , the minimization in eq .
( [ eq : kd>=p+xi+ ] ) needs to be carried out over a single parameter @xmath85 : @xmath87.\ ] ] the absolute minimum of this expression is analyzed in appendix .
however we can simplify the bound in two ways leading to weaker but more compact formulas : _ weaker bound 1 : _ suppose that @xmath88 which is equivalent to @xmath89
. then we have the following estimate : @xmath90 in the last inequality we have used the fact that both @xmath91 and @xmath92 .
_ weaker bound 2 : _ there is a possibility of having another bound with help of subadditivity of the entropy @xmath93
\leq h(p_1+p_2 ) + h(p_1+p_3)$ ] .
@xmath94 for a graphic comparison of the derived formulas see fig .
[ fig : singlewitness ] . derived in eqs .
( [ centralbound ] ) ( solid line ) , ( [ weaker - bound-1 ] ) ( dashed line ) , and ( [ weaker - bound-2 ] ) ( dotted line).,width=226 ] the above considerations are based on the ccq scenario , when alice and bob measure their qubits in the secure key basis @xmath95 , @xmath96 .
however if the shield part is not affected by local unitary the optimal @xmath97 remains unchanged .
for instance , in the case of the four - qubit state whose approximate version was realized experimentally in ref .
@xcite , a two - qubit swap operator @xmath98 used in the privacy witness would give _ exactly _ the value @xmath99 .
note also that if @xmath97 were not hermitian , we could decompose @xmath100 and measure two observables @xmath101 and @xmath102 .
in this section we will show how to characterize the privacy properties of a noisy entangled state from expectation values of two observables .
the first one , given by @xmath103 , characterizes correlations between measurements performed on the subsystems @xmath13 and @xmath14 in the key basis .
the security of the key will be inferred using the expectation value of an observable @xmath104 . in the further discussion
it will be convenient to use the following parameterization of @xmath50 : @xmath105 the nonnegative parameters @xmath80 and @xmath106 can be interpreted as occupation probabilities of the correlated and the anticorrelated subspaces , spanned respectively by pairs or vectors @xmath107 , @xmath108 and @xmath109 . the other two parameters , given explicitly by @xmath110 and @xmath111 , characterize the off - diagonal elements of @xmath50 respectively in the correlated and the anticorrelated sectors and therefore contain information about the privacy properties . because the off - diagonal elements of @xmath50 are nonnegative due to the definition given in eq .
( [ eq : sigmaabdef ] ) and must ensure positive definiteness of @xmath50 , the parameters @xmath112 satisfy the inequality : @xmath113 i.e. the relevant region for pairs @xmath114 has the shape of a square . in the new parameterization ,
the lower bound on the key takes the following form : @xmath115 because the binary entropies @xmath116 and @xmath117 are nonnegative , a necessary condition for this bound to be nontrivial is @xmath118 , otherwise the right hand side is not positive .
this means that @xmath85 must satisfy either @xmath119 or @xmath120 , where @xmath121 is the bigger of two solutions of a transcedental equation @xmath122 on the interval @xmath123
. we will restrict our further discussion only to the case @xmath120 , as the analysis of the second case @xmath119 is completely analogous . let us now discuss how the parameters of @xmath50 are related to measured observables .
the parameters @xmath124 can be evaluated directly from the measured observables as @xmath125 . following discussion after eq .
( [ eq : kdbound ] ) , we will be interested in the regime when @xmath126 .
considering the other regime when @xmath127 effectively boils down to swapping the roles of the correlated and the anticorrelated subspaces .
these two possibilities can be analyzed jointly by defining @xmath128 and using in further discussion @xmath129 the condition @xmath126 can be equivalently written as @xmath130 which defines the minimum value of @xmath131 above which the bound on the key can become nontrivial .
the second quantity we will use in our analysis will be @xmath132 it allows us to bound the parameters @xmath133 and @xmath83 according to the following inequality which will be at the heart of our reasoning : @xmath134 \bigr| \nonumber \\ & \le ||\hat{a}_{00,11 } + \hat{a}_{11,00 }
|| + || \hat{a}_{01,10 } + \hat{a}_{10,01 } || \nonumber \\ & = p_+(2\xi_+ -1 ) + p_-(2\xi_- -1 ) . \label{eq : cxp1p2p3p4}\end{aligned}\ ] ] for fixed @xmath124 ,
this inequality determines the allowed region of @xmath114 in the square defined by eq .
( [ eq : squareofxis ] ) as @xmath135 when evaluating the lower bound on the key rate @xmath136 according to eq .
( [ eq : kdbound ] ) we are interested in the worst case scenario that is consistent with the measurement results .
therefore our task is to minimize the right hand side of eq .
( [ eq : kdbound ] ) under constraints given by eqs .
( [ eq : squareofxis ] ) and ( [ eq : p+xi+p - xi- ] ) .
this is equivalent to maximizing under the same constraints a concave function @xmath137 the lower bound on the key can be written as @xmath138 where @xmath139 is the maximum of @xmath140 over the allowed region of parameters .
it is useful to note that because @xmath140 is a convex linear combination of binary entropies @xmath117 and @xmath116 , within the square given by eq .
( [ eq : squareofxis ] ) decreasing either of the arguments @xmath133 or @xmath83 will always increase the value of @xmath140 .
this in turn implies that @xmath139 is reached on the line @xmath141 to proceed with the maximization , let us start from an observation that if @xmath142 , i.e. the correlated sector of the density matrix has zero off - diagonal elements , no positive key rate can be guaranteed by eq .
( [ eq : kdbound ] ) .
this follows from a straightforward fact that the expression @xmath143 is nonpositive for @xmath120 .
therefore no point with @xmath142 should satisfy eq .
( [ eq : p+xi+p - xi- ] ) . because the slope of the line ( [ eq : linexi - xi+ ] ) is negative , it is sufficient to require that the point @xmath144 is outside the allowed region .
this is equivalent to the inequality : @xmath145 further analysis depends on whether the point @xmath146 is located within the allowed region of parameters .
it is easy to verify that this is determined by the relation between @xmath85 and @xmath147 . if @xmath148 , this point satisfies eq .
( [ eq : p+xi+p - xi- ] ) and the allowed region of parameters has the shape of a trapezoid , as shown in fig .
[ fig : xisquare](a ) .
consequently , all values @xmath149 are allowed .
on the other hand , when @xmath150 , the allowed region is a triangle , as depicted in fig .
[ fig : xisquare](b ) . the minimum allowed value of @xmath133 is then @xmath151 .
we can combine these two cases by defining @xmath152 the maximization of @xmath153 over the line defined in eq .
( [ eq : linexi - xi+ ] ) can now be written as a supremum over a single parameter : @xmath154 which inserted into eq .
( [ eq : kd = fmax ] ) yields the final form of the bound : @xmath155 \label{eq : kdfullbound}\end{gathered}\ ] ] where @xmath156 . the results of a numerical evaluation of the supremum are shown in fig .
[ fig : wxqz](a ) . used to maximize the function @xmath153 defined in eq .
( [ eq : fxi - xi+def]).,width=324 ] and @xmath131 obtained from ( a ) full optimization over the free parameters specified in eq .
( [ eq : kdfullbound ] ) and ( b ) a weaker estimate according to eq .
( [ eq : kdweakerbound ] ) .
the bound is positive in the shaded area , with solid contours drawn at steps @xmath157 starting from @xmath158.,width=288 ] it is also possible to derive a slightly weaker bound that requires no numerical maximization , which we will denote as @xmath159 . because the function @xmath140 is monotonic in each one of its two arguments , as discussed after eq .
( [ eq : kd = fmax ] ) , we have @xmath160 where @xmath161 and @xmath162 are the smallest possible values of @xmath133 and @xmath83 within the allowed region defined by eqs .
( [ eq : squareofxis ] ) and ( [ eq : p+xi+p - xi- ] ) .
the value @xmath161 has been given explicitly in eq .
( [ eq : xi - min ] ) , while it is easy to verify that @xmath163 the simplified bound therefore takes the following explicit form : @xmath164 which is shown in fig . [ fig : wxqz](b ) . finally , let us note that the observable @xmath71 defined in eq .
( [ assumption ] ) can be combined with @xmath131 to yield a stronger bound on the distillable key . to derive this bound ,
let us return to eq .
( [ eq : kdbound ] ) and estimate the last two terms on the right hand side . the inequality shown in eq .
( [ assumption ] ) rewritten in the new parametization provides a lower bound on @xmath83 : @xmath165 because the right hand side is greater or equal @xmath84 , we have @xmath166 .
further , we obviously have @xmath167 .
this yields : @xmath168 - 2 h \left ( p_+ \right ) .
\label{eq : kdwwz}\ ] ] let us note that physical values of @xmath71 and @xmath131 must satisfy the condition @xmath169 , otherwise we would have @xmath170 and consequently @xmath171 . in fig .
[ fig : kdwwz ] we depict the bound ( [ eq : kdwwz ] ) in the physical region of @xmath71 and @xmath131 . )
obtained from observables @xmath71 and @xmath131 .
the physical region is restricted to points @xmath172 satisfying the condition @xmath169 .
the coding of the bound value is the same as in fig .
[ fig : wxqz].,width=288 ]
in this paper we have introduced a privacy witness , i.e. an observable , whose mean value allows one to estimate nontrivially from below the amount of secret key , even in the case when the resulting state exhibits the separation of the secret key and the distillable entanglement . in fact this separation may be quite drastic while the witness is working perfectly . to see this
let us take @xmath7 being @xmath173-level systems and consider a p - bit state @xmath12 in the form @xcite @xmath174\end{aligned}\ ] ] where @xmath175 and @xmath176 stand respectively for bipartite identity and swap operator on the subsystems @xmath7 . in the limit of large @xmath17 the distillable entanglement
is bounded by vanishing log - negativity @xmath177 , while the privacy witness @xmath178 gives us the value of lower bound @xmath179 since @xmath180 and then just either of the weaker bounds given in eqs .
( [ weaker - bound-1 ] ) or ( [ weaker - bound-2 ] ) does the job .
note in particular that since the key part @xmath6 is a two qubit part the above estimate gives the maximum possible value of the secret key @xmath181 despite the fact that the distillable entanglement of the state is almost zero ! in general the complexity of measuring the privacy witness
is related to the hilbert - schmidt decomposition of the hermitian operator @xmath97 used to construct the witness @xcite . in the case of the four qubit state that was studied in the experiment reported in ref .
@xcite the operator in question is the swap operator which is composed of three terms involving products of pauli matrices : @xmath182 taking into account the necessary measurements on the key part this gives in total @xmath183 observables to be measured , each formed by a product of four pauli matrices .
this is dramatically less then the full tomography which requires @xmath1 products of four pauli matrices .
note that in some cases , like the p - bit state discussed above , such an apparently poor measurement has no problem in reporting the key value that lies above the distillable entanglement , which is bounded for our example by the log negativity value @xmath184 .
the above approach may be extended to higher dimensions and other twirling techniques may be applied .
it may be especially useful when the experimentalist has a good guess about the expected p - bit he state in the laboratory then he or she may estimate the high key contents almost perfectly even if there is virtually no distillable entanglement in the system .
finally let us note that the very difficult problem is to find the nonlinear entanglement witness that would capture collective behavior revealing the key in all the cases when any single - copy entropic fuction based on a one - way protocol fails .
it seems that for this one needs quantum secrecy distillation protocols of new generation .
we believe that the present approach will lead to general and systematic development of experimentally friendly methods for detection and estimation of task - dedicated quantum entanglement and other resources .
this work was supported by 7th framework programme future and emerging technologies projects corner ( no .
213681 ) and q - essence ( no .
is supported by grant bw no .
538 - 5300 - 0637 - 1 , p.h .
is supported by nn 202 - 261 - 938 .
we acknowledge useful discusions with r. demkowicz - dobrzaski .
let us denote @xmath185 the derivative with respect to @xmath85 reads : @xmath186 it is easy to see that on the interval @xmath187 the argument of the logarithm function runs from @xmath158 to @xmath188 .
therefore @xmath189 reaches its minimum at a root of a polynomial equation @xmath190 .
this is a cubic equation which can be solved exactly , but a simplified formula can be found by substituting @xmath191 and assuming that @xmath192 which is motivated by numerical analysis .
this yields @xmath193 and consequently the bound on the key in the approximate form @xmath194 this approximate expression turns out to reproduce the original bound quite tightly , as evidenced in fig .
[ fig : difference ] depicting the difference between eq .
( [ eq : kdapprox ] ) and the bound given in eq .
( [ eq : kd>=p+xi+ ] ) . | while it is usually known that the mean value of a single observable is enough to detect entanglement or its distillability , the counterpart of such an approach in the case of quatum privacy has been missing .
here we develop the concept of a privacy witness , i.e. a single observable that may detect presence of the secure key even in the case of bound entanglement .
then we develop the notion of secret key estimation based on few observables and discuss the witness decomposition into local measurements .
the surprising property of the witness is that with the help of a low number of product mesurements involved it may still report the key values that are _ strictly above _
distillable entanglement of the state . for an exemplary four - qubit state studied in a recent experiment [ k. dobek _ et al .
_ , phys .
rev .
lett . * 106 * , 030501 ( 2011 ) ] this means @xmath0 pauli operator product measurements versus @xmath1 needed to carry out the complete quantum state tomography .
the present approach may be viewed as a paradigm for the general program of experimentally friendly detection and estimation of task - dedicated quantum entanglement . |
this paper develops a semiparametric two - stage estimator of preference parameters in the binary choice model where the agent s decision rule is affected by conditional expectations of outcomes which are uncertain at the choice - making stage and the preference shocks are nonparametrically distributed with unknown form of heteroskedasticity .
the pioneering papers of manski(1991,1993 ) established nonparametric identification of agents expectations in the discrete choice model under uncertainty when the expectations are fulfilled and conditioned only on observable variables . utilizing this result , ahn and manski(1993 ) proposed a two - stage estimator for a binary choice model under uncertainty where agent s utility was linear in parameter and the unobserved preference shock had a known distribution .
specifically , they estimated the agent s expectations nonparametrically in the first stage and then the preference parameters in the second stage by maximum likelihood estimation using the choice data and the expectation estimates .
ahn(1995,1997 ) extended the two - step approach further . on one hand , ahn ( 1995 ) considered nonparametric estimation of conditional choice probabilities in the second stage . on the other hand , ahn ( 1997 ) retained the linear index structure of the ahn - manski model but estimated the preference parameters in the second stage using average derivative method hence allowing for unknown distribution of the unobservable . in principle , alternative approaches accounting for nonparametric unobserved preference shock can also be applied in the second step estimation of this framework .
well known methods include cosslett ( 1983 ) , powell et al.(1989 ) , ichimura ( 1993 ) , klein and spady ( 1993 ) , and coppejans ( 2001 ) , among many others .
the aforementioned papers allow for nonparametric setting of the distribution of the preference shock .
but the unobserved shock is assumed either to be independent of or to have specific dependence structure with the covariates .
by contrast , manski(1975,1985 ) considered a binary choice model under the conditional median restriction and thus allowed for general form of heteroskedasticity for the unobserved shock .
it is particularly important , as shown in brown and walker ( 1989 ) , to account for heteroskedasticity in random utility models .
therefore , this paper develops the semiparametric two - stage estimation method for the ahn - manski model where the second stage is based on manski ( 1975,1985 ) s maximum score estimator and thus can accommodate nonparametric preference shock with unknown form of heteroskedasticity . from a methodological perspective
, this paper also contributes to the literature on two - stage m - estimation method with non - smooth criterion functions .
when the true parameter value can be formulated as the unique root of certain population moment equations , the problem of m - estimation can be reduced to that of z - estimation .
chen et al.(2003 ) considered semiparametric non - smooth z - estimation problem with estimated nuisance parameter , while allowing for over - identifying restrictions . chen and pouzo ( 2009,2012 ) developed general estimation methods for semiparametric and nonparametric conditional moment models with possibly non - smooth generalized residuals .
for the general m - estimation problem , ichimura and lee ( 2010 ) assumed some degree of second - order expansion of the underlying objective function and established conditions under which one can obtain a @xmath0-consistent estimator of the finite dimensional parameter where @xmath1 is the sample size when the nuisance parameter at the first stage is estimated at a slower rate .
for more recent papers on two - step semiparametric estimation , see ackerberg et al .
( 2012 ) , chen et al .
( 2013 ) , escanciano et al . ( 2012 , 2013 ) , hahn and ridder ( 2013 ) , and mammen et al .
( 2013 ) , among others . none of the aforementioned papers include the maximum score estimation in the second stage estimation . for this paper ,
the second stage maximum score estimation problem can not be reformulated as a z - estimation problem .
furthermore , even in the absence of nuisance parameter , kim and pollard ( 1990 ) demonstrated that the maximum score estimator can only have the cube root rate of convergence and its asymptotic distribution is non - standard .
the most closely related paper is lee and pun ( 2006 ) who showed that @xmath2 out of @xmath3 bootstrapping can be used to consistently estimate sampling distributions of nonstandard m - estimators with nuisance parameters .
their general framework includes the maximum score estimator as a special case , but allowing for only parametric nuisance parameters .
therefore , established results in the two - stage estimation literature are not immediately applicable and the asymptotic theory developed in this paper may also be of independent interest for non - smooth m - estimation with nonparametrically generated covariates .
the rest of the paper is organized as follows .
section 2 sets up the binary choice model under uncertainty and presents the two - stage maximum score estimation procedure of the preference parameters .
section 3 states regularity assumptions and derives consistency and rate of convergence of the estimator .
in addition , section 3 gives conditions under which the two - stage maximum score estimator is asymptotically equivalent to the infeasible single - stage maximum score estimator with a known first stage input .
section 4 presents monte carlo studies assessing finite sample performance of the estimator .
section 5 gives further applications of maximum score estimation with nonparametrically generated regressors .
section 6 concludes this paper .
proofs of technical results along with some preliminary lemmas are given in the appendices .
suppose an agent must choose between two actions denoted by 0 and 1 .
the utility from choosing action @xmath4 is @xmath5realization of the random vector @xmath6 is known to the agent before the action is chosen and the random vector @xmath7 is realized only after the action is chosen .
random vectors @xmath8 and @xmath9 are not necessarily identical .
distribution of @xmath10 depends on the chosen action and realization of a random vector @xmath11 .
let @xmath12 denote the agent s subjective conditional expectation .
given the realization of @xmath13 , the agent chooses the action @xmath14 that maximizes the expected utility : @xmath15thus the decision rule has the form @xmath16{}^{\prime } \beta { } _ { 2}>\varepsilon \right\ } , \label{e1}\]]where @xmath17 , and @xmath18 is an indicator function whose value is one if the argument is true and zero otherwise . as in ahn and manski ( 1993 )
, suppose that expectations are fulfilled : @xmath19we assume that the researcher does not observe realization of @xmath20 and @xmath21 , but that of @xmath22 .
let @xmath23 and let @xmath24 , where @xmath25 denotes the support of the distribution of @xmath26 .
then , equation ( [ e1 ] ) can be written as @xmath27where @xmath28 is a vector of unknown preference parameters .
the set of assumptions leading to the binary choice model in ( [ eq ] ) is equivalent to that of ahn and manski ( 1993 , equations ( 1)-(3 ) ) .
note that @xmath29 affects the agent s decision only through @xmath30 , and therefore , @xmath29 and @xmath31 can have common elements , as long as the support of the distribution of @xmath26 is not contained in any proper linear subspace of @xmath32 . in this paper , we consider an important deviation from ahn and manski ( 1993 ) s setup where the unobserved preference shock @xmath20 is independent of @xmath33 with a known distribution function .
instead , we consider inference under a flexible specification of the unobserved model component . following manski ( 1985 ) , we impose the restriction : @xmath34the conditional median independence assumption in ( [ m ] ) allows for heteroskedasticity of unknown form , and hence , is substantially weaker than the assumption imposed in ahn and manski ( 1993 ) . given ( [ m ] ) , the model ( [ e1 ] ) then satisfies @xmath35 we may consider sufficient conditions for in terms of the original structural errors @xmath36 and @xmath37 . recall that
. suppose that ( i ) the distribution of @xmath39 is the same as that of @xmath40 conditional on @xmath29 and @xmath31 , and ( ii ) the support of this common conditional distribution is @xmath41 .
this type of condition is called conditional exchangeability assumption .
then this implies that @xmath20 is symmetrically distributed around zero , thereby implying equation ( 2.3 ) . for further discussions regarding conditional exchangeability assumption ,
see fox ( 2007 ) in the context of multinomial discrete - choice models and arellano and honor ( 2001 ) for applications in panel data models , among others .
also , note that the conditional exchangeability assumption is a sufficient ( but not necessary ) condition for equation ( [ m ] ) .
let @xmath42 denote the space of preference parameters , and let @xmath43 , @xmath44 , denote the function space of difference of conditional expectations @xmath45 .
moreover , let @xmath46 and @xmath47 , @xmath44 , denote generic elements of @xmath42 and @xmath48 , respectively .
let @xmath49 and @xmath50 be the space of @xmath51 .
we refer to @xmath52 and @xmath30 as the true finite - dimensional and infinite - dimensional parameters .
suppose that data consist of random samples @xmath53 .
we estimate in the first stage the conditional expectations which are not observed .
let @xmath54 denote an estimate of the difference in conditional expectations . using the estimate @xmath55
, we estimate the preference parameters @xmath56 in the second stage by the method of maximum score estimation of manski ( 1975,1985 ) . for any @xmath57 and @xmath58 ,
define the sample score function@xmath59where @xmath60 is a predetermined weight function to avoid unduly influences from estimated @xmath61 at data points carrying low density .
the two - stage estimator of @xmath56 is now defined as @xmath62 we end this section by commenting on inherent features of the maximum score estimation approach .
the zero conditional median assumption does not require the existence of any error moments and allows heteroskedastic errors of an unknown form . however , the maximum score approach has its drawbacks , mainly due to its weak assumption .
first , in terms of prediction power , it can identify unknown parameters up to scale and also only identify whether the conditional probability of @xmath63 is above or below one half ; hence , the partial effects of covariates are not identified .
second , lack of smoothness in the objective function makes computation of the estimator difficult and lets the estimator converge in probability to the true parameter at a rate of @xmath64 .
let @xmath66 and @xmath67 , respectively , denote the distribution and density of @xmath68 . to simplify the analysis
, we consider fixed trimming such that @xmath69 , where @xmath70 is a predetermined , compact , and convex subset of the support of @xmath29 . for any real vector @xmath57 ,
let @xmath71 denote the euclidean norm of @xmath57 .
for any @xmath72-dimensional vector of functions @xmath73 , let @xmath74 where @xmath75 and @xmath76 denote the @xmath77th component of @xmath78 .
let @xmath79 be the subvector of @xmath31 excluding the first component , say @xmath80 of @xmath31 .
write @xmath81 and @xmath82 .
we assume the following regularity conditions .
[ assumption - consist ] assume that : * @xmath83 , where @xmath84 is a compact subspace of @xmath85and @xmath86 is an interior point of @xmath84 . *
\(a ) the support of the distribution of @xmath26 is not contained in any proper linear subspace of @xmath32 .
( b ) @xmath87 for almost every @xmath26 .
( c ) for almost every @xmath88 , the distribution of @xmath80 conditional on @xmath88 has everywhere positive density with respect to lebesgue measure .
* @xmath89 for almost every @xmath33 .
* there is a positive constant @xmath90 such that @xmath91 for all @xmath92 uniformly over @xmath93 .
* @xmath94 . because the scale of @xmath56 for the model characterized by ( [ e2 ] ) can not be identified ,
assumption c1 imposes scale normalization by requiring that the absolute value of the first coefficient is unity .
assumption c2 implies that @xmath66 is absolutely continuous and has density @xmath67 for each @xmath95 .
assumptions c1 - c3 are standard in the maximum score estimation literature ( see e.g. , manski ( 1985 ) , horowitz ( 1992 ) , and florios and skouras ( 2008 ) ) .
assumption c4 is a mild condition on the distribution of the index variable @xmath68 .
assumption c5 requires uniform consistency of the first stage estimation .
this assumption can be easily verified for standard nonparametric estimators such as series estimators ( newey ( 1997 , theorem 1 ) ) and the kernel regression estimator ( bierens ( 1983 , theorem 1 ) , bierens ( 1987 , theorem 2.3.1 ) and andrews ( 1995 , theorem 1 ) ) . given these regularity conditions , we have the following result .
[ thm - consist ] let assumption [ assumption - consist ] ( * c1 * - * c5 * ) hold .
then the two - stage estimator given by ( [ beta_hat ] ) converges to @xmath56 in probability as @xmath96 .
in addition to consistency , we also study rate of convergence of the estimator @xmath97 . let @xmath98 @xmath99 and @xmath100 .
let @xmath101 and @xmath102 denote the density function of @xmath103 conditional on @xmath88 .
let @xmath104 denote the partial derivative of @xmath105 with respect to @xmath80 .
define the following matrix@xmath106 .\ ] ] since the objective function of ( [ score ] ) is non - smooth , we require the nonparametric parameter of the estimation problem should possess certain degree of smoothness to facilitate derivation of the rate of convergence result .
in particular , we consider the following well known class of smooth functions ( see , e.g. , van der vaart and wellner ( 1996 , section 2.7.1 ) ) : for @xmath107 , let @xmath108 denote the class of functions @xmath109:@xmath110 with @xmath111 where for any @xmath112 dimensional vector of non - negative integers @xmath113,@xmath114where @xmath115 , denotes the greatest integer smaller than @xmath116 , and @xmath117 is the differential operator @xmath118given the norm @xmath119 , for any @xmath72-dimensional vector of functions @xmath73 , let @xmath120 where @xmath121 denote the @xmath77th component of @xmath78 .
note that @xmath122 is a stronger norm than @xmath123 used in condition c5 for the uniform consistency of the first stage estimator .
the regularity conditions imposed for the convergence rate result are stated as follows .
[ assumption - rate]assume that : * the support of @xmath79 is bounded .
* there is a positive constant @xmath124 such that ( i ) for every @xmath80 and for almost every @xmath125 , @xmath126and ( ii ) for non - negative integers @xmath127 and @xmath77 satisfying @xmath128 , @xmath129for every @xmath130 and @xmath80 and for almost every @xmath125 .
* all elements of the vector @xmath131 have finite third absolute moments . *
the matrix @xmath132 is positive definite .
* for each @xmath44 , @xmath133 for some @xmath134 and @xmath135 .
* @xmath136 where @xmath137 is a non - stochastic positive real sequence such that @xmath138 for each @xmath1 .
assumption c6 is standard in deriving asymptotic properties of manski s maximum score estimator ( see , e.g. kim and pollard ( 1990 ) , pp.213 - 216 ) .
assumption c7 requires some smoothness of the density @xmath139 and the distribution @xmath140 .
assumption c8 is mild .
since @xmath141 corresponds to the second order derivative of @xmath142 $ ] with respect to @xmath143 evaluated at true parameter values , assumption c9 is analogous to the classic condition of hessian matrix being non - singular in the m - estimation framework .
assumption c10 imposes smoothness for the nonparametric parameter @xmath51 and hence helps to control complexity of the space @xmath144 .
the requirement @xmath145 is in line with the literature of two - stage semiparametric estimation with non - smooth objective functions ( see , e.g. , chen et.al.(2003 , example 2 , pp.1601 - 1603 ) and ichimura and lee ( 2010 , section 4.1 , pp.258 - 259 ) ) .
assumption c11 requires that the first stage estimator should converge under the norm @xmath146 at a rate no slower than @xmath64 .
note that convergence of @xmath55 to @xmath147 in the norm @xmath146 also implies uniform convergence of derivatives of @xmath55 to those of @xmath147 .
for integer - valued @xmath148 , assumption c11 is fulfilled provided that for vector of non - negative integers @xmath149 that satisfies @xmath150 , @xmath151where @xmath152 denotes the estimate of @xmath153 for @xmath154 .
the condition ( [ der ] ) can also be verified for series estimators ( newey ( 1997 , theorem 1 ) ) and the kernel regression estimator ( andrews ( 1995 , theorem 1 ) ) .
[ thm - rate]in addition to assumption [ assumption - consist ] ( * c1 * - * c5 * ) , let assumption [ assumption - rate ] ( * c6 * - * c11 * ) also hold
. then @xmath155 .
if @xmath147 were priorly known to the researcher , the preference parameters @xmath156 could be estimated by the single stage maximum score estimator @xmath157 , defined as @xmath158kim and pollard(1990 ) showed that @xmath159 converges to @xmath56 at cube root rate and established its asymptotic distribution . in the case of unknown @xmath147 , theorem [ thm - rate ] implies that the two - stage estimator @xmath97 retains the same convergence rate as the infeasible estimator @xmath159 . indeed
if condition c11 is strengthened for faster convergence of first stage estimates , we can establish the oracle property that @xmath160 and @xmath161 have the same limiting distribution .
therefore , the inference on @xmath56 can be carried out by subsampling ( delgado et al.(2001 ) ) since the standard bootstrap can not be used to estimate the distribution of the maximum score estimator consistently ( abrevaya and huang ( 2005 ) ) .
we now state the asymptotic distributional equivalence result in the next theorem .
[ thm - distribution]suppose all assumptions stated in theorem thm - rate hold with the additional restriction that the sequence @xmath162 stated in * c11 * further satisfies @xmath163 .
then @xmath160 is asymptotically equivalent in distribution to @xmath164 .
we employ the following data generating process ( dgp ) in simulation study of the two - stage maximum score estimator:@xmath165where @xmath166 , @xmath167 , @xmath168 and @xmath169 with @xmath170
. the scalar random variable @xmath10 is generated according to @xmath171where @xmath172 are independent of @xmath173 and are jointly normally distributed with @xmath174 , @xmath175 , and @xmath176 . given ( [ y ] ) , @xmath177we consider the following two types of the @xmath178 function:@xmath179the true parameter values are specified as follows : @xmath180 , @xmath181 , @xmath182 , @xmath183 , @xmath184 , @xmath185 , @xmath186 , and @xmath187 . we compare infeasible single - stage estimator using @xmath188 as regressors and also the feasible two - stage estimator using @xmath189 as regressors .
we consider both parametric and nonparametric first stage estimators .
for the former , we estimate @xmath21 by running ols of @xmath10 on @xmath29 with an intercept term using @xmath190 subsamples . for the latter , we implement nadaraya - watson kernel regression estimators .
the nonparametric estimators of @xmath21 , @xmath4 are constructed as @xmath191where @xmath192 is the estimated standard deviation of @xmath193 conditional on @xmath194 , @xmath195 is a univariate kernel function and @xmath196 is a deterministic bandwidth sequence .
we use two types of kernel and bandwidth configurations . for the first type , we use the second - order gaussian kernel and set @xmath196 to be @xmath197 for various values of the bandwidth scale @xmath198 . for the second type ,
we use the following 8th order kernel function ( see , e.g. , bierens ( 1987 , p. 112 ) and andrews ( 1995 , p.567 ) ) : @xmath199 , \label{k}\]]where the constants @xmath200 @xmath201 satisfy @xmath202we specify @xmath203 and then solve @xmath204 as solution of the system of linear equations ( [ ab ] ) . associated with this kernel , the bandwidth @xmath205 is set to be @xmath206 for various values of the scale @xmath198 . for the kernel function is less crucial since its effect on asymptotic variance of the conditional mean estimator can be captured via the bandwidth scale @xmath198 . ] by theorem 1(b ) of andrews ( 1995 ) , kernel regression estimator of @xmath30 based on the second type configuration has convergence property required in ( [ der ] ) with @xmath207 and @xmath208 , thus fulfilling regularity conditions c5 and c11 of section [ main results ] . the first stage estimation with the second - order kernels satisfies condition c5 but may not satisfy c11 ; however , we experiment with the second - order kernels as well since kernel estimates with the second - order kernels often outperform those with the higher - order kernels in small samples . to implement the second - stage estimator using nonparametric first stage estimators , we trim the data by setting @xmath209 where @xmath210 is the weight introduced in ( [ score ] ) . the estimates of @xmath211 and @xmath212 are obtained using grid search method . we report simulation results of @xmath213 for the parameter capturing the agent s uncertainty .
let @xmath214 , @xmath215 , @xmath216 and @xmath217 respectively denote the estimators @xmath218 that are constructed based on the infeasible single - stage , two - stage ( ols first stage ) and two - stage ( kernel regression first stage implemented with the 2nd and 8th order kernels ) maximum score estimators .
we compute bias , median , root mean squared error ( rmse ) , mean absolute deviation ( mean ad ) and median absolute deviation ( median ad ) of these estimators based on 1000 simulation repetitions for sample size @xmath219 .
tables 1 - 6 present simulation results for the four types of estimators of @xmath220 under linear and nonlinear designs of the @xmath30 function . tables 7 and 8 graph the simulated empirical distribution functions ( edf ) for @xmath221 , @xmath222 , @xmath223 and @xmath224 .
as expected , for linear setup of @xmath147 the estimator @xmath215 enjoys the best overall finite - sample performance among all two - stage estimators .
however , this estimator also incurs huge bias when agent s conditional expectation is nonlinear . for the estimators @xmath225 and @xmath226 , the function @xmath147 is nonparametrically estimated at the first stage .
hence regardless of nonlinearity of @xmath147 , we see that the simulated bias , rmse , mean ad and median ad of these estimators generally decrease as sample size grows .
we note that the edf curves of tables 7 and 8 for the ( kernel first - stage ) two - stage estimators broadly match shapes of those for the infeasible estimators .
interestingly , finite sample behavior of the estimator @xmath227 fits that of @xmath228 better than its counterpart implemented with the 8th order kernel .
use of higher order kernels allows for verification of convergence of @xmath229 to @xmath147 in the strong norm @xmath146 .
however , as well known in the literature , the estimates with the higher - order kernels seem to perform poorly in simulations relative to those with the second - order kernels .
the superb performance of @xmath216 suggests that the asymptotic distributional equivalence result in theorem 3 may not give us sharp asymptotics and there is scope to develop further asymptotic theory .
this is an interesting future research topic .
our paper has been motivated by estimation problem in the binary choice model under uncertainty . however , the resulting estimator has wider applicability than just this model . to further motivate our two - step estimation procedure
, this section gives a couple of additional econometric models for which unknown parameters can be estimated by maximum score with nonparametrically generated regressors . we first consider maximum score estimation of an incomplete information games .
aradillas - lopez ( 2012 ) developed a two - step procedure for estimation of incomplete information games with nash equilibrium behavior .
equation ( 2 ) of aradillas - lopez ( 2012 , p.123 ) gives a description of players behavior in a @xmath230 game : @xmath231-\zeta _ { 1}\geq 0\ } , \\
y_{2 } & = 1\{x_{2}^{\prime } \beta _ { 2}+\delta _ { 2}\text{pr}[y_{1}=1|x]-\zeta _ { 2}\geq
0\},\end{aligned}\]]where @xmath232 is the binary action for player @xmath233 , @xmath234 and @xmath235 are observable and unobservable payoff covariates , @xmath236 , and @xmath237 are unknown parameters . aradillas - lopez ( 2012 , assumption a0 , p.122 ) assumed that players behavior corresponds to a bayesian - nash equilibrium with a degenerate selection mechanism .
he further assumed that @xmath238 and @xmath239 are independent of each other , independent of @xmath240 , and of the selection mechanism .
we can make the same assumptions as in aradillas - lopez ( 2012 ) , with one exception .
as in the previous section , we consider @xmath241 almost surely , instead of assuming the full independence between @xmath242 and @xmath240 , where @xmath233 . allowing for dependence between @xmath243 and @xmath240 might be important in applications when we suspect possible interactions between observed covariates and unobserved components that affect players payoffs . then for each @xmath233 , we can estimate @xmath244 by running maximum score regression of @xmath245 on @xmath234 and @xmath246 $ ] with the nonparametric first stage estimation of @xmath247 .
therefore , methodology of the present paper can be applied to extension of aradillas - lopez ( 2012 ) s context allowing unobserved payoffs to exhibit unknown form of heteroskedasticity .
our second application , which is based on fox ( 2007 ) , is maximum score estimation of multinomial discrete - choice models using a subset of choices under endogeneity . fox ( 2007 )
proposed pairwise maximum score estimation of multinomial discrete - choice models using a subset of choices . for simplicity , assume that a researcher has data on only two choice , say @xmath248 and @xmath249 , among @xmath250 alternatives , and also assume that there exists an endogenous covariate . fox ( 2007 , p.1013 ) solved the endogeneity problem by including , instead of the endogenous covariate , fitted values from the ols regression of the endogenous covariate , say price , on a vector of instruments .
we can extend fox ( 2007 ) to allow for nonparametric fitted values .
then this extension again can be accommodated in the framework of maximum score estimation with nonparametrically generated regressors .
this paper has developed maximum score estimation of preference parameters in the binary choice model under uncertainty in which the decision rule is affected by conditional expectations .
the estimation procedure is implemented in two stages : we estimate conditional expectations nonparametrically in the first stage and obtain the maximum score estimate of the preference parameters in the second stage using choice data and the first stage estimates .
the paper has shown consistency and convergence rate of the two - stage maximum score estimator . moreover
, we also establish the oracle property in terms of asymptotic equivalence in distribution of the two - stage estimator and its corresponding infeasible single - stage version .
these results are of independent interest for maximum score estimation with nonparametrically generated regressors .
it would be an alternative approach to develop the second stage estimator using horowitz ( 1992 ) s smoothed maximum score estimator or using a laplace estimator proposed in jun , pinkse , and wan ( 2013 ) .
these alternative methods would produce faster convergence rates but require extra tuning parameters .
alternatively , we might build the second stage estimator based on lewbel ( 2000 ) , who introduced the idea of a special regressor satisfying certain conditional independence restriction .
these are interesting future research topics .
recall that @xmath251 and @xmath252 is the sample score function defined by ( [ score ] ) .
we first state and prove a preliminary lemma that will be invoked in proving theorem [ thm - consist ] of the paper .
[ l1 ] under assumptions c1 , c4 and c5 , @xmath253 note that @xmath254by assumption c1 , @xmath255 for some finite positive constant @xmath256 .
therefore , the right - hand side of the inequality ( [ s ] ) is bounded above by@xmath257where @xmath258 denotes the empirical probability .
note that the term ( [ b ] ) is further bounded above by@xmath259let @xmath260 denote the event @xmath261 for some @xmath262 .
then given @xmath263,@xmath264 + p(e^c_{\eta } ) .\end{aligned}\]]by assumption c5 , @xmath265 as @xmath266 .
hence , to show ( [ s1 ] ) , it remains to establish that as @xmath267,@xmath268 \longrightarrow 0 .
\label{delta}\]]note that by assumption c4 , @xmath269 .
therefore , we have that @xmath270 \notag \\ & \leq & p \left [ \sup\nolimits_{b\in \theta } \left\vert p_{n}\left ( b_{2}\eta \geq \left\vert w^{\prime } b\right\vert \right ) -p\left ( b_{2}\eta \geq \left\vert w^{\prime } b\right\vert \right ) \right\vert > \epsilon -2lb_{2}\eta \right ] , \label{s2}\end{aligned}\]]where @xmath271 is taken to be sufficiently small such that @xmath272 for the given @xmath273 . by lemma 9.6 , 9.7 ( ii ) and 9.12 ( i ) of kosorok ( 2008 ) , the family of sets @xmath274 for @xmath275 forms a vapnik - ervonenkis class .
therefore , by glivenko - cantelli theorem ( see , e.g. theorem 2.4.3 of van der vaart and wellner ( 1996 ) ) , the right - hand side of ( [ s2 ] ) tends to zero as @xmath96 .
hence , the convergence result in ( [ delta ] ) holds and lemma [ l1 ] thus follows .
we now prove theorem [ thm - consist ] for consistency of @xmath97 . for any @xmath276 ,
define @xmath277 .\]]given assumptions c1 - c3 and by manski ( 1985 , lemma 3 , p.321 ) , @xmath156 uniquely satisfies @xmath278 .
we now look at the difference@xmath279where by lemma [ l1 ] , the first term of the right - hand side of ( [ d ] ) converges to zero in probability uniformly over @xmath275 , whilst by manski ( 1985 , lemma 4 , p.321 ) , the second term converges to zero almost surely uniformly over @xmath275 .
therefore , we have that @xmath280by lemma 5 of manski ( 1985 , p.322 ) , @xmath281 is continuous in @xmath57 . given these results , theorem 1 thus follows by application of the consistency theorem in newey and mcfadden ( 1994 , theorem 2.1 ) .
we first present and prove a general lemma establishing the rates of convergence of a general two - stage m - estimator under high level assumptions . in next section ,
we prove theorem [ assumption - rate ] by verifying these assumptions for the particular estimator given by ( [ beta_hat ] ) under the regularity conditions of c1 - c11 . to present a general result , let @xmath282 be measurable functions indexed by parameters @xmath283 .
let @xmath42 and @xmath284 be the space of parameters @xmath285 and @xmath78 , respectively .
let @xmath286 denote the true parameter value .
we assume @xmath287 .
let @xmath288 be the empirical criterion of the m - estimation problem where @xmath289 are i.i.d .
random vectors . suppressing the individual index ,
let @xmath290 $ ] be the population criterion . for a given first stage estimate @xmath291 ,
let the estimator @xmath292 be constructed as @xmath293 let @xmath294 and @xmath295 be non - negative functions measuring discrepancies between @xmath285 and @xmath296 , and @xmath78 and @xmath297 , respectively .
note that @xmath298 and @xmath299 are usually related to but not necessarily the same as the metrics specified for the spaces @xmath42 and @xmath284 . given a non - stochastic positive real sequence @xmath137 ,
define @xmath300 . to simplify the presentation
, we use the notation @xmath301 to denote being bounded above up to a universal constant .
define the recentered criterion @xmath302the following lemma modifies the rate of convergence results developed by van der vaart ( 1998 , theorem 5.55 ) and provides sufficient conditions ensuring that @xmath292 retains the same convergence rate as it would have if @xmath297 were known .
[ rate of convergence lemma ] for any fixed and sufficiently large @xmath303 , assume that for all sufficiently large @xmath1 , @xmath304and there is a sequence of non - stochastic functions @xmath305 such that for all sufficiently small @xmath306 and for every @xmath307 satisfying @xmath308 , @xmath309and @xmath310 \lesssim \frac{\phi _ { n}(\delta ) } { \sqrt{n } } , \label{maximal inequality}\]]where @xmath311 is a sequence of functions defined on @xmath312 and satisfies that @xmath313 is decreasing for some @xmath314 .
suppose @xmath315 , @xmath316 and there is a non - stochastic positive real sequence @xmath317 which tends to zero as @xmath96 and satisfies that @xmath318 and @xmath319 for every @xmath1
. then @xmath320 . based on the peeling technique of van der vaart ( 1998 , theorem 5.55 ) , for each natural number @xmath1
, integer @xmath77 and positive real @xmath321 , construct the set @xmath322then we have that for any @xmath263,@xmath323 \geq 0\right ) \label{b5_2}\end{aligned}\]]where the last inequality follows from the definition of @xmath292 given by ( [ thetahat ] ) .
since @xmath324 , the term @xmath325 tends to zero as @xmath96 .
hence the remaining part of the proof is to bound the terms in the sum ( [ b5_2 ] ) .
let @xmath1 be large enough such that holds and choose @xmath326 to be small enough such that assumptions ( [ quadratic expansion ] ) , ( [ inequality 1 ] ) and ( [ maximal inequality ] ) hold for every @xmath327 .
note that for every sufficiently large @xmath321 , if @xmath328 , then @xmath329 so that by ( [ quadratic
expansion]),@xmath330and thus @xmath331 -\delta _ { n}^{2}2^{2j}.\]]therefore , by markov inequality , each term in the sum ( [ b5_2 ] ) can be bounded above by @xmath332 .
\label{b3}\]]by ( [ inequality 2 ] ) , ( [ inequality 1 ] ) , ( [ maximal inequality ] ) and applying triangular inequality , the term ( [ b3 ] ) is bounded above by@xmath333 .
\label{b4}\]]by the monotonicity property of the mapping @xmath334 , we have that @xmath335 . furthermore , since @xmath336 , the first term in the bracket of ( b4 )
can thus be bounded by @xmath337 . given that @xmath318 , the term ( [ b4 ] ) can be further bounded above by @xmath338 .
using this fact and the condition @xmath314 , it follows that the sum ( [ b5_2 ] ) tends to zero as @xmath339 . since @xmath340
, @xmath341 can be made arbitrarily close to @xmath248 by choosing a sufficiently large value of @xmath342 for every sufficiently large @xmath1 .
therefore , lemma [ rate of convergence lemma ] follows by putting together all these results and noting that @xmath343 .
to establish the convergence rate of @xmath97 , we apply lemma rate of convergence lemma by setting @xmath345 , @xmath346 , @xmath83 , @xmath347 , @xmath348 and @xmath349assumptions ( [ inequality 2 ] ) , ( [ quadratic expansion ] ) , ( inequality 1 ) and ( [ maximal inequality ] ) of lemma [ rate of convergence lemma ] are non - trivial and will be verified using primitive conditions c1 - c11 of the model . assumption ( [ quadratic expansion ] ) is concerned with the quadratic expansion of @xmath350 around @xmath351 by which we obtain the functional form of @xmath352 . recall that @xmath251 , @xmath353 , @xmath354 , @xmath355 , @xmath356 , @xmath357 and @xmath358 .
the following lemma will be used to establish expansion of the population criterion @xmath350 .
[ lemma for sign of partial derivative]under conditions c3 and c7 , the sign of @xmath359 is the same as that of @xmath360 for almost every @xmath361 .
note that the model ( [ eq ] ) implies that@xmath362thus , by c7(ii ) , @xmath105 is differentiable with respect to @xmath80 and @xmath363consider the mapping @xmath364 . by c3 , @xmath365 for almost every @xmath366 .
therefore , lemma [ lemma for sign of partial derivative ] follows from this fact and the monotonicity of @xmath367 in the argument @xmath130 . by assumption c1 ,
the space of the coefficient @xmath368 is @xmath369 and thus @xmath370 when @xmath371 for @xmath372 small enough .
let @xmath373 @xmath374 and @xmath375 .
\label{s1}\]]we now derive the quadratic expansion of @xmath376 around @xmath377 .
[ quadratic expansion of s(b , r)]for sufficiently small @xmath378 and @xmath379 and under conditions c3 , c7 , c8 and c9 , we have that @xmath380and there are constants @xmath381 and @xmath382 such that @xmath383for some function @xmath384 that satisfies @xmath385 we prove lemma [ quadratic expansion of s(b , r ) ] explicitly for the case @xmath386 . proof for the case @xmath387 can be done by similar arguments .
suppose now @xmath388 .
then@xmath389 \right ) .\end{aligned}\]]let @xmath390the first - order and second - order derivatives of @xmath391 are derived as follows:@xmath392 \right\ } \\ & & + e\left ( 2\tau \left [ \left ( 2p(-\lambda ( t),\widetilde{z},x)-1\right ) % \right ] g_{1}(-\lambda ( t)|\widetilde{z},x)(\gamma ( x)-g(x))^{\prime } ( b_{2}-\beta _ { 2})\right ) .\end{aligned}\]]then the second order expansion of @xmath393 takes the form @xmath394where by c7 and c8 , the remainder term has the stated order uniformly over @xmath395 and @xmath51 . given assumption c3 , it follows that @xmath396 for almost every @xmath397 .
let @xmath398then we have that @xmath399where @xmath400under condition c9 , @xmath401 is positive definite , so that @xmath402 for some positive real constant @xmath403 . by lemma [ lemma for sign of partial derivative ] ,
@xmath404 and thus @xmath405 . by cauchy - schwarz inequality , @xmath406 , where @xmath407 , and the function @xmath408 satisfies
that@xmath409hence lemma [ quadratic expansion of s(b , r ) ] follows by noting that when @xmath410 and @xmath411 are sufficiently small , @xmath412and @xmath413 we now verify assumption ( [ maximal inequality ] ) of lemma [ rate of convergence lemma ] .
note that for @xmath372 sufficiently small , assumption c1 implies that @xmath370 when @xmath414 .
therefore we can focus on analyzing ( [ maximal inequality ] ) for the case of @xmath370 and @xmath415 . for any @xmath416 ,
consider the following recentered function @xmath417 \label{recentered m(b , r)}\]]and the class of functions@xmath418let @xmath419 denote the @xmath420 norm such that @xmath421 ^{1/r}$ ] for any measurable function @xmath422 .
for any @xmath263 , let @xmath423}(\epsilon , \digamma , l_{r}(p))$ ] denote the @xmath420 - bracketing number for a given function space @xmath424 .
namely , @xmath423}(\epsilon , \digamma , l_{r}(p))$ ] is the minimum number of @xmath420 - brackets of length @xmath425 required to cover @xmath424 ( see e.g. , van der vaart ( 1998 , p.270 ) ) .
the logarithm of bracketing number for @xmath426 is referred to as the bracketing entropy for @xmath426 .
assumption ( [ maximal inequality ] ) is a stochastic equicontinuity condition concerning the complexity of the function space @xmath427 in terms of its envelope function and bracketing entropy .
let @xmath428 denote an envelope for @xmath427 such that @xmath429 @xmath430 for all @xmath431 and for all @xmath432 .
the next lemma derives the envelope function @xmath428 .
[ envelope]let @xmath372 and @xmath20 be sufficiently small .
then under conditions c1 , c4 , c6 and c10 , for some real constants @xmath433 and @xmath434 , we can take @xmath435and furthermore , @xmath436 note that@xmath437under condition c6 , there is a positive real constant @xmath438 such that @xmath439 with probability 1 .
hence if @xmath440 and @xmath441 , then we have that@xmath442 \geq 0 \\ & \longrightarrow & \delta \left [ \left\vert \widetilde{z}\right\vert _ { e}+\left\vert \gamma \right\vert _ { \infty } \right ] \geq -\left [ z^{\prime } \beta { } _ { 1}+\gamma ( x){}^{\prime } \beta _ { 2}\right ] \text { and } 0\geq w^{\prime } \beta { } + ( \gamma ( x)-g(x)){}^{\prime } \beta _ { 2 } \\ & \longrightarrow & w^{\prime } \beta + ( \gamma ( x)-g(x{}))^{\prime } \beta _ { 2}\geq -\delta \left [ \left\vert \widetilde{z}\right\vert _ { e}+\varepsilon + \left\vert g\right\vert _ { \infty } \right ] \text { and } \varepsilon \left\vert \beta _ { 2}\right\vert _ { e}\geq w^{\prime } \beta { } \\ & \longrightarrow & \delta \left [ b+\varepsilon + \left\vert g\right\vert _ { \infty } \right ] + \varepsilon \left\vert \beta _ { 2}\right\vert _ { e}\geq w^{\prime } \beta \geq -\delta \left [ b+\varepsilon + \left\vert g\right\vert
_ { \infty } \right ] -\varepsilon \left\vert \beta
_ { 2}\right\vert _ { e}\end{aligned}\]]based on similar arguments , it also follows that @xmath443 + \varepsilon \left\vert \beta _ { 2}\right\vert _ { e}\geq w^{\prime } \beta \geq -\delta \left [ b+\varepsilon + \left\vert g\right\vert _ { \infty } \right ] -\varepsilon \left\vert \beta _
{ 2}\right\vert _ { e}\end{aligned}\]]therefore , lemma [ envelope ] follows by noting that for @xmath20 sufficiently small ( e.g. , @xmath444 ) , we can take @xmath435where @xmath445 . by c1 and c10 , @xmath446 and hence by c4 , @xmath447 with @xmath448 where @xmath449 is the positive constant stated in condition c4 .
the following lemma establishes the bound for the bracketing entropy for @xmath450 .
[ entropy bound]given conditions c1 , c4 , c6 , c7 , c8 and c10 , we have that for sufficiently small @xmath372 and @xmath20 , @xmath451}(\epsilon , \digamma _ { \delta , \varepsilon } ,
l_{2}(p))\lesssim ( \max \{\delta , \varepsilon \})^{q/\alpha } \epsilon ^{-2q/\alpha } .\ ] ] for @xmath44 , let @xmath452 and @xmath453 be classes of functions defined as@xmath454assumption c10 implies that both @xmath452 and @xmath455 are @xmath456 .
by corollary 2.7.2 of van der vaart and wellner ( 1996 , p.157 ) , we have that for @xmath44 , @xmath451}(\epsilon ^{2},\widetilde{\lambda } _ { j}(\varepsilon ) , l_{1}(p))\lesssim \epsilon ^{-2q/\alpha } \text { and } \log n_{[]}(\epsilon ^{2},\widetilde{\lambda } _ { j}\mathbf{b}_{j}(\delta , \varepsilon ) , l_{1}(p))\lesssim \epsilon ^{-2q/\alpha } . \label{entropy bound c(alpha,1)}\ ] ] note that for @xmath457 , @xmath458 defined by ( [ recentered m(b , r ) ] ) can be rewritten as @xmath459 + \tau ( 1-d)\left [ 1\{h(s;% \widetilde{b})\leq 0\}-1\{h(s;\widetilde{\beta } ) \leq 0\}\right]\]]where @xmath460consider the following spaces:@xmath461let @xmath462}(\epsilon , \theta _ { i},l_{1}(p))$ ] for @xmath463 and @xmath464}(\epsilon , \theta _ { k , j},l_{1}(p))$ ] for @xmath465 .
let @xmath466.@xmath467 since @xmath468 is a pointwise lipschitz class of functions with envelope @xmath469 . by condition c8
, @xmath470 is finite . thus applying theorem 2.7.11 of van der vaart and wellner ( 1996 , p.164 )
, we have that@xmath471note that for any norm @xmath472 , any fixed real valued @xmath198 , any class of functions @xmath426 , it is straightforward to verify that @xmath473}(\epsilon , c\digamma , \left\vert \cdot \right\vert ) = 1\text { for } c=0 \notag \\ & & n_{[]}(\epsilon , c\digamma , \left\vert \cdot \right\vert ) \leq n_{[]}(\epsilon /\left\vert
c\right\vert , \digamma , \left\vert \cdot \right\vert ) \text { for } c\neq 0 \notag\end{aligned}\]]where @xmath474 .
using this fact , we have that @xmath475}(\epsilon ^{2}/(\varepsilon \delta ) , \widetilde{\lambda } _ { j}\mathbf{b}_{j}(\delta , \varepsilon ) , l_{1}(p))$ ] and @xmath476 for @xmath477 and @xmath478}(\epsilon ^{2}/(\varepsilon \left\vert \beta _ { 2,j}\right\vert ) , \widetilde{\lambda } _ { j}(\varepsilon ) , l_{1}(p))$ ] for @xmath479 .
hence for sufficiently small @xmath372 and @xmath20 ( e.g. , @xmath480 and @xmath444 ) and by ( entropy bound c(alpha,1 ) ) , it follows that @xmath481}(\epsilon ^{2}\psi ^{-2},\widetilde{% \lambda } _ { j}\mathbf{b}_{j}(\delta , \varepsilon ) , l_{1}(p))\lesssim \psi ^{2q/\alpha } \epsilon ^{-2q/\alpha } .\]]using similar arguments , we can also deduce that @xmath482 . by preservation of bracketing metric entropy ( see , e.g. , lemma 9.25 of kosorok ( 2008 , p.169 ) )
, we have that for @xmath483 , @xmath484and @xmath485 .
therefore by the bounds derived above , it follows that @xmath486 , @xmath487 and also @xmath488 .
now let @xmath489}(\epsilon ^{2},\theta _ { 3},l_{1}(p))}^{l}\leq f_{n_{[]}(\epsilon ^{2},\theta _ { 3},l_{1}(p))}^{u}$ ] and @xmath490}(\epsilon ^{2},\theta _ { 4},l_{1}(p))}^{l}\leq g_{n_{[]}(\epsilon ^{2},\theta _ { 4},l_{1}(p))}^{u}$ ] be the @xmath491-brackets with bracket length defined by @xmath492 for the spaces @xmath493 and @xmath494 , respectively . for @xmath495}(\epsilon ^{2},\theta
_ { 3},l_{1}(p))$ ] and @xmath496}(\epsilon ^{2},\theta _ { 4},l_{1}(p))$ ] , define@xmath497 \\ & & + \tau ( 1-d)\left [ 1\{w^{\prime } \beta + g_{j}^{u}(z , x)\leq 0\}-1\{w^{\prime } \beta + f_{k}^{l}(z , x){}\leq 0\}\right ] , \\ m_{jk}^{u}(\tau , d , z , x ) & \equiv & \tau d\left [ 1\{w^{\prime } \beta + g_{j}^{u}(z , x)>0\}-1\{w^{\prime } \beta + f_{k}^{l}(z , x){}>0\}\right ] \\ & & + \tau ( 1-d)\left [ 1\{w^{\prime } \beta + g_{j}^{l}(z , x)\leq 0\}-1\{w^{\prime } \beta + f_{k}^{u}(z , x){}\leq 0\}\right ] .\end{aligned}\]]note
that@xmath498 \text{.}\]]thus@xmath499by condition c1 and given @xmath88 , the mapping @xmath500 is one - to - one .
hence by condition c7 , the density of @xmath501 conditional on @xmath88 is bounded and by ( m_bound ) , it then follows that @xmath502 .
moreover for each @xmath503 , there is a bracket @xmath504 $ ] in which it lies .
therefore , @xmath451}(\epsilon , \digamma _
{ \delta , \varepsilon _ { n}},l_{2}(p))\lesssim n_{3}(\epsilon ^{2})+n_{4}(\epsilon ^{2})\lesssim \psi ^{2q/\alpha } \epsilon ^{-2q/\alpha } \text{.}\ ] ] replacing @xmath283 and @xmath505 with @xmath506 and @xmath507 , respectively in the definition of @xmath508 given by ( recentered criterion ) , we now verify assumption ( [ maximal inequality ] ) in the next lemma .
[ verify maximal inequality]for sufficiently small @xmath372 and @xmath509 , under conditions c1 , c4 , c6 , c7 , c8 and c10 , @xmath510 \lesssim \frac{\sqrt{\max \{\delta , \varepsilon \}}}{\sqrt{n}}.\ ] ] let @xmath466 . by lemmas envelope and [ entropy bound ]
, we have that @xmath511}(\epsilon , \digamma _ { \delta , \varepsilon } , l_{2}(p))}d\epsilon \lesssim \psi ^{q/\alpha } \int\nolimits_{0}^{a_{2}\psi } \epsilon ^{-q/\alpha } d\epsilon \lesssim \psi\]]where the last inequality follows since @xmath134 .
lemma [ verify maximal inequality ] hence follows by applying corollary 19.35 of van der vaart ( 1998 , p.288 ) .
we now prove theorem [ thm - rate ] .
we take @xmath512 , @xmath513 and @xmath514 in the application of lemma [ rate of convergence lemma ] , where @xmath403 and @xmath515 are real constants stated in lemma [ quadratic expansion of s(b , r ) ] . since @xmath381 ,
the norm by the metric @xmath516 is equivalent to the euclidean norm and thus by theorem [ thm - consist ] , @xmath517 .
moreover since @xmath382 , assumption c11 implies that @xmath518 .
given assumption c1 , for sufficiently small @xmath372 , we have that @xmath519 when @xmath520 .
hence for sufficiently small @xmath372 and @xmath137 , by lemma [ quadratic expansion of s(b , r ) ] and noting that @xmath521 is stronger than @xmath522 , assumptions ( [ inequality 2 ] ) , ( [ quadratic expansion ] ) and ( [ inequality 1 ] ) hold .
by lemma [ verify maximal inequality ] and by taking @xmath342 sufficiently large in the definition of @xmath523 of lemma [ rate of convergence lemma ] , assumption ( [ maximal inequality ] ) also holds with @xmath524 .
clearly , @xmath525 is decreasing for some @xmath314 . by assumption
c11 , @xmath526 and thus @xmath527 for every @xmath1 .
therefore , all conditions stated in lemma [ rate of convergence lemma ] are fulfilled and the result of theorem [ thm - rate ] hence follows .
for @xmath303 , define the sets @xmath529where @xmath137 is the sequence stated in the assumptions of theorem [ thm - distribution ] . for each @xmath530 , define the following recentered empirical and population criterion functions @xmath531clearly , @xmath97 and @xmath97 @xmath532 , defined by ( beta_hat ) and ( [ beta_hat_g ] ) , are still maximizers of the objective functions @xmath533 and @xmath534 , respectively
. decompose @xmath535 as follows.@xmath536 + \left [ \overline{s}(b,% \widehat{g})-\overline{s}(b , g)\right ] \label{decomposition}\]]where@xmath537we shall need the following results . for @xmath306 and @xmath538 ,
consider the local neighborhoods @xmath539 and @xmath540 defined as@xmath541recall that @xmath542 , @xmath543 , @xmath544 , @xmath545 and @xmath546 . note that for @xmath372 sufficiently small , assumption c1 implies that @xmath370 when @xmath547 .
therefore we may assume that @xmath548 @xmath370 and @xmath549 where @xmath550 .
for any @xmath551 , consider the following function@xmath552let@xmath553 .\]]define the class of functions@xmath554let @xmath555 denote an envelope for @xmath556 such that @xmath557 @xmath558 for all @xmath431 and for all @xmath559 .
[ envelope copy(1)]let @xmath372 and @xmath20 be sufficiently small .
given conditions c1 , c4 , c6 and c10 , for some positive real constants @xmath403 and @xmath515 , we can take @xmath560and furthermore , @xmath561 note that@xmath562 \text { or } \left [ \overline{m}_{% \widetilde{b},\gamma } ( s)=-1\text { and } \overline{m}_{\widetilde{b},g}(s)=1% \right ] \right\ } .\]]given c1 , c6 and c10 , there is positive real constant @xmath438 such that @xmath563 with probability 1 .
hence if @xmath564 , we have that @xmath565on the other hand,@xmath566therefore , the condition @xmath567 and @xmath568 implies @xmath569,@xmath570 .
based on similar arguments , we can verify that the condition @xmath571 and @xmath572 also implies @xmath573,@xmath574 .
therefore , lemma [ envelope copy(1 ) ] follows by taking @xmath575 and noting that given c4 , inequality ( m_delta_epsilon ) holds for @xmath576 .
[ entropy bound copy(1)]given conditions c1 , c4 , c6 , c7 , c8 and c10 , we have that for sufficiently small @xmath372 and @xmath20 , @xmath577 \lesssim n^{-1/2}(\max \{\delta , \varepsilon \})^{\frac{q}{2\alpha } } ( \min \{\delta , \varepsilon \})^{\frac{\alpha -q}{% 2\alpha } } .\ ] ] define the following two classes of functions @xmath578using lemma 9.25 of kosorok ( 2008 , p.169 ) ) , we have that @xmath451}(\epsilon , \digamma _ { \delta , \varepsilon } , l_{2}(p))\leq \log n_{[]}(\epsilon /2,a_{\delta , \varepsilon } , l_{2}(p))+\log n_{[]}(\epsilon /2,b_{\delta , \varepsilon } , l_{2}(p)).\]]let @xmath466 . by lemma entropy bound , we have that for sufficiently small @xmath372 and @xmath509 , @xmath451}(\epsilon , a_{\delta , \varepsilon } , l_{2}(p))\lesssim \psi ^{2q/\alpha } \epsilon ^{-2q/\alpha } .\]]furthermore by simplifying proof of lemma [ entropy bound ] , it is straightforward to verify that @xmath451}(\epsilon , b_{\delta , \varepsilon } , l_{2}(p))\lesssim \psi ^{2q/\alpha } \epsilon ^{-2q/\alpha } \]]and thus @xmath451}(\epsilon , \digamma _ { \delta , \varepsilon } , l_{2}(p))\lesssim \psi ^{2q/\alpha } \epsilon ^{-2q/\alpha } .
\label{entropy}\]]using inequality ( [ entropy ] ) and lemmas [ envelope copy(1 ) ] , we have that @xmath511}(\epsilon , \digamma _ { \delta , \varepsilon } , l_{2}(p))}d\epsilon \lesssim \psi ^{\frac{q}{\alpha } } \int% \nolimits_{0}^{c_{2}\sqrt{\min \{\delta , \varepsilon \}}}\epsilon ^{-\frac{q% } { \alpha } } d\epsilon \lesssim \psi ^{\frac{q}{\alpha } } ( \min \{\delta , \varepsilon \})^{\frac{\alpha -q}{2\alpha } } \]]where the last inequality follows from the assumption @xmath134 .
lemma entropy bound copy(1 ) hence follows by applying corollary 19.35 of van der vaart ( 1998 , p.288 ) . by kim and pollard(1990 ) , @xmath579 .
hence , by condition c11 and theorem thm - rate , for sufficiently large @xmath303 , probability of the event that @xmath580 @xmath581 and @xmath582 can be made arbitrarily close to @xmath248 .
thus to show the theorem , it suffices to establish that for any fixed sufficiently large @xmath303 , @xmath583given ( [ rate of difference of sn ] ) , we have that @xmath584where the first and third inequalities follow from ( [ rate of difference of sn ] ) and the second inequality follows from the definition of @xmath585 .
therefore by theorem 1.1 of kim and pollard(1990 ) , @xmath586 and @xmath164 are asymptotically equivalent in distribution .
we now verify equation ( [ rate of difference of sn ] ) . given the decomposition ( [ decomposition ] ) , it suffices to show that@xmath587 = o(n^{-2/3 } ) , \label{e11 } \\ & & \sup\limits_{b\in \theta _ { n}(c)}\left\vert \overline{s}(b,\widehat{g})-% \overline{s}(b , g)\right\vert = o_{p}(n^{-2/3 } ) .
\label{e21}\end{aligned}\]]equation ( [ e11 ] ) concerns stochastic equicontinuity of the local recentered process @xmath588 indexed by @xmath589 .
it is satisfied by setting @xmath590 and @xmath591 in the definition of sets @xmath592 and @xmath540 given by ( [ theta_delta ] ) and ( [ h_epsilon ] ) and by invoking lemma entropy bound copy(1 ) with the assumptions @xmath593 and @xmath134 .
we now verify equation ( [ e21 ] ) .
note that for @xmath1 sufficiently large , if @xmath594 , then @xmath370 under condition c1 .
let @xmath595where @xmath376 is defined by ( [ s1 ] ) . hence it suffices to verify @xmath596where @xmath597 @xmath598 .
note that the term @xmath599 is bounded above by @xmath600-[s_{1}(\widetilde{b},g)-s_{1}(\widetilde{\beta } % , g)]\right\vert .
\label{ineq terms}\]]since @xmath593 , by c11 we have that @xmath601 because the norm @xmath602 is stronger than the sup norm @xmath603 .
hence by lemma [ quadratic expansion of s(b , r ) ] , the first term of the sum ( [ ineq terms ] ) is @xmath604 and@xmath605where the terms @xmath606 , @xmath607 and @xmath608 are given by ( [ a1 ] ) , ( [ a2 ] ) and ( e(b , r ) ) , respectively . using
the proof of lemma [ quadratic expansion of s(b , r ) ] , it is also straightforward to verify that @xmath609since @xmath610 and @xmath611 , we have that @xmath612 and @xmath613 .
putting together ( [ t1 ] ) and ( [ t2 ] ) , it follows that the second term of the sum ( [ ineq terms ] ) is also @xmath604 and therefore equation ( [ e21 ] ) holds .
delgado , m.a .
, rodriguez - poo , j.m.,and wolf , m. ( 2001 ) , `` subsampling inference in cube root asymptotics with an application to manski s maximum score estimator '' , _ economics letters _ , 73 ( 2 ) , 241250 .
jun , s. j. , pinkse , j. and wan , y. ( 2013 ) , `` classical laplace estimation for @xmath614{n}$]-consistent estimators : improved convergence rates and rate - adaptive inference '' , working paper , department of economics , pennsylvania state university .
cccccc + @xmath1 & bias & rmse & median & mean ad & median ad + + 300 & -0.112 & 0.410 & 0.894 & 0.329 & 0.288 + 500 & -0.061 & 0.352 & 0.916 & 0.285 & 0.247 + 1000 & -0.031 & 0.264 & 0.961 & 0.211 & 0.182 + + 300 & -0.122 & 0.478 & 0.856 & 0.381 & 0.326 + 500 & -0.070 & 0.376 & 0.908 & 0.304 & 0.259 + 1000 & -0.033 & 0.301 & 0.952 & 0.240 & 0.211 cccccc + @xmath1 & bias & rmse & median & mean ad & median ad + + 300 & -0.056 & 0.330 & 0.918 & 0.262 & 0.216 + 500 & -0.044 & 0.277 & 0.942 & 0.220 & 0.184 + 1000 & -0.020 & 0.212 & 0.966 & 0.169 & 0.139 + + 300 & -0.394 & 0.489 & 0.577 & 0.431 & 0.432 + 500 & -0.413 & 0.469 & 0.568 & 0.424 & 0.432 + 1000 &
-0.400 & 0.435 & 0.587 & 0.402 & 0.412 cccccc + @xmath198 & bias & rmse & median & mean ad & median ad + + 5.4 & -0.063 & 0.639 & 0.868 & 0.502 & 0.417 + 5.6 & 0.041 & 0.653 & 0.966 & 0.500 & 0.400 + 5.8 & 0.138 & 0.718 & 1.067 & 0.544 & 0.427 + + 5.4 & -0.046 & 0.501 & 0.908 & 0.388 & 0.314 + 5.6 & 0.056 & 0.518 & 0.992 & 0.393 & 0.307 + 5.8 & 0.171 & 0.584 & 1.096 & 0.435 & 0.331 + + 5.4 & -0.087 & 0.389 & 0.887 & 0.311 & 0.266 + 5.6 & 0.008 & 0.380 & 0.992 & 0.307 & 0.264 + 5.8 & 0.111 & 0.424 & 1.086 & 0.334 & 0.278 cccccc + @xmath198 & bias & rmse & median & mean ad & median ad + + 5.8 & -0.071 & 0.480 & 0.918 & 0.380 & 0.328 + 6 & 0.053 & 0.523 & 1.028 & 0.408 & 0.340 + 6.2 & 0.147 & 0.577 & 1.132 & 0.448 & 0.372 + + 5.8 & -0.066 & 0.400 & 0.906 & 0.316 & 0.264 + 6 & 0.030 & 0.408 & 1.004 & 0.323 & 0.268 + 6.2 & 0.103 & 0.457 & 1.062 & 0.355 & 0.288 + + 5.8 & -0.125 & 0.298 & 0.865 & 0.243 & 0.211 + 6 & -0.028 & 0.286 & 0.956 & 0.230 & 0.192 + 6.2 & 0.059 & 0.320 & 1.038 & 0.251 & 0.211 cccccc + @xmath198 & bias & rmse & median & mean ad & median ad + + 0.6 & -0.172 & 0.483 & 0.803 & 0.392 & 0.345 + 0.8 & -0.122 & 0.502 & 0.865 & 0.395 & 0.328 + 1 & -0.088 & 0.510 & 0.896 & 0.401 & 0.333 + + 0.6 & -0.111 & 0.391 & 0.880 & 0.315 & 0.276 + 0.8 & -0.073 & 0.394 & 0.913 & 0.316 & 0.268 + 1 & -0.037 & 0.408 & 0.937 & 0.326 & 0.280 + + 0.6 & -0.054 & 0.305 & 0.923 & 0.247 & 0.216 + 0.8 & -0.028 & 0.301 & 0.956 & 0.242 & 0.211 + 1 & 0.002 & 0.313 & 0.980 & 0.250 & 0.216 cccccc + @xmath198 & bias & rmse & median & mean ad & median ad + + 0.6 & -0.112 & 0.440 & 0.865 & 0.347 & 0.297 + 0.8 & -0.057 & 0.443 & 0.918 & 0.351 & 0.302 + 1 & -0.009 & 0.469 & 0.968 & 0.370 & 0.316 + + 0.6 & -0.077 & 0.366 & 0.918 & 0.291 & 0.244 + 0.8 & -0.040 & 0.382 & 0.932 & 0.302 & 0.254 + 1 & -0.010 & 0.397 & 0.966 & 0.313 & 0.264 + + 0.6 & -0.037 & 0.272 & 0.952 & 0.218 & 0.182 + 0.8 & -0.012 & 0.272 & 0.980 & 0.218 & 0.192 + 1 & 0.036 & 0.286 & 1.028 & 0.230 & 0.201 | this paper develops maximum score estimation of preference parameters in the binary choice model under uncertainty in which the decision rule is affected by conditional expectations .
the preference parameters are estimated in two stages : we estimate conditional expectations nonparametrically in the first stage and then the preference parameters in the second stage based on manski ( 1975,1985 ) s maximum score estimator using the choice data and first stage estimates . the paper establishes consistency and derives rate of convergence of the two - stage maximum score estimator .
moreover , the paper also provides sufficient conditions under which the two - stage estimator is asymptotically equivalent in distribution to the corresponding single - stage estimator that assumes the first stage input is known .
these results are of independent interest for maximum score estimation with nonparametrically generated regressors .
the paper also presents some monte carlo simulation results for finite - sample behavior of the two - stage estimator .
* keywords : * _ discrete choice , maximum score estimation , generated regressor , preference parameters , m - estimation , cube root asymptotics _ * jel codes : * c12 , c13 , c14 . |
in this paper and its two companion papers @xcite the relation of the electrons of the one - dimensional ( 1d ) hubbard model to the quantum objects whose occupancy configurations describe its energy eigenstates is investigated for the whole hilbert space .
the study of such a non - perturbative relation is a necessary step for the description of the finite - energy few - electron spectral properties of the many - electron quantum problem .
the study of the one - dimensional ( 1d ) hubbard model @xcite for interacting electrons is of general importance because the understanding of correlated systems and of their finite - energy excitations is still far from complete .
the problem of the 1d hubbard model in the limit of large and infinite on - site coulomb repulsion was previously studied in the literature by many authors @xcite .
our studies of the model for finite values of the on - site repulsion @xmath3 are motivated by the anomalous one - electron and two - electron spectral properties observed in metallic and insulating phases of quasi - one - dimensional materials , which can not be described by the usual fermi - liquid theory @xcite .
recently there has been a renewed experimental interest on the properties of these materials @xcite .
some of these experimental studies observed unusual finite - energy / frequency spectral properties @xcite . since in the case of finite - excitation energy
the luttinger liquid description does not apply @xcite , these finite - energy / frequency spectral properties are far from being well understood . however , there are indications that electronic correlation effects might play an important role in the finite - energy physics of these low - dimensional materials @xcite
. for low values of the excitation energy the microscopic electronic properties of these materials are usually described by systems of coupled chains . on the other hand , for finite values of the excitation energy larger than the transfer integrals for electron hopping between the chains , 1d lattice models taking into account the screened electron - electron coulomb repulsion
are expected to provide a good description of the physics of these materials .
the simplest of these models is the above 1d hubbard model @xcite , which describes such electron - electron interactions by an effective on - site coulomb repulsion @xmath3 .
this model corresponds to a non - perturbative electronic problem . according to the results of the companion paper @xcite ,
its energy eigenstates can be described by occupancy configurations of holons , spinons , and @xmath2 pseudoparticles .
holons and spinons has also been studied for other models @xcite . the 1d hubbard model is often considered a suitable model for the description of the electronic correlation effects and the non - perturbative microscopic mechanisms behind the unusual few - electron spectral properties observed in quasi - one - dimensional materials @xcite .
moreover , recent angle - resolved ultraviolet photoemission spectroscopy revealed very similar spectral fingerprints from both high-@xmath4 superconductors and quasi - one - dimensional compounds @xcite .
the similarity of the ultraviolet data for these two different systems could be evidence of the occurrence of a charge - spin separation associated with holons and spinons .
the anomalous temperature dependence of the spectral function could also indicate a dimensional crossover @xcite .
the results of refs .
@xcite also suggest that the unconventional spectral properties observed in two - dimensional ( 2d ) materials could have a 1d origin .
thus the holons and spinons could play an important role in spectral properties of both 1d and 2d low - dimensional materials . in this second paper
we continue the studies of the first paper of this series , ref .
@xcite . as a preliminary application of the connection of the concept of rotated electron to the quantum numbers that label the energy eigenstates provided by the bethe - ansatz solution and @xmath0-spin and spin symmetries , in this paper we use exact holon and spinon selection rules for rotated - electron operators in the study of the holon and spinon contents of few - electron excitations .
this reveals the dominant holon and spinon microscopic physical processes that generate more than 99% of the spectral weight of few - electron excitations . while our theory also describes the higher - order processes associated with the remaining less than 1% electronic spectral weight , the clarification of the dominant holon and spinon microscopic mechanisms is valuable for the further understanding and description of the few - electron spectral properties observed in real low - dimensional materials
. fortunately , a preliminary application of the theoretical predictions which follow from our study of the holon and spinon contents of few - electron excitations leads to quantitative agreement with the charge and spin spectral branch lines observed by means of angle - resolved photoelectron spectroscopy ( arpes ) in the organic conductor ttf - tcnq @xcite .
the preliminary results reported in ref .
@xcite confirm that from the experimental point of view only the spectral weight associated with the dominant holon and spinon microscopic processes is observed .
moreover , in this paper we introduce the concept of an _ effective electronic lattice_. the expression of the electrons in terms of holons , spinons , and pseudoparticles through the electron - rotated - electron transformation studied here for all values of @xmath3 is a first necessary step for the evaluation of finite - energy few - electron spectral function expressions , as further discussed in sec .
v. in this paper we also discuss and clarify how the relationship of the original electrons to the quantum objects whose occupancy configurations describe the energy eigenstates can be used in a program for evaluation of finite - energy few - electron spectral functions .
the successful fulfilment of such a program needs the concepts of local pseudoparticle and effective pseudoparticle lattice introduced in the third paper of this series , ref .
moreover , in this paper we generalize the concepts of a lower hubbard band and upper hubbard bands to all values of on - site coulombian repulsion .
the paper is organized as follows : in sec .
ii we provide a short introduction to the 1d hubbard model .
the definition of the upper hubbard bands in terms of rotated - electron double occupation , the holon and spinon selection rules for rotated - electron operators , and the concept of an effective electronic lattice are presented and introduced in sec .
the dominant holon and spinon microscopic physical processes that control the few - electron spectral properties are studied in sec .
finally , in sec .
v we present the discussion of our results and the concluding remarks .
this includes the discussion of the application of the concepts introduced in the present series of three papers to the fulfilment of a program for evaluation of finite - energy few - electron spectral functions .
in a chemical potential @xmath5 and magnetic field @xmath6 the 1d hubbard hamiltonian can be written as , @xmath7 + \sum_{\alpha = c,\,s}\mu_{\alpha } \,{\hat{s}}^{\alpha}_z \label{h}\ ] ] where , @xmath8 is the `` simple '' hubbard model .
the operators @xmath9 and @xmath10 on the right - hand side of eq .
( [ hh ] ) are the kinetic energy and the electron double occupation operators , respectively . the operator @xmath11 of eq .
( [ dop ] ) counts the number of spin @xmath12 electrons at real - space lattice site @xmath13 .
the number of lattice sites @xmath14 is even and large and @xmath15 is odd .
we consider periodic boundary conditions .
the operators @xmath16 and @xmath17 which appear in the above equations are the spin @xmath18 electron creation and annihilation operators at site @xmath19 . moreover , on the right - hand side of eq .
( [ h ] ) , @xmath20 , @xmath21 , @xmath22 is the bohr magneton , and the operators @xmath23 and @xmath24 are given in eq .
( 2 ) of ref .
@xcite and are the diagonal generators of the @xmath0-spin @xmath25 and spin @xmath26 @xmath27 algebras @xcite , respectively .
the electron number operator reads @xmath28 , where the operator , @xmath29 counts the number of spin @xmath12 electrons .
the momentum operator reads @xmath30 \ , , \label{popel}\ ] ] and commutes with the hamiltonian introduced in eq .
( [ h ] ) .
there are @xmath31 spin - up electrons and @xmath32 spin - down electrons in the chain of @xmath14 sites , lattice constant @xmath33 , and length @xmath34 $ ] associated with the model ( [ h ] ) .
we introduce fermi momenta which in the present thermodynamic limit @xmath35 are given by @xmath36 and @xmath37/2=\pm \pi n/2 $ ] , where @xmath38 and @xmath39 .
the electronic density can be written as @xmath40 and the spin density is given by @xmath41 .
in general we consider densities in the domains @xmath42 ; @xmath43 and @xmath44 ; @xmath45 , respectively .
however , in the case of the study of transitions whose initial state is a ground state , for simplicity we restrict our considerations to values of the electronic density @xmath46 and spin density @xmath47 such that @xmath42 and @xmath48 , respectively .
the hamiltonian @xmath49 $ ] commutes with the six generators of the @xmath0 spin @xmath50 and spin @xmath51 algebras @xcite , their off - diagonal generators being given in eqs .
( 9 ) and ( 10 ) of ref .
the bethe - ansatz solvability of the 1d hubbard model is restricted to the hilbert subspace spanned by the lowest - weight states ( lwss ) of the @xmath0-spin and spin algebras ,
_ i.e. _ such that @xmath52 @xcite . as in the first paper of this series @xcite
, here we denote the holons ( and spinons ) according to their @xmath0-spin projections @xmath53 ( spin projections @xmath53 ) . for the definition of the holon and spinon numbers , their relations , and other aspects of the holon , spinon , and pseudoparticle description which are useful for the studies of this paper see the companion paper @xcite . an important concept introduced in ref .
@xcite is that of cphs ensemble subspace where cphs stands for @xmath2 pseudoparticle , holon , and spinon .
this is a hilbert subspace spanned by all states with fixed values for the @xmath54 yang holon number @xmath55 , @xmath54 hl spinon number @xmath56 , @xmath2 pseudoparticle number @xmath57 , and for the sets of @xmath58 pseudoparticle numbers @xmath59 and @xmath60 corresponding to the @xmath61 branches .
we note that according to the notation of ref .
@xcite in _ hl spinon _ hl stands for _ heilmann and lieb_.
in this section we generalize the large-@xmath3 concept of upper hubbard band @xcite to all values of the on - site electronic repulsion @xmath3 in terms of rotated - electron double occupation .
moreover , we use the relation of rotated electrons to holons and spinons to introduce rotated - electron exact selection rules .
these rules limit the values of the deviations in holon and spinon numbers associated with excited states generated by application of rotated - electron operators onto any eigenstate of the electron spin @xmath12 number operator ( [ nsi ] ) .
deviation values outside the ranges given by these selection rules correspond to forbidden excited states .
in addition , in this section we introduce the concept of an _ effective electronic lattice_. this concept is further used in the third paper of this series , ref .
@xcite , in the introduction of the concepts of a _ local pseudoparticle _ and an _ effective pseudoparticle
lattice_. the latter concepts play an important role in the studies of finite - energy few - electron spectral functions of refs .
@xcite . as discussed in the companion paper @xcite , the operator @xmath62 represents the rotated electrons .
the relation of such an operator to the electronic operator @xmath63 is such that , @xmath64 here @xmath65 is the electron - rotated - electron unitary operator uniquely defined by eqs .
( 43)-(45 ) of ref .
the rotated - electron double - occupation operator plays an important role in our studies and is given by , @xmath66 where @xmath67 is the electron double occupation operator given in eq .
( [ dop ] ) .
note that @xmath63 and @xmath62 are only identical in the @xmath68 limit where electron double occupation becomes a good quantum number .
let the operator @xmath69 ( or @xmath70 ) have momentum @xmath71 and be a product of a finite number @xmath72 of one - electron ( or one - rotated - electron ) elementary creation and/or annihilation operators . here
the ratio @xmath73 vanishes in the present thermodynamic limit and @xmath74 is the number of electron ( or rotated - electron ) creation and annihilation operators for @xmath75 and @xmath76 , respectively , and with spin down and spin up for @xmath77 and @xmath78 , respectively .
the studies we present below in this paper refer to the transformations generated by the application of the general @xmath79-electron operator @xmath69 ( or @xmath79-rotated - electron operator @xmath70 ) onto any eigenstate of the spin @xmath12 electron number operator ( [ nsi ] ) .
however , we focus our attention on the particular case of states @xmath80 ( or @xmath81 ) which result from application of such an operator onto a ground state @xmath82 .
we note that the number of spin @xmath12 electrons equals the number of spin @xmath12 rotated electrons .
according to the results of the companion paper @xcite , rotated - electron double occupation equals the number of @xmath54 holons .
we emphasize that for electronic densities @xmath46 such that @xmath83 the ground state has no @xmath54 holons @xcite . since rotated - electron double occupation is a good quantum number , in this case we can write a general @xmath79-electron spectral function as follows , @xmath84 where @xmath85 the excitation energy @xmath86 is positive , the @xmath87 summation of eq .
( [ onsf ] ) runs over the values of rotated - electron double occupation @xmath88 of the excited energy eigenstates @xmath89 , the @xmath19 summation of eq .
( [ onmsf ] ) runs over all available excited energy eigenstates @xmath90 with the same value @xmath87 of rotated - electron double occupation , and @xmath91 $ ] are the excitation energies relative to the initial ground state .
examples of few - electron operators @xmath69 whose spectral - function excitation energy is usually considered positive are the one - electron spin @xmath12 creation operator of momentum @xmath71 , the two - electron charge operator of momentum @xmath71 , and the two - electron singlet or triplet cooper pair addition operator of momentum @xmath71 .
we call _ lower hubbard band _ the spectral weight distribution associated with the function @xmath92 defined in eq .
( [ onmsf ] ) .
moreover , we call _
@xmath93 upper hubbard band _ the spectral weight distribution associated with the function @xmath94 defined in eq . ( [ onmsf ] ) such that @xmath95 .
in the ensuing section we find that for one - electron and two - electron operators most spectral weight correspond to the lower hubbard band and first ( or first and second ) upper hubbard band(s ) .
the detailed study of the few - electron spectral functions ( [ onsf ] ) is for finite values of @xmath3 a complex many - electron open problem .
such a problem is studied in ref .
@xcite by the combination of the pseudofermion representation introduced in ref .
@xcite with the holon and spinon description introduced and studied in this paper and in its companion papers @xcite , as discussed in sec .
v. in this paper we use the relation of electrons and rotated electrons to the quantum objects whose occupancy configurations describe the ground state and excited states @xmath90 that appear in eqs .
( [ onsf ] ) and ( [ onmsf ] ) to obtain useful information about the finite - energy few - electron spectral distributions .
an interesting property is that in the metallic phase of the 1d hubbard model the minimum value of the excitation energy @xmath86 for the lower hubbard band is zero .
this justifies the designation _
lower band_. on the other hand , the minimum value of the excitation energy for the @xmath93 upper hubbard band is for finite values of @xmath3 finite .
such a value depends on the minimum excitation - energy value for the creation of a @xmath54 holon . in the case of the hamiltonian ( [ h ] )
, this minimum energy value is for electronic densities @xmath46 such that @xmath96 given by @xmath97 .
for a zero spin density @xmath98 initial ground state , the energy @xmath99 equals twice the chemical potential and is an increasing function of the on - site repulsion @xmath3 with the following limiting values , @xmath100 moreover , this energy parameter for any value of @xmath101 is a decreasing function of the electronic density @xmath46 such that , @xmath102 where @xmath103 is the half - filling mott - hubbard gap @xcite . on the other hand ,
if one uses as initial state the same @xmath98 ground state but considers the excitation energy of the `` simple '' hubbard hamiltonian ( [ hh ] ) , there is a shift in the definition of the ground - state zero - energy level . for this hamiltonian the minimum excitation - energy value for creation of a @xmath54 holon is simply given by @xmath3 . since rotated - electron double occupation equals the number of @xmath54 holons , the minimum excitation - energy value for an excited energy eigenstate with rotated electron double occupation @xmath104 is @xmath105 .
this justifies why all gapless excitation branches are associated with zero rotated - electron double occupation states .
the excitation energy @xmath105 is measured from the ground - state level and corresponds to the hamiltonian ( [ h ] ) .
if instead we consider the ground - state zero - energy level associated with the hamiltonian ( [ hh ] ) , the value of such an excitation energy becomes @xmath106 .
any operator @xmath107 can be expanded as follows , @xmath108 where @xmath109 is such that application of the operator @xmath110 onto any eigenstate of rotated electron double occupation of eigenvalue @xmath111 leads to a final state with rotated electron double occupancy @xmath112\geq 0 $ ] .
also the corresponding rotated operator @xmath113 such that , @xmath114 can be decomposed as , @xmath115 the index @xmath116 of expressions ( [ m ] ) and ( [ mtil ] ) plays the same role as the index @xmath117 of ref .
@xcite . while @xmath118 , the energy @xmath119 only equals @xmath3 in the @xmath68 limit .
since our study refers to all values of @xmath101 , here we use @xmath119 instead of @xmath3 to label the energy of the different upper hubbard bands defined above . for finite values of @xmath101
the choice of ref .
@xcite refers to the zero - energy level of the hamiltonian ( [ hh ] ) , whereas our choice corresponds to the ground - state zero - energy level of the more suitable hamiltonian ( [ h ] ) .
our choice is justified by the fact that in real experiments the excitation energy is , in general , measured from the fermi level .
our notation tells that the minimal amount of excitation energy for a transition which results from application of an operator component @xmath110 onto any eigenstate of rotated electron double occupation is @xmath116
. this excitation energy can be positive ( @xmath95 ) or negative ( @xmath120 ) .
we note that application onto a state of rotated - electron double occupation @xmath121 of a @xmath120 operator component @xmath110 gives zero . since for the present case of electronic densities @xmath46 such that @xmath96 , the ground state has zero rotated - electron double occupation , all the final excited states of the general @xmath79-electron spectral function ( [ onsf])-([onmsf ] ) have @xmath122 .
the operators @xmath123 and @xmath65 associated with the electron - rotated - electron unitary transformation can be written as , @xmath124 these operators are uniquely defined by eqs .
( 44 ) and ( 45 ) of the companion paper @xcite and can be expressed in the form of an expansion in @xmath125 for the operator @xmath126 @xcite .
an important role in such an expansion is played by the rotated kinetic energy operator @xmath127 , associated with the operator ( [ top ] ) . the decomposition ( [ mtil ] ) of this operator leads to three terms , @xmath128 where @xmath129 \ , , \label{ttil0}\end{aligned}\ ] ] @xmath130 and @xmath131 in these expressions
@xmath132 is such that @xmath133 and @xmath134 .
the above - mentioned expansion of the operator @xmath126 of the expressions of eq .
( [ sv ] ) in powers of @xmath125 is obtained by systematic use of eqs .
( 44 ) and ( 45 ) of the companion paper @xcite , with the result , @xmath135 + ... \ , , \label{sexp}\ ] ] where the components of the rotated kinetic energy operator are given in eqs .
( [ ttil0])-([ttil-1 ] ) . while equations ( 44 ) and ( 45 ) of ref .
@xcite uniquely define the electron - rotated - electron unitary operator @xmath65 of eq .
( [ sv ] ) for all values of @xmath101 , it is difficult to extract a closed form expression for that operator from these equations .
let us consider the case of few - electron operators @xmath69 whose spectral - function excitation energy is usually considered negative .
example of such operators are the one - electron spin @xmath12 annihilation operator of momentum @xmath71 and the two - electron singlet or triplet cooper pair annihilation operator of momentum @xmath71 .
we note that according to the results of the companion paper @xcite , the number of rotated - electron empty sites equals the number @xmath136 of @xmath137 holons .
according to eq .
( c24 ) of the same paper , for values of the electronic density @xmath46 such that @xmath96 the ground state has a number of @xmath137 holons whose value is given by @xmath138 $ ] . since the number of rotated - electron empty sites is a good quantum number , we can write a general @xmath79-electron spectral function associated with such a type of few - electron operators as follows , @xmath139 where @xmath140 the excitation energy @xmath86 is negative , the @xmath141 $ ] summation of eq .
( [ onsfb ] ) runs over the values of the deviation in the number of rotated - electron empty sites @xmath142 of the excited energy eigenstates @xmath143 , the @xmath19 summation of eq .
( [ onmsfb ] ) runs over all available excited energy eigenstates @xmath143 with the same value @xmath144 $ ] for the number of rotated - electron empty sites , and @xmath145 $ ] are the excitation energies relative to the initial ground state .
note that in the case of half filling the ground - state value for the number of rotated - electron empty sites is zero and the @xmath146 summation of eq .
( [ onsfb ] ) runs over positive values of rotated - electron empty sites @xmath147 only .
moreover , for electronic densities @xmath148 the role of the numbers of rotated - electron doubly occupied and empty sites is interchanged . for these densities
the summations over the values of rotated - electron doubly occupied and empty sites of eqs .
( [ onsf ] ) and ( [ onsfb ] ) , respectively , correspond to both positive and negative integers and to positive integers , respectively .
this is because for such densities the ground state has no @xmath137 holons . in this paper
we consider mostly the case of electronic densities @xmath149 , where the ground state has no @xmath54 holons . as further discussed in sec .
v , the first step for the evaluation of the general few - electron spectral functions ( [ onsf])-([onmsf ] ) and ( [ onsfb])-([onmsfb ] ) is the expression of the @xmath79-electron operator @xmath150 in terms of the corresponding rotated operator @xmath151 defined in eq .
( [ otil ] ) as follows @xcite , @xmath152 + { 1\over 2}\,[{\hat{s}},\,[{\hat{s}},\,{\hat{o}}_{\cal{n}}(k)\,]\ , ] + ... \ , .
\label{ontil}\end{aligned}\ ] ] according to the results of the companion papers @xcite , the relation of the rotated operators on the right - hand side of eq .
( [ ontil ] ) to the quantum objects whose occupancy configurations describe the ground state and excited states @xmath89 and @xmath143 of the few - electron spectral functions ( [ onsf])-([onmsf ] ) and ( [ onsfb])-([onmsfb ] ) , respectively , is well defined .
expression of the second expression of eq .
( [ ontil ] ) in terms of the creation and annihilation operators for these quantum objects plays a crucial role in the evaluation of the matrix elements of the few - electron spectral functions ( [ onmsf ] ) and ( [ onmsfb ] ) .
this justifies the importance of the holon , spinon , and @xmath2 pseudoparticle description introduced in the companion paper @xcite and further studied here and in the companion paper @xcite for the study of the few - electron spectral properties at finite values of the excitation energy .
solution of this problem is a direct application of the relation of rotated electrons to the quantum numbers of the bethe - ansatz solution and @xmath0-spin and spin symmetries found in the companion paper @xcite .
such a breakthrough provided the relation of electrons and rotated electrons to the above - mentioned quantum objects . in the ensuing section
we find that the first operator term of the second expression on the right - hand side of eq .
( [ ontil ] ) corresponds to the dominant microscopic physical processes that control the few - electron spectral - weight properties . in the case of @xmath153-electron operators
we find that more than 99% of the electronic spectral weight generated by application of the operator @xmath154 onto a ground state corresponds to application of the corresponding @xmath79-rotated - electron operator @xmath155 onto the same state .
this latter operator has a simple expression in terms of holon and spinon elementary operators , and it is associated with the dominant holon and spinon microscopic mechanisms that control the @xmath79-electron spectral properties .
let @xmath156 be any eigenstate of the spin @xmath12 electron number operator ( [ nsi ] ) and @xmath157 ( or @xmath158 ) a general @xmath159-electron ( or @xmath159-rotated - electron ) operator . the corresponding state @xmath160 ( or @xmath161 ) is also an eigenstate of the spin @xmath12 electron number operator ( [ nsi ] ) .
such a state belongs to an electron ensemble space whose electron numbers differ from the numbers of the initial state by deviations @xmath162 and @xmath163 , such that @xmath164 .
since the number of spin @xmath12 electrons equals the number of spin @xmath12 rotated electrons , these deviations also refer to the latter numbers .
these deviations can be expressed in terms of the numbers @xmath74 of eq .
( [ d ] ) as follows , @xmath165 following the results of refs .
@xcite , let us consider the four expectation values @xmath166 , where @xmath167 and @xmath168 and the corresponding operator @xmath169 counts the number of electron doubly - occupied sites , @xmath170 counts the number of electron empty sites , @xmath171 counts the number of spin - down electron singly - occupied sites , and @xmath172 counts the number of spin - up electron singly - occupied sites .
these operators are given in eqs .
( 23 ) and ( 24 ) of the companion paper @xcite . in reference
@xcite it was found that in the limit @xmath68 the deviations @xmath173 , @xmath174 , @xmath175 , and @xmath176 generated by application of a general @xmath79-electron operator @xmath154 onto any eigenstate of the spin @xmath12 electron number operator ( [ nsi ] ) are restricted to the ranges of the inequalities ( 60)-(63 ) of the same reference .
obviously , given one of the four ranges of values defined by these four inequalities , the other three are dependent and follow from the relations given in eq .
( 56 ) of ref .
@xcite .
let us consider the deviations @xmath177 , @xmath178 , @xmath179 , and @xmath180 in the expectation values of the four operators given in eqs .
( 23 ) and ( 24 ) of ref .
@xcite in the case that both the corresponding initial state @xmath181 and final state @xmath182 are energy eigenstates of the 1d hubbard model in the limit @xmath68 .
note that it is assumed that the final energy eigenstate @xmath182 is contained in the state @xmath183 where @xmath154 is the general @xmath79-electron operator under consideration . in the case of these initial and final states the above - mentioned deviations are such that , @xmath184 for all values of @xmath101 . in order to derive this result we used eq .
( 52 ) of the companion paper @xcite .
( 54)-(57 ) of the same paper were used in the evaluation of the equalities on the right - hand side of eq .
( [ drdmee ] ) .
equation ( 52 ) of ref .
@xcite relates the energy eigenstates of the @xmath68 hubbard model to the corresponding energy eigenstates of the hubbard model for any value of @xmath101 .
we note that the last quantity on the right - hand side of eq .
( [ drdmee ] ) is nothing but the deviation @xmath185 in the particular case when the initial state @xmath186 and final state @xmath187 are energy eigenstates of the finite-@xmath101 hubbard model .
the final energy eigenstate @xmath187 is contained in the state @xmath188 , where the rotated operator @xmath155 is related to the operator @xmath154 by eq .
( [ ontil ] ) .
the operator @xmath189 = [ { \hat{s}},\,{\tilde{o}}_{\cal{n}}\ , ] + { 1\over 2}\,[{\hat{s}},\,[{\hat{s}},\,{\hat{o}}_{\cal{n}}\,]\ , ] + ... \ , , \label{dontil}\ ] ] can be expressed as a sum of @xmath190-electron operators such that @xmath191 .
note that eq .
( [ dontil ] ) contains the same information as eq .
( [ ontil ] ) .
the quantities @xmath192 and @xmath193 of eq .
( [ dmasee ] ) are the eigenvalues of the energy eigenstates @xmath186 and @xmath187 , respectively , relative to the holon ( @xmath194 ) or spinon ( @xmath195 ) number operator @xmath196 studied in the companion paper @xcite .
these two states are related to the above initial state @xmath181 and final state @xmath182 by the transformation given in eq .
( 52 ) of the same paper .
the equalities of eq .
( [ drdmee ] ) are valid for all possible @xmath197 initial energy eigenstates @xmath198 of the 1d hubbard model in the limit @xmath68 such that @xmath199 and for final energy eigenstates @xmath182 which are contained in @xmath183 . according to these equalities , the deviations in the eigenvalues of the four operators @xmath200 of eqs .
( 23 ) and ( 24 ) of the companion paper @xcite which count the number of electron doubly - occupied sites , empty sites , spin - down singly - occupied sites , and spin - up singly - occupied sites equal the deviations in the eigenvalues of the four operators @xmath196 given in eq .
( 22 ) of the same paper .
these four operators count the number of @xmath54 holons , @xmath137 holons , @xmath54 spinons , and @xmath137 spinons , respectively .
note that according to eqs .
( 52 ) and ( 54)-(57 ) of ref .
@xcite such an equality is valid because the initial and final states of the latter deviations are the energy eigenstates of the finite-@xmath101 1d hubbard model @xmath201 and @xmath202 , respectively . here
the final state @xmath202 is contained in the state @xmath203 .
the electron double - occupation , no - occupation , spin - down single - occupation , and spin - up single - occupation eigenvalue deviations ( [ drdmee ] ) which result from application onto the state @xmath198 of the above general operator @xmath154 are restricted to the ranges of the inequalities ( 60)-(63 ) of ref .
@xcite . therefore , also the the @xmath54-holon - number , @xmath137-holon - number , @xmath54-spinon - number , and @xmath137-spinon - number eigenvalue deviations ( [ dmasee ] ) which result from application onto the state @xmath204 of the corresponding @xmath79-rotated - electron operator @xmath155 are restricted to the ranges of the following inequalities , @xmath205 @xmath206 @xmath207 and @xmath208 according to eq .
( [ dmasee ] ) , the eigenvalues @xmath192 and @xmath193 of these deviations refer to an initial state which is one of the @xmath197 energy eigenstates @xmath209 of the finite-@xmath101 hubbard model and to a final state which is an energy eigenstate @xmath187 of the same model contained in the state @xmath210 .
however , since the @xmath197 states @xmath204 associated with the deviations of eq .
( [ dmasee ] ) constitute a complete basis for the hilbert space of the 1d hubbard model , the inequalities ( [ mc - range])-([ms+range ] ) also apply to general deviations of the form , @xmath211 provided that the arbitrary states @xmath212 and @xmath213 are eigenstates of the spin @xmath12 electron number operator ( [ nsi ] ) .
it follows from the relations of eq .
( 50 ) of the companion paper @xcite that out of the four inequalities ( [ mc - range])-([ms+range ] ) , only one is independent . by summation of the two inequalities ( [ mc - range ] ) and ( [ mc+range ] ) and of the two inequalities ( [ ms - range ] ) and ( [ ms+range ] )
we find the following inequalities for the deviations in the total number of holons and spinons , @xmath214 note that the limiting values of the inequalities ( [ dmcsrange ] ) are simply @xmath215 and @xmath216 , where @xmath216 is the number of elementary rotated - electron operators of the expression of the operator @xmath155 . finally , through the use of eq .
( 28 ) of ref .
@xcite we find that the validity of these inequalities is equivalent to the inequalities @xmath217 for the deviations @xmath218 in the number @xmath219 of @xmath2 pseudoparticles and in the number @xmath220 of @xmath2 pseudoparticle holes , respectively .
we emphasize that the exact holon and spinon selection rules ( [ mc - range])-([ms+range ] ) and ( [ dmcsrange ] ) refer to deviations generated by @xmath79-rotated - electron operators .
we find below that in the case of deviations generated by the corresponding @xmath79-electron operator the dominant holon and spinon microscopic physical processes lead to excited states that obey the inequalities ( [ mc - range])-([ms+range ] ) and ( [ dmcsrange ] ) .
such dominant microscopic physical processes control most of the @xmath79-electron spectral - weight .
interestingly , as in the @xmath221 case , for @xmath222 these rules are exact for the corresponding @xmath79-electron operator , the dominant processes becoming the only processes contributing to the @xmath79-electron spectral properties . on the other hand , for values of the on - site repulsion such that @xmath223 these rules are not exact for @xmath79-electron operators but correspond to dominant processes that amount for more than 99% of the electronic spectral weight .
for instance , in the case of one - electron creation operators , we find that the spectral weight generated by higher - order holon and spinon processes is extremely small .
such a weight corresponds to states with holon and spinon numbers outside the domains defined by inequalities ( [ mc - range])-([ms+range ] ) and ( [ dmcsrange ] ) and is maximum for half filling and at @xmath223 where it amounts for about 0.25% of the total spectral weight .
however , such a weight decreases for decreasing values of the electronic density .
for instance , at quarter filling it is maximum also for @xmath223 but amounts for 0.005% of the total spectral weight only .
moreover , for small electronic densities the dominant processes become the only processes for @xmath79-electron operators at any value of @xmath101 . in the case of the one - electron creation operators
the states generated by the dominant processes refer to a _ minimum _ amount of relative spectral weight for @xmath224 .
nevertheless , such a _ minimum _ value corresponds for the dominant holon and spinon processes to 99.75% , 99.99% , and 100.00% of the total one - electron addition spectral weight for electronic densities @xmath225 , @xmath226 , and @xmath227 , respectively , as found in sec .
iv . the electron - rotated - electron unitary transformation performs a rotation in hilbert space which maps electrons onto rotated - electrons such that rotated - electron double occupation is a good quantum number . as discussed below , the rotated - electron site distribution configurations that describe the energy eigenstates are independent on the value of the ratio @xmath101 .
in contrast , the electron site distribution configurations that describe these states are dependent on @xmath3 . this is consistent with the @xmath101 dependence of the electron double - occupation quantities studied in ref .
@xcite .
as discussed in the companion paper @xcite , the electronic lattice remains invariant under the electron - rotated - electron unitary transformation .
however , in order to distinguish the rotated - electron from the electronic site distribution configurations it is useful to introduce an _ effective electronic lattice_. the rotated electrons occupy the sites of the effective electronic lattice , whereas the electrons occupy the sites of the _ real - space lattice_. such an effective electronic lattice has the same number of sites @xmath228 , lattice constant @xmath33 , and length @xmath229 as the real - space lattice . in the companion paper @xcite it is found that the rotated - electron site distribution configurations that describe the energy eigenstates include separated charge and spin sequences .
the charge ( and spin ) sequences correspond to distribution configurations of the rotated - electron doubly occupied sites and empty sites ( and spin - down and spin - up singly occupied sites ) . for the effective electronic lattice
the number of @xmath54 holons equals rotated - electron double occupation and thus plays the same role as electron double occupation in the real - space lattice . for finite values of @xmath101 ,
the number of electron doubly - occupied sites , empty sites , spin - down singly - occupied sites , and spin - up singly - occupied sites of the real - space lattice electron site distribution configurations that describe the energy eigenstates , is not a good quantum number .
in contrast , for the effective electronic lattice the number of rotated - electron doubly - occupied sites , empty sites , spin - down singly - occupied sites , and spin - up singly - occupied sites is a good quantum number for all values of @xmath101 .
this number equals the number of @xmath54 holons , @xmath137 holons , @xmath54 spinons , and @xmath137 spinons , respectively , of any energy eigenstate .
a property of major importance is that the effective electronic lattice site distribution configurations of the rotated electrons which describe the energy eigenstate @xmath204 of the relation ( 52)of the companion paper @xcite are the same as the corresponding real - space lattice site distribution configurations of the electrons which describe the state @xmath198 .
since this latter state is an energy eigenstate of the 1d hubbard model in the limit @xmath68 , the rotated - electron site distribution configurations which describe the energy eigenstates are independent of the value of @xmath101 and are precisely the same as the corresponding electron site distribution configurations of the @xmath68 1d hubbard model . in the companion paper
@xcite it is shown that this property is behind the fact that the @xmath2 and @xmath230 pseudoparticle band - momentum occupancy configurations which describe the energy eigenstates are also independent of the value of @xmath101 , only the energy spectrum of such configurations are @xmath101 dependent .
the fact that the @xmath68 selection rule ranges given in eqs .
( 60)-(63 ) of ref .
@xcite are precisely the same as the finite-@xmath101 ranges imposed by the inequalities ( [ mc - range])-([ms+range ] ) results from the @xmath101 independence of the rotated - electron site distribution configurations which describe the energy eigenstates .
another important property is that for electronic densities smaller or equal to one , a ground state has zero rotated - electron double occupation .
the same occurs for the ground state electron site distribution configurations of the real - space lattice in the limit @xmath68 , the ground - state electron double occupation being in that limit given by @xmath231 .
this property is consistent with the above property that the rotated - electron site distribution configurations which describe the energy eigenstates are @xmath101 independent .
note that creation of a rotated electron onto the ground state either transforms an empty site of the effective electronic lattice into a singly - occupied site or a singly - occupied site of the lattice into a doubly - occupied site .
these two alternative transitions lead to @xmath54 holon number deviations such that @xmath232 and @xmath233 , respectively .
therefore , the selection rule of eq .
( [ src ] ) leads to permitted final states such that the values of these deviations are @xmath234 . for the effective electronic lattice the operators @xmath235 , @xmath236 , @xmath237 , and @xmath238 count the number of rotated - electron doubly - occupied sites , empty sites , spin - down singly - occupied sites , and spin - up singly - occupied sites , respectively
each site of the effective electronic lattice can either be doubly occupied by two rotated electrons of opposite spin projection , empty , singly occupied by a spin - down rotated electron , and singly occupied by an spin - up rotated electron .
the result of ref .
@xcite that the @xmath54 holons are zero spin - singlet combinations of two electrons of opposite spin projection whereas the @xmath137 holons are two - electronic hole quantum objects , is consistent with the @xmath54 holons and the @xmath137 holons corresponding to rotated - electron doubly - occupied and empty sites of the effective electronic lattice , respectively .
in the companion paper @xcite it is found that the effective electronic lattice introduced here is related to a set of effective pseudoparticle lattices .
the concept of effective pseudoparticle lattice introduced in that reference arises from a local description of the pseudoparticles , alternative to the band - momentum @xmath239 description extracted directly from the bethe - ansatz solution in the companion paper @xcite .
while in the limit @xmath68 such a local description of the pseudoparticles refers to the real - space lattice , in the case of finite values of @xmath101 the local pseudoparticle character refers to the effective electronic lattice , as discussed in ref .
@xcite . in that reference
the local pseudoparticles are described in terms of rotated - electron site distribution configurations .
the concepts of a effective pseudoparticle elattice and of a local pseudoparticle introduced in the companion paper @xcite are used in the studies of the few - electron spectral properties of the model presented in refs .
the commutation relations ( 60)-(62 ) of the companion paper @xcite imply that the six generators of the @xmath0-spin and spin @xmath27 algebras given in eqs .
( 2 ) , ( 9 ) , and ( 10 ) of the same paper and the momentum operator ( [ popel ] ) have the same expressions both in terms of elementary electronic operators @xmath63 and @xmath17 and elementary rotated - electron operators @xmath62 and @xmath240 , respectively . thus all these seven operators are both two - electron and two - rotated - electron operators .
these two - electron operators are such that all operator terms of the second expression of eq .
( [ ontil ] ) vanish except the first term .
thus the operator ( [ dontil ] ) associated with such operators vanishes .
this result suggests that application of an operator of the form ( [ dontil ] ) associated with a general @xmath216-electron operator onto a ground state leads to very little electronic spectral weight .
the latter spectral weight vanishes in the particular case that the @xmath79-electron operator commutes with the electron - rotated - electron unitary operator , but is expected to be extremely small otherwise , at least for few - electron operators .
the reason is that such general @xmath79-electron and @xmath79-rotated - electron operators are products of the same elementary operators @xmath241 and @xmath242 , respectively , as the above two - electron operators .
application of any of the above seven two - electron operators and of the corresponding two - rotated - electron operators onto a ground state leads to the same final states .
other two - electron or few - electron operators have slightly different expressions in terms of the same elementary electronic operators .
thus it is expected that the application of any other few - electron operator onto a ground state leads to almost the same final states as application of the corresponding few - rotated - electron operator on the same ground state .
this is equivalent to say that application of the operator ( [ dontil ] ) associated with such operators onto a ground state leads to almost no spectral weight .
this property is a direct result of the unitary and canonical character of the electron - rotated - electron transformation .
this prediction is confirmed by the quantitative results obtained below .
indeed , we find that the dominant microscopic physical processes that control the @xmath79-electron spectral - weight distribution of the excitation @xmath243 are associated with the transformation laws of the @xmath79-rotated - electron operator of the second expression of eq .
( [ ontil ] ) . indeed , in the case @xmath153 , over 99% percent of the spectral weight of @xmath243 is found to be contained in @xmath244 .
we start by studying the specific form that the exact holon and spinon selection rules associated with the inequalities ( [ mc - range])-([ms+range ] ) take for transitions generated by application onto a ground state of a general @xmath216-rotated - electron operator .
while these ground - state selection rules refer to @xmath79-rotated - electron operators , spectral functions of physical interest are of the form ( [ onsf])-([onmsf ] ) and ( [ onsfb])-([onmsfb ] ) and involve @xmath79-electron operators . therefore , the main subject of this section is the study and discussion of the consequences of the @xmath79-rotated - electron ground - state selection rules on the @xmath79-electron spectral - weight properties .
we find that the holon and spinon deviation number restrictions of the inequalities ( [ mc - range])-([ms+range ] ) correspond in the case of excitations generated by @xmath79-electron operators to the dominant holon and spinon microscopic physical processes that control the electronic spectral properties . in the quantitative numerical studies presented below
we devote particular attention to the basic @xmath153 electron and rotated - electron operators .
we provide evidence that the general results obtained for one - electron operators also apply to @xmath245 and other few - electron operators . the inequalities ( [ mc - range])-([ms+range ] ) provide the largest domains for permitted values of @xmath53 holon and @xmath246 spinon number deviations generated by application of @xmath216-rotated - electron operators @xmath158 .
according to eq .
( [ d ] ) the number @xmath216 is given by @xmath247 , where the numbers @xmath248 , @xmath249 , @xmath250 , and @xmath251 are defined below such an equation . the deviation permitted values defined by inequalities ( [ mc - range])-([ms+range ] ) are determined uniquely by the values of the four numbers @xmath74 specific to the operator @xmath158 and do not depend on the values of @xmath101 , electronic density @xmath46 , and spin density @xmath47 .
moreover , these inequalities provide the largest domains for permitted deviation values for all values of @xmath101 , electronic densities @xmath96 and @xmath252 and spin densities @xmath253 and @xmath254 , respectively .
however , the domains of such a permitted values can be smaller .
let @xmath255 be the numbers of @xmath246 holons ( @xmath194 ) and @xmath53 spinons ( @xmath195 ) of the initial state .
for instance , in case of deviations which do not obey the following inequalities , @xmath256 the corresponding domains of deviation values are smaller and are contained in those defined by inequalities ( [ mc - range])-([ms+range ] ) . for simplicity in this section
we consider initial ground states corresponding to electronic densities @xmath96 and spin densities @xmath257 . as discussed in the companion paper @xcite
, one can generate from states with electronic densities @xmath96 and spin densities @xmath257 the corresponding towers of states with electronic densities and spin densities belonging to the above extended domains .
this is reached by repetitive application of the off - diagonal generators of the @xmath0-spin and spin algebras given in eqs .
( 9 ) and ( 10 ) of ref .
@xcite , respectively , onto the former states . in the particular case when the initial state is a ground state @xmath258 with values of the density and spin density belonging to the above domains
, the state @xmath259 belongs to an electron ensemble space with electron and rotated - electron numbers , @xmath260 here @xmath261 , @xmath262 , and @xmath263 are the ground - state initial electron - ensemble space electron and rotated - electron numbers and the deviations are given by the general expressions of eq .
( [ nupdodll ] ) .
( we recall that the spin @xmath12 electron numbers remain invariant under the electron - rotated - electron unitary transformation and thus the number of spin @xmath12 electrons equals the number of spin @xmath12 rotated electrons . )
let us first disregard the selection rules associated with the inequality ( [ mc - range ] ) . in this case
the use of eq .
( [ nupdodll ] ) and eqs .
( 49 ) and ( 50 ) of the companion paper @xcite reveals that the deviations leading to the set of final states whose @xmath2-pseudoparticle , @xmath54 holon , and @xmath54 spinon numbers are compatible with the state @xmath264 generated by application onto the ground state of a general @xmath79-rotated - electron operator are such that , @xmath265 the corresponding possible values of the @xmath54 spinon number deviations @xmath266 are determined by the value of the @xmath54 holon number deviation @xmath267 and are given by , @xmath268 for given initial values of the spin @xmath12 electron and rotated - electron numbers , only one of the two deviations @xmath269 and @xmath266 is independent .
( [ dommc ] ) reveals that the number of different cphs ensemble spaces spanned by states whose electron and rotated - electron numbers are the same as the ones of the state @xmath244 is @xmath270 , with @xmath271 in the present thermodynamic limit .
this is confirmed by eqs .
( [ dommc])-([dmsdmc ] ) . according to the expressions given in eq .
( c24 ) of the companion paper @xcite the number of @xmath54 holons vanishes in the case of a ground state and thus the inequality @xmath272 given in eq .
( [ inhs ] ) is not met except when @xmath273 .
this shortens the domain of the deviation values relative to the largest permitted domains given in inequalities ( [ mc - range])-([ms+range ] ) .
when the initial state is the ground state the @xmath54-holon , @xmath137-holon , @xmath54-spinon , and @xmath137-spinon number deviations are restricted to the ranges of the following inequalities , @xmath274 @xmath275 @xmath276 and @xmath277 thus in the case of the @xmath54 holon deviations @xmath269 only the following integer values are permitted , @xmath278 here we used eq .
( 30 ) of the companion paper @xcite to express the deviation @xmath267 in terms of the deviations in the numbers of @xmath54 yang holons and @xmath279 pseudoparticles .
we emphasize that the exact ground - state charge selection rule ( [ src ] ) also limits the value of the deviations @xmath280 in the number of @xmath2 pseudoparticle holes .
indeed such a selection rule is equivalent to the following inequality , @xmath281 note that this inequality is a particular case for an initial ground state of the general @xmath280 inequality ( [ dnc ] ) . in general the ground - state selection rule ( [ src ] )
also implies that the maximum permitted values of the deviations @xmath282 and @xmath283 are @xmath284 and @xmath285 , respectively , provided that the condition @xmath286 is respected .
however , we note that the value @xmath287 can be reached only if the inequalities ( i ) @xmath288 and ( ii ) @xmath289 are met . inequality ( i )
follows from the requirement that generation of a @xmath279 pseudoparticle involves combination of the @xmath290 new created @xmath54 holons with a number @xmath290 of @xmath137 holons @xcite .
the number @xmath291 gives the number of these @xmath137 holons pre - existing in the initial state and @xmath292 gives the number of @xmath137 holons created by annihilation of rotated electrons singly occupying sites in the initial state .
inequality ( ii ) results from two requirements : first the maximum value of @xmath283 is reached when each new created spin - down and/or spin - up rotated electron combines with a spin - up and/or spin - down rotated electron pre - existing in rotated - electron singly occupied sites of the initial state .
this requires that @xmath293 should be larger than or equal to @xmath294 ; second at least @xmath295 of spin @xmath296 rotated electrons singly occupying sites of the initial state should be annihilated and give rise to empty sites associated with the new created @xmath137 holons .
this requires the presence of the extra term @xmath297 on the right - hand side of the above inequality ( ii ) .
for instance , for the half - filling @xmath298 ground state the values of the @xmath279 pseudoparticle number deviations are restricted by the inequality @xmath299 .
however , there might exist other selection rules which further restrict the values of these @xmath279 pseudoparticle deviations .
recent numerical results @xcite reveal that in the case of the half - filling @xmath298 initial ground state , the state generated by application of the @xmath245 charge operator contains no @xmath300 pseudoparticles .
this suggests that there is another selection rule which states that generation of @xmath279 pseudoparticles by new created rotated electrons results from combination of the @xmath54 holons generated by creation of these rotated electrons with pre - existing @xmath137 yang holons in the initial ground state .
since for the half - filling initial ground state we have that @xmath298 , the numerical results of ref .
@xcite suggest that in the case of such an initial ground state @xmath301 and thus eq . ( [ src ] ) simplifies to @xmath302 .
moreover , if such a selection rule is valid for all @xmath79-rotated - electron operators , the maximum permitted value of the deviation @xmath283 becomes @xmath303 .
this gives zero in the case of the half - filling initial ground state , in agreement with the numerical results of ref .
@xcite .
the number @xmath290 equals the length of the ideal charge takahashi s string excitations associated with the @xmath279 pseudoparticle branch @xcite .
we emphasize that an infinite number of the matrix elements between the initial ground state and excited states of the lehmann representation of the @xmath79-rotated - electron spectral function of the operator @xmath158 vanishes exactly as a consequence of the ground - state charge selection rule associated with eq .
( [ src ] ) .
moreover , we find later in this section that the exact @xmath79-rotated - electron rules correspond to the dominant holon and spinon microscopic physical processes generated by application of the corresponding @xmath79-electron operator onto the ground state .
for finite values of @xmath101 less than 1% of the @xmath79-electron spectral weight does not correspond to the final states associated with deviation values obeying the selection rule ( [ src ] ) .
such a small amount of spectral weight corresponds to a few extra final states with deviation values outside the ranges given in eq .
( [ src ] ) .
however , it is found in ref .
@xcite that the absolute value of the matrix elements of the spectral functions ( [ onsf])-([onmsf ] ) and ( [ onsfb])-([onmsfb ] ) decreases rapidly as the number of holon and spinon processes needed for generation of the corresponding excited states increases .
it follows that also for the @xmath79-electron operator an infinite number of possible final states give no measurable contribution to the @xmath79-electron spectral weight .
this property is related to the exact ground - state selection rule ( [ src ] ) for the states generated by the corresponding @xmath79-rotated - electron operator . in the case of @xmath79-rotated - electron operators , the absolute maximum value of the permitted values given in eq .
( [ src ] ) is given by @xmath79 .
the value @xmath304 is reached when the operator @xmath155 can be written as the product of @xmath79 creation rotated - electron operators .
analysis of eq .
( [ src ] ) reveals that the general @xmath79-rotated - electron ( and @xmath79-electron ) excitation @xmath264 has quantum overlap ( and @xmath305 has significant quantum overlap ) only for energy eigenstates described by charge takahashi s ideal string excitations of length @xmath306 . in the particular case when @xmath155 can be expressed as a product of @xmath216 creation rotated - electron operators
this gives @xmath307 .
the number @xmath79 is thus the maximum absolute value for the length of such a charge ideal string excitation that can be generated by application onto the ground state of a general @xmath79-rotated - electron operator . in the case of the excitation generated by application onto the same state of the corresponding @xmath79-electron operator more than 99% of the spectral weight
is exhausted by excited states containing string excitations of length @xmath306 , as confirmed below for one - electron operators .
this also gives @xmath308 if the @xmath79-electron operator can be written as a product of @xmath216 creation electronic operators .
when the initial state is a ground state the permitted values of the @xmath54 spinon deviation @xmath266 generated by application of the ground state of a @xmath79-rotated - electron operator are given by eq .
( [ dmsdmc ] ) , with @xmath269 restricted to the values of eq .
( [ src ] ) .
this leads to a domain of values obeying the inequality ( [ gsms - range ] ) .
however , this exact spinon selection rule does not restrict the length @xmath290 of the spin takahashi s ideal string excitations which have quantum overlap with @xmath79-rotated - electron excitations @xmath244 .
this is because according to the expressions given in eq .
( c25 ) of the companion paper @xcite , the ground state occupancy of @xmath54 spinons is finite and given by @xmath309 .
thus one can generate both @xmath54 hl spinons and @xmath310 pseudoparticles belonging to branches such that @xmath311 by decomposition processes of @xmath312 pseudoparticles which do not change the net number of @xmath54 spinons .
we note that the exact spinon selection rules just limit the value of the deviations of such a net number and do not limit the processes which conserve its value .
the exact selection rule ( [ src ] ) is equivalent to the inequality ( [ gsc - range ] ) .
this latter inequality reveals that the deviation in the number of @xmath2 pseudoparticle holes generated by application onto a ground state of a rotated - electron operator is limited by an exact selection rule .
based on the expressions provided in the companion paper @xcite , one can express the number of @xmath2 pseudoparticle holes and the number of @xmath312 pseudoparticle holes in terms of the values of @xmath0-spin @xmath50 and spin @xmath51 as follows , @xmath313 and @xmath314 respectively .
based on the symmetries between the charge and spin sectors of the 1d hubbard model , one would expect that the values of the deviation in the number of @xmath312 pseudoparticle holes generated by application onto a ground state of a rotated - electron operator should also be limited .
indeed , below we find numerical evidence that within the permitted final states which obey the selection rules ( [ src ] ) and ( [ gsc - range ] ) , there is a sub class of states that describe over 94% of the few - rotated - electron spectral weight and whose deviation values in the number of @xmath312 pseudoparticle holes are such that , @xmath315 note that the exact selection rule ( [ src ] ) limits the number of @xmath54 yang holons and @xmath279 pseudoparticles generated by application onto a ground state of a @xmath79-rotated - electron operator @xmath155 .
also the states whose deviation values obey eq .
( [ gsnhs1-range ] ) have the following restrictions in the number of @xmath54 hl spinons and @xmath310 pseudoparticles belonging to @xmath316 branches , @xmath317 here the quantity @xmath318 is given by , @xmath319 note that the value of this number vanishes for the ground state . equations ( [ gsnhs1-range ] ) and ( [ srs ] ) contain the same information .
let us consider the specific case of one rotated - electron removal and addition . in this case the states obeying the deviation value restrictions of eqs .
( [ gsnhs1-range ] ) and ( [ srs ] ) are such that @xmath320 .
since the rotated - electron spectral weight distributions are independent of the value of @xmath101 and as @xmath68 rotated electrons equal electrons , we can find the latter distributions by evaluating the electron spectral weight for @xmath68 . by use of the method reported in ref .
@xcite , we find that as the number of sites @xmath321 increases the one @xmath312 pseudoparticle hole excitations have for one - electron removal and addition and @xmath68 the relative weights provided in the table below .
we expect that for @xmath35 the value of the relative weights for one - electron removal and addition given in the table for smaller systems are as @xmath68 above 98% and 94% , respectively .
the numbers provided in the table seem to confirm that the deviation value restrictions of eqs .
( [ gsnhs1-range ] ) and ( [ srs ] ) refer to processes that lead to a substantial part of the rotated - electron spectral weight associated with the permitted states whose deviation values obey the exact selection rule ( [ src ] ) .
the @xmath68 results of ref .
@xcite were derived by a scheme where the spin degrees of freedom of the 1d hubbard model were described by a heisenberg isotropic spin model .
within such a scheme the one - electron excitations were in part simulated by a procedure involving a change in the number of sites of the heisenberg chain .
this procedure has similarities with the one used in refs .
@xcite . the deviation value restrictions of eq . (
[ srs ] ) do not correspond to an exact selection rule but are behind the result considered surprising in ref .
@xcite that for @xmath68 , more than 97% and more than 99% of the spectral weight associated with creation of an electron and annihilation of an electron , respectively , is found on the spin branch of faddeev and takhtajan .
this branch dispersion refers to the part of the one - rotated - electron spectral weight associated with the spin degrees of freedom .
it is well known from the bethe - ansatz solution that the holes in the @xmath2 and @xmath312 pseudoparticle bands play a key role in the energy spectrum of the charge and spin excitations of the 1d hubbard model , respectively .
note that the ground - state selection rules ( [ src ] ) and deviation value restrictions ( [ srs ] ) can be expressed in terms of the deviation values of the numbers of @xmath2 and @xmath312 pseudoparticle holes through the inequalities ( [ gsc - range ] ) and ( [ gsnhs1-range ] ) , respectively .
when @xmath322 is the one - rotated - electron creation or annihilation operator , according to the ground state exact selection rule ( [ gsc - range ] ) and the deviation value restrictions of eq .
( [ gsnhs1-range ] ) , the maximum number of holes created in the @xmath2 pseudoparticle and @xmath312 pseudoparticle bands respectively is one .
thus if the initial ground state has zero spin density the single @xmath312 pseudoparticle hole indeed corresponds to the well known spin branch dispersion of faddeev and takhtajan @xcite .
similar results are expected to hold for excitations generated by other few rotated - electron operators .
the numbers provided in the table were obtained by numerical simulations and seem to confirm that , in contrast to eq .
( [ src ] ) , eq .
( [ srs ] ) does not correspond to an exact selection rule for deviations generated by rotated - electron operators .
note that the maximum value on the right - hand side of eq .
( [ srs ] ) is the same as the corresponding @xmath266 maximum value of the general @xmath54 spinon number deviation inequality ( [ ms - range ] ) .
.the relative weight of the one @xmath312 pseudoparticle hole contributions in the @xmath68 limit for finite size systems of @xmath323 electrons .
this also gives the same relative weight in the case of excitations generated by one - rotated - electron operators for any value of @xmath101 . [ cols=">,^,^ " , ] here we confirm that the ground - state charge selection rule ( [ src ] ) for deviations generated by @xmath79-rotated - electron operators also defines the dominant holon and spinon microscopic processes that control the spectral properties of the corresponding @xmath79-electron operator .
the dominant processes correspond to 99.75% - 100.00% of the whole electronic spectral weight and are associated with states whose deviation values generated by application of a @xmath79-electron operator onto a ground state obey eq .
( [ src ] ) . within such dominant processes , there is a sub - group of simple holon and spinon processes that correspond to about 94.00% of the electronic spectral weight . these latter processes are associated with states whose deviation values generated by application of a @xmath79-electron operator onto a ground state obey both eqs .
( [ src ] ) and ( [ srs ] ) .
( we recall that the deviation value restrictions of eq .
( [ srs ] ) are not exact selection rules for rotated - electron operators . ) we start by considering the specific case of one - electron addition .
we recall that according to the results of the companion paper i , the number of holons @xmath324 and the number of spinons @xmath325 are such that @xmath326 $ ] and @xmath327 , respectively . for simplicity ,
let us assume that the initial ground state has zero spin density . in this case
the spectral - weight distribution associated with creation of a spin - up electron has the same form as the one associated with creation of a spin - down electron .
here we consider the former case .
the spin - up electron operator is related by eq .
( [ ontil ] ) to the spin - up rotated electron .
thus in the present one - electron case the selection rule ( [ src ] ) and the deviation value restrictions given in eq . ( [ srs ] ) refer to creation of a spin - up rotated electron . from the use of expressions ( 46 ) , ( 47 ) , ( 49 ) , and ( 50 ) of the companion paper @xcite for the electron and rotated - electron numbers in terms of the holon , spinon , and @xmath2 pseudoparticle numbers , we find that in the case of creation of a spin - up rotated electron the following transitions are permitted by the selection rule ( [ src ] ) and by the equivalent inequality ( [ gsc - range ] ) : \(i ) transitions such that @xmath328 , @xmath329 , @xmath330 , @xmath331 , and @xmath332 .
the minimal excitation energy for such transitions is zero . within these general transitions , the transitions that also obey the restrictions ( [ gsnhs1-range ] ) and ( [ srs ] ) are such that @xmath333 and @xmath334 .
\(ii ) transitions such that @xmath335 , @xmath336 , @xmath337 , @xmath338 , and @xmath339 . the minimal excitation energy for such transitions is @xmath119 . within these general transitions ,
the transitions that also obey the restrictions ( [ gsnhs1-range ] ) and ( [ srs ] ) are such that @xmath340 and @xmath334 . in the case of creation of a spin - up
rotated electron the selection rule ( [ src ] ) imposes that @xmath341 and thus that @xmath342 .
note that the general transitions ( i ) and ( ii ) obey such restrictions . within these general transitions ,
the transitions that also obey the restrictions ( [ srs ] ) are such that @xmath343=0 $ ] and thus that @xmath344 .
the simplest non - permitted transitions involve creation of three holes in the @xmath2 pseudoparticle band and of three holons : \(iii ) transitions such that @xmath345 , @xmath346 , @xmath347 , @xmath348 , and @xmath349 .
the minimal excitation energy for such transitions is @xmath350 .
these transitions are not permitted by the rule ( [ src ] ) because @xmath351 . within these general transitions ,
the transitions that obey the restrictions ( [ gsnhs1-range ] ) and ( [ srs ] ) are such that @xmath352 and @xmath353 . within the permitted transitions of type ( ii ) , the simplest transitions that do not obey the restrictions ( [ gsnhs1-range ] ) and ( [ srs ] ) involve creation of three holes in the @xmath312 pseudoparticle band : ( ii ) transitions such that @xmath335 , @xmath233 , @xmath337 , @xmath354 , and @xmath339 and with @xmath355 , @xmath356 .
the minimal excitation energy for such transitions is @xmath119 .
these transitions do not obey the restrictions ( [ srs ] ) because @xmath357=\delta l_{s,\,-1/2}=1>0 $ ] . in the case of half filling we often shift the ground - state zero - energy level to the middle of the mott - hubbard gap .
in that case the above minimal excitation energies @xmath358 , @xmath119 , and @xmath350 become @xmath358 , @xmath359 , and @xmath360 , respectively , where @xmath361 at @xmath362 . below
we call _ three - holon states _ the final states associated with the transitions ( iii ) because they involve creation of three holons . these states also involve creation of three holes in the @xmath2 pseudoparticle band .
moreover , we call _ three-@xmath312-hole states _ the final states behind transitions ( ii ) , which involve creation of three holes in the @xmath312 pseudoparticle band . in contrast , the final states associated with both the transitions ( i ) and ( ii ) are one - holon and one @xmath312-hole states .
let us now find out what the relative weight of the states ( iii ) ( or states ( iii ) and ( ii ) ) is relative to the weight of the states ( i ) , ( ii ) , and ( ii ) ( or states ( i ) and ( ii ) ) when all these states are generated by application onto the ground state of the spin - up electron creation operator . to assess the importance of the three - holon final states ( iii ) and three-@xmath312-hole states ( ii )
, first we turn to exact diagonalization of small chains .
the small - chain results are not expected to be a good approximation for the evaluation of the thermodynamic - limit weight distribution if we consider the weight associated with each specific final state
. however , here we are mostly interested in the relative spectral - weight sum rules of the permitted states ( i ) , ( ii ) , and ( ii ) versus the _ forbidden _ states ( iii ) .
moreover , we also want to know the relative weight of the states ( ii ) .
fortunately , we find below that in the case of such sum rules , which involve contributions from a whole class of states , the small - chain results provide values for the relative weights which agree up to 99% with the corresponding thermodynamic - limit values .
the full electron addition and removal spectrum for six sites with six electrons ( half filling ) is shown in fig .
[ fig : aw2 ] for a relatively large value of @xmath363 . as mentioned above , in the case of half
filling we define the fermi level in the middle of the mott - hubbard gap .
the hubbard bands at @xmath364 and @xmath365 are well separated , and the weights at @xmath366 are orders of magnitude smaller then the contribution of the main hubbard bands , centered around @xmath364 .
the states centered around @xmath360 are three - holon states of type ( iii ) . as a result of the half - filling particle - hole symmetry
, there is a corresponding structure for electron removal centered around @xmath367 . in the large-@xmath101 limit the latter structure
is associated with creation of two real - space empty sites . for the general @xmath101 case
such a structure results from creation of two rotated - electron empty sites .
we recall that the @xmath146 summations of the few - electron spectral functions ( [ onsfb])-([onmsfb ] ) refer to such a number of rotated - electron empty sites . in fig .
[ fig : sumrule ] we have plotted the contribution of different final states to the sum rule in the case of half filling . for that reason , we have followed adiabatically the weights ( matrix elements ) of different states as we reduced @xmath101 ( which in the present case we did for this relatively small system ) , and summed the weight over the particular family of states . as we can see , the contribution of the three - holon final states ( iii ) to the total sum rule is largest at intermediate values of @xmath368 , and it does not exceed 0.25% in the total sum rule .
for large values of @xmath101 it decreases as @xmath369 , while for small values as @xmath370 .
the three-@xmath312-hole contribution from the states ( iv ) is also less than 0.25 % , and for small values of @xmath101 it goes as @xmath371 .
we find below that the relative spectral weight of the final states ( iii ) decreases with decreasing density .
for instance , at quarter filling such a weight is 2% of that of half filling and vanishes in the limit of vanishing electronic density .
thus at quarter filling and @xmath223 the final states ( i ) , ( ii ) , and ( ii ) correspond to @xmath372 99.99% of the total spectral weight and the states ( iii ) to @xmath372 0.005% of such a weight . on the other hand , we know from the values of the above table that the relative weight of the states ( ii ) increases with increasing values of @xmath101 . as we confirm below ,
this is not so for the states ( iii ) , that remain having very little spectral weight as @xmath35 . ) and removal ( @xmath373 ) spectrum for the half - filled six - site ring , with @xmath374 .
note the logarithmic scale for the weights .
not shown are the five - holon states , whose contributions are extremely small and energies are out of the shown energy window .
[ fig : aw2 ] ] pseudoparticle hole excitations corresponding to the above states ( i ) and ( ii ) . for larger systems
this remains true if we consider the above states ( i ) , ( ii ) , and ( ii ) , which are permitted states for the corresponding rotated - electron problem .
[ fig : sumrule ] ] while we could nt do finite size analysis due to lack of larger system sizes for arbitrary value of @xmath3 , for large @xmath101 we have another method to calculate the contribution of the three - holon states ( iii ) . when @xmath101 is large , it is possible to achieve a more precise statement through the systematic @xmath125 expansion of the electron - rotated - electron unitary operator @xcite
. the electron - rotated - electron canonical transformation maps the hamiltonian onto a rotated hamiltonian .
such a transformation applies to all operators , as given in eq .
( [ otil ] ) .
for instance , it also relates the elementary rotated electron operators to the elementary electron operators when calculating the spectral functions .
so the creation operator according to the expansions ( [ m ] ) and ( [ mtil ] ) can be decomposed into , @xmath375 where the index @xmath87 in @xmath376 refers to the change of rotated - electron double occupation .
although in the present large @xmath101 limit one has that @xmath377 , we use the general notation @xmath116 for labelling each hubbard band .
the operator @xmath378 adds an electron to an unoccupied site . on the other hand
, @xmath379 adds an electron to a site already occupied by an electron of opposite spin , thus promoting the state to the subspace with one more doubly occupied site . in the @xmath68 limit
these operators change the number of @xmath2 pseudoparticles and holons accordingly to the deviation values of the above transitions ( i ) and ( ii)-(ii ) , respectively .
the operator involved in the generation of the three - holon states ( iii ) is @xmath380 .
its first non - vanishing term is proportional to @xmath381 and reads , @xmath382 , { \tilde{c}}^\dagger_{j,\,\sigma } { \tilde{n}}_{j,\,\bar\sigma } \right ] + \frac{1}{2 } \left [ { \tilde{t}}_{e_u } , \left [ { \tilde{t}}_{e_u } , { \tilde{c}}^\dagger_{j,\,\sigma } ( 1-{\tilde{n}}_{j,\,\bar\sigma } ) \right ] \right ] \right),\ ] ] where the rotated kinetic energy operators @xmath383 and @xmath384 are given in eqs .
( [ ttil0 ] ) and ( [ ttil+1 ] ) , respectively . after some algebra
we find , @xmath385 \ , .\end{aligned}\ ] ] the operator only acts when all the @xmath386 , @xmath19 , and @xmath387 sites are singly occupied .
the action of @xmath388 on sites @xmath389 , apart from the @xmath390 coefficient , gives @xmath391 while on any other configuration it will give zero .
here the indexes @xmath392 and @xmath393 stand for doubly - occupied sites and empty sites , respectively , and the @xmath394 and @xmath395 symbols refer to spin - up and spin - down singly - occupied sites , respectively .
we can see that the operator @xmath388 removes three spins with total spin s=1/2 ( for the s=3/2 combination @xmath396 ) .
next , we calculate @xmath397 , the total weight in the _ band _ centered around @xmath398 and of energy width @xmath399 ( each of the @xmath393 and @xmath392 sites contributes @xmath400 to the bandwidth ) .
it is given by the expectation value @xmath401 , where the @xmath402 is the spin - charge factorized wave function @xcite .
first we find that , @xmath403 where the projector @xmath404 ensures the single occupation of site @xmath19 .
replacing the factorized wave function , the partial sum rule reads , @xmath405 let us denote by @xmath406 and @xmath407 the creation and annihilation operators , respectively , of a spin - less fermion at site @xmath19 .
the expectation value to find three neighboring spinless fermions is @xmath408{ccc } \langle f^{\dagger}_{0 } f^{\phantom{\dagger}}_{0 } \rangle & \langle f^{\dagger}_{0 } f^{\phantom{\dagger}}_{1 } \rangle & \langle f^{\dagger}_{0 } f^{\phantom{\dagger}}_{2 } \rangle \\ \langle f^{\dagger}_{1 } f^{\phantom{\dagger}}_{0 } \rangle & \langle f^{\dagger}_{1 } f^{\phantom{\dagger}}_{1 } \rangle & \langle f^{\dagger}_{1 } f^{\phantom{\dagger}}_{2 } \rangle \\ \langle f^{\dagger}_{2 } f^{\phantom{\dagger}}_{0 } \rangle & \langle f^{\dagger}_{2 } f^{\phantom{\dagger}}_{1 } \rangle & \langle f^{\dagger}_{2 } f^{\phantom{\dagger}}_{2 } \rangle \end{array } \right| \nonumber\\ & = & n^3 - \frac{2 n \sin^2(\pi n)}{\pi^2 } + \frac{\sin^2(\pi n ) \sin(2 \pi n)}{\pi^3 } - \frac{n \sin^2(2 \pi n)}{4 \pi^2}\end{aligned}\ ] ] with the limiting behavior , @xmath409{ll } \displaystyle{\frac{4\pi^6}{135}}n^9 & \mbox{if $ n \ll 1 $ ; } \\ & \\ 1 - 3(1-n ) & \mbox{if $ 1-n \ll 1 $ . }
\end{array } \right .
\ , .\ ] ] note that the spectral weigh of these states decreases rapidly away from half filling , _
i.e. _ at quarter filling ( @xmath410 ) it is about 2% of that at half filling .
the expectation values in the thermodynamic limit of the heisenberg model is @xcite , @xmath411 so that @xmath412 the sum rule in the second upper hubbard band defined in sec .
iii is then , @xmath413 for the six - site finite - size cluster , the expectation value in eq .
( [ eq : exps1o2 ] ) is @xmath414 , which is about 1% off from the thermodynamic - limit value .
the asymptotic @xmath415 is shown in fig .
[ fig : sumrule ] as a dashed line .
both our numerical results for general values of @xmath101 and a small system and our analytical large @xmath101 results for the thermodynamic limit confirm that the ground - state selection rule ( [ src ] ) which limits the values of the deviations generated by rotated - electron operators defines the dominant holon and spinon microscopic processes that control the spectral - weight distribution of the corresponding electronic operators . while for @xmath222 and @xmath221
such a selection rule is exact also for these electronic operators , we find that the relative spectral weight of the _ permitted _ above states ( i ) , ( ii ) , and ( ii ) is minimum for @xmath223 . this _ minimum _ value decreases with decreasing density . for half filling it is given by @xmath372 99.75% , whereas for quarter filling it reads @xmath372 99.99% and in the limit of vanishing density it becomes @xmath372 100.00% .
the extremely small amount of missing spectral weight corresponds mainly to the _ forbidden _ three - holon states ( iii ) .
higher order five - holon / five-@xmath2-hole states lead to nearly vanishing spectral weight . while the selection rule ( [ src ] ) is exact for rotated - electron operators , the states obeying the deviation value restrictions of eq .
( [ srs ] ) correspond to rotated - electron processes that generate most spectral weight of the permitted transitions .
although the relative spectral weight of the three-@xmath312-hole states ( ii ) increases with increasing the system length , we expect that for one - electron addition its maximum value is about 6% as @xmath35 for half filling . however , higher order five-@xmath312-hole states lead to nearly vanishing electronic spectral weight .
thus over 99% of the one - electron addition spectral weight corresponds to generation of one holon and one and three @xmath312 pseudoparticle holes .
a similar analysis can be performed for other few - electron operators .
the deviation restrictions of eq .
( [ src ] ) also amount to more than 99% of the spectral weight generated by application onto the ground state of @xmath79-electron operators such that @xmath416 .
importantly , for all few - electron operators these deviation restrictions become exact both for @xmath222 and @xmath221 and refer to the dominant holon and spinon microscopic processes for @xmath223 . the spectral weight associated with excited states
whose deviation values are outside the ranges defined by eq .
( [ src ] ) is for electronic operators in general maximum for half filling , yet it remains small for such an electronic density . as an example we consider the two - electron frequency dependent optical conductivity , which is directly related to the dynamical structure factor .
a preliminary finite - energy study of such a @xmath245 electron problem was presented for the case of the 1d hubbard model in ref .
the ground - state selection rule of eq .
( [ src ] ) implies that in that case the corresponding two - rotated - electron problem the permitted @xmath54 holon number deviations are such that @xmath417 and thus the permitted holon number deviations are @xmath418 , respectively .
the finite - energy absorption considered in that paper results from transitions associated with holon number deviations such that ( @xmath419 ) and ( @xmath420 ) . thus contributions from excited states with holon number deviations such that ( @xmath421 ) are expected to correspond to less than 1% of the two - electron optical conductivity spectral weight . as for the one - electron problem
, it is expected that such an extremely small amount of spectral weight is maximum yet very small at half filling .
such a prediction is confirmed by the results of ref .
there such a finite - energy absorption was investigated for the sine - gordon model .
the absorptions represented for coupling constant @xmath422 by a solid line and a dashed line in fig .
1 of the same reference correspond to one - soliton / one - antisoliton and two - solition / two - antisoliton contributions , respectively .
the latter absorption is very small , being multiplied by @xmath423 in the figure . in the limit of @xmath424
the sine - gordon model acquires a @xmath0-spin @xmath27 symmetry and describes the 1d hubbard model at half filling and small values of @xmath101 .
moreover , in this limit the solitons and antisolitons become @xmath137 holons and @xmath54 holons , respectively . therefore
, the results of that reference are expected to be very similar to those of the half - filling hubbard model . as @xmath424 the above one - soliton / one - antisoliton and two - solition / two - antisoliton transitions correspond to ( @xmath425 ) and ( @xmath426 ) deviations , respectively .
the transitions associated with the latter deviations are forbidden for the corresponding two - rotated - electron operator .
thus in the case of the two - electron operator the contributions from these transitions are expected to amount for less than 1% of the spectral weight .
such a prediction is confirmed by the smallness of the corresponding absorption , which is represented in fig . 1 of ref
similar results hold for other two - electron operators .
we emphasize that equivalent results also hold for one - electron removal .
it is generally accepted that one - electron removal corresponds to final states with a single extra hole in the @xmath2 pseudoparticle band .
however , we find that this general expectation is only true for one - rotated - electron removal . in the case of one - electron removal
less than 1% of the spectral weight goes again to final states with three extra @xmath2 pseudoparticle holes . in the particular case of half filling such final states
lead to the structure of fig .
[ fig : aw2 ] which is centered around @xmath367 .
in this section we summarize and discuss the results obtained in this paper .
moreover , we discuss the application of the concepts and non - perturbative many - electron tools introduced here and in the companion papers @xcite to the accomplishment of a program for the study of few - electron spectral functions at finite values of excitation energy .
such an application program is fulfilled in refs .
@xcite .
the holon , spinon , and @xmath2 pseudoparticle description introduced in the first paper @xcite of this series and further studied in this second and in the third @xcite papers , can be used in the evaluation of matrix elements between the ground state and excited states and in the related problem of the derivation of the line shape for finite - energy few - electron spectral functions @xcite .
the results of the present paper about the dominant microscopic physical processes that amount for more than 99% of the spectral weight generated by application onto the ground state of few - electron operators is useful for the success of such a program .
our theory also describes the higher - order holon and spinon processes associated with the remaining less than 1% of electronic spectral weight .
furthermore , in this paper we used the relation of rotated electrons to the quantum objects whose occupancy configurations describe the energy eigenstates , in the generalization of the concepts of lower hubbard band and upper hubbard bands for all values of the on - site repulsion .
while the lower hubbard band refers to zero rotated - electron double occupation final states , the @xmath93 upper hubbard band corresponds to the spectral weight associated with the excited states of rotated - electron double occupation @xmath87 . for large values of the ratio @xmath101 this definition of lower hubbard band and upper hubbard bands
coincides with the usual one . according to the inequalities given in eq .
( [ dmcsrange ] ) , the maximum number of holons and spinons generated or annihilated by application of @xmath216-rotated - electron operators onto any eigenstate of the spin @xmath12 electron number operator is given by @xmath216 .
in the particular case of excitations whose initial state is the ground state , the ranges provided in eq .
( [ src ] ) refer to an exact selection rule for the deviation values generated by application onto the ground state of such @xmath79-rotated - electron operators .
the occurrence of this selection rule is a direct result of the relation of rotated electrons to the quantum objects whose occupancy configurations describe all energy eigenstates of the model . in this paper we found that the excited states associated with these permitted deviation values correspond to over 99% of the spectral weight generated by application onto the same state of the corresponding @xmath79-electron operators
this means that over 99% of the spectral weight generated by application onto the ground state of a @xmath79-electron operator corresponds to application onto such a state of the first operator term of the second expression of eq .
( [ ontil ] ) .
such an operator term is nothing but the @xmath79-rotated - electron operator associated with the @xmath79-electron operator under consideration .
this interesting result reveals that application onto the ground sate of an operator of the form given in eq .
( [ dontil ] ) leads to very little spectral weight .
the value @xmath216 also determines the absolute maximum value @xmath427 of the quantum number @xmath290 of the @xmath428 pseudoparticle branches which can have finite occupancy in final states generated by @xmath216-rotated - electron operators .
the number @xmath290 is also the length of the ideal charge takahashi s string excitations associated with the @xmath279 pseudoparticle branch .
therefore , the @xmath216-rotated - electron excitations only couple to energy eigenstates described by @xmath2 takahashi s ideal string charge excitations of length @xmath290 such that @xmath308 . for the corresponding @xmath216-electron excitations , the excited states described by @xmath2 takahashi s ideal string charge excitations of length @xmath290 such that @xmath429 amount to less than 1% of the electronic spectral weight .
it follows that the dominant processes generated by application onto the ground state of a @xmath79-electron operator originate a number of @xmath430-holon composite @xmath279 pseudoparticles and @xmath54 yang holons and a number of @xmath2 pseudoparticle holes whose values are within the ranges provided in eqs .
( [ src ] ) and ( [ gsc - range ] ) , respectively .
we recall that for electronic densities @xmath431 there are no @xmath430-holon composite @xmath279 pseudoparticles and no @xmath54 yang holons in the ground state .
higher order processes originating a number of these quantum objects outside the range defined by eqs .
( [ src ] ) and ( [ gsc - range ] ) correspond to less than 1% of the electronic spectral weight .
these processes are generated by application onto the ground state of the operator given in eq .
( [ dontil ] ) . moreover , although the deviation value ranges provided in eqs .
( [ gsnhs1-range ] ) and ( [ srs ] ) are not associated with an exact selection rule for rotated - electron operators , within the dominant processes that include all values for these deviations , they refer to quite important processes for the electronic spectral weight .
these equations refer to ranges for the numbers of generated or annihilated @xmath312 pseudoparticle holes and for the numbers of generated @xmath430-spinon composite @xmath310 pseudoparticles such that @xmath432 and @xmath54 hl spinons , respectively .
we note that for spin densities @xmath433 there are no @xmath430-spinon composite @xmath310 pseudoparticles such that @xmath432 and no @xmath54 hl spinons in the ground state .
for instance , in the case of one - electron addition considered in the previous section , eq .
( [ gsnhs1-range ] ) tells us that these important processes involve generation of one @xmath312 pseudoparticle hole .
we found that these processes lead to about 94% of the electronic spectral weight , whereas excited states involving the creation of three @xmath312 pseudoparticle holes amount for more than 5% of such a spectral weight .
we thus conclude that microscopic processes involving generation of more than three @xmath312 pseudoparticle holes lead to almost no electronic spectral weight .
although in the case of the ideal spin takahashi s string excitations associated with the @xmath310 pseudoparticle branches the ranges of eq .
( [ srs ] ) are not exact , they correspond to a sub - class of excited states that amount to an important part of the electronic spectral weight .
the occurrence of dominant holon and spinon microscopic processes for the electronic spectral weight and how fast the contribution of higher order processes vanishes is related to a property found in ref .
@xcite that plays a central role in the evaluation of the line shape for finite - energy few - electron spectral functions . while the dominant holon and spinon microscopic processes associated with eqs .
( [ src ] ) and ( [ gsc - range ] ) amount for more than 99% of the few - electron spectral weight , the remaining less than 1% of spectral weight is mostly associated with final states involving generation of a few more @xmath279 pseudoparticles , @xmath54 yang holons , and @xmath2 pseudoparticle holes relative to the maximum value of the ranges provided in these equations .
also the contribution to the electronic spectral weight from excited states involving the generation of @xmath310 pseudoparticles such that @xmath432 , @xmath54 hl spinons , and @xmath312 pseudoparticle holes decreases rapidly as the number of these quantum objects increases .
it follows that excited states involving creation of a large number of the above - mentioned quantum objects lead to vanishing spectral weight .
in particular , in the present thermodynamic limit the spectral weight of excited states whose generation involves an infinite number of quantum - object elementary processes vanishes exactly .
thus only processes involving generation of a finite number of quantum objects contribute to the few - electron spectral weight .
this defines the hilbert subspace of few - electron excitations . the pseudofermion description introduced in ref .
@xcite refers to such a hilbert subspace .
the method for evaluation of matrix elements between the ground state and excited states introduced in refs .
@xcite takes into account all possible final states generated by a final number of quantum - object processes , as discussed below .
the holon , spinon , and pseudoparticle description introduced in the companion paper @xcite and further studied here and in the companion paper @xcite is applied in refs .
@xcite to the study of the few - electron spectral functions .
the studies of these references provide the few - electron spectral function line shapes for all values of excitation energy .
the above description is a necessary condition for the successful fulfilment of the program for evaluation of few - electron spectral functions at finite values of the excitation energy .
such a program is fulfilled in refs .
@xcite by application of the new concepts and paradigms introduced in this paper and in its companion papers @xcite as follows : i)- the first step of the evaluation of a few - electron spectral function of the general form ( [ onsf])-([onmsf ] ) or ( [ onsfb])-([onmsfb ] ) is the expression of the corresponding few - electron operator as the expansion in terms of rotated - electron operators given by the second expression of ref .
( [ ontil ] ) ; \ii ) - next , one uses the results of the companion paper @xcite and express the rotated - electron operators in terms of local pseudoparticles operators , yang - holon operators , and hl - spinon operators .
note that after use of the relation ( [ ontil ] ) , the problem of the evaluation of the matrix elements of the few - electron spectral functions ( [ onmsf ] ) and ( [ onmsfb ] ) is equivalent to the computation of matrix elements involving the elementary creation and annihilation operators of these quantum objects ; \iii ) - however , rather than in terms of the interacting pseudoparticles , it is more appropriate to use in the evaluation of such matrix elements the related non - interacting operational pseudofermion description introduced in ref .
fortunately , the few - electron spectral functions can be expressed as a convolution of the pseudofermion spectral functions corresponding to each pseudofermion branch with finite occupancy in the excited states @xmath90 and @xmath434 of the spectral - function expressions ( [ onsf])-([onmsf ] ) and ( [ onsfb])-([onmsfb ] ) , respectively ; \iv ) - finally , the evaluation of the spectral function for each pseudofermion branch with finite occupancy for these excited states uses the method applied to @xmath68 spin - less fermion operators in ref .
the energy spectra of the above excited states appearing in the spectral - function expressions ( [ onmsf ] ) and ( [ onmsfb ] ) is obtained by direct use of the bethe - ansatz solution and @xmath0-spin and spin symmetries .
the state summations of these spectral - function expressions is simplified by the fact that the absolute value of the matrix elements vanishes rapidly as the number of pseudofermion processes involved in the generation from the ground state of the corresponding excited states increases .
such a rapid vanishing is an important property for the evaluation of the finite - energy few - electron spectral functions .
it is directly related to the occurrence of dominant quantum - object microscopic processes .
the use of the relation ( [ ontil ] ) in the evaluation of finite - energy few - electron spectral functions @xcite is a consequence of the breakthrough of the companion paper @xcite concerning the following two issues : first , the clarification of the relation of rotated electrons to the bethe - ansatz quantum numbers that label the energy eigenstates ; second , the association of these quantum numbers with objects whose occupancy configurations describe the energy eigenstates .
the occurrence of dominant processes found in this paper is also important for the studies of refs .
in addition to contributing to the understanding of the non - perturbative microscopic physical mechanisms that control the few - electron spectral properties , our results are related to the rapid vanishing of the matrix elements as the number of processes that generate the corresponding excited states increases , as discussed above .
the complexity of the finite - energy one - electron and two - electron physics of the non - perturbative 1d hubbard model explains why except for @xmath68 investigations @xcite , there have not been many previous studies about this interesting problem . in this paper
we found that excited states generated from the ground state by microscopic processes leading to deviation values in the ranges defined by eq .
( [ src ] ) lead to more than 99% of the few - electron spectral weight .
moreover , a careful comparison of the line shape of the one - electron spectral weight for the trivial @xmath435 limit with the one obtained in refs .
@xcite for the limit @xmath68 , seems to indicate that most spectral weight is located in the vicinity of _ pseudoparticle branch lines_. such lines are generated by a set of final states where one pseudoparticle or pseudoparticle hole is created for all its available band - momentum values @xmath239 and all remaining pseudoparticles or pseudoparticle holes are created at their _
fermi points_. in addition to these elementary processes , the line shape in the vicinity of these branch lines is for electronic and spin densities within the ranges @xmath83 and @xmath436 , respectively , generated by pseudoparticle - pseudoparticle hole processes in the @xmath2 and @xmath312 bands . the one - electron spectral - weight distribution obtained in ref .
@xcite by numerical simulations refers to intermediate values of the ratio @xmath101 such that @xmath437 and displays charge and spin branch lines , in agreement with our prediction .
furthermore , this expectation is confirmed for both one - electron and few - electron spectral functions by the exact results of ref .
@xcite .
consideration of all possible branch lines constructed in this way for the dominant excited states whose deviation values obey the ranges defined by eqs .
( [ src ] ) and ( [ srs ] ) for a given few - electron operator is expected to describe the main spectral lines observed in real quasi - one - dimensional materials .
this expectation is confirmed in the case of the organic conductor ttf - tcnq by the preliminary application of our results presented in ref .
the separated one - electron charge and spin spectral lines observed in ttf - tcnq by photoemission studies agree for the whole energy band width with the theoretical branch lines constructed by the above described procedure .
these preliminary results confirm the interest for the study of the unusual finite - energy spectral properties of these materials of the concepts introduced in this paper . elsewhere
the powerful method introduced in refs .
@xcite is used in a more detailed description of the line shape observed in ttf - tcnq @xcite .
while the preliminary results of ref .
@xcite provide the form of the charge and spin branch lines , the detailed dependence on the excitation energy @xmath86 of the line shape in the vicinity of these lines is presented in ref .
fortunately , this more detailed study also agrees with the line shape observed in the real experiment . elsewhere
the method constructed in refs .
@xcite by means of the concepts and non - perturbative many - electron theoretical tools introduced here and in the companion papers @xcite is applied to the study of the two - electron dynamical structure factor for all values of @xmath101 @xcite . moreover
, a study of the phase diagram of a system of weakly coupled hubbard chains which combines our 1d results with a renormalization group scheme will be also implemented @xcite .
the results of these studies describe many of the anomalous spectral properties observed in real low - dimensional materials .
in particular , they lead to spectral lines similar to the ones observed in finite - energy / frequency experimental investigations of the one - electron spectral weight distribution or organic metals @xcite and two - electron dynamical structure factor of mott - hubbard insulators @xcite .
our theoretical predictions also describe and successfully explain the microscopic mechanisms behind the phase phase diagram observed in quasi - one - dimensional materials @xcite .
finally , the further application of the results obtained in this paper and in its companion paper @xcite to the explicit evaluation of finite - energy few - electron spectral functions requires the introduction of the concepts of _ local pseudoparticle _ and _ effective pseudoparticle lattice_. the introduction of these concepts is the main goal of the third and last paper of this series , ref .
@xcite .
we thank jim w. allen , antnio castro neto , ralph claessen , francisco ( paco ) guinea , katrina hibberd , peter horsch , joo lopes dos santos , lus miguel martelo , and pedro sacramento for stimulating discussions .
we also thank the support of eu money center of excellence ica1-ct-2000 - 70029 and one of us ( k.p . )
thanks the support of otka grants d032689 , t038162 , and the bolyai fellowship .
j. m. p. carmelo , k. penc , and j. m. romn , submitted for publication in phys .
b ( 2002 ) [ cond - mat/0302044 ] .
j. m. p. carmelo , submitted for publication in phys .
b ( 2002 ) [ cond - mat/0305347 ] . elliott h. lieb and f. y. wu , phys .
lett . * 20 * , 1445 ( 1968 ) .
m. takahashi , prog .
. phys . * 47 * , 69 ( 1972 ) .
o. j. heilmann and e. h. lieb , ann . n. y. acad .
sci . * 172 * , 583 ( 1971 ) ; e. h. lieb , phys .
lett . * 62 * , 1201 ( 1989 ) . c. n. yang , phys .
lett . * 63 * , 2144 ( 1989 ) . p. b. ramos and m. j. martins , j. phys .
a * 30 * , l195 ( 1997 ) ; m. j. martins and p.b .
ramos , nucl .
b * 522 * , 413 ( 1998 ) .
j. m. p. carmelo and p. d. sacramento , phys .
b * 67 * ( 2003 ) [ cond - mat/0302042 ] .
m. rasetti , _ the hubbard model , recent results _ , word scientific , singapore ( 1991 ) ; a. montorsi , _ the hubbard model _ , word scientific , singapore ( 1992 ) ; _ the hubbard model its physics and mathematical physics _ , edited by dionys baeriswyl , david k. campbell , jos m. p. carmelo , francisco guinea , and enrique louis ( nato asi series b , vol .
343 , plenum press , new york , 1995 ) .
a. brooks harris and robert v. lange , phys .
rev . * 157 * , 295 ( 1967 ) .
a. h. macdonald , s. m. girvin , and d. yoshioka , phys .
b * 37 * , 9753 ( 1988 ) .
henk eskes and andrzej m. ole , phys .
lett . * 73 * , 1279 ( 1994 ) ; henk eskes , andrzej m. ole , marcel b. j. meinders , and walter stephan , phys .
b * 50 * , 17 980 ( 1994 ) .
f. gebhard , k. bott , m. scheidler , p. thomas , and s. w. koch , phylo .
b * 75 * , 13 ( 1997 ) .
g. beni , t. holstein , and p. pincus , phys .
b * 8 * , 312 ( 1973 ) .
douglas j. klein , phys . rev .
b * 8 * , 3452 ( 1973 ) .
jos carmelo and dionys baeriswyl , phys .
b * 37 * , 7541 ( 1988 ) . f. woynarovich , j. phys .
c * 15 * , 97 ( 1982 ) ; masao ogata and hiroyuki shiba , phys .
b * 41 * , 2326 ( 1990 ) .
s. sorella and a. parola , j. phys .
matter * 4 * , 3589 ( 1992 ) .
r. g. dias and j. m. lopes dos santos , j. de physique i * 2 * , 1889 ( 1992 ) .
karlo penc , frdric mila , and hiroyuki shiba , phys .
* 75 * , 894 ( 1995 ) .
karlo penc , karen hallberg , frdric mila , and hiroyuki shiba , phys .
* 77 * , 1390 ( 1996 ) .
karlo penc , karen hallberg , frdric mila , and hiroyuki shiba , phys .
b * 55 * , 15 475 ( 1997 ) .
d. pines and p. nozires , in _ the theory of quantum liquids _ , ( addison - wesley , redwood city , 1966 and 1989 ) , vol . i. gordon baym and christopher j. pethick , in _ landau fermi - liquid theory concepts and applications _ ,
( john wiley & sons , new york , 1991 ) .
n. e. hussey , m. n. mcbrien , l. balicas , j. s. brooks , s. horii , and h. ikuta , phys .
lett . * 89 * , 086601 ( 2002 ) .
a. menzel , r. beer , and e. bertel , phys .
lett . * 89 * , 076803 ( 2002 ) .
shin - ichi fujimori , akihiro ino , testuo okane , atsushi fujimori , kozo okada , toshio manabe , masahiro yamashita , hideo kishida , and hiroshi okamoto , phys .
lett . * 88 * , 247601 ( 2002 ) .
m. z. hasan , p. a. montano , e. d. isaacs , z .- x .
shen , h. eisaki , s. k. sinha , z. islam , n. motoyama , and s. uchida , phys .
lett . * 88 * , 177403 - 1 ( 2002 ) .
r. claessen , m. sing , u. schwingenschlgl , p. blaha , m. dressel , and c. s. jacobsen , phys .
lett . * 88 * , 096402 ( 2002 ) .
gweon , j. d. denlinger , c. g. olson , h. hochst , j. marcus , and c. schlenker c , physica b * 312 - 313 * , 584 ( 2002 ) ; g .- h .
gweon , j. d. denlinger , j. w. allen , r. claessen , c. g. olson , h. hochst , j. marcus , c. schlenker , and l. f. schneemeyer , j. electron spectrosc .
relat . phenom . *
117 - 118 * , 481 ( 2001 ) .
p. monceau , f. ya .
nad , and s. brazovskii , phys .
lett . * 86 * , 4080 ( 2001 ) .
k. takenaka , k. nakada , a. osuka , s. horii , h. ikuta , i. hirabayashi , s. sugai , and u. mizutani , phys .
lett . * 85 * , 5428 ( 2000 ) .
t. mizokawa , c. kim , z .- x .
shen , a. ino , t. yoshida , a. fujimori , m. goto , h. eisaki , s. uchida , m. tagami , k. yoshida , a. i. rykov , y. siohara , k. tomimoto , s. tajima , yuh yamada , s. horii , n. yamada , yasuji yamada , and i. hirabayashi , phys .
lett . * 85 * , 4779 ( 2000 ) .
j. moser , j. r. cooper , d. jrome , b. alavi , s. e. brown , and k. bechgaard , phys .
lett . * 84 * , 2674 ( 2000 ) .
g. mihly , i. kzsmrki , f. zmborszky , and l. forr , phys .
lett . * 84 * , 2670 ( 2000 ) .
v. vescoli , f. zwick , j. voit , h. berger , m. zacchigna , l. degiorgi , m. grioni , and g. grner , phys .
lett . * 84 * , 1272 ( 2000 ) .
j. d. denlinger , g.h .
gweon , j. w. allen , c. g. olson , j. marcus , c. schlenker , and l .- s .
hsu , phys .
lett . * 82 * , 2540 ( 1999 ) .
h. fujisawa , t. yokoya , t. takahashi , s. miyasaka , m. kibune , and h. takagi , phys .
b * 59 * , 7358 ( 1999 ) .
k. kobayashi , t. mizokawa , a. fujimori , m. isobe , y. ueda , t. tohyama , and s. maekawa , phys .
lett . * 82 * , 803 ( 1999 ) ; k. kobayashi , t. mizokawa , and a. fujimori , _ ibid . _ * 80 * , 3121 ( 1998 ) .
claude bourbonnais and denis jrome , science * 281 * , 1155 ( 1998 ) .
v. vescoli , l. degiorgi , w. henderson , g. grner , k. p. starkey , and l. k. montgomery , science * 281 * , 1181 ( 1998 ) .
f. zwick , d. jrome , g. margaritondo , m. onellion , j. voit , and m. grioni , phys .
lett . * 81 * , 2974 ( 1998 ) ; f. zwick , s. brown , g. margaritondo , c. merlic , m. onellion , j. voit , and m. grioni , _ ibid . _ * 79 * , 3982 ( 1997 ) .
r. neudert , m. knupfer , m. s. golden , j. fink , w. stephan , k. penc , n. motoyama , h. eisaki , and s. uchida , phys .
lett . * 81 * , 657 ( 1998 ) .
t. mori , t. kawamoto , j. yamaura , t. enoki , y. misaki , t. yamabe , h. mori , and s. tanaka , phys .
lett . * 79 * , 1702 ( 1997 ) . c. kim , a. y. matsuura , z .- x .
shen , n. motoyama , h. eisaki , s. uchida , t. tohyama , and s. maekawa , phys .
lett . * 77 * , 4054 ( 1996 ) ; n. motoyama , h. eisaki , and s. uchida , phys .
lett . * 76 * , 3212 ( 1996 ) . a.
luther and i. peschel , phys .
b * 9 * , 2911 ( 1974 ) ; * 12 * , 3908 ( 1975 ) .
j. slyom , adv . phys . * 28 * , 201 ( 1979 )
. f. d. m. haldane , j. phys .
c * 14 * , 2585 ( 1981 ) .
l. d. faddeev and l. a. takhtajan , phys . lett . * 85a * , 375 ( 1981 ) .
j. c. talstra , s. p. strong , and p. w. anderson , phys .
lett . * 74 * , 5256 ( 1995 ) .
j. c. talstra and s. p. strong , phys .
b * 56 * , 6094 ( 1997 ) .
y. mizuno , k. tsutsui , t. tohyama , and s. maekawa , phys .
b * 62 * , r4769 ( 2000 ) ; h. kishida , m. ono , k. miura , h. okamoto , m. izumi , t. manako , m. kawasaki , y. taguchi , y. tokura , t. tohyama , k. tsutsui , and s. maekawa , phys . rev
. lett . * 87 * , 177401 ( 2001 ) .
m. granath , v. oganesyan , d. orgad , and s. a. kivelson , phys .
b * 65 * , 184501 ( 2002 ) .
d. orgad , s. a. kivelson , e. w. carlson , v. j. emery , x. j. zhou , and z. x. shen , phys .
lett . * 86 * , 4362 ( 2001 ) .
e. w. carlson , d. orgad , s. a. kivelson , and v. j. emery , phys .
b * 62 * , 3422 ( 2000 ) .
j. zaanen , o. y. osman , h. v. kruis , z. nussinov , and j. tworzydlo , philos .
b * 81 * , 1485 ( 2001 ) .
a. l. chernyshev , s. r. white , and a. h. castro neto , phys .
b * 65 * , 214527 ( 2002 ) .
m. sing , u. schwingenschlgl , r. claessen , p. blaha , j. m. p. carmelo , l. m. martelo , p. d. sacramento , m. dressel , and c. s. jacobsen , submitted for publication in phys .
b ( 2003 ) [ cond - mat/0304283 ] . fabian h. l. essler , vladimir e. korepin , and kareljan schoutens , phys
lett . * 67 * , 3848 ( 1991 ) ; nucl .
b * 372 * , 559 ( 1992 ) .
j. m. p. carmelo , submitted for publication in phys .
b ( 2003 ) [ cond - mat/0305568 ] . j. m. p.
carmelo and k. penc , to be submitted for publication in phys .
b ( 2003 ) . j. m. p. carmelo , k. penc , l. m. martelo , p. d. sacramento , j. m. b. lopes dos santos , r. claessen , m. sing , and u. schwingenschlgl , to be submitted for publication ( 2003 ) .
j. m. p. carmelo , n. m. r. peres , and p. d. sacramento , phys .
lett . * 84 * , 4673 ( 2000 ) .
j. m. p. carmelo , k. penc , and p. d. sacramento , to be submitted for publication ( 2003 ) .
j. m. p. carmelo , f. guinea , and p. d. sacramento , to be submitted for publication ( 2003 ) .
j. m. p. carmelo , l. m. martelo , and p. d. sacramento , to be submitted for publication ( 2003 ) .
s. j. gu , n. m. r. peres , and j. m. p. carmelo , to be submitted for publication ( 2003 ) .
h. schulz , phys .
lett . * 64 * , 2831 ( 1990 ) . m. takahashi , j. phys .
c * 10 * , 1289 ( 1977 ) .
a. a. belavin , a. m. polyakov , and a. b. zamolodchikov , j. stat . phys . * 34 * , 763 ( 1984 ) ; nucl . phys .
b * 241 * , 333 ( 1984 ) ; holger frahm and v. e. korepin , phys .
b * 42 * , 10 553 ( 1990 ) ; _ ibid . _ * 43 * , 5653 ( 1991 ) .
d. snchal , d. perez , and m. pioro - ladrire , phys .
lett . * 84 * , 522 ( 2000 ) .
d. controzzi , f. h. l. essler , and a. m. tsvelik , phys .
lett . * 86 * , 680 ( 2001 ) . | in the first paper of this series it was found that the @xmath0-spin @xmath1 holons , spin @xmath1 spinons , and @xmath2 pseudoparticles whose occupancy configurations describe the energy eigenstates of the one - dimensional hubbard model emerge from the electron - rotated - electron unitary transformation .
an important breakthrough is that the theory provides relevant information about the relationship of the original electrons to these quantum objects . in this second paper
we discuss and clarify how such a relation can be used in a program for evaluation of finite - energy few - electron spectral functions . as a first step , here we characterize the dominant holon and spinon microscopic physical processes that originate more than 99% of the few - electron spectral weight .
these dominant processes are related to exact selection rules for the values of the number of holons and spinons generated or annihilated by application onto a ground state of rotated - electron operators .
while our theory also describes the higher - order microscopic processes associated with the remaining less than 1% few - electron spectral weight , the clarification and study of the above dominant processes is valuable and useful for the further understanding and description of the few - electron spectral properties observed in real low - dimensional materials .
moreover , in this paper generalize the concepts of a lower hubbard band and upper hubbard bands to all values of on - site coulombian repulsion . |
an understanding of the conditions under which stars formed at all epochs is critical in constraining theoretical models of galaxy evolution .
recent observations of @xmath6 star - forming galaxies suggest that the ism state and/or the hardness of the euv radiation field was more extreme in the past than in the present day .
galaxies at high redshift display larger line ratios ( [ ] /h@xmath1 and/or [ ] /h@xmath2 ) than local galaxies ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* and references therein ) .
these enlarged line ratios have been interpreted in terms of a contribution from an agn or shocks @xcite , a larger nitrogen abundance @xcite , a larger ionization parameter @xcite , and/or a higher ism pressure @xcite .
measurements of the ionization parameter and electron density of the ism using rest - frame optical line ratios support the general picture of a larger ionization parameter or electron density although the useable samples are small @xcite .
however , some high redshift galaxies have electron densities similar to local galaxies ( see * ? ? ?
* ; * ? ? ?
* ) , implying a different cause of the anomalous line ratios .
@xcite used theoretical simulations to show how the [ ] /h@xmath2 and [ ] /h@xmath1 ratios of star forming galaxies and agn may change with redshift , given four sets of extreme assumptions .
if the ism conditions change with redshift , then the standard optical classification schemes and other line ratio diagnostics may not be applicable at high redshift @xcite . here
, we investigate the change in optical line ratios across the intermediate redshift range @xmath0 using the f2 field of the shels galaxy redshift survey .
the shels survey is 95% complete to a limiting magnitude of @xmath7 , allowing the optical line ratios to be investigated as a function of redshift and stellar mass . at high redshift , emission - line selection effects are difficult to remove . galaxies with low [ ] or [ ] luminosities may be missing from high - redshift samples , potentially accounting for a large fraction of the observed change in line ratios with redshift @xcite .
the high level of spectroscopic completeness of the shels sample avoids these issues and allows one to robustly constrain the change in line ratios with time , as well as the cause(s ) of the change in line ratios . in this paper
, we use the shels sample to show that the [ ] /h@xmath1 , [ ] /h@xmath2 , and [ ] /h@xmath2 ratios change systematically from @xmath8 to @xmath9 , and we demonstrate that this change is caused by a change in the number of ionizing photons per unit area within galaxies , as a function of redshift . throughout this paper , we we use the flat @xmath10-dominated cosmology as measured by the 7 year wmap experiment ( @xmath11 , @xmath12 ; * ? ? ?
the shels survey is based on two of the deep lens survey fields @xcite .
we use the f2 field from the shels survey , described in @xcite .
the 4 square degree field contains 12,705 spectroscopic redshifts for galaxies with [email protected] .
the survey is 95% complete to this limit .
the shels spectra were obtained by @xcite using the hectospec multi - fiber spectrograph on the mmt @xcite .
the spectra cover the full optical wavelength range from 3700 - 9100 , at a spectral resolution of @xmath14 .
this high spectral resolution is sufficient to resolve the [ ] doublet , allowing electron density estimates to be made by stacking in bins of stellar mass and redshift .
example individual spectra are shown in @xcite .
we have selected a sub - sample of the shels catalog according to the following criteria : * a redshift in the range @xmath0 .
the lower redshift limit avoids aperture effects , which can be large for @xmath15 @xcite . * the 4000 break index , @xmath16 .
this selection limits the sample to stellar ages @xmath17 gyr , and is insensitive to the metallicity of the stellar population ( e.g. , * ? ? ?
the @xmath18 index is defined as the ratio of the flux ( @xmath19 ) in the @xmath20 and @xmath21 bands @xcite .
the @xmath18 index has been shown to be less sensitive to reddening effects and contains less uncertainties from the stellar continuum than the original @xmath22 index @xcite .
the 4000 break is produced mainly by the heavy element absorption in the atmospheres of the older stellar population- the hotter and younger stars produce a smooth continuum in this region of the spectrum .
thus , the @xmath23 index provides a relative measure of the age of the stellar population . the @xmath18 distribution for shels is described in @xcite and @xcite .
@xcite showed that the shels @xmath18 distribution is bimodal , with @xmath24 selecting for actively star - forming galaxies and @xmath25 selecting for quiescent galaxies . to avoid emission - line selection effects
, we stack the data in bins of stellar mass , as described in @xcite .
we divide our sample into bins of stellar mass in @xmath26 dex increments .
at least @xmath27 galaxies per bin are required to provide measurable [ ] and h@xmath1 emission - lines in each stacked spectrum .
the spectra and observational uncertainties are linearly interpolated to a common rest - frame wavelength vector based on the redshift of the bin . the interpolated rest - frame wavelength vector has a spectral resolution of 1.5 and spans an observed wavelength range of @xmath28 .
the average flux at each resolution element is taken and the errors are added in quadrature .
@xcite conducted an in - depth analysis of the use of stacking for the shels sample .
they showed that stacking does not bias line ratio analysis for shels .
stellar masses were derived for our sample from the broad - band photometry using the _ le phare _ code by arnouts & ilbert ( private communication ) , with the @xcite stellar templates as inputs .
these stellar templates were given for two metallicities and for seven exponentially decreasing star formation models ( @xmath29 ) with @xmath30 gyr over a stellar population age between @xmath31 gyr . to correct for dust attenuation
, we apply the @xcite law with e(b - v)@xmath32 .
the _ le phare _ code provides a mass probability distribution .
we use the median of the mass distribution as the stellar mass of our galaxies .
the ionization parameter , @xmath33 , is defined as the number of hydrogen ionizing photons passing through a unit area per second divided by the number density of hydrogen atoms .
the volume - weighted mean ionization parameter , @xmath33 , can be defined in terms of the strmgren radius @xmath34 for a filled spherical region as @xmath35 where @xmath36 is the total number density , and @xmath37 is the number of ionizing photons produced by the exciting stars .
a larger ionization parameter at high redshift has been proposed as a potential cause of the change in line ratios with redshift ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
however , most high redshift galaxy spectra do not contain sufficient emission - lines to reliably constrain the ionization parameter .
fortunately , we are able to constrain the ionization parameter for shels , and to investigate how the ionization parameter changes both as a function of redshift and as a function of stellar mass .
we measure the ionization parameter using the [ ] / [ ] ratio using the theoretical iterative approach described in @xcite , and updated in @xcite .
we use the equivalent width ratio of [ ] / [ ] in these calculations . @xcite and @xcite showed that the [ ] / [ ] flux ratios are well represented by the equivalent width ratio of [ ] / [ ] . for redshifts
@xmath38 , we have verified that the use of equivalent widths yields the same results as using extinction - corrected line fluxes .
an agn produces a hard ionizing radiation field which leads to elevated [ ] /h@xmath2 and [ ] /h@xmath1 line ratios . a larger relative contribution from an agn at high
redshift has been proposed as the cause of the rise in [ ] /h@xmath1 ratio with redshift @xcite .
sample selection is likely to play a major role in whether agn contribute to the observed line ratios in a given sample .
the effect of residual agn contamination on stacked data is unknown .
however , the shels sample also allows us to investigate the effect of agn on the [ ] /h@xmath2 and [ ] /h@xmath1 optical line ratios ratios for galaxies between @xmath39 and on the [ ] /h@xmath2 ratio for galaxies between @xmath3 .
we remove agn and composites using the [ ] /h@xmath1 versus [ ] /h@xmath2 optical diagnostic diagram , forming a purely star - forming sample @xcite ( figure [ agn_removal ] top panel ) ) . in figure [ agn_removal ] ( middle and lower panels ) , we show how the optical line ratios change with redshift for each stellar mass bin ( colored curves ) .
the right panel shows the effect of agn contamination on the line ratios in the stacked data . the hard ionizing radiation field from an agn strongly affects the [ ] /h@xmath1 ratio because more @xmath40 ions are ionized into @xmath41 ions .
it is clear that if agn are not removed , the [ ] /h@xmath1 ratio is strongly affected by agn for stellar masses larger than @xmath42 .
galaxies with lower stellar masses contain few agn ( e.g. , * ? ? ?
therefore , for @xmath43 where we can not remove the agn we have restricted our analysis to galaxies with @xmath44 to minimise agn contamination .
figure [ oiiihb_mass ] shows that the [ ] /h@xmath1 ratio rises significantly between @xmath0 .
although this ratio is also a strong function of stellar mass ( figure [ oiiihb_mass]a ) nonetheless , the stellar mass does not appear to affect the size of the change in [ ] /h@xmath1 shown in figure [ oiiihb_mass]b for each mass bin .
the [ ] /h@xmath1 ratio ratio rises by @xmath45 dex across the @xmath0 redshift range for all stellar masses in the range @xmath46 .
this rise is consistent with high redshift studies which find large [ ] /h@xmath1 line ratios in star - forming galaxies with @xmath47 @xcite .
we will now briefly investigate possible mechanisms which could produce this observed rise in the [ ] /h@xmath1 ratio .
larger pressures in the regions of galaxies may cause a rise in the [ ] /h@xmath1 ratio .
this pressure can be inferred from the electron density of the gas . in a fully ionized plasma with an isobaric density distribution , the particle density is defined in terms of the ratio of the mean ism pressure , and mean electron temperature , @xmath48 , through @xmath49 , and @xmath36 is related to the electron density through @xmath50 .
for an ionized gas , the electron temperature is @xmath51 k and the electron density is therefore proportional to the ism pressure ( see e.g. , * ? ? ?
* for a discussion ) .
we can trace the electron density in the shels sample for @xmath52 using the [ ] @xmath53 [ ] @xmath54 ratio .
we find that the difference in [ ] ratio between @xmath3 is consistent with no change with redshift ; @xmath55 [ ] @xmath53 [ ] @xmath54@xmath56 from @xmath8 to @xmath57 .
furthermore , a larger ism pressure increases all three [ ] , [ ] , and [ ] ratios simultaneously because the larger pressure suppresses ir fine structure cooling , leading to more collisional excitations of all three [ ] , [ ] , and [ ] lines ( see figure 2 in * ? ? ?
. however , in the shels sample between @xmath3 , the log([]/h@xmath2 ) and log([]/h@xmath2 ) ratios actually decrease with redshift by @xmath58 dex ( panels c and g of figure [ agn_removal ] ) .
we therefore rule out increasing pressure with redshift as the cause of the observed change in shels line ratios across the @xmath3 range .
recently , a hard ionizing radiation field has been proposed as the cause of the large [ ] /h@xmath1 line ratios in high redshift galaxies @xcite .
the effect of a hard ionizing radiation field on the optical line ratios is clearly illustrated in figure [ agn_removal ] by comparing the panels with and without an agn for high mass galaxies ( @xmath59@xmath60 ) .
a hard ionizing radiation field from either an agn , shocks , or from a hot stellar population ( such as wolf - rayet stars ) ionizes more neutral nitrogen and sulphur into @xmath61 and @xmath62 , as well as ionizing @xmath40 into @xmath41 ( e.g. , * ? ? ?
* ; * ? ? ?
* ) . the [ ] emission - line is particularly sensitive to the hardness of the ionizing radiation field . in a hard ionizing radiation field ,
a large , warm and partially ionized zone containing @xmath62 ion extends to the edge of the photoionised nebula , producing enhanced [ ] /h@xmath2 ratios . because both [ ] /h@xmath2 and [ ] /h@xmath2 ratios fall with redshift , we can rule out a hard ionizing radiation as the cause of the change in shels line ratios with redshift .
a larger ionization parameter might produce the observed change in line ratios . in figure [ oiiioii_z ] , we show that the ionization - parameter sensitive line ratio [ ] / [ ] rises with redshift ( by up to @xmath63 dex ) .
this change in [ ] / [ ] ratio corresponds to a change in the ionization parameter of up to @xmath64 @xmath65 .
although an increase in pressure can lead to an apparent rise in @xmath33 of this magnitude @xcite , the pressures needed would be in excess of @xmath66 , which is excluded by an absence of any change in the electron density as measured by the [ ] @xmath53 [ ] @xmath54 ratio with redshift in the shels sample .
thus , any change in the hydrogen ionizing photon flux must be produced either by increasing the hardness ( i.e. changing the slope ) of the ionizing euv radiation field , or by scaling up the whole ionizing radiation field by a some factor .
we have already ruled out a change in the hardness of the ionizing radiation field .
we now investigate the prospect of scaling the radiation field . a larger hydrogen ionizing photon flux on the inner edge of the nebula could either be produced by collective effects such as a larger number of stars within an individual region , or by geometrical effects such as a generally closer proximity of the stars to the ionisation front as in blister regions .
if the first hypothesis is correct , we would expect the ionization parameter to be related to the current star formation rate .
this hypothesis can be tested by comparing the [ ] / [ ] ratio or the derived ionization parameter with the h@xmath1 emission - line equivalent width , because the balmer equivalent width traces the relative fraction of young stars to old stars in a galaxy ( see e.g. , * ? ? ? * for a review ) .
indeed , we find a strong correlation ( figure [ logq_ewhb ] ) between these quantities . the spearman - rank correlation coefficient between the ionization parameter and the h@xmath1 equivalent width is 0.81 , and the probability of obtaining this value by chance is formally zero ( @xmath67 ) .
we conclude that a larger number of young to old stars within regions is likely to be principally responsible for the change in ionization parameter with redshift in the shels sample .
if the ionization parameter increases with redshift , then this will have important implications on the use of line ratios to derive metallicity at high redshift .
in particular , the [ ] /h@xmath2 ratio and the [ ] /h@xmath1 ratios will be impacted strongly by a change in ionization parameter .
both of these line ratios are in widespread use in high - z metallicity studies due to their insensitivity to extinction and flux calibration .
if the ionization parameter is not taken into account , the [ ] /h@xmath2 ratio will underestimate the metallicity .
the effect of ionization parameter on the [ ] /h@xmath1 ratio is more complicated because the [ ] /h@xmath1 ratio is double - valued with metallicity . at high metallicity ( @xmath68@xmath69 ) ,
the [ ] /h@xmath1 metallicity will be underestimated , while at low metallicity ( @xmath68@xmath70 ) , the [ ] /h@xmath1 metallicity will be overestimated .
we recommend that the ionization parameter and metallicity be derived simultaneously to ensure reliable metallicity estimates for non - local galaxies .
we investigate the change in optical line ratios across @xmath0 for the shels survey . for @xmath5 , the [ ] /h@xmath1 and [ ] / [ ] ratios rise ( by @xmath71 dex ) across the full @xmath0 range .
for @xmath3 we show that while the [ ] /h@xmath1 and [ ] / [ ] ratios rise , the [ ] /h@xmath2 and [ ] /h@xmath2 ratios fall significantly ( by @xmath72 dex ) for stellar masses betweeon @xmath4 .
we explore these changes in terms of the electron density ( or ism pressure ) of the gas , the slope of the ionizing radiation field , and the relative fraction of young to old stars .
we rule out a change in the electron density and a change in the hardness of the ionizing radiation field as dominant causes of our observed change in line ratios with redshift .
we examine the effect of agn contamination .
we show that agn substantially contaminates the [ ] /h@xmath1 line ratios at stellar masses ( @xmath73 ) .
agn removal is critical for high - z studies that interpret the [ ] /h@xmath1 ratio in terms of star - forming galaxy properties .
however , contribution by agn does not affect the general trends observed in the shels [ ] /h@xmath2 and [ ] /h@xmath2 line ratios .
it is unclear whether this lack of agn sensitivity will hold at high - z , given potentially different ism conditions and agn properties .
we show that the observed change in shels line ratios across @xmath3 is dominated by a rise in the ionization parameter with redshift .
the geometry , stellar mass contained , and the age of the stellar populations within regions may all contribute to the observed ionization parameter .
the observed change in ionization parameter is strongly correlated with the fraction of young to old stars , as traced by the h@xmath1 equivalent width .
we emphasize that while the ionization parameter dominates the change in line ratios observed in shels , the ionization parameter may not be the dominant cause of a change in line ratios for different redshift ranges or different stellar mass ranges , particularly where shocks or agn contamination may be present .
m.d . and l.k .
are supported by an arc discovery project dp130103925 .
l.k . gratefully acknowledges an arc future fellowship and the 2014 anu academic women s writing workshop .
the smithsonian institution supports the research of m.g . and d.f . , d. g. , hertz , e. n. , szentgyorgyi , a. h. , fata , r. g. , roll , j. b. , & zajac , j. m. 1998 , in society of photo - optical instrumentation engineers ( spie ) conference series , vol . 3355 , optical astronomical instrumentation , ed . s. dodorico , 285296 | we investigate the change in ionizing photons in galaxies between @xmath0 using the f2 field of the shels complete galaxy redshift survey .
we show , for the first time , that while the [ ] /h@xmath1 and [ ] / [ ] ratios rise , the [ ] /h@xmath2 and [ ] /h@xmath2 ratios fall significantly over the @xmath3 redshift range for stellar masses between @xmath4 . the [ ] /h@xmath1 and [ ] / [ ] ratios continue to rise across the full @xmath0 redshift range for stellar masses between @xmath5 .
we conclusively rule out agn contamination , a changing ism pressure , and a change in the hardness of the euv radiation field as the cause of the change in the line ratios between @xmath3 .
we find that the ionization parameter rises significantly with redshift ( by 0.1 to 0.25 dex depending on the stellar mass of the sample ) .
we show that the ionization parameter is strongly correlated with the fraction of young - to - old stars , as traced by the h@xmath1 equivalent width .
we discuss the implications of this result on higher redshift studies , and we consider the implications on the use of standard optical metallicity diagnostics at high redshift . |
we explore the dynamics of lysogenic phage in a bacterial biofilm , for bacterial hosts both with and without crispr immunity .
classical models of phage - bacteria interactions have previously demonstrated that lysogeny can promote the stable co - existence of bacteria and phage @xcite . in agreement with these findings ,
model 1 explores the baseline conditions under which lysogeny exists in a biofilm , and provides conditions for the stability of a lysogenic equilibrium .
the second and third models demonstrate the powerful effect of the crispr - cas system in comparison with the non - crispr bacterial population in the first model .
the second model predicts that at realistic parameter values , only prophage - free crispr - immune bacteria stably survive in a bacterial biofilm . in rare cases ,
the crispr system stably co - exists with the phage , however the crispr system does not allow lysogens to co - exist at this equilibrium . in the third model , there is no crispr system present in the lysogens . in the absence of an external source of lysogens ( case ( a ) ) , crispr - immune bacteria are predicted to dominate the biofilm .
however if an external source of planktonic lysogens contributes to the biofilm , crispr bacteria may co - exist with lysogens or can even be eliminated from the biofilm .
this last result is of clinical relevance because crispr bacteria are highly resistant to phage therapy .
once crispr bacteria are removed from the biofilm , phage therapy has a much higher chance of success .
the existence and stability conditions were found analytically to define the parametric regions in which populations exist or remain stable . in order to visualize the population behaviour in the biologically meaningful parameter space ,
computer simulations were used to verify the stability conditions .
in addition , parametric analysis was used to explore realistic parameter regimes , and to define therapeutic strategies to eradicate crispr - immune bacteria from the biofilm .
these results indicate that large magnitude changes in any one baseline parameter value would be required to achieve that objective .
this suggests that means of varying several parameters simultaneously might hold more therapeutic promise . in the first model ,
the lysogenic equilibrium is typically stable for realistic parameter values .
this equilibrium ( le@xmath5 ) loses stability when @xmath17 .
if we consider the sloughing off rate @xmath11 and prophage induction rate @xmath9 to be negligible compared to the growth rate @xmath18 , then this condition reduces to @xmath19 or @xmath20 which ensures the stability of pfe@xmath5 . in between the regions of stability of le@xmath5 and pfe@xmath5
, there is a small region of length @xmath9 where aee@xmath5 exists .
although the biological relevance of this region is arguable , we provide some numerical explorations in figure [ fig : limitcycle ] .
the second model predicts that the crispr response is sufficiently strong to eliminate lysogens , and therefore crispr - immune bacteria dominate the biofilm over a wide range of parameter values . at high infection rates ( i.e. @xmath21 ) , assuming the sloughing off rate is negligible compared to the bacterial growth rate , a stable crispr equilibrium ce@xmath6 exists .
this predicts that the phage may exist at high adsorption rates , while lysogen survival is unlikely unless the death probability @xmath22 goes to zero and sloughing off rates @xmath23 and @xmath11 become equal .
these two conditions would only hold in the unlikely scenario that the crispr system has no effect on the lysogens .
the third model explored the case of non - crispr lysogens and crispr - immune non - lysogens . at baseline
parametric values , the lysogens are predicted to go extinct , while the crispr - immune bacteria stably exist inside the biofilm . in some more extreme parameter regimes , lysogens also persist at equilibrium .
the corresponding eigenvalues show that the stability of le@xmath7 and pfe@xmath7 have almost complementary conditions , although a small region of bi - stability exists , i.e. @xmath24 .
case ( b ) of model 3 was specifically designed to explore the possibility of using lysogens to penetrate the biofilm and reduce the population of crispr - immune bacteria
. a high rate of biofilm formation is necessary to eliminate crispr - immune bacteria @xmath0 in this case ; both lysogens and phage are always present in the biofilm because of their continuous influx through flow .
although our model treats only one type of virus and corresponding prophage , an interesting possibility here is that diverse prophage could be introduced via the lysogens joining the biofilm . in this way
, lysogens could produce a number of different viruses via induction , and this could ultimately reduce the entire biofilm population .
since older crispr spacers in the bacteria become less efficient , the possibility of crispr failure increases with phage diversity , which helps to eradicate crispr - immune bacteria @xmath0 .
once crispr bacteria have been eliminated , the biofilm can be treated by classical therapeutic techniques .
although the arguments above are highly speculative , the main results of our research are summarized in figure [ fig : all_model ] which is divided into four panels ( vertical lines ) . on the left ,
we use model 2 to simulate a pathogenic biofilm that is resistant to phage therapy because it consists entirely of crispr - immune bacteria . in the second panel , an external , possibly therapeutic source of planktonic lysogens
is applied , and the all existing equilibrium state emerges . when the concentration of lysogens in the external source is further increased ( 3rd panel ) ,
the crispr - immune bacteria are eradicated from the biofilm and the model shows the same behaviour as model 1 with only lysogens and phage surviving . at realistic baseline parameters , the lysogenic equilibrium emerges . once the crispr population has been eliminated , the external source of lysogens can be removed and models 3 and 1 are equivalent .
variations in a number of parameter values can then be used to eliminate the lysogens ( panel 4 ) .
for example , as mentioned above , diversity in the prophages carried by lysogens could produce a number of distinct phage populations inside the biofilm , which could infect bacteria and also increase the prophage induction rate . extending the analysis presented here to include multiple phage types is an intriguing possibility for future work . in the second and third panels respectively . in the fourth panel ,
simulations of model 1 are presented for baseline parameters ( top two curves ) and for varied parameters based on possible therapy , i.e. @xmath25 , @xmath26 and @xmath27 , ( bottom two curves with increased line - width ) .
the dashed lines represent @xmath0 , dash - dot lines represent @xmath1 in model 2 , and @xmath3 in model 1 and 3 , whereas dots symbolize the population density of phage @xmath4 . ]
this work was funded by the natural sciences and engineering research council of canada ( nserc ) . 59 s.t .
abedon , _ bacteriophage ecology , advances in molecular and cellular microbiology _ , camb .
univ . press , cambridge , uk , 2008 .
abedon , _ phage evolution and ecology _ , adv .
microbiol . 67
( 2009 ) , pp .
abedon , _ bacterial `` immunity '' against bacteriophages _ ,
bacteriophage 2 ( 2012 ) , pp .
a.f . andersson and j.f .
banfield , _ virus population dynamics and acquired virus resistance in natural microbial communities _ , science 320 ( 2008 ) , pp .
10471050 .
s. bachellier , j.m .
clment , m. hofnung and e. gilson , _ bacterial interspersed mosaic elements ( bimes ) are a major source of sequence polymorphism in escherichia coli intergenic regions including specific associations with a new insertion sequence _ , genetics 145 ( 1997 ) , pp .
551562 .
m. ballyk , d. jones and h.l .
smith , _ the biofilm model of freter : a review , in structured population models in biology and epidemiology _
, p. magal and s. ruan , eds .
, 2008 , pp . 265302 .
l. barksdale and s.b .
arden , _ persisting bacteriophage infections , lysogeny , and phage conversions _ , ann . rev . of microbiol .
28 ( 1974 ) , pp .
265300 .
r. barrangou , c. fremaux , h. deveau , m. richards , p. boyaval , s. moineau et al .
, _ crispr provides acquired resistance against viruses in prokaryotes _ , science 315 ( 2007 ) , pp .
17091712 .
a. van belkum , s. scherer , l. van alphen and h. verbrugh , _ short - sequence dna repeats in prokaryotic genomes _ , microbiol .
62 ( 1998 ) , pp .
275293 .
e. bester , e.a .
edwards and g.m .
wolfaardt , _ planktonic cell yield is linked to biofilm development _
, can . j. microbiol .
55 ( 2009 ) , pp .
11951206 .
d. bhaya , m. davison and r. barrangou , _ crispr - cas systems in bacteria and archaea : versatile small rnas for adaptive defense and regulation _ , ann . rev . of genet . 45 ( 2011 ) , pp .
273297 .
r. capparelli , m. parlato , g. borriello , p. salvatore and d. iannelli , _ experimental phage therapy against staphylococcus aureus in mice _ ,
agents chemother .
51 ( 2007 ) , pp .
27652773 .
s. chibani - chennoufi , a. bruttin , m.l .
dillmann and h. brssow , _ phage - host interaction : an ecological perspective _ , j. bact .
186 ( 2004 ) , pp .
36773686 .
childs , n.l .
held , m.j .
young , r.j . whitaker and j.s .
weitz , _ multiscale model of crispr - induced coevolutionary dynamics : diversification at the interface of lamarck and darwin _ , evolution 66 ( 2012 ) , pp .
20152029 .
costerton , p.s .
stewart and e.p .
greenberg , _ bacterial biofilms : a common cause of persistent infections _ ,
science 284 ( 1999 ) , pp .
13181322 .
l. deng , r.a .
garrett , s.a .
shah , x. peng and q. she , _ a novel interference mechanism by a type iiib crispr - cmr module in sulfolobus _ , mol .
microbiol . 87 ( 2013 ) , pp .
10881099 .
h. deveau , j.e .
garneau and s. moineau , _ crispr - cas system and its role in phage - bacteria interactions _ , ann .
64 ( 2010 ) , pp .
475493 .
h. echols , _ developmental pathways for the temperate phage : lysis vs lysogeny _
( 1972 ) , pp . 157190 .
r. edgar and u. qimron , _ the escherichia coli crispr system protects from @xmath28 lysogenization , lysogens , and prophage induction _ , j. bacteriol .
192 ( 2010 ) , pp .
62916294 .
emerenini , b.a .
hense , c. kuttler , h.j .
eberl , a mathematical model of quorum sensing induced biofilm detachment .
forestier c , ed .
plos one 10 ( 2015 ) , e0132385 .
r. freter , h. brickner , j. fekete , m.m .
vickerman and k.e .
carey , _ survival and implantation of escherichia coli in the intestinal tract _
39 ( 1983 ) , pp .
686703 .
garneau , m. .
dupuis , m. villion , d.a .
romero , r. barrangou , p. boyaval et al .
, _ the crispr - cas bacterial immune system cleaves bacteriophage and plasmid dna _ , nature 468 ( 2010 ) , pp .
haerter and k. sneppen , _ spatial structure and lamarckian adaptation explain extreme genetic diversity at crispr locus _
, mbio 3 ( 2012 ) , pp . e0012612 .
hale , p. zhao , s. olson , m.o .
duff , b.r .
graveley , l. wells et al .
, _ rna - guided rna cleavage by a crispr rna - cas protein complex _ , cell 139 ( 2009 ) , pp . 945956 .
j. he and m.w .
deem , _ heterogeneous diversity of spacers within crispr ( clustered regularly interspaced short palindromic repeats ) _
( 2010 ) , pp . 128102 .
s. heilmann , k. sneppen and s. krishna , sustainability of virulence in a phage - bacterial ecosystem , j virol 84 ( 2010 ) , pp .
30163022 . n.l .
held , l.m .
childs , m. davison , j.s .
weitz , r.j .
whitaker and d. bhaya , _ crispr - cas systems to probe ecological diversity and host viral interactions , in crispr - cas systems _ , r. barrangou and j. van der oost , eds .
, 2013 , pp . 221250 .
held , a. herrera , h. cadillo - quiroz and r.j .
whitaker , _ crispr associated diversity within a population of sulfolobus islandicus _ , plos one 5 ( 2010 ) , pp .
e12988 .
heussler , k.c .
cady , k. koeppen , s. bhuju , b.a .
stanton and g.a .
otoole , _ clustered regularly interspaced short palindromic repeat - dependent , biofilm - specific death of pseudomonas aeruginosa mediated by increased expression of phage - related genes _ , mbio 6 ( 2015 ) , pp .
e0012915 .
y. ishino , h. shinagawa , k. makino , m. amemura and a. nakata , _ nucleotide sequence of the iap gene , responsible for alkaline phosphatase isozyme conversion in escherichia coli , and identification of the gene product _ , j. bact .
169 ( 1987 ) , pp .
54295433 .
r. jansen , j.d.a .
van embden , w. gaastra and l.m .
schouls , _ identification of genes that are associated with dna repeats in prokaryotes _
microbiol . 43
( 2002 ) , pp .
15651575 .
jones , h.v .
kojouharov , d. le and h.l .
smith , _ the freter model : a simple model of biofilm formation _ , j. math .
47 ( 2003 ) , pp .
137152 .
jones and h.l .
smith , _ bacteriophage and bacteria in a flow reactor _
73 ( 2011 ) , pp . 23572383 .
h. kawaji and y. hayashizaki , _ exploration of small rnas _ , plos genet . 4 ( 2008 ) ,
koonin and y.i .
wolf , _ evolution of the crispr - cas adaptive immunity systems in prokaryotes : models and observations on virus - host coevolution _ ,
11 ( 2015 ) , pp .
e. kutter and h. brussow , _ phage ecology _ , in _ bacteriophages : biology and applications _ , e. kutter and a. sulakvelidze , crc press , 2004 .
labrie , j.e .
samson and s. moineau , _ bacteriophage resistance mechanisms _ , nat .
micro . 8 ( 2010 ) , pp .
317327 .
levin , s. moineau , m. bushman and r. barrangou , _ the population and evolutionary dynamics of phage and bacteria with crispr - mediated immunity _ , plos genet . 9
( 2013 ) , pp .
e1003312 .
lupski , j.r .
roth and g.m .
weinstock , _ chromosomal duplications in bacteria , fruit flies , and humans _ , am
58 ( 1996 ) , pp .
a. mai and h.j .
eberl , _ persistence in a single species cstr model with suspended flocs and wall attached biofilms _ , bull .
74 ( 2012 ) , pp .
10011026 . j. k. miller , j. s. brantner , c. clemons , k. l. kreider , a. milsted , p. wilber et al .
, _ mathematical modelling of pseudomonas aeruginosa biofilm growth and treatment in the cystic fibrosis lung _ , math .
med . bio .
( 2014 ) , pp .
179204 .
f.j.m . mojica and r.a .
garrett , _ discovery and seminal developments in the crispr field , in crispr - cas systems _ ,
r. barrangou and j. van der oost , eds .
, 2013 , pp . 131 .
g. otoole , h.b .
kaplan and r. kolter , _ biofilm formation as microbial development _ , annu .
54 ( 2000 ) , pp .
paul , s. sundarrajan , s.s .
rajagopalan , s. hariharan , n. kempashanaiah , s. padmanabhan et al . ,
_ lysis - deficient phages as novel therapeutic agents for controlling bacterial infection _ , bmc microbiol .
11 ( 2011 ) , pp .
pires , d.v .
boas , s. sillankorva and j. azeredo , _ phage therapy : a step forward in the treatment of pseudomonas aeruginosa infections _ , j. virol .
89 ( 2015 ) , pp .
7449 - 7456 .
rollins , j.t .
schuman , k. paulus , h.s.t .
bukhari and b. wiedenheft , _ mechanism of foreign dna recognition by a crispr rna - guided surveillance complex from pseudomonas aeruginosa _ , nucl .
acids res .
43 ( 2015 ) , pp .
2216 - 2222 .
seed and j.j .
dennis , _ experimental bacteriophage therapy increases survival of galleria mellonella larvae infected with clinically relevant strains of the burkholderia cepacia complex _
, antimicrob .
agents chemother .
53 ( 2009 ) , pp .
22052208 .
singh , a.l .
schaefer , m.r .
parsek , t.o .
moninger , m.j .
welsh and e.p .
greenberg , _
quorum - sensing signals indicate that cystic fibrosis lungs are infected with bacterial biofilms _ , nature 407 ( 2000 ) , pp .
762764 .
m. skurnik and e. strauch , _ phage therapy : facts and fiction _ , int .
296 ( 2006 ) , pp .
r. sorek , c.m .
lawrence and b. wiedenheft , _ crispr - mediated adaptive immune systems in bacteria and archaea _ , ann . rev .
82 ( 2013 ) , pp .
237266 .
stewart and b.r .
levin , _ the population biology of bacterial viruses : why be temperate _ , theor . pop .
26 ( 1984 ) , pp . 93117 .
n. takeuchi , y.i .
wolf , k.s .
makarova and e.v .
koonin , _ nature and intensity of selection pressure on crispr - associated genes _ , j. bacteriol .
194 ( 2012 ) , pp .
12161225 .
tyson and j.f .
banfield , _ rapidly evolving crisprs implicated in acquired resistance of microorganisms to viruses _ , envir . microbiol . 10 ( 2008 ) , pp .
200207 .
vandervoort and g. brelles - mario , _ plasma - mediated inactivation of pseudomonas aeruginosa biofilms grown on borosilicate surfaces under continuous culture system _ , plos one 9 ( 2014 ) , pp .
e108512 .
webb , l.s .
thompson , s. james , t. charlton , t. tolker - nielsen , b. koch et al . , _ cell death in pseudomonas aeruginosa biofilm development _ , j. bacteriol .
185 ( 2003 ) , pp .
45854592 .
weinbauer , _ ecology of prokaryotic viruses _ , fems microbiol .
28 ( 2004 ) , pp .
127181 .
b. wiedenheft , s.h . sternberg and j.a .
doudna , _ rna - guided genetic silencing systems in bacteria and archaea _ , nature 482 ( 2012 ) , pp . 331338 .
yoon , r.f .
hennigan , g.m .
hilliard , u.a .
ochsner , k. parvatiyar , m.c .
kamani et al . , _ pseudomonas aeruginosa anaerobic respiration in biofilms : relationships to cystic fibrosis pathogenesis _ ,
cell 3 ( 2002 ) , pp .
593603 .
zegans , j.c .
wagner , k.c .
cady , d.m .
murphy , j.h .
hammond and g.a .
otoole , _ interaction between bacteriophage dms3 and host crispr region inhibits group behaviors of pseudomonas aeruginosa _ , j. bacteriol .
191 ( 2009 ) , pp .
the jacobian of model 1 , evaluated at @xmath29 provides the following eigenvalues @xmath30 where @xmath31 . clearly , all eigenvalues are real which implies that there can be no oscillation .
furthermore , the first and the third eigenvalues are always negative for @xmath32 . conditions for negativity of the second eigenvalue are discussed in the main text .
the eigenvalues of the jacobian evaluated at the lysogenic equilibrium @xmath33 are @xmath34 where @xmath35 .
when the prophage induction rate is high ( e.g. @xmath36 ) , a narrow range of parameter space allows for the stable existence of aee@xmath5 .
although the biological relevance of this regime is limited , a brief numerical exploration reveals that a stable limit cycle , stable node , or instability are all possible , depending on the initial conditions and the sloughing off rate @xmath11 ( see figure [ fig : limitcycle ] ) .
the eigenvalues of the jacobian for model 2 , evaluated at @xmath37 are : @xmath38 the jacobian evaluated at ce@xmath6 provides the following eigenvalues : @xmath39 since @xmath40 from the existence condition , the sufficient condition for the negativity of the first eigenvalue is @xmath41 , whereas the other two eigenvalues are always negative . evaluating the jacobian at the lysogenic equilibrium le@xmath6 provides the following eigenvalues , @xmath42 where @xmath43 and @xmath44 .
the last two eigenvalues for le@xmath6 are always negative while the first one requires a condition that is provided in table [ table : table3 ] .
eigenvalues for the jacobian of model 3 , evaluated at the crispr equilibrium @xmath45 are given by : @xmath46 the eigenvalues for the jacobian of le@xmath7 are given by : @xmath47 the eigenvalues for the lysogenic equilibrium le@xmath2 are found to be @xmath48 where @xmath49 and @xmath50 .
the stability conditions for the lysogenic equilibrium le@xmath2 in terms of sloughing off rate @xmath11 are : @xmath51 where @xmath52 and @xmath53 .
the induction rate was varied to investigate its effectiveness against the crispr and non - crispr lysogens in the biofilm ( see figure [ fig : alpha ] ) . in the case of competition between non - crispr bacterial populations , i.e. model 1 , variations in @xmath9 had no effect ; the lysogenic equilibrium remains stable .
likewise , in the presence of a crispr system , model 2 , lysogens die out due to the death and sloughing off rates regardless of the value of @xmath9 ; this result holds for model 3a as well . in case ( b ) of model 3 , we see that at low rates of prophage induction , i.e. @xmath54 , the lysogenic equilibrium is stable . as the lysogen lifetime
is reduced due to increases in prophage induction , the population of crispr - immune bacteria also emerges . for the non - crispr populations , model 1 , the lysogenic equilibrium is stable at all biologically meaningful values for the rate of phage loss , as shown in figure [ fig : gamma ] ( top - left plot ) , and
the all existing equilibrium is never stable , even at very high rates of @xmath10 . on the other hand ,
crispr immunity makes it possible for non - lysogens to stably exist at most plausible values of @xmath10 , although for very low @xmath10 the crispr equilibrium ( ce@xmath6 ) is stable in model 2 . in model 3a ,
crispr bacteria again dominate most of the parameter space ( see figure [ fig : gamma ] ) . in case ( b ) of model 3 , the external lysogens joining the biofilm do not stabilize le@xmath2 significantly ; the lysogenic equilibrium is only stable for @xmath55 . for most values of @xmath10 ,
the lysogens exist along with a population of crispr - immune bacteria and a decreasing density of phage in the biofilm , which is an obvious consequence of increasing phage loss .
the sloughing off rate is an important parameter in analyzing the stability of each model since it weakens the bacterial population by increasing its rate of loss from the biofilm .
the stability of the steady states are very similar to those observed in the above parameter analyses . in general
, we see that an increase in the sloughing off rate @xmath11 reduces the population of bacteria and phage in the biofilm , leading at very high rates to the trivial equilibrium ( te ) .
this is because when @xmath11 approaches @xmath18 , all the populations go extinct as shown in figure [ fig : eta ] .
the sloughing off rate @xmath11 affects lysogens either at the same rate as crispr - immune bacteria ( model 3a ) or at the higher rate ( model 2 ) .
moreover , in case ( b ) of model 3 , it can be observed that crispr - immune bacteria are eliminated from the biofilm before @xmath11 reaches @xmath56 , while the lysogens persist because the planktonic lysogens are continuously joining the biofilm .
since there are only two steady states in this case , the parametric value satisfying @xmath57 is sufficient to eradicate the crispr - immune bacteria from the biofilm . among the parameters involved in the inequality
, @xmath12 seems to be an important parameter which can be increased to eradicate @xmath0 even at comparatively low values of @xmath11 . | clustered regularly interspaced short palindromic repeats ( crispr ) , linked with crispr associated ( cas ) genes , play a profound role in the interactions between phage and their bacterial hosts .
it is now well understood that crispr - cas systems can confer adaptive immunity against bacteriophage infections .
however , the possibility of failure of crispr immunity may lead to a productive infection by the phage ( cell lysis ) or lysogeny .
recently , crispr - cas genes have been implicated in changes to group behaviour , including biofilm formation , of the bacterium _ pseudomonas aeruginosa _ when lysogenized . for lysogens with a crispr system
, another recent experimental study suggests that bacteriophage re - infection of previously lysogenized bacteria may lead to cell death .
thus crispr immunity can have complex effects on phage - host - lysogen interactions , particularly in a biofilm . in this contribution
, we develop and analyse a series of models to elucidate and disentangle these interactions . from a therapeutic standpoint ,
crispr immunity increases biofilm resistance to phage therapy .
our models predict that lysogens may be able to displace crispr - immune bacteria in a biofilm , and thus suggest strategies to eliminate phage - resistant biofilms .
+ + keywords : bacteria , bacteriophage , biofilm , crispr - cas system , lysogens .
1.5 1.5 [ cols="^,<,^",options="header " , ] 1.5 in figure [ fig : fig2 ] , the population dynamics of each of these models are illustrated for the same baseline parameter values and initial conditions .
the simulations are run over long times to illustrate stable equilibrium states , and in each case the numerical results were validated by comparison with the analytical expressions .
we plot both population densities and time on a log axis , such that initial transients are visible . in the first model ,
the baseline scenario produces a stable lysogenic equilibrium ( without crispr ) , in which the phage population is maintained in the biofilm through induction .
this is a classical model which supports the idea of co - existence of temperate phage and their bacterial hosts .
this result should generalize to the case of many phage types , since bacteria are capable of carrying several prophage as part of their genome . in the second model , the crispr bacteria @xmath0
clearly dominate , since crispr immunity makes these cells resistant to the phage .
the crispr effects on the lysogens @xmath1 , i.e. the death of lysogens or increase in the sloughing - off rate , make it possible to eliminate these infected bacteria from the biofilm ; their loss is followed by the loss of the phage population .
the second model demonstrates that crispr systems will disrupt the coexistence observed in model 1 , and predicts that for bacteria with crispr systems , stable biofilms are likely to be composed of crispr - immune bacteria only .
the populations in the first two models reach stable equilibria fairly quickly , whereas simulation results for the third model show that the populations take longer to reach their stable states . in the absence of an external source of lysogens ( model 3a ) ,
crispr - immune bacteria dominate , since the crispr system helps the bacteria to maintain the phage - free equilibrium after eliminating non - crispr lysogens from the biofilm . when the external source of lysogens is present ( model 3b ) , we observe stability of the all existing equilibrium for these baseline parameter values . this is because a high biofilm formation rate is required to meet the sufficient condition for the stability of le@xmath2 , thus eliminating the crispr cells from the biofilm . to investigate this further
, we turn to parametric analysis to explore scenarios in which lysogens are able to replace crispr - immune bacteria in the biofilm .
the second goal of the parametric analysis is to find a suitable conditions at which , after the eradication of crispr bacteria from the biofilm , the lysogens themselves could be eradicated . and
@xmath0 represent populations of wild - type bacteria and crispr - immune bacteria ( dashed lines ) , @xmath1 and @xmath3 represent lysogens with and without crispr systems ( dots ) and @xmath4 is the population of bacteriophage ( solid lines ) .
model 1 approaches the lysogenic equilibrium le@xmath5 , model 2 and model 3 ( case ( a ) ) approach the phage - free equilibria ( pfe@xmath6 and pfe@xmath7 ) whereas model 3 ( case ( b ) ) approaches the all existing equilibrium ( aee@xmath2 ) . ] in this section we illustrate several bifurcations that may occur in biologically realistic parameter regimes .
we first illustrate the changes in equilibria and stability that result from variation in the adsorption rate , @xmath8 .
variation in the prophage induction rate , @xmath9 , phage loss rate , @xmath10 , and sloughing off rate @xmath11 have similar qualitative effects and are shown in appendix s4 .
finally , we examine the interesting case of variation in the biofilm formation rate , @xmath12 .
variation in the adsorption rate constant affects the stability of equilibrium states by varying @xmath13 .
figure [ fig : beta ] shows that le@xmath5 remains stable for a wide range of @xmath8 values in model 1 , while the other models each switch stability between two equilibrium states at high rates of adsorption . in model 2 , pfe@xmath6 is replaced by ce@xmath6 at high adsorption rates .
this is due to the strong immune response of crispr bacteria , such that even for high adsorption rates crispr - immune bacteria survive , co - existing with the phage . in model 3 ,
le@xmath6 is stable at high @xmath8 , replacing pfe@xmath7 in case ( a ) and aee@xmath2 in case ( b ) .
we note that the biofilm formation rate @xmath12 does not have a strong influence here , since almost same adsorption rate @xmath8 is required in both cases to eradicate crispr - immune bacteria .
qualitatively , these results demonstrate that for biofilms without crispr systems ( model 1 ) , lysogens and phage are predicted to coexist over a wide parameter range .
in contrast , for biofilms with crispr ( model 2 ) , the stable equilibrium state is typically dominated by crispr - immune bacteria . if crispr - capable lysogens are replaced by lysogens without a functioning crispr system ( model 3 , a and b ) , again crispr - immune bacteria typically dominate , although for parameter regimes that are extremely favourable to phage reproduction , the lysogenic equilibrium may be stable . as shown in appendix s4 , these qualitative conclusions also hold for variations in parameters @xmath9 , @xmath11 and @xmath10 . in the above models .
each colour represents a unique population while the line width is increased to visualize overlapping populations .
bacterial populations are @xmath14 ( green ) and @xmath3 ( blue ) in model 1 , @xmath0 ( black ) and @xmath1 ( purple ) in model 2 and @xmath0 and @xmath3 in model 3 whereas the phage population is represented by @xmath4 ( red ) in all models . ] in model 3 , the rate of biofilm formation by planktonic lysogens is an important parameter that can be regulated either by increasing the density of lysogens near the biofilm or by increasing the adhesion rate .
we illustrate the effect of varying this parameter in panel ( a ) of figure [ fig : phi ] .
as the value @xmath12 increases , the lysogen population increases along with the phage , while the population of crispr - immune bacteria @xmath0 decreases in the biofilm , and is eventually eliminated at high values of @xmath12 .
though a relatively large formation rate is required to eliminate @xmath0 , this threshold value depends on other model parameters . in particular , the critical @xmath12 value to ensure the stability of the lysogenic equilibrium in model 3 ( case ( b ) ) is give by @xmath15 .
in the lower three panels of figure [ fig : phi ] we plot this threshold against @xmath9 , @xmath11 and the carrying capacity @xmath16 .
clearly , an increase in the value of @xmath9 and @xmath16 require an increase in the value of @xmath12 to make the lysogenic equilibrium stable , whereas an increase in the sloughing off rate @xmath11 reduces the threshold value of @xmath12 . in the model 3 ( case ( b ) ) .
bacterial populations @xmath0 ( black ) and @xmath3 ( blue ) and the phage population @xmath4 ( red ) are shown .
panels ( b ) , ( c ) and ( d ) show the critical value of @xmath12 necessary to ensure the stability of the lysogenic equilibrium , versus @xmath9 , @xmath11 and @xmath16 . in each case
, the parameter space above each curve corresponds to a stable lysogenic equilibrium ( le@xmath2 ) while the space under each curve corresponds to stability of the all existing equilibrium ( aee@xmath2 ) . ] |
the first observing runs of the advanced laser interferometer gravitational - wave observatory ( aligo ) are expected in @xmath4 , with advanced virgo following on a similar schedule @xcite .
the primary source for these observatories is the coalescence of binaries containing black holes and/or neutron stars , with predicted rates between a few and several hundred per year at detector design sensitivity @xcite .
the components of these binaries are formed in supernovae when the core of a massive star collapses to a compact remnant , although the exact collapse mechanism remains unknown .
detailed knowledge of the mass distribution of black holes and neutron stars will provide vital clues to their formation as well as explore the equation of state ( eos ) of nuclear matter at high densities .
measuring the upper and lower limits of neutron - star masses allows us to constrain the supernova engine and the nuclear physics of neutron - star remnants @xcite .
for example , the collapse of low - mass stars ( @xmath5 ) is believed to quickly produce explosions with very little mass accreted in a convective engine phase .
the predicted mass of the compact remnant will be less than the chandrasekhar mass by an amount that depends on the collapse model @xcite .
if we can place an upper limit on the mass of a low mass neutron star ( in the @xmath6 range ) , we can distinguish between current models , effectively using these low - mass systems to constrain the eos and neutrino physics in core - collapse . at higher masses , @xcite and @xcite
have argued that there is a gap between @xmath7 where no compact objects exist .
if true , the amount of material that falls back onto the newly formed compact remnant must be small , arguing against engine mechanisms that take a long time ( more than 200ms ) to develop @xcite .
however , this mass gap may simply be an artifact of poor mass resolution of x - ray binaries and poor statistics @xcite .
black - hole mass distributions will allow us to explore the fall - back of material in a weak supernova explosion and black - hole masses in solar metallicity environments will provide clues into stellar mass loss . with an accurate black - hole mass distribution , we can study these open questions in stellar evolution .
the binary s gravitational - wave phasing depends at leading order on its chirp mass @xmath8 , where @xmath9 and @xmath10 are the binary s component masses @xcite ; this quantity will be most accurately measured in a gravitational - wave detection .
the mass ratio @xmath11 enters through higher - order corrections and is less accurately measured ; see e.g. , @xcite .
there is also a partial degeneracy between the mass ratio and the angular momentum @xmath12 of each compact object ( the spin ) , which further limits our ability to measure the binary s component masses .
heuristically , this can be understood as follows : a binary with spins aligned with the orbital angular momentum will inspiral more slowly than a non - spinning system . similarly ,
a binary of the same _ total mass _ but with more extreme mass ratio will inspiral more slowly .
however , a binary with the same _ chirp mass _ but with more extreme mass ratio will inspiral more quickly . the effect on the waveform of decreasing @xmath13 can be mimicked by increasing the component spins .
we investigate the accuracy with which the component masses can be determined from gravitational - wave observations of binary - neutron - star ( bns ) , neutron - star black - hole ( nsbh ) , and binary - black - hole ( bbh ) systems , focussing on systems where the object s spins are aligned with the orbital angular momentum . since the first signals detected by aligo are likely to have signal - to - noise ratios ( snrs ) close to the observable network threshold of @xmath1412 @xcite , we focus on signals with snrs 1020 , which will account for @xmath1480% of observations . for these snrs , we find that the mass - ratio spin degeneracy will prevent us from accurately measuring component masses .
we identify the region of the mass parameter space for which it will not be possible to determine whether the compact objects are black holes or neutron stars using gravitational - wave observations alone , when we can conclusively measure compact - object masses outside the currently observed limits , and show how the observation of an electromagnetic counterpart to an nsbh could be used to constrain the black - hole spin .
fisher - matrix methods show that the binary s chirp mass is recovered well by matched filtering , with accuracies of @xmath15 for typical bns systems in aligo @xcite .
if we assume that the neutron stars are nonspinning @xmath13 can be measured to an accuracy of @xmath16 @xcite .
estimates of the effect of the mass - ratio spin degeneracy were first made by @xcite and @xcite using the fisher approach .
the degeneracy between the mass ratio and the total effective spin @xmath17 degrades the ability to measure the mass ratio and hence the component masses .
we go beyond these studies , using the method introduced in @xcite to equate a confidence interval with a region where the match between the signal and model waveforms exceeds a given threshold .
we use this method to investigate parameter degeneracies for a wide range of binaries and interpret the expected measurement accuracy in the context of the astrophysical questions discussed above .
we model the waveforms with the taylorf2 inspiral approximant @xcite to leading order in amplitude and 3.5 post - newtonian ( pn ) order in phase @xcite with spin - orbit terms to 3.5pn order and spin terms to 2pn order @xcite . for systems with total masses below @xmath18 ,
our results with taylorf2 are consistent with those from phenomenological bbh models that include the merger and ringdown @xcite , calibrated against numerical - relativity waveforms with mass ratios up to 1:4 @xcite . for the higher - mass bbh results in sec .
[ s : bbh ] we use the full merger model . throughout
, we assume for simplicity that the component spins are aligned with the orbital angular momentum . in this case , the binary s distance , orientation and sky location affect only the overall amplitude of the waveform , and we do not consider them here . for two waveforms @xmath19 and @xmath20 the match is given by @xmath21 where @xmath22 is the standard noise - weighted inner product @xmath23 in all cases we use a noise sensitivity @xmath24 corresponding to the zero - detuned high - power configuration of aligo @xcite with a 15 hz low frequency cutoff .
we construct a 90% confidence region for a signal in the @xmath25 space , which corresponds to the three - dimensional region where @xmath26 @xmath27 for an snr of 10 ( 20 ) @xcite .
this method is more accurate at low snrs than the fisher - matrix approach .
observed neutron - star masses currently lie between @xmath28 @xcite and @xmath29 @xcite .
general relativity and causality place a strict upper limit on the maximum neutron - star mass of @xmath30 @xcite ; the actual maximum mass is determined by the as yet unknown neutron - star eos .
the observed masses of stars in double neutron - star systems is narrower with a peak at @xmath31 and a width of @xmath32 @xcite .
we begin by considering a canonical bns system with masses @xmath33 and no spins .
[ fig : bns ] shows the regions of the mass plane that are consistent with this source at 90% confidence for snrs 10 and 20 .
the component masses consistent with the signal lie roughly along a line of constant chirp mass .
however , there is significant spread in the recovered component masses due to the degeneracy between @xmath13 and @xmath34 in the gravitational - wave phase evolution .
typically , the degeneracy persists over a range of @xmath35 in @xmath36 .
if we assume that the compact objects are non - spinning , the component masses are recovered in the ranges @xmath37@xmath31 and @xmath3@xmath38 .
the observed spins of double neutron stars are low , with the minimum observed period of @xmath39 for j0737 - 3039a @xcite , i.e. , @xmath40 , where the neutron - star period can be related to the spin by approximately @xmath41 if we constrain the neutron stars to have a spin @xmath42 , the range of consistent component masses extends to @xmath43@xmath44 and @xmath3@xmath45 . therefore at typical snrs , we would be unable to distinguish the canonical @xmath31 bns from a more exotic bns with @xmath46 .
the fastest spinning pulsar ( psr j1748 - 2446ad ) has a period of 1.4ms , or @xmath47
neutron stars are considered unlikely to have a period less than 1 ms @xcite , although breakup frequencies could be a factor of @xmath48 higher @xcite .
if we allow for larger component spins , the region consistent with the canonical bns system extends to @xmath49@xmath31 and @xmath50@xmath51 .
the maximum total spin of binaries within the 90% confidence region at snr 10 is @xmath140.3 ; see fig .
[ fig : bns ] inset . based on gravitational wave observations alone , a binary of two @xmath31 neutron stars
could not be definitively distinguished from a low - mass neutron star and a black hole or neutron star in the mass gap , even at snr 20 .
if the larger object is a neutron star its spin would be unusually large , but not impossible . to constrain the larger object s mass below 2@xmath52 without assumptions on the spins
would require a signal with an snr of @xmath5340 , expected for @xmath54 of observations .
the 90% confidence region around a 1.35@xmath55 - 1.35@xmath55 bns system at snr 10 ( gray ) and snr 20 ( black ) in the zero - detuned high - power configuration of aligo .
the vertical bars indicate where the snr 10 confidence region would be truncated if we restrict to non spinning neutron stars , and if we restrict to neutron - star spins less than 0.05 .
the inset shows the total effective spin @xmath56 of the waveforms , with respect to the mass of the larger body .
, scaledwidth=50.0% ] in fig .
[ fig : cmlines ] , we consider a family of equal mass , non - spinning bns systems with component masses at the lowest ( @xmath57 ) and highest ( @xmath58 ) observed neutron - star masses , and at the highest theoretical neutron - star mass ( @xmath59 ) .
the @xmath60 binary is shown for reference , and the observed mass gap is indicated by the shaded region .
again , the 90% confidence regions follow lines of approximately constant chirp mass .
we can see that for an exceptionally low or high mass binary , we will be able to identify _ at least one _ of the components as extraordinary .
the observation of a bns with a chirp mass @xmath61 would yield the unambiguous detection of a compact object with mass less than @xmath57 .
the less massive component could have a mass in the range @xmath62@xmath57 , but any mass in that range would challenge our current understanding of neutron stars and their formation in supernovae .
a chirp mass @xmath63 would yield an unambiguous detection of a @xmath64 neutron star .
this constraint would rule out many modern calculations of stellar collapse @xcite and provide a strong validation test for future calculations .
similarly , the observation of a system with @xmath65 would indicate the detection of a neutron star of mass @xmath66 _ provided we can unambiguously identify both components as neutron stars_. however , the observation of a bns system with @xmath67 is consistent with a binary containing a neutron star and a black hole in the mass gap or an exotic , low - mass neutron star and a black hole .
although any configuration along this line is of interest , we can not strongly constrain the component masses with gravitational - wave observations at low snr . in particular , even if we assume that neutron stars are nonspinning and that the minimum bh mass is 3@xmath52 , we can not rule out that the possibility that it is a nsbh system : our assumption would only remove the portion of the confidence interval with @xmath68 , and the observation would be of _ either _ a bns system with well - constrained masses , _ or _ an nsbh .
finally , we consider a binary with component masses at the upper end of the neutron - star mass limit , above which we expect the components to be black holes . in this case
, we could conclusively say that one of the components must be a black hole , although the degeneracy limits our ability to draw strong conclusions on the component masses .
the binary could either be a bbh system with component masses in the mass gap , or an nsbh system with masses consistent with previously observed compact objects .
note that in all the cases we have considered , the equal - mass line provides a hard upper ( lower ) limit on @xmath69 ( @xmath70 ) .
[ fig : cmlines ] shows that it will be difficult to explore the mass gap with single gravitational - wave observations at low snrs . to constrain one of the masses to lie within the mass gap , the system would need to be observed with an snr of greater than 30 ( @xmath71 of observations ) ; in the case of the @xmath67 binary , the mass on one of the objects would then be constrained between @xmath72 and @xmath73 , placing it directly within the mass gap .
the 90% confidence regions for a number of different non - spinning compact - binary configurations ; see text for interpretation .
the number at the end of each confidence region is the chirp mass of the binary .
the gray shaded region indicates the current observational mass gap .
, scaledwidth=50.0% ]
black hole spins can vary between @xmath74 and @xmath75 , with observations of x - ray binaries supporting the full range of values @xcite . as before , the mass - ratio spin degeneracy precludes precise measurement of the component masses .
however , if we observed a bbh with @xmath76m@xmath1 , corresponding to a binary with both components above @xmath77the largest mass observed for a stellar - mass black hole in an x - ray binary @xcite we _ could _ conclude that one of the black holes has a mass of @xmath78 or higher , providing the first observational evidence for a stellar - mass black hole above @xmath77 .
the same is true for any binary with the same or larger chirp mass , @xmath79 .
a bbh will only be unambiguously identified if all binaries within the 90% confidence interval for the measured component masses are also bbh systems .
the shaded region in fig .
[ fig : bhmassrange ] shows the part of parameter space in which bbh systems could be identified as such at snr 10 , assuming a maximum neutron - star mass of @xmath80 .
the upper curve shows the boundary of this region computed using the inspiral only ( taylorf2 ) model .
the lower curve is computed using a phenomenological inspiral - merger - ringdown model @xcite . when the additional information from merger and ringdown is included ,
a larger region of the parameter space can be identified as a bbh system , illustrating the impact of merger - ringdown in these systems .
the shaded region indicates binaries would be unambiguously identified as bbh systems , assuming a maximum neutron star mass of @xmath80 ( dotted line ) .
the light gray region represents the most conservative estimate of this region , using an inspiral - only waveform model , while the additional darker region uses a full merger - ringdown model .
, scaledwidth=50.0% ]
[ fig : bhspin ] shows the 90% confidence regions for an nsbh , where the ns has mass 1.4 @xmath52 and no spin , and the bh has @xmath81 with spin @xmath82 aligned with the orbital angular momentum . the uncertainty in both masses is large : this could be an nsbh system containing a bh between @xmath83 and @xmath84 , or an equal - mass bbh system , with masses @xmath85 .
the possible mass ratio of this system extends from 1:1 through to 1:25 . even at snr 20
the mass ratio ranges from 1:2.4 up to 1:11 .
we note that the waveform models used here only include the waveform s dominant harmonic and do not include merger and ringdown ( the phenomenological model is calibrated only up to mass ratio 1:4 ) .
the confidence region may significantly change if higher - order or merger - ringdown effects are included , emphasizing the need for further waveform development in interpreting gravitational - wave observations as well as for detection .
the 90% confidence regions at snrs 10 and 20 for an nsbh system with masses @xmath86 and @xmath87 , and bh spin @xmath88 .
a grb observation would allow an estimate of the bh mass of @xmath89 , and @xmath90 ( see text ) .
, scaledwidth=50.0% ] the observation of a gamma - ray burst or other electromagnetic signal would unambiguously identify one of the components as a ns and break the mass - ratio spin degeneracy . if we accept that the mass of a neutron star in a binary is @xmath91 @xcite , then the bh mass is restricted to @xmath89 . the further assumption that the neutron - star spin is below 0.4 restricts the possible values of the black - hole spin , allowing an estimate of @xmath90 .
this is shown in fig .
[ fig : bhspin ] ( inset ) , which shows the 90% confidence region restricted to the @xmath92-@xmath69 plane , and the region for which the neutron - star mass is between @xmath93 and @xmath38 . however , due to beaming only a fraction of gravitational - wave observations of bns and nsbh systems are likely to be accompanied by grb observations @xcite .
although gravitational - wave observations will accurately measure the chirp mass @xmath94 of binary mergers , a mass - ratio spin degeneracy prevents the component masses and spins being measured accurately at low snrs . in many cases it will be difficult to determine whether the components of the binary are neutron stars or black holes
however , we have illustrated several cases where significant results can be inferred from gravitational - wave observations .
we illustrate several situations where the binary must contain ( at least one ) compact object that is more or less massive than anything observed to date ; @xmath95 indicates a neutron star smaller than 1@xmath52 , @xmath96 a neutron star larger than 2@xmath52 ( if we can independently verify that it is a bns system ) , and @xmath97 a black hole larger than 35@xmath52 .
observations at snrs higher than 30 will be required to clarify the existence of the mass gap .
the observation of an electromagnetic counterpart will , in certain situations , allow us to identify the system as an nsbh and , if we know the distribution of neutron - star masses , measure the black - hole mass and spin with high accuracy .
our results are qualitatively robust , but the true confidence intervals may vary ; they were generated by calculating waveform mismatches @xcite , rather than full parameter - estimation methods @xcite . in some cases
inclusion of merger - ringdown , higher - order post - newtonian corrections to spin effects , particularly at high mass ratios @xcite , and tidal effects will slightly alter the size of the error regions .
the issues only serve to highlight the urgency of improved waveform models .
the inclusion of precession and higher harmonics will modify our results , but we do not expect these to significantly break the degeneracy we discuss @xcite .
our results highlight the significance of the mass - ratio spin degeneracy in gravitational - wave observations of compact binaries .
this needs to be explored further using more complete parameter - estimation methods , more accurate waveform models , and extended to include an exhaustive study of the parameter space .
results from a population of signals , rather than individual observations , should also be investigated .
understanding the uncertainties in masses and spin measurements will be essential to interpreting gravitational - wave observations .
we thank harald pfeiffer , mark scheel and patrick sutton for helpful discussions .
db is supported by nsf award phy-0847611 and an rcsa cottrell scholar award .
sf is supported by the royal society .
mh is supported by stfc grants st / h008438/1 and st / i001085/1 .
ih is supported by nsf grants phy-0847611 and phy-1205835 .
this work was funded in part under the auspices of the u.s .
dept . of energy , and supported by its contract w-7405-eng-36 to los alamos national laboratory .
we thank the kavli institute for theoretical physics at uc - santa barbara , supported in part by nsf grant phy11 - 25915 , where this work was conceived . | gravitational - wave observations of compact binaries have the potential to uncover the distribution of masses and spins of black holes and neutron stars in the universe .
the binary components physical parameters can be inferred from their effect on the phasing of the gravitational - wave signal , but a partial degeneracy between the components mass ratio and their spins limits our ability to measure the individual component masses . at the typical signal amplitudes expected by the advanced laser interferometer gravitational - wave observatory ( signal - to - noise ratios between 10 and 20 ) ,
we show that it will in many cases be difficult to distinguish whether the components are neutron stars or black holes .
we identify when the masses of the binary components could be unambiguously measured outside the range of current observations : a system with a chirp mass @xmath0 m@xmath1 would unambiguously contain the smallest - mass neutron star observed , and a system with @xmath2 must contain a black hole .
however , additional information would be needed to distinguish between a binary containing two @xmath3 m@xmath1 neutron stars and an exotic neutron - star black - hole binary .
we also identify those configurations that could be unambiguously identified as black - hole binaries , and show how the observation of an electromagnetic counterpart to a neutron - star black - hole binary could be used to constrain the black - hole spin . |
observations of the galactic center ( gc ) reveal a cluster of about 20 stars , mainly main - sequence b stars , that extends outward to about a tenth of a parsec from the central supermassive black hole ( sbh ; * ? ? ?
these stars , usually referred to as `` s - stars , '' follow orbits that are randomly oriented and have a nearly `` thermal '' distribution of eccentricities , @xmath6 @xcite .
the existence of such young stars so close to the gc sbh challenges our understanding of star formation since the strong tidal field of the sbh should inhibit the collapse and fragmentation of molecular clouds @xcite . for this reason
, it is usually assumed that the s - stars formed elsewhere and migrated to their current locations . however , the migration mechanisms proposed in the literature result in orbital distributions that differ substantially from what is observed .
post - migration dynamical evolution due to gravitational interactions with other stars or stellar black holes ( bhs ) has been invoked to bring the predicted orbital distributions more in line with observations ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
the s - stars approach closely enough to sgr a * that relativistic corrections to their equations of motion can be important . in this paper
, we apply recent insights about how relativity interacts with newtonian ( star - star ) perturbations near schwarzschild and kerr sbhs . using an approximate hamiltonian formulation that includes a post - newtonian description of the effects of relativity
, we explore the evolution of the s - star orbits starting from initial conditions that correspond to the different formation models proposed in the literature .
evolving the initial conditions for a time of the order the lifetime of the s - stars , and comparing with the observed distribution of orbital elements , allows us to place constraints on both the parameters of the nuclear cusp and the s - star origin models .
timescales of interest are of order 100 myr , the main - sequence lifetime of a b star , or less .
such times are short compared with two - body ( non - resonant , nr ) relaxation times near the center of the milky way ( e.g. , * ? ? ?
* ; * ? ? ?
* ) , hence we ignore nr relaxation in what follows and assume that orbital energies , i.e. semi - major axes @xmath7 , are unchanged once a star has been deposited near sgr a@xmath8 .
resonant relaxation ( rr ) @xcite acts to change orbital eccentricities in a time @xmath9 the `` incoherent rr time , '' where @xmath10 is the angular momentum of a circular orbit having the same semi - major axis as the test star and @xmath11 is the `` coherence time , '' defined as the time for a typical field - star to change its orbital orientation ; the latter is the shortest of the mass precession time ( due to the distributed mass ) , the relativistic precession time ( due to the 1pn corrections to the newtonian equations of motion ) , and the time for rr itself to reorient orbital planes . for instance , in the case that field - star precession is dominated by relativity , @xmath12 where @xmath13 , @xmath14 is the orbital period of the test star , @xmath15 is the number of field - stars with semi - major axes less than @xmath7 , @xmath16 is the mass of the field stars , and mpc is milliparsecs ; @xmath17 has been assumed .
rr ceases to be effective at changing the eccentricities of stars whose orbits lie below ( at higher eccentricities than ) the `` schwarzschild barrier '' ( sb ) , the locus in the ( @xmath18 ) plane where relativistic precession of the test star acts in a time shorter than the time for the field - star torques to change @xmath19 .
the sb is defined approximately by @xcite @xmath20 orbits above ( at lower @xmath21 than ) the sb evolve in response to rr by undergoing a random walk in @xmath21 .
if such an orbit `` strikes '' the sb , it is `` reflected '' in a time of order the coherence time and random - walks again to lower @xmath21 , in a time @xmath22 , before eventually striking the sb again etc . penetration of the sb from above can occur but only on a timescale that is longer than both the rr and nrr timescales @xcite . if a star should find itself _ below _ the sb , torques from the field stars are still able to change the orientation of its orbital plane ( `` 2d rr '' ) even though changes in eccentricity are suppressed .
the timescale for changes in orientation is @xmath23 where again @xmath17 has been assumed .
however , 2drr itself ceases to be effective for orbits that come sufficiently close to the sbh , where dragging of inertial frames by a spinning sbh induces lense - thirring precession with a period that is shorter than the time for 2drr to randomize orbital planes .
the condition for an orbit to be in this regime is @xcite @xmath24 with @xmath25 the dimensionless spin of the sbh , and @xmath26 the sbh spin angular momentum .
we define @xmath27 , the `` radius of rotational influence '' of the sbh , as the value of @xmath7 that satisfies equation ( [ equation : ak ] ) with @xmath28 ; @xmath27 is roughly @xmath29 mpc for the milky way assuming @xmath30 @xcite .
lllllll & & & p@xmath31 & & & tds@xmath32(% ) + + binary disruption + burst scenario@xmath33 & 5myr & 20myr & 50myr & 100myr & 200myr + @xmath34 & @xmath35 & @xmath36 & @xmath37 & @xmath38 & @xmath39 & 0 + @xmath40 & @xmath41 & @xmath42 & @xmath43 & @xmath44 & @xmath45 & 1.9 + @xmath46 & @xmath47 & @xmath48 & @xmath49 & @xmath50 & 0.201 & 0.36 + + migration from + gaseous disk@xmath33 & 5myr & 20myr & 50myr & 100myr & 200myr + @xmath34 & @xmath51 & @xmath52 & @xmath53 & @xmath54 & @xmath55 & 0 + @xmath40 & @xmath56 & @xmath57 & @xmath58 & @xmath59 & @xmath60 & 0.12 + @xmath46 & @xmath61 & @xmath62 & @xmath63 & @xmath64 & @xmath65 & 0.41 + + binary disruption + continuous scenario@xmath66 & 5myr & 20myr & 50myr & 100myr & 200myr + @xmath34 & @xmath67 & @xmath68 & @xmath69 & @xmath70 & @xmath71 & 0 + @xmath40 & @xmath72 & @xmath73 & @xmath74 & @xmath75 & @xmath76 & @xmath77 + @xmath46 & @xmath78 & @xmath79 & @xmath80 & @xmath81 & @xmath82 & 0 + + + + + while the joint evolution of an ensemble of stars near a spinning sbh can only be convincingly treated using an @xmath0-body code , monte - carlo algorithms have been constructed that faithfully reproduce the eccentricity evolution of single ( test ) stars due to the dynamical mechanisms described above , assuming that @xmath83 for the field - star distribution is not evolving . in this paper
, we use an algorithm similar to that described by @xcite .
the hamiltonian that defines the test - star motion includes terms representing the effects of the spherically - distributed mass ( which results in precession of the argument of periastron ) , 1pn ( schwarzschild ) , 1.5pn ( lense - thirring ) precession due to relativity and dipole and quadruple order terms representing the torquing due to the finite-@xmath0 asymmetry in the field - star distribution ( which induces changes in all the orbital elements ) .
the direction of the torquing field is changed smoothly with time and is randomized in a time of @xmath11 , as described in vb of @xcite .
above the sb , where test - star precession times are comparable with typical field - star precession times , the assumptions underlying the derivation of resonant relaxation are satisfied and the algorithm correctly reproduces the eccentricity evolution predicted by rr , as well as the `` bounce '' observed in @xmath0-body integrations when an orbit strikes the sb . for a test star that finds itself _ below _ the sb ,
the schwarzschild precession time is short compared with field - star precession times . in this regime ,
a test star precesses with period close to @xmath84 the 1pn apsidal precession time . during one precessional period ,
the field - star torques are nearly constant ; as a result , the test - star s angular momentum oscillates with period @xmath85 and with approximate amplitude ( in the small-@xmath86 limit ) [ equation : deltal ] @xmath87 @xcite . here ,
@xmath88 , @xmath89 is the mass in stars at radii @xmath90 , and @xmath91 specifies the inclination of the major axis of the torquing potential with respect to the orbit . by themselves , these periodic variations in @xmath21 do not imply any directed evolution in angular momentum , but random switching of the direction of the torquing potential does result in a random walk in a test star s angular momentum , allowing a star that is initially below the sb to approach it .
dragging of inertial frames results in orbit - averaged rates of change of the argument of periastron , @xmath92 , and the angle of nodes , @xmath93 , of the test star according to : [ equation : omegaomega ] @xmath94 in equations ( [ equation : omegaomega ] ) , the `` reference plane '' for @xmath93 and for @xmath95 ( the orbital inclination ) is the sbh equatorial plane . since little
is known about the distribution of stars and stellar remnants near the galactic center , we explored a range of different models for the field - star distribution . assuming a power - law density profile , the number of stars at radii less than @xmath96 is @xmath97 where @xmath98 .
the parameters @xmath99 then uniquely define the background distribution in which the test particle orbits are evolved . in two models
, we set @xmath100 , and we take either @xmath101 and @xmath102 ( equation ( 5.247 ) in * ? ? ?
* ) , similar to what is inferred from observations ( e.g. , * ? ? ?
* ) , or @xmath103 and @xmath104 ( equation ( 5.246 ) in * ? ? ?
* ) , the expected values for a dynamically - relaxed population of stars @xcite . in another model ,
we adopt @xmath105 and @xmath106 .
this latter choice of parameters approximately reproduces the density of stellar bhs predicted by collisionally relaxed models of a cusp of stars and stellar remnants around sgra * @xcite .
figure [ fig1 ] plots the s stars on the ( @xmath18 ) plane , as well as the location of the sb ; the latter depends on the parameters defining the nuclear cusp through equation ( [ sb ] ) .
dot - dashed lines in the figure delineate the region where frame - dragging torques from a spinning sbh would dominate stellar torques , equation ( [ equation : t2drr ] ) . particularly for large @xmath107 , many of the s stars lie close to this transition region , suggesting that frame dragging could be an important influence on their orbital evolution ; for instance , by inhibiting 2drr . this figure does not give information about timescales , but we note that characteristic times like @xmath108 are functions of the nuclear parameters , which can be important given the limited lifetimes of the s stars . in the two models with a steep cusp , the sb delineates the boundary of the s - star orbits , with only a few stars lying below the minimum @xmath109 for which the barrier exists . in these models ,
the existence of the sb is expected to strongly influence the eccentricity evolution of at least some of the s stars ; for instance , by limiting the maximum eccentricity attainable by a star that starts above the barrier .
furthermore , for some of the nuclear models , figure [ fig1 ] shows that some of the s - stars can lie both above @xmath110 and below the sb .
such a location would be highly unlikely , in a time as short as @xmath111myr , for stars that started above the sb , but is reasonable if the stars were placed initially on such orbits via one of the mechanisms described below .
we considered two models for the origin of the s stars . *
formation of the s - stars in binaries far from the center ( @xmath112pc ) . in this model
, the binaries are scattered onto low - angular - momentum orbits that bring them close enough to the sbh that an exchange interaction can occur , leaving one star on a tightly - bound orbit around sgra * @xcite .
the radius at which the sbh tidally disrupt a binary is typically a few tens of au for main sequence binaries , and the orbital eccentricity is expected to be large .
the initial orbital inclinations of the s - stars will be either randomly distributed if the binaries originated in an isotropic stellar cusp @xcite or highly correlated if they formed in a stellar disk @xcite . *
formation of the s - stars in a disk at roughly their current radius , either one of the known stellar disks , or a pre - existing one .
formation in the disk would be followed by migration to their current locations .
this model predicts initially small eccentricities and inclinations @xcite .
@xmath113{figure3a.ps } & \includegraphics[width=.25\textwidth , angle=0]{figure3b.ps } & \includegraphics[width=.25\textwidth , angle=0]{figure3c.ps } \\
\includegraphics[width=.27\textwidth , angle=0]{figure3d.ps } & \includegraphics[width=.25\textwidth , angle=0]{figure3e.ps } & \includegraphics[width=.25\textwidth , angle=0]{figure3f.ps } \\
\includegraphics[width=.27\textwidth , angle=0]{figure3g.ps } & \includegraphics[width=.25\textwidth , angle=0]{figure3h.ps } & \includegraphics[width=.25\textwidth , angle=0]{figure3i.ps } \end{array}$ ]
we start by assuming that @xmath114 is known : it is given by the observed values of @xmath7 . for each s - star ( i.e. for each value of @xmath7 ) , 100 monte - carlo experiments were carried out using the hamiltonian model described above , in each of the three nuclear models , for an integration time of @xmath115myr .
the initial orbital eccentricities were assigned from a thermal distribution , @xmath116 , over some specified range in @xmath21 . in the case of migration from a gaseous disk we required @xmath117 initially . when considering the binary disruption model , we only considered orbits with initial eccentricities in the range @xmath118 . in this case , we assume either that the s - stars are brought to their current location at the same time ( _ burst scenario _ ) or that they arrive at random times between @xmath119 and @xmath115myr ( _ continuous scenario _ ) .
the former choice corresponds to a burst of s - star formation , for instance in a stellar disk @xcite , while the latter choice assumes that the s - stars form continuously in the isotropic stellar cusp @xcite . in all cases we set @xmath30 .
stars were assumed to be tidally disrupted when they approached the sbh within a distance @xmath120 @xcite , with @xmath121 and @xmath122 .
figure [ fig2 ] compares the initial and final ( after @xmath123 myr ) @xmath124 distributions . in the two ,
steep - cusp models with high initial @xmath21 , most of the stars start off to the left of the sb , where evolution in angular momentum is strongly suppressed by the schwarzschild precession .
nevertheless , it can be seen that , after some time , an initially eccentric population separates into two subpopulations : stars that are so far leftward of the sb initially that their eccentricities hardly evolve ; and stars that either begin rightward of the sb , and remain there , or that are close enough initially to the barrier to cross it .
the latter stars are subject to rr after crossing the barrier and end up with a nearly thermal eccentricity distribution .
a clear `` gap '' between the two populations is evident in several of the frames of figure [ fig2 ] ; the gap extends from the sb on the right , to a somewhat higher eccentricity on the left .
the sb acts like a membrane that is permeable in one direction only , from left to right : having crossed the barrier from left to right , a star moves quickly ( in a time of @xmath22 ) to a region of lower @xmath21 where it remains -body integrations . ] . the observed evolution below the sb is due to reorientation of the torquing potential which results in a random walk in angular momentum , preferentially toward lower @xmath21 ( see vb of * ?
* ) ; this mechanism is qualitatively similar to resonant relaxation but obeys a different set of relations @xcite . on the other hand ,
when eccentricities are initially low , as in the bottom panels of figure [ fig2 ] , they tend to remain low , i.e. , above the sb .
as noted in @xcite , the permeability of the sb tends to increase near @xmath110 and this can be seen in figure [ fig2 ] as well .
time evolution of the cumulative distribution of eccentricities is shown in figure [ fig3 ] .
( in this figure and in the analysis that follows , we do not include s - stars that likely belong to the disk(s ) of o / wr stars , i.e. s66 , s67 , s83 , s87 , s96 , s97 ) .
eccentricity distributions were found to approach a nearly thermal " form in a time of order @xmath125 . based on figure [ fig3 ] , we see that the lifetime of the s - stars may , or may not , be long enough for this to happen , depending on the nuclear model .
we compared the results of the monte - carlo integrations with s - star data by performing 2-sample kolmogorov - simornov ( k - s ) tests on the @xmath21 distributions ( table [ t1 ] ) .
the best match to observations is attained after @xmath126myr of evolution in _ stellar _ cusp models in the continuous scenario with @xmath103 and starting from initially high eccentricities ( k - s test @xmath127-values of @xmath128 ) .
integrations that include stellar bhs also generate orbital distributions which are in agreement with observations after approximately @xmath129 and @xmath126myr of evolution for initially high and low eccentricity distributions respectively .
we tested the degree of randomness of the orbital planes using the rayleigh statistic @xmath130 @xcite , defined as the resultant of the unit vectors @xmath131 , where @xmath132 is perpendicular to the orbital plane of the @xmath133 star and @xmath134 is the total number of monte - carlo data points ( i.e. , @xmath135 ) .
since the test stars were initialized with the same inclination , the orbits are initially strongly correlated and @xmath136 ; over a time of order @xmath137 , 2drr tends to randomize the orbital planes and @xmath130 approaches @xmath138 , the value expected for a random distribution . when the main contribution to dynamical relaxation comes from stellar bhs , @xmath130 reached values consistent with isotropy after @xmath139 and @xmath140myr for initially high and low eccentricity distributions respectively . in the _ stellar_-cusp model with @xmath103 and staring from initial high @xmath21
, @xmath130 reached values consistent with isotropy after @xmath111myr , while in all the other stellar - cusp models we measured a significant departure from randomness ( @xmath141 ) even after @xmath115myr of evolution .
we estimated @xmath130 separately for orbits that any time were below the dot - dashed lines of figure [ fig2 ] and found that this population had a distribution of orbital planes which was less consistent with being random .
evidently , 2drr was somewhat inhibited by frame dragging for orbits with initially large eccentricities .
the fraction of stars that would have been tidally disrupted after @xmath115myr of evolution was never larger than @xmath142 ( table [ t1 ] ) .
as a comparison , in purely newtonian integrations , @xcite found that up to @xmath143% of stars were disrupted after @xmath126myr of evolution .
in this paper we studied the combined effects of newtonian and relativistic perturbations on the angular momentum evolution of the galactic center ( gc ) s - stars . for the first time we have shown that the @xmath124 distribution of the s - stars predicted by the binary disruption model , in which the stars are delivered to the gc on high - eccentricity orbits , is consistent with the observed orbits even when
relativistic effects are considered . in these formation models ,
most of the orbits lie initially below the schwarzschild barrier , the locus in the ( a , e ) plane where resonant relaxation is ineffective at changing eccentricities .
contrary to this basic prediction , we found that orbits starting sufficiently close to the barrier are sometimes able to penetrate it , diffusing above and reaching a nearly thermal e - distribution ; a small fraction of stars remain confined below the barrier at low angular momenta ( @xmath144 ) .
a good match to observations is achieved after @xmath145myr of evolution if the distribution of field stars at the gc follows a nearly dynamically relaxed form .
models that include a mass - segregated population of stellar bhs also generate after @xmath146myr of evolution distributions that are marginally consistent with observations . based on the origin models considered here
, the s - star obits can only be reproduced by postulating dynamically relaxed states ( i.e. , steep density cusps ) for the gc .
this result is interesting given that such models are currently disfavored by observations @xcite and by some theoretical arguments @xcite .
alexander , t. , & merritt , d. 2012 , in preparation , f. , faber , j. , gualandris , a. , & merritt , d. 2010 , , 713 , 90 antonini , f. , lombardi , j. c. , jr . , & merritt , d. 2011a , apj , 731 , 128 , f. , capuzzo - dolcetta , r. , mastrobuono - battisti , a. , & merritt , d. 2011b , apj , 750 , 111 antonini , f. & merritt , d. 2012 , apj , 745 , 83 bahcall , j. n. , & wolf , r. a. 1977 , , 216 , 883 buchholz , r. m. , schdel , r. , & eckart , a. 2009 , a&a , 499 , 483 do , t. , ghez , a. m. , morris , m. r. , lu , j. r. , matthews , k. , yelda , s. , & larkin , j. 2009 , apj , 703 , 1323 hills , j. g. 1988 , nature , 331 , 687 hopman , c. , & alexander , t. , 2006 , apj , l645 ghez , a. m. , et al .
2008 , apj , 689 , 1044 gillessen , s. , eisenhauer , f. , trippe , s. , alexander , t. , genzel , r. , martins , f. , 2009 , apj , 692 , 1075 gualandris , a. , & merritt , d. 2012 , apj , 744 , 74 levin , y. 2007 , mnras , 374 , l515 madigan , a. , levin , y. , & hopman , c. 2009 , apj , 697 , l44 madigan , a. , hopman , c. , & levin , y. 2011 , apj , 738 , 99 merritt , d. , gualandris , a , & mikkola , s. 2009 , , 693 , 35 merritt , d. , 2010 , apj , 718 , 739 merritt , d. , alexander , t. , mikkola , s. , & will , c. 2010,phrvd , 81 , 062002 merritt , d. , alexander , t. , mikkola , s. , & will , c. 2011,phrvd , 84 , 4024 merritt , d. , 2013 , dynamics and evolution of galactic nuclei ( princeton : princeton university press ) , p. 337 .
merritt , d. , & vasiliev , e. 2012 , arxiv:1208.6274 morris , m. 1993 , apj , 408 , 496 perets , h. b. , hopman , c. , & alexander , t. 2007 , apj , 656 , 709 perets , h. b. , gualandris , a. , kupi , g. , merritt , d. , & alexander , t. 2009 , apj , 702 , 884 perets , h. b. , gualandris , a. , 719 , 220 rauch , k. p. , & tremaine , s. 1996 , new astronomy , 1 , 149 rayleigh , l. 1919 , phil . mag .
, 37 , 321 zhang , f. , lu , y. , & yu , q. 2012 , arxiv:1210.1901 yu , q. , & tremaine , s. 2003 , apj , 599 , 1129 | we consider the orbital evolution of the s - stars , the young main - sequence stars near the supermassive black hole ( sbh ) at the galactic center ( gc ) , and put constraints on competing models for their origin .
our analysis includes for the first time the joint effects of newtonian and relativistic perturbations to the motion , including the dragging of inertial frames by a spinning sbh as well as torques due to finite-@xmath0 asymmetries in the field - star distribution ( resonant relaxation , rr ) .
the evolution of the s - star orbits is strongly influenced by the schwarzschild barrier ( sb ) , the locus in the ( @xmath1 ) plane where rr is ineffective at driving orbits to higher eccentricities .
formation models that invoke tidal disruption of binary stars by the sbhtend to place stars below ( i.e. , at higher eccentricities than ) the sb ; some stars remain below the barrier , but most stars are able to penetrate it , after which they are subject to rr and achieve a nearly thermal distribution of eccentricities .
this process requires roughly @xmath2myr in nuclear models with relaxed stellar cusps , or @xmath3myr , regardless of the initial distribution of eccentricities , in nuclear models that include a dense cluster of @xmath4 black holes .
we find a probability of @xmath5 for any s - star to be tidally disrupted by the sbh over its lifetime . |
ultra - luminous x - ray sources ( ulxs ) are extra - nuclear sources with luminosities in the range of @xmath9 erg / sec in other galaxies , and have been observed by rosat , asca , recently by xmm - newton and chandra observatory in large numbers . as compared to the cases of the x - ray binaries in our galaxy , which are powered by accretion onto neutron stars or stellar mass black holes and
have luminosities of @xmath10 erg / sec , the luminosities of ulxs require accreting compact objects of masses @xmath11
@xmath12 @xmath2 if they emit at @xmath13 of the eddington luminosity , typical of galactic x - ray binaries . while the required masses could be much larger if they emit at much less than @xmath13 of the eddington luminosity , as in the cases of some low luminosity active galactic nuclei ( agn ) , the masses can not be much greater than @xmath14 @xmath2 for these extra - nuclear sources to survive the dynamic friction over a few gigayears ( colbert 1999 ) .
such intermediate mass compact objects can only be black holes , and if they exist , are the missing links between stellar mass black holes and super - massive black holes in the nuclei of galaxies . while the explanation with intermediate mass black holes is simple , intriguing and astrophysically interesting , such black holes are not predicted by ordinary stellar evolution theories . it is suggested that black holes of a few hundred @xmath2 can form from the death of pop iii stars , and more massive ones might form from multiple stellar interactions in some dense clusters , hencing manifest as ultra - luminous x - ray sources ( portegies zwart et al .
alternatively , these x - ray sources could be stellar mass black holes or neutron stars whose apparent super - eddington luminosities are due to some special processes .
one suggestion is that radiation pressure - dominated accretion disks with photon - bubble instabilities are able to emit truly super - eddington luminosities ( begelman 2002 ) .
another suggestion is that beaming effects can produce the observed luminosities of ulxs ( king et al .
2001 ) .
the leading goals in ulx studies are to determine the masses of the primary , to understand how and where they form , and to find out how they emit at such high luminosities . in the last few years
many efforts have been made to address these questions , and important clues have been revealed .
however , these studies mainly focus on some well - known objects and galaxies ( e.g. , m81 x-9 , ngc5204 x-1 , antenna galaxy , cartwheel galaxy ) except for a few works ( e.g. , colbert & ptak , 2002 , hereafter cp2002 ; swartz et al .
2004 ; ptak & colbert 2004 ) , and now it is time to define a complete sample of ulxs to study the properties of the phenomenon and test critical ideas .
one resource to rely on for this purpose is the data archive of the rosat high resolution imager ( hri ) , which includes 5403 observations in the ten years from june 1990 to february 1999 .
the hri is a multichannel plate detector of @xmath15 square field of view , large enough to contain all individual nearby galaxies other than lmc , smc , m31 , and m33 in our local group .
its spatial resolution is suitable for extra - galactic point source studies , with on - axis fwhm @xmath16 and a pixel scale of @xmath17/pixel , adequate to resolve point sources in most cases .
also , the archival observations have reasonable sky coverage for survey purposes . for example , the 5393 pointed hri observations used in the first rosat hri source catalog ( 1rxh ; rosat scientific team , 2000 ) covers about 2% of the sky .
the large database of rosat hri observations has not been thoroughly exploited for complete samples of ulxs in nearby galaxies .
roberts & warwick ( 2000 ; hereafter rw2000 ) have used the hri archive to study the x - ray properties of nearby galaxies , and detected in 83 galaxies 187 discrete x - ray sources of all luminosity ranges , among which 27 non - nuclear sources have @xmath18@xmath4@xmath0 erg / sec and can be taken as ulxs .
they used the nearby galaxy sample by ho , filippenko & sargent ( 1995 ) , which was constructed to search for dwarf seyfert nuclei in nearby galaxies and contains 486 bright northern galaxies .
many nearby galaxies with hri observations are not included in this sample , and the hri archive is far from being fully utilized for surveying ulx .
recently colbert & ptak ( 2002 ) made an effort to analyze the hri observations to search for ulxs in a sample of 9999 galaxies in the third reference catalog of galaxies ( rc3 ; de vaucouleurs et al .
1991 ) with @xmath19@xmath20@xmath21 km / sec .
they found 87 ulxs in 54 galaxies , with 37 in early - type galaxies .
however , many ulxs in the cp2002 catalog are projected far from the host galaxies , and may be false ulxs from foreground stars or background agn / qsos .
for example , irwin et al .
( 2004 ) pointed out that the radial distribution of ulxs in early - type galaxies in the cp2002 catalog is consistent with a random distribution , thus these ulxs are probably not physically associated with these early - type galaxies . here
we present our study of ulxs in nearby galaxies with the wealth of hri archive . to fully utilize the hri archive , we choose all observations associated with any rc3 galaxies within 40 mpc with isophotal diameters @xmath41 arcminute .
the rc3 galaxy sample , the selected hri observations and the survey galaxies are described in section 2 . in our analysis
a wavelet algorithm is used for point source detection , and in section 3 we discuss its performance on hri images through simulations . in section 4
, we describe the analysis procedures applied on the data , including the point source detection , the variability tests , astrometric corrections , associations with galaxies and computation of luminosities .
these efforts lead to a catalog of 562 extragalactic x - ray sources above @xmath22 detection within @xmath23 of 173 out of the 313 survey galaxies , including 106 ulx candidates within the @xmath3 isophotes of 63 galaxies and 110 ulx candidates between 12 @xmath8 of 64 galaxies .
the identifications of these sources are given in section 5 .
as a final product , we construct a clean sample of ulxs to include ulxs within , and a few outside the @xmath3 isophotes , with identified foreground stars and background qsos excluded .
the comments and finding charts are given for these ulxs in section 6 . in section 7
we summarize the procedures and preliminary results of the survey . in a subsequent paper
, we will study the occurrence rates and luminosity functions for ulxs in different types of galaxies .
to search for the ultra - luminous x - ray sources in the nearby galaxies from the rosat hri archival observations , we cross correlated a nearby galaxy sample with the list of hri archival observations to select observations with nearby galaxies within their field of views . here
we describe the nearby galaxy sample , the selected hri observations , and the survey galaxy sample .
the nearby galaxy sample is extracted from rc3 , one of the most complete catalog for nearby galaxies having apparent diameters larger than 1 arcminute at the @xmath3 isophotal level , total b - band magnitudes @xmath24 brighter than 15.5 , and redshifts not in excess of 15,000 km / sec .
distances to galaxies are collected from the literature , including the hst key project ( designated as kp ; freedman et al .
2002 ) , the surface brightness fluctuation method ( sbf ; tonry et al .
2001 ) , the nearby galaxy flow model by tully ( 1992 ; t92 ) , and the nearby galaxy catalog ( t88 ; tully 1988 ) .
distances are also computed using the hubble relation @xmath25 with @xmath26 km / s / mpc .
the ned service provides the recessional velocity @xmath27 for most galaxies , and distances computed from them are designated as ned .
another source of @xmath27 is rc3 , and the distances computed are designated as v3k , since we use the velocities corrected to the rest frame defined by the 3k cosmic microwave background . when multiple distance measurements are available we use the best distance measurement , which is kp , followed by sbf , t92 , t88 , ned , and v3k .
the positions of the galaxies are taken from rc3 .
ned also gives galaxy positions that are consistent with rc3 to a few arcseconds for most galaxies . in the rare occasions when the positions from the two sources differ by @xmath4@xmath5 , we verify the positions by overlaying the isophotal ellipses on the optical images from the digital sky survey ( dss )
the @xmath3 isophotes , the elliptical contour best corresponding to the 25 mag / arcsec@xmath28 blue isophote , are also taken from rc3 .
the @xmath3 isophote is a quantitative description of the domain of a galaxy , though the galaxies can extend beyond @xmath3 in some cases but usually within @xmath29 as seen from dss images . for our ulx study
, we choose galaxies with isophotal diameters @xmath41 arcminute , for which rc3 is reasonably complete for bright nearby galaxies .
given hri position uncertainties typically below @xmath30 , this choice of galaxy size leaves space for extra - nuclear sources to be distinguished from nuclear ones . for a short hri observation of 10 kiloseconds , the count threshold for 3@xmath31 detection for on - axis point sources
is @xmath32 photons based on the simulations in section 3 , and for the x - ray spectrum used in section 4 this corresponds to a flux level of @xmath33 erg / sec / cm@xmath28 , which in turn corresponds to a luminosity of @xmath35 erg / sec for sources in galaxies at 10 mpc .
sources with extremely high luminosities , e.g. , @xmath18@xmath4@xmath36 erg / sec , are usually very revealing about the nature of ulxs , and can still be detected in galaxies at larger distances . here
we choose to survey galaxies within 40 mpc , with the hope of detecting these extreme ultra - luminous x - ray sources from the large number of galaxies between 10 mpc and 40 mpc .
galaxies in our local group usually have large projected areas and have severe contamination problems from foreground stars , we thus exclude them from our study here .
our above selection criteria lead to a sample of 4434 rc3 galaxies .
a complete list of rosat hri observations was extracted from the mpe rosat site .
the list of hri observations with exposures @xmath41000 seconds was cross correlated with the above rc3 galaxy list with a correlation radius of @xmath37 .
galaxies only partly observed by the hri are excluded from our survey for their incompleteness and large off - axis angles .
also excluded are galaxies behind the galactic plane or the magellanic clouds for their dense foreground stellar fields to reduce the confusion problems for identifying x - ray sources . in total ,
our survey includes 467 observations for 313 galaxies .
these observations were downloaded from the mpe ftp site and processed with the uniform procedures described in section 4 . in the hri galaxy sample ,
313 galaxies were surveyed by 467 hri observations , about 7.1% of the rc3 galaxy sample .
this fraction is comparable to that in the work of ptak & colbert ( 2004 ) , in which 766 ( @xmath388% ) of the 9452 rc3 galaxies with @xmath19@xmath20@xmath21 km / sec are surveyed by hri observations . in table 1
the survey galaxies are listed with galactic positions , sizes , the galactic hi column densities , the distances , the galaxy types , the b - band , the iras 60@xmath39 m and fir luminosities .
this hri galaxy sample is compared to the rc3 galaxy sample to reveal possible preferences in the survey galaxies .
the blue luminosity distribution of the survey galaxies resembles that of the rc3 galaxy sample to a large extent , as shown in figure 1 , except for a slight over - sampling of very bright galaxies .
this over - sampling is related to the fact that the galaxies surveyed by the hri tend to be larger and closer ones , as demonstrated in figure 2 and figure 3 .
the morphological type distributions of the galaxies are compared between these two samples , and an over - abundance of ellipticals , and over - abundances of lenticulars and s0/a sc early spirals to a lesser extent , are found in the hri galaxy sample with respect to the rc3 galaxy sample as shown in figure 4 .
in contrast , there is an under - abundance of dwarf spirals and irregulars in the survey sample , reflecting a bias in selecting the hri targets that is against the dwarf spirals and irregulars .
to summarize , the hri survey galaxy sample is representative for the galaxies within 40 mpc , although there are preferences for galaxies that are larger , closer , optically or x - ray brighter .
a point source detection method suitable for the hri images is needed to study the x - ray point sources in hri observations .
two packages for two detection algorithms are widely used in the x - ray astronomy community .
one is celldetect , which uses the sliding cell method to check for local maxima and detect unresolved sources .
this was the method adopted by 1rxh .
another popular package is wavdetect , which employs a wavelet method to detect point sources . for our purpose of detecting x - ray point sources in the hri images , it is more effective to use wavdetect rather than celldetect , since celldetect does not treat diffuse emission well , and tends to split diffuse emission into multiple point sources .
for example , celldetect reports 67 point sources for the 10 kilo - second hri observation rh600463a02 of an elliptical galaxy ngc 4406 , while there is really only one nuclear source plus diffuse x - ray emission , as correctly reported by wavdetect .
the wavdetect package was developed for use with the chandra observatory data , and can also be used with other x - ray instruments , including the rosat hri . however , when using wavdetect with hri images , we need to address the questions about the false detection rate , the count ( rate ) threshold for @xmath22 detection , and the correction factor between the detected counts and the true counts . here
we study these questions by running wavdetect on simulated hri images . for this study
, the hri images are simulated with the task qpsim in the pros package .
the psf shape is chosen to be roshri , the in - flight psf for the hri detector .
the background rate is set to a constant of 5 count / second , which is typical for hri observations and equivalent to @xmath40 count / second / pixel .
artificial sources are evenly spaced on the images , with off - axis angles ( @xmath41 ) ranging from 0 to @xmath42 .
an exposure time of @xmath43 seconds and a source count are assigned to all artificial sources on one image . for our simulations
, we use a series of source counts from 3 to @xmath12 , and @xmath43 from @xmath11 to @xmath44 seconds . the central 4098 pixels ( @xmath45 for a pixel scale of @xmath46/pixel ) of the simulated hri images
are binned by a factor of four to get a 1024x1024 image to run wavdetect on . during
the wavdetect runs , the parameter of significance threshold ( sigthresh ) is set to 1e-6 for the 1024x1024 images , which would cause about one out of the total 1e6 pixels be erroneously identified with a source .
the result of a wavdetect run depends on the scale set used in search of point sources .
the scales chosen should include the range of source sizes across the image , since a source can not be found by scales much larger or smaller than the source size .
for rosat hri images , the source size changes with the off - axis angle .
one estimate of the source size is @xmath47 , the radius encircling 50% of the total source count , which can be expressed in a closed form as @xmath48^{0.5 } arcsec$ ] , with @xmath49 , @xmath50 , @xmath51 , @xmath52 , and @xmath53 .
this radius changes from an on - axis @xmath54 to @xmath55 at an off - axis angle of @xmath56 . note that the source size is supposed to enclose ( almost ) all source counts , and should be larger than @xmath47 by a factor of a few . here
five sets of scales ( in binned pixels ) , i.e. , ( 1 2 4 8) , ( 2 4 8 16 ) , ( 3 6 12 ) , ( 4 8 16 ) , and ( 5 10 20 ) , are tried to determine the best scale set for hri images .
one indicator for choosing the best scale set is the false detection rate .
a source detected @xmath4@xmath57 away from any artificial sources is defined as a false source .
our simulations show that there are fewer false detections for larger scales . in figure 5
we show the false detection percentages for five scale sets .
while none of the five scale sets give false detections above @xmath58 , the first three scale sets give many more false detections below @xmath58 than the last two .
the total numbers of detected sources from the same hri images also depend on the scale sets , with smaller scale sets detecting fewer sources .
this trend , shown in figure 6 , arises because smaller scales are sensative to smaller source regions and tend to miss more source counts , leading to a lower detection significance or even a non - detection .
these results make the last two scale sets favorable over the first three of smaller scales . here
we choose ( 4 8 16 ) as the best scale set to use in following simulations and data analysis . while the scale set ( 5 10 20 )
is also favorable , we choose ( 4 8 16 ) since scales of @xmath59 are computationally convenient for the wavelet algorithm . for the chosen scale
set , the false detection rate is about 2% for sources with 34@xmath31 , 0.4% for sources with 45@xmath31 , and there is no false detections above 5@xmath31 ( figure 5 ) .
the false detection rate is @xmath410% for detections below 3@xmath31 , which are discarded in our analysis of the hri observations .
an important quantity to determine for an observation is the count threshold for @xmath22 detection . completely different from the parameter of significance threshold in wavdetect
, this quantity could mean the count level at which a source will be detected in a 99.7% chance , or the count level at which a source will be detected with a significance of @xmath22 , i.e. , with a source count equal to 3 times the background error .
though related , these two count levels are not always the same , and in our analysis we use the latter since it is easier to calculate both theoretically and through simulations . by comparing the thresholds from simulations with the theoretical thresholds as a function of the source size and the background rate , one would be able to study how the source size changes across the images , with which the count thresholds across the image of an observation can be computed given its background rate .
the count threshold can be computed theoretically given the background level and the source region .
as adopted by the wavdetect package , the gehrels error @xmath60 of the background count enclosed in a source region with an area of @xmath61 pixels is @xmath62 for a uniform background rate of @xmath63 count / second / pixel and an exposure of @xmath43 seconds .
the source count threshold for the @xmath22 detection is then @xmath64 at large off - axis angles the source region deviates from a circle and is usually described as an ellipse due to the elongation of the psf .
however , for the purpose of computing the background count , only the total area of the source region matters for a uniform background rate .
the background rates in most of the observations in our hri survey are reasonably uniform across the hri images , because the background vignetting is mild for the central @xmath65@xmath66@xmath65 square field of view of the hri detector , and observations with foreground diffuse emission from the galactic plane , lmc and smc are excluded from our survey .
thus , for the purpose of calculating the area , the source region can be approximated as a circle with radius in unit of @xmath47 as @xmath67 , and the source area @xmath68 . here
@xmath69 should be @xmath70 , and its exact value may vary for different off - axis angles .
the count threshold can also be obtained by simulation runs on images of artificial sources spanning a large range of count levels . in our simulations , when the artificial source counts change from the smallest 3 counts to more and more counts , the sources change from being missed , to being detected with 1@xmath31 , to 2@xmath31 , to 3@xmath31 , to a few hundred @xmath31 . in figure 7
we show how this count threshold is determined with a series of count values for certain off - axis angles and background levels .
these sources are sorted in ascending order of counts and connected in the count@xmath31 plane , the point where the 3@xmath31 line crosses this connecting line marks the count value for 3@xmath31 detection . for artificial sources with the same counts that have different detection @xmath31 , two methods are available to sort these sources , one in ascending order of detection @xmath31 and another in descending order of detection @xmath31 . since the two methods may lead to different results , both methods are tried and their average is used as the method - independent count value of 3@xmath31 detection .
the situation can be rather complex , in which there are more than one crossing , i.e. , detections above 3@xmath31 and detections below 3@xmath31 or non - detections alternate in a range of count values due to the complexity of the implementation of the detection algorithm .
in this situation , the simulated count threshold is taken as the average of count values from all crossings .
the count thresholds are obtained from the simulations for a grid of off - axis angles and background levels to determine the dependencies on these two parameters .
for a fixed off - axis angle @xmath41 , the simulated count thresholds change with the background level the same way as thresholds computed for a fixed source region , shown in figure 8 , and consistent with the expectation that the source region is fixed for a fixed off - axis angle . with the source region approximated as a circle and its area computed as @xmath71 , the value of @xmath69 can be obtained by fitting the simulated count thresholds to the count thresholds computed using equation ( 1 ) for that off - axis angle with a least square method . for example , for the off - axis angle of @xmath72 in figure 8 , the best fit @xmath69 is 2.6 , and the equivalent source radius is @xmath73 = @xmath74 . the fitted @xmath69 shows a dependency on off - axis angle @xmath41 that can be fitted to a 4-th polynomial , @xmath75 . with this analytical form , the equivalent radius , thus the area , of the source region can be computed for any off - axis angle , as shown in figure 9 .
the equivalent source radius does not change much within @xmath76 , outside which the source radius increases rapidly with the off - axis angle .
note that the change can be attributed to mostly the change of point source sizes and partly the change of the effective area as described by the vignetting function .
equipped with the source area , the count thresholds can be computed with equation ( 1 ) given the background rate and the exposure time . in figure 10 , the count rate thresholds for @xmath22 detection , or the observation sensitivity , for a few typical exposures are calculated and plotted . thus computed thresholds are consistent with the simulated ones to within 20% .
the detected source counts are usually not the true source counts due to a number of reasons , e.g. , the mismatching between the source region sizes and the wavelet scales used by wavdetect , vignetting and other nonuniformities of the detector . to correct the detected counts to the true counts , one must apply a correction factor ( @xmath77 ) , which we define as @xmath78 .
our simulations here , with known true source count , detection significance and detected source count for each source , provide us a chance to study the correction factor that can be applied to real data .
the simulations reveal strong dependencies of the correction factor on the off - axis angle and the detection significance .
the dependency of the correction factor upon the detection significance for a fixed off - axis angle is shown in figure 11 . for sources with high detection significance , the factor is close to unity and is approximately constant .
it begins to decrease below a certain significance and drops to a few tenth at @xmath79 .
the details of this dependency vary with the off - axis angle @xmath41 . for sources on the edge of the image ,
the factor begins to drop at @xmath80 , and drops to @xmath81 at @xmath79 . for sources at smaller off - axis angles ,
the factor begins to drop at smaller significances , and drops to higher values .
in contrast , there is no significant drop in the factor for on - axis sources . for a fixed off - axis angle @xmath41 and a fixed detection significance @xmath31 ,
sources show slightly different correction factors in the simulations .
an average correction factor @xmath82 can be computed from these simulated sources , with its error representing the dispersion of the simulated factors . based on the simulations ,
average correction factors are computed for a grid of off - axis angles and detection significances . for a source with any combination of @xmath41 and @xmath31 , the correction factor @xmath82
can be interpolated from this grid , and applied to the detected count to obtain the true count .
the interpolated correction factors , plotted as a function of off - axis angles for @xmath83,10 , and 100 in figure 12 , are approximately constant within @xmath76 , and decrease outward with the off - axis angle .
this decrease means more source photons are missed by the used wavelet scales at larger off - axis angles , which can be attributed to the increase of point source sizes and the vignetting function with the off - axis angle .
the interpolated correction factors are tested by applying them to compute `` corrected '' counts from detected counts and comparing to the true counts of the simulated sources .
the results are shown in figure 12 . for sources with a high detection significance , the detected counts can be corrected to within a few percents of the true counts .
for example , the detected counts for 67% ( @xmath84 ) of the simulated sources can be corrected to within 1% of their true counts for detection @xmath85 , i.e. , the ( @xmath84 ) fractional error of the corrected counts is 1% .
this error is 1.2% for sources with @xmath86 , 2.7% for @xmath87 , and 7% for @xmath88 .
the error is considerably larger for sources with lower @xmath31 , e.g. , 12% for @xmath89 , 20% for @xmath90 , and up to 30% for @xmath83 . for sources with @xmath91 , the corrected counts are significantly lower than their true counts .
this systematic inaccuracy is caused by the small number of low @xmath31 sources in the simulation , and will not affect the applicability of the interpolated correction factors on our hri data , since we only work on sources above @xmath22 .
the selected hri observations are processed to extract point sources with uniform procedures in light of the simulations in section 3 .
each hri observation is binned by a factor of 4 into a 1024x1024 image , on which wavdetect is run with the scale set ( 4 8 16 ) , i.e. , ( 8@xmath92 16@xmath92 32@xmath92 ) .
the source list from wavdetect includes the position , detection significance @xmath31 and the detected net count for each source .
the off - axis angle @xmath41 is computed for a source from its position on the image , and @xmath82 is computed based on the simulations to correct the detected count to the true count .
the corrected count rate , computed as the corrected count divided by the exposure time , is converted to flux with the conversion factor computed with pimms by assuming a power - law spectrum between 0.38 kev with a photon index of 1.7 , with the galactic hi column density derived from the hi map by dickey & lockman ( 1990 ) .
such a spectrum , also used by cp2002 , was adopted to enable comparison with recent chandra observations . for each source
, a kolmogorov - smirnov test is also run to check for the variability during the observation . for each observation
, efforts have been made to improve the source positions by cross correlating x - ray sources with sources in optical , radio or infrared .
sources are also cross correlated to galaxy isophotal ellipses to determine whether they are associated with galaxies , and if they are , their luminosities are calculated . in the following subsections
we describe the data analysis procedures , which result in a catalog of point sources in other galaxies ( table 2 ) , and a list of individual observations for ulxs ( table 3 ) .
the temporal variations of an x - ray source can be visualized by its light curve . in our study , the background - subtracted light curves are constructed for bright sources .
the event list for an x - ray source is extracted from the @xmath22 elliptical source region as taken from wavdetect results , which contains 95% of the source count for an assumed 2-d gaussian distribution . in figure 13
we show the light curve for a ulx in the circinus galaxy , in comparison to the scaled background . for this observation , multiple observation intervals ( obis ) of total @xmath93 hours scatter over a @xmath94 day period , and only a few days with significant exposure time
are plotted .
the light curve shows clear variations over these days .
however , some temporal features are missing from the light curves of hri sources due to the discontinuous observation scheme .
the ulx in figure 13 is an x - ray eclipsing source with a period of 7.5 hours as revealed by chandra observations ( bauer et al .
2001 ) , yet the periodicity is not obvious in the hri light curve due to the wide separations between obis . the variability can be described quantitatively with the kolmogorov - smirnov test .
the kolmogorov - smirnov test is a widely used nonparametric method to test whether two distributions differ significantly , and here it is used to test the null hypothesis that the source is constant during an observation .
a source can be viewed as constant if the null hypothesis probability ( listed in table 3 ) is of the order of 1 .
a source can be viewed as variable if the probability is much smaller , for example , @xmath950.1 . to be conservative , we define a source as variable if the null hypothesis probability is @xmath200.01 , i.e. , the source is variable with a significance of @xmath499% .
the timescale for this variability is the duration of the observation , which ranges from a few hours to a few months . in figure 14 , we plot for the above mentioned ulx the cumulative probability curves of the kolmogorov - smirnov tests .
the null hypothesis probability is 11% for the ulx to be constant over the 22 day observation for this source , and the probability for it to be variable is 89% .
we caution here that even if the source is constant by a high probability during obis , the source could be variable during the wide separations between obis .
the positional accuracy of an x - ray source is important in our study for determining whether the source is associated with a galaxy and whether it is a nuclear source , and for identifying the source in other wavelengths . for sources from hri observations ,
the positions usually have an uncertainty of @xmath96 attributable to the smearing and elongation of hri psf due to residual errors in the aspect solutions associated with the rosat wobble or the reacquisition of the guide star ( morse , 1994 ) .
the detected positions may offset from the true positions by up to @xmath97 in rare occasions of very bad aspect solutions .
the astrometric solution for an observation can be improved if some x - ray sources can be identified with objects whose positions have been accurately catalogued . in our study
, the x - ray sources are cross correlated with the two micron all sky survey ( 2mass ; cutri et al . 2003 ) and the faint images of the radio sky at twenty centimeters ( first ) survey catalog ( becker et al .
2003 ) to identify optical and radio counterparts .
the correlation procedure works iteratively by re - correlating after adjusting the source positions based on the identified sources in the previous correlation , until the number of identified sources and the adjusted positions stabilize .
the identifications are visually checked after the iterative procedure stabilizes .
this process was carried out for 416 observations , and the distribution of the offsets @xmath98 between the hri and catalogued positions for all identified sources is plotted in figure 15 . assuming a 2-d normal distribution of the hri positions relative to the catalogued positions , @xmath99 , a least square fit to the plotted distribution results in @xmath100 .
this is consistent with what was found by morse ( 1994 ) . with the identified sources registered to their accurate positions ,
the positions of unidentified sources are corrected using their positions relative to the identified ones on the x - ray image .
such corrections are carried out when there are x - ray sources identified with objects within @xmath101 of their uncorrected positions to reduce the misidentifications .
this process removes the possible large offsets induced by aspect solution errors .
the position error after correction is mainly attributeable to statistical errors of centroiding and plate scale variations . in table 2
the number of identified sources used to correct the position is listed for each source . a number @xmath1020 indicates the source position error is less than @xmath103 . in the rare occasions when this number is zero , the position error could be up to @xmath30 .
note that the identification based merely on positional coincidence may be insecure , and the astrometric correction based on the identifications may be in error .
extra cautions should be exercised when there are only one such identification for an observation , especially when this identification is for the ulx in the obaservation .
after the above astrometric corrections , multiple observations for the same galaxies are aligned to identify the same source in different observations .
the position of a source , as listed in table 2 , is computed from averaging the positions in these observations . if a source is not detected but within the field of view of an observation , the @xmath22 upper limit at its supposed position is computed based on the simulations and listed in table 3 .
one question for studying the ulxs in other galaxies is to determine whether an x - ray source is associated with a galaxy , and further whether it is an extra - nuclear source .
once a source is known to be associated with a galaxy , its luminosity can be calculated from its flux using the distance of the galaxy it is associated with .
the association of a source with a galaxy is determined by means of the blue @xmath3 isophote in our study .
the separation @xmath98 between the galaxy center and the source is computed and compared to the elliptical radius @xmath104 of the @xmath3 isophotal ellipse along the great arc connecting the galaxy center and the source .
this elliptical radius @xmath104 has a minimum value of the length of the semi - minor axis when the source is along the minor axis , and has a maximum value of the length of the semi - major axis when the source is along the major axis .
a source with @xmath105 , i.e. , within the @xmath3 isophote , is considered as associated with the galaxy .
visual inspections of the dss images show that the @xmath3 isophotes are a good delimiter of the optical domain of the galaxies , though galactic features extend apparently beyond @xmath3 , but within 2 @xmath8 of some galaxies .
for example , in ngc1313 there are dusty star forming regions between 1 2 @xmath8 , and in ngc1316 there are dust loops and strips between 1 2 @xmath8 . to not miss any ulxs , a source with @xmath106 is tentatively associated with the galaxy . in follow - up studies for the statistical properties , sources within the @xmath3 isophote and sources between 1
2 @xmath8 are treated differently .
nuclear x - ray sources powered by accretion onto central supermassive black holes of galaxies should be excluded from the class of ulxs . in this work
the source positions after astrometric corrections are better than @xmath101 , and the positions of optical nuclei are better than a few arcseconds for most galaxies
. an offset limit of @xmath30 is thus chosen to tentatively identify nuclear and non - nuclear sources .
the identification is verified by examining the x - ray positions on dss images , to account for possible large errors in the positions of optical nuclei .
a nuclear source is labeled as ` n ' in table 2 .
we note that rw2000 used @xmath57 to identify nuclear sources based on the positions without astrometric corrections .
this will miss ulxs near the nuclear region .
for example , the ulx in ngc 5204 is @xmath107 from the nucleus , and was classified as nuclear source in rw2000 .
data analysis of the 467 hri observations in our survey as described in section 4 leads to 562 x - ray sources with detection significance above @xmath22 within @xmath23 of 173 out of the 313 survey galaxies .
these include 371 sources within the @xmath3 isophotes in 155 galaxies , and 191 sources between 12 @xmath8 in 88 galaxies .
these extragalactic sources are listed in table 2 by the host galaxy and the source number within the galaxy in the order of the nuclear offsets .
the positions are corrected with the help of @xmath108 ( col 6 ) x - ray sources identified with objects that have accurate cataloged positions in the same field , and are generally reliable to better than @xmath101 for @xmath109 .
the position of a source with respect to its host galaxy is described by the nuclear offsets in arcseconds and in unit of the elliptical radius @xmath104 . for each source ,
also listed are the maximum detection significance , the maximum luminosity , and number of observations during which the source was variable by a probability @xmath110% .
inspection of the source environments shows that some sources are located in dusty regions such as dust rings / loops / patches / strips / lanes or spiral arms that usually harbor star - forming activities and young massive stars .
sources in such dusty environments are flagged as ` d ' in the table .
particularly , those on spiral arms are flagged as ` s ' , and those on dust rings are flagged as ` r ' .
efforts have been made to identify the sources in other wavelengths by cross correlating with the databases of simbad , ned and vizier . while based primarily on positional coincidence , some identifications are strengthened by the rarity and strong x - ray emission power of the objects .
the x - ray sources are identified to a variety of objects .
some sources are identified with objects not physically associated with the host galaxy , such as background qso / agns , or bright foreground stars ( denoted as ` * ' , prefixed with their spectral type if available ) .
some sources are identified with objects within the host galaxy , such as supernovae ( remnants ) , globular clusters ( gc ) , young massive stars , or nuclei of the host galaxies .
these identifications are given in table 2 enclosed by parenthesis .
many sources are identified with objects for which the current limited knowledge can not dictate whether they are foreground stars , objects in the galaxies , or background qso / agns .
these include faint point sources from optical catalogs such as the usno - b1 catalog ( monet et al .
2003 ) , which are denoted as ` p ' in table 2 , or point sources from radio catalogs such as the first survey catalog , which are denoted as ` pr ' .
the sources are denoted as ` mp ' when there are multiple faint objects close to the x - ray position .
usually magnitudes for these point sources are quoted from the catalogs to illustrate their optical / infrared brightness .
visual inspections of the optical images reveal some sources positionally coincident with faint fuzzy non - pointlike features , and they are denoted as ` z / mz ' .
these p / pr / z sources make promising targets for future identification programs with better positional accuracies and spatial resolutions with instruments such as chandra and hst to better understand their nature . here
we define ulx candidates as those with @xmath111 erg / sec in at least one observation , and with x - ray positions offset from optical galactic nuclei by @xmath112 .
note that for a few x - ray nuclear sources the nuclear offset listed in table 2 are larger than @xmath30 due to the inaccuracy of optical nuclear positions in rc3 .
ulxs are grouped by their nuclear offsets .
ulxs within the @xmath3 isophote are defined as ` 1ulx ' , and those between 12 @xmath113 defined as ` 2ulx ' .
the most extreme ulxs which show @xmath114 erg / sec are defined as ` eulx ' .
if a source is above @xmath0 erg / sec in the observations with the deepest sensitivity for the host galaxies , it is defined as ` ulxd ' . in our survey
there are 216 ulxs found in 95 galaxies , with 106 ` 1ulx ' ( of which 14 are ` eulx ' ) in the @xmath3 isophotes of 63 galaxies , and 110 ` 2ulx ' ( of which 19 are ` eulx ' ) between 12 @xmath113 of 64 galaxies .
the individual observations for the 216 ulx candidates from table 2 are listed in table 3 . for an individual observation of a ulx candidate
, we list the observation date , the exposure time , the duration of the observation , the detection @xmath31 , the correction factor @xmath82 , the corrected count and its percentage error , the flux , the luminosity , and the probability of being constant during the observation from kolmogorov - smirnov tests .
if a source is not detected in an observation , we list the upper limits computed from the count thresholds for @xmath22 detection based on the simulations .
the information in this table can be used for long term variability studies , possibly along with recent observations from instruments on the chandra and xmm - newton observatories .
the ulx sample defined in the above section is based on the proximity of the sources to the host galaxies , and is subject to contamination from background or foreground objects , as demonstrated by the identifications of some ulxs to background qso / agns or foreground stars . to form a basis for future studies on ulxs environments and optical identifications , here
we construct a clean sample of ulxs to minimize contamination .
for this clean sample , we exclude identified qsos and foreground stars . since the ulx candidates between 12 @xmath8 of host galaxies are outside the optical domain of most galaxies and less likely to associate with the galaxies , they are also excluded from the clean sample , except for a few which show apparent connections to the host galaxies by means of dust loops / strips or spiral arms . in total , this clean sample includes 109 ulxs within 61 galaxies , for which the names are given as ` uname ' in table 2 , and the finding charts are shown in figures 1675 . in the following we comments on the host galaxies , the environments , possible identifications and variabilities for these ulxs . the ixo number for a ulx
is given if it is listed in cp2002 .
this sc spiral galaxy ( at a distance of 3.0 mpc ) is a prototype starburst galaxy .
ulx1 is on the outer edge of the galaxy . during seven years
, the source was detected once at [email protected] @xmath116 erg / sec , and it was below 0.3@xmath116 erg / sec in the other five observations .
ngc891 is an edge - on sb galaxy at a distance of 8.36 mpc .
ulx1 is close to but not the nucleus and on the edge of the dust lane .
ulx2 is on the dust lane and is identified as sn1986j .
ulx3 ( ixo 3 ) is on the tip of the dust lane , and showed a steady luminosity increase from @xmath117 to @xmath118 erg / sec in three years .
ulx1 ( ixo 4 in cp2002 ) is on the tip of the spiral arm of this face - on scd spiral galaxy , for which the distance is 8.4 mpc .
ngc1068 at a distance of 14.4 mpc is a sb spiral galaxy with an outer ring .
the luminosities of ulx1 changed from below @xmath119 erg / sec to @xmath120 erg / sec between four observations over five years .
ulx2 is positioned on the outer ring , and showed variability during an 18 day observation .
ngc1073 ( at a distance of 15.2 mpc ) is a barred sc spiral galaxy .
ulx1 ( ixo 5 ) is close to a star forming knot on the spiral arm , and its luminosities increased by more than 50% between two observations separated by half a year .
ulx1 ( ixo 6 ) is located on the outer ring - like spiral arm of this barred s0/a spiral galaxy , for which the distance is 8.6 mpc .
three ulxs are associated with the barred sd galaxy ngc1313 ( at a distance of 3.7 mpc ) , which shows scattered star - forming regions outside its @xmath3 isophote .
ulx1 ( ixo 7 ) is close to but definitely not the nucleus , and it exhibits extremely high luminosities with @xmath114 erg / s . during six years of observations ,
its luminosities vary by more than 50% and showed a dramatic decrease from @xmath36 to @xmath120 erg / sec in less than a month .
pakull & mirioni ( 2002 ; hereafter pm2002 ) found a @xmath121 nebula around this ulx .
ulx2 is identified as sn1978k in a star forming region outside the @xmath3 isophote .
ulx3 ( ixo 8) is in another star forming region outside the @xmath3 isophote . during the observations , it has shown variations of @xmath122% .
it has been identified with a r=21.6 mag stellar - like object ( zampieri et al .
2003 ) within a bubble nebula ( pm2002 ) .
both ulx1 and ulx3 are variable on time scales of a few tens of days .
this peculiar s0 ( lenticular ) radio galaxy at a distance of 21.48 mpc is a member of the fornax cluster , and shows pronounced dusty patches / loops / shells outside its @xmath3 isophote , reminiscent of recent merger events and possibly star forming activities .
ulx1 ( ixo 11 ) and ulx2 ( ixo 10 ) are within the isophote .
ulx2 is close to a faint point source with b2 = 20.24 mag and r2 = 19.90 mag .
assuming v@xmath3820 , the x - ray - to - optical ratio @xmath123 , and indicates it might be a background agn ( stocke et al .
however , due to the large uncertainties in the optical magnitudes ( up to 0.5 mag ) and the x - ray spectral shape , this could be a red old cluster with @xmath124 mag in ngc1316 , or a m dwarf in our galaxy .
five ulxs are on the dusty features outside the @xmath3 isophote .
ulx3 is ixo 12 in cp2002 .
ulx4 is close to a faint point source with b2 = 20.25 mag and r2 = 19.52 mag .
ulx6 ( ixo 13 ) is close to a faint point source with b2 = 21.56 mag and r2 = 19.93 mag .
multiple faint optical objects are found around ulx7 .
ngc1365 is a two armed sb galaxy with a giant bar .
ulx1 is on the edge of the bar .
its luminosity dropped from @xmath125 to below @xmath126 erg / sec in one year .
two ulxs are outside the @xmath3 isophote but lie on the extension of a spiral arm .
ulx2 ( ixo 16 ) is close to a faint point source with b2 = 18.73 and r2 = 19.32 .
this could be a very young massive star cluster ( @xmath127 mag ) in ngc1365 , or a background qso .
the luminosity of ulx3 ( ixo 15 ) dropped from @xmath128 to below @xmath126 erg / sec in one year .
ngc1380 is a lenticular galaxy at a distance of 7.2 mpc .
ulx1 is on the edge of the galaxy .
this peculiar e1 elliptical is the central and brightest galaxy in the fornax cluster , and is known to have globular clusters four times more abundant than a typical elliptical and 15 times more abundant than a typical spiral .
ulx1 ( ixo 18 ) and ulx2 ( ixo 17 ) are identified with globular clusters from chandra and wfpc2 observations ( angelini et al .
ulx1 showed a change in luminosities from @xmath129 to below @xmath130 erg / sec in less than a year .
ulx3 is located around some faint fuzzy features .
this magellanic irregular galaxy is an asymmetric barred galaxy with many small knots at a distance of 16.9 mpc in the fornax i cluster .
ulx1 ( ixo 21 ) is located on the dusty bar .
pgc13826 ( ic342 ) is a face - on scd spiral at a distance of 3.9 mpc with a very bright nucleus and weak surface brightness spiral arms .
three ulxs are all on spiral arms .
ulx1 is close to but not the second nuclear source .
ulx2 was highly variable during the two day hri observation .
ulx3 ( ixo 22 ) is coincident with a tooth - shaped nebula ( pm2002 ) , and is quite variable and showed spectral state transitions reminiscent of black hole candidates during asca observations ( kobuta et al .
2001 ) .
ulx1 and ulx2 are within the @xmath3 isophote in this lenticular galaxy at a distance of 18.54 mpc .
ulx1 and ulx2 are both nearby knots on the dusty spiral arms of this scd spiral galaxy at a distance of 13.6 mpc .
this face - on barred sbc spiral galaxy ( at a distance of 13.4 mpc ) shows clear spiral patterns .
ulx1 ( ixo 24 ) and ulx2 are on the thin spiral arms , and ulx3 is on the edge of a spiral arm . during five years of observations ,
all three ulxs showed variations in luminosity by more than 50% .
ulx3 showed variability during a five day observation .
ulx1 and ulx2 ( ixo 26 and 27 ) are on the spiral arms of this barred sb spiral galaxy at a distance of 14.5 mpc .
ngc1792 is a sbc spiral galaxy at a distance of 10.1 mpc .
ulx1 and ulx2 ( ixo 28 ) are both close to knots on the spiral arms . between two observations in two years
, ulx1 showed a change in luminosity from below @xmath131 to @xmath126 erg / s .
this sc spiral galaxy at a distance of 36.8 mpc shows unusual asymmetric spiral patterns perhaps because of a tidal encounter with its e3 companion ngc2300 .
there are many bright knots on the spiral arms , some of them are known to be recent supernovae .
ulx1 is an extreme ulx on the spiral arm and close to a bright knot .
its luminosity increased from 24@xmath1154 @xmath116 to 34@xmath1158 @xmath116 erg / sec in half a year .
ngc2403 is a sc galaxy with open spiral arms at a distance of 3.133 mpc .
ulx1 is on the amorphous spiral arms .
pgc23324 ( holmberg ii ) is a magellanic irregular galaxy at a distance of 4.5 mpc with numerous hii regions and blue stellar complexes .
ulx1 ( ixo 31 ) is an extreme ulx with @xmath132 erg / s . it is located in an hii complex that shows strong he ii @xmath1334686 that requires ionizing x - ray emission of @xmath134 erg / sec , thus arguing against beaming of the x - ray emission ( pm2002 ) .
this region also exhibits flat spectrum radio emission indicative of a supernova remnant ( tongue & westpfahl 1995 ) .
this nearly face - on ring sab spiral galaxy ( at a distance of 17.0 mpc ) shows multiple arms of fine structure .
ulx1 is on the outer edge of the galaxy and close to a faint object with b2 = 21.58 mag and r2 = 20.15 mag .
this ringed sa spiral galaxy ( at a distance of 37.3 mpc ) is a starburst galaxy with tidal plumes / tails .
ulx1 is an extreme ulx with @xmath135 erg / sec , and is located on the dusty plume near the disk .
this barred sbc spiral is a starburst galaxy at a distance of 7.4 mpc .
ulx1 is on the edge of a spiral arm .
ngc3031 ( m81 ) is an sc spiral galaxy at a distance of 3.42 mpc .
ulx1 and ulx2 are both on a thin spiral arm .
ulx1 is identified with sn1993j .
ulx2 is identified as a black hole binary system with a o8v secondary based on chandra and hst / wfpc2 observations ( liu et al .
2003 ) .
pgc28757 ( holmberg ix ) is a dwarf irregular companion to m81 with a low surface brightness and very blue color , suggesting this is a young galaxy just condensing out of the hi tidal material .
the ulx ( ixo 34 ) is outside the @xmath3 isophote , and surrounded by a barrel shaped @xmath121 emission nebula ( pm2002 ) ; this source is m81-x9 ( fabbiano et al . 1989 ) .
it is an extreme ulx with luminosities varying between 513 @xmath136 erg / sec during six years of observations .
its luminosity once dropped from 13 @xmath136 to 5 @xmath136 erg / sec between two observations in half a year .
la perola et al .
( 2001 ) showed that the x - ray emission is highly variable over 20 years with spectral changes reminiscent of black hole candidates .
this nearly face - on sbc spiral galaxy at a distance of 18.7 mpc shows spiral arms with sporadic star forming knots . ulx1 and ulx2 ( ixo 38 ) are both around knots on spiral arms .
ngc3623 ( m65 ) is an sa spiral galaxy at a distance of 12.3 mpc with a pronounced dust arm in its silhouette against the disk .
ulx1 is positioned on the dust arm .
ngc3627 ( m66 ) is an sb spiral galaxy at a distance of 8.75 mpc with great quantity of dust throughout the disk .
ulx1 is on the outer edge of a spiral arm and close to a very faint feature on the dss image .
ngc3628 is an edge - on sb spiral galaxy at the distance of 10.6 mpc with a starburst nucleus , and distorted dust lanes due to its interaction with other leo triplet galaxies ngc3627 and ngc3623 .
ulx1 is located on the tip of the dust lane where distortions occur due to tidal interactions .
note that the aspect solution of the hri observation for this galaxy was offset by up to @xmath137 . this asymmetric sbc spiral galaxy at a distance of 17.0 mpc shows distorted massive knotty arms with many hii regions .
ulx1 ( ixo 42 ) is located at a bright knot on the spiral arm , and is coincident with a radio point source from the first survey catalog .
ngc4136 is a face - on sc spiral galaxy at a distance of 9.7 mpc .
ulx1 is located on a spiral arm and surrounded by several knots of emission .
this almost face - on ringed sab galaxy ( at a distance of 20.3 mpc ) hosts a prototype seyfert 1.5 agn , and its low surface brightness spiral arms extend slightly beyond the isophote .
ulx1 ( ixo 44 ) is within the @xmath3 isophote and close to a faint point source with b1 = 19.70 mag and r1 = 19.40 mag .
three ulxs are on the spiral arms outside the @xmath3 isophote .
multiple faint features are around ulx4 on the dss image .
ulx4 showed a luminosity drop from @xmath138 to below @xmath139 erg / sec between two observations separated by two years . this peculiar magellanic irregular galaxy at a distance of 2.8 mpc show a moderately high surface brightness indicative of a high recent star - formation rate .
ulx1 is located @xmath97 east of the optical nucleus .
ngc4254 ( m99 ) is a face - on sc grand design spiral galaxy at a distance of 16.8 mpc in the virgo cluster .
ulx1 is located on a thin spiral arm , while ulx2 ( ixo 46 ) is an extreme ulx with @xmath140 erg / sec which is located @xmath141 north of a foreground star .
ngc4258 ( m106 ) is a large barred sbc spiral galaxy at the distance of 7.727 mpc which hosts a weak seyfert 2 nucleus .
ulx1 is on the edge of a spiral arm , and ulx2 is on a thin spiral arm .
the luminosities of both ulx1 and ulx2 changed from below @xmath142 to @xmath143 erg / sec in 500 days .
ulx1 showed variability during an 20 day observation .
ngc4303 ( m61 ) is a face - on sc spiral galaxy with a seyfert 2 nucleus at a distance of 15.2 mpc in the virgo cluster .
ulx1 is located on a thin knotty spiral arm .
ngc4321 ( m100 ) is a grand - design sbc spiral galaxy at a distance of 14.13 mpc .
ulx1 is identified with sn1979c .
ulx2 is on the tenuous tip of a spiral arm .
this is a magellanic spiral galaxy at a distance of 3.6 mpc hosting a seyfert 1 nucleus .
ulx1 ( ixo 53 ) is located on a knotty spiral arm .
the magellanic irregular ngc4485 ( at a distance of 9.3 mpc ) and the sd spiral ngc4490 ( at a distance of 7.8 mpc ) are interacting , and their morphologies are distorted .
there are three ulxs in ngc4490 , all located on irregular dusty regions . within the isophote of ngc4485 is ngc4485-ulx1 ( ixo 62 ) , located on a dust lane .
its luminosity decreased steadily from @xmath144 to @xmath145 erg / sec between three observations separated by 600 days .
ngc4501 ( m88 ) is a sb spiral galaxy at a distance of 16.8 mpc .
ulx1 is an extreme ulx with @xmath140 erg / s .
this ulx is located on a spiral arm , and showed a luminosity drop from @xmath146 to @xmath147 erg / sec between two observations separated by half a year .
this scd spiral galaxy , at a distance of 5.8 mpc , hosts two ulxs , ngc4559-ulx1 and ulx2 .
ulx1 ( ixo 66 ) is @xmath30 east of the optical nucleus , while ulx2 ( ixo 65 ) is coincident with some blue fuzzy objects on the extension of a spiral arm .
ngc4565 is an edge - on sb spiral galaxy at a distance of 17.46 mpc .
ulx1 ( ixo 67 ) is an extreme ulx with @xmath148 erg / s .
it has been identified to a globular cluster in the halo of ngc4565 with chandra and hst / wfpc2 observations ( wu et al . 2002 ) .
ngc4594 is a nearly edge - on sa spiral galaxy with pronounced dust lanes at a distance of 9.77 mpc .
ulx1 and ulx2 are both on the dust lane .
ngc4631 is an edge - on sd galaxy at a distance of 6.9 mpc .
ulx1 ( ixo 68 ) is is located on the dust lanes .
the magellanic spiral ngc4656 at a distance of 7.2 mpc is interacting with its companion ngc4657 to its north - east .
ulx1 is on the tenuous tip of the tidal tail , and is coincident with a very faint object in the dss image .
this e6 elliptical galaxy at a distance of 11.75 mpc has an appreciable globular cluster population .
ulx1 is on the outer edge of the galaxy and coincident with a faint object with b1 = 18.81 mag , r1 = 19.15 mag , b2 = 19.39 mag , and r2 = 19.04 mag .
it showed a luminosity drop from @xmath149 to below @xmath150 erg / sec between two observations over one year .
this magellanic spiral galaxy at a distance of 17.8 mpc has no regular structure but appears to be a strand of hii regions , with a huge hii complex or ob association as the bright spot at the south end .
ulx1 ( ixo 73 ) is on the strand of hii regions , and ulx2 ( ixo 72 ) is coincident with the ob association .
ulx1 is an extreme ulx , and its luminosity increased from @xmath151 to @xmath152 erg / sec between two observations in half a year .
this sbc spiral galaxy at a distance of 8.5 mpc show flocculent spiral arms in the dss image .
ulx1 is close to but not the nucleus . ulx2 and ulx3 are on the spiral arms .
ulx4 ( ixo 74 ) is on the outer edge of the galaxy .
ulx4 showed a luminosity drop from @xmath153 to @xmath154 erg / sec between two observations separated by three years .
ngc5128 ( centarus a ) is a peculiar lenticular galaxy at a distance of 4.21 mpc with a warped lane of gas and dust .
it is also a bright radio galaxy with an x - ray jet .
ulx1 ( ixo 76 ) is located on the edge of the bulge .
this ulx showed luminosities @xmath129 erg / sec in five observations in a ten day time window , but it was below @xmath155 erg / sec in the other three observations over eight years .
ngc5194 ( m51 ) is a grand - design sbc galaxy at a distance of 7.7 mpc which is interacting with its companion ngc5195 .
six ulxs are associated with the galaxy , with ulx6 on the outer edge of a spiral arm , and all the other five right on the thin knotty spiral arms .
ulx1 ( ixo 79 ) , ulx2 and ulx3 ( ixo 80 ) are near bright knots on the dss image that are probably young star clusters .
all ulxs have exhibited variations by more than 50% during six years of observations .
ulx3 showed a luminosity drop from @xmath144 to below @xmath131 erg / sec in the observations .
this source appears to have two - hour periodic variations in a recent chandra observation ( liu et al .
ulx4 is ixo 78 in cp2002 .
ulx5 ( ixo 81 ) showed variability during a six day observation .
ngc5204 is a magellanic irregular galaxy at a distance of 4.3 mpc .
ulx1 ( ixo 77 ) is positioned @xmath107 east of the optical nucleus .
it showed high variability within a one day observation .
this ulx has been identified with a b0ib supergiant presumably as a secondary in a black hole x - ray binary from chandra and hst observations ( liu et al .
2004 ) .
ngc5236 ( m83 ) is a starburst barred sc spiral galaxy at a distance of 4.7 mpc with very luminous spiral arms .
ulx1 ( ixo 82 ) is located on the edge of a spiral arm , and its luminosity showed an increase from @xmath145 to @xmath156 erg / sec between two observations separated by 600 days .
ngc5457 ( m101 ) is a face - on prototype sc spiral galaxy at a distance of 6.855 mpc .
ulx1 is coincident with an object with b = 17.80 mag and r = 15.00 mag , and showed variability within a two day observation .
ulx2 and ulx3 ( ixo 83 ) are both on spiral arms , both showed variability within a 26 day observation , and the luminosities of both varied by more than 50% during the four observations in five years .
ngc5774 is a face - on sd spiral with bright blue knots in the arms at a distance of 26.8 mpc .
ulx1 ( ixo 84 ) is close to knotty features on a chain of knots across the disk .
ngc6946 is a scd spiral at a distance of 5.5 mpc with recent star formation throughout the spiral arm structure and a mild starburst at the center .
ulx1 ( ixo 85 ) and ulx2 are on knotty spiral arms .
ulx1 showed a steady luminosity decrease from @xmath157 to below @xmath131 erg / sec in 800 days .
ulx3 is on the edge of a spiral arm and coincident with a cocoon shaped supernova remnant ( dunne et al . 2001 ) .
its extreme luminosity @xmath158 erg / sec either comes from the colliding snr or from accretion onto a newborn black hole ( roberts & colbert , 2003 ) .
this sbc spiral , at a distance of 12.6 mpc , has a seyfert 1.9 nucleus and hii regions all over its spiral arms .
ulx1 ( ixo 86 ) is on the outer edge of the galaxy .
this sbc spiral at a distance of 17.3 mpc has a seyfert 2 nucleus and very bright spiral arms in the inner part .
ulx1 ( ixo 87 ) is close to knotty features on the spiral arm .
its luminosity showed a drop from @xmath159 to @xmath149 erg / sec in a year .
this sc spiral at a distance of 36.9 mpc is interacting with its dwarf irregular companion ngc7715 .
ulx1 is an extreme ulx with @xmath160 erg / sec , and is located on the tenuous plume on the outer edge of a spiral arm in ngc7714 .
this sb spiral at a distance of 22.2 mpc has an extremely bright nucleus and many poorly resolved knotty arms .
ulx1 is on the outer edge of the galaxy .
in this paper , we report our archival rosat hri survey of extragalactic x - ray point sources and ultraluminous x - ray sources in a sample of 313 nearby galaxies with @xmath161 within 40 mpc .
the survey was carried out with well defined data reduction procedures . to detect point sources from hri observations we utilize the wavdetect package , an implementation of a wavelet algorithm , for which simulations were run to better understand its characteristics such as the count thresholds for @xmath22 detection and correction factors . for sources detected above @xmath22 from the hri observations in our survey ,
we compute their count rate and fluxes , and test whether they are variable during the observations with kolmogorov - smirnov tests .
the x - ray positions were corrected by registering x - ray sources to the accurate positions of their counterparts in other wavelengths , and the positional accuracy is better than @xmath101 for most x - ray sources .
x - ray sources within @xmath23 of a galaxy are considered associated with the galaxy and their luminosities are computed by placing them at the distance for the galaxy .
the uniform data reduction procedures lead to 562 extragalactic x - ray sources in 173 nearby galaxies spanning a luminosity range of @xmath162
@xmath7 erg / sec , with 371 sources within the @xmath3 isophotes in 155 galaxies , and 191 sources between 12 @xmath8 in 88 galaxies . in our survey
we define those extra - nuclear sources with @xmath163 erg / sec as ulx candidates .
this leads to a sample of 216 ulx candidates , which includes a sample of 106 ulxs within the @xmath3 isophotes in 63 galaxies ( 1ulx ) , and a sample of 116 ulxs between 12 @xmath8 in 64 galaxies ( 2ulx ) .
thus defined ulx candidates are possibly not in the host galaxies , and ten ( 16 ) in the 1ulx ( 2ulx ) sample are already identified as background qsos or foreground stars . to minimize such contaminations
, we constructed a clean sample of 109 ulxs in 61 galaxies by excluding those identified stars and qsos , and excluding most of the 2ulx sample since they are more likely contaminating sources due to their large separations from the galaxies .
this clean sample forms a good basis for studies on ulx s environments and identifications .
a few conclusions can be drawn from examinations of the clean sample .
ulxs are preferentially found in late - type galaxies .
for example , 49 out of 181 spiral galaxies in our survey host 89 ulxs , while 4 out of 93 early - type galaxies host 7 ulxs if we exclude the 8 ulxs in two peculiar lenticular galaxies ngc1316 and ngc5128 . there is also a strong tendency for ulxs to occur in dusty regions of star formation , with 84 of the 109 ulxs in the clean sample in such regions .
specifically , 51 ulxs occur on thin dusty spiral arms .
also note that the seven ulxs in ngc1316 are associated with dust loops / shells / patches reminiscent of star forming activities .
the above statistics reveals a strong connection between the ulx phenomenon and star formation , as reported by previous authors .
some ulxs have demonstrated great variability over a range of time scales from days to up to 10 years .
fifteen out of 109 ulxs in the clean sample have shown variability during at least one observation lasting from days to months .
these ulxs are all located in star forming regions in late - type galaxies . for ulxs with more than one hri observations ,
a variety of temporal behaviors have been exhibited .
these include sudden bursts with small duty cycles ( e.g. , ngc5128-ulx1 ) , steady increases ( e.g. , ngc891-ulx3 ) , steady decreases ( e.g. , ngc6946-ulx1 ) , or occasional sharp drops from almost constant high luminosity ( e.g. , ngc1313-ulx1 ) .
ulxs show a very diverse nature seen from the identifications for the clean sample .
five ulxs are identified as recent supernovae ( remnants ) .
many ulxs on knotty spiral arms in the dss images are clearly identified with hii regions / nebulae in recent observations with better resolutions ( pm2002 ) .
there are two confirmed cases in which the ulxs are identified with very young massive stars : ngc3031-ulx2 with an o8v star ( liu et al .
2003 ) ; and ngc5204-ulx1 with a b0ib star ( liu et al .
these identifications are consistent with the strong link between ulxs and star formation .
however , there are also ulxs identified with old globular clusters lacking recent star formation .
for example , ngc4565-ulx1 is identified with a halo globular cluster , and two ulxs are identified with globular clusters in the elliptical galaxy ngc1399 .
these ulxs in the globular clusters may have been formed differently from those in star forming regions .
while many qualitative properties have been revealed from studying the clean sample of ulxs , the quantative properties such as the occurrence rates of ulxs derived from the clean sample are not strictly accurate because the clean sample is not a complete sample in the statistical sense , and the contamination , though small , is not known . in a subsequent paper , statistical properties , such as the occurrence frequencies and luminosity functions , of ulxs in different types of galaxies will be studied quantatively with the contaimnation calrefully calculated and subtracted , in comparison to statistical studies of ulxs in 766 nearby galaxies with @xmath19@xmath20@xmath21 km / sec in hri observations by ptak & colbert ( 2004 ) and of ulxs in 82 nearby galaxies in chandra observations by swartz et al .
( 2004 ) .
we are grateful to the ned , vizier services .
we would like to thank dan harris , samantha stevenson , renato dupke , james irwin , ed lloyd - daves and eric miller for helpful discussions .
we thank the referee for his constructive suggestions .
we gratefully acknowledge support for this work from nasa under grants hst - go-09073 .
ngc247 & 113.95 & -83.56 & 21.4 & 6.9 & 1.55e20 & 2.8[t92 ] & sab(s)d & 7 & 1.6e09 & 7.8e-01 & -16.7 & + ngc253 & 97.36 & -87.96 & 27.5 & 6.8 & 1.42e20 & 3.0[t92 ] & sab(s)c & 5 & 6.8e09 & 7.6e+02 & -21.8 & sbrst + ngc300 & 299.20 & -79.42 & 21.9 & 15.5 & 3.23e20 & 2.023[kp ] & sa(s)d & 7 & 2e09 & @xmath206.8e+00 & -17.1 & + ngc404 & 127.03 & -27.01 & 3.5 & 3.5 & 5.31e20 & 3.27[sbf ] & sa(s)0- & -3 & 5.9e08 & 2.3e+00 & -15.4 & + ngc520 & 138.70 & -58.06 & 4.5 & 1.8 & 3.27e20 & 27.8[t88 ] & peculiar & 99 & 1.9e10 & 3.2e+01 & -22.8 & + ngc613 & 229.07 & -80.30 & 5.5 & 4.2 & 1.63e20 & 26.7[t92 ] & sb(rs)bc & 4 & 5.6e10 & 2.0e+01 & -22.4 & + ngc720 & 173.02 & -70.36 & 4.7 & 2.4 & 1.54e20 & 27.67[sbf ] & e5 & -5 & 4.1e10 & @xmath202.2e-01 & @xmath20 - 17.3 & + ngc891 & 140.38 & -17.42 & 13.5 & 2.5 & 7.64e20 & 8.36[sbf ] & sa(s)b & 3 & 6.3e09 & 3.4e+01 & -21.3 & hii + ngc1042 & 182.08 & -58.17 & 4.7 & 3.6 & 3.12e20 & 8.4[t92 ] & sab(rs)cd & 6 & 2.1e09 & 1.2e+00 & -17.3 & + ngc1047 & 181.70 & -57.96 & 1.3 & 0.6 & 3.07e20 & 17.9[ned ] & s0 + & -0.5 & & @xmath201.5e-02 & @xmath20 - 13.4 & + ngc1052 & 182.02 & -57.93 & 3.0 & 2.1 & 3.07e20 & 19.41[sbf ] & e4 & -5 & 1.5e10 & 9.4e-01 & -18.1 & + ngc1068 & 172.10 & -51.94 & 7.1 & 6.0 & 3.53e20 & 14.4[t88 ] & sa(rs)b & 3 & 4.3e10 & 1.9e+02 & -23.2 & + ngc1073 & 170.91 & -50.73 & 4.9 & 4.5 & 4.23e20 & 15.2[t88 ] & sb(rs)c & 5 & 8.3e09 & 1.1e+00 & -18.5 & + ngc1097 & 226.91 & -64.68 & 9.3 & 6.3 & 1.85e20 & 16.8[t92 ] & sb(s)b & 3 & 4.5e10 & 4.6e+01 & -22.3 & + ngc1291 & 247.52 & -57.05 & 9.8 & 8.1 & 2.12e20 & 8.6[t88 ] & sb(s)0/a & 0 & 1.9e10 & 1.7e+00 & -17.8 & + ngc1313 & 283.36 & -44.64 & 9.1 & 6.9 & 3.89e20 & 3.7[t88 ] & sb(s)d & 7 & 2.8e09 & 1.0e+01 & -18.8 & hii + ngc1316 & 240.16 & -56.69 & 12.0 & 8.5 & 1.89e20 & 21.48[sbf ] & sab(s)0o & -2 & 1.2e11 & 3.1e+00 & -19.9 & + ngc1317 & 239.97 & -56.69 & 2.8 & 2.4 & 1.91e20 & 16.9[t88 ] & sab(r)a & 1 & 8e09 & 3.6e+00 & -19.6 & + ngc1332 & 212.18 & -54.37 & 4.7 & 1.4 & 2.23e20 & 22.91[sbf ] & s(s)0- & -3 & 2.4e10 & 5.2e-01 & -18.3 & + ngc1365 & 237.95 & -54.60 & 11.2 & 6.2 & 1.39e20 & 17.38[kp ] & sb(s)b & 3 & 3.8e10 & 7.8e+01 & -23.0 & + ngc1369 & 238.10 & -53.96 & 1.5 & 1.4 & 1.40e20 & 18.9[ned ] & sb(rs)0/a & 0 & 1.8e09 & @xmath204.3e-02 & @xmath20 - 14.7 & + ngc1380 & 235.93 & -54.06 & 4.8 & 2.3 & 1.31e20 & 17.62[sbf ] & sa0 & -2 & 1.7e10 & 1.1e+00 & -18.4 & + ngc1380a & 235.52 & -53.99 & 2.4 & 0.7 & 1.29e20 & 15.85[sbf ] & s0o & -2 & 1.8e09 & @xmath203.4e-02 & @xmath20 - 14.0 & + ngc1382 & 236.29 & -53.92 & 1.5 & 1.3 & 1.31e20 & 22.59[sbf ] & sab(s)0- & -2.7 & 2.5e09 & @xmath204.0e-02 & @xmath20 - 15.0 & + ngc1386 & 237.66 & -53.97 & 3.4 & 1.3 & 1.37e20 & 16.52[sbf ] & sb(s)0 + & -0.6 & 5.3e09 & 5.9e+00 & -19.8 & + ngc1395 & 216.21 & -52.12 & 5.9 & 4.5 & 1.99e20 & 24.10[sbf ] & e2 & -5 & 3.9e10 & @xmath205.2e-01 & @xmath20 - 17.9 & + ngc1398 & 221.53 & -52.79 & 7.1 & 5.4 & 9.46e19 & 22.2[t92 ] & sb(r)ab & 2 & 4.8e10 & 9.3e-01 & -19.5 & + ngc1399 & 236.71 & -53.64 & 6.9 & 6.5 & 1.34e20 & 19.95[sbf ] & e1 & -5 & 4.9e10 & @xmath208.9e-01 & @xmath20 - 18.1 & + ngc1404 & 236.95 & -53.56 & 3.3 & 3.0 & 1.36e20 & 20.99[sbf ] & e1 & -5 & 2.6e10 & @xmath202.0e-01 & @xmath20 - 16.6 & + ngc1427a & 236.99 & -53.29 & 2.3 & 1.5 & 1.36e20 & 16.9[t88 ] & ib(s)m & 10 & 1.9e09 & 2.1e-01 & @xmath20 - 15.0 & + ngc1433 & 255.69 & -51.20 & 6.5 & 5.9 & 1.44e20 & 11.6[t88 ] & sb(r)ab & 2 & 1e10 & 3.3e+00 & -19.0 & + ngc1481 & 214.00 & -47.81 & 1.0 & 0.7 & 3.73e20 & 23.1[ned ] & sa0- & -3.3 & 1.4e09 & 3.6e-01 & @xmath20 - 13.9 & + ngc1482 & 214.12 & -47.80 & 2.5 & 1.4 & 3.69e20 & 19.6[t88 ] & sa0 + & -0.8 & 3.3e09 & 3.2e+01 & -22.0 & hii + ngc1531 & 233.13 & -46.60 & 1.3 & 0.9 & 1.68e20 & 13.0[t88 ] & s0- & -3.2 & 1.4e09 & @xmath202.5e-02 & @xmath20 - 13.3 & + ngc1532 & 233.17 & -46.58 & 12.6 & 3.3 & 1.68e20 & 16.8[t92 ] & sb(s)b & 2.7 & 1.5e10 & @xmath208.3e-01 & -19.7 & + ngc1549 & 265.41 & -43.80 & 4.9 & 4.1 & 1.46e20 & 19.68[sbf ] & e0 & -5 & 3.3e10 & 1.3e-01 & @xmath20 - 17.2 & + ngc1553 & 265.63 & -43.69 & 4.5 & 2.8 & 1.50e20 & 18.54[sbf ] & sa(r)0o & -2 & 4.2e10 & 5.2e-01 & -17.4 & + ngc1559 & 274.51 & -41.20 & 3.5 & 2.0 & 2.65e20 & 13.6[t92 ] & sb(s)cd & 6 & 9.2e09 & 2.4e+01 & @xmath20 - 15.2 & + ngc1566 & 264.31 & -43.39 & 8.3 & 6.6 & 1.34e20 & 13.4[t88 ] & sab(s)bc & 4 & 2.5e10 & 1.3e+01 & -20.6 & + ngc1574 & 266.89 & -42.58 & 3.4 & 3.1 & 1.64e20 & 19.86[sbf ] & sa(s)0- & -2.7 & 1.4e10 & 3.1e-01 & @xmath20 - 16.5 & + ngc1640 & 218.87 & -37.22 & 2.6 & 2.0 & 2.99e20 & 19.1[t88 ] & sb(r)b & 3 & 5.8e09 & 1.1e+00 & -18.7 & + ngc1672 & 268.79 & -38.99 & 6.6 & 5.5 & 2.28e20 & 14.5[t88 ] & sb(s)b & 3 & 1.6e10 & 3.5e+01 & -21.6 & + ngc1705 & 261.08 & -38.74 & 1.9 & 1.4 & 4.19e20 & 6.0[t88 ] & sa0- & -3 & 4.2e08 & 1.1e+00 & -15.7 & hii + ngc1792 & 241.69 & -36.45 & 5.2 & 2.6 & 2.58e20 & 10.1[t92 ] & sa(rs)bc & 4 & 7.4e09 & 2.5e+01 & -20.6 & + ngc2146 & 135.66 & 24.90 & 6.0 & 3.4 & 7.30e20 & 17.2[t88 ] & sb(s)ab & 2 & 1.8e10 & 1.4e+02 & -23.3 & + ngc2273 & 154.97 & 23.31 & 3.2 & 2.5 & 6.97e20 & 28.4[t88 ] & sb(r)a & 0.5 & 1.4e10 & 6.3e+00 & -21.1 & + ngc2276 & 127.67 & 27.71 & 2.8 & 2.7 & 5.32e20 & 36.8[t88 ] & sab(rs)c & 5 & 2.6e10 & 1.2e+01 & -22.5 & + ngc2300 & 127.71 & 27.81 & 2.8 & 2.0 & 5.27e20 & 31.0[t88 ] & sa0o & -2 & 2.4e10 & @xmath201.2e-01 & @xmath20 - 16.8 & + ngc2366 & 146.42 & 28.54 & 8.1 & 3.3 & 4.08e20 & 2.7[t92 ] & ib(s)m & 10 & 3.2e08 & @xmath205.4e-01 & -15.2 & + ngc2403 & 150.57 & 29.19 & 21.9 & 12.3 & 4.13e20 & 3.133[kp ] & sab(s)cd & 6 & 3.4e09 & 1.8e+01 & -18.8 & hii + ngc2442 & 281.30 & -21.50 & 5.5 & 4.9 & 1.18e21 & 17.1[t88 ] & sab(s)bc & 3.7 & 1.7e10 & 5.3e+00 & @xmath20 - 17.2 & + ngc2715 & 134.73 & 33.32 & 4.9 & 1.7 & 1.94e20 & 20.4[t88 ] & sab(rs)c & 5 & 1.2e10 & 1.7e+00 & -19.7 & + ngc2775 & 223.27 & 33.99 & 4.3 & 3.3 & 4.07e20 & 17.0[t88 ] & sa(r)ab & 2 & 1.6e10 & 1.7e+00 & @xmath20 - 16.5 & + ngc2782 & 182.15 & 43.68 & 3.5 & 2.6 & 1.76e20 & 37.3[t88 ] & sab(rs)a & 1 & 3.4e10 & 8.5e+00 & -22.0 & sbrst + ngc2853 & 182.05 & 45.44 & 1.7 & 0.9 & 1.60e20 & 28.2[t88 ] & sb0 & -2 & 2.4e09 & 2.6e-01 & @xmath20 - 15.2 & + ngc2903 & 208.71 & 44.54 & 12.6 & 6.0 & 3.14e20 & 7.4[t92 ] & sab(rs)bc & 4 & 1.2e10 & 2.8e+01 & -20.7 & hii + ngc2974 & 239.51 & 35.01 & 3.5 & 2.0 & 3.72e20 & 21.48[sbf ] & e4 & -5 & 1.2e10 & 4.8e-01 & -18.0 & + ngc2992 & 249.71 & 28.78 & 3.5 & 1.1 & 5.26e20 & 30.5[t88 ] & sa & 1 & 1.3e10 & 6.9e+00 & -21.4 & + ngc2993 & 249.76 & 28.77 & 1.3 & 0.9 & 5.27e20 & 30.5[t88 ] & sa & 1 & 7.9e09 & 1.0e+01 & -21.7 & hii + ngc3031 & 142.09 & 40.90 & 26.9 & 14.1 & 4.16e20 & 3.42[kp ] & sa(s)ab & 2 & 1.2e10 & 7.0e+00 & -19.1 & + ngc3034 & 141.40 & 40.57 & 11.2 & 4.3 & 3.98e20 & 5.2[t88 ] & i0 & 90 & 8.3e09 & 1.2e+03 & -23.0 & sbrst + ngc3049 & 227.57 & 44.72 & 2.2 & 1.4 & 3.03e20 & 19.9[ned ] & sb(rs)ab & 2 & 3.7e09 & 2.7e+00 & -19.4 & sbrst + ngc3073 & 157.99 & 48.25 & 1.3 & 1.2 & 8.22e19 & 33.73[sbf ] & sab0- & -2.5 & 6e09 & @xmath203.1e-02 & @xmath20 - 15.6 & + ngc3079 & 157.81 & 48.36 & 7.9 & 1.4 & 7.89e19 & 17.3[t92 ] & sb(s)c & 7 & 1.3e10 & 4.3e+01 & -22.2 & + ngc3115 & 247.78 & 36.78 & 7.2 & 2.5 & 4.32e20 & 9.68[sbf ] & s0- & -3 & 2.2e10 & 2.1e-01 & @xmath20 - 15.5 & + ngc3185 & 213.22 & 54.70 & 2.3 & 1.6 & 2.12e20 & 21.3[t88 ] & sb(r)a & 1 & 5.6e09 & 1.6e+00 & -19.1 & + ngc3187 & 212.93 & 54.79 & 3.0 & 1.3 & 2.12e20 & 26.1[t88 ] & sb(s)c & 5 & 5.5e09 & 9.1e-01 & -19.1 & hii + ngc3190 & 213.03 & 54.85 & 4.4 & 1.5 & 2.12e20 & 22.4[t88 ] & sa(s)a & 1 & 1.4e10 & 3.5e+00 & -20.2 & + ngc3193 & 212.97 & 54.93 & 3.0 & 2.7 & 2.12e20 & 34.04[sbf ] & e2 & -5 & 2.6e10 & @xmath201.6e-01 & @xmath20 - 17.4 & + ngc3226 & 216.93 & 55.44 & 3.2 & 2.8 & 2.14e20 & 23.55[sbf ] & e2 & -5 & 7.9e09 & @xmath201.8e-01 & @xmath20 - 16.7 & + ngc3227 & 217.00 & 55.45 & 5.4 & 3.6 & 2.15e20 & 20.6[t88 ] & sab(s)a & 1 & 1.5e10 & 8.0e+00 & -20.8 & + ngc3252 & 134.80 & 40.05 & 2.0 & 0.6 & 4.09e20 & 20.8[t88 ] & sbd & 6.7 & 1.5e09 & 9.4e-01 & -18.5 & + ngc3294 & 184.62 & 59.84 & 3.5 & 1.8 & 1.65e20 & 26.7[t88 ] & sa(s)c & 5 & 2.4e10 & 6.1e+00 & -21.2 & + ngc3310 & 156.61 & 54.06 & 3.1 & 2.4 & 1.13e20 & 18.7[t88 ] & sab(r)bc & 4 & 2.1e10 & 3.3e+01 & -21.9 & hii + ngc3370 & 225.35 & 59.67 & 3.2 & 1.8 & 2.64e20 & 23.4[t88 ] & sa(s)c & 5 & 1e10 & 3.6e+00 & -20.3 & + ngc3377 & 231.18 & 58.31 & 5.2 & 3.0 & 2.86e20 & 11.22[sbf ] & e5 & -5 & 7.9e09 & @xmath203.2e-01 & @xmath20 - 15.7 & + ngc3377a & 230.96 & 58.29 & 2.2 & 2.1 & 2.87e20 & 5.1[t88 ] & sab(s)m & 9 & 1.5e08 & @xmath209.6e-02 & @xmath20 - 12.7 & + ngc3379 & 233.49 & 57.63 & 5.4 & 4.8 & 2.75e20 & 10.57[sbf ] & e1 & -5 & 1.3e10 & @xmath205.2e-01 & @xmath20 - 16.1 & + ngc3384 & 233.52 & 57.75 & 5.5 & 2.5 & 2.73e20 & 11.59[sbf ] & sb(s)0- & -3 & 1.2e10 & @xmath202.8e-01 & @xmath20 - 15.6 & + ngc3389 & 233.72 & 57.74 & 2.8 & 1.4 & 2.72e20 & 22.5[t88 ] & sa(s)c & 5 & 9.2e09 & @xmath207.5e-02 & @xmath20 - 15.7 & + ngc3395 & 192.93 & 63.14 & 2.1 & 1.2 & 1.98e20 & 27.4[t88 ] & sab(rs)cd & 6 & 1.2e10 & 8.2e+00 & @xmath20 - 15.7 & + ngc3396 & 192.90 & 63.16 & 3.1 & 1.2 & 1.98e20 & 21.8[t92 ] & ibm & 10 & 6.5e09 & 8.2e+00 & @xmath20 - 15.6 & + ngc3398 & 152.01 & 54.36 & 1.3 & 0.6 & 7.84e19 & 38.2[ned ] & sa & 1 & 1.7e09 & 8.2e-01 & @xmath20 - 15.1 & + ngc3440 & 149.55 & 53.50 & 2.1 & 0.5 & 5.78e19 & 25.4[ned ] & sbb & 3 & 2.5e09 & 5.5e-01 & -18.4 & + ngc3486 & 202.07 & 65.49 & 7.1 & 5.2 & 1.91e20 & 12.3[t92 ] & sab(r)c & 5 & 1.1e10 & 4.6e+00 & @xmath20 - 16.8 & + ngc3513 & 273.08 & 33.31 & 2.8 & 2.2 & 5.58e20 & 14.7[t92 ] & sb(rs)c & 5 & 4.9e09 & 2.6e+00 & -19.0 & + ngc3516 & 133.24 & 42.40 & 1.7 & 1.4 & 3.05e20 & 38.9[t88 ] & sb(s)0o & -2 & 2.9e10 & 1.7e+00 & -20.3 & + ngc3622 & 135.77 & 47.55 & 1.2 & 0.5 & 1.28e20 & 17.5[ned ] & s & & 1.2e09 & 1.1e+00 & -18.3 & + ngc3623 & 241.33 & 64.22 & 9.8 & 2.9 & 2.16e20 & 12.3[t92 ] & sab(rs)a & 1 & 2.4e10 & 2.1e+00 & -19.1 & + ngc3627 & 241.95 & 64.42 & 9.1 & 4.2 & 2.43e20 & 8.75[kp ] & sab(s)b & 3 & 1.8e10 & 3.4e+01 & -21.1 & + ngc3628 & 240.85 & 64.78 & 14.8 & 3.0 & 2.22e20 & 10.6[t92 ] & sb & 3 & 1.2e10 & 4.8e+01 & -21.4 & + ngc3642 & 142.55 & 54.53 & 5.4 & 4.5 & 8.05e19 & 27.5[t88 ] & sa(r)bc & 4 & 2.9e10 & 1.5e+00 & -19.8 & + ngc3658 & 175.38 & 68.45 & 1.6 & 1.5 & 2.06e20 & 27.3[ned ] & sa(r)0o & -2.3 & 6.1e09 & @xmath204.8e-02 & @xmath20 - 15.6 & + ngc3665 & 174.71 & 68.49 & 2.5 & 2.0 & 2.06e20 & 32.4[t88 ] & sa(s)0o & -2 & 3.4e10 & 2.0e+00 & -20.5 & + ngc3738 & 144.56 & 59.32 & 2.5 & 1.9 & 1.08e20 & 4.3[t88 ] & i m & 10 & 5.5e08 & 2.0e+00 & -15.8 & hii + ngc3972 & 138.85 & 60.06 & 3.9 & 1.1 & 1.20e20 & 17.0[t88 ] & sa(s)bc & 4 & 3.5e09 & 1.0e+00 & -18.4 & + ngc3982 & 138.83 & 60.27 & 2.3 & 2.0 & 1.22e20 & 17.0[t88 ] & sab(r)b & 3 & 8.6e09 & 6.9e+00 & -20.2 & + ngc3990 & 138.25 & 60.04 & 1.4 & 0.8 & 1.21e20 & 10.28[sbf ] & s0- & -3 & 6e08 & @xmath202.3e-02 & @xmath20 - 12.7 & + ngc3998 & 138.17 & 60.06 & 2.7 & 2.2 & 1.22e20 & 14.13[sbf ] & sa(r)0o & -2 & 8.6e09 & 4.4e-01 & @xmath20 - 15.2 & + ngc4036 & 132.98 & 54.25 & 4.3 & 1.7 & 1.89e20 & 24.6[t88 ] & s0- & -3 & 2.4e10 & 5.4e-01 & @xmath20 - 16.6 & + ngc4041 & 132.70 & 54.05 & 2.7 & 2.5 & 1.97e20 & 22.7[t88 ] & sa(rs)bc & 4 & 1.6e10 & 1.4e+01 & -21.6 & + ngc4051 & 148.88 & 70.08 & 5.2 & 3.9 & 1.32e20 & 17.0[t88 ] & sab(rs)bc & 4 & 1.3e10 & 8.3e+00 & -20.5 & + ngc4085 & 140.59 & 65.17 & 2.8 & 0.8 & 2.00e20 & 17.0[t88 ] & sab(s)c & 5 & 3.1e09 & 5.5e+00 & -20.1 & + ngc4088 & 140.33 & 65.01 & 5.8 & 2.2 & 2.00e20 & 17.0[t88 ] & sab(rs)bc & 4 & 1.6e10 & 1.7e+01 & @xmath20 - 16.4 & + ngc4105 & 291.46 & 32.11 & 2.7 & 2.0 & 4.99e20 & 26.55[sbf ] & e3 & -5 & 2.4e10 & 2.3e-01 & @xmath20 - 16.4 & + ngc4106 & 291.48 & 32.11 & 1.6 & 1.3 & 4.98e20 & 30.6[t88 ] & sb(s)0 + & -1 & 2.1e10 & 2.3e-01 & @xmath20 - 15.7 & + ngc4136 & 193.63 & 80.34 & 4.0 & 3.7 & 1.62e20 & 9.7[t88 ] & sab(r)c & 5 & 3.1e09 & 1.3e+00 & -17.4 & hii + ngc4138 & 147.29 & 71.40 & 2.6 & 1.7 & 1.36e20 & 13.80[sbf ] & sa(r)0 + & -1 & 4.3e09 & @xmath208.8e-02 & @xmath20 - 14.8 & + ngc4144 & 143.17 & 69.01 & 6.0 & 1.3 & 1.28e20 & 5.7[t92 ] & sab(s)cd & 6 & 7.2e08 & 1.4e+00 & -16.2 & hii + ngc4150 & 190.45 & 80.47 & 2.3 & 1.6 & 1.62e20 & 13.74[sbf ] & sa(r)0o & -2 & 3e09 & 1.2e+00 & -17.8 & + ngc4151 & 155.07 & 75.06 & 6.3 & 4.5 & 1.98e20 & 20.3[t88 ] & sab(rs)ab & 2 & 2.8e10 & @xmath205.6e-01 & @xmath20 - 17.6 & + ngc4156 & 154.75 & 75.04 & 1.4 & 1.1 & 1.99e20 & 15.5[t88 ] & sb(rs)b & 3 & 8.6e08 & @xmath203.2e-02 & @xmath20 - 13.9 & + ngc4190 & 160.62 & 77.59 & 1.7 & 1.5 & 1.54e20 & 2.8[t88 ] & i m & 10 & 6.1e07 & 3.7e-01 & @xmath20 - 10.7 & + ngc4203 & 173.03 & 80.08 & 3.4 & 3.2 & 1.19e20 & 15.14[sbf ] & sab0- & -3 & 8.5e09 & 6.4e-01 & -17.6 & + ngc4214 & 160.25 & 78.07 & 8.5 & 6.6 & 1.49e20 & 3.5[t88 ] & iab(s)m & 10 & 1.6e09 & 1.4e+01 & -17.5 & hii + ngc4235 & 279.17 & 68.47 & 4.2 & 0.9 & 1.53e20 & 35.1[t88 ] & sa(s)a & 1 & 1.8e10 & 3.2e-01 & -18.4 & + ngc4236 & 127.41 & 47.36 & 21.9 & 7.2 & 1.76e20 & 2.2[t88 ] & sb(s)dm & 8 & 5.4e08 & @xmath203.2e+00 & -15.2 & + ngc4242 & 140.77 & 70.31 & 5.0 & 3.8 & 1.18e20 & 7.5[t88 ] & sab(s)dm & 8 & 2.1e09 & @xmath203.8e-01 & @xmath20 - 15.1 & + ngc4244 & 154.56 & 77.16 & 16.6 & 1.9 & 1.67e20 & 4.2[t92 ] & sa(s)cd & 6 & 1.7e09 & 4.5e-01 & -16.9 & hii + ngc4248 & 138.71 & 68.68 & 3.0 & 1.1 & 1.20e20 & 7.3[t88 ] & i0 & 90 & 3e08 & @xmath206.6e-02 & @xmath20 - 13.1 & + ngc4254 & 270.42 & 75.19 & 5.4 & 4.7 & 2.72e20 & 16.8[t88 ] & sa(s)c & 5 & 3.7e10 & 2.3e+01 & -22.2 & + ngc4257 & 281.72 & 67.26 & 1.3 & 0.4 & 1.55e20 & 36.7[ned ] & sab & 2 & 2.3e09 & @xmath209.1e-03 & @xmath20 - 14.4 & + ngc4258 & 138.32 & 68.84 & 18.6 & 7.2 & 1.16e20 & 7.727[kp ] & sab(s)bc & 4 & 1.8e10 & @xmath202.7e+00 & -20.0 & + ngc4261 & 281.80 & 67.37 & 4.1 & 3.6 & 1.55e20 & 31.62[sbf ] & e2 & -5 & 4.4e10 & @xmath203.0e-01 & @xmath20 - 17.9 & + ngc4269 & 281.91 & 67.59 & 1.1 & 0.8 & 1.54e20 & 27.7[ned ] & s0 + & -1 & 3.1e09 & @xmath201.8e-02 & @xmath20 - 14.6 & + ngc4278 & 193.79 & 82.77 & 4.1 & 3.8 & 1.77e20 & 16.07[sbf ] & e1 & -5 & 1.5e10 & 5.7e-01 & -17.5 & + ngc4283 & 193.44 & 82.81 & 1.5 & 1.5 & 1.77e20 & 15.70[sbf ] & e0 & -5 & 2.5e09 & @xmath204.5e-02 & @xmath20 - 14.3 & + ngc4286 & 193.02 & 82.88 & 1.6 & 1.0 & 1.79e20 & 8.6[ned ] & sa(r)0/a & 0 & 2.6e08 & @xmath203.2e-02 & @xmath20 - 12.6 & + ngc4291 & 125.55 & 41.60 & 1.9 & 1.6 & 2.96e20 & 26.18[sbf ] & e & -5 & 1.3e10 & @xmath206.0e-02 & @xmath20 - 15.8 & + ngc4292 & 283.90 & 66.35 & 1.7 & 1.1 & 1.66e20 & 30.1[ned ] & sb(r)0o & -1.7 & 5e09 & @xmath203.8e-02 & @xmath20 - 15.6 & + ngc4298 & 272.36 & 75.67 & 3.2 & 1.8 & 2.68e20 & 16.8[t88 ] & sa(rs)c & 5 & 6.1e09 & 4.2e+00 & -19.6 & + ngc4302 & 272.51 & 75.69 & 5.5 & 1.0 & 2.69e20 & 16.8[t88 ] & sc & 5 & 3.8e09 & @xmath201.1e-01 & -20.0 & + ngc4303 & 284.37 & 66.27 & 6.5 & 5.8 & 1.67e20 & 15.2[t88 ] & sab(rs)bc & 4 & 2.9e10 & 2.4e+01 & -21.8 & hii + ngc4303a & 284.62 & 66.40 & 1.5 & 1.3 & 1.67e20 & 17.0[ned ] & sab(s)cd & 5.5 & 1.6e09 & 5.5e-01 & -17.4 & + ngc4319 & 125.44 & 41.66 & 3.0 & 2.3 & 2.91e20 & 28.2[t88 ] & sb(r)ab & 2 & 9.3e09 & 4.0e-01 & @xmath20 - 16.8 & + ngc4321 & 271.14 & 76.90 & 7.4 & 6.3 & 2.39e20 & 14.13[kp ] & sab(s)bc & 4 & 2.4e10 & 1.8e+01 & -21.5 & hii + ngc4328 & 271.53 & 76.95 & 1.3 & 1.2 & 2.39e20 & 6.4[ned ] & sa0- & -3 & 1.2e08 & @xmath203.3e-02 & @xmath20 - 12.0 & + ngc4374 & 278.20 & 74.48 & 6.5 & 5.6 & 2.60e20 & 18.37[sbf ] & e1 & -5 & 4e10 & 5.1e-01 & @xmath20 - 17.7 & + ngc4387 & 278.84 & 74.47 & 1.8 & 1.1 & 2.60e20 & 21.38[sbf ] & e & -5 & 4.3e09 & @xmath203.9e-02 & @xmath20 - 14.9 & + ngc4388 & 279.12 & 74.34 & 5.6 & 1.3 & 2.60e20 & 16.8[t88 ] & sa(s)b & 3 & 7.8e09 & 1.1e+01 & -20.6 & + ngc4395 & 162.08 & 81.54 & 13.2 & 11.0 & 1.35e20 & 3.6[t88 ] & sa(s)m & 9 & 8.5e08 & 2.1e+00 & -16.5 & + ngc4402 & 278.79 & 74.79 & 3.9 & 1.1 & 2.65e20 & 3.1[ned ] & sb & 3 & 1.1e08 & 5.8e+00 & -16.5 & + ngc4406 & 279.08 & 74.64 & 8.9 & 5.8 & 2.62e20 & 17.14[sbf ] & e3 & -5 & 4.8e10 & @xmath201.0e+00 & @xmath20 - 17.9 & + ngc4411a & 283.86 & 70.82 & 2.0 & 1.9 & 1.68e20 & 16.8[t88 ] & sb(rs)c & 5 & 1.7e09 & @xmath207.9e-02 & -16.9 & + ngc4411b & 284.07 & 70.85 & 2.5 & 2.5 & 1.68e20 & 16.8[t88 ] & sab(s)cd & 6 & 3e09 & 4.2e-01 & -17.3 & + ngc4413 & 279.82 & 74.35 & 2.3 & 1.5 & 2.61e20 & 16.8[t88 ] & sb(rs)ab & 2 & 2.8e09 & 1.0e+00 & -18.3 & + ngc4435 & 280.18 & 74.89 & 2.8 & 2.0 & 2.66e20 & 16.8[t88 ] & sb(s)0o & -2 & 9.2e09 & 2.1e+00 & -18.8 & hii + ngc4438 & 280.34 & 74.83 & 8.5 & 3.2 & 2.66e20 & 16.8[t88 ] & sa(s)0/a & 0 & 2.1e10 & 4.1e+00 & -19.7 & + ngc4449 & 136.85 & 72.40 & 6.2 & 4.4 & 1.37e20 & 3.0[t88 ] & ibm & 10 & 1.2e09 & @xmath205.4e-01 & @xmath20 - 13.4 & hii + ngc4464 & 286.50 & 70.32 & 1.1 & 0.8 & 1.66e20 & 16.8[t88 ] & s & & 1.4e09 & @xmath201.8e-02 & @xmath20 - 13.5 & + ngc4467 & 286.73 & 70.17 & 1.2 & 0.9 & 1.66e20 & 16.8[t88 ] & e2 & -5 & 8e08 & @xmath202.2e-02 & @xmath20 - 13.7 & + ngc4470 & 286.95 & 70.01 & 1.3 & 0.9 & 1.65e20 & 31.2[ned ] & sa & 1 & 7.8e09 & 1.9e+00 & -20.1 & hii + ngc4472 & 286.92 & 70.19 & 10.2 & 8.3 & 1.66e20 & 16.29[sbf ] & e2 & -5 & 7.8e10 & @xmath201.7e+00 & @xmath20 - 18.4 & + ngc4476 & 283.09 & 74.39 & 1.7 & 1.2 & 2.54e20 & 17.22[sbf ] & sa(r)0- & -3 & 2.9e09 & 5.9e-01 & -17.7 & + ngc4478 & 283.38 & 74.39 & 1.9 & 1.6 & 2.53e20 & 18.11[sbf ] & e2 & -5 & 6.8e09 & @xmath206.2e-02 & @xmath20 - 15.0 & + ngc4485 & 137.98 & 74.81 & 2.3 & 1.6 & 1.78e20 & 9.3[t88 ] & ib(s)m & 10 & 1.4e09 & @xmath207.5e-02 & @xmath20 - 13.7 & hii + ngc4486 & 283.78 & 74.49 & 8.3 & 6.6 & 2.54e20 & 16.07[sbf ] & e0 + & -4 & 5.5e10 & 4.3e-01 & @xmath20 - 17.9 & + ngc4490 & 138.00 & 74.87 & 6.3 & 3.1 & 1.78e20 & 7.8[t88 ] & sb(s)d & 7 & 6.9e09 & 4.0e+01 & @xmath20 - 15.2 & + ngc4501 & 282.33 & 76.51 & 6.9 & 3.7 & 2.48e20 & 16.8[t88 ] & sa(rs)b & 3 & 3.3e10 & 1.4e+01 & -21.5 & + ngc4523 & 283.12 & 77.36 & 2.0 & 1.9 & 2.33e20 & 16.8[t88 ] & sab(s)m & 8.5 & 1.2e09 & 3.4e-01 & -17.3 & + ngc4526 & 290.16 & 70.14 & 7.2 & 2.4 & 1.65e20 & 16.90[sbf ] & sab(s)0o & -2 & 3e10 & 5.9e+00 & -20.1 & + ngc4527 & 292.60 & 65.18 & 6.2 & 2.1 & 1.87e20 & 13.5[t88 ] & sab(s)bc & 4 & 7.7e09 & 2.6e+01 & -21.2 & hii + ngc4533 & 292.85 & 64.86 & 2.1 & 0.4 & 1.82e20 & 23.4[ned ] & sad & 7.1 & 1.5e09 & @xmath201.7e-02 & @xmath20 - 14.2 & + ngc4536 & 292.95 & 64.73 & 7.6 & 3.2 & 1.82e20 & 14.32[kp ] & sab(rs)bc & 4 & 1.3e10 & 3.0e+01 & -21.2 & hii + ngc4539 & 278.55 & 80.27 & 3.3 & 1.4 & 2.15e20 & 18.6[ned ] & sb(s)a & 0.5 & 3.3e09 & @xmath209.0e-02 & @xmath20 - 15.5 & + ngc4540 & 283.68 & 77.80 & 1.9 & 1.5 & 2.29e20 & 16.8[t88 ] & sab(rs)cd & 6 & 4.4e09 & 1.4e+00 & -18.7 & + ngc4545 & 126.09 & 53.50 & 2.5 & 1.5 & 1.76e20 & 28.9[t92 ] & sb(s)cd & 5.5 & 8.2e09 & 1.0e+00 & -19.4 & + ngc4552 & 287.93 & 74.97 & 5.1 & 4.7 & 2.57e20 & 15.35[sbf ] & e0 & -5 & 1.9e10 & @xmath204.8e-01 & @xmath20 - 16.8 & hii + ngc4559 & 198.35 & 86.47 & 10.7 & 4.4 & 1.49e20 & 5.8[t92 ] & sab(rs)cd & 6 & 3.9e09 & 5.3e+00 & -18.4 & hii + ngc4562 & 231.85 & 86.23 & 2.5 & 0.8 & 1.34e20 & 9.7[t88 ] & sb(s)dm & 8 & 3.6e08 & @xmath203.9e-02 & @xmath20 - 13.1 & + ngc4565 & 230.73 & 86.44 & 15.8 & 2.1 & 1.30e20 & 17.46[sbf ] & sa(s)b & 3 & 4.3e10 & 6.0e+00 & -21.1 & + ngc4569 & 288.47 & 75.62 & 9.5 & 4.4 & 2.49e20 & 16.8[t88 ] & sab(rs)ab & 2 & 3.2e10 & 7.2e+00 & -20.7 & + ngc4586 & 294.64 & 66.98 & 4.0 & 1.3 & 1.88e20 & 17.5[t88 ] & sa(s)a & 1 & 4.5e09 & 5.5e-01 & -17.6 & + ngc4593 & 297.48 & 57.40 & 3.9 & 2.9 & 2.31e20 & 39.5[t88 ] & sb(rs)b & 3 & 5.2e10 & 2.8e+00 & -21.0 & + ngc4594 & 298.46 & 51.15 & 8.7 & 3.5 & 3.77e20 & 9.77[sbf ] & sa(s)a & 1 & 5.5e10 & 3.2e+00 & -19.0 & + ngc4605 & 125.33 & 55.47 & 5.8 & 2.2 & 1.43e20 & 4.0[t88 ] & sb(s)c & 5 & 1.1e09 & 1.2e+01 & -17.7 & hii + ngc4627 & 142.96 & 84.18 & 2.6 & 1.8 & 1.27e20 & 7.2[ned ] & e4 & -5 & 5.4e08 & @xmath209.2e-02 & @xmath20 - 13.4 & + ngc4631 & 142.82 & 84.22 & 15.5 & 2.7 & 1.27e20 & 6.9[t88 ] & sb(s)d & 7 & 1e10 & 5.2e+01 & -21.0 & + ngc4636 & 297.75 & 65.47 & 6.0 & 4.7 & 1.81e20 & 14.66[sbf ] & e0 & -5 & 1.9e10 & @xmath205.6e-01 & @xmath20 - 16.9 & + ngc4638 & 295.14 & 74.18 & 2.2 & 1.4 & 2.17e20 & 21.68[sbf ] & s0- & -3 & 1.1e10 & @xmath206.2e-02 & @xmath20 - 15.4 & + ngc4639 & 294.30 & 75.99 & 2.8 & 1.9 & 2.35e20 & 20.8[kp ] & sab(rs)bc & 4 & 9.6e09 & 1.4e+00 & -19.3 & + ngc4647 & 295.75 & 74.34 & 2.9 & 2.3 & 2.20e20 & 16.8[t88 ] & sab(rs)c & 5 & 6.5e09 & 5.0e+00 & -20.1 & + ngc4649 & 295.88 & 74.31 & 7.4 & 6.0 & 2.20e20 & 16.83[sbf ] & e2 & -5 & 6e10 & @xmath209.0e-01 & @xmath20 - 17.7 & + ngc4651 & 293.07 & 79.12 & 4.0 & 2.6 & 1.99e20 & 16.8[t88 ] & sa(rs)c & 5 & 1.3e10 & 5.2e+00 & -20.1 & + ngc4656 & 140.32 & 84.70 & 15.1 & 3.0 & 1.23e20 & 7.2[t88 ] & sb(s)m & 9 & 4.8e09 & 2.2e+00 & -18.1 & + ngc4666 & 299.54 & 62.37 & 4.6 & 1.3 & 1.74e20 & 14.1[t88 ] & sabc & 5 & 6.3e09 & 2.6e+01 & @xmath20 - 15.2 & + ngc4668 & 299.75 & 62.30 & 1.4 & 0.8 & 1.74e20 & 21.6[ned ] & sb(s)d & 7 & 3e09 & 7.2e-01 & -18.3 & + ngc4697 & 301.63 & 57.06 & 7.2 & 4.7 & 2.12e20 & 11.75[sbf ] & e6 & -5 & 1.9e10 & @xmath206.8e-01 & @xmath20 - 16.7 & + ngc4736 & 123.36 & 76.01 & 11.2 & 9.1 & 1.44e20 & 5.20[sbf ] & sa(r)ab & 2 & 1.4e10 & 5.6e+01 & -20.0 & + ngc4757 & 303.50 & 52.56 & 1.5 & 0.4 & 3.46e20 & 11.3[ned ] & s0 & -2 & & @xmath201.1e-02 & @xmath20 - 12.1 & + ngc4758 & 304.53 & 78.72 & 3.1 & 0.7 & 1.96e20 & 16.2[t88 ] & i m & 10 & 1.2e09 & 9.2e-01 & -18.2 & + ngc4772 & 304.14 & 65.04 & 3.4 & 1.7 & 1.79e20 & 16.3[t88 ] & sa(s)a & 1 & 4.4e09 & @xmath201.2e-01 & @xmath20 - 15.4 & + ngc4826 & 315.71 & 84.42 & 10.0 & 5.4 & 2.57e20 & 7.48[sbf ] & sa(rs)ab & 2 & 2e10 & 3.7e+01 & -20.2 & + ngc4861 & 111.52 & 82.10 & 4.0 & 1.5 & 1.19e20 & 17.8[t88 ] & sb(s)m & 9 & 4.7e09 & 1.8e+00 & @xmath20 - 15.7 & sbrst + ngc4904 & 308.15 & 62.75 & 2.2 & 1.4 & 1.55e20 & 15.6[ned ] & sb(s)cd & 6 & 2.9e09 & 2.5e+00 & -19.0 & sbrst + ngc5005 & 101.61 & 79.25 & 5.8 & 2.8 & 1.08e20 & 21.3[t88 ] & sab(rs)bc & 4 & 3.1e10 & 1.9e+01 & -22.0 & + ngc5033 & 98.05 & 79.45 & 10.7 & 5.0 & 9.96e19 & 15.2[t92 ] & sa(s)c & 5 & 2.1e10 & 1.3e+01 & -21.1 & + ngc5035 & 311.02 & 46.01 & 1.4 & 1.1 & 4.99e20 & 29.1[ned ] & sab(r)0 + & -1 & & 3.7e-01 & @xmath20 - 15.3 & + ngc5037 & 311.05 & 45.91 & 2.2 & 0.7 & 5.10e20 & 29.0[t88 ] & sa(s)a & 1 & 6.9e09 & 7.1e-01 & -19.3 & + ngc5044 & 311.23 & 46.10 & 3.0 & 3.0 & 4.93e20 & 31.19[sbf ] & e0 & -5 & 2.8e10 & @xmath201.8e-01 & @xmath20 - 17.3 & + ngc5049 & 311.43 & 46.07 & 1.9 & 0.6 & 4.98e20 & 39.3[t88 ] & s0 & -2 & & @xmath202.2e-02 & @xmath20 - 15.5 & + ngc5055 & 106.01 & 74.28 & 12.6 & 7.2 & 1.32e20 & 8.5[t92 ] & sa(rs)bc & 4 & 1.7e10 & 2.8e+01 & -20.9 & hii + ngc5076 & 313.52 & 49.55 & 1.3 & 0.7 & 3.23e20 & 39.9[ned ] & sb(rs)0 + & -1 & & @xmath201.9e-02 & @xmath20 - 15.4 & + ngc5079 & 313.57 & 49.59 & 1.6 & 1.0 & 3.19e20 & 38.3[ned ] & sb(rs)bc & 3.5 & & 2.4e+00 & -20.9 & + ngc5088 & 313.87 & 49.68 & 2.6 & 0.8 & 3.09e20 & 24.5[t88 ] & sab(s)bc & 3.7 & 4.9e09 & 1.8e+00 & -19.5 & + ngc5089 & 61.41 & 83.07 & 1.7 & 0.9 & 1.12e20 & 35.1[t88 ] & s & & 4.7e09 & 6.8e-01 & -19.2 & + ngc5117 & 44.43 & 82.93 & 2.2 & 1.0 & 1.18e20 & 38.3[t88 ] & sbcd & 6 & 6.2e09 & 3.2e-01 & @xmath20 - 16.3 & + ngc5128 & 309.52 & 19.42 & 25.7 & 19.9 & 8.62e20 & 4.21[sbf ] & s0 & -2 & 2.3e10 & 1.7e+02 & -20.9 & + ngc5194 & 104.85 & 68.56 & 11.2 & 6.9 & 1.57e20 & 7.7[t88 ] & sa(s)bc & 4 & 2.1e10 & 3.2e+01 & -21.6 & hii + ngc5195 & 104.89 & 68.48 & 5.8 & 4.6 & 1.56e20 & 7.66[sbf ] & i0 & 90 & 5e09 & 1.0e+01 & @xmath20 - 15.4 & + ngc5204 & 113.50 & 58.01 & 5.0 & 3.0 & 1.39e20 & 4.3[t92 ] & sa(s)m & 9 & 6.2e08 & 2.4e+00 & -16.1 & hii + ngc5236 & 314.58 & 31.97 & 12.9 & 11.5 & 3.82e20 & 4.7[t88 ] & sab(s)c & 5 & 1.6e10 & 1.0e+02 & -21.4 & sbrst + ngc5238 & 107.40 & 64.19 & 1.7 & 1.4 & 1.10e20 & 4.9[t88 ] & sab(s)dm & 8 & 1e08 & @xmath204.9e-02 & @xmath20 - 11.9 & + ngc5273 & 74.35 & 76.25 & 2.8 & 2.5 & 9.40e19 & 16.52[sbf ] & sa(s)0o & -2 & 4.8e09 & @xmath201.4e-01 & @xmath20 - 15.7 & + ngc5322 & 110.28 & 55.49 & 5.9 & 3.9 & 1.81e20 & 31.19[sbf ] & e3 & -5 & 6e10 & 4.1e-01 & @xmath20 - 18.3 & + ngc5350 & 82.84 & 71.59 & 3.2 & 2.3 & 9.79e19 & 37.8[t88 ] & sb(r)b & 3 & 2.7e10 & 2.2e+00 & -21.0 & sbrst + ngc5353 & 82.61 & 71.63 & 2.2 & 1.1 & 9.76e19 & 37.8[t88 ] & s0 & -2 & 3.2e10 & 3.0e-01 & @xmath20 - 16.3 & + ngc5354 & 82.66 & 71.62 & 1.4 & 1.3 & 9.77e19 & 34.3[ned ] & s0 & -2 & 1.8e10 & @xmath203.8e-02 & @xmath20 - 15.8 & + ngc5355 & 82.62 & 71.55 & 1.2 & 0.7 & 9.86e19 & 31.3[ned ] & s0 & -2 & 3.9e09 & @xmath201.7e-02 & @xmath20 - 14.8 & + ngc5358 & 82.37 & 71.56 & 1.1 & 0.3 & 9.87e19 & 32.0[ned ] & s0/a & 0 & 2.1e09 & @xmath207.4e-03 & @xmath20 - 13.9 & + ngc5371 & 82.17 & 71.19 & 4.4 & 3.5 & 1.03e20 & 37.8[t88 ] & sab(rs)bc & 4 & 6.4e10 & 2.6e+00 & -21.4 & + ngc5457 & 102.04 & 59.77 & 28.8 & 26.9 & 1.16e20 & 6.855[kp ] & sab(rs)cd & 6 & 4.6e10 & 3.9e+00 & -21.1 & + ngc5506 & 339.15 & 53.81 & 2.8 & 0.9 & 3.81e20 & 28.7[t88 ] & sa & 1 & 6.4e09 & 8.8e+00 & -21.3 & + ngc5507 & 339.23 & 53.85 & 1.7 & 0.9 & 3.84e20 & 34.3[t88 ] & sab(r)0o & -2.3 & 7.4e09 & @xmath203.2e-02 & @xmath20 - 15.7 & + ngc5774 & 359.41 & 52.49 & 3.0 & 2.5 & 3.46e20 & 26.8[t88 ] & sab(rs)d & 7 & 8.4e09 & 5.6e-01 & -19.1 & + ngc5775 & 359.43 & 52.42 & 4.2 & 1.0 & 3.48e20 & 26.7[t88 ] & sbc & 5 & 1.2e10 & 1.5e+01 & -22.2 & + ngc5813 & 359.18 & 49.85 & 4.2 & 3.0 & 4.21e20 & 32.21[sbf ] & e1 & -5 & 3.2e10 & @xmath202.5e-01 & @xmath20 - 17.8 & + ngc5839 & .20 & 49.01 & 1.3 & 1.2 & 4.28e20 & 22.59[sbf ] & sab(rs)0o & -2 & 2.8e09 & @xmath203.0e-02 & @xmath20 - 14.7 & + ngc5846 & .43 & 48.80 & 4.1 & 3.8 & 4.26e20 & 24.89[sbf ] & e0 & -5 & 3.2e10 & @xmath203.1e-01 & @xmath20 - 17.4 & hii + ngc5850 & .52 & 48.64 & 4.3 & 3.7 & 4.26e20 & 28.5[t88 ] & sb(r)b & 3 & 1.9e10 & 7.3e-01 & -19.5 & + ngc5951 & 23.52 & 50.45 & 3.5 & 0.8 & 3.33e20 & 20.4[t92 ] & sbc & 5.3 & 2.9e09 & 6.0e-01 & -18.2 & + ngc5953 & 23.92 & 50.35 & 1.6 & 1.3 & 3.27e20 & 33.0[t88 ] & saa & 1 & & 1.0e+01 & -22.0 & + ngc5954 & 23.94 & 50.34 & 1.3 & 0.6 & 3.26e20 & 32.1[t88 ] & sab(rs)cd & 6 & & 1.0e+01 & @xmath20 - 14.7 & + ngc6217 & 111.32 & 33.36 & 3.0 & 2.5 & 4.04e20 & 23.9[t88 ] & sb(rs)bc & 4 & 1.3e10 & 1.0e+01 & -21.4 & + ngc6434 & 102.97 & 31.43 & 2.3 & 1.0 & 3.73e20 & 33.1[ned ] & sbbc & 4 & 8.8e09 & 1.8e+00 & -20.2 & hii + ngc6503 & 100.57 & 30.64 & 7.1 & 2.4 & 4.09e20 & 6.1[t88 ] & sa(s)cd & 6 & 2.2e09 & 7.2e+00 & -18.3 & hii + ngc6654 & 103.97 & 27.84 & 2.6 & 2.1 & 5.42e20 & 29.5[t88 ] & sb(s)0/a & 0 & 1.2e10 & 4.0e-01 & @xmath20 - 16.7 & + ngc6764 & 81.50 & 18.23 & 2.3 & 1.3 & 6.07e20 & 25.9[t92 ] & sb(s)bc & 3.5 & 6e09 & 6.5e+00 & -21.0 & + ngc6946 & 95.72 & 11.67 & 11.5 & 9.8 & 2.11e21 & 5.5[t88 ] & sab(rs)cd & 6 & 6.5e09 & 5.2e+01 & -21.1 & hii + ngc7172 & 15.13 & -53.06 & 2.5 & 1.4 & 1.65e20 & 33.9[t88 ] & sa & 1.4 & 1.4e10 & 6.0e+00 & -21.5 & + ngc7173 & 14.96 & -53.08 & 1.2 & 0.9 & 1.64e20 & 31.33[sbf ] & e+ & -4.1 & 2.2e10 & @xmath202.1e-02 & @xmath20 - 15.0 & + ngc7174 & 14.93 & -53.09 & 2.3 & 1.2 & 1.64e20 & 35.5[t88 ] & sab & 1.8 & 2.4e10 & 3.5e+00 & -21.1 & + ngc7176 & 14.93 & -53.10 & 1.0 & 0.8 & 1.64e20 & 32.3[t88 ] & e & -4.6 & 2.3e10 & 3.5e+00 & @xmath20 - 14.8 & + ngc7213 & 349.59 & -52.58 & 3.1 & 2.8 & 2.04e20 & 22.0[t88 ] & sa(s)a & 1 & 1.9e10 & 2.6e+00 & -19.9 & + ngc7217 & 86.51 & -19.71 & 3.9 & 3.2 & 1.06e21 & 16.0[t88 ] & sa(r)ab & 2 & 1.3e10 & 4.8e+00 & -20.0 & + ngc7314 & 27.13 & -59.74 & 4.6 & 2.1 & 1.46e20 & 12.9[t92 ] & sab(rs)bc & 4 & 5.7e09 & 3.4e+00 & -19.2 & + ngc7332 & 87.39 & -29.67 & 4.1 & 1.1 & 4.57e20 & 23.01[sbf ] & s0 & -2 & 1.2e10 & @xmath209.1e-02 & @xmath20 - 15.9 & + ngc7339 & 87.46 & -29.73 & 3.0 & 0.7 & 4.57e20 & 19.3[t88 ] & sab(s)bc & 4 & 3.4e09 & @xmath204.5e-02 & @xmath20 - 14.8 & + ngc7463 & 88.04 & -39.35 & 2.9 & 0.7 & 6.04e20 & 33.5[t88 ] & sabb & 3 & 8.1e09 & @xmath203.9e-02 & @xmath20 - 15.8 & + ngc7465 & 88.07 & -39.39 & 1.2 & 0.8 & 6.03e20 & 27.2[t88 ] & sb(s)0o & -1.5 & 5.4e09 & 2.7e+00 & -20.3 & + ngc7507 & 23.44 & -68.04 & 2.8 & 2.7 & 2.47e20 & 25.00[sbf ] & e0 & -5 & 2.1e10 & @xmath201.5e-01 & @xmath20 - 16.6 & + ngc7552 & 348.15 & -65.24 & 3.4 & 2.7 & 1.85e20 & 19.5[t88 ] & sb(s)ab & 2 & 1.7e10 & 7.3e+01 & -22.9 & hii + ngc7582 & 348.08 & -65.70 & 5.0 & 2.1 & 1.93e20 & 17.6[t88 ] & sb(s)ab & 2 & 1.5e10 & 4.8e+01 & -22.2 & + ngc7590 & 348.23 & -65.85 & 2.7 & 1.0 & 1.96e20 & 17.3[t88 ] & sa(rs)bc & 4 & 6.1e09 & 7.4e+00 & -20.4 & + ngc7599 & 348.09 & -65.91 & 4.4 & 1.3 & 1.98e20 & 20.2[t88 ] & sa(s)c & 5 & 1.1e10 & 6.1e+00 & -20.6 & + ngc7714 & 88.22 & -55.57 & 1.9 & 1.4 & 4.86e20 & 36.9[t88 ] & sb(s)b & 3 & 1.4e10 & 1.1e+01 & -22.1 & hii + ngc7715 & 88.27 & -55.58 & 2.6 & 0.5 & 4.86e20 & 36.3[t88 ] & i m & 10 & 3.3e09 & @xmath202.6e-02 & @xmath20 - 15.6 & + ngc7742 & 97.43 & -48.75 & 1.7 & 1.7 & 5.14e20 & 22.2[t88 ] & sa(r)b & 3 & 8e09 & 3.0e+00 & -19.9 & hii + pgc4013 & 131.57 & -61.53 & 1.1 & 0.9 & 3.13e20 & 8.9[ned ] & s & & 1.3e08 & @xmath202.1e-02 & @xmath20 - 12.3 & + pgc5208 & 138.72 & -57.99 & 1.2 & 0.9 & 3.30e20 & 28.7[ned ] & i m & 10 & & @xmath202.2e-02 & @xmath20 - 14.9 & + pgc12827 & 212.16 & -54.42 & 2.3 & 0.7 & 2.23e20 & 28.3[ned ] & s & & 8e08 & @xmath203.3e-02 & @xmath20 - 15.3 & + pgc13381 & 215.95 & -52.26 & 1.3 & 0.6 & 2.02e20 & 17.9[ned ] & s0 & & 4.7e08 & @xmath201.6e-02 & @xmath20 - 13.5 & + pgc13449 & 236.58 & -53.49 & 1.3 & 1.0 & 1.30e20 & 11.0[ned ] & sab(s)0o & -2 & & @xmath202.7e-02 & @xmath20 - 13.0 & + pgc13452 & 237.17 & -53.46 & 1.1 & 1.1 & 1.40e20 & 12.0[ned ] & e0 & -5 & & @xmath202.4e-02 & @xmath20 - 13.1 & + pgc13826 & 138.17 & 10.58 & 21.4 & 20.9 & 3.02e21 & 3.9[t88 ] & sab(rs)cd & 6 & 5.4e09 & 8.6e+01 & -21.0 & hii + pgc14656 & 233.10 & -46.47 & 1.1 & 0.6 & 1.73e20 & 16.8[ned ] & sab(s)0o & -2 & 6.2e08 & @xmath201.4e-02 & @xmath20 - 13.2 & + pgc14664 & 233.38 & -46.44 & 1.7 & 1.1 & 1.72e20 & 11.7[ned ] & ib(s)m & 10 & 2.7e08 & @xmath203.9e-02 & @xmath20 - 13.5 & + pgc15455 & 256.76 & -42.55 & 2.6 & 0.4 & 1.71e20 & 21.5[t88 ] & sb & 3 & 3.2e09 & 1.4e+00 & -19.1 & + pgc16790 & 242.20 & -35.94 & 3.5 & 2.7 & 2.96e20 & 11.1[t88 ] & sb(s)dm & 8 & 1.1e09 & 2.3e-01 & @xmath20 - 15.2 & + pgc20844 & 143.69 & 28.08 & 1.8 & 1.1 & 3.76e20 & 38.8[v3k ] & scd & 6 & 1.3e10 & 8.0e-01 & -20.0 & + pgc21406 & 281.13 & -21.37 & 1.6 & 1.5 & 1.19e21 & 19.4[ned ] & ibm & 10 & & @xmath204.8e-02 & @xmath20 - 14.9 & + pgc21457 & 281.30 & -21.33 & 1.9 & 1.0 & 1.17e21 & 20.4[ned ] & sb(s)0/a & -0.5 & 2.6e09 & @xmath204.0e-02 & @xmath20 - 14.8 & + pgc23324 & 144.28 & 32.69 & 7.9 & 6.3 & 3.42e20 & 4.5[t88 ] & i m & 10 & 1.4e09 & @xmath201.0e+00 & -15.4 & + pgc23852 & 207.51 & 29.32 & 1.5 & 0.9 & 3.79e20 & 27.7[ned ] & sb(s)cd & 6 & 2.7e09 & 3.0e-01 & @xmath20 - 14.9 & + pgc26068 & 181.09 & 43.81 & 1.7 & 1.4 & 1.80e20 & 33.3[ned ] & sab(s)d & 7 & 2.3e09 & @xmath204.6e-02 & @xmath20 - 16.0 & + pgc28757 & 141.98 & 41.06 & 2.5 & 2.0 & 4.07e20 & 3.42[kp ] & i m & 10 & & @xmath201.0e-01 & @xmath20 - 11.9 & + pgc29584 & 158.67 & 49.93 & 1.4 & 0.9 & 7.53e19 & 14.7[ned ] & s & & & @xmath202.6e-02 & @xmath20 - 13.6 & + pgc30819 & 140.20 & 43.60 & 13.2 & 5.4 & 2.37e20 & 3.5[t92 ] & sab(s)m & 9 & 9e08 & @xmath201.4e+00 & -16.0 & + pgc31586 & 269.55 & 26.99 & 2.6 & 1.8 & 5.08e20 & 30.2[ned ] & e4 + & -4 & 1.1e10 & @xmath209.4e-02 & @xmath20 - 16.6 & + pgc32390 & 193.39 & 63.07 & 1.3 & 0.9 & 2.01e20 & 21.8[ned ] & sb(s)m & 9.3 & 9.7e08 & 3.4e-01 & @xmath20 - 14.3 & + pgc32604 & 151.76 & 54.33 & 1.3 & 1.1 & 7.80e19 & 15.9[ned ] & scd & 6 & 4.3e08 & @xmath202.8e-02 & @xmath20 - 13.8 & + pgc35286 & 127.84 & 37.33 & 1.4 & 0.8 & 3.83e20 & 1.4[t88 ] & peculiar & 99 & 4.8e06 & 3.8e-01 & @xmath20 - 8.4 & + pgc35450 & 181.07 & 70.67 & 1.5 & 0.6 & 2.13e20 & 29.6[ned ] & i m & 10 & & @xmath201.7e-02 & @xmath20 - 14.7 & + pgc35684 & 151.91 & 63.28 & 1.2 & 0.7 & 1.54e20 & 4.0[t88 ] & i m & 10 & 6.4e07 & @xmath201.6e-02 & @xmath20 - 10.3 & + pgc37532 & 138.38 & 59.83 & 1.4 & 0.4 & 1.17e20 & 18.1[ned ] & sdm & 8 & 4.3e08 & @xmath201.2e-02 & @xmath20 - 13.2 & + pgc37580 & 288.51 & 33.65 & 1.1 & 0.9 & 6.21e20 & 23.8[ned ] & sb(r)dm & 7.9 & 1.2e09 & @xmath202.0e-02 & @xmath20 - 14.4 & + pgc37951 & 132.68 & 53.85 & 1.6 & 0.5 & 2.00e20 & 21.2[t88 ] & i m & 10 & 1.1e09 & 2.5e-01 & @xmath20 - 13.9 & + pgc38811 & 154.65 & 75.15 & 1.0 & 0.9 & 1.97e20 & 13.3[ned ] & sbm & 9 & & @xmath201.9e-02 & @xmath20 - 13.0 & + pgc40045 & 270.47 & 76.79 & 1.1 & 0.8 & 2.42e20 & 17.3[ned ] & i & & & @xmath201.9e-02 & @xmath20 - 13.6 & + pgc41054 & 281.29 & 74.82 & 1.3 & 0.4 & 2.64e20 & 6.3[ned ] & s0 + & -1 & 6.6e07 & @xmath201.1e-02 & @xmath20 - 10.8 & + pgc41258 & 287.14 & 70.14 & 1.1 & 0.9 & 1.65e20 & 3.7[ned ] & i m & 10 & 2.5e07 & @xmath202.0e-02 & @xmath20 - 10.3 & + pgc41763 & 283.00 & 77.55 & 1.5 & 1.1 & 2.31e20 & 31.1[ned ] & sb(rs)c & 4.7 & 3.5e09 & 3.8e-01 & @xmath20 - 15.5 & + pgc42081 & 288.28 & 75.71 & 2.2 & 1.1 & 2.48e20 & 16.8[t88 ] & ibm & 10 & 1.3e09 & 6.1e-01 & -17.6 & + pgc45561 & 117.40 & 62.87 & 1.6 & 0.6 & 1.57e20 & 33.0[ned ] & sb & & 3.6e09 & 3.6e-01 & @xmath20 - 14.9 & + pgc46093 & 106.40 & 74.36 & 1.6 & 0.6 & 1.32e20 & 5.2[ned ] & i m & 10 & & @xmath201.8e-02 & @xmath20 - 11.0 & + pgc53683 & 359.57 & 49.78 & 1.4 & 0.5 & 4.22e20 & 16.6[ned ] & sb(rs)dm & 8 & 6.1e08 & 2.5e-01 & @xmath20 - 13.2 & + pgc55095 & 67.94 & 55.22 & 1.6 & 0.5 & 1.98e20 & 37.1[t88 ] & ibm & 10 & & @xmath201.5e-02 & @xmath20 - 15.0 & + pgc58052 & 96.22 & 39.82 & 1.4 & 0.7 & 2.62e20 & 36.8[ned ] & sdm & 8 & 2.1e09 & @xmath202.0e-02 & @xmath20 - 15.3 & + pgc59551 & 104.03 & 33.74 & 1.8 & 0.3 & 4.02e20 & 14.4[ned ] & sb & 3 & 8e08 & 1.1e+00 & -18.0 & hii + pgc67352 & 10.04 & -50.17 & 1.4 & 1.4 & 2.04e20 & 33.3[t88 ] & sb0- & -3 & 8.2e09 & @xmath204.0e-02 & @xmath20 - 15.8 & + pgc67707 & 344.49 & -49.43 & 1.5 & 0.9 & 2.17e20 & 36.0[ned ] & ibm & 10 & 1.2e09 & @xmath202.6e-02 & @xmath20 - 15.5 & + pgc67910 & 14.94 & -53.22 & 1.2 & 0.4 & 1.63e20 & 30.9[ned ] & s & & 1.8e09 & @xmath209.1e-03 & @xmath20 - 14.1 & + pgc68498 & 352.26 & -54.53 & 1.3 & 1.0 & 2.24e20 & 22.5[t88 ] & ib(s)m & 10 & 8.6e08 & @xmath202.6e-02 & @xmath20 - 14.5 & + pgc68618 & 350.32 & -54.84 & 8.5 & 3.9 & 1.80e20 & 11.1[t88 ] & sb(rs)cd & 6 & 4.3e09 & 3.3e-01 & -17.6 & + pgc69404 & 60.93 & -51.99 & 2.6 & 0.9 & 4.01e20 & 38.9[ned ] & sab(s)dm & 7.6 & & @xmath204.6e-02 & @xmath20 - 16.3 & + pgc70070 & 5.19 & -64.04 & 4.1 & 0.8 & 1.19e20 & 20.5[t88 ] & sb(rs)cd & 5.5 & 2.7e09 & @xmath206.4e-02 & @xmath20 - 15.3 & + pgc70081 & 4.48 & -64.03 & 2.5 & 0.5 & 1.18e20 & 25.9[ned ] & sab & 2.3 & 4.4e09 & 5.1e-01 & -18.7 & + pgc70090 & 4.67 & -64.11 & 5.2 & 3.8 & 1.18e20 & 29.24[sbf ] & e3 & -5 & 6e10 & 4.2e-01 & @xmath20 - 18.1 & + pgc70110 & 5.58 & -64.31 & 1.8 & 0.8 & 1.21e20 & 26.9[t88 ] & sab(rs)0 & -1.8 & 5.9e09 & @xmath202.9e-02 & @xmath20 - 15.0 & + pgc70117 & 5.94 & -64.39 & 3.2 & 0.6 & 1.23e20 & 25.7[ned ] & sb(r)cd & 5.7 & 6.5e09 & 3.1e+00 & -20.4 & hii + pgc70285 & 88.05 & -39.26 & 1.4 & 0.3 & 6.07e20 & 27.0[ned ] & i m & 10 & & @xmath208.0e-03 & @xmath20 - 13.6 & + pgc71120 & 87.53 & -48.51 & 1.2 & 1.1 & 5.04e20 & 37.5[ned ] & sm & 9 & & @xmath202.6e-02 & @xmath20 - 15.6 & + ngc247 & x1 & 00:47:10.30 & -20:47:10.7 & 0 & 95.4 & 0.16 & 5.9 & 0.133 & & & & & + ngc247 & x2 & 00:47:04.00 & -20:47:45.7 & 1 & 143.8 & 0.42 & 32.4 & 0.853 & & d & & & + ngc247 & x3 & 00:46:51.78 & -20:43:27.2 & 1 & 271.1 & 1.10 & 17.9 & 0.427 & & & qso phl6625 & & + ngc247 & x4 & 00:47:27.57 & -20:47:24.6 & 1 & 285.2 & 1.23 & 12.3 & 0.265 & & & p b2/18.83 & & + ngc247 & x5 & 00:47:03.69 & -20:40:15.5 & 0 & 330.1 & 0.54 & 3.7 & 0.0628 & & & & & + ngc247 & x6 & 00:46:47.98 & -20:38:36.9 & 1 & 511.8 & 1.37 & 5.3 & 0.095 & & & mz & & + ngc247 & x7 & 00:47:03.31 & -20:36:57.1 & 1 & 526.4 & 0.83 & 9.3 & 0.243 & & d & & & + ngc253 & x1 & 00:47:33.55 & -25:17:24.4 & 1 & 8.9 & 0.04 & 15.4 & 1.74 & & & ( ngc253 ) & n & + ngc253 & x2 & 00:47:32.97 & -25:17:50.2 & 0 & 32.3 & 0.12 & 23.5 & 0.823 & & & & & + ngc253 & x3 & 00:47:29.69 & -25:17:03.4 & 0 & 48.5 & 0.20 & 4.0 & 0.157 & & & & & + ngc253 & x4 & 00:47:34.87 & -25:15:11.0 & 0 & 129.3 & 0.44 & 11.0 & 0.3 & & & & & + ngc253 & x5 & 00:47:24.93 & -25:19:47.0 & 0 & 185.6 & 0.33 & 3.5 & 0.063 & & & & & + ngc253 & x6 & 00:47:42.58 & -25:15:00.4 & 0 & 188.3 & 0.27 & 13.6 & 0.322 & & & & & + ngc253 & x7 & 00:47:17.37 & -25:18:12.7 & 1 & 220.3 & 0.50 & 12.3 & 0.451 & & & & & + ngc253 & x8 & 00:47:18.10 & -25:19:13.6 & 0 & 234.0 & 0.33 & 5.1 & 0.131 & & & & & + ngc253 & x9 & 00:47:22.43 & -25:20:51.8 & 1 & 258.1 & 0.49 & 12.6 & 0.631 & & & & & + ngc253 & x10 & 00:47:10.68 & -25:20:55.1 & 0 & 373.5 & 0.46 & 13.0 & 0.419 & & & & & + ngc253 & x11 & 00:47:09.27 & -25:14:01.5 & 0 & 378.4 & 1.75 & 4.3 & 0.0786 & & & & & + ngc253 & x12 & 00:47:05.29 & -25:19:44.0 & 0 & 404.4 & 0.74 & 5.6 & 0.122 & & & & & + ngc253 & x13 & 00:47:09.08 & -25:21:25.0 & 0 & 408.7 & 0.50 & 10.3 & 0.203 & & & & & + ngc253 & x14 & 00:47:22.89 & -25:10:52.2 & 0 & 409.9 & 1.93 & 6.6 & 0.149 & & & p b1/19.42 & & + ngc253 & x15 & 00:47:31.49 & -25:09:58.2 & 0 & 440.3 & 1.80 & 3.4 & 0.0652 & & & & & + ngc253 & x16 & 00:47:10.86 & -25:23:33.9 & 0 & 481.9 & 0.79 & 3.5 & 0.0589 & & & p & & + ngc253 & x17 & 00:47:50.55 & -25:08:39.0 & 0 & 570.5 & 1.43 & 4.3 & 0.11 & & & p b2/21.06 & & + ngc253 & x18 & 00:48:00.26 & -25:09:51.8 & 1 & 578.7 & 0.92 & 5.8 & 0.21 & & & qso j0047.9 - 2509 & & + ngc253 & x19 & 00:46:47.16 & -25:21:51.6 & 0 & 680.3 & 1.15 & 6.8 & 0.159 & & & mp & & + ngc253 & x20 & 00:48:19.99 & -25:10:09.8 & 0 & 767.0 & 0.96 & 38.6 & 2.37 & & & & 1ulx & ulx1 + ngc253 & x21 & 00:48:29.81 & -25:08:40.1 & 0 & 927.6 & 1.17 & 3.3 & 0.194 & & & & & + ngc300 & x1 & 00:54:40.33 & -37:40:46.0 & 1 & 160.3 & 0.25 & 4.2 & 0.0535 & & s & & & + ngc300 & x2 & 00:55:10.07 & -37:42:16.2 & 1 & 208.7 & 0.32 & 24.2 & 0.367 & 1 & s & & & + ngc300 & x4 & 00:55:11.18 & -37:48:41.1 & 1 & 507.8 & 0.95 & 9.5 & 0.157 & & d & & & + ngc300 & x5 & 00:54:44.86 & -37:51:07.1 & 1 & 619.3 & 1.32 & 4.2 & 0.0511 & & & & & + ngc300 & x6 & 00:55:41.74 & -37:35:33.5 & 1 & 655.0 & 1.26 & 5.1 & 0.0855 & & & p b2/20.34 & & + ngc300 & x7 & 00:55:26.82 & -37:31:26.2 & 1 & 692.6 & 1.47 & 17.4 & 0.314 & 1 & & g5 * & & + ngc300 & x8 & 00:54:14.18 & -37:29:45.2 & 1 & 820.4 & 1.43 & 7.0 & 0.122 & & & & & + ngc300 & x9 & 00:56:01.65 & -37:32:47.9 & 1 & 943.0 & 1.83 & 4.2 & 0.151 & & & & & + ngc404 & x1 & 01:09:27.01 & 35:43:04.2 & 2 & 0.2 & 0.00 & 4.7 & 0.178 & & & ( ngc404 ) & n & + ngc613 & x1 & 01:34:18.17 & -29:25:06.1 & 2 & 12.0 & 0.07 & 20.6 & 76.8 & & & ( ngc613 ) & n & + ngc720 & x1 & 01:53:00.48 & -13:44:18.7 & 1 & 2.5 & 0.03 & 25.3 & 73.2 & & & ( ngc720 ) & n & + ngc720 & x2 & 01:53:03.50 & -13:46:54.6 & 1 & 160.1 & 1.37 & 9.3 & 12.7 & & & & 2eulxd & + ngc720 & x3 & 01:52:49.38 & -13:42:17.4 & 1 & 202.6 & 1.53 & 4.5 & 5.86 & & & & 2ulxd & + ngc891 & x1 & 02:22:31.39 & 42:20:30.3 & 8 & 25.9 & 0.16 & 11.4 & 1.53 & & d & & 1ulxd & ulx1 + ngc891 & x2 & 02:22:31.27 & 42:19:58.8 & 8 & 53.2 & 0.13 & 27.0 & 3.77 & & & ( sn1986j ) & 1ulxd & ulx2 + ngc891 & x3 & 02:22:30.42 & 42:19:14.5 & 8 & 98.1 & 0.26 & 3.8 & 0.367 & & d & & & + ngc891 & x4 & 02:22:24.31 & 42:21:40.0 & 8 & 110.5 & 1.45 & 5.7 & 0.633 & & & & & + ngc891 & x5 & 02:22:26.31 & 42:18:32.4 & 8 & 155.1 & 0.46 & 4.6 & 0.496 & & d & & & + ngc891 & x6 & 02:22:32.47 & 42:26:24.2 & 8 & 336.3 & 1.91 & 9.8 & 1.27 & & & p b2/21.32 & 2ulxd & + ngc891 & x7 & 02:22:45.55 & 42:25:57.8 & 8 & 339.1 & 0.85 & 11.9 & 2.28 & & d & & 1ulxd & ulx3 + ngc891 & x8 & 02:22:07.97 & 42:11:45.4 & 8 & 610.1 & 1.67 & 3.8 & 0.354 & & & p b/17.00 & & + ngc1042 & x1 & 02:40:25.51 & -08:24:30.5 & 2 & 94.3 & 0.68 & 17.1 & 4.27 & & s & & 1ulxd & ulx1 + ngc1042 & x2 & 02:40:21.97 & -08:21:40.0 & 2 & 263.7 & 1.89 & 4.3 & 0.789 & & & p b1/20.68 & & + ngc1052 & x1 & 02:41:04.98 & -08:15:20.2 & 2 & 2.2 & 0.04 & 21.4 & 32.8 & & & ( ngc1052 ) & n & + ngc1068 & x1 & 02:42:40.88 & -00:00:45.3 & 3 & 10.6 & 0.05 & 635.3 & 1330 & & & ( ngc1068 ) & n & + ngc1068 & x2 & 02:42:37.65 & -00:01:21.0 & 2 & 50.6 & 0.24 & 8.6 & 4.26 & & & & 1ulxd & ulx1 + ngc1068 & x3 & 02:42:42.64 & -00:03:28.2 & 3 & 164.4 & 0.91 & 4.5 & 2.59 & 1 & r & & 1ulxd & ulx2 + ngc1068 & x4 & 02:42:44.56 & -00:06:35.0 & 3 & 353.1 & 1.95 & 44.1 & 25.9 & 1 & & * v/12.505 & 2eulxd & + ngc1068 & x5 & 02:43:00.59 & -00:04:03.4 & 2 & 362.9 & 1.90 & 4.7 & 1.46 & & & & 2ulxd & + ngc1068 & x7 & 02:43:06.59 & 00:01:31.0 & 2 & 419.6 & 1.98 & 4.4 & 1.27 & & & & 2ulxd & + ngc1073 & x1 & 02:43:39.69 & 01:21:06.9 & 2 & 89.8 & 0.61 & 5.8 & 3.16 & & & qso 0241 + 011 & 1ulxd & + ngc1073 & x2 & 02:43:38.48 & 01:24:10.2 & 2 & 98.5 & 0.69 & 4.0 & 3.63 & & s & & 1ulxd & ulx1 + ngc1073 & x3 & 02:43:33.53 & 01:22:18.2 & 2 & 104.5 & 0.77 & 3.3 & 1.61 & & & qso n1073u1 & 1ulxd & + ngc1073 & x4 & 02:43:33.49 & 01:21:33.9 & 2 & 120.8 & 0.86 & 13.9 & 8.4 & & & qso n1073u2 & 1ulxd & + ngc1073 & x6 & 02:43:46.70 & 01:25:10.5 & 2 & 181.1 & 1.24 & 4.2 & 2.91 & & & p b/17.30 & 2ulxd & + ngc1097 & x1 & 02:46:18.96 & -30:16:30.2 & 2 & 9.2 & 0.04 & 50.3 & 134 & 1 & & ( ngc1097 ) & n & + ngc1291 & x1 & 03:17:17.99 & -41:06:22.4 & 0 & 7.1 & 0.03 & 13.5 & 7 & & & ( ngc1291 ) & n & + ngc1291 & x2 & 03:17:13.54 & -41:10:32.5 & 0 & 248.8 & 0.85 & 10.9 & 3.03 & & r & & 1ulxd & ulx1 + ngc1291 & x3 & 03:17:45.95 & -41:01:26.5 & 0 & 440.2 & 1.67 & 5.1 & 1.62 & & & & 2ulxd & + ngc1313 & x1 & 03:18:18.46 & -66:30:00.2 & 1 & 20.0 & 0.09 & 8.0 & 0.417 & & d & & & + ngc1313 & x2 & 03:18:19.80 & -66:29:09.6 & 1 & 48.8 & 0.20 & 139.3 & 10.8 & 7 & d & & 1eulxd & ulx1 + ngc1313 & x3 & 03:18:20.59 & -66:30:34.1 & 0 & 52.7 & 0.21 & 16.6 & 0.789 & & & & & + ngc1313 & x4 & 03:18:04.65 & -66:30:09.3 & 1 & 67.4 & 0.32 & 4.6 & 0.177 & & & & & + ngc1313 & x5 & 03:18:18.64 & -66:32:28.9 & 1 & 159.1 & 0.58 & 4.1 & 0.126 & & & * b2/13.69 & & + ngc1313 & x6 & 03:17:38.23 & -66:33:03.4 & 1 & 294.3 & 1.28 & 59.0 & 3.12 & & & ( sn1978k ) & 2ulxd & ulx2 + ngc1313 & x7 & 03:18:22.00 & -66:36:04.3 & 1 & 375.2 & 1.38 & 54.1 & 3.98 & 3 & d & ( p r/21.56 ) & 2ulxd & ulx3 + ngc1313 & x8 & 03:17:16.21 & -66:26:37.6 & 1 & 404.3 & 1.85 & 4.1 & 0.168 & & & & & + ngc1316 & x1 & 03:22:41.71 & -37:12:30.4 & 2 & 2.8 & 0.01 & 44.2 & 110 & & & ( ngc1316 ) & n & + ngc1316 & x3 & 03:22:51.40 & -37:09:47.1 & 2 & 199.0 & 0.57 & 5.4 & 4.45 & & & & 1ulxd & ulx1 + ngc1316 & x4 & 03:22:40.44 & -37:16:41.8 & 2 & 254.2 & 0.87 & 7.4 & 9.69 & & & p b2/20.24 & 1ulxd & ulx2 + ngc1316 & x5 & 03:22:57.45 & -37:15:56.0 & 2 & 281.3 & 1.10 & 4.7 & 4.53 & & d & & 2ulxd & ulx3 + ngc1316 & x6 & 03:22:15.49 & -37:16:59.5 & 2 & 413.5 & 1.15 & 3.5 & 3.05 & & d & p b2/20.25 & 2ulxd & ulx4 + ngc1316 & x7 & 03:22:39.13 & -37:20:04.5 & 2 & 457.4 & 1.57 & 3.5 & 2.76 & & d & & 2ulxd & ulx5 + ngc1316 & x8 & 03:23:14.33 & -37:16:54.5 & 2 & 473.0 & 1.82 & 4.3 & 3.55 & & d & p b2/21.56 & 2ulxd & ulx6 + ngc1316 & x9 & 03:23:28.86 & -37:10:10.9 & 2 & 581.2 & 1.76 & 4.6 & 4.11 & & d & mp & 2ulxd & ulx7 + ngc1317 & x1 & 03:22:44.41 & -37:06:12.6 & 2 & 4.3 & 0.05 & 9.7 & 7.99 & & & & n & + ngc1332 & x1 & 03:26:17.24 & -21:20:07.2 & 1 & 2.0 & 0.02 & 17.5 & 59.6 & & & ( ngc1332 ) & n & + ngc1365 & x1 & 03:33:36.56 & -36:08:22.7 & 2 & 5.6 & 0.02 & 30.0 & 67.4 & & & ( ngc1365 ) & n & + ngc1365 & x2 & 03:33:34.60 & -36:09:35.0 & 1 & 81.6 & 0.26 & 5.9 & 7.29 & & d & & 1ulx & ulx1 + ngc1365 & x3 & 03:33:12.85 & -36:11:50.4 & 2 & 358.1 & 1.22 & 6.2 & 5.98 & & & p b2/18.73 & 2ulxd & ulx2 + ngc1365 & x4 & 03:33:11.63 & -36:11:32.7 & 1 & 360.2 & 1.27 & 5.7 & 6.31 & & & & 2ulx & ulx3 + ngc1365 & x5 & 03:34:07.88 & -36:03:57.2 & 2 & 459.6 & 1.60 & 9.1 & 9.32 & & & & 2ulxd & + ngc1386 & x1 & 03:36:46.20 & -35:59:57.3 & 1 & 2.5 & 0.06 & 14.7 & 27.1 & & & ( ngc1386 ) & n & + ngc1380 & x1 & 03:36:27.23 & -34:58:32.7 & 4 & 4.1 & 0.06 & 20.0 & 19.3 & & & ( ngc1380 ) & n & + ngc1380 & x2 & 03:36:25.21 & -34:59:19.8 & 4 & 51.2 & 0.40 & 3.0 & 1.43 & & d & & 1ulxd & ulx1 + ngc1395 & x1 & 03:38:29.66 & -23:01:37.4 & 2 & 2.7 & 0.02 & 11.7 & 31.4 & & & ( ngc1395 ) & n & + ngc1395 & x2 & 03:38:17.64 & -23:02:30.5 & 2 & 172.6 & 1.27 & 3.9 & 4.8 & & & p b1/20.04 & 2ulxd & + ngc1395 & x3 & 03:38:31.19 & -23:05:03.3 & 2 & 204.5 & 1.16 & 3.5 & 4.51 & & & & 2ulxd & + ngc1395 & x4 & 03:38:11.29 & -23:00:38.3 & 2 & 260.2 & 1.92 & 4.1 & 6.13 & & & z & 2ulxd & + ngc1395 & x5 & 03:38:29.55 & -23:07:04.8 & 2 & 324.8 & 1.84 & 4.2 & 6.45 & & & z & 2ulx & + ngc1398 & x1 & 03:38:51.71 & -26:20:09.3 & 2 & 2.2 & 0.01 & 9.2 & 15.4 & & & ( ngc1398 ) & n & + ngc1398 & x2 & 03:39:09.37 & -26:20:12.9 & 2 & 236.2 & 1.12 & 4.6 & 5.51 & & & & 2ulxd & + ngc1398 & x3 & 03:39:11.94 & -26:19:57.5 & 2 & 271.0 & 1.30 & 4.9 & 5.91 & & & & 2ulxd & + ngc1399 & x1 & 03:38:28.97 & -35:27:01.1 & 4 & 3.2 & 0.02 & 109.7 & 257 & & & ( ngc1399 ) & n & + ngc1399 & x2 & 03:38:32.96 & -35:27:05.2 & 3 & 48.9 & 0.25 & 10.6 & 7.69 & & & ( gc ) & 1ulxd & ulx1 + ngc1399 & x3 & 03:38:31.99 & -35:26:02.4 & 4 & 66.4 & 0.33 & 8.5 & 5.87 & & & ( gc ) & 1ulxd & ulx2 + ngc1399 & x4 & 03:38:25.53 & -35:25:19.6 & 1 & 107.1 & 0.52 & 3.9 & 2.35 & & & mz & 1ulx & ulx3 + ngc1399 & x5 & 03:38:31.86 & -35:30:58.1 & 4 & 242.6 & 1.17 & 4.7 & 1.96 & & & & 2ulxd & + ngc1399 & x6 & 03:38:25.53 & -35:22:40.9 & 1 & 260.6 & 1.26 & 3.1 & 1.32 & & & & 2ulx & + ngc1399 & x7 & 03:38:48.99 & -35:28:36.4 & 3 & 263.3 & 1.35 & 3.3 & 1.51 & & & & 2ulxd & + ngc1399 & x8 & 03:38:51.61 & -35:26:41.9 & 4 & 276.8 & 1.43 & 14.1 & 9.03 & 1 & & z & 2ulxd & + ngc1399 & x9 & 03:38:41.42 & -35:31:34.7 & 4 & 315.6 & 1.55 & 44.3 & 42.3 & 1 & & p b2/20.11 & 2eulxd & + ngc1399 & x10 & 03:38:31.71 & -35:33:43.3 & 1 & 406.7 & 1.96 & 3.4 & 1.5 & & & & 2ulx & + ngc1404 & x1 & 03:38:52.07 & -35:35:39.8 & 4 & 5.9 & 0.06 & 108.0 & 288 & & & ( ngc1404 ) & n & + ngc1404 & x2 & 03:38:43.66 & -35:33:48.3 & 4 & 145.7 & 1.55 & 5.7 & 3.65 & & & p b2/18.72 & 2ulxd & + ngc1427a & x1 & 03:40:12.20 & -35:37:34.7 & 4 & 36.8 & 0.69 & 8.6 & 8.81 & & d & & 1ulx & ulx1 + ngc1433 & x1 & 03:42:01.49 & -47:13:18.9 & 1 & 2.7 & 0.01 & 5.0 & 2.33 & & & ( ngc1433 ) & n & + pgc13826 & x1 & 03:46:48.82 & 68:05:49.5 & 3 & 6.7 & 0.01 & 17.9 & 2.57 & & & ( pgc13826 ) & n & + pgc13826 & x2 & 03:46:57.17 & 68:06:22.4 & 3 & 56.1 & 0.09 & 10.6 & 1.21 & & s & & 1ulxd & ulx1 + pgc13826 & x3 & 03:46:45.54 & 68:09:51.7 & 3 & 247.7 & 0.39 & 13.7 & 1.49 & 1 & s & & 1ulxd & ulx2 + pgc13826 & x4 & 03:46:06.31 & 68:07:07.2 & 3 & 256.3 & 0.41 & 6.2 & 0.641 & & & & & + pgc13826 & x6 & 03:45:55.17 & 68:04:58.6 & 3 & 308.7 & 0.49 & 21.5 & 2.77 & & s & ( hii ) & 1ulxd & ulx3 + pgc13826 & x7 & 03:46:15.70 & 68:11:19.7 & 3 & 384.8 & 0.60 & 7.2 & 0.723 & & & & & + pgc13826 & x8 & 03:45:41.40 & 68:02:42.2 & 3 & 424.1 & 0.67 & 3.0 & 0.252 & & & & & + pgc13826 & x9 & 03:48:06.58 & 68:04:55.4 & 3 & 433.1 & 0.69 & 4.5 & 0.467 & & & & & + pgc13826 & x10 & 03:47:17.97 & 67:51:11.7 & 3 & 887.7 & 1.39 & 9.7 & 2.23 & & & * bt/12.152 & 2ulxd & + ngc1482 & x1 & 03:54:38.95 & -20:30:14.6 & 1 & 8.5 & 0.18 & 9.3 & 32.3 & & & ( ngc1482 ) & n & + ngc1532 & x1 & 04:12:05.86 & -32:52:18.4 & 2 & 10.6 & 0.03 & 3.9 & 1.88 & & & ( ngc1532 ) & n & + ngc1549 & x1 & 04:15:45.87 & -55:35:29.6 & 2 & 6.7 & 0.05 & 13.2 & 13 & & & ( ngc1549 ) & n & + ngc1549 & x2 & 04:15:16.86 & -55:34:55.7 & 2 & 242.0 & 1.77 & 4.3 & 3.01 & & & * j/12.857 & 2ulx & + ngc1553 & x1 & 04:16:11.07 & -55:46:49.2 & 2 & 5.9 & 0.07 & 20.1 & 20.7 & & & ( ngc1553 ) & n & + ngc1553 & x2 & 04:16:03.51 & -55:46:54.1 & 2 & 58.2 & 0.64 & 5.1 & 2.6 & & & & 1ulxd & ulx1 + ngc1553 & x3 & 04:16:08.19 & -55:48:16.4 & 2 & 87.5 & 0.85 & 3.0 & 1.39 & & & & 1ulx & ulx2 + ngc1559 & x1 & 04:17:42.27 & -62:46:40.7 & 2 & 40.7 & 0.40 & 5.8 & 3.73 & & d & & 1ulxd & ulx1 + ngc1559 & x2 & 04:17:30.39 & -62:47:24.7 & 2 & 52.3 & 0.51 & 5.8 & 2.97 & & d & & 1ulxd & ulx2 + ngc1559 & x3 & 04:17:28.00 & -62:48:20.1 & 2 & 99.7 & 1.10 & 6.0 & 2.89 & & & z & 2ulxd & + ngc1566 & x1 & 04:20:00.41 & -54:56:16.7 & 1 & 1.3 & 0.01 & 454.1 & 371 & 1 & & ( ngc1566 ) & n & + ngc1566 & x2 & 04:19:56.11 & -54:56:38.3 & 1 & 42.2 & 0.17 & 12.6 & 4.96 & & s & & 1ulxd & ulx1 + ngc1566 & x3 & 04:19:56.73 & -54:55:23.0 & 1 & 63.5 & 0.32 & 7.4 & 3.43 & & s & & 1ulxd & ulx2 + ngc1566 & x4 & 04:20:10.12 & -54:56:42.1 & 1 & 87.1 & 0.40 & 9.6 & 3.6 & 1 & s & & 1ulxd & ulx3 + ngc1566 & x5 & 04:20:16.22 & -54:59:51.9 & 1 & 253.6 & 1.28 & 4.5 & 1.17 & & d & mz & 2ulxd & + ngc1574 & x1 & 04:21:58.69 & -56:58:29.4 & 2 & 4.7 & 0.05 & 16.1 & 25.8 & & & ( ngc1547 ) & n & + ngc1672 & x1 & 04:45:41.98 & -59:14:53.3 & 1 & 4.1 & 0.02 & 25.8 & 18.6 & & & ( ngc1672 ) & n & + ngc1672 & x2 & 04:45:34.11 & -59:14:44.2 & 1 & 63.4 & 0.38 & 7.5 & 3.76 & & s & & 1ulxd & ulx1 + ngc1672 & x3 & 04:45:52.87 & -59:14:59.1 & 1 & 81.8 & 0.49 & 8.2 & 3.53 & & s & & 1ulxd & ulx2 + ngc1672 & x4 & 04:46:22.77 & -59:15:25.5 & 1 & 312.4 & 1.88 & 3.8 & 1.56 & & & p b2/20.70 & 2ulx & + ngc1672 & x5 & 04:45:28.64 & -59:09:35.3 & 1 & 338.1 & 1.71 & 4.7 & 2 & & & p b2/21.97 & 2ulxd & + ngc1705 & x1 & 04:54:19.32 & -53:20:44.8 & 1 & 78.3 & 1.39 & 3.9 & 0.441 & & & & & + ngc1792 & x1 & 05:05:15.83 & -37:58:29.7 & 2 & 19.3 & 0.23 & 5.0 & 3.04 & & s & & 1ulx & ulx1 + ngc1792 & x2 & 05:05:11.01 & -37:58:44.6 & 1 & 48.4 & 0.48 & 6.4 & 1.53 & & s & & 1ulxd & ulx2 + ngc2146 & x1 & 06:18:40.48 & 78:21:32.1 & 1 & 13.1 & 0.11 & 12.7 & 46.2 & & & ( ngc2146 ) & n & + ngc2273 & x1 & 06:50:08.57 & 60:50:44.9 & 1 & 0.2 & 0.00 & 8.2 & 17.6 & & & ( ngc2273 ) & n & + ngc2273 & x2 & 06:50:12.76 & 60:48:36.7 & 1 & 131.9 & 1.71 & 10.4 & 23.6 & & & z & 2eulxd & + ngc2273 & x3 & 06:50:22.68 & 60:53:00.4 & 1 & 170.0 & 1.78 & 5.9 & 10.9 & & & z & 2eulxd & + ngc2276 & x1 & 07:26:49.45 & 85:45:48.3 & 1 & 38.2 & 0.47 & 7.9 & 33.8 & & s & & 1eulxd & ulx1 + ngc2276 & x2 & 07:25:22.75 & 85:44:18.2 & 1 & 137.0 & 1.65 & 9.7 & 27.5 & & & f8 * & 2eulxd & + ngc2300 & x1 & 07:32:22.84 & 85:42:26.3 & 1 & 6.8 & 0.09 & 28.8 & 144 & & & ( ngc2300 ) & n & + ngc2366 & x1 & 07:28:51.67 & 69:07:26.0 & 0 & 326.3 & 1.77 & 6.6 & 0.131 & & & & & + ngc2366 & x2 & 07:29:30.58 & 69:19:38.8 & 0 & 449.9 & 1.85 & 5.6 & 0.111 & & & p b2/19.68 & & + ngc2403 & x1 & 07:36:56.05 & 65:35:38.6 & 1 & 21.7 & 0.04 & 10.7 & 0.38 & & d & & & + ngc2403 & x2 & 07:36:25.95 & 65:35:38.9 & 1 & 178.0 & 0.38 & 32.5 & 1.4 & & s & & 1ulxd & ulx1 + ngc2403 & x3 & 07:37:02.83 & 65:39:35.5 & 1 & 223.6 & 0.57 & 10.0 & 0.326 & & s & & & + ngc2403 & x5 & 07:35:09.17 & 65:40:25.3 & 1 & 704.5 & 1.14 & 44.3 & 2.29 & & & g0 * & 2ulxd & + ngc2403 & x6 & 07:35:38.68 & 65:52:09.4 & 1 & 1078.0 & 1.98 & 7.8 & 0.585 & 1 & & & & + ngc2442 & x1 & 07:36:23.51 & -69:31:51.7 & 1 & 2.6 & 0.02 & 6.1 & 5.93 & & & ( ngc2442 ) & n & + ngc2442 & x2 & 07:37:05.67 & -69:32:06.5 & 1 & 219.7 & 1.49 & 4.8 & 3.72 & & & & 2ulxd & + pgc23324 & x1 & 08:19:29.90 & 70:42:18.6 & 0 & 122.8 & 0.65 & 123.5 & 28.4 & 1 & d & ( hii ) & 1eulxd & ulx1 + ngc2775 & x1 & 09:10:20.63 & 07:02:17.3 & 3 & 2.6 & 0.02 & 3.5 & 1.97 & & & ( ngc2775 ) & n & + ngc2775 & x3 & 09:10:14.17 & 07:02:49.7 & 3 & 99.1 & 0.90 & 3.7 & 1.83 & & & p b2/21.58 & 1ulxd & ulx1 + ngc2775 & x4 & 09:10:32.29 & 07:03:44.5 & 3 & 195.2 & 1.97 & 5.9 & 2.97 & & & * b1/14.07 & 2ulxd & + ngc2775 & x5 & 09:10:27.08 & 06:59:09.5 & 3 & 213.3 & 1.67 & 7.9 & 5.65 & & & p b2/21.28 & 2ulxd & + ngc2775 & x6 & 09:10:19.90 & 07:06:00.4 & 3 & 221.6 & 1.82 & 4.9 & 2.42 & & & & 2ulxd & + ngc2782 & x1 & 09:14:05.01 & 40:06:51.4 & 1 & 5.6 & 0.07 & 17.0 & 98.5 & & & ( ngc2782 ) & n & + ngc2782 & x2 & 09:14:03.93 & 40:07:41.6 & 1 & 52.8 & 0.53 & 5.8 & 19.4 & & d & & 1eulxd & ulx1 + ngc2903 & x1 & 09:32:09.95 & 21:30:05.4 & 0 & 4.8 & 0.02 & 15.8 & 5.26 & & & ( ngc2903 ) & n & + ngc2903 & x2 & 09:32:02.04 & 21:31:19.2 & 0 & 131.8 & 0.70 & 4.7 & 1.5 & & s & & 1ulx & ulx1 + ngc2974 & x1 & 09:42:33.27 & -03:41:56.7 & 2 & 2.3 & 0.03 & 21.1 & 31.7 & & & ( ngc2974 ) & n & + ngc2992 & x1 & 09:45:41.95 & -14:19:34.8 & 1 & 4.7 & 0.08 & 87.2 & 429 & & & ( ngc2992 ) & n & + ngc2993 & x1 & 09:45:48.27 & -14:22:05.1 & 1 & 3.5 & 0.10 & 14.4 & 51.1 & & & ( ngc2993 ) & n & + ngc3031 & x1 & 09:55:33.38 & 69:03:54.3 & 4 & 5.8 & 0.01 & 411.5 & 74.4 & 8 & & ( ngc3031 ) & n & + ngc3031 & x2 & 09:55:42.15 & 69:03:35.1 & 4 & 52.7 & 0.09 & 8.5 & 0.472 & & & & & + ngc3031 & x3 & 09:55:36.29 & 69:02:38.5 & 1 & 82.9 & 0.11 & 5.0 & 0.209 & & & & & + ngc3031 & x4 & 09:55:22.13 & 69:05:11.2 & 4 & 93.7 & 0.13 & 4.9 & 0.209 & & & & & + ngc3031 & x5 & 09:55:50.05 & 69:05:32.6 & 4 & 128.2 & 0.29 & 8.1 & 0.445 & & & & & + ngc3031 & x6 & 09:55:09.59 & 69:04:08.8 & 3 & 128.4 & 0.28 & 13.1 & 0.504 & & & & & + ngc3031 & x7 & 09:55:10.43 & 69:05:01.7 & 4 & 138.1 & 0.25 & 8.0 & 0.31 & & & & & + ngc3031 & x8 & 09:55:24.74 & 69:01:11.8 & 4 & 174.7 & 0.31 & 44.2 & 2.61 & & & ( sn1993j ) & 1ulx & ulx1 + ngc3031 & x9 & 09:55:33.08 & 69:00:31.4 & 4 & 208.6 & 0.31 & 64.7 & 3.37 & & s & ( o8 * ) & 1ulx & ulx2 + ngc3031 & x10 & 09:54:51.25 & 69:02:51.3 & 2 & 236.7 & 0.56 & 9.0 & 0.427 & & & & & + ngc3031 & x11 & 09:56:09.25 & 69:01:03.7 & 4 & 260.5 & 0.39 & 6.5 & 0.227 & & & & & + ngc3031 & x12 & 09:55:00.06 & 69:07:45.5 & 4 & 287.8 & 0.39 & 5.9 & 0.226 & & & & & + ngc3031 & x13 & 09:55:12.84 & 69:08:27.7 & 2 & 289.6 & 0.36 & 4.8 & 0.18 & & & & & + ngc3031 & x14 & 09:56:26.67 & 69:05:11.1 & 4 & 293.5 & 0.69 & 3.2 & 0.114 & & & & & + ngc3031 & x15 & 09:55:49.64 & 68:58:33.0 & 4 & 338.3 & 0.43 & 4.6 & 0.288 & & s & & & + ngc3031 & x17 & 09:55:24.33 & 69:09:59.9 & 4 & 363.2 & 0.49 & 17.8 & 0.682 & & & & & + ngc3031 & x18 & 09:56:36.81 & 69:00:27.8 & 4 & 400.6 & 0.68 & 4.6 & 0.154 & & & & & + ngc3031 & x19 & 09:56:14.29 & 68:57:20.8 & 4 & 455.5 & 0.57 & 9.4 & 0.348 & 1 & & & & + ngc3031 & x20 & 09:55:01.81 & 68:56:18.9 & 4 & 491.6 & 0.91 & 6.1 & 0.229 & & & & & + ngc3031 & x21 & 09:54:45.40 & 68:56:55.8 & 4 & 496.8 & 1.02 & 5.8 & 0.208 & & & & & + ngc3031 & x22 & 09:53:56.71 & 69:03:56.0 & 2 & 518.8 & 1.16 & 3.5 & 0.0946 & & & & & + ngc3031 & x23 & 09:56:09.60 & 69:12:49.0 & 2 & 563.0 & 1.04 & 3.2 & 0.115 & & & & & + ngc3031 & x24 & 09:57:01.74 & 68:54:57.1 & 4 & 721.1 & 1.00 & 7.3 & 0.314 & & & & & + ngc3031 & x25 & 09:58:02.47 & 68:57:08.6 & 4 & 900.0 & 1.61 & 7.6 & 0.444 & & & & & + ngc3034 & x1 & 09:55:52.79 & 69:40:37.0 & 2 & 21.1 & 0.13 & 74.7 & 61.7 & & & ( ngc3034 ) & n & + ngc3034 & x2 & 09:56:00.98 & 69:41:06.5 & 2 & 37.6 & 0.12 & 4.5 & 0.516 & & & & & + ngc3034 & x3 & 09:55:52.07 & 69:42:02.4 & 2 & 66.2 & 0.50 & 6.5 & 0.957 & & & & & + ngc3034 & x5 & 09:55:26.84 & 69:39:24.5 & 2 & 169.1 & 0.53 & 3.0 & 0.155 & & & & & + ngc3034 & x7 & 09:55:14.99 & 69:36:09.5 & 2 & 352.3 & 1.64 & 3.7 & 0.164 & & & & & + ngc3034 & x8 & 09:57:32.60 & 69:46:01.2 & 2 & 596.0 & 1.83 & 4.4 & 0.214 & & & & & + pgc28757 & x1 & 09:57:36.72 & 69:02:23.1 & 1 & 33.5 & 0.51 & 3.0 & 0.0804 & & & & & + pgc28757 & x2 & 09:57:54.05 & 69:03:47.3 & 4 & 753.4 & 1.67 & 131.8 & 13.4 & 1 & d & & 2eulx & ulx1 + ngc3079 & x1 & 10:01:59.27 & 55:40:49.0 & 2 & 10.9 & 0.24 & 17.2 & 19.4 & & & ( ngc3079 ) & n & + ngc3115 & x1 & 10:05:14.51 & -07:43:08.7 & 0 & 6.3 & 0.08 & 3.8 & 4.88 & & & ( ngc3115 ) & n & + ngc3190 & x1 & 10:18:05.35 & 21:50:00.5 & 3 & 5.1 & 0.05 & 7.7 & 8.79 & & & ( ngc3190 ) & n & + ngc3226 & x1 & 10:23:26.94 & 19:53:56.5 & 1 & 6.7 & 0.08 & 11.5 & 18 & & & ( ngc3226 ) & n & + ngc3226 & x3 & 10:23:25.74 & 19:56:17.6 & 1 & 144.5 & 1.56 & 5.1 & 5.46 & & & p b1/18.84 & 2ulxd & + ngc3227 & x1 & 10:23:30.51 & 19:51:54.0 & 1 & 15.2 & 0.12 & 166.2 & 489 & 2 & & ( ngc3227 ) & n & + pgc30819 & x1 & 10:28:42.94 & 68:28:17.1 & 3 & 227.9 & 0.73 & 8.0 & 0.385 & & & & & + pgc30819 & x2 & 10:29:24.60 & 68:25:50.8 & 3 & 346.4 & 1.35 & 3.4 & 0.136 & & & & & + pgc30819 & x3 & 10:29:32.44 & 68:30:13.7 & 2 & 497.4 & 1.26 & 4.5 & 0.166 & & & & & + pgc30819 & x4 & 10:27:21.97 & 68:18:43.2 & 3 & 503.3 & 1.34 & 5.6 & 0.282 & & & & & + ngc3294 & x1 & 10:36:12.26 & 37:18:48.2 & 2 & 60.7 & 1.08 & 3.5 & 9.41 & & & & 2ulxd & + pgc31586 & x1 & 10:37:47.38 & -27:05:09.0 & 3 & 14.6 & 0.19 & 4.9 & 12.3 & & & ( pgc31586 ) & n & + pgc31586 & x2 & 10:37:39.69 & -27:05:24.6 & 3 & 111.0 & 1.98 & 26.9 & 84.3 & 1 & & pr & 2ulxd & + ngc3310 & x1 & 10:38:45.71 & 53:30:10.6 & 2 & 4.3 & 0.06 & 57.3 & 83.1 & & & ( ngc3310 ) & n & + ngc3310 & x2 & 10:38:46.25 & 53:30:43.2 & 1 & 35.2 & 0.38 & 10.2 & 5.41 & & s & & 1ulxd & ulx1 + ngc3310 & x3 & 10:38:42.86 & 53:31:06.6 & 1 & 65.4 & 0.75 & 8.9 & 6.04 & & s & & 1ulxd & ulx2 + ngc3377 & x1 & 10:47:42.58 & 13:59:10.5 & 1 & 16.6 & 0.11 & 4.7 & 1.5 & & & ( ngc3377 ) & n & + ngc3379 & x1 & 10:47:49.80 & 12:34:58.3 & 1 & 1.9 & 0.01 & 14.1 & 6.35 & 1 & & ( ngc3379 ) & n & + ngc3396 & x2 & 10:49:47.96 & 33:00:30.2 & 2 & 123.1 & 1.85 & 10.2 & 15.9 & & & ? * b2/17.33 & 2eulxd & + pgc32390 & x1 & 10:49:24.90 & 32:46:25.9 & 2 & 10.6 & 0.32 & 4.1 & 6.79 & & & ( pgc32390 ) & n & + ngc3516 & x1 & 11:06:47.47 & 72:34:07.1 & 1 & 4.9 & 0.09 & 395.9 & 5140 & 1 & & ( ngc3516 ) & n & + ngc3623 & x1 & 11:18:55.79 & 13:05:38.2 & 3 & 7.9 & 0.09 & 8.6 & 3.03 & & & ( ngc3623 ) & n & + ngc3623 & x2 & 11:18:58.41 & 13:05:33.6 & 3 & 45.5 & 0.52 & 6.3 & 1.75 & & s & & 1ulxd & ulx1 + ngc3623 & x3 & 11:18:52.39 & 13:03:35.2 & 3 & 127.1 & 0.75 & 4.5 & 1.07 & & & * b/12.163 & 1ulxd & + ngc3627 & x1 & 11:20:14.44 & 12:59:31.3 & 1 & 8.5 & 0.06 & 5.9 & 2.21 & & & ( ngc3627 ) & n & + ngc3627 & x2 & 11:20:20.88 & 12:58:48.2 & 1 & 95.1 & 0.67 & 6.3 & 2.06 & & s & p & 1ulxd & ulx1 + ngc3627 & x3 & 11:20:22.04 & 12:52:49.4 & 1 & 412.6 & 1.56 & 6.9 & 2.55 & & & mz & 2ulxd & + ngc3628 & x1 & 11:20:15.43 & 13:35:19.8 & 0 & 12.8 & 0.06 & 28.3 & 11.9 & & d & ( ngc3628 ) & n & + ngc3628 & x2 & 11:20:37.17 & 13:34:32.0 & 1 & 308.3 & 0.75 & 7.9 & 2.9 & & d & & 1ulxd & ulx1 + ngc3628 & x3 & 11:20:39.76 & 13:36:26.6 & 0 & 348.1 & 1.79 & 3.9 & 1.22 & & & qso j1120.6 + 1336 & 2ulxd & + ngc3628 & x4 & 11:20:41.49 & 13:35:57.9 & 0 & 369.1 & 1.60 & 3.4 & 0.986 & & & qso wee52 & & + ngc3642 & x1 & 11:22:17.88 & 59:04:28.3 & 1 & 7.0 & 0.05 & 5.1 & 11.4 & & & ( ngc3642 ) & n & + ngc3642 & x2 & 11:22:40.36 & 59:02:37.1 & 1 & 205.7 & 1.31 & 3.6 & 7.9 & & & p b1/19.60 & 2ulxd & + ngc3665 & x1 & 11:24:43.38 & 38:45:42.9 & 1 & 4.3 & 0.06 & 9.1 & 33.5 & & & ( ngc3665 ) & n & + pgc35286 & x1 & 11:28:03.12 & 78:59:56.5 & 0 & 57.2 & 1.36 & 37.3 & 0.386 & & & & & + ngc3738 & x1 & 11:35:45.42 & 54:33:15.4 & 1 & 117.1 & 1.57 & 7.4 & 0.371 & & & z & & + ngc3982 & x1 & 11:56:28.21 & 55:07:29.8 & 1 & 1.2 & 0.02 & 4.5 & 5.13 & & & ( ngc3982 ) & n & + ngc3998 & x1 & 11:57:55.92 & 55:27:13.5 & 0 & 6.0 & 0.09 & 186.9 & 259 & & & ( ngc3998 ) & n & + ngc3998 & x2 & 11:57:49.69 & 55:28:30.2 & 0 & 95.4 & 1.18 & 4.3 & 2.88 & & & & 2ulxd & + ngc4036 & x1 & 12:01:26.89 & 61:53:38.1 & 3 & 4.1 & 0.06 & 4.7 & 15.7 & & & ( ngc4036 ) & n & + ngc4041 & x1 & 12:02:12.13 & 62:08:15.3 & 3 & 5.7 & 0.07 & 13.0 & 21.5 & & & ( ngc4041 ) & n & + ngc4051 & x1 & 12:03:09.46 & 44:31:54.3 & 0 & 2.0 & 0.01 & 626.1 & 2770 & 2 & & ( ngc4051 ) & n & + ngc4088 & x1 & 12:05:32.02 & 50:32:48.8 & 2 & 36.4 & 0.52 & 4.9 & 5.86 & & s & pr & 1ulxd & ulx1 + ngc4105 & x1 & 12:06:40.29 & -29:45:40.8 & 2 & 5.5 & 0.08 & 9.4 & 13.4 & & & ( ngc4105 ) & n & + ngc4105 & x2 & 12:06:40.97 & -29:48:03.0 & 2 & 141.1 & 1.89 & 6.9 & 6.34 & & & * v/11.03 & 2ulxd & + ngc4106 & x1 & 12:06:45.67 & -29:46:04.7 & 2 & 12.7 & 0.26 & 5.5 & 6.63 & & & ( ngc4106 ) & n & + ngc4136 & x1 & 12:09:22.50 & 29:55:53.8 & 0 & 59.6 & 0.53 & 3.5 & 3.96 & & s & & 1ulxd & ulx1 + ngc4144 & x1 & 12:10:09.94 & 46:27:10.6 & 1 & 110.0 & 0.67 & 4.2 & 0.507 & & d & & & + ngc4150 & x1 & 12:10:34.69 & 30:23:57.9 & 1 & 17.4 & 0.28 & 41.0 & 40.3 & & & * j/16.049 & & + ngc4151 & x1 & 12:10:32.57 & 39:24:20.9 & 4 & 8.6 & 0.05 & 720.4 & 1330 & 1 & & ( ngc4151 ) & n & + ngc4151 & x2 & 12:10:22.45 & 39:23:14.6 & 4 & 142.7 & 0.76 & 4.8 & 1.84 & & & p b1/19.70 & 1ulxd & ulx1 + ngc4151 & x3 & 12:10:47.48 & 39:23:10.7 & 4 & 184.8 & 1.32 & 5.4 & 2.13 & & s & & 2ulxd & ulx2 + ngc4151 & x4 & 12:10:21.81 & 39:21:53.1 & 4 & 202.0 & 1.08 & 4.4 & 1.62 & & s & & 2ulxd & ulx3 + ngc4151 & x5 & 12:10:17.76 & 39:21:24.5 & 3 & 254.7 & 1.36 & 5.9 & 9.84 & & s & & 2ulx & ulx4 + ngc4151 & x6 & 12:10:07.90 & 39:23:10.6 & 4 & 301.1 & 1.73 & 9.0 & 3.86 & & & p b1/18.49 & 2ulxd & + ngc4156 & x1 & 12:10:49.63 & 39:28:23.5 & 4 & 0.9 & 0.03 & 13.3 & 4.32 & & & ( ngc4156 ) & n & + pgc38811 & x1 & 12:11:13.71 & 39:24:33.3 & 4 & 41.0 & 1.42 & 4.9 & 0.872 & & & & & + ngc4190 & x1 & 12:13:45.41 & 36:37:55.2 & 1 & 9.8 & 0.21 & 52.1 & 2.31 & & d & & 1ulxd & ulx1 + ngc4203 & x1 & 12:15:04.97 & 33:11:51.3 & 1 & 4.5 & 0.05 & 122.4 & 109 & & & ( ngc4203 ) & n & + ngc4203 & x2 & 12:15:09.23 & 33:09:55.4 & 1 & 127.5 & 1.28 & 98.7 & 78.7 & & & qso ton1480 & 2eulxd & + ngc4203 & x3 & 12:15:15.67 & 33:10:11.9 & 1 & 164.9 & 1.72 & 4.9 & 2.3 & & & p b2/20.09 & 2ulxd & + ngc4214 & x1 & 12:15:38.06 & 36:19:17.6 & 1 & 27.6 & 0.12 & 11.4 & 0.223 & & & & & + ngc4214 & x2 & 12:15:59.10 & 36:22:26.8 & 1 & 290.2 & 1.36 & 7.1 & 0.138 & & & p b2/19.76 & & + ngc4214 & x3 & 12:15:43.01 & 36:26:28.5 & 1 & 411.7 & 1.62 & 5.4 & 0.109 & & & z & & + ngc4235 & x1 & 12:17:10.06 & 07:11:30.2 & 1 & 17.3 & 0.44 & 198.6 & 1270 & 1 & & ( ngc4235 ) & n & + ngc4236 & x1 & 12:17:29.73 & 69:15:16.8 & 1 & 797.8 & 1.22 & 11.5 & 0.466 & 1 & & k5 * & & + ngc4242 & x1 & 12:17:46.79 & 45:36:57.8 & 2 & 175.6 & 1.51 & 5.7 & 3.03 & & & p & 2ulxd & + ngc4242 & x2 & 12:17:21.57 & 45:34:46.1 & 2 & 179.4 & 1.20 & 5.3 & 3.16 & & & p b1/18.68 & 2ulxd & + ngc4242 & x3 & 12:17:42.71 & 45:39:28.6 & 2 & 183.0 & 1.27 & 3.7 & 2.15 & & & p b1/18.67 & 2ulxd & + ngc4244 & x1 & 12:17:48.64 & 37:50:22.0 & 1 & 248.9 & 1.19 & 3.2 & 0.239 & & & & & + ngc4254 & x1 & 12:18:52.13 & 14:24:52.3 & 1 & 42.3 & 0.30 & 4.6 & 4.71 & & s & & 1ulxd & ulx1 + ngc4254 & x2 & 12:18:56.34 & 14:24:18.1 & 1 & 112.1 & 0.78 & 15.6 & 15.3 & & s & & 1eulxd & ulx2 + ngc4254 & x3 & 12:18:34.12 & 14:23:55.1 & 1 & 233.3 & 1.64 & 3.9 & 2.78 & & & & 2ulxd & + ngc4258 & x1 & 12:19:05.06 & 47:18:30.7 & 3 & 74.3 & 0.33 & 7.7 & 1.63 & 1 & s & & 1ulx & ulx1 + ngc4258 & x2 & 12:18:54.70 & 47:16:48.9 & 3 & 93.0 & 0.35 & 3.7 & 1.59 & & s & & 1ulx & ulx2 + ngc4258 & x3 & 12:18:57.67 & 47:15:58.1 & 3 & 137.9 & 0.39 & 5.7 & 0.701 & & s & & & + ngc4258 & x4 & 12:18:43.79 & 47:17:21.4 & 3 & 153.6 & 0.70 & 3.5 & 0.408 & & & & & + ngc4258 & x5 & 12:18:56.10 & 47:21:18.9 & 3 & 183.8 & 0.46 & 5.0 & 0.674 & & & & & + ngc4258 & x6 & 12:18:46.18 & 47:14:16.8 & 1 & 267.3 & 1.06 & 4.2 & 0.572 & & & & & + ngc4258 & x7 & 12:18:45.42 & 47:24:11.6 & 3 & 377.6 & 0.74 & 5.4 & 0.701 & & & & & + ngc4258 & x8 & 12:18:47.85 & 47:25:24.1 & 3 & 440.1 & 0.95 & 4.5 & 0.585 & & & & & + ngc4258 & x9 & 12:19:22.98 & 47:09:33.8 & 3 & 581.3 & 1.05 & 14.6 & 2.38 & & & mp & 2ulxd & + ngc4258 & x10 & 12:18:16.31 & 47:27:10.2 & 3 & 681.1 & 1.29 & 4.5 & 0.438 & & & & & + ngc4258 & x11 & 12:19:09.39 & 47:05:27.5 & 1 & 777.4 & 1.84 & 4.4 & 0.759 & & & qso 12166 + 4722 & & + ngc4258 & x12 & 12:18:28.04 & 47:30:18.1 & 3 & 783.2 & 1.46 & 4.7 & 0.537 & & & & & + ngc4261 & x1 & 12:19:22.96 & 05:49:30.0 & 1 & 6.5 & 0.05 & 37.1 & 229 & & & ( ngc4261 ) & n & + ngc4278 & x1 & 12:20:06.94 & 29:16:51.9 & 1 & 6.0 & 0.05 & 24.1 & 42.8 & & & ( ngc4278 ) & n & + ngc4291 & x1 & 12:20:20.41 & 75:22:09.6 & 3 & 14.2 & 0.27 & 11.1 & 57.3 & & & ( ngc4291 ) & n & + ngc4319 & x1 & 12:21:44.53 & 75:18:35.9 & 3 & 52.1 & 0.60 & 385.9 & 2210 & 1 & & qso mark205 & 1eulxd & + ngc4319 & x2 & 12:22:06.80 & 75:19:03.2 & 3 & 90.5 & 1.22 & 4.2 & 13.3 & & & p b1/20.89 & 2eulxd & + ngc4292 & x1 & 12:21:14.49 & 04:36:51.9 & 2 & 70.9 & 1.62 & 6.8 & 25.5 & & & g0 * & 2eulxd & + ngc4303 & x1 & 12:21:55.00 & 04:28:25.6 & 2 & 7.1 & 0.04 & 11.1 & 8.51 & & & ( ngc4303 ) & n & + ngc4303 & x2 & 12:21:58.41 & 04:28:14.3 & 1 & 55.7 & 0.32 & 3.6 & 2.46 & & s & & 1ulx & ulx1 + ngc4303 & x4 & 12:21:37.55 & 04:29:07.8 & 2 & 260.9 & 1.50 & 3.5 & 1.6 & & & & 2ulxd & + ngc4303 & x5 & 12:22:03.76 & 04:34:07.0 & 1 & 372.5 & 1.95 & 4.2 & 4.85 & & & & 2ulx & + ngc4321 & x1 & 12:22:54.60 & 15:49:19.7 & 3 & 9.2 & 0.04 & 25.7 & 17.2 & & & ( ngc4321 ) & n & + ngc4321 & x2 & 12:22:58.49 & 15:47:51.8 & 3 & 102.8 & 0.52 & 4.7 & 1.57 & & s & ( sn1979c ) & 1ulxd & ulx1 + ngc4321 & x3 & 12:22:48.01 & 15:51:56.9 & 3 & 185.6 & 0.95 & 4.0 & 1.44 & & s & & 1ulxd & ulx2 + ngc4321 & x4 & 12:23:08.47 & 15:51:22.9 & 1 & 225.9 & 1.06 & 3.2 & 1.6 & & & & 2ulx & + ngc4321 & x6 & 12:22:42.80 & 15:44:00.5 & 3 & 368.8 & 1.66 & 6.2 & 2.23 & & & p b1/20.31 & 2ulxd & + ngc4321 & x7 & 12:22:48.23 & 15:43:05.8 & 3 & 390.4 & 1.78 & 4.3 & 1.29 & & & z & 2ulxd & + ngc4374 & x1 & 12:25:04.07 & 12:53:16.0 & 2 & 5.4 & 0.03 & 48.3 & 137 & & & ( ngc4374 ) & n & + ngc4374 & x2 & 12:25:12.13 & 12:51:55.7 & 2 & 146.5 & 0.76 & 16.7 & 19.4 & & & qso & 1eulxd & + ngc4374 & x3 & 12:25:15.52 & 12:56:05.6 & 2 & 242.8 & 1.44 & 8.8 & 7.33 & & & & 2ulxd & + ngc4374 & x4 & 12:25:18.49 & 12:49:03.0 & 2 & 332.1 & 1.72 & 3.9 & 5.43 & & & p b2/20.23 & 2ulxd & + ngc4374 & x5 & 12:24:40.79 & 12:54:16.2 & 2 & 340.5 & 1.85 & 4.6 & 3.49 & & & & 2ulxd & + ngc4388 & x1 & 12:25:47.12 & 12:39:37.6 & 2 & 4.8 & 0.11 & 12.0 & 23 & & & ( ngc4388 ) & n & + ngc4406 & x1 & 12:26:11.85 & 12:56:53.8 & 2 & 4.9 & 0.03 & 13.3 & 43 & 1 & & ( ngc4406 ) & n & + ngc4395 & x1 & 12:26:01.52 & 33:31:34.7 & 1 & 161.8 & 0.43 & 32.1 & 1.73 & & d & & 1ulxd & ulx1 + ngc4395 & x2 & 12:25:32.31 & 33:25:34.0 & 1 & 484.7 & 1.42 & 7.1 & 0.308 & & & & & + ngc4395 & x3 & 12:25:49.58 & 33:41:05.3 & 1 & 499.3 & 1.34 & 6.0 & 0.288 & & & & & + ngc4395 & x4 & 12:26:28.92 & 33:36:30.7 & 1 & 536.9 & 1.63 & 5.3 & 0.23 & & & & & + ngc4435 & x1 & 12:27:40.45 & 13:04:48.1 & 2 & 2.2 & 0.04 & 7.1 & 6.67 & & & ( ngc4435 ) & n & + ngc4438 & x1 & 12:27:45.53 & 13:00:30.5 & 3 & 5.5 & 0.04 & 19.3 & 25.7 & & & ( ngc4438 ) & n & + ngc4438 & x2 & 12:27:57.22 & 13:02:32.6 & 3 & 207.2 & 1.27 & 16.6 & 16.4 & & & p b2/19.54 & 2eulxd & + ngc4449 & x1 & 12:28:09.36 & 44:05:52.0 & 3 & 25.0 & 0.19 & 4.7 & 0.107 & & & & & + ngc4449 & x2 & 12:28:09.28 & 44:05:09.2 & 3 & 38.3 & 0.21 & 14.7 & 0.492 & & & & & + ngc4449 & x3 & 12:28:12.76 & 44:06:22.5 & 3 & 44.9 & 0.27 & 12.4 & 0.301 & & & & & + ngc4449 & x4 & 12:28:11.07 & 44:06:45.3 & 3 & 65.5 & 0.44 & 14.2 & 0.45 & 1 & & & & + ngc4449 & x5 & 12:28:17.74 & 44:06:34.9 & 3 & 87.5 & 0.48 & 9.9 & 0.289 & & & & & + ngc4449 & x6 & 12:28:10.63 & 44:03:36.4 & 3 & 123.9 & 0.80 & 11.3 & 0.247 & & & & & + ngc4449 & x7 & 12:28:01.81 & 44:08:04.5 & 3 & 177.7 & 1.35 & 4.6 & 0.103 & & & & & + ngc4449 & x8 & 12:27:47.56 & 44:02:06.0 & 3 & 334.4 & 1.82 & 4.6 & 0.093 & & & & & + ngc4472 & x1 & 12:29:46.92 & 08:00:06.1 & 0 & 10.3 & 0.04 & 63.9 & 181 & & & ( ngc4472 ) & n & + ngc4472 & x2 & 12:30:06.35 & 08:02:05.3 & 0 & 321.2 & 1.29 & 4.5 & 2.91 & & & & 2ulxd & + ngc4472 & x3 & 12:29:39.73 & 07:53:30.6 & 0 & 400.3 & 1.43 & 18.0 & 13 & & & p b2/20.98 & 2eulxd & + ngc4472 & x4 & 12:29:27.63 & 08:06:40.5 & 0 & 490.5 & 1.61 & 7.4 & 4.9 & & & mp b2/19.96 & 2ulxd & + ngc4472 & x5 & 12:29:23.63 & 07:53:60.0 & 0 & 493.6 & 1.93 & 15.4 & 13.3 & & & p b2/20.11 & 2eulxd & + ngc4472 & x6 & 12:29:34.63 & 08:09:37.4 & 0 & 605.6 & 1.98 & 4.9 & 4.9 & & & & 2ulxd & + ngc4476 & x1 & 12:29:58.11 & 12:21:48.1 & 1 & 55.8 & 1.26 & 4.6 & 4.31 & & & z & 2ulx & + ngc4486 & x1 & 12:30:49.34 & 12:23:39.5 & 2 & 16.3 & 0.07 & 159.5 & 1740 & 1 & & ( ngc4486 ) & n & + ngc4486 & x2 & 12:30:41.99 & 12:17:59.1 & 1 & 343.9 & 1.42 & 3.1 & 3.13 & & & & 2ulx & + ngc4485 & x1 & 12:30:30.34 & 41:41:41.0 & 2 & 25.1 & 0.39 & 9.2 & 3.87 & & d & & 1ulxd & ulx1 + ngc4490 & x1 & 12:30:36.47 & 41:38:39.8 & 2 & 17.0 & 0.15 & 15.3 & 3.42 & & & ( ngc4490 ) & n & + ngc4490 & x2 & 12:30:32.45 & 41:39:16.3 & 2 & 71.5 & 0.41 & 9.3 & 1.39 & & d & & 1ulxd & ulx1 + ngc4490 & x3 & 12:30:43.17 & 41:38:13.7 & 2 & 73.1 & 0.50 & 7.6 & 1.57 & & d & & 1ulxd & ulx2 + ngc4490 & x4 & 12:30:30.42 & 41:39:06.4 & 2 & 82.7 & 0.44 & 7.2 & 1.32 & & d & & 1ulxd & ulx3 + ngc4501 & x1 & 12:32:00.71 & 14:24:41.6 & 2 & 38.9 & 0.20 & 5.9 & 14.7 & & s & & 1eulxd & ulx1 + ngc4523 & x1 & 12:33:52.93 & 15:11:44.8 & 2 & 118.6 & 1.98 & 5.2 & 4.84 & & & & 2ulxd & + pgc41763 & x1 & 12:33:51.98 & 15:21:48.6 & 2 & 77.0 & 1.90 & 8.4 & 19.2 & & & mp & 2eulxd & + ngc4526 & x1 & 12:34:02.65 & 07:42:04.3 & 1 & 5.0 & 0.03 & 5.0 & 7.16 & & & ( ngc4526 ) & n & + ngc4536 & x1 & 12:34:27.11 & 02:11:15.6 & 0 & 4.6 & 0.02 & 4.0 & 5.76 & & & ( ngc4536 ) & n & + ngc4536 & x2 & 12:34:33.81 & 02:13:07.6 & 0 & 150.1 & 1.54 & 7.2 & 10.5 & & & z & 2eulxd & + ngc4552 & x1 & 12:35:39.60 & 12:33:23.7 & 0 & 4.6 & 0.03 & 45.7 & 82.6 & & & ( ngc4552 ) & n & + ngc4552 & x2 & 12:35:29.42 & 12:31:05.9 & 0 & 207.0 & 1.42 & 3.9 & 2.67 & & & & 2ulx & + ngc4559 & x1 & 12:35:58.63 & 27:57:39.0 & 1 & 11.3 & 0.08 & 43.3 & 3.16 & & d & & 1ulxd & ulx1 + ngc4559 & x2 & 12:35:58.07 & 27:58:07.1 & 1 & 31.3 & 0.16 & 10.2 & 0.46 & & & & & + ngc4559 & x3 & 12:36:03.23 & 27:58:03.7 & 1 & 77.2 & 0.58 & 5.5 & 0.314 & & & & & + ngc4559 & x4 & 12:35:56.57 & 27:59:22.4 & 1 & 107.7 & 0.43 & 5.2 & 0.207 & & s & & & + ngc4559 & x5 & 12:35:51.58 & 27:56:01.7 & 1 & 125.2 & 0.91 & 66.9 & 5.05 & & & mp & 1ulxd & ulx2 + ngc4559 & x6 & 12:35:56.29 & 27:55:30.2 & 1 & 127.4 & 0.69 & 4.9 & 0.216 & & & & & + ngc4559 & x7 & 12:35:40.91 & 27:58:28.0 & 1 & 229.8 & 1.37 & 3.7 & 0.158 & & & & & + ngc4559 & x8 & 12:36:14.41 & 27:55:43.3 & 1 & 247.3 & 1.21 & 4.2 & 0.178 & & & & & + ngc4559 & x9 & 12:35:33.57 & 28:00:08.5 & 1 & 355.4 & 1.79 & 8.6 & 0.464 & & & & & + ngc4559 & x10 & 12:36:19.00 & 27:48:16.2 & 1 & 626.4 & 1.96 & 5.0 & 0.258 & & & & & + ngc4565 & x1 & 12:36:20.79 & 25:59:14.2 & 1 & 9.5 & 0.13 & 6.8 & 12.1 & & & ( ngc4565 ) & n & + ngc4565 & x2 & 12:36:17.45 & 25:58:53.1 & 1 & 44.1 & 0.61 & 12.6 & 25.2 & & & ( gc ) & 1eulxd & ulx1 + ngc4565 & x3 & 12:36:31.35 & 25:59:36.2 & 1 & 148.3 & 1.97 & 3.5 & 5.34 & & & p b2/20.55 & 2ulxd & + ngc4569 & x1 & 12:36:49.76 & 13:09:44.8 & 3 & 6.0 & 0.03 & 13.5 & 13.3 & & & ( ngc4569 ) & n & + ngc4569 & x2 & 12:37:09.48 & 13:15:09.7 & 3 & 428.5 & 1.76 & 9.6 & 6.37 & & & * j/11.028 & 2ulxd & + ngc4593 & x1 & 12:39:39.47 & -05:20:34.7 & 1 & 5.9 & 0.05 & 275.7 & 15600 & 1 & & ( ngc4593 ) & n & + ngc4594 & x1 & 12:39:59.55 & -11:37:24.1 & 4 & 3.1 & 0.02 & 30.7 & 22 & 1 & & ( ngc4594 ) & n & + ngc4594 & x2 & 12:40:01.84 & -11:37:22.8 & 4 & 35.8 & 0.14 & 10.8 & 5.83 & & d & & 1ulxd & ulx1 + ngc4594 & x3 & 12:39:48.69 & -11:37:17.8 & 4 & 157.4 & 0.60 & 3.6 & 1.25 & & d & & 1ulxd & ulx2 + ngc4594 & x4 & 12:39:45.22 & -11:38:54.7 & 4 & 228.0 & 1.18 & 14.2 & 5.84 & & & g0 * & 2ulxd & + ngc4594 & x5 & 12:40:22.64 & -11:39:33.4 & 4 & 365.8 & 1.80 & 3.7 & 1.04 & & & & 2ulxd & + ngc4627 & x1 & 12:42:04.07 & 32:34:07.7 & 0 & 59.2 & 1.10 & 4.0 & 0.584 & & & p b2/20.29 & & + ngc4631 & x1 & 12:41:57.29 & 32:32:01.4 & 0 & 134.2 & 0.36 & 4.7 & 0.569 & & d & & & + ngc4631 & x2 & 12:41:55.43 & 32:32:14.3 & 0 & 155.8 & 0.34 & 13.2 & 2.42 & & d & ( hii ) & 1ulxd & ulx1 + ngc4636 & x1 & 12:42:49.73 & 02:41:13.0 & 3 & 4.1 & 0.03 & 42.6 & 148 & & & ( ngc4636 ) & n & + ngc4649 & x1 & 12:43:39.81 & 11:33:12.1 & 0 & 15.8 & 0.08 & 65.4 & 229 & & & ( ngc4649 ) & n & + ngc4649 & x2 & 12:43:36.27 & 11:30:07.9 & 0 & 180.1 & 0.99 & 7.3 & 10.3 & & & p b1/17.84 & 1eulxd & + ngc4649 & x3 & 12:44:09.05 & 11:33:35.9 & 0 & 424.1 & 1.96 & 13.1 & 12.1 & & & p b1/19.79 & 2eulxd & + ngc4639 & x1 & 12:42:52.38 & 13:15:26.8 & 1 & 4.6 & 0.08 & 26.2 & 113 & & & ( ngc4639 ) & n & + ngc4651 & x1 & 12:43:42.69 & 16:23:35.2 & 2 & 4.9 & 0.06 & 4.6 & 3.45 & & & ( ngc4651 ) & n & + ngc4656 & x1 & 12:43:41.11 & 32:04:58.0 & 0 & 385.2 & 0.85 & 10.5 & 1.39 & & d & z & 1ulxd & ulx1 + ngc4666 & x1 & 12:45:08.36 & -00:27:37.6 & 2 & 9.7 & 0.20 & 3.4 & 2.08 & & & ( ngc4666 ) & n & + ngc4697 & x1 & 12:48:35.87 & -05:48:02.6 & 2 & 0.7 & 0.00 & 6.5 & 1.29 & 1 & & ( ngc4697 ) & n & + ngc4697 & x2 & 12:48:32.92 & -05:47:42.4 & 2 & 48.7 & 0.29 & 3.2 & 0.517 & & & & & + ngc4697 & x3 & 12:48:39.58 & -05:48:08.7 & 2 & 55.3 & 0.29 & 4.2 & 0.777 & & & & & + ngc4697 & x5 & 12:48:47.01 & -05:48:49.3 & 0 & 172.4 & 0.97 & 3.1 & 3.33 & & & p b1/18.81 & 1ulx & ulx1 + ngc4697 & x6 & 12:48:45.45 & -05:45:37.6 & 2 & 202.9 & 1.05 & 3.5 & 0.638 & & & & & + ngc4697 & x7 & 12:48:47.92 & -05:45:08.1 & 2 & 249.8 & 1.28 & 5.2 & 1.01 & & & & 2ulxd & + ngc4697 & x8 & 12:48:20.11 & -05:51:55.3 & 2 & 331.6 & 1.70 & 3.2 & 0.422 & & & p b1/20.29 & & + ngc4736 & x1 & 12:50:53.02 & 41:07:17.2 & 1 & 9.7 & 0.03 & 136.4 & 13.3 & & & ( ngc4736 ) & n & + ngc4736 & x2 & 12:50:47.24 & 41:05:15.0 & 1 & 135.5 & 0.49 & 5.2 & 0.121 & & & & & + ngc4736 & x3 & 12:50:44.25 & 41:04:59.3 & 1 & 168.1 & 0.60 & 4.8 & 0.121 & & & & & + ngc4736 & x4 & 12:50:59.90 & 41:10:59.7 & 1 & 240.5 & 0.88 & 6.2 & 0.386 & & & & & + ngc4736 & x5 & 12:50:52.19 & 41:02:55.2 & 1 & 255.2 & 0.93 & 5.0 & 0.106 & & & & & + ngc4736 & x6 & 12:50:23.24 & 41:07:53.0 & 1 & 345.8 & 1.03 & 8.4 & 0.275 & & & & & + ngc4736 & x7 & 12:51:03.77 & 41:00:11.6 & 1 & 433.9 & 1.52 & 5.9 & 0.142 & & & & & + ngc4736 & x8 & 12:50:27.55 & 41:13:19.4 & 1 & 472.3 & 1.53 & 4.8 & 0.119 & & & & & + ngc4736 & x9 & 12:50:58.56 & 40:59:08.7 & 1 & 484.6 & 1.73 & 6.3 & 0.132 & & & & & + ngc4736 & x10 & 12:51:11.45 & 41:15:14.9 & 1 & 525.1 & 1.91 & 5.6 & 0.241 & & & & & + ngc4736 & x11 & 12:50:34.66 & 41:15:27.5 & 1 & 541.4 & 1.85 & 4.7 & 0.0973 & & & & & + ngc4736 & x12 & 12:50:02.29 & 41:09:27.9 & 1 & 595.7 & 1.77 & 3.2 & 0.0717 & & & & & + ngc4826 & x1 & 12:56:43.54 & 21:40:56.3 & 1 & 13.7 & 0.08 & 11.8 & 4.49 & & & ( ngc4826 ) & n & + ngc4861 & x1 & 12:59:02.00 & 34:51:12.0 & 1 & 34.3 & 0.30 & 25.1 & 31.4 & & d & & 1eulxd & ulx1 + ngc4861 & x2 & 12:59:00.45 & 34:50:41.3 & 1 & 69.0 & 0.59 & 7.2 & 8.4 & & & ( hii ) & 1ulxd & ulx2 + ngc5005 & x1 & 13:10:56.16 & 37:03:25.2 & 0 & 3.8 & 0.04 & 26.6 & 58.3 & & & ( ngc5005 ) & n & + ngc5033 & x1 & 13:13:27.56 & 36:35:36.5 & 0 & 5.5 & 0.04 & 54.1 & 135 & & & ( ngc5033 ) & n & + ngc5055 & x1 & 13:15:49.30 & 42:01:45.5 & 2 & 20.5 & 0.09 & 7.3 & 2.2 & & & ( ngc5055 ) & n & + ngc5055 & x2 & 13:15:46.71 & 42:01:58.4 & 2 & 29.9 & 0.10 & 7.5 & 1.96 & & s & & 1ulxd & ulx1 + ngc5055 & x3 & 13:15:40.69 & 42:01:48.2 & 2 & 97.6 & 0.30 & 4.9 & 1.18 & & s & & 1ulxd & ulx2 + ngc5055 & x4 & 13:15:47.13 & 42:00:25.1 & 2 & 103.8 & 0.48 & 4.5 & 1.05 & & s & & 1ulxd & ulx3 + ngc5055 & x5 & 13:16:02.49 & 42:01:55.4 & 2 & 147.3 & 0.40 & 3.7 & 0.762 & & s & & & + ngc5055 & x6 & 13:15:30.27 & 42:03:14.0 & 2 & 222.6 & 0.59 & 5.9 & 1.44 & & & k0 * & 1ulxd & + ngc5055 & x7 & 13:15:19.43 & 42:03:03.2 & 2 & 337.6 & 0.90 & 28.2 & 9.36 & & & & 1ulxd & ulx4 + ngc5055 & x8 & 13:15:18.12 & 42:04:08.6 & 2 & 368.3 & 0.98 & 3.8 & 0.785 & & & & & + ngc5055 & x9 & 13:15:08.51 & 42:01:13.9 & 2 & 457.4 & 1.37 & 4.2 & 1.02 & & & * j/9.457 & 2ulxd & + ngc5055 & x10 & 13:14:54.58 & 42:02:39.1 & 2 & 610.6 & 1.69 & 4.8 & 1.19 & & & & 2ulx & + ngc5128 & x1 & 13:25:40.68 & -43:01:18.8 & 6 & 129.5 & 0.21 & 3.0 & 0.0953 & & & & & + ngc5128 & x2 & 13:25:38.23 & -43:03:09.5 & 6 & 164.3 & 0.27 & 3.3 & 0.115 & & & & & + ngc5128 & x3 & 13:25:40.83 & -43:02:42.8 & 6 & 165.5 & 0.28 & 3.4 & 0.11 & & & & & + ngc5128 & x4 & 13:25:19.94 & -43:03:15.8 & 3 & 168.2 & 0.22 & 39.0 & 9.27 & & & & 1ulx & ulx1 + ngc5128 & x5 & 13:25:12.89 & -43:01:13.9 & 6 & 177.2 & 0.27 & 3.1 & 0.103 & & & & & + ngc5128 & x7 & 13:25:07.59 & -43:01:13.5 & 6 & 235.2 & 0.36 & 5.2 & 0.41 & & & & & + ngc5128 & x8 & 13:25:22.33 & -42:57:15.5 & 6 & 236.2 & 0.37 & 6.0 & 0.742 & & & & & + ngc5128 & x9 & 13:25:06.31 & -43:02:23.3 & 6 & 262.3 & 0.38 & 7.1 & 0.308 & & & & & + ngc5128 & x10 & 13:25:33.44 & -43:05:26.4 & 6 & 270.8 & 0.41 & 3.3 & 0.105 & & & f0 * & & + ngc5128 & x12 & 13:25:54.60 & -42:59:20.9 & 6 & 297.8 & 0.43 & 5.0 & 0.154 & & & & & + ngc5128 & x13 & 13:25:07.34 & -43:04:07.9 & 6 & 302.8 & 0.40 & 45.6 & 2.68 & 1 & & * j/10.358 & 1ulxd & + ngc5128 & x14 & 13:25:02.40 & -43:02:42.6 & 6 & 309.2 & 0.44 & 10.4 & 0.485 & & & & & + ngc5128 & x16 & 13:25:58.84 & -43:04:29.8 & 6 & 388.6 & 0.65 & 14.8 & 0.974 & & & & & + ngc5128 & x17 & 13:26:01.49 & -43:05:28.0 & 6 & 445.7 & 0.74 & 12.0 & 0.808 & & & & & + ngc5128 & x18 & 13:24:57.08 & -43:05:36.6 & 6 & 446.0 & 0.59 & 7.6 & 0.287 & & & & & + ngc5128 & x19 & 13:25:00.68 & -43:06:43.7 & 6 & 463.0 & 0.60 & 11.1 & 0.481 & 1 & & & & + ngc5128 & x20 & 13:25:35.40 & -42:52:57.4 & 6 & 487.7 & 0.67 & 3.9 & 0.13 & & & & & + ngc5128 & x21 & 13:24:52.31 & -43:05:43.2 & 3 & 491.8 & 0.66 & 3.8 & 0.793 & & & & & + ngc5128 & x22 & 13:25:49.32 & -42:52:38.8 & 6 & 548.6 & 0.72 & 6.2 & 0.236 & & & & & + ngc5128 & x23 & 13:25:46.47 & -43:09:42.6 & 6 & 556.6 & 0.87 & 4.5 & 0.155 & & & & & + ngc5128 & x24 & 13:25:41.96 & -43:10:42.9 & 6 & 599.9 & 0.91 & 8.7 & 0.408 & & & & & + ngc5128 & x25 & 13:25:30.64 & -43:11:15.8 & 6 & 616.1 & 0.89 & 4.6 & 0.15 & & & & & + ngc5128 & x26 & 13:24:57.97 & -43:09:46.7 & 6 & 626.8 & 0.81 & 5.1 & 0.17 & & & & & + ngc5128 & x27 & 13:25:45.94 & -43:11:01.8 & 6 & 629.8 & 0.97 & 6.8 & 0.347 & & & & & + ngc5128 & x28 & 13:26:26.18 & -43:02:55.4 & 6 & 637.4 & 1.03 & 9.2 & 0.961 & 1 & & & & + ngc5128 & x29 & 13:25:05.64 & -43:11:09.4 & 6 & 660.9 & 0.87 & 3.9 & 0.131 & & & & & + ngc5128 & x30 & 13:24:32.66 & -42:56:11.4 & 6 & 682.3 & 1.13 & 14.4 & 0.844 & & & & & + ngc5128 & x31 & 13:25:00.96 & -43:12:00.1 & 6 & 728.0 & 0.95 & 4.7 & 0.195 & & & & & + ngc5128 & x32 & 13:26:12.65 & -43:10:52.1 & 6 & 761.0 & 1.25 & 7.3 & 0.338 & & & & & + ngc5128 & x33 & 13:26:04.47 & -42:47:03.2 & 6 & 923.1 & 1.21 & 6.5 & 1.01 & & & * b2/14.11 & 2ulx & + ngc5128 & x35 & 13:24:28.45 & -42:46:01.4 & 6 & 1118.2 & 1.83 & 3.7 & 0.381 & & & & & + ngc5128 & x36 & 13:24:14.82 & -43:14:15.6 & 2 & 1136.9 & 1.49 & 3.2 & 0.621 & & & & & + ngc5128 & x37 & 13:26:26.16 & -43:17:15.6 & 3 & 1158.8 & 1.88 & 3.4 & 1.96 & & & & 2ulx & + ngc5128 & x38 & 13:26:56.81 & -42:49:53.6 & 2 & 1172.3 & 1.58 & 3.7 & 1.97 & & & & 2ulx & + ngc5194 & x1 & 13:29:52.63 & 47:11:45.9 & 4 & 7.1 & 0.03 & 52.9 & 17.2 & & & ( ngc5194 ) & n & + ngc5194 & x3 & 13:29:45.79 & 47:10:45.2 & 4 & 99.0 & 0.46 & 4.8 & 0.516 & & & * j/12.261 & & + ngc5194 & x4 & 13:29:43.18 & 47:11:36.6 & 4 & 103.8 & 0.49 & 5.8 & 1.65 & & s & & 1ulx & ulx1 + ngc5194 & x5 & 13:29:47.60 & 47:09:43.7 & 2 & 137.2 & 0.54 & 4.7 & 1.12 & & s & & 1ulx & ulx2 + ngc5194 & x6 & 13:30:01.08 & 47:13:45.4 & 4 & 141.7 & 0.59 & 18.0 & 3.95 & & s & & 1ulxd & ulx3 + ngc5194 & x7 & 13:29:39.89 & 47:12:42.5 & 4 & 147.1 & 0.62 & 10.4 & 1.81 & & s & & 1ulxd & ulx4 + ngc5194 & x8 & 13:30:07.71 & 47:11:07.1 & 4 & 152.5 & 0.67 & 23.4 & 3.65 & 1 & s & & 1ulxd & ulx5 + ngc5194 & x9 & 13:29:53.54 & 47:14:40.0 & 4 & 172.0 & 0.55 & 4.8 & 1.31 & & s & & 1ulx & ulx6 + ngc5194 & x10 & 13:30:06.10 & 47:08:36.5 & 4 & 231.8 & 0.74 & 4.4 & 0.415 & & & & & + ngc5194 & x11 & 13:29:58.71 & 47:05:29.5 & 4 & 382.5 & 1.16 & 4.8 & 0.43 & & & & & + ngc5194 & x12 & 13:29:38.53 & 47:18:58.7 & 4 & 456.2 & 1.36 & 4.2 & 0.882 & & & p b1/20.07 & & + ngc5195 & x1 & 13:29:59.28 & 47:16:02.1 & 4 & 19.9 & 0.14 & 13.2 & 3.27 & & & ( ngc5195 ) & n & + ngc5195 & x2 & 13:30:06.23 & 47:15:45.3 & 4 & 84.5 & 0.54 & 5.5 & 0.488 & & & & & + ngc5195 & x4 & 13:30:29.11 & 47:16:11.2 & 4 & 309.7 & 1.82 & 4.3 & 0.417 & & & & & + ngc5204 & x1 & 13:29:38.75 & 58:25:07.5 & 1 & 18.8 & 0.20 & 54.0 & 5.15 & 1 & & ( b0ib * ) & 1ulxd & ulx1 + ngc5236 & x1 & 13:37:00.73 & -29:51:56.5 & 3 & 9.4 & 0.03 & 46.3 & 7.13 & & & ( ngc5236 ) & n & + ngc5236 & x2 & 13:37:04.68 & -29:51:21.7 & 3 & 71.0 & 0.20 & 7.8 & 0.644 & & & & & + ngc5236 & x3 & 13:37:01.42 & -29:53:23.0 & 3 & 80.3 & 0.21 & 3.9 & 0.268 & & & & & + ngc5236 & x4 & 13:37:05.67 & -29:53:06.6 & 3 & 93.8 & 0.26 & 4.6 & 0.281 & & & & & + ngc5236 & x6 & 13:36:59.85 & -29:49:55.5 & 3 & 128.6 & 0.33 & 4.9 & 0.323 & & & & & + ngc5236 & x7 & 13:37:04.45 & -29:54:05.3 & 3 & 132.7 & 0.35 & 6.7 & 0.451 & & s & & & + ngc5236 & x8 & 13:36:49.30 & -29:52:58.7 & 3 & 153.2 & 0.44 & 4.8 & 0.302 & & & & & + ngc5236 & x9 & 13:36:46.01 & -29:52:01.9 & 3 & 185.9 & 0.54 & 3.4 & 0.217 & & & & & + ngc5236 & x11 & 13:37:20.12 & -29:53:47.7 & 3 & 277.9 & 0.80 & 32.5 & 2.63 & 1 & s & & 1ulxd & ulx1 + ngc5236 & x12 & 13:37:37.21 & -29:48:21.0 & 3 & 529.6 & 1.51 & 4.7 & 0.265 & & & & & + ngc5273 & x1 & 13:42:08.18 & 35:39:14.4 & 0 & 3.6 & 0.05 & 5.5 & 4.72 & & & ( ngc5273 ) & n & + ngc5322 & x1 & 13:49:15.36 & 60:11:24.2 & 0 & 5.2 & 0.04 & 9.7 & 29.7 & & & ( ngc5322 ) & n & + ngc5350 & x1 & 13:53:21.94 & 40:21:52.2 & 0 & 5.8 & 0.06 & 5.0 & 21.4 & & & ( ngc5350 ) & n & + ngc5353 & x1 & 13:53:26.36 & 40:17:03.6 & 0 & 5.2 & 0.13 & 18.8 & 106 & & & ( ngc5353 ) & n & + ngc5457 & x1 & 14:03:12.66 & 54:20:54.9 & 4 & 1.4 & 0.00 & 8.2 & 0.466 & 1 & & ( ngc5457 ) & n & + ngc5457 & x2 & 14:03:21.58 & 54:19:45.2 & 3 & 105.8 & 0.13 & 5.2 & 0.282 & 1 & s & & & + ngc5457 & x3 & 14:03:24.80 & 54:19:47.6 & 4 & 126.9 & 0.15 & 4.5 & 0.321 & 1 & s & & & + ngc5457 & x4 & 14:02:52.99 & 54:21:12.6 & 3 & 171.5 & 0.21 & 3.9 & 0.187 & & & & & + ngc5457 & x5 & 14:03:32.17 & 54:21:02.7 & 3 & 172.1 & 0.21 & 6.2 & 0.362 & & & & & + ngc5457 & x6 & 14:03:31.11 & 54:21:55.8 & 4 & 173.6 & 0.21 & 4.2 & 0.335 & & & & & + ngc5457 & x7 & 14:03:35.94 & 54:19:23.1 & 3 & 224.7 & 0.28 & 6.3 & 0.302 & 1 & s & & & + ngc5457 & x8 & 14:03:19.70 & 54:17:15.9 & 4 & 227.9 & 0.26 & 6.9 & 0.694 & & s & & & + ngc5457 & x9 & 14:03:41.19 & 54:19:02.5 & 3 & 275.0 & 0.34 & 4.5 & 0.307 & & s & & & + ngc5457 & x11 & 14:03:54.01 & 54:21:58.8 & 4 & 368.4 & 0.46 & 8.1 & 0.48 & 1 & s & & & + ngc5457 & x12 & 14:02:29.67 & 54:21:18.1 & 4 & 375.1 & 0.46 & 26.8 & 1.57 & 1 & s & p b/17.80 & 1ulxd & ulx1 + ngc5457 & x13 & 14:03:50.83 & 54:24:14.2 & 4 & 389.7 & 0.47 & 3.8 & 0.178 & & & & & + ngc5457 & x14 & 14:03:16.03 & 54:27:28.4 & 4 & 394.6 & 0.46 & 3.9 & 0.309 & & & & & + ngc5457 & x15 & 14:03:03.93 & 54:27:35.8 & 4 & 407.8 & 0.47 & 21.8 & 1.54 & 1 & s & & 1ulxd & ulx2 + ngc5457 & x16 & 14:02:46.82 & 54:26:53.8 & 4 & 423.1 & 0.50 & 4.5 & 0.398 & & & & & + ngc5457 & x17 & 14:02:28.43 & 54:16:24.0 & 4 & 471.4 & 0.57 & 14.0 & 0.891 & 1 & s & & & + ngc5457 & x18 & 14:02:21.90 & 54:17:56.9 & 4 & 477.1 & 0.59 & 6.3 & 0.374 & & & & & + ngc5457 & x19 & 14:02:28.53 & 54:26:31.8 & 4 & 510.8 & 0.61 & 6.3 & 0.541 & 1 & & & & + ngc5457 & x20 & 14:02:14.21 & 54:21:57.2 & 2 & 513.3 & 0.63 & 4.1 & 0.173 & 1 & & & & + ngc5457 & x21 & 14:04:22.14 & 54:19:21.8 & 4 & 616.1 & 0.76 & 12.6 & 1.25 & 2 & & * j/10.732 & 1ulx & + ngc5457 & x22 & 14:02:03.23 & 54:18:28.7 & 3 & 623.3 & 0.77 & 4.6 & 0.23 & & & & & + ngc5457 & x23 & 14:04:14.40 & 54:26:05.9 & 4 & 623.6 & 0.76 & 43.7 & 3.17 & 1 & s & & 1ulxd & ulx3 + ngc5457 & x24 & 14:04:16.85 & 54:16:10.9 & 3 & 630.8 & 0.77 & 8.1 & 0.424 & 1 & & & & + ngc5457 & x25 & 14:02:27.93 & 54:12:39.0 & 4 & 631.1 & 0.75 & 11.7 & 0.773 & 1 & & & & + ngc5457 & x26 & 14:04:29.41 & 54:23:56.7 & 3 & 696.1 & 0.86 & 10.0 & 0.615 & 1 & & & & + ngc5457 & x27 & 14:04:01.23 & 54:11:27.6 & 2 & 710.1 & 0.84 & 3.6 & 0.165 & 1 & & & & + ngc5457 & x28 & 14:02:09.88 & 54:30:18.4 & 4 & 784.9 & 0.94 & 4.5 & 0.549 & & & & & + ngc5457 & x29 & 14:04:00.25 & 54:09:15.6 & 3 & 815.0 & 0.96 & 4.2 & 0.228 & & & & & + ngc5457 & x30 & 14:03:05.64 & 54:07:06.5 & 3 & 830.7 & 0.96 & 4.0 & 0.327 & & & & & + ngc5457 & x31 & 14:02:07.72 & 54:10:30.2 & 2 & 844.1 & 1.01 & 5.6 & 0.376 & 1 & & & & + ngc5457 & x32 & 14:01:34.04 & 54:20:42.5 & 2 & 860.9 & 1.07 & 7.4 & 1.26 & & & & 2ulx & + ngc5457 & x34 & 14:01:19.30 & 54:18:36.1 & 4 & 999.8 & 1.24 & 18.7 & 2.26 & 1 & & * j/10.453 & 2ulxd & + ngc5457 & x35 & 14:03:03.93 & 54:04:14.7 & 3 & 1003.1 & 1.16 & 7.1 & 0.891 & 2 & & & & + ngc5457 & x36 & 14:01:23.46 & 54:15:19.1 & 2 & 1011.8 & 1.24 & 4.3 & 0.529 & 1 & & & & + ngc5506 & x1 & 14:13:14.83 & -03:12:26.6 & 1 & 1.1 & 0.02 & 216.6 & 455 & 1 & & ( ngc5506 ) & n & + ngc5774 & x1 & 14:53:45.07 & 03:33:31.0 & 1 & 89.5 & 1.00 & 4.7 & 7.72 & & s & & 2ulxd & ulx1 + ngc5774 & x2 & 14:53:43.19 & 03:37:06.9 & 1 & 131.9 & 1.58 & 3.8 & 5.3 & & & & 2ulxd & + ngc6434 & x1 & 17:36:36.16 & 72:05:29.3 & 3 & 57.2 & 0.83 & 16.1 & 78.6 & & & qso j17366 + 7205 & 1eulxd & + ngc6500 & x1 & 17:55:59.58 & 18:20:15.9 & 2 & 13.0 & 0.27 & 4.6 & 64.3 & & & ( ngc6500 ) & n & + ngc6503 & x1 & 17:49:12.42 & 70:09:28.4 & 2 & 91.1 & 0.43 & 3.9 & 0.554 & & d & & & + ngc6654 & x1 & 18:24:03.60 & 73:09:29.0 & 3 & 92.8 & 1.19 & 4.8 & 13.5 & & & p & 2eulxd & + ngc6654 & x2 & 18:24:36.45 & 73:10:51.8 & 3 & 120.2 & 1.91 & 5.8 & 13.8 & & & p & 2eulxd & + ngc6764 & x1 & 19:08:16.30 & 50:56:00.6 & 4 & 7.6 & 0.19 & 31.4 & 68.2 & & & ( ngc6764 ) & n & + ngc6946 & x1 & 20:34:52.62 & 60:09:11.9 & 7 & 5.6 & 0.02 & 13.9 & 1.67 & & & ( ngc6946 ) & n & + ngc6946 & x2 & 20:35:00.33 & 60:09:07.0 & 7 & 62.7 & 0.21 & 20.5 & 1.8 & 1 & s & & 1ulxd & ulx1 + ngc6946 & x3 & 20:35:01.52 & 60:10:07.2 & 7 & 88.1 & 0.29 & 4.5 & 0.421 & & & & & + ngc6946 & x4 & 20:34:36.53 & 60:09:30.4 & 7 & 116.5 & 0.40 & 15.8 & 1.6 & & s & & 1ulxd & ulx2 + ngc6946 & x5 & 20:34:34.86 & 60:10:33.3 & 7 & 150.0 & 0.49 & 4.2 & 0.358 & & & & & + ngc6946 & x6 & 20:35:01.05 & 60:11:31.8 & 7 & 152.5 & 0.46 & 71.1 & 11 & & s & ( snr ) & 1eulxd & ulx3 + ngc6946 & x7 & 20:35:11.81 & 60:07:30.9 & 7 & 180.9 & 0.59 & 3.8 & 0.309 & & & & & + ngc6946 & x8 & 20:34:49.59 & 60:12:21.8 & 7 & 187.7 & 0.55 & 3.5 & 0.244 & & & & & + ngc6946 & x9 & 20:34:25.95 & 60:09:09.7 & 7 & 194.5 & 0.66 & 8.5 & 0.584 & & & & & + ngc6946 & x10 & 20:34:49.04 & 60:05:55.1 & 7 & 201.2 & 0.59 & 5.1 & 0.392 & & & * b/12.18 & & + ngc6946 & x11 & 20:35:17.88 & 60:10:56.2 & 7 & 218.0 & 0.72 & 4.5 & 0.355 & & & & & + ngc6946 & x12 & 20:35:36.64 & 60:06:42.0 & 7 & 366.8 & 1.22 & 4.3 & 0.25 & & & & & + ngc6946 & x13 & 20:33:51.83 & 60:10:48.8 & 7 & 458.7 & 1.55 & 5.2 & 0.414 & & & & & + ngc6946 & x14 & 20:34:27.22 & 60:02:06.7 & 7 & 466.7 & 1.40 & 6.2 & 0.45 & & & & & + ngc6946 & x15 & 20:35:11.97 & 60:01:19.0 & 7 & 498.8 & 1.47 & 5.8 & 0.432 & & & & & + ngc6946 & x16 & 20:34:09.46 & 60:16:15.9 & 7 & 527.0 & 1.63 & 4.8 & 0.327 & & & & & + ngc6946 & x17 & 20:33:46.22 & 60:14:18.2 & 7 & 576.5 & 1.89 & 6.0 & 0.555 & & & & & + ngc7172 & x1 & 22:02:01.46 & -31:52:08.5 & 1 & 2.3 & 0.05 & 7.8 & 21.8 & & & ( ngc7172 ) & n & + ngc7213 & x1 & 22:09:16.24 & -47:09:59.9 & 1 & 7.0 & 0.08 & 526.8 & 3080 & & & ( ngc7213 ) & n & + ngc7217 & x1 & 22:07:52.59 & 31:21:38.3 & 1 & 5.9 & 0.05 & 5.6 & 12.5 & & & ( ngc7217 ) & n & + ngc7217 & x2 & 22:07:36.99 & 31:21:15.4 & 1 & 195.8 & 1.69 & 3.3 & 4.67 & & & & 2ulxd & + pgc68618 & x1 & 22:20:37.01 & -46:02:28.6 & 5 & 213.6 & 1.51 & 5.8 & 1.28 & & & p b/18.40 & 2ulxd & + ngc7314 & x1 & 22:35:46.21 & -26:03:01.6 & 1 & 5.8 & 0.09 & 201.5 & 202 & 1 & & ( ngc7314 ) & n & + ngc7314 & x2 & 22:35:48.13 & -26:01:28.2 & 1 & 99.8 & 0.82 & 5.6 & 2.34 & & s & & 1ulxd & ulx1 + ngc7314 & x4 & 22:35:41.51 & -26:00:30.6 & 1 & 163.0 & 1.50 & 4.4 & 1.81 & & & & 2ulxd & + ngc7314 & x5 & 22:35:48.03 & -25:58:33.9 & 1 & 270.8 & 1.98 & 11.3 & 5.21 & & & p b1/20.44 & 2ulxd & + pgc70090 & x1 & 22:57:10.32 & -36:27:43.0 & 2 & 2.9 & 0.03 & 29.5 & 140 & & & ( pgc70090 ) & n & + ngc7465 & x1 & 23:02:00.95 & 15:57:53.3 & 1 & 3.5 & 0.12 & 60.5 & 331 & 1 & & ( ngc7465 ) & n & + ngc7552 & x1 & 23:16:10.85 & -42:35:05.6 & 2 & 4.6 & 0.05 & 19.7 & 80.8 & & & ( ngc7552 ) & n & + ngc7582 & x1 & 23:18:23.40 & -42:22:14.1 & 3 & 3.3 & 0.02 & 27.9 & 41.3 & & & ( ngc7582 ) & n & + ngc7582 & x2 & 23:18:30.15 & -42:20:35.9 & 1 & 121.6 & 1.74 & 7.9 & 7.81 & & & p b2/20.23 & 2ulxd & + ngc7590 & x1 & 23:18:56.48 & -42:13:57.6 & 3 & 25.4 & 0.32 & 5.0 & 6.54 & & & & 1ulx & ulx1 + ngc7714 & x1 & 23:36:14.12 & 02:09:18.1 & 1 & 8.6 & 0.20 & 20.2 & 91 & & & ( ngc7714 ) & n & + ngc7714 & x2 & 23:36:15.75 & 02:09:28.5 & 1 & 18.9 & 0.41 & 13.7 & 41.8 & & d & & 1eulxd & ulx1 + ngc7742 & x1 & 23:44:15.34 & 10:46:07.2 & 1 & 10.2 & 0.20 & 4.7 & 14.8 & & & ( ngc7742 ) & n & + ngc7742 & x2 & 23:44:17.47 & 10:45:51.0 & 1 & 25.3 & 0.48 & 3.4 & 7.39 & & & & 1ulxd & ulx1 + ngc253-ulx1 & rh600088a00.u & 48599 & 3.1 & 1.2d & @xmath203 & & & @xmath2012.5 & & @xmath202.64e-13 & @xmath200.285 & + & rh600088a01.u+ & 48779 & 25.4 & 1.2d & @xmath203 & & & @xmath2030.2 & & @xmath207.88e-14 & @xmath200.0851 & + & rh600714n00.u & 49720 & 10.9 & 4.1d & @xmath203 & & & @xmath2019.7 & & @xmath201.19e-13 & @xmath200.129 & + & rh600714a01.u & 49882 & 19.7 & 23.9d & @xmath203 & & & @xmath2023.0 & & @xmath207.72e-14 & @xmath200.0834 & + & rh601111n00.1 & 50802 & 17.3 & 6.9d & 38.6 & 15.49 & 0.85 & 566.1 & 3.92 & 2.19e-12 & 2.37 & 0.42 + & rh601113n00.u & 50996 & 2.1 & 0.6h & @xmath203 & & & @xmath2016.5 & & @xmath205.36e-13 & @xmath200.579 & + ngc720-x2 & rh600472n00.15 + & 49360 & 56.8 & 4.4d & 9.3 & 2.61 & 0.95 & 116.3 & 13.56 & 1.38e-13 & 12.7 & 0.67 + ngc720-x3 & rh600472n00.17 + & 49360 & 56.8 & 4.4d & 4.5 & 3.36 & 0.96 & 53.8 & 26.40 & 6.38e-14 & 5.86 & 0.52 + ngc891-ulx1 & rh600690n00.10 + & 49744 & 97.2 & 8.7d & 11.4 & 0.68 & 0.94 & 178.5 & 11.13 & 1.82e-13 & 1.53 & 0.034 + & rh500501n00.6 & 50663 & 41.5 & 4.0d & 5.5 & 0.69 & 0.98 & 45.0 & 23.31 & 1.07e-13 & 0.897 & + & rh500539n00.u & 50867 & 5.0 & 5.4h & @xmath203 & & & @xmath2011.3 & & @xmath202.22e-13 & @xmath201.86 & + ngc891-ulx2 & rh600690n00.11 + & 49744 & 97.2 & 8.7d & 27.0 & 1.15 & 0.89 & 440.0 & 5.69 & 4.49e-13 & 3.77 & 0.99 + & rh500501n00.7 & 50663 & 41.5 & 4.0d & 12.4 & 0.17 & 0.93 & 128.1 & 10.36 & 3.05e-13 & 2.56 & 0.54 + & rh500539n00.u & 50867 & 5.0 & 5.4h & @xmath203 & & & @xmath2011.3 & & @xmath202.22e-13 & @xmath201.86 & + ngc891-x6 & rh600690n00.9 + & 49744 & 97.2 & 8.7d & 9.8 & 5.40 & 0.95 & 147.7 & 13.06 & 1.51e-13 & 1.27 & 0.19 + & rh500501n00.5 & 50663 & 41.5 & 4.0d & 7.1 & 6.53 & 0.99 & 57.6 & 17.76 & 1.37e-13 & 1.15 & 0.036 + & rh500539n00.u & 50867 & 5.0 & 5.4h & @xmath203 & & & @xmath2011.2 & & @xmath202.21e-13 & @xmath201.85 & + ngc891-ulx3 & rh600690n00.7 + & 49744 & 97.2 & 8.7d & 11.9 & 5.41 & 0.94 & 168.3 & 10.71 & 1.72e-13 & 1.44 & 0.085 + & rh500501n00.4 & 50663 & 41.5 & 4.0d & 11.6 & 6.67 & 0.95 & 103.3 & 10.93 & 2.46e-13 & 2.06 & 0.91 + & rh500539n00.2 & 50867 & 5.0 & 5.4h & 3.5 & 6.70 & 1.00 & 13.8 & 26.06 & 2.72e-13 & 2.28 & + ngc1042-ulx1 & rh600469n00.2 & 49569 & 3.4 & 8.1d & 5.6 & 8.95 & 1.02 & 19.7 & 23.10 & 4.61e-13 & 3.9 & + & rh600469a01.2 + & 49911 & 21.9 & 16.9d & 17.1 & 8.93 & 0.95 & 139.9 & 7.99 & 5.04e-13 & 4.27 & 0.71 + ngc1068-ulx1 & rh150020n00.u & 48097 & 19.9 & 1.3d & @xmath203 & & & @xmath2015.6 & & @xmath206.38e-14 & @xmath201.59 & + & rh700347n00.u & 49548 & 22.4 & 3.8d & @xmath203 & & & @xmath2016.6 & & @xmath206.02e-14 & @xmath201.5 & + & rh701352n00.u & 49562 & 2.2 & 21.0h & @xmath203 & & & @xmath207.6 & & @xmath202.78e-13 & @xmath206.92 & + & rh701352a01.10 + & 49914 & 71.4 & 17.9d & 8.6 & 1.06 & 0.96 & 149.7 & 14.49 & 1.71e-13 & 4.26 & 0.55 + ngc1068-ulx2 & rh150020n00.u & 48097 & 19.9 & 1.3d & @xmath203 & & & @xmath2015.3 & & @xmath206.25e-14 & @xmath201.56 & + & rh700347n00.5 & 49548 & 22.4 & 3.8d & 4.0 & 2.85 & 0.95 & 28.6 & 26.41 & 1.04e-13 & 2.59 & + & rh701352n00.u & 49562 & 2.2 & 21.0h & @xmath203 & & & @xmath207.5 & & @xmath202.75e-13 & @xmath206.84 & + & rh701352a01.8 + & 49914 & 71.4 & 17.9d & 4.5 & 2.95 & 0.96 & 57.5 & 26.40 & 6.57e-14 & 1.63 & 0.00019 + ngc1068-x4 & rh150020n00.2 & 48097 & 19.9 & 1.3d & 32.8 & 6.15 & 0.89 & 254.8 & 4.63 & 1.04e-12 & 25.9 & 3.9e-06 + & rh700347n00.4 & 49548 & 22.4 & 3.8d & 21.6 & 6.02 & 0.90 & 175.6 & 6.77 & 6.37e-13 & 15.9 & 0.77 + & rh701352n00.1 & 49562 & 2.2 & 21.0h & 8.1 & 6.09 & 0.97 & 24.6 & 15.45 & 9.03e-13 & 22.5 & + & rh701352a01.7 + & 49914 & 71.4 & 17.9d & 44.1 & 6.07 & 0.89 & 669.0 & 3.37 & 7.64e-13 & 19 & 0.66 + ngc1068-x5 & rh150020n00.u & 48097 & 19.9 & 1.3d & @xmath203 & & & @xmath2015.5 & & @xmath206.32e-14 & @xmath201.57 & + & rh700347n00.u & 49548 & 22.4 & 3.8d & @xmath203 & & & @xmath2016.4 & & @xmath205.95e-14 & @xmath201.48 & + & rh701352n00.u & 49562 & 2.2 & 21.0h & @xmath203 & & & @xmath207.6 & & @xmath202.77e-13 & @xmath206.89 & + & rh701352a01.5 + & 49914 & 71.4 & 17.9d & 4.7 & 6.07 & 1.00 & 51.2 & 26.39 & 5.85e-14 & 1.46 & 0.098 + ngc1068-x7 & rh150020n00.u & 48097 & 19.9 & 1.3d & @xmath203 & & & @xmath2015.7 & & @xmath206.40e-14 & @xmath201.59 & + & rh700347n00.u & 49548 & 22.4 & 3.8d & @xmath203 & & & @xmath2016.6 & & @xmath206.03e-14 & @xmath201.5 & + & rh701352n00.u & 49562 & 2.2 & 21.0h & @xmath203 & & & @xmath207.6 & & @xmath202.79e-13 & @xmath206.94 & + & rh701352a01.4 + & 49914 & 71.4 & 17.9d & 4.4 & 6.82 & 1.00 & 44.9 & 26.40 & 5.12e-14 & 1.27 & + ngc1073-x1 & rh600999n00.4 + & 50659 & 27.5 & 9.5d & 5.8 & 1.69 & 0.96 & 36.8 & 22.10 & 1.14e-13 & 3.16 & + & rh600999a01.1 & 50828 & 11.8 & 2.7d & 3.7 & 1.72 & 0.95 & 15.6 & 26.35 & 1.12e-13 & 3.11 & + ngc1073-ulx1 & rh600999n00.5 + & 50659 & 27.5 & 9.5d & 3.5 & 1.54 & 0.95 & 21.4 & 26.16 & 6.63e-14 & 1.84 & + & rh600999a01.2 & 50828 & 11.8 & 2.7d & 4.0 & 1.43 & 0.96 & 18.3 & 26.41 & 1.31e-13 & 3.63 & + ngc1073-x3 & rh600999n00.6 + & 50659 & 27.5 & 9.5d & 3.3 & 1.86 & 0.94 & 18.7 & 25.80 & 5.80e-14 & 1.61 & + & rh600999a01.u & 50828 & 11.8 & 2.7d & @xmath203 & & & @xmath2012.8 & & @xmath209.17e-14 & @xmath202.54 & + ngc1073-x4 & rh600999n00.7 + & 50659 & 27.5 & 9.5d & 13.9 & 2.19 & 0.92 & 97.7 & 9.36 & 3.03e-13 & 8.4 & 0.045 + & rh600999a01.3 & 50828 & 11.8 & 2.7d & 7.2 & 2.19 & 0.96 & 39.2 & 17.55 & 2.82e-13 & 7.82 & + ngc1073-x6 & rh600999n00.10 + & 50659 & 27.5 & 9.5d & 4.2 & 2.82 & 0.95 & 33.8 & 26.41 & 1.05e-13 & 2.91 & + & rh600999a01.u & 50828 & 11.8 & 2.7d & @xmath203 & & & @xmath2012.6 & & @xmath209.08e-14 & @xmath202.52 & + ngc1291-ulx1 & rh600828n00.3 + & 50085 & 14.7 & 9.3h & 10.9 & 4.00 & 0.94 & 69.3 & 11.48 & 3.41e-13 & 3.03 & 0.1 + ngc1291-x3 & rh600828n00.1 + & 50085 & 14.7 & 9.3h & 5.1 & 7.54 & 1.01 & 37.1 & 24.97 & 1.83e-13 & 1.62 & + ngc1313-ulx1 & rh400065n00.3 & 48730 & 5.4 & 36.0d & 65.1 & 6.88 & 0.89 & 423.4 & 1.97 & 6.58e-12 & 10.8 & 1.4e-05 + & rh600505n00.3 & 49527 & 22.4 & 22.9d & 107.2 & 4.35 & 0.87 & 1327.8 & 1.21 & 4.94e-12 & 8.12 & 0.018 + & rh500403n00.2 & 49749 & 13.4 & 10.7d & 110.6 & 5.67 & 0.88 & 984.2 & 1.15 & 6.08e-12 & 9.99 & 0.00031 + & rh500404n00.3 & 49750 & 27.1 & 8.9d & 139.3 & 5.66 & 0.88 & 2063.4 & 1.06 & 6.32e-12 & 10.4 & 8.5e-06 + & rh600505a01.3 & 49819 & 20.2 & 8.1d & 122.8 & 4.37 & 0.87 & 1504.1 & 1.07 & 6.19e-12 & 10.2 & 1.2e-16 + & rh500404a01.3 & 49846 & 18.9 & 72.6d & 76.1 & 5.68 & 0.88 & 829.5 & 1.54 & 3.64e-12 & 5.98 & 3.5e-15 + & rh500403a01.2 + & 49846 & 31.0 & 72.4d & 76.3 & 5.68 & 0.88 & 1114.2 & 1.55 & 2.98e-12 & 4.9 & 1.1e-07 + & rh500492n00.3 & 50722 & 22.8 & 6.9d & 124.6 & 5.64 & 0.88 & 1738.2 & 1.05 & 6.33e-12 & 10.4 & 0.0015 + & rh500550n00.2 & 50893 & 23.9 & 28.4d & 68.4 & 5.67 & 0.88 & 807.5 & 1.81 & 2.80e-12 & 4.6 & 0.2 + ngc1313-ulx2 & rh400065n00.4 & 48730 & 5.4 & 36.0d & 21.2 & 5.03 & 0.90 & 89.9 & 6.84 & 1.40e-12 & 2.3 & 0.72 + & rh600505n00.5 & 49527 & 22.4 & 22.9d & 47.0 & 2.10 & 0.88 & 427.0 & 3.03 & 1.59e-12 & 2.61 & 0.52 + & rh500403n00.3 & 49749 & 13.4 & 10.7d & 37.9 & 0.10 & 0.88 & 307.8 & 4.07 & 1.90e-12 & 3.12 & 0.15 + & rh500404n00.6 & 49750 & 27.1 & 8.9d & 48.1 & 0.09 & 0.88 & 525.6 & 2.95 & 1.61e-12 & 2.64 & 0.81 + & rh600505a01.5 & 49819 & 20.2 & 8.1d & 52.3 & 2.13 & 0.88 & 423.1 & 2.75 & 1.74e-12 & 2.86 & 0.32 + & rh500404a01.5 & 49846 & 18.9 & 72.6d & 44.1 & 0.08 & 0.88 & 405.2 & 3.37 & 1.78e-12 & 2.92 & 0.31 + & rh500403a01.6 + & 49846 & 31.0 & 72.4d & 59.0 & 0.11 & 0.87 & 692.1 & 2.28 & 1.85e-12 & 3.04 & 0.41 + & rh500492n00.5 & 50722 & 22.8 & 6.9d & 47.0 & 0.10 & 0.88 & 388.1 & 3.03 & 1.41e-12 & 2.32 & 0.092 + & rh500550n00.4 & 50893 & 23.9 & 28.4d & 42.5 & 0.05 & 0.88 & 424.2 & 3.47 & 1.47e-12 & 2.41 & 0.98 + ngc1313-ulx3 & rh400065n00.2 & 48730 & 5.4 & 36.0d & 24.7 & 0.31 & 0.89 & 114.5 & 6.13 & 1.78e-12 & 2.92 & 0.076 + & rh600505n00.1 & 49527 & 22.4 & 22.9d & 27.2 & 3.84 & 0.89 & 224.2 & 5.64 & 8.35e-13 & 1.37 & 0.79 + & rh500403n00.1 & 49749 & 13.4 & 10.7d & 51.1 & 5.39 & 0.88 & 392.6 & 2.75 & 2.42e-12 & 3.98 & 0.47 + & rh500404n00.1 & 49750 & 27.1 & 8.9d & 54.1 & 5.37 & 0.88 & 558.5 & 2.54 & 1.71e-12 & 2.81 & 0.00035 + & rh600505a01.1 & 49819 & 20.2 & 8.1d & 36.0 & 3.76 & 0.88 & 276.7 & 4.32 & 1.14e-12 & 1.87 & 0.39 + & rh500404a01.2 & 49846 & 18.9 & 72.6d & 36.0 & 5.38 & 0.89 & 332.1 & 4.32 & 1.46e-12 & 2.4 & 3.7e-07 + & rh500403a01.1 + & 49846 & 31.0 & 72.4d & 40.4 & 5.38 & 0.89 & 414.4 & 3.75 & 1.11e-12 & 1.82 & 0.035 + & rh500492n00.2 & 50722 & 22.8 & 6.9d & 53.9 & 5.36 & 0.88 & 432.8 & 2.57 & 1.58e-12 & 2.6 & 0.58 + & rh500550n00.1 & 50893 & 23.9 & 28.4d & 46.1 & 5.30 & 0.88 & 437.3 & 3.16 & 1.51e-12 & 2.48 & 1.2e-13 + ngc1316-ulx1 & rh600255n00.2 & 49367 & 11.0 & 3.6h & 2.5 & 3.55 & 0.92 & 10.8 & 24.72 & 6.85e-14 & 3.79 & + & rh600255a01.5 + & 49540 & 29.4 & 3.2d & 5.4 & 3.52 & 0.97 & 33.5 & 23.84 & 8.04e-14 & 4.45 & + ngc1316-ulx2 & rh600255n00.5 & 49367 & 11.0 & 3.6h & 4.8 & 4.14 & 0.97 & 27.4 & 26.06 & 1.75e-13 & 9.69 & + & rh600255a01.8 + & 49540 & 29.4 & 3.2d & 7.4 & 4.09 & 0.97 & 63.8 & 16.90 & 1.53e-13 & 8.47 & 0.66 + ngc1316-ulx3 & rh600255n00.u & 49367 & 11.0 & 3.6h & @xmath203 & & & @xmath2013.8 & & @xmath208.78e-14 & @xmath204.86 & + & rh600255a01.4 + & 49540 & 29.4 & 3.2d & 4.7 & 4.72 & 0.98 & 34.1 & 26.39 & 8.18e-14 & 4.53 & + ngc1316-ulx4 & rh600255n00.u & 49367 & 11.0 & 3.6h & @xmath203 & & & @xmath2014.1 & & @xmath208.98e-14 & @xmath204.97 & + & rh600255a01.10 + & 49540 & 29.4 & 3.2d & 3.5 & 6.66 & 0.99 & 22.9 & 26.05 & 5.50e-14 & 3.05 & + ngc1316-ulx5 & rh600255n00.u & 49367 & 11.0 & 3.6h & @xmath203 & & & @xmath2014.3 & & @xmath209.14e-14 & @xmath205.06 & + & rh600255a01.9 + & 49540 & 29.4 & 3.2d & 3.5 & 7.48 & 1.00 & 20.8 & 26.05 & 4.98e-14 & 2.76 & + ngc1316-ulx6 & rh600255n00.u & 49367 & 11.0 & 3.6h & @xmath203 & & & @xmath2014.5 & & @xmath209.26e-14 & @xmath205.13 & + & rh600255a01.3 + & 49540 & 29.4 & 3.2d & 4.3 & 7.96 & 1.01 & 26.8 & 26.40 & 6.42e-14 & 3.55 & + ngc1316-ulx7 & rh600255n00.u & 49367 & 11.0 & 3.6h & @xmath203 & & & @xmath2016.1 & & @xmath201.02e-13 & @xmath205.65 & + & rh600255a01.2 + & 49540 & 29.4 & 3.2d & 4.6 & 9.90 & 1.02 & 31.0 & 26.39 & 7.42e-14 & 4.11 & + ngc1365-ulx1 & rh701297n00.2 & 49554 & 9.8 & 14.7d & 5.9 & 1.29 & 0.97 & 30.0 & 22.05 & 2.01e-13 & 7.29 & + & rh701297a02.u+ & 49902 & 9.8 & 1.1d & @xmath203 & & & @xmath2011.5 & & @xmath207.78e-14 & @xmath202.82 & + ngc1365-ulx2 & rh701297n00.3 & 49554 & 9.8 & 14.7d & 6.2 & 5.86 & 0.99 & 24.5 & 20.68 & 1.65e-13 & 5.98 & + & rh701297a02.4 + & 49902 & 9.8 & 1.1d & 5.7 & 5.83 & 0.99 & 20.0 & 22.51 & 1.35e-13 & 4.89 & + ngc1365-ulx3 & rh701297n00.5 & 49554 & 9.8 & 14.7d & 5.7 & 5.87 & 0.99 & 26.0 & 22.56 & 1.74e-13 & 6.31 & + & rh701297a02.u+ & 49902 & 9.8 & 1.1d & @xmath203 & & & @xmath2011.4 & & @xmath207.69e-14 & @xmath202.79 & + ngc1365-x5 & rh701297n00.u & 49554 & 9.8 & 14.7d & @xmath203 & & & @xmath2012.5 & & @xmath208.39e-14 & @xmath203.04 & + & rh701297a02.2 + & 49902 & 9.8 & 1.1d & 9.1 & 7.77 & 0.98 & 38.0 & 13.84 & 2.57e-13 & 9.32 & + ngc1380-ulx1 & rh600676n00.7 + & 49924 & 42.1 & 1.6d & 3.0 & 0.57 & 0.98 & 24.8 & 25.51 & 3.83e-14 & 1.43 & + ngc1395-x2 & rh600492n00.u & 49386 & 17.6 & 3.2d & @xmath203 & & & @xmath2014.9 & & @xmath206.02e-14 & @xmath204.2 & + & rh600492a01.4 + & 49553 & 19.9 & 34.8d & 3.9 & 2.65 & 0.95 & 19.3 & 26.42 & 6.89e-14 & 4.8 & + ngc1395-x3 & rh600492n00.u & 49386 & 17.6 & 3.2d & @xmath203 & & & @xmath2014.8 & & @xmath205.99e-14 & @xmath204.17 & + & rh600492a01.1 + & 49553 & 19.9 & 34.8d & 3.5 & 3.32 & 0.94 & 18.1 & 26.10 & 6.47e-14 & 4.51 & + ngc1395-x4 & rh600492n00.u & 49386 & 17.6 & 3.2d & @xmath203 & & & @xmath2014.8 & & @xmath205.97e-14 & @xmath204.16 & + & rh600492a01.5 + & 49553 & 19.9 & 34.8d & 4.1 & 4.17 & 0.96 & 24.6 & 26.41 & 8.80e-14 & 6.13 & + ngc1395-x5 & rh600492n00.2 & 49386 & 17.6 & 3.2d & 4.2 & 5.25 & 0.99 & 22.8 & 26.41 & 9.25e-14 & 6.45 & + & rh600492a01.u+ & 49553 & 19.9 & 34.8d & @xmath203 & & & @xmath2015.3 & & @xmath205.46e-14 & @xmath203.81 & + ngc1398-x2 & rh500518n00.5 + & 50650 & 17.2 & 37.5d & 4.6 & 3.34 & 0.96 & 26.1 & 26.39 & 9.32e-14 & 5.51 & + ngc1398-x3 & rh500518n00.3 + & 50650 & 17.2 & 37.5d & 4.9 & 3.89 & 0.97 & 28.0 & 25.79 & 9.99e-14 & 5.91 & + ngc1399-ulx1 & rh600256n00.u & 49035 & 7.3 & 7.1h & @xmath203 & & & @xmath2012.2 & & @xmath201.10e-13 & @xmath205.25 & + & rh600220n00.u & 49035 & 8.5 & 3.4d & @xmath203 & & & @xmath2013.0 & & @xmath201.01e-13 & @xmath204.82 & + & rh600831n00.14 & 50087 & 72.7 & 49.1d & 8.2 & 0.38 & 0.96 & 148.8 & 15.26 & 1.34e-13 & 6.4 & 0.21 + & rh600831a01.10 + & 50272 & 87.6 & 49.4d & 10.6 & 0.35 & 0.94 & 215.5 & 11.87 & 1.61e-13 & 7.69 & 0.64 + & rh600940n00.u & 50475 & 63.1 & 7.4d & @xmath203 & & & @xmath2033.9 & & @xmath203.53e-14 & @xmath201.69 & + ngc1399-ulx2 & rh600256n00.u & 49035 & 7.3 & 7.1h & @xmath203 & & & @xmath2012.2 & & @xmath201.10e-13 & @xmath205.25 & + & rh600220n00.u & 49035 & 8.5 & 3.4d & @xmath203 & & & @xmath2014.0 & & @xmath201.08e-13 & @xmath205.16 & + & rh600831n00.17 & 50087 & 72.7 & 49.1d & 7.3 & 0.88 & 0.97 & 136.1 & 17.28 & 1.23e-13 & 5.87 & 0.84 + & rh600831a01.11 + & 50272 & 87.6 & 49.4d & 8.5 & 0.97 & 0.96 & 163.0 & 14.74 & 1.22e-13 & 5.83 & 0.4 + & rh600940n00.12 & 50475 & 63.1 & 7.4d & 6.8 & 10.34 & 1.00 & 109.0 & 18.53 & 1.13e-13 & 5.4 & 0.35 + ngc1399-ulx3 & rh600256n00.u & 49035 & 7.3 & 7.1h & @xmath203 & & & @xmath2012.1 & & @xmath201.09e-13 & @xmath205.21 & + & rh600220n00.u & 49035 & 8.5 & 3.4d & @xmath203 & & & @xmath2016.0 & & @xmath201.24e-13 & @xmath205.92 & + & rh600831n00.21 & 50087 & 72.7 & 49.1d & 3.9 & 2.02 & 0.95 & 54.7 & 26.41 & 4.93e-14 & 2.35 & 0.83 + & rh600831a01.u+ & 50272 & 87.6 & 49.4d & @xmath203 & & & @xmath2033.9 & & @xmath202.53e-14 & @xmath201.21 & + & rh600940n00.u & 50475 & 63.1 & 7.4d & @xmath203 & & & @xmath2043.9 & & @xmath204.58e-14 & @xmath202.19 & + ngc1399-x5 & rh600256n00.u & 49035 & 7.3 & 7.1h & @xmath203 & & & @xmath2011.9 & & @xmath201.07e-13 & @xmath205.11 & + & rh600220n00.u & 49035 & 8.5 & 3.4d & @xmath203 & & & @xmath2011.9 & & @xmath209.22e-14 & @xmath204.4 & + & rh600831n00.16 & 50087 & 72.7 & 49.1d & 3.4 & 4.01 & 0.95 & 37.6 & 25.93 & 3.39e-14 & 1.62 & 0.48 + & rh600831a01.12 + & 50272 & 87.6 & 49.4d & 4.7 & 4.01 & 0.97 & 54.8 & 26.39 & 4.10e-14 & 1.96 & 0.46 + & rh600940n00.11 & 50475 & 63.1 & 7.4d & 3.1 & 6.25 & 0.99 & 33.9 & 25.52 & 3.54e-14 & 1.69 & 0.016 + ngc1399-x6 & rh600256n00.u & 49035 & 7.3 & 7.1h & @xmath203 & & & @xmath2011.9 & & @xmath201.07e-13 & @xmath205.11 & + & rh600220n00.u & 49035 & 8.5 & 3.4d & @xmath203 & & & @xmath2024.5 & & @xmath201.90e-13 & @xmath209.07 & + & rh600831n00.20 & 50087 & 72.7 & 49.1d & 3.1 & 4.45 & 0.95 & 30.7 & 25.58 & 2.77e-14 & 1.32 & + & rh600831a01.u+ & 50272 & 87.6 & 49.4d & @xmath203 & & & @xmath2033.3 & & @xmath202.49e-14 & @xmath201.19 & + & rh600940n00.u & 50475 & 63.1 & 7.4d & @xmath203 & & & @xmath2071.2 & & @xmath207.42e-14 & @xmath203.54 & + ngc1399-x7 & rh600256n00.u & 49035 & 7.3 & 7.1h & @xmath203 & & & @xmath2011.9 & & @xmath201.07e-13 & @xmath205.11 & + & rh600220n00.u & 49035 & 8.5 & 3.4d & @xmath203 & & & @xmath2012.0 & & @xmath209.32e-14 & @xmath204.45 & + & rh600831n00.u & 50087 & 72.7 & 49.1d & @xmath203 & & & @xmath2029.8 & & @xmath202.69e-14 & @xmath201.28 & + & rh600831a01.7 + & 50272 & 87.6 & 49.4d & 3.3 & 3.96 & 0.94 & 42.4 & 25.81 & 3.17e-14 & 1.51 & 0.31 + & rh600940n00.u & 50475 & 63.1 & 7.4d & @xmath203 & & & @xmath2030.7 & & @xmath203.20e-14 & @xmath201.53 & + ngc1399-x8 & rh600256n00.u & 49035 & 7.3 & 7.1h & @xmath203 & & & @xmath2011.9 & & @xmath201.07e-13 & @xmath205.11 & + & rh600220n00.u & 49035 & 8.5 & 3.4d & @xmath203 & & & @xmath2012.6 & & @xmath209.81e-14 & @xmath204.69 & + & rh600831n00.11 & 50087 & 72.7 & 49.1d & 10.2 & 4.17 & 0.95 & 128.6 & 12.42 & 1.16e-13 & 5.54 & 0.0012 + & rh600831a01.6 + & 50272 & 87.6 & 49.4d & 14.1 & 4.14 & 0.92 & 252.0 & 9.23 & 1.89e-13 & 9.03 & 0.019 + & rh600940n00.7 & 50475 & 63.1 & 7.4d & 8.7 & 8.75 & 0.99 & 114.4 & 14.30 & 1.19e-13 & 5.68 & 0.24 + ngc1399-x9 & rh600256n00.2 & 49035 & 7.3 & 7.1h & 16.1 & 5.03 & 0.92 & 82.6 & 8.34 & 7.45e-13 & 35.6 & 0.28 + & rh600220n00.3 & 49035 & 8.5 & 3.4d & 20.5 & 4.52 & 0.90 & 114.1 & 6.96 & 8.85e-13 & 42.3 & 0.69 + & rh600831n00.13 & 50087 & 72.7 & 49.1d & 21.3 & 5.02 & 0.90 & 339.3 & 6.83 & 3.06e-13 & 14.6 & 2.6e-05 + & rh600831a01.9 + & 50272 & 87.6 & 49.4d & 44.3 & 5.07 & 0.88 & 786.4 & 3.35 & 5.88e-13 & 28.1 & 0.95 + & rh600940n00.9 & 50475 & 63.1 & 7.4d & 21.4 & 4.52 & 0.90 & 271.8 & 6.81 & 2.83e-13 & 13.5 & 0.43 + ngc1399-x10 & rh600256n00.u & 49035 & 7.3 & 7.1h & @xmath203 & & & @xmath2012.2 & & @xmath201.10e-13 & @xmath205.25 & + & rh600220n00.u & 49035 & 8.5 & 3.4d & @xmath203 & & & @xmath2011.7 & & @xmath209.10e-14 & @xmath204.35 & + & rh600831n00.15 & 50087 & 72.7 & 49.1d & 3.4 & 6.76 & 1.00 & 34.9 & 26.01 & 3.14e-14 & 1.5 & + & rh600831a01.u+ & 50272 & 87.6 & 49.4d & @xmath203 & & & @xmath2034.3 & & @xmath202.57e-14 & @xmath201.23 & + & rh600940n00.u & 50475 & 63.1 & 7.4d & @xmath203 & & & @xmath2029.8 & & @xmath203.10e-14 & @xmath201.48 & + ngc1404-x2 & rh600256n00.u & 49035 & 7.3 & 7.1h & @xmath203 & & & @xmath2012.3 & & @xmath201.11e-13 & @xmath205.87 & + & rh600220n00.u & 49035 & 8.5 & 3.4d & @xmath203 & & & @xmath2011.8 & & @xmath209.17e-14 & @xmath204.85 & + & rh600831n00.12 & 50087 & 72.7 & 49.1d & 4.2 & 7.29 & 1.01 & 48.7 & 26.41 & 4.39e-14 & 2.32 & + & rh600831a01.8 + & 50272 & 87.6 & 49.4d & 4.3 & 7.32 & 1.01 & 48.7 & 26.40 & 3.64e-14 & 1.92 & + & rh600940n00.8 & 50475 & 63.1 & 7.4d & 5.7 & 2.54 & 0.96 & 66.2 & 22.49 & 6.90e-14 & 3.65 & 0.15 + ngc1427a - ulx1 & rh600220n00.u & 49035 & 8.5 & 3.4d & @xmath203 & & & @xmath2040.0 & & @xmath203.11e-13 & @xmath2010.7 & + & rh600940n00.18 & 50475 & 63.1 & 7.4d & 8.6 & 16.24 & 0.53 & 246.8 & 14.62 & 2.57e-13 & 8.81 & 0.27 + pgc13826-ulx1 & rh600022n00.3 + & 48300 & 19.0 & 1.9d & 10.6 & 0.76 & 0.94 & 72.8 & 11.77 & 6.64e-13 & 1.21 & 0.91 + pgc13826-ulx2 & rh600022n00.5 + & 48300 & 19.0 & 1.9d & 13.7 & 3.87 & 0.92 & 89.2 & 9.46 & 8.14e-13 & 1.49 & 1.3e-06 + pgc13826-ulx3 & rh600022n00.8 + & 48300 & 19.0 & 1.9d & 21.5 & 5.22 & 0.90 & 166.8 & 6.79 & 1.52e-12 & 2.77 & 0.63 + pgc13826-x10 & rh600022n00.10 + & 48300 & 19.0 & 1.9d & 9.7 & 15.05 & 0.70 & 133.7 & 13.18 & 1.22e-12 & 2.23 & 0.83 + ngc1549-x2 & rh600490n00.u & 49365 & 3.5 & 1.1d & @xmath203 & & & @xmath2010.3 & & @xmath201.97e-13 & @xmath209.16 & + & rh600479n00.u+ & 49518 & 46.1 & 72.3d & @xmath203 & & & @xmath2049.4 & & @xmath207.17e-14 & @xmath203.33 & + & rh600490a01.16 & 49527 & 39.3 & 61.5d & 4.3 & 8.92 & 1.02 & 38.0 & 26.40 & 6.47e-14 & 3.01 & + ngc1553-ulx1 & rh600490n00.u & 49365 & 3.5 & 1.1d & @xmath203 & & & @xmath209.6 & & @xmath201.84e-13 & @xmath207.59 & + & rh600479n00.12 + & 49518 & 46.1 & 72.3d & 5.1 & 0.84 & 0.98 & 43.5 & 24.89 & 6.31e-14 & 2.6 & 0.74 + & rh600490a01.u & 49527 & 39.3 & 61.5d & @xmath203 & & & @xmath2020.1 & & @xmath203.42e-14 & @xmath201.41 & + ngc1553-ulx2 & rh600490n00.u & 49365 & 3.5 & 1.1d & @xmath203 & & & @xmath209.7 & & @xmath201.86e-13 & @xmath207.67 & + & rh600479n00.u+ & 49518 & 46.1 & 72.3d & @xmath203 & & & @xmath2021.2 & & @xmath203.08e-14 & @xmath201.27 & + & rh600490a01.6 & 49527 & 39.3 & 61.5d & 3.0 & 6.54 & 0.99 & 19.8 & 25.47 & 3.37e-14 & 1.39 & + ngc1559-ulx1 & rh600717n00.10 + & 49992 & 17.5 & 7.2d & 5.8 & 1.03 & 0.97 & 38.6 & 22.14 & 1.68e-13 & 3.73 & 0.12 + ngc1559-ulx2 & rh600717n00.11 + & 49992 & 17.5 & 7.2d & 5.8 & 0.61 & 0.98 & 30.7 & 22.31 & 1.34e-13 & 2.97 & + ngc1559-x3 & rh600717n00.3 + & 49992 & 17.5 & 7.2d & 6.0 & 1.30 & 0.97 & 30.0 & 21.38 & 1.30e-13 & 2.89 & + ngc1566-ulx1 & rh700024n00.u & 48631 & 9.3 & 21.3h & @xmath203 & & & @xmath2011.5 & & @xmath208.13e-14 & @xmath201.75 & + & rh701668n00.u & 49988 & 2.1 & 0.7h & @xmath203 & & & @xmath209.6 & & @xmath203.00e-13 & @xmath206.46 & + & rh600860n00.4 & 50320 & 15.0 & 17.6d & 6.0 & 0.58 & 0.97 & 37.0 & 21.48 & 1.61e-13 & 3.47 & 0.5 + & rh600962n00.7 + & 50469 & 40.4 & 5.1d & 12.6 & 0.62 & 0.93 & 141.7 & 10.14 & 2.30e-13 & 4.96 & 0.19 + ngc1566-ulx2 & rh700024n00.u & 48631 & 9.3 & 21.3h & @xmath203 & & & @xmath2011.5 & & @xmath208.10e-14 & @xmath201.75 & + & rh701668n00.u & 49988 & 2.1 & 0.7h & @xmath203 & & & @xmath209.6 & & @xmath202.99e-13 & @xmath206.44 & + & rh600860n00.u & 50320 & 15.0 & 17.6d & @xmath203 & & & @xmath2014.4 & & @xmath206.27e-14 & @xmath201.35 & + & rh600962n00.6 + & 50469 & 40.4 & 5.1d & 7.4 & 1.15 & 0.96 & 97.9 & 17.08 & 1.59e-13 & 3.43 & 0.24 + ngc1566-ulx3 & rh700024n00.u & 48631 & 9.3 & 21.3h & @xmath203 & & & @xmath2011.4 & & @xmath208.08e-14 & @xmath201.74 & + & rh701668n00.u & 49988 & 2.1 & 0.7h & @xmath203 & & & @xmath209.6 & & @xmath202.98e-13 & @xmath206.42 & + & rh600860n00.2 & 50320 & 15.0 & 17.6d & 3.2 & 1.54 & 0.94 & 14.5 & 25.65 & 6.33e-14 & 1.36 & + & rh600962n00.4 + & 50469 & 40.4 & 5.1d & 9.6 & 1.45 & 0.95 & 103.2 & 13.26 & 1.67e-13 & 3.6 & 0.0083 + ngc1566-x5 & rh700024n00.u & 48631 & 9.3 & 21.3h & @xmath203 & & & @xmath2011.2 & & @xmath207.93e-14 & @xmath201.71 & + & rh701668n00.u & 49988 & 2.1 & 0.7h & @xmath203 & & & @xmath209.4 & & @xmath202.93e-13 & @xmath206.31 & + & rh600860n00.u & 50320 & 15.0 & 17.6d & @xmath203 & & & @xmath2014.0 & & @xmath206.11e-14 & @xmath201.32 & + & rh600962n00.3 + & 50469 & 40.4 & 5.1d & 4.5 & 4.14 & 0.97 & 33.5 & 26.40 & 5.43e-14 & 1.17 & + ngc1672-ulx1 & rh701022n00.5 & 48798 & 24.3 & 78.8d & 4.8 & 1.18 & 0.97 & 29.2 & 26.06 & 8.82e-14 & 2.23 & + & rh601011n00.5 + & 50476 & 34.6 & 33.6d & 7.5 & 1.22 & 0.96 & 70.2 & 16.66 & 1.49e-13 & 3.76 & 0.98 + ngc1672-ulx2 & rh701022n00.3 & 48798 & 24.3 & 78.8d & 7.0 & 1.30 & 0.96 & 44.8 & 18.02 & 1.36e-13 & 3.43 & + & rh601011n00.2 + & 50476 & 34.6 & 33.6d & 8.2 & 1.20 & 0.96 & 66.1 & 15.24 & 1.40e-13 & 3.53 & 0.85 + ngc1672-x4 & rh701022n00.1 & 48798 & 24.3 & 78.8d & 3.8 & 5.12 & 0.98 & 20.4 & 26.42 & 6.18e-14 & 1.56 & + & rh601011n00.u+ & 50476 & 34.6 & 33.6d & @xmath203 & & & @xmath2021.1 & & @xmath204.47e-14 & @xmath201.13 & + ngc1672-x5 & rh701022n00.u & 48798 & 24.3 & 78.8d & @xmath203 & & & @xmath2016.7 & & @xmath205.07e-14 & @xmath201.28 & + & rh601011n00.6 + & 50476 & 34.6 & 33.6d & 4.7 & 5.80 & 1.00 & 37.3 & 26.39 & 7.91e-14 & 2 & + ngc1792-ulx1 & rh600433n00.u+ & 48878 & 25.9 & 15.6h & @xmath203 & & & @xmath2018.7 & & @xmath205.46e-14 & @xmath200.668 & + & rh600584n00.2 & 49607 & 7.8 & 3.8h & 5.0 & 0.85 & 0.98 & 25.5 & 25.26 & 2.48e-13 & 3.04 & + ngc1792-ulx2 & rh600433n00.3 + & 48878 & 25.9 & 15.6h & 6.4 & 0.18 & 0.97 & 42.8 & 19.95 & 1.25e-13 & 1.53 & + & rh600584n00.u & 49607 & 7.8 & 3.8h & @xmath203 & & & @xmath2012.4 & & @xmath201.21e-13 & @xmath201.48 & + ngc2273-x2 & rh601000n00.4 + & 50531 & 34.2 & 7.9d & 10.4 & 2.38 & 0.94 & 86.2 & 12.09 & 2.44e-13 & 23.6 & 0.18 + & rh601000a01.u & 50882 & 3.4 & 8.4h & @xmath203 & & & @xmath208.8 & & @xmath202.48e-13 & @xmath2024 & + ngc2273-x3 & rh601000n00.2 + & 50531 & 34.2 & 7.9d & 5.9 & 2.60 & 0.96 & 39.9 & 22.02 & 1.13e-13 & 10.9 & + & rh601000a01.u & 50882 & 3.4 & 8.4h & @xmath203 & & & @xmath208.8 & & @xmath202.48e-13 & @xmath2024 & + ngc2276-ulx1 & rh600498n00.5 + & 49429 & 51.7 & 3.9d & 7.9 & 0.46 & 0.96 & 83.8 & 15.83 & 1.46e-13 & 23.7 & 0.74 + & rh600498a01.4 & 49593 & 21.5 & 1.3d & 5.8 & 0.51 & 0.98 & 49.8 & 22.47 & 2.08e-13 & 33.8 & 0.86 + ngc2276-x2 & rh600498n00.6 + & 49429 & 51.7 & 3.9d & 9.7 & 2.37 & 0.95 & 97.2 & 13.14 & 1.69e-13 & 27.5 & 0.17 + & rh600498a01.3 & 49593 & 21.5 & 1.3d & 3.9 & 2.37 & 0.95 & 27.2 & 26.41 & 1.14e-13 & 18.5 & + ngc2403-ulx1 & rh600767n00.4 + & 49978 & 26.2 & 4.1d & 32.5 & 3.02 & 0.88 & 368.3 & 4.70 & 1.19e-12 & 1.4 & 0.36 + ngc2403-x5 & rh600767n00.6 + & 49978 & 26.2 & 4.1d & 44.3 & 11.79 & 0.94 & 603.0 & 3.35 & 1.94e-12 & 2.29 & 0.055 + ngc2442-x2 & rh600953n00.3 + & 50770 & 40.4 & 23.6d & 4.8 & 3.70 & 0.96 & 37.7 & 26.31 & 1.06e-13 & 3.72 & + pgc23324-ulx1 & rh600745n00.1 + & 49643 & 7.8 & 6.6d & 123.5 & 0.31 & 0.87 & 1122.0 & 1.05 & 1.17e-11 & 28.4 & 0.0092 + ngc2775-ulx1 & rh500385n00.u & 49832 & 10.8 & 22.9h & @xmath203 & & & @xmath2013.7 & & @xmath201.07e-13 & @xmath203.71 & + & rh600826n00.u & 50046 & 3.1 & 2.3h & @xmath203 & & & @xmath208.6 & & @xmath202.36e-13 & @xmath208.18 & + & rh600826a01.8 + & 50198 & 48.1 & 16.3d & 3.7 & 1.91 & 0.94 & 30.0 & 26.38 & 5.27e-14 & 1.83 & + ngc2775-x4 & rh500385n00.1 & 49832 & 10.8 & 22.9h & 2.8 & 3.85 & 0.93 & 13.2 & 25.18 & 1.03e-13 & & + & rh600826n00.u & 50046 & 3.1 & 2.3h & @xmath203 & & & @xmath208.6 & & @xmath202.34e-13 & @xmath208.12 & + & rh600826a01.4 + & 50198 & 48.1 & 16.3d & 5.9 & 2.94 & 0.96 & 48.9 & 21.99 & 8.57e-14 & 2.97 & + ngc2775-x5 & rh500385n00.2 & 49832 & 10.8 & 22.9h & 4.7 & 3.35 & 0.96 & 20.8 & 26.39 & 1.63e-13 & 5.65 & + & rh600826n00.u & 50046 & 3.1 & 2.3h & @xmath203 & & & @xmath208.6 & & @xmath202.34e-13 & @xmath208.12 & + & rh600826a01.5 + & 50198 & 48.1 & 16.3d & 7.9 & 3.51 & 0.96 & 74.1 & 15.78 & 1.30e-13 & 4.51 & 0.64 + ngc2775-x6 & rh500385n00.u & 49832 & 10.8 & 22.9h & @xmath203 & & & @xmath2013.4 & & @xmath201.05e-13 & @xmath203.64 & + & rh600826n00.u & 50046 & 3.1 & 2.3h & @xmath203 & & & @xmath208.6 & & @xmath202.34e-13 & @xmath208.12 & + & rh600826a01.6 + & 50198 & 48.1 & 16.3d & 4.9 & 3.63 & 0.97 & 39.8 & 25.78 & 6.98e-14 & 2.42 & + ngc2782-ulx1 & rh700462n00.3 + & 48731 & 21.5 & 1.5d & 5.8 & 0.50 & 0.98 & 36.0 & 22.17 & 1.16e-13 & 19.4 & + ngc2903-ulx1 & rh600602n00.u+ & 49470 & 13.5 & 1.1d & @xmath203 & & & @xmath2013.8 & & @xmath208.08e-14 & @xmath200.531 & + & rh600602a01.2 & 49663 & 8.2 & 8.2h & 4.7 & 2.12 & 0.96 & 23.8 & 26.39 & 2.29e-13 & 1.5 & + ngc3031-ulx1 & rh600247n00.u+ & 48919 & 26.3 & 3.7d & @xmath203 & & & @xmath2017.8 & & @xmath205.73e-14 & @xmath200.0804 & + & rh180015n00.u & 49094 & 1.7 & 1.3h & @xmath203 & & & @xmath207.6 & & @xmath203.80e-13 & @xmath200.533 & + & rh600247a01.12 & 49095 & 21.1 & 26.4d & 44.2 & 3.08 & 0.88 & 463.5 & 3.36 & 1.86e-12 & 2.61 & 0.096 + & rh600739n00.12 & 49644 & 19.9 & 2.1d & 20.6 & 3.12 & 0.90 & 159.9 & 6.94 & 6.81e-13 & 0.956 & 0.64 + & rh600740n00.7 & 49821 & 19.0 & 20.9d & 18.7 & 3.09 & 0.90 & 172.1 & 7.40 & 7.68e-13 & 1.08 & 0.34 + & rh600881n00.6 & 50003 & 14.8 & 12.8d & 24.9 & 3.15 & 0.89 & 187.8 & 6.09 & 1.07e-12 & 1.5 & 0.31 + & rh600882n00.6 & 50189 & 18.3 & 21.7d & 25.9 & 3.09 & 0.89 & 220.0 & 6.01 & 1.02e-12 & 1.43 & 0.41 + & rh600882a01.5 & 50384 & 5.1 & 13.2d & 11.6 & 3.16 & 0.93 & 51.0 & 10.89 & 8.47e-13 & 1.19 & + & rh601001n00.8 & 50537 & 19.2 & 2.5d & 26.1 & 3.07 & 0.89 & 214.6 & 5.95 & 9.45e-13 & 1.33 & 0.72 + & rh601002n00.7 & 50721 & 19.6 & 15.4d & 24.5 & 3.15 & 0.89 & 202.6 & 6.18 & 8.77e-13 & 1.23 & 0.39 + & rh601095n00.5 & 50898 & 12.4 & 17.7h & 12.3 & 3.10 & 0.93 & 92.3 & 10.43 & 6.28e-13 & 0.881 & 0.69 + ngc3031-ulx2 & rh600247n00.7 + & 48919 & 26.3 & 3.7d & 64.7 & 3.70 & 0.87 & 736.2 & 2.01 & 2.37e-12 & 3.33 & 0.42 + & rh180015n00.u & 49094 & 1.7 & 1.3h & @xmath203 & & & @xmath207.6 & & @xmath203.80e-13 & @xmath200.533 & + & rh600247a01.10 & 49095 & 21.1 & 26.4d & 58.4 & 3.65 & 0.87 & 562.4 & 2.30 & 2.26e-12 & 3.17 & 0.99 + & rh600739n00.11 & 49644 & 19.9 & 2.1d & 56.8 & 3.73 & 0.88 & 564.4 & 2.42 & 2.40e-12 & 3.37 & 0.56 + & rh600740n00.5 & 49821 & 19.0 & 20.9d & 48.1 & 3.67 & 0.88 & 508.8 & 2.95 & 2.27e-12 & 3.19 & 0.13 + & rh600881n00.5 & 50003 & 14.8 & 12.8d & 50.7 & 3.73 & 0.88 & 369.8 & 2.74 & 2.11e-12 & 2.96 & 0.51 + & rh600882n00.5 & 50189 & 18.3 & 21.7d & 48.1 & 3.65 & 0.88 & 430.1 & 2.95 & 1.99e-12 & 2.79 & 0.56 + & rh600882a01.3 & 50384 & 5.1 & 13.2d & 23.5 & 3.74 & 0.89 & 120.3 & 6.49 & 2.00e-12 & 2.81 & 0.033 + & rh601001n00.7 & 50537 & 19.2 & 2.5d & 55.9 & 3.63 & 0.88 & 511.4 & 2.44 & 2.25e-12 & 3.16 & 0.87 + & rh601002n00.5 & 50721 & 19.6 & 15.4d & 47.4 & 3.73 & 0.88 & 441.8 & 2.99 & 1.91e-12 & 2.68 & 0.026 + & rh601095n00.4 & 50898 & 12.4 & 17.7h & 24.9 & 3.66 & 0.89 & 203.7 & 6.09 & 1.39e-12 & 1.95 & 0.39 + pgc28757-ulx1 & rh600247n00.1 + & 48919 & 26.3 & 3.7d & 81.6 & 12.59 & 0.93 & 1205.1 & 1.47 & 3.88e-12 & 5.45 & 0.035 + & rh180015n00.u & 49094 & 1.7 & 1.3h & @xmath203 & & & @xmath2012.3 & & @xmath206.16e-13 & @xmath200.865 & + & rh600247a01.1 & 49095 & 21.1 & 26.4d & 97.4 & 12.55 & 0.93 & 1416.9 & 1.31 & 5.70e-12 & 8 & 0.00026 + & rh600739n00.1 & 49644 & 19.9 & 2.1d & 59.7 & 12.63 & 0.94 & 716.1 & 2.25 & 3.05e-12 & 4.28 & 0.37 + & rh600740n00.1 & 49821 & 19.0 & 20.9d & 80.1 & 12.56 & 0.93 & 1129.9 & 1.43 & 5.04e-12 & 7.07 & 0.13 + & rh600881n00.1 & 50003 & 14.8 & 12.8d & 69.1 & 12.59 & 0.93 & 717.1 & 1.77 & 4.10e-12 & 5.75 & 0.89 + & rh600882n00.1 & 50189 & 18.3 & 21.7d & 82.1 & 12.54 & 0.93 & 1172.0 & 1.46 & 5.42e-12 & 7.61 & 0.48 + & rh600882a01.1 & 50384 & 5.1 & 13.2d & 64.7 & 12.61 & 0.93 & 562.9 & 2.01 & 9.37e-12 & 13.2 & 0.68 + & rh601001n00.1 & 50537 & 19.2 & 2.5d & 131.8 & 12.57 & 0.93 & 2170.9 & 1.11 & 9.56e-12 & 13.4 & 0.41 + & rh601002n00.1 & 50721 & 19.6 & 15.4d & 60.8 & 12.57 & 0.94 & 789.5 & 2.21 & 3.42e-12 & 4.8 & 0.71 + & rh601095n00.1 & 50898 & 12.4 & 17.7h & 48.1 & 12.55 & 0.94 & 568.7 & 2.95 & 3.87e-12 & 5.43 & 0.065 + ngc3226-x3 & rh701299n00.5 + & 49478 & 30.0 & 5.6d & 5.1 & 4.89 & 0.98 & 34.0 & 25.01 & 8.21e-14 & 5.46 & + & rh701870n00.u & 49869 & 7.5 & 1.7d & @xmath203 & & & @xmath2011.0 & & @xmath201.07e-13 & @xmath207.12 & + ngc3294-x1 & rh500389n00.5 + & 49860 & 11.2 & 16.3h & 3.5 & 1.39 & 0.95 & 18.0 & 26.15 & 1.10e-13 & 9.41 & + pgc31586-x2 & rh800741n00.4 + & 49866 & 32.4 & 15.7d & 26.9 & 1.47 & 0.89 & 279.7 & 5.71 & 7.70e-13 & & 0.0012 + ngc3310-ulx1 & rh600685n00.u & 49680 & 4.1 & 6.6h & @xmath203 & & & @xmath208.4 & & @xmath201.31e-13 & @xmath205.5 & + & rh600685a01.7 + & 49826 & 41.5 & 27.9d & 10.2 & 0.75 & 0.95 & 84.4 & 12.49 & 1.29e-13 & 5.41 & 0.94 + ngc3310-ulx2 & rh600685n00.u & 49680 & 4.1 & 6.6h & @xmath203 & & & @xmath208.4 & & @xmath201.31e-13 & @xmath205.5 & + & rh600685a01.9 + & 49826 & 41.5 & 27.9d & 8.9 & 1.18 & 0.95 & 94.2 & 14.04 & 1.44e-13 & 6.04 & 0.4 + ngc3396-x2 & rh600771n00.6 + & 49859 & 24.7 & 22.2h & 10.2 & 1.67 & 0.94 & 96.2 & 12.37 & 2.78e-13 & 15.9 & 0.12 + ngc3623-ulx1 & rh600827n00.u & 50249 & 4.3 & 1.1d & @xmath203 & & & @xmath209.5 & & @xmath201.60e-13 & @xmath202.9 & + & rh600827a01.7 + & 50426 & 36.6 & 6.4d & 6.3 & 0.85 & 0.97 & 48.7 & 20.41 & 9.66e-14 & 1.75 & + ngc3623-x3 & rh600827n00.u & 50249 & 4.3 & 1.1d & @xmath203 & & & @xmath209.4 & & @xmath201.59e-13 & @xmath202.89 & + & rh600827a01.9 + & 50426 & 36.6 & 6.4d & 4.5 & 1.96 & 0.95 & 29.8 & 26.40 & 5.92e-14 & 1.07 & + ngc3627-ulx1 & rh600836a01.4 + & 50423 & 9.4 & 1.5d & 6.3 & 1.80 & 0.96 & 28.1 & 20.36 & 2.24e-13 & 2.06 & + ngc3627-x3 & rh600836a01.3 + & 50423 & 9.4 & 1.5d & 6.9 & 6.95 & 0.99 & 34.9 & 18.23 & 2.78e-13 & 2.55 & + ngc3628-ulx1 & rh700009n00.3 + & 48598 & 13.5 & 2.2d & 7.9 & 4.99 & 0.97 & 39.8 & 15.87 & 2.15e-13 & 2.9 & + & rh700009a01.1 & 49495 & 5.3 & 1.3d & 3.3 & 5.19 & 0.97 & 12.7 & 25.85 & 1.77e-13 & 2.39 & + ngc3628-x3 & rh700009n00.2 + & 48598 & 13.5 & 2.2d & 3.9 & 5.68 & 0.99 & 16.7 & 26.41 & 9.05e-14 & 1.22 & + & rh700009a01.u & 49495 & 5.3 & 1.3d & @xmath203 & & & @xmath2010.7 & & @xmath201.48e-13 & @xmath202 & + ngc3642-x2 & rh701003n00.1 + & 48918 & 9.9 & 2.9d & 3.6 & 3.51 & 0.95 & 14.4 & 26.21 & 8.71e-14 & 7.9 & + ngc3998-x2 & rh600430n00.3 + & 49121 & 13.5 & 29.2d & 4.3 & 1.18 & 0.97 & 25.2 & 26.40 & 1.20e-13 & 2.88 & + ngc4088-ulx1 & rh500391n00.1 + & 49700 & 9.9 & 10.0h & 4.9 & 0.43 & 0.98 & 23.5 & 25.91 & 1.69e-13 & 5.86 & + ngc4105-x2 & rh600975n00.6 + & 50618 & 81.1 & 8.1d & 6.9 & 2.41 & 0.96 & 68.5 & 18.30 & 7.49e-14 & 6.34 & 0.7 + ngc4136-ulx1 & rh702734n00.1 + & 50259 & 2.0 & 0.8h & 3.5 & 1.25 & 0.96 & 10.1 & 26.09 & 3.51e-13 & 3.96 & + ngc4151-ulx1 & rh700023n00.u & 49124 & 9.2 & 7.1h & @xmath203 & & & @xmath2012.3 & & @xmath209.59e-14 & @xmath204.74 & + & rh701707n00.20 + & 49856 & 107.2 & 13.8d & 4.8 & 2.53 & 0.96 & 56.0 & 26.26 & 3.72e-14 & 1.84 & 0.061 + & rh701692n00.u & 49870 & 1.5 & 0.7h & @xmath203 & & & @xmath207.6 & & @xmath203.62e-13 & @xmath2017.9 & + & rh701989n00.16 & 50254 & 102.4 & 3.9d & 3.4 & 2.52 & 0.94 & 42.1 & 25.98 & 2.93e-14 & 1.45 & 0.23 + ngc4151-ulx2 & rh700023n00.u & 49124 & 9.2 & 7.1h & @xmath203 & & & @xmath2012.3 & & @xmath209.55e-14 & @xmath204.72 & + & rh701707n00.9 + & 49856 & 107.2 & 13.8d & 4.8 & 3.05 & 0.96 & 55.1 & 26.15 & 3.66e-14 & 1.81 & 0.65 + & rh701692n00.u & 49870 & 1.5 & 0.7h & @xmath203 & & & @xmath207.6 & & @xmath203.60e-13 & @xmath2017.8 & + & rh701989n00.6 & 50254 & 102.4 & 3.9d & 5.4 & 3.03 & 0.96 & 62.1 & 24.00 & 4.31e-14 & 2.13 & 0.92 + ngc4151-ulx3 & rh700023n00.u & 49124 & 9.2 & 7.1h & @xmath203 & & & @xmath2012.3 & & @xmath209.52e-14 & @xmath204.71 & + & rh701707n00.21 + & 49856 & 107.2 & 13.8d & 4.4 & 3.52 & 0.96 & 49.4 & 26.40 & 3.28e-14 & 1.62 & + & rh701692n00.u & 49870 & 1.5 & 0.7h & @xmath203 & & & @xmath207.6 & & @xmath203.60e-13 & @xmath2017.8 & + & rh701989n00.u & 50254 & 102.4 & 3.9d & @xmath203 & & & @xmath2032.9 & & @xmath202.28e-14 & @xmath201.13 & + ngc4151-ulx4 & rh700023n00.4 & 49124 & 9.2 & 7.1h & 5.4 & 4.42 & 0.98 & 25.6 & 24.01 & 1.99e-13 & 9.84 & + & rh701707n00.u+ & 49856 & 107.2 & 13.8d & @xmath203 & & & @xmath2032.7 & & @xmath202.17e-14 & @xmath201.07 & + & rh701692n00.u & 49870 & 1.5 & 0.7h & @xmath203 & & & @xmath207.5 & & @xmath203.60e-13 & @xmath2017.8 & + & rh701989n00.18 & 50254 & 102.4 & 3.9d & 5.9 & 4.42 & 0.98 & 73.6 & 21.96 & 5.11e-14 & 2.53 & 0.43 + ngc4151-x6 & rh700023n00.u & 49124 & 9.2 & 7.1h & @xmath203 & & & @xmath2012.3 & & @xmath209.55e-14 & @xmath204.72 & + & rh701707n00.27 + & 49856 & 107.2 & 13.8d & 9.0 & 5.18 & 0.96 & 117.5 & 13.93 & 7.80e-14 & 3.86 & 0.024 + & rh701692n00.u & 49870 & 1.5 & 0.7h & @xmath203 & & & @xmath207.6 & & @xmath203.60e-13 & @xmath2017.8 & + & rh701989n00.22 & 50254 & 102.4 & 3.9d & 4.0 & 5.05 & 0.98 & 53.9 & 26.41 & 3.75e-14 & 1.85 & 0.18 + ngc4190-ulx1 & rh600733n00.2 & 49860 & 3.4 & 8.8d & 22.0 & 0.48 & 0.90 & 93.1 & 6.71 & 1.83e-12 & 1.72 & 0.65 + & rh600968n00.2 + & 50773 & 10.7 & 3.8d & 52.1 & 0.45 & 0.88 & 389.3 & 2.76 & 2.46e-12 & 2.31 & 0.049 + ngc4203-x2 & rh600221n00.4 + & 49150 & 25.2 & 2.2d & 98.7 & 2.26 & 0.87 & 1128.4 & 1.28 & 2.86e-12 & 78.7 & 0.3 + ngc4203-x3 & rh600221n00.3 + & 49150 & 25.2 & 2.2d & 4.9 & 2.89 & 0.96 & 32.9 & 25.91 & 8.35e-14 & 2.3 & + ngc4242-x1 & rh600943n00.1 + & 50767 & 2.4 & 14.5h & 5.7 & 2.76 & 0.96 & 17.0 & 22.64 & 4.49e-13 & 3.03 & + ngc4242-x2 & rh600943n00.3 + & 50767 & 2.4 & 14.5h & 5.3 & 2.90 & 0.96 & 17.7 & 24.07 & 4.68e-13 & 3.16 & + ngc4242-x3 & rh600943n00.2 + & 50767 & 2.4 & 14.5h & 3.7 & 3.09 & 0.95 & 12.0 & 26.38 & 3.19e-13 & 2.15 & + ngc4254-ulx1 & rh600972n00.7 + & 50629 & 20.8 & 6.2d & 4.6 & 0.50 & 0.98 & 37.7 & 26.39 & 1.39e-13 & 4.71 & 0.31 + ngc4254-ulx2 & rh600972n00.1 + & 50629 & 20.8 & 6.2d & 15.6 & 1.66 & 0.92 & 122.6 & 8.55 & 4.52e-13 & 15.3 & 0.36 + ngc4254-x3 & rh600972n00.3 + & 50629 & 20.8 & 6.2d & 3.9 & 4.19 & 0.96 & 22.3 & 26.42 & 8.22e-14 & 2.78 & + ngc4258-ulx1 & rh701008n00.u+ & 49122 & 27.3 & 1.6d & @xmath203 & & & @xmath2018.1 & & @xmath204.21e-14 & @xmath200.302 & + & rh600599n00.u & 49492 & 2.4 & 11.0h & @xmath203 & & & @xmath209.0 & & @xmath202.38e-13 & @xmath201.71 & + & rh600598n00.u & 49519 & 1.7 & 3.7h & @xmath203 & & & @xmath207.7 & & @xmath202.81e-13 & @xmath202.01 & + & rh600599a01.16 & 49680 & 25.3 & 19.4d & 7.7 & 8.62 & 1.00 & 89.9 & 16.28 & 2.27e-13 & 1.63 & 0.0013 + ngc4258-ulx2 & rh701008n00.u+ & 49122 & 27.3 & 1.6d & @xmath203 & & & @xmath2018.1 & & @xmath204.21e-14 & @xmath200.302 & + & rh600599n00.2 & 49492 & 2.4 & 11.0h & 2.7 & 7.74 & 0.99 & 8.4 & 25.02 & 2.22e-13 & 1.59 & + & rh600598n00.u & 49519 & 1.7 & 3.7h & @xmath203 & & & @xmath207.8 & & @xmath202.85e-13 & @xmath202.04 & + & rh600599a01.6 & 49680 & 25.3 & 19.4d & 3.7 & 7.69 & 1.01 & 26.4 & 26.33 & 6.68e-14 & 0.479 & + ngc4258-x9 & rh701008n00.1 + & 49122 & 27.3 & 1.6d & 14.6 & 9.53 & 0.96 & 142.7 & 8.99 & 3.32e-13 & 2.38 & 0.48 + & rh600599n00.u & 49492 & 2.4 & 11.0h & @xmath203 & & & @xmath2023.8 & & @xmath206.28e-13 & @xmath204.5 & + & rh600598n00.u & 49519 & 1.7 & 3.7h & @xmath203 & & & @xmath207.7 & & @xmath202.82e-13 & @xmath202.02 & + & rh600599a01.15 & 49680 & 25.3 & 19.4d & 5.8 & 15.98 & 0.48 & 77.9 & 22.46 & 1.97e-13 & 1.41 & 0.84 + ngc4319-x1 & rh600441n00.4 & 48917 & 12.3 & 3.5d & 297.4 & 0.59 & 0.87 & 3671.1 & 0.46 & 2.32e-11 & 2.21e+03 & 0.06 + & rh600834n00.4 + & 50039 & 35.0 & 10.7d & 385.9 & 6.68 & 0.88 & 9373.4 & 0.44 & 2.09e-11 & 1.99e+03 & 2.5e-08 + ngc4319-x2 & rh600441n00.2 & 48917 & 12.3 & 3.5d & 3.9 & 1.64 & 0.95 & 21.9 & 26.42 & 1.39e-13 & 13.3 & + & rh600834n00.2 + & 50039 & 35.0 & 10.7d & 4.2 & 7.66 & 1.01 & 36.2 & 26.41 & 8.09e-14 & 7.72 & 0.42 + ngc4292-x1 & rh600854n00.6 & 50267 & 15.9 & 6.0d & 6.8 & 13.41 & 0.82 & 54.6 & 18.38 & 2.35e-13 & 25.5 & + & rh600854a01.9 + & 50608 & 27.3 & 14.2d & 6.1 & 13.36 & 0.81 & 58.6 & 20.93 & 1.47e-13 & 16 & 0.15 + & rh600854a02.u & 50819 & 5.6 & 13.2h & @xmath203 & & & @xmath2017.0 & & @xmath202.10e-13 & @xmath2022.8 & + ngc4303-ulx1 & rh600854n00.2 & 50267 & 15.9 & 6.0d & 3.6 & 0.76 & 0.98 & 20.5 & 26.18 & 8.86e-14 & 2.46 & + & rh600854a01.u+ & 50608 & 27.3 & 14.2d & @xmath203 & & & @xmath2018.6 & & @xmath204.68e-14 & @xmath201.3 & + & rh600854a02.u & 50819 & 5.6 & 13.2h & @xmath203 & & & @xmath209.8 & & @xmath201.22e-13 & @xmath203.38 & + ngc4303-x4 & rh600854n00.u & 50267 & 15.9 & 6.0d & @xmath203 & & & @xmath2014.4 & & @xmath206.22e-14 & @xmath201.72 & + & rh600854a01.6 + & 50608 & 27.3 & 14.2d & 3.5 & 4.54 & 0.96 & 22.9 & 26.12 & 5.76e-14 & 1.6 & + & rh600854a02.u & 50819 & 5.6 & 13.2h & @xmath203 & & & @xmath209.6 & & @xmath201.19e-13 & @xmath203.3 & + ngc4303-x5 & rh600854n00.u & 50267 & 15.9 & 6.0d & @xmath203 & & & @xmath2014.7 & & @xmath206.33e-14 & @xmath201.75 & + & rh600854a01.u+ & 50608 & 27.3 & 14.2d & @xmath203 & & & @xmath2018.5 & & @xmath204.64e-14 & @xmath201.29 & + & rh600854a02.2 & 50819 & 5.6 & 13.2h & 4.2 & 6.32 & 1.00 & 14.2 & 26.41 & 1.75e-13 & 4.85 & + ngc4321-ulx1 & rh600731n00.3 + & 49873 & 42.4 & 23.2d & 4.7 & 1.55 & 0.96 & 37.4 & 26.39 & 6.55e-14 & 1.57 & + & rh500422n00.u & 49899 & 8.9 & 5.6h & @xmath203 & & & @xmath2012.4 & & @xmath201.03e-13 & @xmath202.47 & + & rh500542n00.u & 50801 & 24.8 & 16.4d & @xmath203 & & & @xmath2018.1 & & @xmath205.42e-14 & @xmath201.3 & + ngc4321-ulx2 & rh600731n00.9 + & 49873 & 42.4 & 23.2d & 4.0 & 3.25 & 0.95 & 34.3 & 26.41 & 6.00e-14 & 1.44 & + & rh500422n00.u & 49899 & 8.9 & 5.6h & @xmath203 & & & @xmath2012.1 & & @xmath201.01e-13 & @xmath202.42 & + & rh500542n00.u & 50801 & 24.8 & 16.4d & @xmath203 & & & @xmath2017.8 & & @xmath205.32e-14 & @xmath201.27 & + ngc4321-x4 & rh600731n00.u+ & 49873 & 42.4 & 23.2d & @xmath203 & & & @xmath2022.2 & & @xmath203.88e-14 & @xmath200.93 & + & rh500422n00.u & 49899 & 8.9 & 5.6h & @xmath203 & & & @xmath2012.1 & & @xmath201.01e-13 & @xmath202.42 & + & rh500542n00.3 & 50801 & 24.8 & 16.4d & 3.2 & 3.90 & 0.94 & 22.2 & 25.71 & 6.67e-14 & 1.6 & + ngc4321-x6 & rh600731n00.6 + & 49873 & 42.4 & 23.2d & 6.2 & 5.95 & 0.99 & 43.5 & 20.82 & 7.62e-14 & 1.83 & + & rh500422n00.u & 49899 & 8.9 & 5.6h & @xmath203 & & & @xmath2012.2 & & @xmath201.01e-13 & @xmath202.42 & + & rh500542n00.7 & 50801 & 24.8 & 16.4d & 4.5 & 5.95 & 1.00 & 31.1 & 26.40 & 9.32e-14 & 2.23 & + ngc4321-x7 & rh600731n00.5 + & 49873 & 42.4 & 23.2d & 4.3 & 6.30 & 1.00 & 30.6 & 26.40 & 5.37e-14 & 1.29 & + & rh500422n00.u & 49899 & 8.9 & 5.6h & @xmath203 & & & @xmath2012.2 & & @xmath201.01e-13 & @xmath202.42 & + & rh500542n00.u & 50801 & 24.8 & 16.4d & @xmath203 & & & @xmath2018.1 & & @xmath205.42e-14 & @xmath201.3 & + ngc4374-x2 & rh700192n00.5 & 48598 & 11.2 & 1.9d & 6.2 & 15.13 & 0.60 & 60.0 & 20.53 & 4.06e-13 & 16.4 & + & rh600463n00.u & 48801 & 1.5 & 0.4h & @xmath203 & & & @xmath2017.7 & & @xmath208.67e-13 & @xmath2035.1 & + & rh600463a02.u & 49516 & 10.4 & 16.3d & @xmath203 & & & @xmath2040.7 & & @xmath202.96e-13 & @xmath2012 & + & rh600493n00.5 + & 49699 & 26.2 & 7.3d & 16.7 & 2.31 & 0.91 & 165.6 & 8.15 & 4.78e-13 & 19.4 & 0.86 + & rh600463a03.2 & 49882 & 8.4 & 8.2h & 4.3 & 15.49 & 0.48 & 43.5 & 26.40 & 3.95e-13 & 16 & + ngc4374-x3 & rh600463n00.u & 48801 & 1.5 & 0.4h & @xmath203 & & & @xmath2013.2 & & @xmath206.45e-13 & @xmath2026.1 & + & rh600463a02.u & 49516 & 10.4 & 16.3d & @xmath203 & & & @xmath2028.7 & & @xmath202.09e-13 & @xmath208.46 & + & rh600493n00.4 + & 49699 & 26.2 & 7.3d & 8.8 & 3.70 & 0.96 & 62.8 & 14.22 & 1.81e-13 & 7.33 & 0.13 + & rh600463a03.u & 49882 & 8.4 & 8.2h & @xmath203 & & & @xmath2026.3 & & @xmath202.38e-13 & @xmath209.64 & + ngc4374-x4 & rh700192n00.4 & 48598 & 11.2 & 1.9d & 3.9 & 11.86 & 0.92 & 19.8 & 26.42 & 1.34e-13 & 5.43 & + & rh600463n00.u & 48801 & 1.5 & 0.4h & @xmath203 & & & @xmath2017.2 & & @xmath208.41e-13 & @xmath2034.1 & + & rh600463a02.u & 49516 & 10.4 & 16.3d & @xmath203 & & & @xmath2039.3 & & @xmath202.86e-13 & @xmath2011.6 & + & rh600493n00.2 + & 49699 & 26.2 & 7.3d & 3.9 & 5.50 & 0.99 & 24.4 & 26.41 & 7.06e-14 & 2.86 & + & rh600463a03.u & 49882 & 8.4 & 8.2h & @xmath203 & & & @xmath2035.9 & & @xmath203.25e-13 & @xmath2013.2 & + ngc4374-x5 & rh600493n00.9 + & 49699 & 26.2 & 7.3d & 4.6 & 5.93 & 1.00 & 29.9 & 26.39 & 8.63e-14 & 3.49 & + ngc4395-ulx1 & rh702725n00.2 + & 50258 & 11.3 & 7.2h & 32.1 & 2.49 & 0.88 & 189.0 & 4.80 & 1.11e-12 & 1.73 & 0.55 + ngc4438-x2 & rh600608n00.2 + & 49703 & 21.5 & 3.8d & 16.6 & 2.92 & 0.91 & 136.3 & 8.18 & 4.84e-13 & 16.4 & 0.81 + & rh701657n00.1 & 49907 & 5.8 & 2.4h & 5.6 & 3.48 & 0.97 & 24.7 & 22.99 & 3.24e-13 & 11 & + ngc4472-x2 & rh600216n00.u & 48967 & 7.0 & 3.5d & @xmath203 & & & @xmath2010.6 & & @xmath201.04e-13 & @xmath203.31 & + & rh600216a01.1 + & 49522 & 27.1 & 19.8d & 4.5 & 5.57 & 0.99 & 36.2 & 26.40 & 9.13e-14 & 2.91 & + ngc4472-x3 & rh600216n00.2 & 48967 & 7.0 & 3.5d & 9.3 & 6.70 & 0.97 & 41.8 & 13.63 & 4.08e-13 & 13 & + & rh600216a01.4 + & 49522 & 27.1 & 19.8d & 18.0 & 6.63 & 0.92 & 161.2 & 7.65 & 4.07e-13 & 13 & 0.85 + ngc4472-x4 & rh600216n00.u & 48967 & 7.0 & 3.5d & @xmath203 & & & @xmath2011.1 & & @xmath201.08e-13 & @xmath203.44 & + & rh600216a01.6 + & 49522 & 27.1 & 19.8d & 7.4 & 8.01 & 1.00 & 60.9 & 16.91 & 1.54e-13 & 4.9 & 0.99 + ngc4472-x5 & rh600216n00.4 & 48967 & 7.0 & 3.5d & 8.6 & 8.13 & 0.99 & 42.7 & 14.56 & 4.18e-13 & 13.3 & + & rh600216a01.7 + & 49522 & 27.1 & 19.8d & 15.4 & 8.08 & 0.94 & 133.1 & 8.61 & 3.36e-13 & 10.7 & 0.056 + ngc4472-x6 & rh600216n00.3 & 48967 & 7.0 & 3.5d & 4.0 & 9.98 & 1.01 & 15.7 & 26.41 & 1.54e-13 & 4.9 & + & rh600216a01.5 + & 49522 & 27.1 & 19.8d & 4.9 & 9.99 & 1.01 & 36.3 & 25.82 & 9.16e-14 & 2.92 & + ngc4476-x1 & rh700214n00.u & 48780 & 14.0 & 13.1h & @xmath203 & & & @xmath2040.0 & & @xmath202.15e-13 & @xmath207.65 & + & rh701712n00.u+ & 49877 & 44.3 & 1.9d & @xmath203 & & & @xmath2062.3 & & @xmath201.06e-13 & @xmath203.77 & + & rh701713n00.u & 50064 & 40.4 & 6.8d & @xmath203 & & & @xmath2062.8 & & @xmath201.17e-13 & @xmath204.16 & + & rh702081n00.u & 50267 & 34.2 & 5.8d & @xmath203 & & & @xmath2062.3 & & @xmath201.37e-13 & @xmath204.87 & + & rh702774n00.u & 50438 & 38.2 & 3.2d & @xmath203 & & & @xmath2063.0 & & @xmath201.24e-13 & @xmath204.41 & + & rh702775n00.6 & 50606 & 30.8 & 3.2d & 4.6 & 12.97 & 0.80 & 49.6 & 26.39 & 1.21e-13 & 4.31 & + & rh704000n00.u & 50796 & 8.0 & 24.1h & @xmath203 & & & @xmath2030.5 & & @xmath202.87e-13 & @xmath2010.2 & + & rh704000a01.u & 50818 & 20.6 & 1.2d & @xmath203 & & & @xmath2046.4 & & @xmath201.70e-13 & @xmath206.05 & + ngc4486-x2 & rh700214n00.u & 48780 & 14.0 & 13.1h & @xmath203 & & & @xmath2021.7 & & @xmath201.17e-13 & @xmath203.63 & + & rh701712n00.u+ & 49877 & 44.3 & 1.9d & @xmath203 & & & @xmath2032.8 & & @xmath205.58e-14 & @xmath201.73 & + & rh701713n00.13 & 50064 & 40.4 & 6.8d & 3.1 & 5.89 & 0.98 & 45.8 & 25.63 & 8.53e-14 & 2.64 & 0.4 + & rh702081n00.u & 50267 & 34.2 & 5.8d & @xmath203 & & & @xmath2032.8 & & @xmath207.22e-14 & @xmath202.24 & + & rh702774n00.u & 50438 & 38.2 & 3.2d & @xmath203 & & & @xmath2033.2 & & @xmath206.54e-14 & @xmath202.03 & + & rh702775n00.u & 50606 & 30.8 & 3.2d & @xmath203 & & & @xmath2030.2 & & @xmath207.37e-14 & @xmath202.28 & + & rh704000n00.u & 50796 & 8.0 & 24.1h & @xmath203 & & & @xmath2017.0 & & @xmath201.60e-13 & @xmath204.96 & + & rh704000a01.8 & 50818 & 20.6 & 1.2d & 2.9 & 5.71 & 0.98 & 27.7 & 25.30 & 1.01e-13 & 3.13 & 0.58 + ngc4485-ulx1 & rh600697n00.3 & 49857 & 6.6 & 23.4h & 6.5 & 2.70 & 0.96 & 35.3 & 19.51 & 3.73e-13 & 3.87 & + & rh600855n00.7 + & 50253 & 26.7 & 1.7d & 9.2 & 2.75 & 0.95 & 85.5 & 13.77 & 2.23e-13 & 2.31 & 0.36 + & rh600855a01.7 & 50414 & 24.6 & 1.2d & 7.2 & 2.67 & 0.96 & 49.9 & 17.43 & 1.41e-13 & 1.46 & 0.99 + ngc4490-ulx1 & rh600697n00.u & 49857 & 6.6 & 23.4h & @xmath203 & & & @xmath2011.6 & & @xmath201.22e-13 & @xmath200.891 & + & rh600855n00.5 + & 50253 & 26.7 & 1.7d & 7.6 & 0.48 & 0.96 & 72.7 & 16.46 & 1.89e-13 & 1.38 & 0.51 + & rh600855a01.5 & 50414 & 24.6 & 1.2d & 9.3 & 0.31 & 0.95 & 67.4 & 13.63 & 1.91e-13 & 1.39 & 0.5 + ngc4490-ulx2 & rh600697n00.2 & 49857 & 6.6 & 23.4h & 4.0 & 2.56 & 0.95 & 20.4 & 26.41 & 2.15e-13 & 1.57 & + & rh600855n00.3 + & 50253 & 26.7 & 1.7d & 6.3 & 2.31 & 0.96 & 57.6 & 20.46 & 1.50e-13 & 1.1 & 0.15 + & rh600855a01.3 & 50414 & 24.6 & 1.2d & 7.6 & 2.40 & 0.96 & 54.3 & 16.55 & 1.54e-13 & 1.12 & 0.1 + ngc4490-ulx3 & rh600697n00.u & 49857 & 6.6 & 23.4h & @xmath203 & & & @xmath2011.6 & & @xmath201.22e-13 & @xmath200.891 & + & rh600855n00.6 + & 50253 & 26.7 & 1.7d & 7.2 & 0.29 & 0.97 & 69.4 & 17.54 & 1.81e-13 & 1.32 & 0.27 + & rh600855a01.6 & 50414 & 24.6 & 1.2d & 6.6 & 0.12 & 0.97 & 42.2 & 19.09 & 1.19e-13 & 0.869 & 0.012 + ngc4501-ulx1 & rh601003n00.3 & 50638 & 4.3 & 2.3h & 5.9 & 5.76 & 0.99 & 24.9 & 21.84 & 4.33e-13 & 14.7 & + & rh601003a01.5 + & 50815 & 6.0 & 4.7d & 4.5 & 5.83 & 1.00 & 19.2 & 26.40 & 2.40e-13 & 8.13 & + ngc4523-x1 & rh800879n00.12 + & 50620 & 44.9 & 17.8d & 5.2 & 14.02 & 0.71 & 87.5 & 24.53 & 1.43e-13 & & 0.32 + & rh800879a01.u & 50800 & 21.3 & 15.4d & @xmath203 & & & @xmath2043.2 & & @xmath201.50e-13 & @xmath205.08 & + pgc41763-x1 & rh800879n00.3 + & 50620 & 44.9 & 17.8d & 8.4 & 4.01 & 0.96 & 100.8 & 14.86 & 1.65e-13 & 19.2 & 0.58 + & rh800879a01.3 & 50800 & 21.3 & 15.4d & 5.0 & 4.15 & 0.97 & 45.1 & 25.45 & 1.56e-13 & 18.1 & + ngc4536-x2 & rh601117n00.1 + & 50825 & 3.8 & 2.2h & 7.2 & 2.54 & 0.96 & 23.2 & 17.52 & 4.28e-13 & 10.5 & + ngc4552-x2 & rh600491n00.3 & 49543 & 18.0 & 1.5d & 3.9 & 3.71 & 0.95 & 22.4 & 26.42 & 9.43e-14 & 2.67 & + & rh600491a01.u+ & 49883 & 29.9 & 2.6d & @xmath203 & & & @xmath2019.2 & & @xmath204.85e-14 & @xmath201.37 & + ngc4559-ulx1 & rh600861n00.8 + & 50245 & 53.3 & 18.9d & 43.3 & 0.25 & 0.88 & 624.0 & 3.42 & 7.83e-13 & 3.16 & 0.016 + ngc4559-ulx2 & rh600861n00.12 + & 50245 & 53.3 & 18.9d & 66.9 & 2.03 & 0.87 & 992.8 & 1.81 & 1.25e-12 & 5.05 & 0.014 + ngc4565-ulx1 & rh600691n00.3 + & 49903 & 5.3 & 1.5d & 12.6 & 1.06 & 0.93 & 55.8 & 10.20 & 6.90e-13 & 25.2 & 0.52 + ngc4565-x3 & rh600691n00.1 + & 49903 & 5.3 & 1.5d & 3.5 & 2.19 & 0.94 & 11.8 & 26.13 & 1.46e-13 & 5.34 & + ngc4569-x2 & rh600603n00.u & 49544 & 2.1 & 0.8h & @xmath203 & & & @xmath208.3 & & @xmath202.97e-13 & @xmath2010.1 & + & rh600603a01.3 + & 49887 & 21.7 & 2.9d & 9.6 & 7.66 & 0.98 & 54.2 & 13.28 & 1.88e-13 & 6.37 & 0.75 + ngc4594-ulx1 & rh600044n00.u & 48633 & 1.7 & 0.5h & @xmath203 & & & @xmath207.2 & & @xmath203.61e-13 & @xmath204.14 & + & rh600044a01.u & 49352 & 3.6 & 8.0d & @xmath203 & & & @xmath209.8 & & @xmath202.22e-13 & @xmath202.54 & + & rh600044a02.3 + & 49542 & 15.7 & 7.8d & 10.8 & 0.40 & 0.94 & 96.4 & 11.53 & 5.09e-13 & 5.83 & 0.2 + ngc4594-ulx2 & rh600044n00.u & 48633 & 1.7 & 0.5h & @xmath203 & & & @xmath207.2 & & @xmath203.57e-13 & @xmath204.09 & + & rh600044a01.u & 49352 & 3.6 & 8.0d & @xmath203 & & & @xmath209.6 & & @xmath202.18e-13 & @xmath202.5 & + & rh600044a02.5 + & 49542 & 15.7 & 7.8d & 3.6 & 2.82 & 0.94 & 20.6 & 26.22 & 1.09e-13 & 1.25 & + ngc4594-x4 & rh600044n00.u & 48633 & 1.7 & 0.5h & @xmath203 & & & @xmath207.1 & & @xmath203.56e-13 & @xmath204.08 & + & rh600044a01.u & 49352 & 3.6 & 8.0d & @xmath203 & & & @xmath209.5 & & @xmath202.17e-13 & @xmath202.49 & + & rh600044a02.6 + & 49542 & 15.7 & 7.8d & 14.2 & 4.00 & 0.92 & 96.7 & 9.20 & 5.10e-13 & 5.84 & 0.16 + ngc4594-x5 & rh600044n00.u & 48633 & 1.7 & 0.5h & @xmath203 & & & @xmath207.2 & & @xmath203.59e-13 & @xmath204.11 & + & rh600044a01.u & 49352 & 3.6 & 8.0d & @xmath203 & & & @xmath209.7 & & @xmath202.19e-13 & @xmath202.51 & + & rh600044a02.1 + & 49542 & 15.7 & 7.8d & 3.7 & 5.93 & 0.99 & 17.2 & 26.40 & 9.08e-14 & 1.04 & + ngc4631-ulx1 & rh600193a00.1 & 48597 & 7.0 & 3.1d & 5.4 & 2.43 & 0.96 & 24.1 & 23.70 & 2.23e-13 & 1.27 & + & rh600193a01.4 & 48779 & 14.2 & 3.0d & 10.0 & 2.52 & 0.95 & 57.6 & 12.75 & 2.63e-13 & 1.5 & 0.22 + & rh600193a02.3 & 48969 & 11.9 & 1.1d & 13.2 & 2.43 & 0.93 & 78.1 & 9.76 & 4.24e-13 & 2.42 & 0.78 + & rh600934n00.4 + & 50626 & 23.9 & 6.9d & 13.2 & 2.48 & 0.93 & 109.2 & 9.82 & 2.96e-13 & 1.69 & 0.021 + ngc4649-x2 & rh600480n00.2 & 49545 & 2.2 & 0.8h & 2.5 & 3.13 & 0.92 & 7.4 & 24.74 & 2.45e-13 & 8.33 & + & rh600620n00.3 & 49546 & 4.8 & 2.3h & 5.0 & 3.04 & 0.96 & 19.8 & 25.33 & 3.02e-13 & 10.3 & + & rh600480a01.4 & 49886 & 11.8 & 17.7h & 6.9 & 3.10 & 0.96 & 41.1 & 18.14 & 2.54e-13 & 8.63 & + & rh600620a01.4 + & 49889 & 19.7 & 4.2d & 7.3 & 3.06 & 0.96 & 52.6 & 17.24 & 1.94e-13 & 6.59 & 0.56 + ngc4649-x3 & rh600480n00.u & 49545 & 2.2 & 0.8h & @xmath203 & & & @xmath208.1 & & @xmath202.70e-13 & @xmath209.18 & + & rh600620n00.1 & 49546 & 4.8 & 2.3h & 3.8 & 6.96 & 1.00 & 13.4 & 26.42 & 2.05e-13 & 6.97 & + & rh600480a01.1 & 49886 & 11.8 & 17.7h & 7.1 & 6.96 & 0.99 & 30.5 & 17.66 & 1.88e-13 & 6.39 & + & rh600620a01.1 + & 49889 & 19.7 & 4.2d & 13.1 & 6.96 & 0.94 & 96.0 & 9.84 & 3.55e-13 & 12.1 & 0.47 + ngc4656-ulx1 & rh600605n00.3 + & 49502 & 27.4 & 1.0d & 10.5 & 9.72 & 0.99 & 95.4 & 11.93 & 2.24e-13 & 1.39 & 0.71 + ngc4697-ulx1 & rh600825n00.1 & 50284 & 3.7 & 5.4h & 3.1 & 2.86 & 0.93 & 10.2 & 25.60 & 2.01e-13 & 3.33 & + & rh600825a01.u+ & 50619 & 77.9 & 30.0d & @xmath203 & & & @xmath2029.4 & & @xmath202.72e-14 & @xmath200.451 & + ngc4697-x7 & rh600825n00.u & 50284 & 3.7 & 5.4h & @xmath203 & & & @xmath209.2 & & @xmath201.81e-13 & @xmath203 & + & rh600825a01.5 + & 50619 & 77.9 & 30.0d & 5.2 & 4.13 & 0.97 & 65.6 & 24.55 & 6.08e-14 & 1.01 & 0.061 + ngc4861-ulx1 & rh600273n00.3 + & 48799 & 11.2 & 2.4d & 7.1 & 0.39 & 0.97 & 33.7 & 17.69 & 1.94e-13 & 7.38 & + & rh600273a01.4 & 48967 & 10.1 & 2.2d & 25.1 & 0.43 & 0.89 & 130.0 & 6.09 & 8.26e-13 & 31.4 & 0.9 + ngc4861-ulx2 & rh600273n00.4 + & 48799 & 11.2 & 2.4d & 7.2 & 1.05 & 0.97 & 38.6 & 17.35 & 2.21e-13 & 8.4 & + & rh600273a01.5 & 48967 & 10.1 & 2.2d & 6.8 & 0.97 & 0.97 & 33.1 & 18.33 & 2.10e-13 & 7.98 & + ngc5055-ulx1 & rh600742n00.4 + & 49691 & 12.2 & 7.4h & 7.5 & 0.68 & 0.96 & 42.2 & 16.63 & 2.26e-13 & 1.96 & + & rh601110n00.2 & 50813 & 11.0 & 1.6d & 5.5 & 0.67 & 0.98 & 30.8 & 23.65 & 1.83e-13 & 1.59 & + ngc5055-ulx2 & rh600742n00.5 + & 49691 & 12.2 & 7.4h & 4.9 & 1.79 & 0.96 & 25.5 & 25.91 & 1.36e-13 & 1.18 & + & rh601110n00.10 & 50813 & 11.0 & 1.6d & 3.2 & 1.77 & 0.94 & 15.0 & 25.72 & 8.96e-14 & 0.777 & + ngc5055-ulx3 & rh600742n00.3 + & 49691 & 12.2 & 7.4h & 4.5 & 1.58 & 0.96 & 22.7 & 26.40 & 1.21e-13 & 1.05 & + & rh601110n00.u & 50813 & 11.0 & 1.6d & @xmath203 & & & @xmath2012.9 & & @xmath207.70e-14 & @xmath200.668 & + ngc5055-x6 & rh600742n00.6 + & 49691 & 12.2 & 7.4h & 5.5 & 3.98 & 0.97 & 25.6 & 23.28 & 1.36e-13 & 1.18 & + & rh601110n00.3 & 50813 & 11.0 & 1.6d & 5.9 & 3.93 & 0.97 & 27.8 & 21.81 & 1.66e-13 & 1.44 & + ngc5055-ulx4 & rh600742n00.7 + & 49691 & 12.2 & 7.4h & 28.2 & 5.86 & 0.90 & 201.3 & 5.45 & 1.08e-12 & 9.36 & 0.61 + & rh601110n00.4 & 50813 & 11.0 & 1.6d & 16.8 & 5.84 & 0.92 & 94.7 & 8.14 & 5.64e-13 & 4.89 & 0.59 + ngc5055-x9 & rh600742n00.9 + & 49691 & 12.2 & 7.4h & 4.2 & 7.80 & 1.01 & 22.0 & 26.41 & 1.18e-13 & 1.02 & + & rh601110n00.u & 50813 & 11.0 & 1.6d & @xmath203 & & & @xmath2013.3 & & @xmath207.92e-14 & @xmath200.687 & + ngc5055-x10 & rh600742n00.u+ & 49691 & 12.2 & 7.4h & @xmath203 & & & @xmath2016.3 & & @xmath208.71e-14 & @xmath200.755 & + & rh601110n00.7 & 50813 & 11.0 & 1.6d & 4.8 & 10.35 & 1.01 & 23.1 & 26.01 & 1.37e-13 & 1.19 & + ngc5128-ulx1 & rh150004n00.u & 48099 & 19.5 & 19.8h & @xmath203 & & & @xmath2015.1 & & @xmath207.93e-14 & @xmath200.169 & + & rh701571n00.u+ & 49571 & 64.1 & 3.4d & @xmath203 & & & @xmath2027.0 & & @xmath204.31e-14 & @xmath200.0917 & + & rh701924n00.6 & 49911 & 4.9 & 3.9h & 32.0 & 2.62 & 0.88 & 197.8 & 4.82 & 4.13e-12 & 8.78 & 0.077 + & rh701925n00.7 & 49912 & 5.3 & 3.9h & 39.0 & 2.60 & 0.88 & 227.0 & 3.88 & 4.36e-12 & 9.27 & 0.12 + & rh701926n00.6 & 49916 & 5.7 & 3.8h & 36.6 & 2.57 & 0.88 & 216.1 & 4.27 & 3.87e-12 & 8.23 & 0.73 + & rh701927n00.5 & 49920 & 4.1 & 2.2h & 23.9 & 2.61 & 0.89 & 115.6 & 6.40 & 2.89e-12 & 6.15 & 0.14 + & rh701928n00.3 & 49921 & 4.1 & 2.2h & 29.2 & 2.59 & 0.89 & 157.6 & 5.31 & 3.97e-12 & 8.44 & 0.35 + & rh704206n00.u & 50850 & 17.2 & 1.6d & @xmath203 & & & @xmath2016.7 & & @xmath209.93e-14 & @xmath200.211 & + ngc5128-x13 & rh150004n00.11 & 48099 & 19.5 & 19.8h & 17.2 & 4.85 & 0.91 & 126.6 & 7.96 & 6.64e-13 & 1.41 & 0.47 + & rh701571n00.26 + & 49571 & 64.1 & 3.4d & 45.6 & 4.90 & 0.88 & 683.4 & 3.23 & 1.09e-12 & 2.32 & 0.0004 + & rh701924n00.7 & 49911 & 4.9 & 3.9h & 13.6 & 4.84 & 0.93 & 60.2 & 9.55 & 1.26e-12 & 2.68 & 0.84 + & rh701925n00.8 & 49912 & 5.3 & 3.9h & 12.0 & 4.85 & 0.94 & 54.4 & 10.64 & 1.04e-12 & 2.21 & 0.83 + & rh701926n00.7 & 49916 & 5.7 & 3.8h & 7.7 & 4.85 & 0.97 & 34.4 & 16.34 & 6.16e-13 & 1.31 & + & rh701927n00.6 & 49920 & 4.1 & 2.2h & 3.4 & 4.89 & 0.96 & 13.4 & 25.95 & 3.36e-13 & 0.715 & + & rh701928n00.4 & 49921 & 4.1 & 2.2h & 7.5 & 4.83 & 0.97 & 32.5 & 16.66 & 8.19e-13 & 1.74 & + & rh704206n00.9 & 50850 & 17.2 & 1.6d & 14.9 & 4.91 & 0.92 & 108.2 & 8.81 & 6.43e-13 & 1.37 & 0.5 + ngc5128-x33 & rh150004n00.16 & 48099 & 19.5 & 19.8h & 3.2 & 15.58 & 0.41 & 35.9 & 25.68 & 1.88e-13 & 0.4 & + & rh701571n00.3 + & 49571 & 64.1 & 3.4d & 6.4 & 15.62 & 0.54 & 145.9 & 19.92 & 2.33e-13 & 0.496 & 0.068 + & rh701924n00.u & 49911 & 4.9 & 3.9h & @xmath203 & & & @xmath2028.7 & & @xmath206.00e-13 & @xmath201.28 & + & rh701925n00.u & 49912 & 5.3 & 3.9h & @xmath203 & & & @xmath2029.5 & & @xmath205.67e-13 & @xmath201.21 & + & rh701926n00.u & 49916 & 5.7 & 3.8h & @xmath203 & & & @xmath2029.9 & & @xmath205.35e-13 & @xmath201.14 & + & rh701927n00.u & 49920 & 4.1 & 2.2h & @xmath203 & & & @xmath2026.6 & & @xmath206.64e-13 & @xmath201.41 & + & rh701928n00.u & 49921 & 4.1 & 2.2h & @xmath203 & & & @xmath2025.9 & & @xmath206.52e-13 & @xmath201.39 & + & rh704206n00.14 & 50850 & 17.2 & 1.6d & 6.5 & 15.51 & 0.55 & 79.6 & 19.58 & 4.73e-13 & 1.01 & + ngc5128-x37 & rh701927n00.8 & 49920 & 4.1 & 2.2h & 3.4 & 19.14 & 0.26 & 36.9 & 26.02 & 9.20e-13 & 1.96 & + ngc5128-x38 & rh701924n00.9 & 49911 & 4.9 & 3.9h & 3.7 & 19.69 & 0.27 & 44.4 & 26.42 & 9.25e-13 & 1.97 & + ngc5194-ulx1 & rh600062a00.u & 48597 & 8.5 & 33.3d & @xmath203 & & & @xmath2011.4 & & @xmath209.08e-14 & @xmath200.646 & + & rh600062a01.6 & 48765 & 8.7 & 13.7d & 5.8 & 1.63 & 0.96 & 28.8 & 22.33 & 2.22e-13 & 1.58 & + & rh600062a03.u & 49494 & 9.3 & 1.1d & @xmath203 & & & @xmath2012.7 & & @xmath209.24e-14 & @xmath200.657 & + & rh600601n00.13 + & 49521 & 36.0 & 5.8d & 4.9 & 1.62 & 0.96 & 37.7 & 25.99 & 7.08e-14 & 0.504 & + & rh601115n00.5 & 50809 & 8.0 & 3.3d & 5.8 & 1.70 & 0.96 & 27.5 & 22.13 & 2.32e-13 & 1.65 & + ngc5194-ulx2 & rh600062a00.u & 48597 & 8.5 & 33.3d & @xmath203 & & & @xmath2011.3 & & @xmath209.01e-14 & @xmath200.641 & + & rh600062a01.5 & 48765 & 8.7 & 13.7d & 4.7 & 2.45 & 0.96 & 20.4 & 26.39 & 1.58e-13 & 1.12 & + & rh600062a03.u & 49494 & 9.3 & 1.1d & @xmath203 & & & @xmath2012.6 & & @xmath209.16e-14 & @xmath200.652 & + & rh600601n00.u+ & 49521 & 36.0 & 5.8d & @xmath203 & & & @xmath2019.9 & & @xmath203.74e-14 & @xmath200.266 & + & rh601115n00.u & 50809 & 8.0 & 3.3d & @xmath203 & & & @xmath2011.9 & & @xmath201.00e-13 & @xmath200.711 & + ngc5194-ulx3 & rh600062a00.2 & 48597 & 8.5 & 33.3d & 10.8 & 2.16 & 0.94 & 57.9 & 11.60 & 4.61e-13 & 3.28 & 0.043 + & rh600062a01.2 & 48765 & 8.7 & 13.7d & 12.0 & 2.29 & 0.93 & 71.8 & 10.65 & 5.55e-13 & 3.95 & 0.14 + & rh600062a03.u & 49494 & 9.3 & 1.1d & @xmath203 & & & @xmath2012.6 & & @xmath209.18e-14 & @xmath200.653 & + & rh600601n00.7 + & 49521 & 36.0 & 5.8d & 18.0 & 2.22 & 0.91 & 193.6 & 7.65 & 3.64e-13 & 2.59 & 0.021 + & rh601115n00.u & 50809 & 8.0 & 3.3d & @xmath203 & & & @xmath2011.9 & & @xmath201.00e-13 & @xmath200.711 & + ngc5194-ulx4 & rh600062a00.4 & 48597 & 8.5 & 33.3d & 3.8 & 2.34 & 0.94 & 16.4 & 26.42 & 1.31e-13 & 0.932 & + & rh600062a01.7 & 48765 & 8.7 & 13.7d & 6.0 & 2.21 & 0.96 & 32.8 & 21.61 & 2.54e-13 & 1.81 & + & rh600062a03.u & 49494 & 9.3 & 1.1d & @xmath203 & & & @xmath2012.6 & & @xmath209.17e-14 & @xmath200.652 & + & rh600601n00.14 + & 49521 & 36.0 & 5.8d & 10.4 & 2.27 & 0.94 & 92.6 & 12.06 & 1.74e-13 & 1.24 & 0.76 + & rh601115n00.u & 50809 & 8.0 & 3.3d & @xmath203 & & & @xmath2011.9 & & @xmath201.00e-13 & @xmath200.711 & + ngc5194-ulx5 & rh600062a00.1 & 48597 & 8.5 & 33.3d & 8.4 & 2.64 & 0.95 & 40.7 & 14.91 & 3.25e-13 & 2.31 & + & rh600062a01.1 & 48765 & 8.7 & 13.7d & 10.6 & 2.74 & 0.94 & 50.5 & 11.90 & 3.91e-13 & 2.78 & + & rh600062a03.2 & 49494 & 9.3 & 1.1d & 7.0 & 2.70 & 0.96 & 34.0 & 17.82 & 2.47e-13 & 1.76 & + & rh600601n00.4 + & 49521 & 36.0 & 5.8d & 23.4 & 2.75 & 0.90 & 273.2 & 6.51 & 5.13e-13 & 3.65 & 0.00041 + & rh601115n00.1 & 50809 & 8.0 & 3.3d & 5.7 & 2.78 & 0.96 & 25.7 & 22.61 & 2.17e-13 & 1.54 & + ngc5194-ulx6 & rh600062a00.u & 48597 & 8.5 & 33.3d & @xmath203 & & & @xmath2011.3 & & @xmath208.99e-14 & @xmath200.64 & + & rh600062a01.3 & 48765 & 8.7 & 13.7d & 4.4 & 2.65 & 0.95 & 23.8 & 26.40 & 1.84e-13 & 1.31 & + & rh600062a03.u & 49494 & 9.3 & 1.1d & @xmath203 & & & @xmath2012.6 & & @xmath209.14e-14 & @xmath200.65 & + & rh600601n00.10 + & 49521 & 36.0 & 5.8d & 4.8 & 2.62 & 0.96 & 34.6 & 26.32 & 6.50e-14 & 0.462 & + & rh601115n00.u & 50809 & 8.0 & 3.3d & @xmath203 & & & @xmath2011.8 & & @xmath209.99e-14 & @xmath200.711 & + ngc5204-ulx1 & rh702723n00.1 + & 50083 & 14.7 & 2.8d & 50.4 & 0.41 & 0.88 & 428.2 & 2.77 & 1.93e-12 & 4.28 & 0.45 + & rh702723a01.1 & 50206 & 13.6 & 1.0d & 54.0 & 0.37 & 0.87 & 476.2 & 2.55 & 2.32e-12 & 5.15 & 2.3e-13 + ngc5236-ulx1 & rh600024n00.2 & 49008 & 23.3 & 1.5d & 19.1 & 4.58 & 0.90 & 158.7 & 7.28 & 5.65e-13 & 1.5 & 0.64 + & rh600024a01.1 + & 49564 & 23.9 & 3.4d & 32.5 & 4.63 & 0.89 & 286.4 & 4.70 & 9.94e-13 & 2.63 & 0.0026 + ngc5457-ulx1 & rh600092n00.8 & 48631 & 18.5 & 2.0d & 13.4 & 5.85 & 0.93 & 81.0 & 9.69 & 2.79e-13 & 1.57 & 0.00021 + & rh600383n00.11 & 48967 & 32.4 & 3.8d & 15.4 & 5.88 & 0.92 & 109.7 & 8.62 & 2.16e-13 & 1.22 & 0.95 + & rh600820n00.11 + & 50217 & 107.9 & 26.0d & 26.8 & 5.84 & 0.90 & 439.3 & 5.76 & 2.59e-13 & 1.46 & 0.031 + & rh600820a01.15 & 50408 & 68.2 & 2.7d & 17.9 & 5.91 & 0.91 & 220.8 & 7.68 & 2.06e-13 & 1.16 & 0.57 + ngc5457-ulx2 & rh600092n00.7 & 48631 & 18.5 & 2.0d & 6.4 & 7.24 & 1.00 & 30.4 & 20.09 & 1.05e-13 & 0.592 & + & rh600383n00.9 & 48967 & 32.4 & 3.8d & 18.2 & 7.28 & 0.93 & 138.6 & 7.57 & 2.73e-13 & 1.54 & 0.6 + & rh600820n00.9 + & 50217 & 107.9 & 26.0d & 21.8 & 7.23 & 0.92 & 407.7 & 6.73 & 2.41e-13 & 1.36 & 0.0001 + & rh600820a01.13 & 50408 & 68.2 & 2.7d & 14.7 & 7.20 & 0.94 & 159.9 & 8.87 & 1.49e-13 & 0.84 & 0.51 + ngc5457-x21 & rh600092n00.1 & 48631 & 18.5 & 2.0d & 9.2 & 10.70 & 0.99 & 63.9 & 13.71 & 2.21e-13 & 1.25 & 0.0002 + & rh600383n00.1 & 48967 & 32.4 & 3.8d & 6.2 & 10.68 & 1.00 & 48.6 & 20.58 & 9.56e-14 & 0.539 & 0.13 + & rh600820n00.2 + & 50217 & 107.9 & 26.0d & 9.9 & 10.63 & 0.99 & 153.6 & 12.88 & 9.07e-14 & 0.511 & 4e-05 + & rh600820a01.2 & 50408 & 68.2 & 2.7d & 12.6 & 10.59 & 0.98 & 153.8 & 10.17 & 1.44e-13 & 0.812 & 0.053 + ngc5457-ulx3 & rh600092n00.2 & 48631 & 18.5 & 2.0d & 14.8 & 11.04 & 0.97 & 113.0 & 8.86 & 3.90e-13 & 2.2 & 0.21 + & rh600383n00.2 & 48967 & 32.4 & 3.8d & 24.8 & 11.00 & 0.95 & 285.6 & 6.09 & 5.62e-13 & 3.17 & 0.2 + & rh600820n00.4 + & 50217 & 107.9 & 26.0d & 43.7 & 11.08 & 0.94 & 951.1 & 3.41 & 5.61e-13 & 3.16 & 1.9e-07 + & rh600820a01.3 & 50408 & 68.2 & 2.7d & 22.8 & 11.00 & 0.96 & 325.7 & 6.56 & 3.04e-13 & 1.71 & 0.1 + ngc5457-x32 & rh600092n00.13 & 48631 & 18.5 & 2.0d & 7.4 & 13.88 & 0.79 & 64.8 & 17.06 & 2.24e-13 & 1.26 & 0.039 + & rh600383n00.u & 48967 & 32.4 & 3.8d & @xmath203 & & & @xmath2043.8 & & @xmath208.62e-14 & @xmath200.486 & + & rh600820n00.u+ & 50217 & 107.9 & 26.0d & @xmath203 & & & @xmath2076.9 & & @xmath204.54e-14 & @xmath200.256 & + & rh600820a01.u & 50408 & 68.2 & 2.7d & @xmath203 & & & @xmath2065.4 & & @xmath206.11e-14 & @xmath200.345 & + ngc5457-x34 & rh600092n00.14 & 48631 & 18.5 & 2.0d & 8.0 & 16.09 & 0.53 & 116.0 & 15.73 & 4.00e-13 & 2.26 & 0.46 + & rh600383n00.17 & 48967 & 32.4 & 3.8d & 9.4 & 16.13 & 0.56 & 155.9 & 13.55 & 3.07e-13 & 1.73 & 0.55 + & rh600820n00.18 + & 50217 & 107.9 & 26.0d & 18.7 & 16.18 & 0.71 & 666.2 & 7.42 & 3.93e-13 & 2.22 & 2.2e-05 + & rh600820a01.19 & 50408 & 68.2 & 2.7d & 12.7 & 16.21 & 0.62 & 384.2 & 10.13 & 3.59e-13 & 2.02 & 0.89 + ngc5774-ulx1 & rh600964n00.2 + & 50646 & 34.8 & 8.2d & 4.7 & 3.41 & 0.96 & 38.4 & 26.39 & 8.96e-14 & 7.72 & + ngc5774-x2 & rh600964n00.3 + & 50646 & 34.8 & 8.2d & 3.8 & 5.97 & 0.99 & 26.5 & 26.42 & 6.15e-14 & 5.3 & + ngc6434-x1 & rh601018n00.2 & 50433 & 20.6 & 4.7d & 15.5 & 0.88 & 0.92 & 149.0 & 8.58 & 5.98e-13 & 78.6 & 0.39 + & rh601018a01.4 + & 50473 & 23.8 & 1.0d & 16.1 & 0.89 & 0.91 & 123.2 & 8.37 & 4.27e-13 & 56.1 & 0.24 + ngc6654-x1 & rh600124a00.18 + & 48512 & 36.5 & 23.0h & 4.8 & 12.79 & 0.83 & 51.8 & 26.00 & 1.29e-13 & 13.5 & 0.11 + & rh600124a01.u & 48731 & 18.1 & 1.3d & @xmath203 & & & @xmath2026.7 & & @xmath201.34e-13 & @xmath2014 & + ngc6654-x2 & rh600124a00.11 + & 48512 & 36.5 & 23.0h & 5.8 & 13.17 & 0.82 & 53.2 & 22.43 & 1.32e-13 & 13.8 & + & rh600124a01.u & 48731 & 18.1 & 1.3d & @xmath203 & & & @xmath2028.9 & & @xmath201.44e-13 & @xmath2015 & + ngc6946-ulx1 & rh600501n00.8
+ & 49486 & 59.9 & 15.1d & 20.5 & 0.95 & 0.90 & 208.4 & 6.96 & 4.97e-13 & 1.8 & 0.0097 + & rh600718n00.2 & 49943 & 21.5 & 3.0d & 5.7 & 1.29 & 0.97 & 44.0 & 22.67 & 2.92e-13 & 1.06 & 0.64 + & rh500476n00.u & 50305 & 8.4 & 9.7h & @xmath203 & & & @xmath2011.8 & & @xmath202.02e-13 & @xmath200.733 & + ngc6946-ulx2 & rh600501n00.12 + & 49486 & 59.9 & 15.1d & 15.8 & 2.06 & 0.92 & 152.6 & 8.46 & 3.64e-13 & 1.32 & 0.21 + & rh600718n00.4 & 49943 & 21.5 & 3.0d & 9.3 & 1.78 & 0.95 & 66.6 & 13.57 & 4.42e-13 & 1.6 & 0.4 + & rh500476n00.u & 50305 & 8.4 & 9.7h & @xmath203 & & & @xmath2011.7 & & @xmath202.00e-13 & @xmath200.726 & + ngc6946-ulx3 & rh600501n00.7 + & 49486 & 59.9 & 15.1d & 71.1 & 2.69 & 0.87 & 922.0 & 1.74 & 2.20e-12 & 7.99 & 0.83 + & rh600718n00.1 & 49943 & 21.5 & 3.0d & 47.7 & 2.36 & 0.88 & 455.5 & 2.97 & 3.02e-12 & 11 & 0.031 + & rh500476n00.1 & 50305 & 8.4 & 9.7h & 23.7 & 0.14 & 0.90 & 145.8 & 6.47 & 2.50e-12 & 9.07 & 0.45 + ngc7217-x2 & rh702933n00.u & 50612 & 4.8 & 2.0d & @xmath203 & & & @xmath209.2 & & @xmath202.06e-13 & @xmath206.33 & + & rh702933a01.1 + & 50773 & 10.7 & 14.7h & 3.3 & 2.94 & 0.94 & 14.9 & 25.80 & 1.52e-13 & 4.67 & + pgc68618-x1 & rh500493n00.7 + & 50582 & 34.0 & 1.4d & 5.8 & 6.02 & 0.99 & 42.5 & 22.37 & 8.69e-14 & 1.28 & + ngc7314-ulx1 & rh701300n00.4 + & 49665 & 19.1 & 14.0d & 5.6 & 1.66 & 0.96 & 33.5 & 23.11 & 1.17e-13 & 2.34 & + ngc7314-x4 & rh701300n00.8 + & 49665 & 19.1 & 14.0d & 4.4 & 2.68 & 0.95 & 26.0 & 26.40 & 9.07e-14 & 1.81 & + ngc7314-x5 & rh701300n00.5 + & 49665 & 19.1 & 14.0d & 11.3 & 4.50 & 0.94 & 74.8 & 11.20 & 2.61e-13 & 5.21 & 0.18 + ngc7582-x2 & rh800189n00.u & 48780 & 7.5 & 3.1d & @xmath203 & & & @xmath2032.8 & & @xmath203.12e-13 & @xmath2011.6 & + & rh800189a01.u & 49124 & 4.2 & 23.9d & @xmath203 & & & @xmath2024.5 & & @xmath204.15e-13 & @xmath2015.4 & + & rh701710n00.3 + & 49860 & 17.3 & 18.0h & 7.9 & 1.93 & 0.96 & 51.4 & 15.81 & 2.10e-13 & 7.81 & 0.39 + & rh702055n00.u & 50418 & 11.2 & 8.6h & @xmath203 & & & @xmath2013.4 & & @xmath208.48e-14 & @xmath203.15 & + & rh702055a01.u & 50758 & 12.2 & 2.9d & @xmath203 & & & @xmath2013.8 & & @xmath208.02e-14 & @xmath202.98 & + ngc7590-ulx1 & rh800189n00.u & 48780 & 7.5 & 3.1d & @xmath203 & & & @xmath2011.9 & & @xmath201.14e-13 & @xmath204.09 & + & rh800189a01.u & 49124 & 4.2 & 23.9d & @xmath203 & & & @xmath209.7 & & @xmath201.64e-13 & @xmath205.89 & + & rh701710n00.u+ & 49860 & 17.3 & 18.0h & @xmath203 & & & @xmath2017.0 & & @xmath206.97e-14 & @xmath202.5 & + & rh702055n00.4 & 50418 & 11.2 & 8.6h & 5.0 & 10.21 & 1.01 & 28.7 & 25.31 & 1.82e-13 & 6.54 & + & rh702055a01.5 & 50758 & 12.2 & 2.9d & 3.5 & 10.09 & 1.01 & 16.2 & 26.14 & 9.45e-14 & 3.39 & + ngc7714-ulx1 & rh600738n00.4 + & 49713 & 45.5 & 2.9d & 13.7 & 0.82 & 0.92 & 132.6 & 9.51 & 2.56e-13 & 41.8 & 0.92 + ngc7742-ulx1 & rh500386n00.1 + & 49713 & 9.6 & 2.0d & 3.4 & 0.43 & 0.98 & 13.5 & 25.91 & 1.25e-13 & 7.39 & + | x - ray observations have revealed in other galaxies a class of extra - nuclear x - ray point sources with x - ray luminosities of @xmath0@xmath1 erg / sec , exceeding the eddington luminosity for stellar mass x - ray binaries . these ultra - luminous x - ray sources ( ulxs ) may be powered by intermediate mass black hole of a few thousand @xmath2 or stellar mass black holes with special radiation processes . in this paper , we present a survey of ulxs in 313 nearby galaxies with @xmath3@xmath4@xmath5 within 40 mpc with 467 rosat hri archival observations .
the hri observations are reduced with uniform procedures , refined by simulations that help define the point source detection algorithm employed in this survey .
a sample of 562 extragalactic x - ray point sources with @xmath6@xmath7 erg / sec is extracted from 173 survey galaxies , including 106 ulx candidates within the @xmath3 isophotes of 63 galaxies and 110 ulx candidates between 12@xmath8 of 64 galaxies , from which a clean sample of 109 ulxs is constructed to minimize the contamination from foreground or background objects .
the strong connection between ulxs and star formation is confirmed based on the striking preference of ulxs to occur in late - type galaxies , especially in star forming regions such as spiral arms .
ulxs are variable on time scales over days to years , and exhibit a variety of long term variability patterns .
the identifications of ulxs in the clean sample show some ulxs identified as supernovae ( remnants ) , hii regions / nebulae , or young massive stars in star forming regions , and a few other ulxs identified as old globular clusters . in a subsequent paper ,
the statistic properties of the survey will be studied to calculate the occurrence frequencies and luminosity functions for ulxs in different types of galaxies to shed light on the nature of these enigmatic sources . |
many if not all high - luminous infrared galaxies ( ulirgs , @xmath1 ) possess regions hidden by huge amounts of dust .
this makes it difficult to ascertain whether this enormous energy output is due to a starburst activity or an accretion process onto a supermassive black hole .
one of the best known objects to study this relationship is the nearby ulirg ngc 6240 ( assuming @xmath2 ) .
infrared observations favour an energy source dominated by starburst processes , whereas observations in the x - ray range point to an agn as the central engine ( @xmath3 ) .
we have analyzed the data of ngc 6240 taken from an 24 ksec observation with _ xmm - newton _ using the epic - pn and epic - mos instruments . in order to investigate the fe line complex around 6.4 kev and the 0.3 - 10.0 kev spectrum as a whole the high sensitivity and
therefore the good photon statistics - especially in the 6.4 kev range - in combination with a higher energy resolution enables us to examine this feature in unprecedented detail .
table 1 summarizes some basic parameters ( powerlaw - @xmath4 , line energies ) of different models ( first column ) after fitting to the data .
the first of the leading three models includes line profiles with no line width ( @xmath5 ) , whereas eachone of the last two models uses a second powerlaw , but with a different number of line profiles .
each model contains a 6.4kev line as an indication of an agn contribution .
a prove of an compton - thick agn has been reported by vignati et al .
( 1999 ) using bepposax and by ikebe et al .
( 2000 ) using rxte . however , the last model seems to have the best statistical acceptance ( see fig . 1 , left ) . emission lines & powerlaw & & @xmath6 + & @xmath4 & energy - line 1 & energy - line 2 & energy - line 3 & d.o.f .
+ lines : @xmath7 & -0.18 & @xmath8 & @xmath9 & @xmath10 & 38.5/53 + lines : @xmath11 & -0.16 & @xmath12 & @xmath13 & @xmath14 & 38.4/51 + lines : 2nd broad & -0.27 & @xmath15 & @xmath16 & @xmath17 & 43.1/53 + emission lines + & powerlaw & & @xmath6 + absorp . edge : & @xmath4 & energy - line 1 & energy - line 2 & energy - line 3 & d.o.f .
+ po + 2 lines & 0.47 & @xmath18 & @xmath19 & - & 39.7/54 + po + 3 lines & 0.47 & @xmath12 & @xmath19 & @xmath20 & 39.1/54 + the analysis of the spectral data ( @xmath21 ) indicates at least two models providing an statistically acceptable fit : each of them contains two thin thermal plasmas ( @xmath22 and @xmath23 ) , a direct component ( absorbed powerlaw with @xmath24 and @xmath25 , both fixed ) as well as a reflection component ( absorbed powerlaw , either reflected from neutral matter or not ) .
finally , three gaussian lines have been added to the models ( neutral + ionized k @xmath0 and k @xmath26 ) . the right plot of fig .
1 shows the components of the second model ( incl .
reflection ) and their deviations from the data points . | a recently performed _
xmm - newton _ observation of the ulirg ngc 6240 clearly indicates the presence of an agn contribution to its x - ray spectrum . in the 5.0 - 7.0 kev energy range
there is a clear signature of the fluorescent fe k @xmath0 lines at 6.4 , 6.7 and 6.9 kev , respectively .
the line strength of the 6.4 kev line can not be produced by a thermal component .
the 0.3 - 10.0 kev spectral energy distribution is characterized by the following components : ( i ) two hot thermal components ( the starburst ) , ( ii ) one direct component ( heavily absorbed ; agn is hidden ) , ( iii ) one reflection component ( the agn ) , ( iv ) three narrow fe lines
. the model parameters for the broad - band spectral energy distribution are consistent with the results of previously works .
# 1_#1 _ # 1_#1 _ = # 1 1.25 in .125 in .25 in |
the connection between long - duration gamma - ray bursts ( grbs ) and sne is well established in the local universe .
following the first cases of coincidences @xcite , now for almost every grb at redshift @xmath2 a corresponding sn has been identified @xcite . despite the diverse properties of their associated grbs , all grb / sne observed so far are luminous , broad - lined type ic sne ( no h , no he , * ? ? ?
the very broad lines indicate a high expansion velocity of the ejecta , and point to a high explosion kinetic energy ( @xmath3 ) @xcite .
detailed models of grb / sne yield a typical sn @xmath3 of a few @xmath0erg ( depending on the asphericity of the sn ) , an ejected mass @xmath4@xmath5 , and a @xmath6ni mass of @xmath7 .
this places grb / sne at the luminous , energetic and massive end of sneic ( e.g. * ? ? ?
* ) and points to a massive star origin ( e.g. * ? ? ?
two recent events confirm and reinforce this trend : sn2013cq
/ grb130427a @xcite and sn2013dx / grb130702a @xcite .
although the two grbs are very different in energy , the former extremely powerful , similar to cosmological ones , the latter a normal grb , the sne are again similar ( * ? ? ?
* delia et al . , in prep . ) .
it has been proposed that long grbs are produced by the collapsar mechanism , where a massive star collapses directly to a black hole ( bh ) .
accretion on the bh releases energy in the form of a relativistic jet which may explode the star and produce a visible sn if @xmath6ni is synthesised in the dense neutrino wind emanating from the accretion disc @xcite .
sneic associated with x - ray flashes ( xrfs ) have a smaller @xmath3 , more similar to ordinary sneic , and are not as luminous @xcite .
models indicate progenitor stars of @xmath8 , which are expected to collapse to neutron stars ( ns ) .
their @xmath3 ( a few @xmath9 erg ) appears to be consistent with energy injection from a magnetar , a rapidly spinning magnetised ns @xcite .
this mechanism taps the energy in the magnetic field and may also give rise to a relativistic jet ( see e.g. , * ? ? ?
* ; * ? ? ?
observational and theoretical evidence has been mounting that more massive stars can also collapse to ns @xcite .
magnetar jets and their potential as a source of grbs have been investigated in various papers , suggesting that magnetar energy can be used to energise grbs or xrfs @xcite .
it has also been proposed that very rapidly spinning magnetars can explain the much brighter light curves of grb / sne @xcite .
this may conflict with the observation in sn1998bw of strong emission lines of fe , which indicate a high @xmath6ni yield @xcite . on the other hand ,
only sn1998bw could be followed late enough to observe fe lines .
one of the most interesting unsolved questions in grb science is what actually drives the event . in the collapsar model
the jet generated by the bh explodes the star , but is its energy sufficient to impart a high @xmath3 to the sn ?
simulations have so far not tested this , but the energy needed for the jet to emerge from the star and unbind it ( @xmath10erg , * ? ? ? * ) appears to be much smaller than the sn @xmath3 . in the magnetar scenario ,
if the coupling is large energy may be extracted from the ns and added to the sn @xmath3 , which would otherwise derive from the classical neutrino mechanism .
the sub - relativistic outflow may not be highly collimated , as indicated by the distribution of sn material @xcite . in this scenario energy production
would be limited by the ns spin rate .
we analyse the global properties of the grbs and their sne in order to look for indications of a preferred mechanism .
we compare the energies of grbs , xrfs , and their accompanying sne . in section 2.1
we estimate the intrinsic energy of low - redshift grbs ( @xmath11 ) with associated sne by applying a correction for the jet opening angle to the observed @xmath12-ray energies . in section 2.2
we estimate the energy in relativistic ejecta as probed by radio data . in section 2.3
we compare both of these to the sn @xmath3 as derived from modelling . in section 3
we present our results . in section 4
we extend the comparison to all grbs at higher redshift for which a sn was reported and discuss our findings .
isotropic - equivalent energies ( @xmath13 ) of nearby grbs connected with well - studied sne are extremely diverse .
grb980425 had a very low @xmath13 , which was one of the aspects that raised doubts on the reality of the first grb / sn association . on the other hand grb030329 , associated with sn2003dh , was similar to many long grbs .
grb130427a has @xmath13@xmath14erg , comparable to cosmological grbs .
however , @xmath13is unlikely to be the real jet energy .
the true energy of the jet , @xmath15 , can be estimated from @xmath13 , adopting a correction for collimation .
alternatively , radio energy is thought to be a good proxy for the energy of relativistic material , assuming that this energy is completely used up in the interaction with circumstellar material and radiated isotropically at later times ( jet radio calorimetry ) .
a model - dependent estimate of @xmath15 can be obtained from the timing of the break in the afterglow light curve .
an achromatic break may indicate that the edge of the jet swept past our viewing point .
this information is however not always available .
its absence may indicate lack of collimation but also just be due to incomplete data . once @xmath13has been corrected for jet collimation , which can be quite uncertain ( see e.g. * ? ? ?
* ) , it can be compared with the sn @xmath3 and with the radio energies .
cccccccccc grb / sn & @xmath16 & t90 & @xmath13 & @xmath17 & @xmath15 & sn @xmath3 & m(@xmath6ni ) & @xmath18 & refs .
+ & & [ s ] & [ @xmath19 erg ] & [ deg ] & [ @xmath19 erg ] & [ @xmath19 erg ] & [ @xmath20 & [ @xmath19 erg ] & + 1 & 2 & 3 & 4 & 5 & 6 & 7 & 8 & 9 & 10 + 980425 / 1998bw & 0.0085 & 30 & @xmath21 & 180 & @xmath21 & @xmath22 & @xmath23 & @xmath240.2 & 1 - 3 + 030329 / 2003dh & 0.1685 & 23 & @xmath25 & @xmath26 & @xmath27 & @xmath28 & @xmath29 & @xmath30 & 2,4 - 7 + 031203 / 2003lw & 0.1055 & 40 & @xmath31 & 180 & @xmath32 & @xmath33 & @xmath34 & @xmath35 & 1,2,8 + 060218 / 2006aj & 0.0335 & 2000 & @xmath36 & 180 & @xmath37 & @xmath38 & @xmath39 & @xmath40 & 4,9 - 11 + 100316d / 2010bh & 0.059 & @xmath411300 & @xmath42 & 180 & @xmath43 & @xmath44 & @xmath45 & @xmath240.2 & 12 - 14 + 120422a / 2012bz & 0.283 & 5 & @xmath46 & @xmath47 & @xmath48 & @xmath28 & @xmath49 & & 15,16 + 130427a / 2013cq & 0.3399 & 160 & @xmath50 & @xmath51 & @xmath52 & @xmath53 & @xmath29 & @xmath26 & 17 - 20 + 130702a / 2013dx & 0.145 & 59 & @xmath54 & @xmath55 & @xmath56 & @xmath57 & @xmath58 & @xmath59 & 21 - 23 + + + + + + + values of @xmath13 of grb / sne are listed in table 1 .
if an estimate of the jet opening angle @xmath17 was available in the literature or could be derived from the afterglow multiwavelength light curves ( as outlined by * ? ? ?
* ) , we reported this angle and computed the collimation - corrected energy @xmath15 .
the optical light curves of grb130702a steepen at @xmath60 days , and the x - ray light curve is compatible with a steepening at about the same time @xcite .
if that is a jet break , the jet opening angle is @xmath2414 degrees and @xmath61 erg .
a sn similar to sn1998bw was indeed detected in coincidence with grb130702a ( delia et al .
, in prep . ) .
no correction is possible for grbs 980425 and 031203 or the two xrfs .
we list grb / sn energies from radio measurements ( @xmath18 ) in table 1 .
whenever possible , we took estimates from the literature . for grb130427a / sn2013cq and
grb130702a / sn2013dx @xmath18was estimated from the available radio data following @xcite . in the case of grb130702a / sn2013dx there may be a significant contribution from the afterglow , because radio measurements were taken only 2 rest - frame days after the explosion .
since it is impossible to disentangle the contribution to the radio emission by the grb from that of the sn , these values must be regarded as upper limits to the sn energy .
no radio observations are available for grb120422a / sn2012bz .
we obtained sn @xmath3 ( table 1 ) from models or from spectroscopic analogues .
in particular , @xmath3 of sne1998bw , 2003dh and 2003lw are from the re - analysis of @xcite . for sn2012bz @xmath3
is taken to be equal to that of sn2003dh based on the similarity of the spectra and the light curve @xcite ; m(@xmath6ni ) is only 15% less than for sn2003dh . for sn2013cq we used the @xmath3 estimate of @xcite , who find a value similar to that of sn1998bw .
this is also supported by the fact that the bolometric light curve maximum of sn2013cq , which was accurately measured with hst @xcite , is consistent with that of sn1998bw . for sn2013dx ,
whose light curve is similar to those of other grb / sne ( delia et al .
2014 , in preparation ) , we took the average of the @xmath3 of sn2010ah @xcite and sn1998bw based on the spectroscopic similarity . for sn2006aj
@xmath3 was obtained through modelling @xcite .
for sn2010bh it was estimated by @xcite .
all these @xmath3 assume spherical symmetry .
however , we know from the distribution of elements ( in particular fe and o as observed through their nebular emission lines ) that at least sn1998bw was significantly aspherical , and was observed near the direction of most rapid expansion , which is consistent with the detection of the grb @xcite .
therefore , spherically symmetric @xmath3are likely to be overestimated . using 2d explosion models and 3d radiation transport calculations , @xcite and @xcite found that the real sn @xmath3 may be a factor of 2 - 5 smaller .
this correction , which is not shown in fig .
1 , would cause the six grb / sne to cluster around @xmath3@xmath62erg .
the @xmath3 of the two xrf / sne do not require a correction , because there is no evidence for asymmetry in sn2006aj @xcite .
we can now compare the various energies . in fig .
1a the collimation - corrected grb energy , @xmath15 , is compared to the sn @xmath3 , not corrected for asphericity .
grb @xmath13values range over 6 orders of magnitude .
@xmath15 values still cover 3 orders of magnitude , and are always significantly smaller than any sn @xmath3 ( the diagonal line is @xmath3 @xmath63 ) .
this suggests that the grb jet is unlikely to be the driving phenomenon behind grb / sne , as the sn carries most of the energy , as already noted by @xcite .
1b shows @xmath18vs the sn @xmath3 .
again , the sn energy is always much larger . finally , fig .
1c shows the @xmath18v .
@xmath15 . the grb energies estimated from the jet break and from the radio , which rely on different wavebands and on observations taken at completely different times , are in general agreement , but the sn @xmath3 are much larger than both .
this confirms that @xmath18is a good proxy for @xmath15 , but not for either the sn @xmath3or the total energy of the event . [
cols="^,^,^ " , ]
well - studied grb / sne have a roughly standard energy , @xmath3@xmath24(1 - 2)@xmath64erg , if account is taken for asphericity .
xrf / sne have a smaller @xmath3by about a factor of 10 .
all sn @xmath3 are much larger than all of the grb / xrf @xmath15 , which seem to be capped at a few @xmath19erg .
this suggests that the grb / sn phenomenon is driven by the sn , not the grb jet .
this evidence challenges a picture in which the relativistic jet explodes the star .
simulations @xcite have shown that the jet can unbind the star , but it is not clear how the jet can transfer the required large @xmath3 to the star once it has escaped without @xmath15 increasing to unreasonably large values .
also , it may be less natural for grb jets to produce sne with always the same total @xmath3 and the same amount of @xmath6ni to such a degree @xcite .
on the other hand , the energetics of grb / sne are strikingly similar to the rotational energy of a millisecond magnetar .
the rotational energy of a rapidly rotating ns is @xmath65 , where @xmath66 is the ns mass , @xmath67 its radius and @xmath68 its spin period .
rotational energy can be tapped by rapid spindown on a grb timescale if the magnetic field is in the magnetar range ( @xmath69 g ; * ? ? ?
* ; * ? ? ?
this leads us to propose that in grb / sne the exploding star gives birth to a highly magnetized millisecond ns .
deposited magnetar energy can further energize the sn , and @xmath3 @xmath70erg is a limit to the total intrinsic energy of grb / sne .
this limit has not been violated by any grb / sn so far , if they are all aspherical .
magnetar outflows can be focussed into magnetic jets by interaction with the stellar envelope because hoop stress tends to collimate the flow after it comes into pressure equilibrium with the shocked stellar cavity from which the magnetar formed .
the collimated magnetic wind ( which is sometimes called a `` magnetic tower '' and is similar to the collimation of pulsar wind nebulae such as the crab nebula ) can burrow its way out of the star .
a very small fraction of the total energy is seen to emerge in the relativistic jet .
if a large fraction of the magnetar energy can be transferred to the progenitor star , mostly near the jet axis @xcite , it can be added to the sn energy .
the energy deposited also contributes to increasing the isotropic component of the sn @xmath3
the sn can take on an increasingly aspherical shape the higher the energy contribution from the magnetar ( grb / sne are more aspherical than xrf / sne , * ? ? ?
in this scenario , @xmath6ni may be produced as the expanding magnetar wind shocks the inner star .
if this happens quasi - spherically , before the star expands too much , sufficient material can be shocked to produce the several @xmath71 of @xmath6ni in an almost spherical distribution required by grb / sn light curves @xcite . the collimated magnetic wind may produce some more @xmath6ni at high velocities , as also required by the rapid rise of grb / sn light curves .
the late - time deposition of magnetar energy may also contribute to the sn light curve , along with @xmath6ni .
late - time spectra of more grb / sne would be necessary to clarify how much @xmath6ni is actually produced through the observation of emission lines of fe . presently this information is only available for the nearest event , sn1998bw @xcite .
the range of grb prompt emission energy could be produced by interaction of the jet as it propagates through the stellar envelope .
a range of several orders of magnitude in @xmath15 may be possible , since the jet may be slowed down to variable degrees by the development of instabilities or by interaction with extended outer layers of the star .
small amounts of baryons mixed into the jet can pollute " it and reduce its @xmath12-ray luminosity .
an extended envelope may even block the jet altogether @xcite .
magnetars have been proposed to energise x - ray flashes ( xrf ) and their associated sneic @xcite .
xrf / sne have less extreme properties than grb / sne , in particular they have smaller @xmath3 ( a few @xmath9erg ) , luminosities [ m(@xmath6ni)@xmath72 , only marginally larger than in ordinary core - collapse sne ] , and progenitor masses ( @xmath73 , * ? ? ?
they are less aspherical than grb / sne @xcite
. they may be the result of lower - spin magnetars .
the progenitors of grb / sne are thought to be stars of m@xmath74 .
if grb / sne are also powered by magnetars then at least some of these stars also collapse to ns .
, if the remnant is a ns instead masses may have to be revised downwards slightly . ] since grbs and xrfs exhibit a continuum of properties , this picture reconciles their appearance with their origin as a single mechanism .
indeed , @xcite find that jets are always produced when a proto - ns is formed , if the magnetic field is very high .
direct collapse to a bh may not necessarily lead to a luminous sn .
the @xmath6niproduced by the disk wind @xcite could be highly variable and may accrete into the bh , in the spirit of the initial proposal of a failed sn " @xcite .
this may be the case of the 2 low - redshift grbs , 060614 and 060505 , which showed no sn down to m(@xmath6ni ) @xmath75 @xcite .
fallback of @xmath6ni onto the bh is one possibility @xcite . on the other hand ,
both of these grbs have @xmath15 well below the magnetar limit .
magnetars have also been proposed as the energy source for grbs @xcite , for grb / sne and luminous sneib / c @xcite and for the peculiar snib 2005bf @xcite .
@xcite find that most grbs are compatible with being energised by a magnetar . on the other hand ,
@xcite find that the energetics of three out of five well - observed high - redshift swift grbs have energies similar to the maximum energy provided by a spinning ns ( @xmath0erg ) even after correction for collimation .
only one of these , grb080319b , may show a bump in its light curve , but any sn would be very faint @xcite
. these may indeed all be collapsars .
we checked all other grbs , at any redshift , for which a sn was reported .
their @xmath13 almost ever exceeds a few @xmath0 erg .
five ( 991208 , 000911 , 011121 , 020405 , 090618 ) have @xmath13@xmath76erg @xcite , but in four of these the optical afterglows exhibit potential jet breaks , leading to substantial energy collimation corrections @xcite .
no breaks are reported in the optical afterglow of grb000911 @xcite , but the light curve could admit a break at t@xmath245 rest - frame days , leading to an energy collimation correction factor of at least 100 . for grb111211 @xcite , no @xmath12-ray energy or fluence is available . from the grb peak flux ( @xmath77 ergs@xmath78@xmath79 in 20 - 60 kev ) , duration ( 15 s ) and
redshift ( @xmath80 , * ? ? ?
* ) we estimate @xmath13@xmath81erg .
all these events may be driven by magnetars .
we thank the yukawa institute of kyoto university and the national astronomical observatory of japan for the hospitality while some of this work was carried out .
ep acknowledges support from asi inaf i/088/06/0 , inaf prin 2011 and prin miur 2010/2011 .
sew has been supported by the nasa theory program ( nnx09ak36 g ) and the high energy physics program of the doe ( de - sc0010676 ) . | the kinetic energy of supernovae ( sne ) accompanied by gamma - ray bursts ( grbs ) tends to cluster near @xmath0erg , with @xmath1erg an upper limit to which no compelling exceptions are found ( assuming a certain degree of asphericity ) , and it is always significantly larger than the intrinsic energy of the grb themselves ( corrected for jet collimation ) .
this energy is strikingly similar to the maximum rotational energy of a neutron star rotating with period 1 ms .
it is therefore proposed that all grbs associated with luminous sne are produced by magnetars .
grbs that result from black hole formation ( collapsars ) may not produce luminous sne .
x - ray flashes ( xrfs ) , which are associated with less energetic sne , are produced by neutron stars with weaker magnetic field or lower spin .
[ firstpage ] supernovae : general stars : magnetars gamma - ray bursts : general |
one of the most attractive physical systems for generating large entangled states and realizing a quantum computer @xcite is a collection of cold trapped atomic ions @xcite .
the ion trap quantum computer stores one or more quantum bits ( qubits ) in the internal states of each trapped ion , and quantum logic gates ( implemented by interactions with externally applied laser beams ) can couple qubits through a collective quantized mode of motion of the ion coulomb crystal .
loss of coherence of the internal states of trapped ions is negligible under proper conditions but heating of the motion of the ion crystal may ultimately limit the fidelity of logic gates of this type .
in fact , such heating is currently a limiting factor in the nist ion - trap quantum logic experiments @xcite .
+ + electric fields from the environment readily couple to the motion of the ions , heating the ion crystal @xcite .
if the ion trap is much larger than the ion crystal size , we expect these electric fields to be nearly uniform across the crystal .
uniform fields will heat only modes that involve center - of - mass motion ( com motion ) , in which the crystal moves as a rigid body .
motional modes orthogonal to the com motion , for instance the collective breathing mode , require field gradients to excite their motion .
the heating of these modes is therefore suppressed @xcite . however , even if quantum logic operations use such a `` cold '' mode , the heating of the com motion can still indirectly limit the fidelity of logic operations .
since the laser coupling of an internal qubit and a motional mode depends on the total wavepacket spread of the ion containing the qubit , the thermal com motion can reduce the logic fidelity @xcite .
+ + in this paper , we examine sympathetic cooling @xcite in a particular scheme for which we can continuously laser - cool the com motion while leaving undisturbed the coherences of both the internal qubits and the mode used for quantum logic . in this method ,
one applies continuous laser cooling to only the center ion of a coulomb - coupled string of an odd number of ions .
one can address the center ion alone if the center ion is of a different ion species than that composing the rest of the string @xcite .
alternatively , one can simply focus the cooling beams so that they affect only the center ion . in either case
, the cooling affects only the internal states of the center ion , leaving all other internal coherences intact .
if the logic operations use a mode in which the center ion remains at rest , the motional coherences in that mode are also unaffected by the cooling . on the other hand ,
the sympathetic cooling keeps the com motion cold , reducing the thermal wavepacket spread of the ions . in the following
, we will discuss the dynamics of an ion string in which all ions are identical except the center ion , assuming heating by a uniform electric field .
our results give guidelines for implementing the sympathetic cooling scheme .
similar results would apply to two- and three - dimensional ion crystals @xcite .
we consider a crystal of @xmath0 ions , all of charge @xmath1 , in a linear rf trap @xcite .
the linear rf trap is essentially an rf quadrupole mass filter with a static confining potential along the filter axis @xmath2 .
if the radial confinement is sufficiently strong compared to the axial confinement , the ions will line up along the @xmath3-axis in a string configuration @xcite .
there is no rf electric field along @xmath2 , so we can write the axial confining potential as @xmath4 for @xmath5 a constant .
the potential energy of the string is then given by @xmath6 for @xmath7 the position of the @xmath8th ion in the string ( counting from the end of the string ) .
the first term in the potential energy expresses the influence of the static confining potential along the @xmath3-axis , while the second arises from the mutual coulomb repulsion of the ions . for a single ion of mass @xmath9 ,
the trap frequency along @xmath3 is just @xmath10 . + + we can compute the equilibrium positions of the ions in the string by minimizing the potential energy of eq . [
eq : pot ] . defining a length scale @xmath11 by @xmath12 and normalizing the ion positions by @xmath13
gives a set of equations for the @xmath14 as @xmath15 which has analytic solutions only up to @xmath16 .
steane @xcite and james @xcite have computed the equilibrium positions of ions in strings with @xmath0 up to 10 .
the potential energy is independent of the mass , so the equilibrium positions of ions in a string are independent of the elemental composition of the string if all the ions have the same charge .
+ + in a real ion trap the ions will have some nonzero temperature and will move about their equilibrium positions . if the ions are sufficiently cold , we can write their positions as a function of time as @xmath17 , where @xmath18 is small enough to allow linearizing all forces .
we specialize to the case of an odd number of ions @xmath0 , where all ions have mass @xmath9 , except for the one at the center of the string which has mass @xmath19 .
the ions are numbered @xmath20 , with the center ion labeled by @xmath21 .
following james @xcite , the lagrangian for the resulting small oscillations is @xmath22 where @xmath23 + we define a normalized time as @xmath24 . in treating the case of two ion species ,
we write @xmath25 for the mass ratio of the two species and normalize the amplitude of the ion vibrations @xmath18 as @xmath26 , @xmath27 , @xmath28 .
the lagrangian becomes @xmath29 where @xmath30 generalizing the result of james @xcite .
+ the lagrangian is now cast in the canonical form for small oscillations in the coordinates @xmath31 . to find the normal modes , we solve the eigenvalue equation @xmath32 for the frequencies @xmath33 and ( orthonormal ) eigenvectors @xmath34 of the @xmath0 normal modes .
because of our normalization of the lagrangian ( [ eq : axl ] ) , the @xmath33 are normalized to @xmath35 and the @xmath34 are expressed in terms of the normalized coordinates @xmath31 . in terms of the physical time
@xmath36 , the frequency of the @xmath37th mode is @xmath38 .
if the @xmath37th mode is excited with an amplitude @xmath39 , we have @xmath40 \hspace{12pt } i\not = n_c\\ q_{n_c}(t ) & = & { \rm re}\:[c\frac{1}{\sqrt{\mu } } v_{n_c}^{(k ) } e^{i(\zeta_k\omega_zt+\phi_k)}]\end{aligned}\ ] ] in terms of the physical coordinates @xmath18 .
+ + we can solve for the normal modes analytically for @xmath16 .
exact expressions for the normal - mode frequencies are @xmath41^{\frac{1}{2}}\\ \zeta_2 & = & \sqrt{3}\\ \zeta_3 & = & \left [ \frac{13}{10}+\frac{1}{10\mu}(21 + \sqrt{441 - 34\mu+169\mu^2 } ) \right]^{\frac{1}{2}}\end{aligned}\ ] ] normalized to @xmath35 .
the mode eigenvectors are @xmath42 in terms of @xmath31 . here
@xmath43 and @xmath44 are normalization factors . in the case of three identical ions ( @xmath45 )
, we can express the mode eigenvectors in terms of the @xmath31 as @xmath46 , @xmath47 , and @xmath48 .
the mode eigenvectors , in this special case , also give the ion oscillation amplitudes in terms of the physical coordinates @xmath18 .
for three identical ions , then , pure axial com motion constitutes a normal mode .
( this result holds for an arbitrary number of identical ions . )
we also note that the center ion does not move in mode # 2 ; hence the frequency and eigenvector of mode # 2 are independent of @xmath49 . for any odd number @xmath0 of ions
there are @xmath50 modes for which the center ion does not move .
these modes will likewise have frequencies and eigenvectors independent of @xmath49 .
moreover , they have @xmath51 and so they are orthogonal to the com motion and do not couple to uniform electric fields
. the center ion moves in the other @xmath52 modes , and unless @xmath45 , each of these @xmath52 modes has a component of axial com motion and therefore couples to uniform electric fields .
+ + for @xmath53 and higher , the normal mode frequencies depend on @xmath49 in a complicated way . however , it is easy to find the frequencies numerically .
1 shows the mode frequencies for @xmath54 , 5 , 7 , and 9 as a function of @xmath49 for @xmath55 .
the modes are numbered in order of increasing frequency ( at @xmath45 ) , and are normalized to @xmath35 . in each case
, the lowest - lying mode has all ions moving in the same direction and consists of pure com motion for @xmath45 .
the even - numbered modes correspond to the @xmath50 modes for which the center ion does not move .
their frequencies are therefore independent of @xmath49 . for both very large and very small @xmath49 the modes pair up , as shown in fig .
1 . for each pair
there is some value @xmath56 for which the modes become degenerate .
the relative spacing between modes in a pair is also smaller in the large-@xmath49 limit than in the small-@xmath49 limit .
+ + in selecting a normal mode of motion for logic operations , we want to ensure that the mode is well resolved from all other normal modes .
however , when modes are nearly degenerate , as for @xmath57 and @xmath58 , transfer of energy can occur between the modes in the presence of an appropriate coupling , for instance if the static confining potential contains small terms of order @xmath59 @xcite .
this coupling can lead to a loss of coherence of the logic mode .
also , the need to resolve the logic mode from a nearby spectator mode can force a reduction in gate speed .
these effects limit the usefulness of the sympathetic cooling scheme for @xmath49 very large .
evidently it is best to use a cooling ion that is of the same mass or lighter than the logic ions . in this case mode
# 2 is well - separated from all other modes , as shown in fig .
we now consider the motion of the ions transverse to the @xmath3-axis . the ions experience an rf potential @xmath60 for a suitable choice of axes @xmath61 and @xmath62 perpendicular to @xmath3 , where @xmath63 is the frequency of the rf field and @xmath64 is a constant .
the static confining potential can be written @xmath65 at the position of the ions ( with @xmath66 a constant ) , so there is also a transverse static electric field .
to analyze the ion motion , we work in the pseudopotential approximation , in which one time - averages the motion over a period of the rf drive to find the ponderomotive force on the ion . if the static potential is negligible , the rf drive gives rise to an effective transverse confining potential of @xmath67 , where @xmath68 for an ion of mass @xmath9 .
if we include the effects of the static field , the transverse potential becomes @xmath69 , where @xmath70 , @xmath71 .
below we will assume @xmath72 , so that @xmath73 . in any case
, the transverse potential is that of a simple harmonic oscillator , as we saw also for the axial potential .
however , the transverse potential depends directly on the ion s mass , so the center ion of a string feels a different trap potential than the others for @xmath74 . + + we define @xmath75 , so that @xmath76 . then the normalized lagrangian for the motion along @xmath61 is @xmath77 where @xmath78 for @xmath79 and @xmath80 are normalized ion vibration amplitudes along @xmath61 . here
@xmath81 and @xmath82 we can describe the normal mode frequencies and oscillation amplitudes in terms of the eigenvectors and eigenvalues of @xmath83 , just as for the axial case above .
the normalizations of the time and position coordinates remain the same as in the axial case .
+ + in the previous section , we assumed that the radial confinement of the ions was strong enough that the configuration of ions in a string along the @xmath3-axis was always stable .
however , for sufficiently small @xmath84 , the string configuration becomes unstable .
the stable configurations for different values of @xmath84 can be calculated @xcite , and several of these configurations have been observed for small numbers of ions @xcite .
rather than review the theory of these configurations , we will simply find the range of validity of our small - oscillation lagrangian for the string configuration .
the string will remain stable for all @xmath84 greater than some @xmath85 ; @xmath86 also varies with @xmath0 . on the boundary between stable and unstable regions , the frequency of
some mode goes to zero . recalling that the determinant of a matrix is equal to the product of its eigenvalues
, we see that @xmath87 is the maximum value of @xmath84 satisfying @xmath88 for @xmath49 fixed .
2 shows @xmath87 as a function of @xmath49 for 3 , 5 , 7 , and 9 ions . in each case
, there is a cusp in @xmath87 corresponding to the crossing of the two largest solutions to @xmath88 .
the position of the cusp varies with the number of ions , but lies between @xmath89 and @xmath45 for @xmath90 .
only the cusp for @xmath16 is clearly visible in fig . 2 , but numerical study indicates the presence of a cusp for all four values of @xmath0 . for @xmath49
greater than the value at the cusp , @xmath91 corresponds to instability of the zigzag mode , so that the string breaks into a configuration in which each ion is displaced in the opposite direction to its neighbors @xcite . for @xmath49
smaller than the value at the cusp , @xmath86 is independent of @xmath49 . in this regime , @xmath92 creates an instability in a mode similar to the zigzag mode , except that the center ion remains fixed .
+ + we can proceed to calculate the frequencies of the transverse modes for values @xmath93 . again , these frequencies are normalized to the axial frequency of a single ion of mass @xmath9 . fig .
3 shows the transverse mode frequencies for 3 , 5 , 7 , and 9 ions as a function of @xmath49 , where @xmath84 is taken equal to @xmath94 .
the modes are numbered in order of increasing frequency at @xmath45 ( all ions identical ) . in this numbering scheme ,
the central ion moves in odd - numbered modes but not in even - numbered modes .
the frequencies of the even - numbered modes appear to depend on @xmath49 because they are calculated at a multiple of @xmath87 ; for constant @xmath84 these frequencies are independent of @xmath49 .
the cusps in the mode frequencies in fig . 3 arise from the cusps of @xmath87 at the crossover points between the two relevant solutions of @xmath95 .
mode frequencies plotted for a constant value of @xmath84 do not exhibit these cusps . as in the case of axial motion ,
the mode frequencies form pairs of one even- and one odd - numbered mode for small @xmath49 . however , for large @xmath49 all but one of the transverse modes become degenerate .
the only nondegenerate transverse mode in this case is the zigzag mode . in general , the modes are most easily resolved from their neighbors for @xmath45 , as in the case of axial motion . increasing @xmath84 reduces the frequency spacing between nearly degenerate modes . at @xmath96 and @xmath45 , for instance , the fractional spacing between the cold transverse mode of 3 ions and its nearest neighbor is 0.20 , but for @xmath97 the same spacing is 0.09 .
+ + the near - degeneracy of the modes for large or small @xmath49 and for @xmath98 significantly greater than 1 limits the usefulness of these modes because of possible mode cross - coupling , just as for the axial modes . resolving a particular transverse mode
requires operating the trap near the point at which the string configuration becomes unstable , i.e. , @xmath84 near @xmath99 . in this regime ,
the collective motion of the ions is quite sensitive to uncontrolled perturbations , which may pose significant technical problems for using a transverse mode in quantum logic operations .
stochastic electric fields present on the ion trap electrodes , for instance from fluctuating surface potentials , can heat the various normal modes of motion incoherently . for ion trap characteristic dimension
@xmath100 much larger than the size of the ion crystal @xmath101 , these fields are approximately uniform across the ion crystal , so they couple only to the com motion .
the @xmath50 even - numbered modes are orthogonal to the com motion , so they are only heated by fluctuating electric field gradients .
the heating rates of these modes are reduced by a factor of at least @xmath102 as compared to the heating of the other modes @xcite . in the following , therefore , we will neglect the effects of fluctuating field gradients , so that the even - numbered modes do not heat at all .
+ + the analysis of sections 2 and 3 shows that the motion of a crystal of @xmath0 ions is separable into the @xmath103 normal modes , each of which is equivalent to a simple harmonic oscillator .
hence we can quantize the crystal motion by quantizing the normal modes .
the @xmath37th normal mode gives rise to a ladder of energy levels spaced by @xmath104 , with @xmath103 such ladders in all .
if we now write the uniform electric field power spectral density as @xmath105 , we can generalize the result of @xcite to give @xmath106 for the heating rate of the @xmath37th mode , expressed in terms of the average number of quanta gained per second . recall that @xmath107 is the oscillation amplitude of the @xmath8th ion in the @xmath37th normal mode , expressed in the normalized coordinates .
it is useful to normalize the heating rate in equation ( [ eq : heat ] ) to the heating rate of the lowest - lying axial mode of a string of identical ions .
this normal mode consists entirely of com motion and we write @xmath108 for all ions .
the normalized heating rate of the @xmath37th mode is then @xmath109 where we have assumed that the spectral density @xmath105 is constant over the frequency range of the normal modes , i.e. , @xmath110 .
+ + fig .
4 shows plots of the normalized heating rates of the axial modes for @xmath54 , 5 , 7 , and 9 as a function of @xmath49 . fig .
5 is the same , but for the transverse modes , with @xmath111 .
the numbering of modes on the plots of heating rate matches the numbering on the corresponding plots of mode frequency ( figs . 1 and 2 )
. + + in both axial - mode and transverse - mode plots , the even - numbered modes have the center ion at rest , while the center ion moves for all odd - numbered modes .
we see from figs . 4 and 5 that the modes for which the center ion is fixed can never heat , while all the other modes always heat to some extent for @xmath74
. we will refer to these modes as cold " and hot " modes , respectively .
if the ions are identical , only the modes with all ions moving with the same amplitude ( com modes ) can heat .
there are three such modes , one along @xmath112 , one along @xmath113 , and one along @xmath2 . in interpreting figs . 4 and 5 , it is important to recall that the normalized heating rate defined in eq .
( [ eq : hnorm ] ) is inversely proportional to the mode frequency .
for instance , the @xmath49-dependence of the heating rate of the highest - frequency transverse mode can be largely ascribed to variations in the mode frequency , rather than to changes in the coupling of the mode to the electric field .
heating reduces logic gate fidelity in two ways .
the logic mode itself can be heated , but by choosing a cold mode , we can render this effect negligible . on the other hand ,
the rabi frequency of the transition between logic - mode motional states depends on the total wavepacket spread of the ion involved in the transition @xcite .
heating on modes other than the logic mode can thus lead to unknown , uncontrolled changes in this rabi frequency , resulting in overdriving or underdriving of the transition .
the purpose of sympathetic cooling is to remove this effect by cooling the center ion and thus all hot modes .
+ + for sympathetic cooling to be useful , we must find a cold mode suitable for use in quantum logic
. the cold mode must be spectrally well separated from any other modes in order to prevent unwanted mode cross - coupling .
we can use the lowest - lying cold axial mode as the logic mode for @xmath114 . in this mode , called the breathing mode , the center ion remains fixed and the spacings between ions expand and contract in unison . unless the trap is operated very close to the instability point of the string configuration , the breathing mode is better separated from its neighbors than are any of the cold transverse modes . for @xmath115 any cold mode , either axial or transverse , is nearly degenerate with a hot mode . in this regime
one must make a specific calculation of mode frequencies in order to find the best - resolved cold mode .
even so , the cold axial modes are again better separated from their neighbors than are the cold transverse modes , except for @xmath84 very close to @xmath87 .
it seems best to select a cold axial mode as the logic mode in most cases .
+ + by selecting our laser - beam geometry appropriately , we can ensure that the rabi frequency of the motional transition on the axial mode used for logic depends chiefly on the spread of the ion wavepacket along @xmath3 . in this case
, heating of the axial modes will affect logic - gate fidelity , but heating of the transverse modes will have little effect .
if the mass of the central ion is nearly the same as that of the others ( @xmath116 ) , only the lowest axial mode will heat significantly , and we can continuously cool this mode by cooling only the central ion , ensuring that all ions remain in the lamb - dicke limit @xcite . if @xmath49 is not near 1 , we must cool all @xmath52 hot modes ( again by addressing the central ion ) to keep all ions in the lamb - dicke limit .
+ + the analysis above indicates that , all other things being equal , we are best off if our substituted ion is identical to , or is an isotope of , the logic ions
. however , sympathetic cooling can still be useful if the two ion species have different masses .
for example , we can consider sympathetic cooling using the species @xmath117be@xmath118 and @xmath119mg@xmath120 .
linear traps constructed at nist have demonstrated axial secular frequencies of over 10 mhz for single trapped @xmath117be@xmath118 ions . for three ions with @xmath119mg@xmath120 as the central ion , @xmath121 mhz yields a spacing of 1.6 mhz between the cold axial breathing mode and its nearest neighbor .
if we reverse the roles of the ions ( @xmath122 mhz ) , the spacing increases to 6.2 mhz .
the transverse modes are much harder to resolve from each other . for three ions with @xmath119mg@xmath120 in the center
, we require @xmath123 mhz to obtain @xmath111 , and the spacing between the cold transverse zigzag mode and its nearest neighbor is only 560 khz . reversing the roles of the ions , we find @xmath111 at @xmath124 mhz with a spacing of 1.1 mhz . for this combination of ion species , the cold axial breathing mode seems most appropriate for logic . for a string of 3 or 5 ions
, sympathetic cooling would require driving transitions on 2 or 3 axial - mode sidebands , respectively .
from this example we see that sympathetic cooling can be useful even for ion mass ratios of nearly 3 to 1 .
we have investigated a particular sympathetic cooling scheme for the case of an ion string confined in a linear rf trap .
we have numerically calculated the mode frequencies of the axial and transverse modes as functions of the mass ratio @xmath49 and trap anisotropy @xmath84 for 3 , 5 , 7 , and 9 ions .
we have also calculated the heating rates of these modes relative to the heating rate of a single ion , assuming that the heating is driven by a uniform stochastic electric field .
the results indicate that the scheme is feasible for many choices of ion species if we use a cold axial mode as the logic mode .
the optimal implementation of the scheme employs two ion species of nearly equal mass .
however , a demonstration of sympathetic cooling using @xmath117be@xmath118 and @xmath119mg@xmath120 appears well within the reach of current experimental technique .
+ + a. steane , appl .
b 64 , 623 ( 1997 ) . j.i .
cirac and p. zoller , phys .
74 , 4091 ( 1995 ) .
d.j . wineland
_ et al_. , j. res .
stand . technol .
103 , 259 ( 1998 ) .
b.e . king _
et al _ , phys .
81 , 1525 ( 1998 ) .
s. k. lamoreaux , phys .
a 56 , 4970 ( 1997 ) .
james , phys .
lett 81 , 317 ( 1998 ) . c. henkel and m. wilkins , quant - ph/9902009 .
et al_. , phys .
57 , 70 ( 1986 ) .
we assume that the cooling transitions of the center ion are well - resolved from all relevant transitions of the logic ions .
et al_. , phys .
a 45 , 6493 ( 1992 ) .
h. walther , adv . at .
31 , 137 ( 1993 ) .
m. drewsen _
et al_. , phys .
81 , 2878 ( 1998 ) .
et al_. , science 282 , 1290 ( 1998 ) .
james , appl .
b 66 , 181 ( 1998 ) . p.k .
ghosh , ion traps ( oxford : clarendon ) , 1995 .
r. rafac _ et al_. , proc .
usa 88 , 483 ( 1991 ) .
schiffer , phys .
70 , 818 ( 1993 ) .
d.j . wineland _ et al_. , fortschr .
46 , 363 ( 1998 ) . | one limit to the fidelity of quantum logic operations on trapped ions arises from heating of the ions collective modes of motion .
sympathetic cooling of the ions during the logic operations may eliminate this source of errors .
we discuss benefits and drawbacks of this proposal , and describe possible experimental implementations .
we also present an overview of trapped - ion dynamics in this scheme . |
let @xmath2 be a locally compact topological group , with haar measure @xmath6 .
a discrete subgroup @xmath7 is a _ lattice _ if @xmath8 carries a finite @xmath2invariant measure , and is _ uniform _ if @xmath8 is compact .
some basic questions are : 1 .
[ q : existence ] does @xmath2 admit a ( uniform or nonuniform ) lattice ? 2 .
[ q : covolumes ] what is the set of _ covolumes _ of lattices in @xmath2 , that is , the set of positive reals @xmath9 3 .
[ q : generation ] are lattices in @xmath2 finitely generated ?
these questions have been well - studied in classical cases .
for example , suppose @xmath2 is a reductive algebraic group over a local field @xmath10 of characteristic @xmath11 .
then @xmath2 admits a uniform lattice , constructed by arithmetic means ( borel harder @xcite ) , and a nonuniform lattice only if @xmath10 is archimedean ( tamagawa @xcite ) .
if @xmath2 is a semisimple real lie group , the set @xmath12 is in most cases discrete ( see @xcite and its references ) . if in addition @xmath2 is higher - rank , then @xmath2 and hence its lattices have kazhdan s property ( t ) ( see , for example , @xcite ) .
since countable groups with property ( t ) are finitely generated , it follows that all lattices in @xmath2 are finitely generated .
a nonclassical case is @xmath2 the automorphism group of a locally finite tree @xmath13 .
the study of lattices in @xmath14 was initiated by bass and lubotzky , and has yielded many surprising differences from classical results ( see the survey @xcite and the reference @xcite ) .
for example , the set @xmath12 is in many cases nondiscrete , and nonuniform tree lattices are never finitely generated .
in fact , the automorphism group @xmath2 of any locally finite polyhedral complex @xmath15 is naturally a locally compact group ( see section [ ss : lattices ] ) . for many @xmath15 with @xmath16 ,
there is greater rigidity than for trees , as might be expected in higher dimensions .
for instance , burger
mozes @xcite proved a ` normal subgroup theorem ' for products of trees ( parallel to that of margulis @xcite for higher - rank semisimple lie groups ) , and bourdon
pajot @xcite and xie @xcite established quasi - isometric rigidity for certain fuchsian buildings . on the other hand , lattices in @xmath17
can exhibit the same flexibility as tree lattices .
for example , the set @xmath12 is nondiscrete for certain right - angled buildings @xcite and fuchsian buildings @xcite .
another example is density of commensurators of uniform lattices in @xmath2 , proved by haglund @xcite for certain @xmath5dimensional davis complexes , and by haglund @xcite and kubena barnhill
thomas @xcite for right - angled buildings .
apart from right - angled buildings , very little is known for @xmath15 of dimension @xmath18 .
almost nothing is known for @xmath15 not a building . in this paper
we consider questions above for lattices in @xmath19 , where @xmath0 is the davis complex for a coxeter system @xmath1 ( see @xcite and section [ ss : davis_complexes ] below ) .
the davis complex is a locally finite , piecewise euclidean @xmath20 polyhedral complex , and the coxeter group @xmath3 may be regarded as a uniform lattice in @xmath2 .
our results are the main theorem and its corollaries [ c : nondiscreteness ] and [ c : infinite_generation ] below , which establish tree - like properties for lattices in many such @xmath2 .
after stating these results , we discuss how they apply to ( barycentric subdivisions of ) buildings and non - buildings , and to many davis complexes @xmath0 with @xmath21 . to state the main theorem , recall that for a coxeter system @xmath1 with @xmath22 , and any @xmath23 , the _ special subgroup _ @xmath24 is the subgroup of @xmath3 generated by the elements @xmath25 .
a special subgroup @xmath24 is _ spherical _ if it is finite , and the set of spherical special subgroups of @xmath3 is partially ordered by inclusion .
the poset of nontrivial spherical special subgroups is an abstract simplicial complex @xmath26 , called the _
nerve _ of @xmath1 .
we identify each generator @xmath27 with the corresponding vertex @xmath28 of @xmath26 , and denote by @xmath29 the group of _ label - preserving automorphisms _ of @xmath26 , that is , the group of automorphisms @xmath30 of @xmath26 such that @xmath31 for all @xmath32 .
the group @xmath19 is nondiscrete if and only if there is an @xmath33 such that @xmath30 fixes the star in @xmath26 of some vertex @xmath34 , and @xmath35 ( haglund paulin @xcite ) .
[ t : existence ] let @xmath1 be a coxeter system , with nerve @xmath26 and davis complex @xmath0 .
let @xmath29 be the group of label - preserving automorphisms of @xmath26 .
assume that there are vertices @xmath36 and @xmath37 of @xmath26 , and elements @xmath38 of prime order , such that for @xmath39 : 1 .
[ c : fix ] @xmath40 fixes the star of @xmath41 in @xmath26 , and @xmath42 ; 2 .
[ c : orbit ] for all @xmath43 such that @xmath44 is the image of @xmath45 under some power of @xmath46 , @xmath47 ; and 3 .
[ c : halvable ] all spherical special subgroups @xmath24 with @xmath48 are _ halvable along @xmath45 _
( see definition [ d : halvable ] below ) . then @xmath19 admits : * a nonuniform lattice @xmath4 ; and * an infinite family of uniform lattices @xmath49 , such that @xmath50 , where @xmath6 is haar measure on @xmath2 .
[ c : nondiscreteness ] the set of covolumes of lattices in @xmath2 is nondiscrete . [
c : infinite_generation ] the group @xmath2 admits a lattice which is not finitely generated .
corollary [ c : infinite_generation ] follows from the proof of the main theorem and theorem [ t : group_action_intro ] below . by the discussion above , corollary [ c : infinite_generation ]
implies that the group @xmath2 in the main theorem does not have property ( t ) .
this was already known for @xmath19 , where @xmath0 is any davis complex ( haglund paulin @xcite ) ; our results thus provide an alternative proof of this fact in some cases .
we describe several infinite families of examples of davis complexes @xmath0 to which our results apply in section [ s : examples ] below . to establish these applications , we use properties of spherical building ( in @xcite ) , and some results of graph theory @xcite . in two dimensions ,
examples include the fuchsian buildings considered in @xcite , and some of the highly symmetric _ platonic polygonal complexes _ investigated by witkowski @xcite .
platonic polygonal complexes are not in general buildings , and even the existence of lattices ( other than @xmath3 ) in their automorphism groups was not previously known .
an example of a platonic polygonal complex is the ( unique ) @xmath20 @xmath5complex with all @xmath5cells squares , and the link of every vertex the petersen graph ( figure [ f : petersen ] below ) .
the main theorem and its corollaries also apply to many higher - dimensional @xmath0 , including both buildings and non - buildings . to prove the main theorem
, we construct the lattices @xmath51 and @xmath4 as fundamental groups of complexes of groups with universal covers @xmath0 ( see @xcite and section [ ss : complexes_of_groups ] below ) .
the construction is given in section [ s : proof ] below , where we also prove corollary [ c : infinite_generation ] .
complexes of groups are a generalisation to higher dimensions of graphs of groups .
briefly , given a polyhedral complex @xmath52 , a ( simple ) complex of groups @xmath53 over @xmath52 is an assignment of a _ local group _
@xmath54 to each cell @xmath55 of @xmath52 , with monomorphisms @xmath56 whenever @xmath57 , so that the obvious diagrams commute .
the action of a group @xmath2 on a polyhedral complex @xmath15 induces a complex of groups @xmath53 over @xmath58 .
a complex of groups is _ developable _ if it is isomorphic to a complex of groups constructed in this way .
a developable complex of groups @xmath53 has a simply - connected _ universal cover _
@xmath59 , equipped with a canonical action of the _ fundamental group of the complex of groups _
@xmath60 .
a key difference from graphs of groups is that complexes of groups are not in general developable .
in addition , even if @xmath53 is developable , with universal cover say @xmath15 , if @xmath61 it may be impossible to identify @xmath15 using only local data such as the links of its vertices ( see ballmann
brin @xcite and haglund @xcite ) . to ensure that our complexes of groups are developable with universal covers @xmath0
, we use covering theory for complexes of groups ( see @xcite and @xcite , and section [ ss : definitions ] below ) . the main result needed is that if there is a covering of complexes of groups @xmath62 , then @xmath53 is developable if and only if @xmath63 is developable , and the universal covers of @xmath53 and @xmath63 are isometric ( see theorem [ t : coverings ] below ) .
the other main ingredient in the proof of the main theorem is theorem [ t : group_action_intro ] below , which introduces a theory of group actions on complexes of groups " .
this is a method of manufacturing new complexes of groups with a given universal cover , by acting on previously - constructed complexes of groups .
given a complex of groups @xmath53 , and the action of a group @xmath64 on @xmath52 , the @xmath64action _ extends to an action on @xmath53 _ if there is a homomorphism from @xmath64 to @xmath65 .
roughly , this means that for each cell @xmath55 of @xmath52 , each @xmath66 induces a group isomorphism @xmath67 , so that the obvious diagrams commute ( see section [ ss : definitions ] below for definitions ) . in section [ s : group_actions ] below we prove : [ t : group_action_intro ] let @xmath53 be a ( simple ) complex of groups over @xmath52 , and suppose that the action of a group @xmath64 on @xmath52 extends to an action on @xmath53 .
then the @xmath64action induces a complex of groups @xmath63 over @xmath68 such that there is a covering of complexes of groups @xmath62 .
moreover the fundamental group @xmath69 splits as @xmath70 .
theorem [ t : group_action_intro ] is also used in @xcite , and we expect this result to be of independent interest . to our knowledge , automorphisms of complexes of groups have not previously been considered . in @xcite , bass jiang determined the structure of the full automorphism group of a graph of groups , but did not define or study the graph of groups induced by a group action on a graph of groups . a more precise statement of theorem [ t : group_action_intro ] , including some additional facts about @xmath63 ,
is given as theorem [ t : group_action ] below .
the main theorem is then proved as follows .
the action of the coxeter group @xmath3 on @xmath0 induces a complex of groups @xmath71 over @xmath72 , with local groups the spherical special subgroups of @xmath3 .
we construct a family of finite complexes of groups @xmath73 and @xmath74 , and infinite complexes of groups @xmath75 and @xmath76 , so that there are coverings of complexes of groups as sketched in figure [ f : coverings ] below .
the fundamental groups of @xmath74 and @xmath76 are , respectively , the uniform lattices @xmath51 , and the nonuniform lattice @xmath4 , in @xmath19 . for the local groups of @xmath73 and @xmath75
, we use condition in the main theorem to replace certain spherical special subgroups @xmath24 by the subgroup @xmath77 , defined as follows : [ d : halvable ] let @xmath24 be a spherical special subgroup of @xmath3 , and suppose @xmath25 .
then @xmath24 is _ halvable along @xmath34 _ if the union @xmath78 generates an index @xmath5 subgroup , denoted @xmath77 , of @xmath24 . the complexes of groups @xmath74 and @xmath76 are induced by group actions on , respectively , @xmath73 and @xmath75 .
to construct these group actions , we use the automorphisms @xmath79 and @xmath80 of @xmath26 .
i am grateful to benson farb for introducing me to these questions , and for his continuing encouragement and advice .
i also thank g. christopher hruska and kevin wortman for many useful discussions .
this particular work was inspired by conversations with tadeusz januszkiewicz and jacek witkowski , and much of this project was carried out at the mathematical sciences research institute , where i benefited from talking with angela kubena barnhill , michael w. davis , jonathan p. mccammond and damian osajda .
i would also like to thank karen vogtmann for helpful comments on this manuscript .
in this section we present brief background . in section [ ss : lattices ] we describe the natural topology on @xmath2 the group of automorphisms of a locally finite polyhedral complex @xmath15 , and characterise uniform and nonuniform lattices in @xmath2 .
section [ ss : davis_complexes ] constructs the davis complex @xmath0 for a coxeter system @xmath1 , following @xcite . in section [ ss : complexes_of_groups ] we recall the basics of haefliger s theory of complexes of groups ( see @xcite ) , and describe the complex of groups
@xmath71 induced by the action of @xmath3 on @xmath0 .
let @xmath2 be a locally compact topological group .
we will use the following normalisation of haar measure @xmath6 on @xmath2 .
[ t : scovolumes ] suppose that a locally compact group @xmath2 acts on a set @xmath81 with compact open stabilisers and a finite quotient @xmath82 .
then there is a normalisation of the haar measure @xmath6 on @xmath2 , depending only on the choice of @xmath2set @xmath81 , such that for each discrete subgroup @xmath83 of @xmath2 we have @xmath84 suppose @xmath15 is a connected , locally finite polyhedral complex .
let @xmath17 . with the compact - open topology ,
@xmath2 is naturally a locally compact topological group , and the @xmath2stabilisers of cells in @xmath15 are compact and open . hence if @xmath85 is finite , there are several natural choices of sets @xmath81 in theorem [ t : scovolumes ] above . by the same arguments as for tree lattices ( @xcite , chapter 1 )
, it can be shown ( for any suitable set @xmath81 ) that a discrete subgroup @xmath86 is a lattice if and only if the series @xmath87 converges , and @xmath4 is uniform if and only if this is a sum with finitely many terms . in section [ ss : davis_complexes ] below we describe our choice of @xmath2set @xmath81 when @xmath2 is the group of automorphisms of a davis complex @xmath0 .
in this section we recall the construction of the davis complex for a coxeter system .
we follow the reference @xcite .
a _ coxeter group _ is a group @xmath3 with a finite generating set @xmath88 and presentation of the form @xmath89 with @xmath90 an integer @xmath91 or @xmath92 , meaning that @xmath93 has infinite order .
the pair @xmath1 is called a _ coxeter system_.
this example will be followed throughout this section , and also referred to in sections [ ss : complexes_of_groups ] and [ s : proof ] below .
let @xmath94 @xmath95 where @xmath96 and @xmath97 are integers @xmath91 .
let @xmath1 be a coxeter system .
a subset @xmath13 of @xmath88 is _ spherical _ if the corresponding special subgroup @xmath24 is spherical , that is , @xmath24 is finite . by convention , @xmath98 is the trivial group .
denote by @xmath99 the set of all spherical subsets of @xmath88 .
the set @xmath99 is partially ordered by inclusion , and the poset @xmath100 is the nerve @xmath26 of @xmath1 ( this is equivalent to the description of @xmath26 in the introduction ) . by definition ,
a nonempty set @xmath13 of vertices of @xmath26 spans a simplex @xmath101 in @xmath26 if and only if @xmath13 is spherical .
we identify the generator @xmath27 with the vertex @xmath102 of @xmath26 .
the vertices @xmath34 and @xmath103 of @xmath26 are joined by an edge in @xmath26 if and only if @xmath90 is finite , in which case we label this edge by the integer @xmath90 .
the nerve @xmath26 of example 1 above , with its edge labels , is sketched in figure [ f : nerve ] below . of example 1 , with edge labels ]
we denote by @xmath10 the geometric realisation of the poset @xmath99 .
equivalently , @xmath10 is the cone on the barycentric subdivision of the nerve @xmath26 of @xmath1 .
note that @xmath10 is compact and contractible , since it is the cone on a finite simplicial complex .
each vertex of @xmath10 has _ type _ a spherical subset of @xmath88 , with the cone point having type @xmath104 .
for example 1 above , @xmath10 and the types of its vertices are sketched on the left of figure [ f : k ] . and
types of vertices ( left ) and mirrors ( right ) for example 1 ] for each @xmath27 let @xmath105 be the union of the ( closed ) simplices in @xmath10 which contain the vertex @xmath34 but not the cone point .
in other words , @xmath105 is the closed star of the vertex @xmath34 in the barycentric subdivision of @xmath26 .
note that @xmath105 and @xmath106 intersect if and only if @xmath90 is finite .
the family @xmath107 is a _ mirror structure _ on @xmath10 , meaning that @xmath108 is a family of closed subspaces of @xmath10 , called _
mirrors_. we call @xmath105 the _ @xmath34mirror _ of @xmath10 . for example 1 above , the mirrors @xmath109 are shown in heavy lines on the right of figure [ f : k ] . for each @xmath110 , put @xmath111 now define an equivalence relation @xmath112 on the set @xmath113 by @xmath114 if and only if @xmath115 and @xmath116 .
the _ davis complex _
@xmath0 for @xmath1 is then the quotient space : @xmath117 the types of vertices of @xmath10 induce types of the vertices of @xmath0 , and the natural @xmath3action on @xmath113 descends to a type - preserving action on @xmath0 , with compact quotient @xmath10 , so that the @xmath3stabiliser of a vertex of @xmath0 of type @xmath118 is a conjugate of the spherical special subgroup @xmath24 .
we identify @xmath10 with the subcomplex @xmath119 of @xmath0 , and write @xmath120 for the translate @xmath121 , where @xmath122 .
any @xmath120 is called a _ chamber _ of @xmath0 .
the mirrors @xmath105 of @xmath10 , or any of their translates by elements of @xmath3 , are called the _ mirrors _ of @xmath0 .
two distinct chambers of @xmath0 are _ @xmath34adjacent _
if their intersection is an @xmath34mirror , and are _ adjacent _ if their intersection is an @xmath34mirror for some @xmath27 .
note that the chambers @xmath120 and @xmath123 are @xmath34adjacent if and only if @xmath124 , equivalently @xmath125 and @xmath126 .
for example 1 above , part of the davis complex @xmath0 for @xmath1 is shown in figure [ f : davis ] .
there are @xmath127 copies of @xmath10 glued around the vertices of types @xmath128 , for @xmath129 , since @xmath130 has order @xmath127 . similarly , there are @xmath131 copies of @xmath10 glued around the vertices of types @xmath132 , for @xmath129 .
the davis complex @xmath0 may be metrised with a piecewise euclidean structure , such that @xmath0 is a complete @xmath20 space ( moussong , see theorem 12.3.3 of @xcite ) . from now on we will assume that @xmath0 is equipped with this metric .
, for example 1 ] suppose that @xmath19 is the group of automorphisms of a davis complex @xmath0 .
since @xmath3 acts cocompactly on @xmath0 , with finite stabilisers , it may be regarded as a uniform lattice in @xmath2 .
we take as the set @xmath81 in theorem [ t : scovolumes ] above the set of vertices of @xmath0 of type @xmath104 .
then the covolume of @xmath3 is @xmath133 , since @xmath3 acts simply transitively on @xmath81 .
we now outline the basic theory of complexes of groups .
the definitions of the more involved notions of morphisms and coverings of complexes of groups are postponed to section [ ss : definitions ] below .
all references to @xcite in this section are to chapter iii.@xmath134 . in the literature , a complex of groups @xmath53
is constructed over a space or set @xmath52 belonging to various different categories , including simplicial complexes , polyhedral complexes , or , most generally , _
scwols _ ( small categories without loops ) . in each case
there is a set of vertices , and a set of oriented edges with a composition law .
the formal definition is : [ d : scwol ] a _ scwol _
@xmath15 is the disjoint union of a set @xmath135 of vertices and a set @xmath136 of edges , with each edge @xmath137 oriented from its initial vertex @xmath138 to its terminal vertex @xmath139 , such that @xmath140 .
a pair of edges @xmath141 is _ composable _ if @xmath142 , in which case there is a third edge @xmath143 , called the _ composition _ of @xmath137 and @xmath144 , such that @xmath145 , @xmath146 , and if @xmath147 and @xmath148 then @xmath149 ( associativity ) .
we will always assume scwols are _ connected _ ( see section 1.1 , @xcite ) .
[ d : action_on_scwol ] an _ action _ of a group @xmath2 on a scwol @xmath15 is a homomorphism from @xmath2 to the group of automorphisms of the scwol ( see section 1.5 of @xcite ) such that for all @xmath150 and all @xmath151 : 1 .
@xmath152 ; and 2 .
[ i : no_inversions ] if @xmath153 then @xmath154 .
condition in this definition ensures that if the scwol @xmath15 is the barycentric subdivision of a polyhedral complex @xmath52 , the action of @xmath2 on @xmath15 corresponds to an action _ without inversions _ on @xmath52 , meaning that if @xmath151 fixes a cell @xmath55 of @xmath52 setwise , then @xmath155 fixes @xmath55 pointwise .
suppose @xmath0 is the davis complex for a coxeter system @xmath1 , as defined in section [ ss : davis_complexes ] above .
recall that each vertex @xmath156 has type @xmath13 a spherical subset of @xmath88 .
the edges @xmath157 are then naturally oriented by inclusion of type .
that is , if the edge @xmath137 joins the vertex @xmath158 of type @xmath13 to the vertex @xmath159 of type @xmath160 , then @xmath161 and @xmath162 exactly when @xmath163 .
it is clear that the sets @xmath164 and @xmath157 satisfy the properties of a scwol .
moreover , if @xmath52 is a subcomplex of @xmath0 , then the sets @xmath165 and @xmath166 also satisfy definition [ d : scwol ] above . by abuse of notation , we identify @xmath0 and @xmath52 with the associated scwols .
we now define complexes of groups over scwols . a _ complex of groups _ @xmath167 over a scwol @xmath52 is given by : 1 . a group @xmath168 for each @xmath169 , called the _ local group _ at @xmath158 ; 2 . a monomorphism @xmath170 along the edge @xmath137 for each @xmath171 ; and 3 . for each pair of composable edges , a twisting element @xmath172 , such that @xmath173 where @xmath174 is conjugation by @xmath175 in @xmath176 , and for each triple of composable edges @xmath177 the following _ cocycle condition
_ holds @xmath178 a complex of groups is _ simple _ if each @xmath175 is trivial .
this example will be followed throughout this section , and used in the proof of the main theorem in section [ s : proof ] below .
let @xmath1 be a coxeter system with nerve @xmath26 and let @xmath10 be the cone on the barycentric subdivision of @xmath26 , as in section [ ss : davis_complexes ] above .
put @xmath179 , with the orientations on edges discussed above .
we construct a simple complex of groups @xmath71 over @xmath180 as follows .
let @xmath181 .
then @xmath55 has type a spherical subset @xmath13 of @xmath88 , and we define @xmath182 .
all monomorphisms along edges of @xmath180 are then natural inclusions , and all @xmath175 are trivial . for @xmath1 as in example 1 of section [ ss : davis_complexes ] above , the complex of groups @xmath71 is shown in figure [ f : gwk ] below . in this figure ,
@xmath183 and @xmath184 are the dihedral groups of orders @xmath127 and @xmath131 respectively , and @xmath185 is the cyclic group of order @xmath5 . , for example 1 of section [ ss : davis_complexes ] ]
suppose a group @xmath2 acts on a scwol @xmath15 , as in definition [ d : action_on_scwol ] above
. then the quotient @xmath58 also has the structure of a scwol , and the action of @xmath2 induces a complex of groups @xmath53 over @xmath52 ( this construction is generalised in section [ ss : induced_cogs ] below ) .
let @xmath186 with @xmath187 the natural projection . for each @xmath169
, choose a lift @xmath188 such that @xmath189 .
the local group @xmath168 of @xmath53 is then defined to be the stabiliser of @xmath190 in @xmath2 , and the monomorphisms @xmath191 and elements @xmath175 are defined using further choices .
the resulting complex of groups @xmath53 is unique up to isomorphism ( see definition [ d : morphism ] below ) .
a complex of groups is _ developable _ if it is isomorphic to a complex of groups @xmath53 induced , as just described , by such an action .
complexes of groups , unlike graphs of groups , are not in general developable .
the complex of groups @xmath71 above is developable , since it is the complex of groups induced by the action of @xmath3 on @xmath0 , where for each @xmath181 , with @xmath55 of type @xmath13 , we choose @xmath190 in @xmath0 to be the vertex of @xmath192 of type @xmath13 .
let @xmath53 be a complex of groups .
fundamental group of the complex of groups _
@xmath60 is defined so that if @xmath52 is simply connected and @xmath53 is simple , @xmath60 is isomorphic to the direct limit of the family of groups @xmath168 and monomorphisms @xmath191 .
for example , since @xmath193 is contractible and @xmath71 is a simple complex of groups , the fundamental group of @xmath71 is @xmath3 .
if @xmath53 is a developable complex of groups , then it has a _ universal cover _ @xmath59 .
this is a connected , simply - connected scwol , equipped with an action of @xmath60 , so that the complex of groups induced by the action of the fundamental group on the universal cover is isomorphic to @xmath53 .
for example , the universal cover of @xmath71 is @xmath0 .
let @xmath53 be a developable complex of groups over @xmath52 , with universal cover @xmath15 and fundamental group @xmath4 .
we say that @xmath53 is _ faithful _ if the action of @xmath4 on @xmath15 is faithful , in which case we may identify @xmath4 with its image in @xmath194 . if @xmath15 is locally finite , then with the compact - open topology on @xmath194 , by the discussion in section [ ss : lattices ] above the subgroup @xmath195 is discrete if and only if all local groups of @xmath53 are finite .
moreover , if @xmath194 acts cocompactly on @xmath15 , a discrete @xmath196 is a uniform lattice in @xmath194 if and only if @xmath197 is finite , and is a nonuniform lattice if and only if @xmath52 is infinite and the series @xmath87 converges , for some @xmath198 on which @xmath17 acts according to the hypotheses of theorem [ t : scovolumes ] above .
in this section we introduce a theory of group actions on complexes of groups .
the main result is theorem [ t : group_action ] below , which makes precise and expands upon theorem [ t : group_action_intro ] of the introduction .
the terms used in theorem [ t : group_action ] which were not discussed in section [ ss : complexes_of_groups ] above are defined in section [ ss : definitions ] below , where we also introduce some notation . in section [ ss : induced_cogs ]
we construct the complex of groups induced by a group action on a complex of groups , and in section [ ss : induced_covering ] we construct the induced covering . using these results , in section [ ss : fund_group ] , we consider the structure of the fundamental group of the induced complex of groups . as in section [ ss : complexes_of_groups ] above
, all references to @xcite are to chapter iii.@xmath134 .
we will need only actions on _ simple _ complexes of groups @xmath53 by _ simple _ morphisms ; this case is already substantially technical .
if in addition the action has a strict fundamental domain in @xmath52 , it is possible to make choices so that the induced complex of groups is also simple , and many of the proofs in this section become much easier .
however , in our applications , the group action will not necessarily have a strict fundamental domain .
[ t : group_action ] let @xmath53 be a ( simple ) complex of groups over a connected scwol @xmath52 , and suppose that the action of a group @xmath64 on @xmath52 extends to an action ( by simple morphisms ) on @xmath53 . then the @xmath64action induces a complex of groups @xmath63 over @xmath68 , with @xmath63 well - defined up to isomorphism of complexes of groups , such that there is a covering of complexes of groups @xmath199 moreover the fundamental group @xmath69 splits as @xmath200 finally , if @xmath53 is faithful and the @xmath64action on @xmath53 is faithful then @xmath63 is faithful
. we will need the following general result on functoriality of coverings ( see @xcite and @xcite ) .
[ t : coverings ] let @xmath53 and @xmath63 be complexes of groups over scwols @xmath52 and @xmath201 .
suppose there is a covering of complexes of groups @xmath202 .
then @xmath53 is developable if and only if @xmath63 is developable .
moreover , @xmath203 induces a monomorphism of fundamental groups @xmath204 and an equivariant isomorphism of universal covers @xmath205 we gather here the definitions and notation needed for the statement and proof of theorem [ t : group_action ] above . throughout this section , @xmath52 and @xmath201 are scwols , @xmath206 is a simple complex of groups over @xmath52 , and @xmath207 is a complex of groups over @xmath201 .
[ d : morphism ] let @xmath208 be a morphism of scwols ( see section 1.5 of @xcite ) .
a _ morphism _ @xmath209 over @xmath210 consists of : 1 . a homomorphism @xmath211 for each @xmath169 , called the _ local map _
at @xmath55 ; and 2 . [
i : commuting ] an element @xmath212 for each @xmath171 , such that the following diagram commutes @xmath213 ^ -{\phi_{i(a ) } } \ar[rrr]^{\psi_a } & & & g_{t(a ) } \ar[d]^-{\phi_{t(a ) } } \\
h_{f(i(a ) ) } \ar[rrr]^{\operatorname{ad}(\phi(a))\circ \theta_{f(a ) } } & & & h_{f(t(a ) ) } } \ ] ] and for all pairs of composable edges @xmath141 in @xmath166 , @xmath214 a morphism is _ simple _ if each element @xmath215 is trivial .
if @xmath210 is an isomorphism of scwols , and each @xmath216 an isomorphism of the local groups , then @xmath203 is an _ isomorphism of complexes of groups_. we introduce the following , expected , definitions .
an _ automorphism _ of @xmath53 is an isomorphism @xmath217 .
it is not hard to check that the set of automorphisms of @xmath53 forms a group under composition , which we denote @xmath65 ( see section 2.4 of @xcite for the definition of composition of morphisms ) .
[ d : action ] a group _ @xmath64 acts on @xmath53 _ if there is a homomorphism @xmath218 our notation is as follows .
suppose @xmath64 acts on @xmath53 .
then in particular , @xmath64 acts on the scwol @xmath52 in the sense of definition [ d : action_on_scwol ] above .
we write the action of @xmath64 on @xmath52 as @xmath219 and @xmath220 , for @xmath66 , @xmath221 and @xmath171 .
the element @xmath66 induces the automorphism @xmath222 of @xmath53 .
the data for @xmath222 is a family @xmath223 of group isomorphisms @xmath224 , and a family of elements @xmath225 with @xmath226 , satisfying the definition of morphism above ( definition [ d : morphism ] ) .
we say that the @xmath64action is _ by simple morphisms _ if each @xmath222 is simple , that is , if each @xmath226 is the trivial element .
explicitly , for each @xmath171 and each @xmath66 , the following diagram commutes .
^ -{\phi^h_{i(a ) } } \ar[rrr]^{\psi_a } & & & g_{t(a ) } \ar[d]^-{\phi^h_{t(a ) } } \\ g_{h.i(a ) } \ar[rrr]^{\psi_{h.a } } & & & g_{h.t(a ) } } \ ] ] we note also that the composition of simple morphisms @xmath227 is the simple morphism @xmath228 with local maps @xmath229 finally we recall the definition of a covering of complexes of groups .
[ d : covering ] a morphism @xmath202 over a nondegenerate morphism of scwols @xmath230 ( see section 1.5 of @xcite ) is a _ covering of complexes of groups _ if further : 1 .
each @xmath216 is injective ; and 2 .
[ i : covbijection ] for each @xmath169 and @xmath231 such that @xmath232 , the map on cosets @xmath233 induced by @xmath234 is a bijection .
suppose that a group @xmath64 acts by simple morphisms on a simple complex of groups @xmath206 . in this section
we construct the complex of groups @xmath63 induced by this action , prove that @xmath63 is well - defined up to isomorphism of complexes of groups , and discuss faithfulness .
let @xmath201 be the quotient scwol @xmath68 and let @xmath235 be the natural projection .
for each vertex @xmath236 choose a representative @xmath237 such that @xmath238 .
let @xmath239 be the subgroup of @xmath64 fixing @xmath240 and let @xmath241 be the local group of @xmath53 at @xmath240 . since the @xmath64action is by simple morphisms , by equation above there is a group homomorphism @xmath242 given by @xmath243 .
for each @xmath236 we then define the local group @xmath244 to be the corresponding semidirect product of @xmath241 by @xmath239 , that is , @xmath245 for each edge @xmath246 with @xmath247 there is , since @xmath64 acts on @xmath52 in the sense of definition [ d : action_on_scwol ] above , a unique edge @xmath248 such that @xmath249 and @xmath250 . for
each @xmath246 choose an element @xmath251 such that @xmath252 .
[ l : monom ] let @xmath253 and @xmath254 .
then the map @xmath255 is a monomorphism @xmath256 .
we will show that @xmath257 is a group homomorphism .
since @xmath258 , @xmath259 and the conjugation @xmath260 are all injective , the conclusion follows .
let @xmath261 and @xmath262 .
note that since @xmath263 and @xmath264 fix @xmath265 , they fix the edge @xmath266 and hence fix the vertex @xmath267 as well .
we have @xmath268 while @xmath269 after applying equation above to the map @xmath270 , and some cancellations , it remains to show that @xmath271 this follows from the fact that @xmath222 is a simple morphism with @xmath272 . to complete the construction of @xmath63 , for each composable pair of edges @xmath141 in @xmath273 ,
define @xmath274 one checks that @xmath275 hence @xmath276 . by abuse of notation
we write @xmath277 for @xmath278 . the data @xmath279 is a complex of groups .
given lemma [ l : monom ] above , it remains to show that for each pair of composable edges @xmath141 in @xmath273 , @xmath280 and that the cocycle condition holds .
let @xmath281 .
we compute @xmath282 while @xmath283 by definition of @xmath277 it remains to show equality in the first component . by equation and the definition of @xmath277 ,
@xmath284 hence it suffices to prove @xmath285 since @xmath53 is a simple complex of groups , and @xmath286 is the composition of the edges @xmath287 and @xmath288 , we have @xmath289 applying this , and the fact that @xmath290 is a simple morphism on the edge @xmath291 , we have @xmath292 hence equation holds
. the cococycle condition follows from the definition of @xmath277 .
we conclude that @xmath63 is a complex of groups .
we now have a complex of groups @xmath63 induced by the action of @xmath64 on @xmath53 .
this construction depended on choices of lifts @xmath240 and of elements @xmath251 .
we next show ( in a generalisation of section 2.9(2 ) of @xcite ) that : [ l : well_defined ] the complex of groups @xmath63 is well - defined up to isomorphism of complexes of groups .
suppose we made a different choice of lifts @xmath293 and elements @xmath294 , resulting in a complex of groups @xmath295 .
an isomorphism @xmath296 from @xmath63 to @xmath297 over the identity map @xmath298 is constructed as follows . for each @xmath236 ,
choose an element @xmath299 such that @xmath300 , and define a group isomorphism @xmath301 by @xmath302 for each @xmath246 , define @xmath303 .
note that by 2.9(2 ) of @xcite , @xmath304 .
the verification that @xmath305 is an isomorphism of complexes of groups is straightforward .
if @xmath53 is faithful and the @xmath64action on @xmath52 is faithful then @xmath63 is faithful .
this follows from the construction of @xmath63 , and the characterisation of faithful complexes of groups in proposition 38 of @xcite .
suppose @xmath64 acts by simple morphisms on a simple complex of groups @xmath53 , inducing a complex of groups @xmath63 as in section [ ss : induced_cogs ] above . in this section
we construct a covering of complexes of groups @xmath306 over the quotient map @xmath235 . for @xmath221 , the local maps @xmath307 are defined as follows .
recall that for each vertex @xmath236 we chose a lift @xmath237 .
now for each @xmath221 , we choose @xmath308 such that @xmath309 .
hence @xmath310 is an isomorphism @xmath311 .
the local map @xmath312 is then defined by @xmath313 note that each @xmath314 is injective . for each edge
@xmath171 , define @xmath315 where @xmath316 .
note that , since @xmath64 acts on @xmath52 in the sense of definition [ d : action_on_scwol ] above , we have @xmath317 hence @xmath318 fixes @xmath319 .
thus @xmath320 as required .
[ p : covering ] the map @xmath305 is a covering of complexes of groups .
it may be checked that @xmath321 is a morphism of complexes of groups . as noted ,
each of the local maps @xmath314 is injective .
it remains to show that for each @xmath169 and @xmath322 such that @xmath323 , the map on cosets @xmath324 induced by @xmath325 is a bijection .
we first show that @xmath326 is injective .
suppose @xmath137 and @xmath327 are in @xmath328 with @xmath329 , and suppose @xmath330 with @xmath155 representing a coset of @xmath331 in @xmath54 and @xmath332 a coset of @xmath333 in @xmath54 .
assume that @xmath334 and @xmath335 belong to the same coset of @xmath336 in @xmath244 .
looking at the second component of the semidirect product @xmath244 , it follows from the definition of @xmath337 ( lemma [ l : monom ] above ) that for some @xmath338 , @xmath339 thus @xmath340 fixes @xmath341 .
hence @xmath342 fixes @xmath343 , and so @xmath342 fixes @xmath137 .
thus @xmath344 , hence @xmath345
. looking now at the first component of @xmath334 and @xmath346 in the semidirect product @xmath244 , by definition of @xmath337 , for some @xmath347 we have @xmath348 since @xmath349 is an isomorphism , and @xmath350 , this implies @xmath351 as required .
thus the map @xmath326 is injective . to show that @xmath326 is surjective , let @xmath352 and @xmath353 , so that @xmath354 .
let @xmath137 be the unique edge of @xmath52 with @xmath355 and such that @xmath356 .
let @xmath332 be the unique element of @xmath54 such that @xmath357 .
we claim that @xmath358 lies in the same coset as @xmath359 . now
@xmath360 so it suffices to show that @xmath361 .
equivalently , we wish to show that @xmath362 fixes @xmath341 .
we have @xmath363 by definition , and the result follows by our choice of @xmath137 .
thus @xmath326 is surjective .
hence @xmath321 is a covering of complexes of groups .
suppose @xmath64 acts by simple morphisms on a simple complex of groups @xmath53 , inducing a complex of groups @xmath63 as in section [ ss : induced_cogs ] above . in this section
we show that the fundamental group of @xmath63 is the semidirect product of the fundamental group of @xmath53 by @xmath64 ( proposition [ p : fund_gp_splits ] below ) .
let @xmath364 be a vertex of @xmath52 and let @xmath235 be the natural projection .
we refer the reader to section 3 of @xcite for the definition of the _ fundamental group of g(y ) at @xmath364 _ , denoted @xmath365 .
we will use notation and results from that section in the following proof .
let @xmath366 be the fundamental group of @xmath63 at @xmath367 .
[ p : fund_gp_splits ] there is a group isomorphism @xmath368 we will construct a short exact sequence which splits @xmath369 to obtain a monomorphism @xmath370 , we use the morphism of complexes of groups @xmath306 defined in section [ ss : induced_covering ] above . by proposition 3.6 of @xcite
, @xmath321 induces a natural homomorphism @xmath371 since @xmath321 is a covering ( proposition [ p : covering ] above ) , theorem [ t : coverings ] above implies that this map @xmath372 is in fact injective .
we next define a surjection @xmath373 . the group @xmath64 may be regarded as a complex of groups over a single vertex .
there is then a canonical morphism of complexes of groups @xmath374 , defined as follows . for each @xmath236 ,
the local map @xmath375 is trivial on @xmath376 , and is inclusion on @xmath377 . for each edge @xmath144 of @xmath201 ,
@xmath378 . by proposition 3.6 of @xcite ,
the morphism @xmath203 induces a homomorphism of fundamental groups @xmath379 by 3.14 and corollary 3.15 of @xcite , if @xmath53 were a complex of trivial groups , this map would be surjective .
since the image of @xmath380 does not in fact depend on the local groups of @xmath53 , we have that in all cases , @xmath380 is surjective , as required .
it follows from definitions that the image of the monomorphism @xmath372 is the kernel of the surjection @xmath380 , hence the sequence above is exact .
given the definition of @xmath203 above , a section @xmath381 is not hard to construct .
this completes the proof of theorem [ t : group_action ] .
we now prove the main theorem and corollary [ c : infinite_generation ] , stated in the introduction . throughout this section
, we adopt the notation of the main theorem , and assume that the vertices @xmath36 and @xmath37 of the nerve @xmath26 , and the elements @xmath79 and @xmath80 of the group @xmath29 of label - preserving automorphisms of @xmath26 , satisfy conditions of its statement . in section [ ss : underlying ] we introduce notation , and construct a family of finite polyhedral complexes @xmath382 , for @xmath383 , and an infinite polyhedral complex @xmath384 .
we then in section [ ss : gyn ] construct complexes of groups @xmath73 and @xmath75 over these spaces , and show that there are coverings of complexes of groups @xmath385 and @xmath386 . in section [ ss : hn ] we define the action of a finite group @xmath387 on @xmath382 , and of an infinite group @xmath388 on @xmath384 , and then in section [ ss : action ] we show that these actions extend to actions on the complexes of groups @xmath73 and @xmath75 . in section [ ss : conclusion ] we combine these results with theorem [ t : group_action ] above to complete the proof of the main theorem .
corollary [ c : infinite_generation ] is proved in section [ ss : corollary_proof ] . in this section
we construct a family of finite polyhedral complexes @xmath382 and an infinite polyhedral complex @xmath384 .
we first set up some notation . for @xmath39 ,
let @xmath389 be the order of @xmath40 , assumed prime .
it will be convenient to put , for all @xmath390 , @xmath391 and @xmath392 , and similarly @xmath393 , @xmath394 , @xmath395 and @xmath396 .
conditions of the main theorem then become : 1 . for all @xmath383
, @xmath397 fixes the star of @xmath398 in @xmath26 , and @xmath399 ; 2 . for all @xmath383 , and all @xmath400 such that @xmath401 is the image of @xmath402 under some power of @xmath397 , @xmath403 ; and 3 . for all @xmath383 ,
all spherical special subgroups of @xmath3 which contain @xmath402 are halvable along @xmath402 .
we now use the sequences @xmath404 and @xmath405 to define certain elements and subsets of @xmath3 .
let @xmath406 be the trivial element of @xmath3 and for @xmath407 let @xmath408 be the product @xmath409 denote by @xmath410 the one - element set @xmath411 . for @xmath407 , and @xmath412
, in order to simplify notation , write @xmath413 for the composition of automorphisms @xmath414 where @xmath415 for @xmath416
. let @xmath417 be the element of @xmath3 : @xmath418 now for @xmath407 and @xmath412 , define @xmath419 note that if @xmath420 then @xmath421 .
let @xmath1 be the coxeter system in example 1 of section [ ss : davis_complexes ] above , with nerve @xmath26 shown in figure [ f : nerve ] above .
for @xmath39 , let @xmath422 be the automorphism of @xmath26 which fixes the star of @xmath41 in @xmath26 and interchanges @xmath45 and @xmath423 .
then if @xmath96 and @xmath97 are both even , the main theorem applies to this example .
( if @xmath424 then @xmath24 is halvable along @xmath34 with @xmath77 the trivial group . if @xmath425 then @xmath24 is the dihedral group of order @xmath426 , and @xmath24 is halvable along @xmath34 if and only if @xmath90 is even , in which case @xmath77 is the dihedral group of order @xmath90 . )
note that @xmath427 , and so , for instance , @xmath428 the following lemma establishes key properties of the sets @xmath429 .
[ l : disjoint ] for all @xmath383 : 1 . the sets @xmath430 , @xmath431 , , @xmath410 are pairwise disjoint ; and 2 . for all @xmath412 , if @xmath432 ( where @xmath415 for @xmath416 ) then @xmath433 , @xmath434 , , and @xmath435 .
given @xmath436 , with @xmath415 for @xmath416 and @xmath437 for @xmath438 , suppose @xmath439 then @xmath440@xmath441 by condition above , for each @xmath416 , the automorphism @xmath40 fixes @xmath442 , thus @xmath443 also by condition , since @xmath40 fixes the star of @xmath442 but @xmath444 , we have @xmath445 . since @xmath446 is a label - preserving automorphism , it follows that the product of the two generators @xmath447 has infinite order , for each @xmath416 .
similarly for each @xmath448 .
thus the only way for equation to hold is if @xmath449 , and for each @xmath416 , @xmath450 .
since @xmath40 is of prime order @xmath389 and we specified @xmath415 , the result follows . for @xmath383 , and @xmath451 ,
define @xmath452 to be the set of chambers @xmath453 recall that we are writing @xmath120 for the pair @xmath121 . by lemma [ l : disjoint ]
above , for fixed @xmath454 , the sets @xmath455 are pairwise disjoint .
we now define @xmath382 to be the polyhedral complex obtained by gluing together " the chambers in @xmath455 , using the same relation @xmath112 as in the davis complex @xmath0 for @xmath1 .
more precisely , @xmath456 where , for @xmath457 , we have @xmath114 if and only if @xmath115 and @xmath458 .
note that @xmath459 . to define @xmath384 , for each @xmath460 , noting that @xmath429 is only defined for @xmath451 , put @xmath461 then @xmath462 is the set of chambers @xmath463 similarly to the finite case , the sets @xmath464 are pairwise disjoint , and we define @xmath465 for the same relation @xmath112 .
note that there are natural strict inclusions as subcomplexes @xmath466 ( in fact , @xmath382 and @xmath384 are subcomplexes of the davis complex @xmath0 , but we will not adopt this point of view . ) we define a mirror of @xmath382 or @xmath384 to be an _ interior mirror _ if it is contained in more than one chamber .
let @xmath1 , @xmath79 and @xmath80 be as in the previous example of this section .
to indicate the construction of @xmath382 and @xmath384 in this case , figure [ f : y4 ] below depicts the dual graph for @xmath467 , that is , the graph with vertices the chambers of @xmath467 , and edges joining adjacent chambers .
the edges are labelled with the type of the corresponding interior mirror .
figure [ f : yinfty ] sketches the dual graph for @xmath384 .
, with vertices and edges labelled ] ] we now describe features of @xmath382 and @xmath384 which will be needed below .
the first lemma follows from the construction of @xmath382 and @xmath384 and lemma [ l : disjoint ] above .
[ l : adjacency ] let @xmath468 .
all of the chambers of @xmath382 to which @xmath469 is adjacent are described by the following . 1 .
[ i : up ] for @xmath383 and @xmath412 , the chamber @xmath120 is adjacent to exactly one chamber of @xmath470 , namely it is @xmath471adjacent to the chamber @xmath472 of @xmath470 .
[ i : down ] for @xmath407 and @xmath451 , the chamber @xmath120 is adjacent to exactly @xmath473 distinct chambers of @xmath474 , namely for each @xmath475 , the chamber @xmath120 is @xmath476adjacent to the chamber @xmath477 of @xmath474 .
similarly for @xmath384 .
[ c : mirrors ] 1 .
[ i : vertex ] any vertex of @xmath382 is contained in at most two distinct chambers of @xmath382 , and similarly for @xmath384 .
[ i : mirrors ] any two interior mirrors of @xmath382 or @xmath384 are disjoint .
suppose @xmath55 is a vertex of @xmath382 , contained in the chamber @xmath120 , where @xmath478 is as in lemma [
l : adjacency ] above . if @xmath55 is contained in more than one chamber of @xmath382 or @xmath384 , then @xmath55 is contained in an interior mirror @xmath105 , for some @xmath27 . by the construction of @xmath382 and
lemma [ l : adjacency ] above , @xmath34 is either an image of @xmath479 , or one of @xmath473 distinct images of @xmath480 , under some element of @xmath29 .
suppose @xmath34 is in the image of @xmath479 .
condition of the main theorem implies that @xmath481 .
hence the mirror @xmath105 is disjoint from each of the @xmath473 mirrors of types the @xmath473 images of @xmath480 .
therefore the only chambers of @xmath382 which contain @xmath55 are the two chambers @xmath120 and @xmath482 .
now suppose @xmath34 is one of the @xmath473 images of @xmath480 under some element of @xmath29 .
condition of the main theorem implies that the mirrors of types each of these images are pairwise disjoint , and so again @xmath55 is contained in only two distinct chambers of @xmath382 .
similarly , any two interior mirrors of @xmath382 or @xmath384 are disjoint .
[ l : subyn ] for all @xmath407 , there are @xmath483 disjoint subcomplexes of @xmath382 , denoted @xmath484 for @xmath485 , each isomorphic to @xmath486 , and with @xmath487 . for each @xmath485 , the subcomplex @xmath484 is attached to the chamber @xmath488 of @xmath382 along its mirror of type @xmath489 .
an isomorphism @xmath490 is given by sending the chamber @xmath491 to the chamber @xmath492 , and the vertex of @xmath493 of type @xmath13 to the vertex of @xmath494 of type @xmath495 , for each spherical subset @xmath13 of @xmath88 .
by induction on @xmath454 , using lemma [ l : adjacency ] and corollary [ c : mirrors ] above .
we now construct complexes of groups @xmath73 over each @xmath382 , and @xmath75 over @xmath384 , and show that there are coverings @xmath385 and @xmath386 . to simplify notation , write @xmath52 for @xmath382 or @xmath384 . to define the local groups of @xmath53 ,
let @xmath55 be a vertex of @xmath52 , of type @xmath13 . by corollary
[ c : mirrors ] above , @xmath55 is contained in at most two distinct chambers of @xmath52 .
if @xmath55 is only contained in one chamber of @xmath52 , put @xmath496 .
if @xmath55 is contained in two distinct chambers of @xmath52 , then by corollary [ c : mirrors ] above @xmath55 is contained in a unique interior mirror @xmath105 , with @xmath497 . by the construction of @xmath52 , @xmath34 is in the @xmath29orbit of some @xmath402 , @xmath383 . by condition of the main theorem , it follows that the group @xmath24 is halvable along @xmath34 .
we define the local group at @xmath55 to be @xmath498 .
the monomorphisms between local groups are defined as follows .
let @xmath137 be an edge of @xmath52 , with @xmath138 of type @xmath13 and @xmath139 of type @xmath160 , so that @xmath499 . if both of the vertices @xmath138 and @xmath139 are contained in a unique chamber of @xmath52 , then the monomorphism @xmath191 along this edge is defined to be the natural inclusion @xmath500 . if @xmath138 is contained in two distinct chambers , then @xmath138 is contained in a unique interior mirror @xmath105 , with @xmath25 .
thus @xmath501 as well , and so @xmath139 is also contained in the mirror @xmath105 . from the definitions of @xmath77 and @xmath502 , it follows that there is a natural inclusion @xmath503 , and we define @xmath191 be this inclusion .
finally suppose @xmath138 is contained in a unique chamber of @xmath52 but @xmath139 is contained in two distinct chambers of @xmath52 .
then for some @xmath460 , @xmath138 is in a chamber of @xmath452 ( respectively , @xmath462 ) , and @xmath139 is either in @xmath474 or in @xmath470 ( respectively , in @xmath504 or @xmath505 ) .
moreover @xmath139 is contained in a unique interior mirror @xmath105 , with @xmath506 .
if @xmath139 is in @xmath474 ( respectively , @xmath504 ) , then we define @xmath191 to be the natural inclusion @xmath507 .
if @xmath139 is in @xmath470 ( respectively , @xmath505 ) , then we define @xmath191 to be the monomorphism defined on the generators @xmath508 of @xmath24 by @xmath509 , that is , @xmath510 .
it is not hard to verify that for all pairs of composable edges @xmath141 in @xmath52 , @xmath511 .
hence we have constructed simple complexes of groups @xmath73 and @xmath75 over @xmath382 and @xmath384 respectively .
note that these complexes of groups are faithful , since by construction the local group at each vertex of type @xmath104 is trivial .
note also that @xmath71 is the same complex of groups as constructed in section [ ss : complexes_of_groups ] above , which has fundamental group @xmath3 and universal cover @xmath0 .
let @xmath1 , @xmath79 and @xmath80 be as in the examples in section [ ss : underlying ] above .
the complex of groups @xmath512 is sketched in figure [ f : gy2 ] . from left to right , the three chambers here are @xmath10 , @xmath513 and @xmath514 .
we denote by @xmath183 the dihedral group of order @xmath127 , with @xmath515 the dihedral group of order @xmath96 , and similarly for @xmath184 and @xmath516 ( recall that @xmath96 and @xmath97 are even ) . ]
there are coverings of complexes of groups @xmath385 and @xmath386 .
let @xmath517 and @xmath518 be the maps sending each vertex of @xmath382 or @xmath384 of type @xmath13 to the unique vertex of @xmath519 of type @xmath13 . then by construction of @xmath382 and @xmath384 ,
the maps @xmath520 and @xmath521 are nondegenerate morphisms of scwols .
we define coverings @xmath522 and @xmath523 over @xmath520 and @xmath521 respectively . to simplify notation , write @xmath52 for respectively @xmath382 or @xmath384 , @xmath210 for respectively @xmath520 or @xmath521 , and @xmath203 for respectively @xmath524 or @xmath525 .
let @xmath55 be a vertex of @xmath52 , of type @xmath13 .
if the local group at @xmath55 is @xmath496 then the map of local groups @xmath526 is the identity map . if the local group at @xmath55 is @xmath77 , for some @xmath25 , then @xmath527 is the natural inclusion as an index @xmath5 subgroup . to define elements @xmath215 ,
if the monomorphism @xmath191 in @xmath53 is natural inclusion , define @xmath528 .
if @xmath191 is @xmath529 , then define @xmath530 .
it is then easy to check that , by construction , @xmath203 is a morphism of complexes of groups . to show that @xmath203 is a covering of complexes of groups , we first observe that each of the local maps @xmath531 is injective .
now fix @xmath158 a vertex of @xmath52 , of type @xmath160 , and @xmath144 an edge of @xmath532 such that @xmath533 , with @xmath534 of type @xmath13 ( hence @xmath499 ) .
we must show that the map @xmath535 induced by @xmath536 is a bijection , where @xmath54 and @xmath537 are the local groups of @xmath53 .
first suppose that @xmath55 is contained in a unique chamber of @xmath52 .
then by construction , there is a unique edge @xmath137 of @xmath52 with @xmath138 of type @xmath13 and @xmath538 , hence a unique edge @xmath539 with @xmath538 .
moreover , @xmath540 , @xmath541 , the monomorphism @xmath191 is natural inclusion hence @xmath528 , and @xmath542 is the identity map .
hence @xmath543 is a bijection in this case .
now suppose that @xmath55 is contained in two distinct chambers of @xmath52 .
then @xmath55 is contained in a unique interior mirror @xmath105 of @xmath52 , with @xmath501 .
assume first that @xmath25 as well .
then there is a unique edge @xmath137 of @xmath52 with @xmath138 of type @xmath13 and @xmath538 .
this edge is also contained in the mirror @xmath105 .
hence there is a unique @xmath539 with @xmath538 . by construction
, we have @xmath544 , the map @xmath542 is natural inclusion as an index @xmath5 subgroup , @xmath545 , the map @xmath191 is natural inclusion , and @xmath215 trivial .
since the index @xmath546 = [ \operatorname{half}_s(w_{t'}):\operatorname{half}_s(w_t)]$ ] is finite , it is enough to verify that the inclusion @xmath547 induces an injective map on cosets @xmath548 for this , suppose that @xmath549 and that @xmath550 in @xmath551
. then @xmath552 . by definitions
, it follows that @xmath553 , as required .
now assume that @xmath55 is contained in the interior mirror @xmath105 , with @xmath554 .
there are then two edges @xmath555 such that @xmath556 . without loss of generality ,
@xmath557 is natural inclusion @xmath558 and @xmath559 , while @xmath560 with @xmath561 . since the index @xmath562 = \frac{1}{2}[w_{t'}:w_t]$ ] is finite , it is enough to show that the map on cosets @xmath543 is surjective
. let @xmath563 . if @xmath564 , then the image of the coset @xmath565 in @xmath54 is the coset @xmath566 in @xmath551 . if @xmath567 , then since @xmath502 has index @xmath5 in @xmath551 , and @xmath568 , there is a @xmath569 such that @xmath570 .
the image of the coset @xmath571 in @xmath502 is then the coset @xmath572 in @xmath551 .
thus @xmath543 is surjective , as required .
we conclude that @xmath203 is a covering of complexes of groups . in this section
we construct the action of a finite group @xmath387 on @xmath382 in the sense of definition [ d : action_on_scwol ] above , and that of an infinite group @xmath388 on @xmath384 .
we first define the groups @xmath387 and @xmath388 . for each @xmath383 , let @xmath573 denote the cyclic group of order @xmath574 .
we define @xmath575 to be the trivial group and @xmath576 . for @xmath577 , we define @xmath387 to be the wreath product @xmath578 that is , @xmath387 is the semidirect product by @xmath579 of the direct product of @xmath483 copies of @xmath580 , where @xmath579 acts on this direct product by cyclic permutation of coordinates .
note that @xmath387 is a finite group of order @xmath581 we define @xmath388 to be the infinite iterated ( unrestricted ) wreath product @xmath582 we then have natural inclusions @xmath583 the following lemma will be needed for the proof of corollary [ c : infinite_generation ] in section [ ss : corollary_proof ] below .
[ l : not_fg ] the group @xmath388 is not finitely generated . by definition of @xmath388 , for any nontrivial @xmath584
there is an @xmath383 such that @xmath585 .
we now define the actions of @xmath387 and @xmath388 on @xmath382 and @xmath384 respectively .
this uses the label - preserving automorphisms @xmath586 .
note that the action of @xmath29 on the nerve @xmath26 extends to the chamber @xmath10 , fixing the vertex of type @xmath104 .
this action does not in general have a strict fundamental domain .
inconveniently , this action also does not satisfy condition of definition [ d : action_on_scwol ] above , since for any nontrivial @xmath33 , there is an edge @xmath137 of @xmath10 with @xmath138 of type @xmath104 but @xmath587 .
however , to satisfy definition [ d : action_on_scwol ] , it suffices to define actions on @xmath382 and @xmath384 , and then extend in the obvious way to the scwols which are the barycentric subdivisions of these spaces , with naturally oriented edges . for each @xmath383 fix a generator @xmath588 for the cyclic group @xmath573 . recall that @xmath589 has order @xmath574 .
thus for any @xmath33 , there is a faithful representation @xmath590 , given by @xmath591 .
recall also that @xmath397 fixes the star in @xmath26 of the vertex @xmath398 , and that @xmath592 . hence @xmath591 induces an action of @xmath573 on the chamber @xmath10 , which fixes pointwise the mirror of type @xmath593 , and permutes cyclically the set of mirrors of types @xmath594 , for @xmath595 .
we define the action of @xmath387 on @xmath382 inductively , as follows .
the group @xmath575 is trivial . for @xmath596 ,
assume that the action of @xmath580 on @xmath486 has been given .
the subgroup @xmath579 of @xmath387 then fixes the chamber @xmath488 of @xmath382 setwise , and acts on this chamber via @xmath597 . by the discussion above
, this action fixes pointwise the mirror of type @xmath598 of @xmath599 , and permutes cyclically the @xmath483 mirrors of types @xmath600 , with @xmath485 , along which ( by lemma [ l : subyn ] above ) , @xmath483 disjoint subcomplexes of @xmath382 , each isomorphic to @xmath486 , are attached . by induction , a copy of @xmath580 in @xmath387 acts on each of these copies of @xmath486 in @xmath382 .
more precisely , for @xmath485 , the @xmath601st copy of @xmath580 in @xmath387 acts on the subcomplex @xmath602 of lemma [ l : subyn ] above .
this action is given by conjugating the ( inductively defined ) action of @xmath580 on @xmath603 by the isomorphism @xmath604 in lemma [ l : subyn ] . by definition ,
the action of @xmath579 cyclically permutes the subcomplexes @xmath602 , and so we have defined an action of @xmath387 on @xmath382 .
the action of @xmath388 on @xmath384 is similar .
we now describe the fundamental domains for these actions . for each @xmath383 and each @xmath605 , observe that @xmath387 acts transitively on the set of chambers @xmath452 .
let @xmath606 , and for @xmath596 let @xmath607 be the quotient of the chamber @xmath608 by the action of @xmath609 as defined above . in @xmath607 , the mirrors of types @xmath600 , for @xmath485 , have been identified . by abuse of notation , we refer to these identified mirrors as the mirror of type @xmath610 of @xmath607 .
note also that @xmath609 fixes pointwise the mirror of type @xmath402 of @xmath599 , and so we may speak of the mirror of type @xmath402 of @xmath607 . then a fundamental domain for the action of @xmath387 on @xmath382 is the finite complex @xmath611 where @xmath112 means we identify the @xmath612mirrors of @xmath613 and @xmath614 , for @xmath615 .
similarly , a fundamental domain for the action of @xmath388 on @xmath384 is the infinite complex @xmath616 finally we describe the stabilisers in @xmath387 and @xmath388 of the vertices of @xmath382 and @xmath384 .
let @xmath120 be a chamber of @xmath382 or @xmath384 .
then there is a smallest @xmath460 such that @xmath617 . by construction
, it follows that the stabiliser in @xmath387 or @xmath388 of any vertex in the chamber @xmath120 is a subgroup of the finite group @xmath618 .
hence @xmath387 and @xmath388 act with finite stabilisers .
note also that for every @xmath383 , the action of @xmath387 fixes the vertex of type @xmath104 in the chamber @xmath599 .
we may thus speak of the vertex of type @xmath104 in the quotient @xmath607 defined above .
in fact , in the fundamental domains @xmath619 and @xmath620 defined above , the vertex of type @xmath104 in @xmath607 , for @xmath383 , has a lift in @xmath382 or @xmath384 with stabiliser the finite group @xmath387 .
we observe also that the actions of @xmath387 and @xmath388 are faithful , since the stabiliser of the vertex of type @xmath104 of @xmath621 is the trivial group @xmath575 .
figure [ f : zinfty ] shows @xmath620 and the stabilisers of ( lifts of ) its vertices of type @xmath104 for the example in section [ ss : underlying ] above . ] in this section we show that the actions of @xmath387 and @xmath388 on @xmath382 and @xmath384 , defined in section [ ss : hn ] above , extend to actions ( by simple morphisms ) on the complexes of groups @xmath73 and @xmath75 . to simplify notation , write @xmath64 for @xmath387 or @xmath388 , @xmath52 for @xmath382 or @xmath384 , and @xmath201 for @xmath619 or @xmath620 . technically , instead of working with @xmath53 , we work with the corresponding naturally defined complex of groups over the barycentric subdivision of @xmath52 , so that the action of @xmath64 satisfies definition [ d : action_on_scwol ] above . by abuse of notation
we will however continue to write @xmath53 .
recall that for @xmath55 a vertex of @xmath52 of type @xmath13 , the local group @xmath54 is either @xmath24 or @xmath77 , and the latter occurs if and only if @xmath55 is contained in an interior @xmath34mirror of @xmath52 with @xmath25 . let @xmath120 be a chamber of @xmath52 and let @xmath66 . by definition of the @xmath64action
, there is an @xmath33 such that for each vertex @xmath55 in @xmath120 , with @xmath55 of type @xmath13 , the vertex @xmath622 of @xmath623 has type @xmath624 .
moreover , if @xmath55 is contained in an interior @xmath34mirror then @xmath622 is contained in an interior @xmath625mirror .
we may thus define the local map @xmath626 by @xmath627 for each @xmath508 , and ( if @xmath628 ) , @xmath629 .
then @xmath630 is an isomorphism either @xmath631 , or @xmath632 , as appropriate .
it is not hard to verify that these local maps define an action of @xmath64 on @xmath53 by simple morphisms . in this section
we combine the results of sections [ ss : underlying][ss : action ] above to complete the proof of the main theorem .
recall that @xmath71 is developable with universal cover @xmath0 ( see section [ ss : complexes_of_groups ] ) . by proposition [ p : covering ] and theorem [ t : coverings ] above , it follows that the complexes of groups @xmath73 and @xmath75 are developable with universal cover @xmath0 .
let @xmath74 be the complex of groups induced by @xmath387 acting on @xmath73 , and @xmath76 that induced by @xmath388 acting on @xmath75 . by theorem [ t : group_action ] above , there are coverings of complexes of groups @xmath633 and @xmath634 .
hence ( by theorem [ t : coverings ] above ) each @xmath74 and @xmath76 is developable with universal cover @xmath0 .
let @xmath51 be the fundamental group of @xmath74 and @xmath4 the fundamental group of @xmath76 .
since the complexes of groups @xmath73 and @xmath75 are faithful , and the actions of @xmath387 and @xmath388 are faithful , theorem [ t : group_action ] above implies that @xmath74 and @xmath76 are faithful complexes of groups .
thus @xmath51 and @xmath4 may be identified with subgroups of @xmath19 .
now @xmath73 and @xmath75 are complexes of finite groups , and the @xmath387 and @xmath388actions have finite vertex stabilisers . hence by construction , @xmath74 and @xmath76 are complexes of finite groups .
therefore @xmath51 and @xmath4 are discrete subgroups of @xmath2 .
since the fundamental domain @xmath619 is finite , it follows that each @xmath51 is a uniform lattice . to show that @xmath4 is a nonuniform lattice
, we use the normalisation of haar measure @xmath6 on @xmath19 defined in section [ ss : lattices ] above , with the @xmath2set @xmath81 the set of vertices of @xmath0 of type @xmath104 . since the local groups of @xmath76 at the vertices of type @xmath104 in @xmath620 are @xmath575 , @xmath635 ,
, we have @xmath636 this series converges ( see equation above for the order of @xmath387 , and note that each @xmath637 ) . we conclude that @xmath4 is a nonuniform lattice in @xmath2 .
moreover , as the covolumes of the uniform lattices @xmath51 are the partial sums of this series , we have @xmath50 , as required .
this completes the proof of the main theorem .
the nonuniform lattice @xmath4 is the fundamental group of the complex of groups @xmath76 induced by the action of @xmath388 on @xmath75 . by theorem [ t : group_action ] above , there is an isomorphism @xmath638 .
hence there is a surjective homomorphism @xmath639 .
since @xmath388 is not finitely generated ( lemma [ l : not_fg ] above ) , we conclude that @xmath4 is not finitely generated .
in this section we describe several infinite families of examples to which the main theorem applies . by the _ dimension _ of the davis complex @xmath0 for a coxeter system @xmath1 , we mean the maximum cardinality of a spherical subset of @xmath88 .
we note that there may be maximal spherical special subgroups @xmath24 with @xmath640 strictly less than @xmath641 . if @xmath642 then the nerve of the coxeter system @xmath1 is a graph @xmath26 with vertex set @xmath88 and two vertices @xmath34 and @xmath103 joined by an edge if and only if @xmath90 is finite .
assume for simplicity that for some integer @xmath643 all finite @xmath644 .
then @xmath0 is the barycentric subdivision of a polygonal complex @xmath15 , with all @xmath5cells of @xmath15 regular euclidean @xmath127gons , and the link of every vertex of @xmath15 the graph @xmath26 .
such an @xmath15 is called a _
@xmath645complex_. condition of the main theorem can hold only if @xmath96 is even , and so we also assume this .
it is then not hard to find graphs @xmath26 so that , for some pair @xmath36 and @xmath37 of non - adjacent vertices of @xmath26 , and for some nontrivial elements @xmath646 , conditions and of the main theorem also hold .
we present three infinite families of examples .
let @xmath26 be the complete bipartite graph @xmath647 , with @xmath648 .
if @xmath649 then there is a vertex @xmath36 of @xmath26 so that some element @xmath650 , without loss of generality of prime order , fixes the star of @xmath36 in @xmath26 , but @xmath651 .
let @xmath37 be a vertex of @xmath26 such that @xmath652 .
the main theorem then applies provided @xmath96 is even . if @xmath653 then @xmath0 is the barycentric subdivision of the product of trees @xmath654 , where @xmath655 is the @xmath656regular tree . in particular ,
if @xmath657 in example 1 of section [ ss : davis_complexes ] above , then @xmath0 is the barycentric subdivision of @xmath658 .
if @xmath659 , then by theorem 12.6.1 of @xcite the complex @xmath0 may be metrised as a piecewise hyperbolic @xmath660 polygonal complex . with this metric , if @xmath661 and @xmath662 then @xmath0 is the barycentric subdivision of bourdon s building @xmath663 ( studied in , for example , @xcite and @xcite ) , which is the unique @xmath5complex with all @xmath5cells regular right - angled hyperbolic @xmath664gons @xmath665 , and the link of every vertex the complete bipartite graph @xmath666 .
bourdon s building is a right - angled hyperbolic building , of type @xmath667 where @xmath668 is the coxeter group generated by reflections in the sides of @xmath665 . a fuchsian
building is a @xmath5dimensional hyperbolic building .
bourdon s building @xmath663 is a ( right - angled ) fuchsian building . for fuchsian buildings
which are not right - angled see , for example , @xcite and @xcite .
to show that the main theorem applies to certain fuchsian buildings which are not right - angled , let @xmath26 be the finite building of rank @xmath5 associated to a chevalley group @xmath669 ( see @xcite ) .
then @xmath26 is a bipartite graph , with vertex set say @xmath670 , and for some @xmath671 , @xmath26 has girth @xmath672 and diameter @xmath673 .
figure [ f : proj_plane ] depicts the building @xmath26 for the group @xmath674 , for which @xmath675 .
the white vertices of this building may be identified with the set of one - dimensional subspaces of the vector space @xmath676 , and the black vertices with the set of two - dimensional subspaces of @xmath81 .
two vertices are joined by an edge if those two subspaces are incident .
the group @xmath669 acts on @xmath26 , preserving the type of vertices , with quotient an edge .
suppose @xmath678 , and let @xmath679 be a vertex at distance @xmath673 from @xmath36 . since @xmath26 is a thick building , there is more than one such vertex @xmath37 . for @xmath39 ,
the stabiliser @xmath680 of @xmath45 in @xmath669 acts transitively on the set of vertices of @xmath26 at distance @xmath673 from @xmath45 .
now , by theorem 6.18 of @xcite , @xmath680 has a levi decomposition @xmath681 where @xmath682 is the subgroup of @xmath680 fixing the vertex @xmath41 . moreover , by lemma 6.5 of @xcite , @xmath683 fixes the star of @xmath45 in @xmath26 .
hence , we may find elements @xmath684 , without loss of generality of prime order , for which condition of the main theorem holds .
condition of the main theorem follows since @xmath26 is bipartite and the action of @xmath669 preserves the type of vertices . for example , for @xmath26 as in figure [ f : proj_plane ] , if @xmath36 is the vertex @xmath685 , we may choose @xmath37 to be the vertex @xmath686 , and then choose @xmath687 suppose now that @xmath26 as above is the nerve of a coxeter system @xmath1 . by theorem 12.6.1 of @xcite ,
since @xmath26 has girth @xmath688 , the corresponding davis complex @xmath0 may also be metrised as a piecewise hyperbolic @xmath660 polygonal complex . with this metrisation
, @xmath0 is then the barycentric subdivision of a fuchsian building , with the link of every vertex @xmath26 and all @xmath5cells regular hyperbolic @xmath127gons ( of vertex angle @xmath689 ) .
we call such a building a _ @xmath645building_. in general , there may be uncountably many isomorphism classes of @xmath645buildings ( see for instance @xcite ) .
in fact , the davis complex @xmath0 is the barycentric subdivision of the unique locally reflexive @xmath645building with trivial holonomy ( see haglund @xcite ) .
a polygonal complex @xmath15 is _ platonic _ if @xmath194 acts transitively on the set of flags ( vertex , edge , face ) in @xmath15 .
any platonic polygonal complex is a @xmath690complex , with @xmath691 and @xmath26 a graph such that @xmath692 acts transitively on the set of oriented edges in @xmath26 . in @xcite ,
witkowski studied cat(0 ) platonic polygonal complexes @xmath15 , where @xmath26 is a trivalent graph .
such complexes are not in general buildings .
a graph @xmath26 is said to be _
@xmath454arc regular _ , for some @xmath383 , if @xmath692 acts simply transitively on the set of edge paths of length @xmath454 in @xmath26 .
for example , the petersen graph in figure [ f : petersen ] above is @xmath693arc regular .
any finite , connected , trivalent graph @xmath26 , with @xmath692 transitive on the set of oriented edges of @xmath26 , is @xmath454arc regular for some @xmath694 ( tutte @xcite ) .
witkowski @xcite showed that if @xmath695 , then for all @xmath696 there is a unique @xmath690complex @xmath15 , with @xmath15 platonic .
thus if @xmath697 is even , the barycentric subdivision of @xmath15 is the davis complex @xmath0 for @xmath1 , where @xmath1 has nerve @xmath26 and all finite @xmath644 .
now suppose @xmath26 is a finite , connected , trivalent , @xmath454arc regular graph with @xmath698 .
choose vertices @xmath36 and @xmath37 of @xmath26 at distance two in @xmath26 if @xmath699 , and at distance three in @xmath26 if @xmath700 .
then by propositions 35 of djokovi
miller @xcite , for @xmath39 there are involutions @xmath701 such that @xmath40 fixes the star of @xmath41 in @xmath26 , and @xmath444 is not adjacent to @xmath45
. thus if @xmath96 is even , the main theorem applies to @xmath19 .
we first discuss when condition in the main theorem can hold .
suppose @xmath24 is a spherical special subgroup of @xmath3 , with @xmath702 .
if @xmath24 is irreducible , then from the classification of spherical coxeter groups ( see , for example , @xcite ) , it is not hard to verify that @xmath24 is halvable along @xmath34 if and only if @xmath24 is of type @xmath703 , with @xmath25 the unique generator so that @xmath704 for all @xmath705 ; in this case @xmath77 is of type @xmath706 .
many reducible spherical special subgroups @xmath24 are also halvable along @xmath34 , for example @xmath707 .
thus an easy source of examples in dimension @xmath708 is right - angled coxeter systems @xmath1 ( meaning that @xmath709 for all @xmath710 with @xmath711 ) .
the davis complex @xmath0 will then be the barycentric subdivision of a cubical complex . 1 .
let @xmath26 be the join of @xmath673 sets of points , of cardinalities respectively @xmath712 .
then @xmath0 is the barycentric subdivision of the product of trees @xmath713 .
provided at least one of @xmath712 is greater than @xmath5 , vertices @xmath36 and @xmath37 in @xmath26 and elements @xmath714 can be found so that the main theorem applies .
2 . let @xmath26 be a finite building of rank @xmath691 associated to a chevalley group @xmath669 , and choose vertices @xmath36 and @xmath37 in @xmath26 which are _ opposite _ ( see @xcite ) . by the same arguments as in section [ sss : fuchsian ] above , there are elements @xmath715 such that conditions and of the main theorem hold .
we do not know of any hyperbolic buildings of dimension @xmath716 to which the main theorem applies ( for the @xmath693dimensional constructions of haglund paulin in @xcite , certain of the @xmath90 must be equal to @xmath693 , so condition of the main theorem will not hold ) .
slight modifications of the above examples , for example by adding a vertex @xmath34 to @xmath26 with @xmath92 for all @xmath717 , produce nerves which are not buildings , hence examples of @xmath0 which are not buildings .
f. haglund and f. paulin , _ simplicit de groupes dautomorphismes despaces courbure ngative _ , the epstein birthday schrift , 181248 ( electronic ) , geom .
, 1 , geom .
coventry , 1998 .
a. lubotzky , _ tree - lattices and lattices in lie groups _ , combinatorial and geometric group theory ( edinburgh , 1993 ) , 217232 , london math .
lecture note ser .
, 204 , cambridge univ . press , cambridge , 1995 . | let @xmath0 be the davis complex for a coxeter system @xmath1 .
the automorphism group @xmath2 of @xmath0 is naturally a locally compact group , and a simple combinatorial condition due to haglund
paulin determines when @xmath2 is nondiscrete .
the coxeter group @xmath3 may be regarded as a uniform lattice in @xmath2 .
we show that many such @xmath2 also admit a nonuniform lattice @xmath4 , and an infinite family of uniform lattices with covolumes converging to that of @xmath4 .
it follows that the set of covolumes of lattices in @xmath2 is nondiscrete .
we also show that the nonuniform lattice @xmath4 is not finitely generated .
examples of @xmath0 to which our results apply include buildings and non - buildings , and many complexes of dimension greater than @xmath5 .
to prove these results , we introduce a new tool , that of group actions on complexes of groups " , and use this to construct our lattices as fundamental groups of complexes of groups with universal cover @xmath0 . |
cyclic universe has always been a burning topic in the field of theoretical cosmology , since it is expected to avoid the initial singularity by providing an infinitely oscillating universe .
however cyclic universe confront a serious problem of black holes(bhs ) .
if the bhs formed during the expanding phase survives into the next cycle they will grow even larger from one cycle to the next and act as a serious defect in an otherwise nearly uniform universe . with the passage of time the bhs will occupy the entire horizon and then the cyclic models will break away . in this paper
we investigate the possibility of an oscillating universe in two of the well known models of brane - world gravity , namely , rsii brane and dgp brane models .
randall and sundrum @xcite proposed a bulk - brane model to explain the higher dimensional theory , popularly known as rs ii brane model . according to this model we live in a four dimensional world ( called 3-brane , a domain wall ) which is embedded in a 5d space time ( bulk ) .
all matter fields are confined in the brane whereas gravity can only propagate in the bulk .
the consistency of this brane model with the expanding universe has given popularity to this model of late in the field of cosmology .
a simple and effective model of brane - gravity is the dvali - gabadadze - porrati ( dgp ) braneworld model @xcite which models our 4-dimensional world as a frw brane embedded in a 5-dimensional minkowski bulk .
it explains the origin of dark energy(de ) as the gravity on the brane leaking to the bulk at large scale . on the 4-dimensional brane
the action of gravity is proportional to @xmath2 whereas in the bulk it is proportional to the corresponding quantity in 5-dimensions .
the model is then characterized by a cross over length scale @xmath3 such that gravity is 4-dimensional theory at scales @xmath4 where matter behaves as pressureless dust , but gravity leaks out into the bulk at scales @xmath5 and matter approaches the behaviour of a cosmological constant .
moreover it has been shown that the standard friedmann cosmology can be firmly embedded in dgp brane . to explain the latest cosmic acceleration one usually assumes the existence of dark energy ( de ) with a negative pressure .
in general one can assume a perfect fluid with state equation @xmath6 , with @xmath7 , in order to realize the cosmic acceleration .
most models of de in the present universe predict that its effective equation of state satisfies the null energy condition ( nec ) , i.e. , @xmath8 , where @xmath9 and @xmath10 are the effective de density and pressure , respectively .
however , the observations do not rule out that de is phantom , i.e. , it violates nec .
observations from wmap indicates the value @xmath11 @xcite , which means that our universe would be dominated by phantom energy ( @xmath12 ) .
it has been shown in @xcite that phantom dark energy can be successfully accomodated within framework of general relativity ( gr ) . in the context of bhs and phantom energy accretion on bh
, it should be mentioned that babichev et al @xcite has shown that bh mass decrease with phantom energy accretion on it .
jamil et al @xcite studied charged bhs in phantom cosmology .
jamil in @xcite has shown the evolution of a schwarzschild black hole in phantom - like chaplygin gas cosmologies .
primordial bhs in phantom cosmology and accretion of phantom de on btz bhs were also studied by jamil et al in @xcite .
nayak in @xcite investigated the effect of vacuum energy on the evolution of primordial bhs in einstein gravity .
paolis in @xcite studied bhs in bulk viscous cosmology . in the context of cyclic cosmology
, it should be mentioned that saridakis in @xcite studied cyclic universes from general collisionless braneworld models .
cai et al in @xcite investigated cyclic extension of the non - singular cosmology in a model of non - relativistic gravity .
cai et al in @xcite investigated cyclic and singularity - free evolutions in a universe governed by lagrange - multiplier modified gravity .
moreover cai et al in @xcite showed that gravity described by an arbitrary function of the torsion scalar , can provide a mechanism for realizing bouncing cosmologies , thereby avoiding the big bang singularity .
non - singular cyclic cosmology without phantom menace was also studied by cai et al in @xcite .
we intend to study the effects and consequences of phantom energy accretion onto bhs in a cyclic scenario of the universe described by dgp and rsii branes .
our motivation is to find out if there is any remanent mass of bh when it undergoes a turnaround in a cyclic scenario .
as mentioned earlier babichev et al @xcite has shown that bh mass decrease with phantom energy accretion on it .
hence the bh will disappear before the turnaround in an oscillating universe .
but sun @xcite provided a mechanism which showed that in an universe described by modified friedmann cosmology the destruction of bhs is totally out of question , as there is always a remanent mass of a bh facing a turnaround . in this paper
our motivation is to testify the above fact for brane - world cosmology and find out the fate of a bh undergoing phantom energy accretion in an oscillating universe .
the paper is organised as follows : in section 2 we discuss the mechanism of cyclic universe in rsii brane model .
section 3 deals with an identical mechanism for dgp brane model . in section 4 , we present an argument regarding hawking evaporation of remanent bhs .
finally the paper ends with some concluding remarks in section 5 .
the novel feature of the rs models compared to previous higher - dimensional models is that the observable 3 dimensions are protected from the large extra dimension ( at low energies ) by curvature rather than straightforward compactification . in rs
ii model the effective equations of motion on the 3-brane embedded in 5d bulk having @xmath13-symmetry are given by @xcite @xmath14 where @xmath15 @xmath16 and @xmath17 and @xmath18 is the electric part of the 5d weyl tensor . here
@xmath19 and @xmath1 are respectively the 5d gravitational coupling constant , 5d cosmological constant , the brane tension ( vacuum energy ) , brane energy - momentum tensor and effective 4d cosmological constant .
the explicit form of the above modified einstein equations in flat universe for rsii brane are @xmath20 and @xmath21 the dark radiation @xmath22 obeys @xmath23 where @xmath24 and @xmath25 are the total energy density and pressure respectively .
we consider an homogeneous and isotropic universe filled with de fluid , with de density @xmath24 and pressure @xmath25 . for an asymptotic observer the black hole mass ,
@xmath26 changes at the rate of @xcite @xmath27 here the overdot denotes the derivative with respect to cosmic time .
moreover it has been considered that @xmath28 .
we consider an universe dominated by de .
the friedmann equation for the expanding universe is given by @xmath29 where @xmath30 is the hubble parameter .
the conservation equation for de is given by @xmath31 phantom de has the equation of state @xmath32 .
we get that @xmath33 , which shows that @xmath24 increases with the expansion of the universe . in this section
we will investigate the dynamics of cyclic universe in rsii brane model .
the modified friedmann equation for rsii brane model is given by equation ( [ cyclic5.5 ] ) .
putting @xmath34 , the equation becomes , @xmath35 in the expanding phase of the universe , the phantom energy density , @xmath24 increases . from the above equation
we see that at the turnaround the phantom energy density is given by , @xmath36 where @xmath37 . from the above value we see that when @xmath38 , then @xmath39 , and when @xmath40 , then @xmath41 .
this shows that either the brane tension or the bulk cosmological constant has to be negative , so that the universe undergoes a bounce , and a possibility for the cyclic scenario is evident .
here @xmath43 stands for critical density . after
the turnaround the universe begins to contract .
the reason behind this being , that in the contracting phase the non - phantom components of the universe increase and begins to dominate the evolution . in this phase , again when the dominant energy density reaches the critical value given by equation ( [ cyclic5.12 ] ) , a bounce occurs .
thus we get an oscillating scenario .
we intend to study the variation of bh mass with phantom energy accretion around it , in this oscillating cosmological model . using equations ( [ cyclic5.8 ] ) and ( [ cyclic5.10 ] )
we get @xmath44 before turnaround , we have @xmath45 now substituting the above value of @xmath46 in equation ( [ cyclic5.13 ] ) we get , @xmath47 where @xmath48 and @xmath43 is given by equation ( [ cyclic5.12 ] ) . now integrating equation ( [ cyclic5.15 ] ) we get , @xmath49 here @xmath50 and @xmath51 denotes respectively the phantom energy density and the black hole mass at the moment when the phantom energy density just begins to dominate the evolution of the universe . in general , @xmath52 and @xmath53 . so using equation ( [ cyclic5.12 ] ) we obtain @xmath54 .
hence from equation ( [ cyclic5.16 ] ) , we obtain @xmath55 at the turnaround , @xmath56 and hence the black hole mass at the turnaround becomes , @xmath57 this shows that there is a remnant mass of the bh when the turnaround occurs .
hence this result is different from the result obtained by zhang @xcite .
it is quite clear from the above equations that initially , through phantom energy accretion , the bh mass decreases , until it reaches the minimum value @xmath58 at the turnaround in the expanding phase . for @xmath59 ,
@xmath58 becomes independent of @xmath51 @xmath60 after the turnaround as expected the the universe will contract and consequently the phantom energy density @xmath24 starts decreasing . during the contraction it is obvious that the time derivative of the scale factor will become negative , and as a result @xmath46 becomes negative
so , here we take the value of @xmath46 as , @xmath61 substituting the above value of @xmath46 in equation ( [ cyclic5.13 ] ) we get , @xmath62 now integrating the above equation we get , @xmath63 where @xmath64 .
the above equation shows that as the universe contracts , the bh mass continue to decrease .
when @xmath65 , then the bh mass is , @xmath66 for @xmath67 , using equation ( [ cyclic5.19 ] ) we find that the final mass of bhs is @xmath68 hence @xmath69 is independent of @xmath51 .
+ fig .
1 fig .
2 + fig 1 : the mass of the black hole is plotted against the increasing density of phantom dark energy before turnaround .
other parameters are fixed at @xmath70 , @xmath71 + fig 2 : the mass of the black hole is plotted against the decreasing density of phantom dark energy after turnaround .
other parameters are fixed at @xmath72 + in figure 1 , plot is generated for the mass of the bh against the density of phantom de , in the expanding cycle before the turnaround .
we know that in the expanding phase , the phantom energy components dominates over the non - phantom counterparts . as a result
, phantom energy begins accreting on the bhs in this cycle . from the figure , it is seen that with the increase in the phantom energy density before turnaround , the bh mass gradually decreases , due to more and more accreting phenomenon going on the bh .
this result is in complete accordance with babichev et al @xcite . in figure 2 ,
the bh mass is plotted against the phantom energy density , in the contracting cycle after the turnaround .
it is clearly evident from the figure that as the non - phantom components begin to dominate in the contracting phase , the bh mass cease to diminish .
this is quite an expected result .
while flat , homogeneous and isotropic brane is being considered , the friedmann equation in dgp brane model is modified to the equation @xmath73 where @xmath74 is the hubble parameter , @xmath24 is the total cosmic fluid energy density and @xmath75 is the cross - over scale which determines the transition from 4d to 5d behaviour and @xmath76 ( choosing @xmath77 ) . for @xmath78
, we have standard dgp@xmath79 model which is self accelerating model without any form of de , and effective @xmath80 is always non - phantom .
however for @xmath81 , we have dgp@xmath82 model which does not self accelerate but requires de on the brane .
it experiences 5d gravitational modifications to its dynamics which effectively screen de . like the original friedmann equation , and contrary to rsii brane equation ,
the dgp brane equation does not readily support an oscillating universe undergoing turnarounds .
this fact is quite obvious from the equation ( [ cyclic5.25 ] ) .
we can see that there is no possibility of turnaround .
therefore we will propose a modified dgp brane equation , that can support a cyclic universe .
sun et al in @xcite proposed a method , by which they were able to modify the friedmann equation into a form that avoids the big rip singularity and gives a bouncing cosmological model .
their investigations led to the fact that the character of physics changes remarkably near the planck scale .
we know that a de sitter universe with a cosmological constant , @xmath83 is similar to a bh .
also it has a temperature , @xmath84 .
hence by conjecturing physics at the planck scale , they actually modified the definition of hawking temperature , and subsequently obtained the modified friedmann equation .
sun in @xcite used this modified equation successfully to study the phantom energy accretion on bhs . hence , taking a leaf out of their book , we proceed to modify the dgp brane equation as follows : the proposed modified dgp brane equation is @xmath85 from the above equation , we see that the turnaround occurs for @xmath86 .
the expression for @xmath46 for the expanding phase is given by , @xmath87 using equation ( [ cyclic5.13 ] ) and ( [ cyclic5.27 ] ) we get , @xmath88 where @xmath89 . now integrating the above equation we get , @xmath90 @xmath91 @xmath92 @xmath93\ ] ] where @xmath94 , and @xmath95 now as in the case of rsii brane , @xmath96 , which implies that @xmath97 .
hence from the above equation we obtain , @xmath98 @xmath91 @xmath99 @xmath100\ ] ] at the turnaround @xmath86 .
hence we get , @xmath101 @xmath102 @xmath103\ ] ] the above equation gives the remanent mass of the bh at the turnaround .
hence we see that the mass of bh gradually decrease during the expanding phase of cyclic universe , due to phantom energy accretion and finally becomes minimum at the turnaround , when it is called the critical mass , given by equation([cyclic5.30 ] ) . for @xmath67
, @xmath58 becomes independent of @xmath51 and is given by the following relation , @xmath104 @xmath102 @xmath105\ ] ] just like the rsii model , in dgp model as well , the universe starts to contract after turnaround , and the non - phantom components starts to dominate the evolution .
the expression for @xmath46 is given by , @xmath106 using equation ( [ cyclic5.13 ] ) and ( [ cyclic5.33 ] ) we get , @xmath107 integrating the above equation we get , @xmath108 @xmath91 @xmath92 @xmath109\ ] ] now as in the previous section , @xmath96 , which implies that @xmath97 .
hence from the above equation we obtain , @xmath110 @xmath91 @xmath99 @xmath111\ ] ] at the turnaround @xmath86 .
hence we get , @xmath112 @xmath102 @xmath113\ ] ] as before the above equation gives the remanent mass of the bh at the turnaround . for @xmath67
, @xmath58 becomes independent of @xmath51 and is given by the following relation , @xmath114 @xmath115 @xmath116\ ] ]
from the above evaluation it is clear that there is always a remanent mass of the bhs at the turnaround in an oscillating universe .
so contrary to existing literature we conclude that there is no possibility of destruction of bh .
the bhs formed during the expanding cycle of the cyclic universe survive into the next cycle and eventually grow in size .
therefore they create undesired non - uniformity in a nearly uniform universe .
eventually the bhs will occupy the entire volume of the horizon and will be responsible for the destruction of the cyclic models .
this is a serious defect indeed !
hence the problem posed by the bhs in a cyclic universe still stands un - eliminated .
but in @xcite it has been argued that for a bh with mass @xmath117 , hawking evaporation takes place in time @xmath118 sec and ultimately the bh becomes non - existent , thus causing no problems . here
@xmath119 represents planck mass . in our above calculations
we have considered @xmath120 . rewriting equation ( [ cyclic5.18 ] ) , for rsii brane we get , @xmath121 @xmath122 is taken as a constant of the order unity and @xmath123 .
then we have @xmath124 . for a bh with @xmath67 , @xmath125 ,
i.e. , the remanent mass of the bh is of the order of planck mass .
hence the remanent bh undergoes hawking evaporation in time @xmath126 sec at the order of planck time .
a similar evaluation is possible for dgp brane model as well .
so fortunately , the remanent bhs do not cause any problems .
this gives a possible solution of the bh problem in cyclic universe described by brane - world scenario .
a serious problem is posed by the existence of black holes in an oscillating universe . in @xcite
brown et al suggested that in an oscillating cosmology the black holes keep losing mass due to phantom energy accretion before totally disappearing before the turnaround .
babichev et al in @xcite devised a successful mechanism which was in accordance with the result given by brown et al . in this paper
we have investigated the outcome of phantom energy accretion on black holes in a cyclic universe described by brane - worlds .
it is seen that rsii brane model readily incorporates the oscillating nature in its framework .
the only condition being the negativity of the brane tension , @xmath0 or the bulk cosmological constant , @xmath1 .
but unlike rsii brane model , dgp brane does not readily support the cyclic nature of the universe .
so a modified dgp brane equation , that supports cyclic turnarounds has been proposed for our evaluation .
it is found that for rsii brane model , during the expanding phase the black hole mass gradually decrease with the increase of phantom energy density and finally reaches a critical value at which the turnaround occurs .
this result is absolutely consistent with the known results in literature . in the contracting phase the black hole mass again decreases with the increase in the non - phantom components of the universe
this is however contrary to our expectations .
so it is understood that our evaluations after the turnaround are not really rigid .
we see that the black holes in a cyclic universe reaches a remanent mass , @xmath58 before turnaround .
so , the remanent mass implies that the destruction of black holes is not a real possibility in the cyclic cosmology , with phantom energy turnarounds , for a universe characterized by brane gravity .
however fortunately we find that the remanent masses of black holes at turnaround do not cause problems .
the reason being that these remanent black holes hawking evaporate in a time @xmath126 .
randall , l. , sundrum , r. : -_phys .
lett . _ * 83 * , 4690(1999b ) .
randall , l. , sundrum , r.:-_phys .
lett . _ * 83 * , 3770(1999a ) .
dvali , g. r. , gabadadze , g. , porrati , m.:- _ phys.lett .
b _ * 485 * 208(2000)[_arxiv_:*000506*(hep - th ) ] .
e. kamatsu et al . , astrophys .
192 , 18 ( 2011 ) armendariz - picon , jcap 0407 , 007 ( 2004 ) [ arxiv : astro - ph/0405267 ] ; l. senatore , phys . rev .
d 71 , 043512 ( 2005 ) ; p. creminelli , m. a. luty , a. nicolis and l. senatore , jhep 0612 , 080 ( 2006 ) ; v. a. rubakov , theor .
149 , 1651 ( 2006 ) [ arxiv : hep - th/0604153 ] ; m. libanov , v. rubakov , e. papantonopoulos , m. sami and s. tsujikawa , jcap 0708 , 010 ( 2007 ) arxiv:0704.1848 [ hep - th ] ; a. nicolis , r. rattazzi and e. trincherini , phys . rev .
d 79 , 064036 ( 2009 ) [ arxiv:0811.2197 [ hep - th ] ] ; c. deffayet , g. esposito - farese and a. vikman , phys .
d 79 , 084003 ( 2009 ) [ arxiv:0901.1314 [ hep - th ] ] ; c. deffayet , s. deser and g. esposito- farese , phys . rev .
d 80 , 064015 ( 2009 ) [ arxiv:0906.1967 [ gr - qc ] ] ; p. creminelli , a. nicolis and e. trincherini , jcap 1011 , 021 ( 2010 ) [ arxiv:1007.0027 [ hep - th ] ] .
babichev , e. , dokuchaev , v. , eroshenko , y. : - _ phys .
lett . _ * 93 * 021102(2004 ) [ gr - qc/0402089 ] jamil , m. , rashid , m. , a. , qadir , a. : - _ eur.phys.j.c _ * 58 * 325 ( 2008 ) [ _ arxiv_:*0808.1152v4*[astro - ph ] ] jamil , m. : - _ eur.phys.j.c _ * 62 * 609 ( 2009 ) [ _ arxiv_:*0906.2875v2*[gr - qc ] ] jamil , m. , qadir , a. : - _ gen.rel.grav _ * 43 * 1069 ( 2011 ) [ _ arxiv_:*0908.0444v3*[gr - qc ] ] jamil , m. , akbar , m. : - _ gen.rel.grav _ * 43 * 1061 ( 2011 ) [ _ arxiv_:*1005.3444v2*[gr - qc ] ] nayak , b. , jamil , m. : - _ phys .
b _ * 709 * 118 ( 2012 ) [ _ arxiv_:*1107.2025v1*[gr - qc ] ] paolis , f. , d. , jamil , m. , qadir , a. : - _ int.j.theor.phys . _ * 49 * 621 ( 2010 ) [ _ arxiv_:*0802.1264v4*[astro - ph ] ] saridakis , e. , n. : - _ nucl.phys.b _ * 808 * 224 ( 2009 ) [ _ arxiv_:*0710.5269v2*[hep - th ] ] .
cai , yi - fu . ,
saridakis , e. , n. : - _ jcap _ * 0910 * 020 ( 2009 ) [ _ arxiv_:*0906.1789v2*[hep - th ] ] .
cai , yi - fu . , saridakis , e. , n. : - _ class .
quantum grav .
_ * 28 * 035010 ( 2011 ) [ _ arxiv_:*1007.3204v2*[astro - ph.co ] ] .
cai , yi - fu . ,
chen , shih - hung . , dent , j. , b. , dutta , s. , saridakis , e. , n. : - _ class .
quantum grav .
_ 28 215011 ( 2011)[_arxiv_:*1104.4349 * [ astro-ph.co ] ] cai , yi - fu . ,
saridakis , e. , n. : - _ j.cosmol . _
* 17 * 7238 ( 2011 ) [ _ arxiv_:*1108.6052 * [ gr - qc ] ] sun , c. , y. : - [ _ arxiv_:*0806.1080v2*(hep - th ) ] ( 2008 ) .
maartens , r. : -_phys .
d _ * 62 * 084023 ( 2000 ) .
maartens , r. : -_living rev . relativity _
* 7 * 7(2004)[_arxiv_:*0312059*(gr - qc ) ] .
shiromizu , t. , maeda , k. and sasaki , m. : -_phys .
* 62 * 024012 ( 2000 ) .
maeda , k. and wands , d. : - _ phys .
d _ * 62 * 124009(2000 ) .
sasaki , m. , shiromizu , t. and maeda , k. : - _ phys .
d _ * 62 * 024008 ( 2000 ) .
zhang , x. : - [ _ arxiv_:*0708.1408*(gr - qc ) ] .
sun , c. , y. , zhang , d. , h. : - [ _ arxiv_:*0506218v1*(hep - th ) ] ( 2005 ) .
brown , m. g. , freese , k. , kinney , w. , h. : - _ jcap _ * 0803 * 002 ( 2008 ) [ _ arxiv_:*0405353*(astro - ph ) ] . | a universe described by braneworlds is studied in a cyclic scenario .
as expected such an oscillating universe will undergo turnarounds , whenever the phantom energy density reaches a critical value from either side .
it is found that a universe described by rsii brane model will readily undergo oscillations if , either the brane tension , @xmath0 or the bulk cosmological constant , @xmath1 is negative .
the dgp brane model does not readily undergo cyclic turnarounds .
hence for this model a modified equation is proposed to incorporate the cyclic nature .
it is found that there is always a remanent mass of a black hole at the verge of a turnaround . hence contrary to known results in literature , it is found that the destruction of black holes at the turnaround is completely out of question . finally to alleviate ,
if not solve , the problem posed by the black holes , it is argued that the remanent masses of the black holes do not act as a serious defect of the model because of hawking evaporation
. 0.2 cm |
the experimental setup ( fig . [ fig1 ] ) consisted of a brass container filled with solid carbon dioxide ( dry ice ) .
a clean glass slide was placed over the brass container , where a drop ( @xmath2 ) of deionized and degassed water was deposited using a syringe pump . to increase contrast and observe the freezing front
, red food dye was added to the water .
the process was recorded from the side using a long distance microscope ( vzm1000 edmund optics ) mounted on a color camera , at a frame rate of 50 frames per second .
we used both backlight and bottom light illumination provided by optic fiber lamps .
the resolution obtained was @xmath3 pixels with approximately @xmath4 .
we measured the plate temperature near the droplet using a standard thermocouple . | in this fluid dynamics video we show how a drop of water ( @xmath0 ) freezes into a singular shape when deposited on a cold surface ( @xmath1 ) .
the process of solidification can be observed very clearly due to the change in refraction when water turns into ice .
the drop remains approximately spherical during most of the process , with a freezing front moving upwards and smoothly following the interface .
however , at the final stage of freezing , when the last cap of liquid turns into ice , a singular tip develops spontaneously .
interestingly , the sharp tip of the ice drop acts as a preferential site for deposition of water vapour , and a beautiful `` tree '' of ice crystals develops right at the tip .
the tip singularity attracts the vapour in analogy to a sharp lightning rod attracting lightning . |
in mathematically modelling infectious diseases , often compartment models are used .
compartment models divide the population under consideration into disjunct sets of individuals with the same biological characteristics .
prominent examples in infectious disease modelling are the _
si _ , _ sis _ and _ sirs _ models , see for example @xcite .
the models have in common that they depend on one or more parameters .
for instance , all these models need a parameter , mostly called _ transmission rate _ @xmath0 that describes how effective contacts between susceptible and infected persons are with respect to spreading the disease . biological , chemical and physical properties of infectious agents as well as the behaviour of hosts , susceptible or infected , lead to a variety of possible values of @xmath1 even within the same class of disease the transmission characteristics may vary considerably , which was shown for example in influenza @xcite .
this may impose practical problems in estimating and predicting the parameters . in this work we analyse the temporal dynamics of the prevalence of infectious diseases in a non - parametric way .
the temporal change of the prevalence is expressed in terms of the incidence and the recovery rate .
we start with a simple compartment model that divides the population into those who are not infected ( suceptible ) , and those who are diseased ( figure [ fig : compmodel ] ) . the numbers of persons in the states _ susceptible _ and _ diseased _ are denoted by @xmath2 and @xmath3 ( cases ) .
the transition rates between the states are the incidence rate @xmath4 and the recovery rate @xmath5 , which depend on the time variable @xmath6 the equations characterising the changes of @xmath2 and @xmath3 in the compartment model of figure [ fig : compmodel ] are : [ e : balance ] @xmath7 by applying the quotient rule to the prevalence @xmath8 and inserting these equations we get the following scalar ordinary differential equation ( ode ) @xmath9 the linear ode shows that the temporal change @xmath10 of the prevalence is a convex combination of the incidence rate @xmath4 and the recovery rate @xmath5 . the solution of with the initial condition @xmath11 is @xmath12 where @xmath13 [ [ remark-1 ] ] remark 1 : + + + + + + + + + if @xmath4 and @xmath5 are constant and the disease is in equilibrium , i.e. @xmath14 equation in case of @xmath15 reads as @xmath16 this is the well - known result that the prevalence odds @xmath17 equals the product of incidence and mean duration of the disease .
[ [ remark-2 ] ] remark 2 : + + + + + + + + + for later use we define the triangle function @xmath18 let @xmath19 and @xmath20 then set @xmath21 the function has a triangular shape with a peak of height @xmath22 at @xmath23 an example of a triangle function is the red curve in figure [ fig : triangle ] .
in this section we illustrate equations and by some examples .
the first example assumes a rectangular time course of the incidence ( figure [ fig : rectangle ] ) .
the _ support _ of the incidence ( i.e. , the set @xmath24 ) is @xmath25 , the support of the recovery rate is @xmath26 . on these intervals the values of the incidence and recovery
are assumed to be @xmath27 and 0.85 ( per week ) , respectively . ) in example 1 . the resulting prevalence according to equation is
is the blue curve.,scaledwidth=95.0% ] the resulting prevalence ( calculated by numerically integrating equation ) is shown as blue curve in figure [ fig : rectangle ] . as we shall see in the next section
, the time course of the incidence in a wave of influenza does not have a rectangular shape .
it is ( approximately ) symmetric and has a peak in the middle .
compared to the previous example , a triangular shape of the incidence is more realistic .
we assume a wave of influenza having the incidence as shown by the red curve of figure [ fig : triangle ] : @xmath28 with @xmath29 ( per week ) .
the recovery rate @xmath5 is assumed to be a triangle function , too , with @xmath30 and peak height 1.7 ( per week ) .
the time course of the associated prevalence ( blue line in figure [ fig : triangle ] ) has been calculated by equation . from figure
[ fig : triangle ] it is apparent that the prevalence starts to increase later than the incidence . at about week 7 , the prevalence has overtaken the incidence . in week 10 the incidence peaks at @xmath31 ( per week ) , whereas the prevalence peaks at about 0.3 weeks later at the value @xmath32 in summary , we can see that the prevalence is delayed compared to the incidence and overshoots the peak of the incidence . to illustrate equation ,
we have chosen @xmath33 for @xmath34 and @xmath35 for @xmath36 .
the associated duration of the disease is @xmath37 ) and prevalence ( blue ) in example 3.,scaledwidth=80.0% ] beginning at @xmath38 the prevalence is constant ( equilibrium ) .
it holds @xmath39 for all @xmath36 .
figure [ fig : stationary ] shows the course of the associated prevalence ( blue ) . for use in the next section we solve the following problem : given the triangular incidence @xmath40 what has to be the minimal @xmath41 in a triangular recovery @xmath42 and @xmath43 such that @xmath44 for all @xmath45 ? with other words : what is the minimal peak height @xmath46 of a triangular recovery rate @xmath5 that follows after a triangular incidence @xmath4 with height @xmath47 such that the disease is eradicated at @xmath48 for @xmath45 it holds @xmath49 with @xmath50 it is easy to see that @xmath51 for all @xmath45 .
thus , we may speak of the _ terminal prevalence_. as the terminal prevalence @xmath52 is the product of two positive factors , the prevalence is positive for all @xmath48 hence , the only aim we can achieve is to bring @xmath52 below a prescribed threshold .
that this is possible , can be seen by the following calculation @xmath53 the inequality holds true , because @xmath54 is monotonically increasing .
from @xmath55 it follows that @xmath56 as @xmath57 for a given incidence @xmath58 and @xmath59 we are interested in the minimal @xmath22 such that the terminal prevalence @xmath52 is below a prescribed threshold . to solve this problem
, we examine the function @xmath60 where @xmath61 figure [ fig : terminalp ] gives an example of the terminal prevalence @xmath52 for @xmath62 ( incidence in figure [ fig : triangle ] ) and @xmath63 as a function of the peak height @xmath22.,scaledwidth=80.0% ] we will choose the threshold @xmath22 such that the terminal prevalence is one per mille of the peak incidence , i.e. @xmath64 then , we assume that the wave of influenza is eradicated after @xmath65 the corresponding peak height @xmath41 in case of @xmath62 and @xmath66 is @xmath67
figure [ fig : influenza ] shows the incidence of influenza in germany from 2001 to 2013 .
the abscissa and ordinate represent calendar time and age , respectively .
the colour indicates the incidence rate , the associated numerical values are coded as shown in the rightmost part of figure [ fig : influenza ] . by incidence
we mean the incidence reported to the national influenza register at the robert - koch - institute , @xcite .
in germany , all confirmed influenza cases statutorily have to be reported to the robert - koch - institute ( influenza a , b , c according to the reference definition ) . from figure
[ fig : influenza ] it becomes apparent that influenza usually appears in the first quarter of the year and vanishes ( nearly ) completely from the second to the fourth quarter .
an exception is the epidemic in the last quarter of 2009 .
then , the swine flu ( h1n1 influenza ) became pandemic .
we also see that not all age groups are affected equally from one wave of influenza to the other .
it seems that as calendar time progresses , the more older age groups get involved . in figure
[ fig : influenza ] a whitish line is visible in the lower third of the image .
this is the trajectory of a hypothetical birth cohort born in september 1998 , which has been followed from september 2001 to september 2013 .
the values of the incidence rate along the line is shown in figure [ fig : cohort ] .
the seasonal variability and the enormous peak during the swine flu pandemic are clearly visible .
.,scaledwidth=95.0% ] during follow - up , the birth cohort faces twelve waves of influenza with different intensities ( see figure [ fig : cohort ] ) .
three of them are analysed in more detail : the wave with relatively low incidence at 3.5 years of age ( spring 2002 ) , the moderate wave at age 8.5 ( spring 2007 ) and the swine flu at age 11.3 ( autumn 2009 ) .
the corresponding incidences are shown as black curves in figure [ fig : incidencethreewaves ] .
: spring 2002 ( left ) , spring 2007 ( middle ) and autumn 2009 ( right ) .
the black curves are the raw incidence rates as reported to the robert - koch - institute @xcite .
the triangle functions ( red ) are the approximated incidence rates .
note the different scalings of the ordinate.,scaledwidth=95.0% ] the raw incidence data are approximated by triangle functions @xmath68 @xmath69 these have been calculated by an ordinary least squares approach . as in example 4
( see above ) we assume triangle functions for the recovery rates @xmath70 the support of the recovery rate @xmath5 is assumed to be @xmath71 and @xmath72 this corresponds to a mean delay of 10 with range 3 - 17 ( days ) .
the peak heights @xmath73 of the recovery rates @xmath74 are calculated as in example 4 by forcing the terminal prevalence to be less than one per mille of the peak incidence .
the results are presented in table [ tab : waves ] .
c | c c | c c influenza & & + wave & support @xmath75 & peak height & support @xmath76 & peak height + spring 2002 & @xmath77 & @xmath78 & @xmath79 & 190.35 + spring 2007 & @xmath80 & @xmath81 & @xmath82 & 167.61 + autumn 2009 & @xmath83 & @xmath84 & @xmath85 & 137.85 + the time courses of the prevalence during the three waves of influenza are depicted in figure [ fig : prevalencethreewaves ] .
the shapes are very similar but the peak values differ considerably . during the waves in spring 2002 and 2007 the peak prevalence was @xmath86 and @xmath87 respectively .
the maximum of the prevalence during the swine flu epidemic in autumn was @xmath88 roughly speaking , the three peak prevalences differ by about one magnitude .
this is consistent with the incidence rates , which approximately differ by a factor of 10 .
in this work we have described a disease model that links the temporal change of the prevalence to the incidence and the recovery rates .
the model is non - parametric in the sense that it does not depend on biological , behavioural or disease - specific parameters .
it is only based on the common epidemiological measures incidence and recovery rate . in that respect
the model is very flexible and is not restricted to a specific class of infectious disease . after the introduction of the disease model , the characterising equations and some examples ,
the model has been applied to incidence of influenza in germany during 2001 - 2013 .
the incidence data stem from the robert - koch - institute , which is the official authority each confirmed case of influenza in germany by law has to be reported to .
since data about recovery rates are not published , assumptions had to be made . with these assumptions
the prevalence of influenza during three waves has been calculated for a hypothetical birth cohort . during the three waves of influenza , the resulting prevalence portions in the birth cohort are low .
there are mainly two reasons : the first is the short duration of symptoms of averagely 10 days ( range 3 - 17 days ) , which implies a high recovery rate with onset soon after the start of the wave of influenza .
the second reason lies in the data itself . presumably , the cases reported to the robert - koch - institute are only the most severe cases .
it is very likely that a lot of patients with the symptoms of influenza have not been examined by a medical doctor at all , or have not been examined in detail ( for example by pcr ) .
those cases have not been _ confirmed influenza cases _ that statutorily have to be reported .
thus , they are not covered by the incidence rates in this article .
the fraction of unreported cases is difficult to access and is beyond the scope of this work .
although the equations are mathematically correct in the context they were developed for ( no mortality , no migration ) , due to this coverage issue the calculated prevalence portions have to be interpreted very carefully .
another limitation of this article lies in the mathematical models for the incidence and recovery rates . here
, we have used rectangle and triangle functions . in countries with seasonal waves of influenza , incidence and
recovery rates vanish in certain periods .
thus , in modelling single waves of influenza , functions with bounded support would be preferable . here ,
we have chosen triangle and rectangle functions , but other functions may be possible as well , for example b - splines . in regions where influenza is present during the whole year , incidence and remission do not have bounded support .
so far , the model in figure [ fig : compmodel ] does not include the impact of mortality .
one may do so by changing the equations in : @xmath89 then , equation changes to @xmath90 with @xmath91 , @xcite . note that equation becomes equation in case of @xmath92 . here
an interesting point becomes obvious : equation is a consequence of equation .
however , while equation implyies @xmath93 i.e. the population size remains constant , equation does not imply this .
as it is a special case of equation , equation holds true in the presence of mortality with @xmath94 | in this work we describe a non - parametric disease model that links the temporal change of the prevalence of an infectious disease to the incidence and the recovery rates .
the model is only based on the common epidemiological measures incidence and recovery rate . as an application ,
the model is used to calculate the prevalence of influenza in germany for a hypothetical birth cohort during 2001 and 2013 . |
planets outside the solar system have been detected since 1992 @xcite from the doppler shift of spectral features in emission from the star , which determines the orbital period and places a constraint on the mass of the planet @xcite . since 1999 , the presence of exoplanets has also been deduced from their transit across the face of the parent star @xcite .
the fraction of light blocked by the planet allows the radius of the planet to be deduced as a function of the radius of the star @xcite .
subsequently , several hundred exoplanets have been detected at different distances from their stars , and the precision with which mass and radius have been deduced has increased for some exoplanets to better than 10% in both mass and radius @xcite . in 2001 , light from an exoplanet was detected directly @xcite , opening the window to studies of exoplanet structure through the composition and temperature of the surface or atmosphere .
however , inferences about the composition and structure rely on the comparison of mass and radius with planets within the solar system . with the exception of earth , which is currently the only planet for which seismic data exist , interpretations of the internal structure of the planets
rely in turn on assumptions about the composition and temperature profiles through the planet
. theories of planetary formation can be investigated by comparing the structure of exoplanets with those within the solar system .
another motivation is to estimate the occurrence of earth - like planets , in terms of mass and composition , and also those that might be habitable for life .
deductions about planetary structures , i.e. the composition profile , depend on the compressibility of the possible compositions thought to occur .
the compressibility is needed over the range of pressures and temperatures occurring within each planet .
the compressibility is found from the derivative of the pressure - density relation at the appropriate temperature , which can be determined from the equation of state ( eos ) for the particular composition of matter of interest .
the development of eos has been driven most by research in weapons ( explosives and projectile impacts ) ( for instance * ? ? ?
* ) , geophysics ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) , and inertial confinement fusion @xcite .
there is a perception that experimental and theoretical methods for determining eos are not available in regimes necessary to understand the internal structure of planets for pressures between 200gpa and 10tpa , i.e. from the limit of diamond anvil data to the onset of the thomas - fermi - dirac ( tfd ) regime @xcite . some studies ( e.g. * ?
* ) have considered sub - tfd eos with care , but it is common practice even when accurate theoretical calculations are available to represent the material with _ ad hoc _ functional forms , potentially leading to problems when extrapolating beyond the bounds of the constraining data and in switching over to tfd at high pressures .
although there is a definite need for more theoretical and experimental studies , appropriate theoretical techniques are just as accurate above 200gpa as below , and are more than adequate to draw inferences about the structure of exoplanets ; moreover , shock and ramp loading experiments can readily explore states up to @xmath35tpa and are suitable to test and calibrate eos . in this paper
, we study the validity of electronic structure methods for predicting eos up to the @xmath4tpa pressures apparently occurring in exoplanets , and the capability of dynamic loading experiments to measure relevant states .
we calculate mass - radius relations for several compositions of matter representing different classes of , and layers in , planets , discussing the validity of the eos used .
finally , we compare the mass - radius relations with representative planets and exoplanets .
the quasistatic structure of self - gravitating bodies depends on the scalar eos , which can be described by an appropriate free energy expressed as a function of its natural variables , such as the helmholtz free energy @xmath5 , where @xmath6 is mass density and @xmath7 temperature . in principle
, one could consider the composition as a set of explicit parameters in the eos . in practice ,
planetary structures are thought to comprise layers in each of which a single composition , or range of compositions , dominates the eos , such as fe in the core of rocky planets .
therefore , we consider a separate eos for each layer . as with dynamic loading situations ,
the pressure @xmath8 is the most directly useful aspect of the eos for calculating and interpreting planetary structures .
almost invariably , a thermodynamically incomplete eos is used : @xmath9 or @xmath10 where @xmath11 is the specific internal energy .
planetary structures may be constrained to follow a specified temperature profile , in which @xmath9 is the more convenient form , or an isentrope , for which @xmath10 is convenient since @xmath12 where @xmath13 . in planets , brown dwarfs , and main sequence stars ,
the eos is dominated by electrostatic forces and pauli exclusion among the electrons and ions , rather than by strong - force interactions between the nuclei . in stars , the radiation pressure
must be included , and the entropy is much higher , exploring a different region of the eos . in this section ,
we describe key theoretical methods and approximations used when predicting eos , and comment on their applicability to states likely to occur in exoplanets .
the relevant eos of matter can be calculated using electronic structure theory .
for a representative set of atoms defining the composition , thermodynamic potentials can be calculated as a function of @xmath6 ( i.e. system volume ) and @xmath7 . because the mass of the electrons is so much less than that of the nuclei ,
the state of the electrons can generally be considered with respect to the instantaneous positions of the nuclei the born - oppenheimer approximation .
forces on the nuclei can also be calculated with respect to their positions and the distribution of electrons .
the energy of the system can be thought of as comprising the ground state energy for stationary ions , thermal motion of the ions , and thermal excitation of the electrons out of their ground state . in principle , all contributions should be calculated self - consistently . however , for eos , it is generally a good approximation to partition the free energy into the cold compression curve , thermal motion of the ions , and thermal excitation of the electrons , @xmath14 ( see * ? ? ?
to describe the material state and the dynamics of most atoms , the quantum nature of the nuclei can be ignored , allowing their treatment as point - like masses exhibiting newtonian dynamics .
the exception is h , for which the quantum behavior of the proton can be represented as a correction to the potential experienced by the electrons .
the motion of the nuclei can be calculated with respect to time .
this technique , known as first - principles molecular dynamics ( fpmd ) or quantum molecular dynamics ( qmd ) @xcite , is particularly appropriate for fluids , unknown or ill - defined crystal structures , and multi - species compositions of matter where it is not clear _ a priori _ where particular species will be located .
in addition to the eos , fpmd has been used to predict transport properties , including the electric and thermal conductivities @xcite and viscosity @xcite , which are important in understanding the formation and thermal profile of planets as well as the generation of their magnetic fields .
fpmd has also been used in studies of the free energy profiles of mixtures to determine possible miscibility gaps which could have a large impact on planetary evolution and structure @xcite . for crystalline structures ,
the motion of the nuclei can be calculated in terms of oscillations about their equilibrium positions , i.e. phonon modes . in principle , phonon modes should be calculated self - consistently with electron excitations ; in practice , for most conditions encountered in planets , the ion - thermal motion can be calculated from the phonon density of states at @xmath15 , and the electron - thermal contribution can often be ignored .
@xmath16 comprises the ground state energy of the system as a function of compression , @xmath17 , and any configurational entropy that may be associated with different structural polymorphs or the entropy of mixing in an impure system @xcite .
one may calculate @xmath18 by setting up a configuration of nuclei , and finding the ground state of the hamiltonian of the electrons .
the configuration of the nuclei may be altered or relaxed under the net force they experience to find the lowest - energy structure , but they are often held fixed in likely structures . for well - defined crystalline structures , phonon modes can be predicted by calculating the electrostatic forces on the nuclei as each is displaced from equilibrium .
explicit calculation of the phonons is desirable below the debye temperature , where the zero - point motion of the nuclei and the freezing out of modes may have a significant effect on the eos @xcite .
thermal excitation of the electrons can be calculated from the band structure , which is the set of eigenstates of the electronic hamiltonian . for sufficiently high temperatures ,
the eigenstates must be calculated self - consistently with their population , but accurate calculations can be made to @xmath3ev temperatures using the @xmath15 band structure . the key physics , and thus the source of limitations in the accuracy and validity of the eos , is the electronic structure calculation .
the challenge is in representing the multi - fermion nature of the electrons accurately , but with a method that is computationally tractable for real materials .
path integral methods , such as path - integral monte - carlo ( pimc ) , can be used for direct , rigorous calculations of multi - particle states @xcite . however
, these methods are practicable mostly for low-@xmath19 materials , and have not been applied systematically ( or , generally , at all ) to compositions of matter relevant to ice and rocky planets .
there is , however , no reason to suspect that these techniques are unsuitable under exoplanet conditions _ per se_.
although tfd - based electronic structure theory @xcite is often regarded as an adequate treatment for pressures over @xmath310tpa ( for instance * ? ? ?
* ; * ? ? ?
* ) , it does not capture the effects of electronic shell structure which are expected at high pressures @xcite .
however , shell effects such as pressure ionization are captured by the electronic structure techniques considered here .
most theoretical eos calculations use density functional theory ( dft ) @xcite and its variants to treat exchange and correlation between the electrons via functionals of the electron density , calibrated to reproduce the same system energy as techniques that treat the electron wavefunctions more directly .
dft functionals are calibrated against calculations of idealized electron gases , which may be performed up to arbitrarily high densities : they should be no less accurate for the relatively modest absolute compressions occurring in the cores of exoplanets than at @xmath20 .
dft calculations typically reproduce the observed mass density to within a few percent ( for instance * ? ? ?
* ; * ? ? ?
* ) , which is an observable discrepancy when compared with dynamic or quasistatic loading measurements .
calculations can be corrected by adjusting the energy to reproduce the mass density at zero pressure , which is most accurately measured .
the resulting eos which we have termed ` ab fere initio ' to distinguish them from uncorrected ab initio eos then typically reproduce high pressure measurements to within their uncertainties @xcite . in many situations ,
electrons closer to the nuclei are not affected by changes in the compression or temperature , and the states of the other electrons may be calculated more efficiently by subsuming the inner electrons into a modified nuclear potential : the pseudopotential method @xcite .
pseudopotential calculations can become inaccurate when the nuclei are compressed sufficiently closely together .
the validity can be checked by comparing against all - electron calculations , and may be accurate to severalfold compressions or pressures of terapascals @xcite . with these caveats , pseudopotential techniques and the dft
construct in general are not inherently unsuitable for predicting eos in the planetary structure regime .
however , detailed calculations of many - species compounds are computationally expensive , and predictions of polymorphic structures are sensitive to relatively small inaccuracies in the computational methods .
it is highly desirable to compare eos predictions against experimental measurements . for the most part
, we have used previously - developed eos based on experimental data or validated electronic structure calculations up to the limit of available data , and blending into tfd predictions at high compressions @xcite .
these eos incorporate thermal effects , allowing us to investigate the sensitivity of planetary structures to temperature profile .
for fe and fe - ni , we have compared predictions from such eos with 3d electronic structure calculations extended to higher pressures than have been reported previously .
previous theoretical studies include very careful treatments of the eos of fe up to pressures and temperatures representative of the earth s core ( in particular * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* and references therein ) , and it is very desirable to perform equivalent studies to higher pressures and temperatures necessary for the study of exoplanets . for our present purposes ,
it is most useful to compare eos for different compositions of matter constructed according to consistent prescriptions .
the use of wide - ranging eos constructed using optimized algorithms , and wide - ranging sets of electronic structure calculations made using a consistent method that reproduces tfd at extreme compressions , allows us to avoid any reliance on extrapolating using _
ad hoc _ functional forms such as vinet and birch - murnaghan for pressure - density relations , which can give unquantified uncertainties outside the range of the fitting data .
the canonical experimental technique for studying the properties at matter at high pressure is shock loading , using a variety of methods to induce a shock wave .
shock measurements of eos are often performed relative to a reference material , but an attraction of shock loading is that experiments can in principle be configured to yield absolute measurements , if the shock is induced by the impact of a projectile with a target of the same material .
indeed , the pressure standards in static compression apparatus such as diamond anvil cells are ultimately calibrated against absolute shock measurements .
although the time scales in dynamic loading experiments are typically nanoseconds to microseconds , typical equilibration times for electrons and atomic vibrations are much shorter , so inferred states used for testing and calibrating eos are in thermodynamic equilibrium and thus equivalent to quasistatic compression measurements made in presses such as diamond anvil cells .
indeed , the difference in time scale between dynamic and quasistatic loading is less than the difference between the latter and planetary ages .
although eos measurements are in thermodynamic equilibrium with respect to a given phase of matter , the time dependence of phase transitions must be considered : phase changes often occur with a significant degree of superpressurization or superheating in dynamic loading experiments and indeed in quasistatic loading , compared with the equilibrium phase boundary .
the effects of time dependence are also evident as hysteresis in the location of the phase change on loading compared with unloading .
in shock loading , the entropy increases with compression , so the temperature rises faster with compression than it does along an isentrope .
the grneisen parameter and heat capacity of a material , which may be estimated theoretically or experimentally , can be used to predict the eos away from the locus of shock states .
measurements along the principal shock hugoniot have been used in this way for many years to estimate the principal isentrope and the cold curve @xcite . while a direct measurement of a relevant state is preferable , shock - derived eos are likely to be adequate for exoplanets over a wide range of pressures . furthermore ,
if a theoretical eos is validated by shock experiments , this provides reasonable confidence that the eos is valid at lower temperatures .
a more serious limitation with shock experiments is that , for a given material and starting state , there is a limit to the compression that can be achieved by the passage of a single shock , and therefore a limit to the range of compressions that can be deduced along isentropes .
however , this compression ( discussed further below for different compositions of matter ) is adequate to explore matter into the massive exoplanet regime .
shock measurements have been made to pressures of up to 10tpa using radiation from underground nuclear explosions @xcite .
the compression limit can be circumvented through the use of multiple shocks , which induce lower temperatures than a single shock to the same pressure . the ultimate limit for an infinite sequence of infinitesimal shocks
is a ramp compression .
ramp compression of condensed matter was first demonstrated using the expanding reaction products from a chemical explosive @xcite , and has subsequently been demonstrated using pulsed magnetic fields at the z accelerator @xcite and high power lasers . with lasers ,
the pulse energy may be used to vaporize a foil which then loads the sample in an analogous way to the chemical reaction products @xcite , the pulse may ablate the sample directly with an intensity history designed to induce a ramp in pressure @xcite , or the sample may be driven from a hohlraum with a power history chosen to control the hohlraum temperature history such that a ramp is induced .
ramp loading using the last variant has been demonstrated to @xmath31tpa on the omega laser @xcite , with near - term experiments on fe planned to reach 2tpa at the national ignition facility ( nif ) .
experimental techniques have thus been available to develop and test eos under conditions relevant to exoplanets , and planned developments should provide the first direct measurements of matter under exoplanet core conditions .
if the rotation rate is not extreme , self - gravitating bodies are close to spherical .
spherical structures are certainly an adequate starting place for studies of exoplanets .
the condition for isostatic equilibrium is that the stress induced by pressure variations is balanced by the gravitational acceleration @xmath21 : @xmath22 for newtonian gravitation , by gauss theorem , @xmath21 can be expressed in terms of the mass enclosed within a given radius @xmath23 : @xmath24 so @xmath25 @xmath23 can be calculated simply from the distribution of mass density , @xmath26 or @xmath27 when the total mass @xmath28 , outer radius @xmath29 , and surface composition are known with reasonable confidence , and the objective is to determine an aspect of the internal structure such as the core radius , the equations can be integrated from the surface inward , starting at the known @xmath30 ( hence @xmath31 ) and @xmath32 @xcite .
the internal structure parameter is found as the solution of a shooting problem to ensure that @xmath33 as @xmath34 . to calculate generic mass - radius relations for a uniform composition
, it is most efficient to start at the center , where @xmath35 and @xmath36 , choose the core pressure @xmath37 , and integrate outward until @xmath20 . in this way
, the mass - radius condition is obtained in terms of @xmath38 as an independent parameter : @xmath39 .
these solution methods are unwieldy when it is desired to constrain the overall composition as a mixture of components , such as a fixed ratio between fe , rock , and h / he . in this case
, the isostatic structure may be found more conveniently by starting with the desired masses of each component , and solving the system of equations as an elliptic problem .
a numerical solution may be found iteratively by taking an approximate solution ( which may initially be concentric shells of material at their @xmath20 density ) and improving it by calculating the net stress gradient at each radius @xmath40 and @xmath41 ( ignoring the relatively slow change of @xmath42 with deformation , which is valid near equilibrium ) , and hence estimating a change in radius for each shell such that the net pressure is zero .
the radial variation was represented discretely by values at a finite series of radii .
multiple iterations are necessary to allow the effect of a net pressure at some radius to be balanced through equilibration of the whole structure , but standard techniques may be used for solving elliptic equations efficiently , such as successive over - relaxation @xcite .
a series of solutions describing a mass - radius relation can be found efficiently by working upward in mass , and increasing the density at each radius by a constant factor , then calculating the modified equilibrium structure .
we verified that this method of solution gave the same result as integration from the center to the surface , by comparison with structures comprising pure fe . to close the system of equations , a @xmath43 relation is needed for the material .
the choice of @xmath43 relation implies a choice for the entropy or temperature profile of the body . in the present study we usually chose an isentrope passing through a reasonable surface state , such as stp for rock or metal , or a few atmospheres pressure and cryogenic temperatures for compositions that are gaseous at stp .
isentropes were calculated by numerical integration of the relation @xmath44 using a procedure valid for eos of arbitrary form @xcite . for a few sample compositions ,
structures were calculated for @xmath43 along the cold curve .
the precision of @xmath43 curves necessary to be useful for exoplanet structures is lower than for solar system planets , and the difference in structures between the stp isentrope and cold curve was found to be negligible over the range of pressures considered , as shown below .
mass - radius relations were calculated for a series of compositions of matter relevant to ( exo)planets , using as a baseline eos developed from and for shock wave applications , in particular the sesame library @xcite , for which the constituent assumptions and calibrations are reasonably well documented .
sesame eos are tabulated functions @xmath45 .
most sesame eos are tabulated over a wide enough range of states to be valid ( at least in principle , to some finite precision ) for massive exoplanets and star formation .
they were constructed to incorporate experimental data , mainly shock measurements , where available .
experimental measurements not used in the construction of the eos , including shock states , release from shock , and ramp compression , are frequently compared against the sesame eos , so there is a relatively rich literature on experimental validation .
many of the eos were constructed using the same theoretical and modeling approaches , so conclusions drawn from the use of one eos are more likely to apply to other materials .
when these eos are compared with subsequent shock or ramp measurements at higher pressures , it is quite usual to find significant differences , i.e. inaccuracies in the eos , but they are usually at the level of a few times the experimental uncertainty ( for instance * ? ? ?
* ; * ? ? ?
* ) , which would not have much effect on the mass - radius relation . for all undifferentiated compositions
, the low - mass limit behaves as @xmath46 and deviates as matter within the planet is compressed gravitationally above its zero - pressure density @xmath47 .
the deviations become substantial as the central pressure approaches the bulk modulus . as the mass increases further
, an increasing proportion is compressed to significantly higher density , until eventually any increase in mass is accommodated entirely by density increase , leading to a maximum in radius . for fe , which is the dominant component of interest at the highest pressures
, we are developing eos better optimized to exoplanet core states , to complement nif experiments @xcite .
thermodynamically complete versions of these eos were not available for this study : only the cold curves .
the effect of temperature was assessed by comparing structures calculated with @xmath43 along an isentrope compared with the cold curve , for thermodynamically complete fe eos , and was found to be negligible , in agreement with previous work using a simplified thermal contribution @xcite .
the mass - radius relation derived from our new fe cold curve , which was calculated with dft using the local density approximation and troullier - martins pseudopotentials @xcite , agreed well with the relations from sesame eos 2140 @xcite and 2150 @xcite ( fig . [ fig : femr ] ) .
fe@xmath48ni was considered as a representative fe - ni composition , calculated using the same dft method as for previous studies of intermetallic compounds @xcite .
the pseudopotential for ni was used previously for constructing eos for nial @xcite . comparing @xmath43 relations ,
fe@xmath48ni was significantly softer than fe or a segregated fe - ni mixture of the same composition , but the mass - radius relations did not differ significantly .
the mass - radius relation was also calculated for the principal isentrope from the linear grneisen eos found previously to perform well for rocky planets @xcite ; it deviated from the other fe relations for masses above approximately twice that of the earth ( m@xmath49 ) , demonstrating the inaccuracy of mass - radius relations deduced from eos extrapolated from low - pressure measurements .
mass - radius relations deduced from the dft cold curves were very similar to those from the sesame eos up to around 1r@xmath49 or 3m@xmath49 ( @xmath31tpa ) , above which the properties calculated with dft became progressively softer .
this observation highlights the importance of ramp - loading experiments to several terapascals . with the exception of the linear grneisen eos , these relations are similar , but not identical , to relations used recently in the interpretation of exoplanet corot-7b @xcite and for fe mass - radius relations @xcite
interestingly , the relation by valencia et al follows the linear grneisen curve up to around 2m@xmath49 , then approaches the @xmath15 dft relations above 10m@xmath49 ; the fortney et al relation is stiffer ( larger radius ) up to around 1m@xmath49 , then follows the relation for sesame 2150 very closely .
the mass - radius relations derived from the sesame eos were denser above 1m@xmath49 .
( figs [ fig : femr ] and [ fig : femp ] . ) ni . ] ni . ] for fe , the maximum shock compression deduced from the sesame eos was a factor of around 5.2 , calculated using a solution method valid for general forms of the eos @xcite .
at the same mass density on the principal isentrope , the pressure is approximately 11tpa .
thus single shock experiments , corrected appropriately for thermal effects , can plausibly explore planetary pressures to this regime .
impact - induced shock experiments , with the projectile driven by gas , chemical propellant , or high explosive , have been performed on fe to around 1tpa by several groups @xcite . at higher pressures , data are very rare .
nuclear - driven shock impedance mismatch measurements @xcite reached a pressure which we estimate to be @xmath50tpa .
further measurements above 1tpa are desirable .
high quality theoretical and experimental studies have been performed for mantle constituents such as mgsio@xmath48 and sio@xmath0 up to pressures around 150gpa relevant to earth s core - mantle boundary , and the effect of fe ions ( see * ? ? ?
? * ; * ? ? ?
* ; * ? ? ?
as compositions representative of the variations in rocks , mass - radius relations were calculated for sio@xmath0 , mgo , and basalt .
sio@xmath0 was represented by sesame eos 7383 @xcite .
mgo was represented by sesame eos 7460 @xcite , and also by a dft - derived two phase eos including quasiharmonic phonons @xcite , but extending only to @xmath30.4tpa .
the mass - radius relations for the different mgo models were indistinguishable , demonstrating insensitivity of the mass - radius relation to the inclusion of phase transitions , for current exoplanet applications .
basalt was represented by sesame eos 7530 @xcite .
` basalt ' refers to the composition ( by number of atoms : 60.11% o , 18.26% si , 5.96% al , 4.01% ca , 3.48% fe , 3.39% mg , 2.18% h , 1.56% na , 0.55% ti , 0.38% k , 0.07% p , 0.06% mn ) and initial state : the eos was derived from shock data including phase changes , and was blended into tfd predictions at high pressures .
we also compared the linear grneisen eos for basalt found previously to give better than expected structures for the rocky planets @xcite ; its mass - radius relation was similar to that of the sesame eos to several tenths of a terapascal , and then diverged significantly .
the mass - radius and pressure - radius relations were similar but distinguishable on scales extending to exoplanets , giving an indication of the possible uncertainty in inferences about rocky planet structure in the absence of constraints on the composition : the mass - radius relation does show sensitivity to the composition , but the sensitivity is small compared to current uncertainties in _ exoplanet _ observables .
recent mass - radius calculations by @xcite for mg - silicate planets were almost indistinguishable from the mgo relation ; slightly older calculations by @xcite using a linear mixture of 38% sio@xmath0 , 25% mgo , 25% fes , and 12% feo followed our calculation for basalt ( sesame ) very closely to several tens of m@xmath49 .
( figs [ fig : rockmr ] and [ fig : rockmp ] . ) , mgo , and basalt , including a simplified , linear grneisen eos ( lg ) , and comparing with previous calculations @xcite . ] , mgo , and basalt , including a simplified , linear grneisen eos ( lg ) . ] _ ab initio _ treatments of mixtures of molecules of low molecular weight , usually called ices ( regardless of being in the solid or fluid state ) , are complicated and computationally expensive because of the possibility of segregation of species and the formation of a wide variety of chemical species . at issue here is the uncertainty of the equilibrium concentration of the different molecules .
such calculations have been performed for relatively small numbers of thermodynamic states @xcite , and will in future allow wide - ranging eos to be predicted as a function of composition .
a complementary approach is to calculate the equilibrium chemical composition given the thermodynamic state and atomic composition , using measured or estimated thermodynamic potentials for each possible chemical specie .
chemical equilibrium calculations have been used to predict the eos of the reaction products of chemical explosives up to temperatures of several hundred gigapascals and temperatures of @xmath3ev , including many species in common with planetary ices @xcite .
however , the thermodynamic potentials are not necessarily well - constrained for the hot , compressed states occurring within exoplanets . as compositions representative of the variations in ice giants ,
mass - radius relations were calculated for h@xmath0o , nh@xmath48 , and ch@xmath1 , for isentropes chosen to pass through a condensed phase so that the planet surface could be defined as zero pressure .
the calculated radius is thus a lower bound for planets with a thick atmosphere , as is likely to be the case for those orbiting close to their star .
several eos were available for h@xmath0o ; sesame eos 7154 was the most appropriate for these pressures @xcite .
the stp isentrope was used .
the only sesame eos available for nh@xmath48 was 5520 @xcite , fitted to national bureau of standards ( nbs ) gas phase data , and was tabulated to a maximum mass density of 0.765 g @xmath51 , which is barely greater than typical for liquid ammonia .
an empirical linear grneisen eos was constructed from shock wave data on liquid ammonia at an initial temperature of 203k @xcite , using an approximate treatment for off - hugoniot states .
the corresponding mass - radius calculation should be regarded with caution as it is not constrained by experimental data for pressures above a few tens of gigapascals .
the mass - radius relation and central pressure were in reasonable agreement with predictions using sesame eos 5520 up to a mass @xmath30.1@xmath52 above which the isentrope from the tabular eos became unusable .
for ch@xmath1 , three sesame eos were available : 5500 and 5501 @xcite , both tabulated to 2.5 g @xmath51 and respectively with a maxwell construction and van der waals loops in the liquid - vapor region ; and 5502 @xcite , tabulated to 0.47 g @xmath51 and based on nbs gas phase data .
the mass - radius relation was calculated for 5500 , which is the most relevant for planetary structures , for an isentrope passing through 70k at zero pressure .
the difference between h@xmath0o and ch@xmath1 was considerable : it is difficult to infer a composition for an ice giant from the mass - radius measurement alone to a factor @xmath32 in mass or @xmath325% in radius , without further constraints on the composition .
( figs [ fig : icemr ] and [ fig : icemp ] . )
o , nh@xmath48 , and ch@xmath1 ices . ] o , nh@xmath48 , and ch@xmath1 ices .
the crossing of the ch@xmath1 by the nh@xmath48 curve is at a mass density much greater than the shock data used to construct the nh@xmath48 eos , and should not be regarded as a meaningful prediction .
the crossing of the h@xmath0o and nh@xmath48 curves is in a regime where the eos should be adequately accurate . ]
jovian planets are thought to be composed predominantly of a mixture of h with approximately 5 - 10% he by number of atoms .
h is theoretically interesting because quantum mechanical uncertainty in the position of the proton is more important than for other elements , and may affect the eos .
significant work has been devoted to predicting the eos of h and h / he mixtures @xcite .
these h / he mixtures are challenging because even the relatively small amount of helium involved is thought to greatly impact the eos .
specifically , the insulator - to - metal transition that occurs at @xmath4gpa pressures in pure h @xcite is thought to be suppressed to even higher pressures and temperatures with the inclusion of trace amounts of he @xcite .
furthermore , helium is thought to segregate ( i.e. form droplets ) in the metallic region of the planet@xcite , where the resulting he rain is thought to produce heat as the droplets fall down the gravitational well .
this effect could have a disproportionately large effect on the mean eos and on the differentiation of a planet by gravitational separation .
we did not find a h / he eos of similar pedigree to the other eos above , so the mass - radius relation was calculated for pure h ( sesame eos 5251 : @xcite ) as a limiting case of low density .
this eos was isotopically scaled from sesame eos 5263 for d@xmath0 , which is wide - ranging and includes dissociation and ionization @xcite , and has been found to give good agreement with static and shock compression data . as with the ices ,
the isentrope was chosen such that a well - defined planetary surface was calculated , and is thus a lower bound on the radius of planets with a thick atmosphere .
for the eos used , the isentrope passed through 10k at zero pressure , and exhibited a kink at around 100m@xmath49 caused by metallization at around 200gpa .
interestingly , the resulting mass - radius relation is quite close to the tfd prediction @xcite .
our calculation has a maximum in radius , 79000 km at a mass of @xmath53 kg .
this result is slightly higher in mass and lower in radius than the tfd prediction .
( figs [ fig : hmr ] and [ fig : hmp ] . ) mass - radius relations were calculated for differentiated two - component planets comprising an fe core and a basalt ( composition ) mantle , with fixed mass ratios , using the elliptic method of solution .
fe and basalt were represented by sesame eos 2150 and 7530 respectively . because of its low density , the addition of an outer basalt layer gave a disproportionate change in the radius , to a greater extent than indicated in previous work @xcite .
[ fig : febasaltmr ] . )
for comparative purposes , we chose to represent each type of material by a single mass - radius curve corresponding to a single eos : fe - ni by the electronic structure cold curve for fe@xmath48ni ( which is similar to the fe sesame 2150 curve but extends to higher pressure ) , rock by sesame 5530 for basalt , ice by sesame 7154 for h@xmath0o , and h / he by sesame 5251 for h. plotting the planets of the solar system , they fall in the expected places with respect to the curves : the rocky planets between fe - ni and rock with mercury closest to fe - ni and mars closest to rock , jupiter and saturn close to the h / he curve , and uranus and neptune close to the ch@xmath1 curve .
minor planets pluto and eris are also shown , lying on the icy side of the basalt curve .
we also compare exoplanets at extremes of size and mass : ` super - earths ' kepler-10b @xcite , corot-7b @xcite , and gj 1214b @xcite and the ` super - jupiters ' hat - p-2b @xcite and corot-3b @xcite , whose core pressures are representative of states it would be useful to explore in future experiments on material properties , and the anomalously large corot-2b .
( figs [ fig : mrcmp2 ] and [ fig : mrcmp ] ; exoplanet parameters from @xcite . ) measurements of exoplanet kepler-10b constrain its radius to @xmath54@xmath52 and mass to @xmath55@xmath56 .
the nominal value lies on our mass - radius relation for fe - basalt differentiated planets with 2/3 of the total mass in the core . with this structure ,
the central pressure would be @xmath57tpa .
considering two - layer differentiated structures at the @xmath58 level , the planet could have a core mass fraction between 25% and 95% , with central pressure between about 1.5 and 3.7tpa .
measurements of exoplanet corot-7b constrain its radius to @xmath59r@xmath49 and mass to @xmath60m@xmath49@xcite , though other researchers have deduced different , and inconsistent , masses from the same doppler shift data @xcite .
given the current discrepancy in mass deduced by different groups , the validity of deduced compositions and central pressures is unclear , but it is informative to assess the values and uncertainties derived from the quoted uncertainties in mass and radius .
the nominal value by @xcite lies very close to our mass - radius relation for basalt , for which the central pressure would be @xmath61tpa . considering two - layer differentiated structures at the @xmath58 level , the planet could be fe / rock with a core of radius up to 4500 km , or rock / ice with a surface layer of ( for example ) h@xmath0o up to 2500 km thick . the central pressure is calculated to increase rapidly with fe , giving @xmath62tpa . for three - layer differentiated structures ,
the ice layer would be thicker as the core size increased .
the uncertainty in deduced composition , core radius , and central pressure is nonlinear with respect to the uncertainty in mass and radius .
measurements of doppler shift in the stellar spectrum and dimming in the stellar brightness are essentially independent , so we calculate uncertainty contours in radius and central pressure and smooth them to infer the @xmath58 uncertainty .
the planet is very unlikely to be fe - only , in agreement with the previous conclusions @xcite .
in fact , corot-7b is likely to have a smaller core in proportion to its size than do the terrestrial planets .
the presence of ice is thought to be unlikely because of the proximity of the planet to its star , corot-7 , resulting in high surface temperatures .
the nominal mass - radius measurement is consistent with a composition of rock only ; given the uncertainties in the mass - radius measurement and relations , a metallic core can not be ruled out .
recent measurements of exoplanet gj 1214b give a radius of @xmath63r@xmath49 and mass of @xmath64m@xmath49 @xcite or @xmath65m@xmath49 @xcite , lying between the mass - radius curves for h@xmath0o and ch@xmath1 . the closest analogue in the solar system
is the icy giants ; the mass and radius of gj 1214b are consistent with a higher proportion of h@xmath0o , or with a proportionately larger amount of rock or fe , than neptune or uranus .
the observations do not of themselves suggest an ` ocean planet ' as the planet s density seems implausibly low . furthermore , as has been pointed out previously @xcite , a mean density consistent with a ` water world ' could arise from a rocky interior with an h / he atmosphere .
recent parameters for the super - massive exoplanet hat - p-2b are a mass of @xmath66m@xmath67 and a radius of @xmath68r@xmath67 @xcite .
these parameters are consistent with a massive hydrogen - rich planet , as concluded previously .
proportionately , the radius is smaller with respect to our predicted mass - radius curve for hydrogen than are jupiter or saturn , suggesting a larger proportion of rock or fe .
a pure hydrogen planet of the nominal mass would have a radius of 1.12mj and a central pressure of 214tpa .
the actual proportion of rock that would be consistent with the observed mass and radius is difficult to estimate using the available eos , because they roll over at lower masses ; future studies will include theoretical prediction and experimental investigation of the eos under relevant conditions .
the super - massive exoplanet corot-3b has a mass of @xmath69m@xmath67and @xmath70r@xmath67 @xcite , lying close to our mass - radius curve for h. our mass - radius relation for h confirms the anomalously large radius of corot-2b . as proposed recently @xcite ,
a plausible explanation is that this planet is young and hot , from accretion , tidal heating , or internal processes .
the radius would be larger for a warmer isentrope , but is relatively insensitive to temperature .
another potential source of uncertainty , particularly for warm bodies with a significant proportion of volatiles , is the sensitivity to definition of the outer surface of the body .
if the low - density gas in the outer region of the atmosphere has a higher opacity than expected , the planet would appear larger .
using the location of a given planet with respect to the mass - radius relation for different compositions , it is possible to estimate the pressure in the center of the core from the corresponding mass - pressure relation ( fig .
[ fig : mpcmp ] ) .
thus the central pressure of kepler-10b is likely to be between 1.5 and 2.7tpa , and that of corot-7b is likely to be between 0.7 and 2tpa , which are accessible using planar ramp - loading experiments at the omega and nif facilities .
the core pressure in gj 1214b is approximately 1tpa : the same mass of h@xmath0o gives a pressure of around 1.5tpa , and that mass of ch@xmath1 , 0.5tpa
. it would be most useful to improve eos of ice compositions into this regime , by qmd simulations and dynamic loading experiments ; again , planar ramp experiments at nif should be able to access relevant states .
the core pressure in hat - p-7b is approximately 20tpa , which may be accessible with subsequent nif experiments using convergent compression .
the core of corot-3b is at the highest pressure known for non - stellar matter : 1900tpa according to the eos used .
such pressures are achieved in inertial confinement fusion experiments , but will require very significant experimental developments before eos experiments can be performed .
mass - radius relations were derived using eos widely used for , and thus validated against , dynamic compression experiments .
the results were similar to relations used previously to infer planetary structures , but were not identical .
mass - radius relations for fe were inferred using eos calculated to higher pressures than reported previously , using density functional theory .
the difference between the cold curve and stp isentrope had a negligible effect on the mass - radius relation in the exoplanet regime .
the effect of ni in fe was significant for the eos , but negligible for the mass - radius relation of exoplanets . in a differentiated planet ,
the radius was more sensitive to small proportions of rock than has been concluded previously .
these results support the conclusion that corot-7b , with measurements analyzed by @xcite , is ` earth - like ' in that it is likely to have a rocky layer over an fe - ni core , though the present uncertainty in the exoplanet s mass is too large to constrain the mass of fe significantly : there may be very little .
gj 1214b lies between the curves for h@xmath0o and ch@xmath1 , which are compositions where improvements to the eos are particularly desirable . the composition can not be constrained meaningfully from present data , and suggestions that it might be an ` ocean planet ' are speculative .
the present results also support the conclusion that hat - p-2b is a gas giant , though the density is proportionately larger than the gas giants in the solar system , indicating a higher proportion of matter of higher atomic number .
the mass - radius relations imply core pressures of @xmath71tpa for corot-7b , a range range 0.5 - 1.5tpa for gj 1214b , over 210tpa in hat - p-2b , and around 1900tpa in corot-3b .
based on the constituent physics and approximations , we conclude that current theoretical techniques for electronic structure , with care in the use of pseudopotentials , should be as suitable for predicting eos in the exoplanet regime as they are at low pressures .
shock measurements are relevant in terms of the compressions that can be explored , though interpretation is involved to account for the elevated temperatures .
ramp loading techniques are very appropriate , and planned nif experiments should provide data on fe in the region where dft mass - radius relations deviate from sesame eos . in both experiment and theory ,
the limiting cases of fe ( or fe - ni ) and h are the subject of active research .
rocks and ices are more complicated in terms of composition and structure , and there is clearly a need for future research into the properties of these materials at compressions relevant for planets and exoplanets .
the authors would like to acknowledge the invaluable contributions of g.i .
kerley , j.d .
johnson at los alamos national laboratory , w.m .
howard at lawrence livermore national laboratory , and k. rice at the royal observatory , edinburgh .
this work was performed under the auspices of the u.s .
department of energy under contract de - ac52 - 07na27344 . | for planets other than earth , particularly exoplanets , interpretation of the composition and structure depends largely on comparing the mass and radius with the composition expected given their distance from the parent star .
the composition implies a mass - radius relation which relies heavily on equations of state calculated from electronic structure theory and measured experimentally on earth .
we lay out a method for deriving and testing equations of state , and deduce mass - radius and mass - pressure relations for key , relevant materials whose equation of state is reasonably well established , and for differentiated fe / rock .
we find that variations in the equation of state , such as may arise when extrapolating from low pressure data , can have significant effects on predicted mass - radius relations , and on planetary pressure profiles .
the relations are compared with the observed masses and radii of planets and exoplanets , broadly supporting recent inferences about exoplanet structures .
kepler-10b is apparently ` earth - like , ' likely with a proportionately larger core than earth s , nominally 2/3 of the mass of the planet .
corot-7b is consistent with a rocky mantle over an fe - based core which is likely to be proportionately smaller than earth s .
gj 1214b lies between the mass - radius curves for h@xmath0o and ch@xmath1 , suggesting an ` icy ' composition with a relatively large core or a relatively large proportion of h@xmath0o .
corot-2b is less dense than the hydrogen relation , which could be explained by an anomalously high degree of heating or by higher than assumed atmospheric opacity .
hat - p-2b is slightly denser than the mass - radius relation for hydrogen , suggesting the presence of a significant amount of matter of higher atomic number .
corot-3b lies close to the hydrogen relation .
the pressure at the center of kepler-10b is @xmath2tpa .
the central pressure in corot-7b is probably close to 0.8tpa , though may be up to 2tpa .
these pressures are accessible by planar shock and ramp loading experiments at large laser facilities .
the center of hat - p-2b is probably around 210tpa , in the range of planned national ignition facility experiments , and that of corot-3b around 1900tpa . |
the nature of dark matter and dark energy is one of the most important issues today in physics
. there are strong observational evidences in astrophysics and cosmology for the existence of these two components of the cosmic energy budget , indicating that about @xmath7 of the universe is composed by dark matter ( about @xmath8 ) and by dark energy ( about @xmath9 ) , but no direct detection has been reported until now .
the usual candidates * for * dark matter ( neutralinos and axions , for example ) and dark energy ( cosmological constant , quintessence , etc . ) lead to very robust scenarios , but at same time they must face theoretical and observational issues . for recent reviews on the subject ,
see for example @xcite .
the strongest issue is perhaps the one regarding dark energy as the vacuum expectation value of some quantum field , which would be a natural candidate , but whose correct theoretical value could be predicted only in the framework of a complete theory of quantum gravity , which still we do not possess . nevertheless , it is possible , at least , to guess some of the features of this theory .
in particular , the holographic principle @xcite may shed some light on the dark energy problem . according to this principle , in presence of gravity the number of the degrees of freedom of a local quantum system
would be related to the area of its boundary , rather than to the volume of the system ( as expected when gravity is absent ) . following this *
idea * , in @xcite the authors suggested an entanglement relation between the infrared and ultraviolet * cutoffs * due to the limitation set by the formation of a black hole , which sets an upper bound for the vacuum energy .
we can then interpret the ultraviolet cutoff as the vacuum density value , but still we need an ansatz for the infrared cutoff . as a candidate for such distance ,
in @xcite the authors propose and investigate the future event horizon , tested against type ia supernovae data and cosmic microwave background anisotropies in @xcite .
we shall present more detail on this in sec . [
sec : holde ] .
adding new components of dark energy to the whole energy budget in order to explain the current observation is a way , but not the only one .
since general relativity has been thoroughly tested up to solar system scales , it may be possible that the einstein - hilbert action contain corrections on larger , cosmological , scales thereby candidating as possible explanation of the evolution of the universe .
such modifications should be , in principle , composed by higher order curvature invariant terms ( such as @xmath10 , @xmath11 , etc ) but also by non - trivial coupling between matter or fields and geometry .
see for example @xcite for some reviews on the subject ( especially on @xmath6 theory ) .
it is also worth pointing out that these terms should naturally emerge as quantum corrections in the low energy effective action of quantum gravity or string theory @xcite . in this paper
we connect these two approaches , considering a @xmath0 theory of gravity , where @xmath1 is the ricci scalar , whereas @xmath2 is the trace of the stress - energy momentum .
this modified gravity theory has been recently introduced in @xcite , where the authors derived the field equations and considered several cases , relevant in cosmology and astrophysics . as for the former
, @xmath0 models have been constructed describing the transition from the matter dominated phase to the late times accelerated one @xcite .
our task here , is to find out which form the function @xmath0 has to have in order to reproduce the same properties of the holographic dark energy proposed in @xcite . to this purpose
, we employ the same reconstruction scheme proposed and employed in @xcite . for reference , in order to track the contribution of the @xmath2 part of the action in the reconstruction , we consider two special @xmath0 models : in the first instance , we investigate the modification @xmath3 , i.e. the usual einstein - hilbert term plus a @xmath4 correction . in the second instance
we consider a @xmath5 theory , i.e. a @xmath2 correction to the renown @xmath6 gravity . in both cases , we consider dark energy accompanied by a pressureless matter component ( which would determine @xmath2 ) .
the paper is organised as follows . in sec .
[ sec : holde ] , the equations of motion are established and the holographic dark energy introduced . in sec .
[ sec : simpl ] and [ sec : complcase ] the above mentioned cases are analysed .
finally , sec .
[ sec : discconcl ] is devoted to discussion and conclusions .
we use @xmath12 units and adopt the metric formalism , i.e. * the variation of the action is considered with respect to the metric quantities . *
in @xcite , the following modification of * einstein s * theory is proposed : @xmath13 where @xmath0 is an arbitrary function of the ricci scalar @xmath1 and of the trace @xmath2 of the energy - momentum tensor , defined as @xmath14 where @xmath15 is the matter lagrangian density .
we assume the matter lagrangian to depend on the metric , so that @xmath16 varying action with respect to the metric @xmath17 , one obtains @xcite @xmath18 where the * subscripts * @xmath1 or @xmath2 * imply * derivation with respect that quantity and we have also defined @xmath19 * planning * a cosmological application , we assume matter to be described by a perfect fluid energy - momentum tensor @xmath20 and that @xmath21 , so that we have @xmath22 and eq .
simplifies as @xmath23 in order to compare * it * with * einstein s * , we cast the above equation as follows : @xmath24 where @xmath25 is the einstein tensor .
now we can identify @xmath26 as the _ effective _ matter energy - momentum tensor and @xmath27 as the energy - momentum tensor of a `` geometric '' matter component .
we now assume a * background * described by the friedmann - lematre - robertson - walker metric @xmath28 with spatially flat hypersurfaces , and find a form for the function @xmath0 which is able to reconstruct * holographic dark energy*. according to the holographic principle @xcite an entanglement relation between the infrared ( ir ) and ultraviolet ( uv ) cut - offs of a quantum theory , due to the limitation set by the formation of a black hole , sets an upper bound for the vacuum energy @xcite : @xmath29 where @xmath30 is a free parameter and the ir ( large scales ) cutoff @xmath31 needs to be specified by an ansatz .
we are interested in the one proposed in @xcite : @xmath32 i.e. the future event horizon , that is the distance covered by a photon from now until the remote future .
note that the very presence of a vacuum energy component makes the above integration finite .
we consider a model composed by holographic dark energy plus ordinary pressureless matter , i.e. @xmath33 where @xmath34 is the hubble parameter and the dot denotes derivation with respect to the cosmic time .
introduce the critical energy density @xmath35 * * , we define * * @xmath36 using eqs . and , it is easy to show that @xmath37 the holographic dark energy density @xmath38 evolves according to the conservation law @xmath39 because in eq .
we have implicitly assumed the matter component to conserve separately .
now , using eqs . , and , one can find @xmath40 comparing with one can read off @xmath41 moreover , combining eq .
with eqs . and , the evolution for @xmath42 is determined by the following equation : @xmath43 where the prime denotes derivation with respect to the redshift @xmath44 .
testing this model against type ia supernovae and cosmic microwave background anisotropies , @xmath30 turns out to be constrained around unity , with the case @xmath45 favoured @xcite .
note that , from eq .
, @xmath45 means that the universe will end up in a phantom phase . for more comprehensive analysis of holographic dark energy models ,
we refer the reader to @xcite .
in the next section we investigate a simple case of reconstruction of @xmath0 .
we now consider a single perfect fluid model with density @xmath46 and pressure @xmath47 , together with the following ansatz ( one of the first considered in @xcite ) : @xmath48 i.e. the action is given by the same einstein - hilbert one plus a function of @xmath2 .
this is a particularly interesting choice since , from eqs . and
, we get @xmath49 for @xmath50 one has @xmath51 and , choosing @xmath52 one can construct a model with an effective cosmological constant @xcite . from eq . one can read off the effective energy density and pressure of the universe content : @xmath53 and therefore a dark energy component may appear , even if we are considering a single perfect fluid model . from eqs . and
, it is clear that we can pick out a `` fictitious '' component , due to @xmath4 , described by @xmath54 and , provided @xmath55 positive , it may well describe a dark energy component , since its pressure is negative . in order to reconstruct the function @xmath55 starting from the holographic principle
, we note that the equation of state parameter of the dark component induced by @xmath55 is @xmath56 and we identify it with @xmath57 , the one provided by the holographic dark energy in eq .
: @xmath58 for the standard model given in eq . , consider the fluid component to be pressureless matter , i.e. @xmath50 .
we are left to solve the following system of equations : @xmath59 where we have used @xmath51 , because we are considering pressureless matter . in order to solve the above system
, we have to fix some initial conditions . clearly , @xmath60 , and we choose @xmath61 , accordingly with current cosmological observation . for the initial condition on @xmath55 , from eq .
( with @xmath50 ) we write @xmath62\omega_{\rm m0 } + \frac{f(\rho_0)}{3h_0 ^ 2 } = 1\;.\ ] ] evaluating eq . today and combining it with eq .
we find the following algebraic equation determining the initial condition on @xmath55 : @xmath63 as we expected , when @xmath64 , then @xmath65 and eq
. implies that @xmath55 is identically vanishing .
right panel : evolution of @xmath42 .
the cases considered are @xmath66 ( solid black , dashed red , dot - dashed blue and dotted green , respectively ) .
we have chosen as initial conditions in @xmath61 the values @xmath67 and @xmath68 given by eq . .
note that @xmath55 and @xmath46 are normalised to @xmath69 .
the vertical lines in the plots represent @xmath70 , i.e. the present instant.,title="fig : " ] .
right panel : evolution of @xmath42 .
the cases considered are @xmath66 ( solid black , dashed red , dot - dashed blue and dotted green , respectively ) .
we have chosen as initial conditions in @xmath61 the values @xmath67 and @xmath68 given by eq . .
note that @xmath55 and @xmath46 are normalised to @xmath69 .
the vertical lines in the plots represent @xmath70 , i.e. the present instant.,title="fig : " ] + in we plot the solution of the system of differential equations and .
note that we normalise @xmath55 and @xmath46 to @xmath69 . as expected from eq .
, for large values of @xmath46 , i.e. far in the past , @xmath71 because the dark energy component is subdominant .
the actual difference among the various choices of @xmath30 takes place at late times , for small values of @xmath46 . again from inspection of eq .
, we can see that for large values of @xmath30 the linear evolution @xmath71 is again solution .
that is why in the left panel of the curve seems to `` straighten up '' for increasing @xmath30 . in
we plot the same quantities , but as functions of the redshift , in order to make clearer their cosmological evolution . as independent variable.,title="fig : " ] as independent variable.,title="fig : " ] a final remark about the future evolution .
it is clear from eq .
, that when @xmath72 , in the remote future , the solution for @xmath55 gets the asymptotic form @xmath73 therefore , we would have a future singularity for @xmath74 , as it appears for the relevant cases of and . the special case @xmath75 implies an asymptotically constant @xmath55 .
we now turn our discussion on a more general case , where the curvature @xmath1 * comes into * the action * as * a function to be determined .
now we turn our attention to the special case @xmath76 i.e. a @xmath2-linear correction to the class of @xmath6 theories . with the ansatz the matter content
is `` corrected '' as follows : @xmath77 whereas the geometry induced stress - energy tensor is @xmath78g_{\mu\nu } - \frac{1}{2}g_{\mu\nu } r\;.\ ] ] our aim is now to reconstruct the form of the @xmath6 which is able to reproduce the holographic dark energy paradigm .
we again consider a pressureless perfect fluid with density @xmath46 and again assume metric . from eqs . and
the effective density and pressure are the following : @xmath79 from eq .
it appears that the energy density of the perfect fluid is rescaled by a factor @xmath80 .
looking at eq . , we can extract a form for @xmath38 in the following way : @xmath81 whereas the form of @xmath82 is already given in eq .
, since our fluid is pressureless . from eqs .
and we can write the following differential equation for @xmath83 : @xmath84f_r = -\rho(1 + \lambda)\;.\ ] ] note , as a cross - check , that for @xmath85 the above equation simplifies to @xmath86 which possesses the particular solution @xmath87 , i.e. @xmath88 , the original einstein - hilbert action plus an integration `` cosmological '' constant .
we expect this solution to be the only one , otherwise there would exist an alternative @xmath6 theory which would behave exactly as general relativity .
let us speculate a bit more on this point . if @xmath85 , i.e. for a pure einstein - de sitter universe , we have from eq .
@xmath89 where @xmath90 is the present cosmic time ( i.e. the age of the universe ) . considering the homogeneous part of eq . and looking for a solution of the form @xmath91
, we find : @xmath92 which gives : @xmath93 and the general solution of eq .
can be written as : @xmath94 now , the initial conditions we adopt here are @xmath95 and @xmath96 .
the reason is essentially not spoiling the agreement between general relativity and solar system tests , see @xcite .
however , we stress here that these two conditions also imply @xmath97 and therefore restore the general relativity limit @xmath87 when @xmath85 . changing the variable to the redshift and employing eqs .
and one can recast eq . in the following compact form :
@xmath98f_r ' - 3\left[1 - \omega_{\rm v}\left(\frac{1}{3 } + \frac{2\sqrt{\omega_{\rm v}}}{3b}\right)\right]f_r = \nonumber\\ - 3(1 + \lambda)(1 - \omega_{\rm v})\;,\end{aligned}\ ] ] where again the prime denotes derivation with respect to the redshift @xmath44 .
the curvature @xmath1 can be easily found as @xmath99\;.\ ] ] the initial conditions @xmath95 and @xmath96 , translated to the redshift variable , are @xmath100 finally , the initial condition on @xmath55 can be extracted by eq . , being that @xmath101 .
thus , we have @xmath102 note the correction due to the @xmath103 term . as a function of the curvature @xmath1 .
left panel : @xmath104 , i.e. we are considering a pure @xmath6 theory , and @xmath66 ( solid black , dashed red , dot - dashed blue and dotted green , respectively ) .
right panel : @xmath105 and @xmath106 ( solid black , dashed red , dot - dashed blue and dotted green , respectively .
the curves appear superposed . ) .
we have chosen as initial conditions in @xmath61 the values @xmath67 .
note that @xmath55 and @xmath1 are normalised to @xmath69 and the redshift interval chosen is @xmath107.,title="fig : " ] as a function of the curvature @xmath1 .
left panel : @xmath104 , i.e. we are considering a pure @xmath6 theory , and @xmath66 ( solid black , dashed red , dot - dashed blue and dotted green , respectively ) .
right panel : @xmath105 and @xmath106 ( solid black , dashed red , dot - dashed blue and dotted green , respectively .
the curves appear superposed . ) .
we have chosen as initial conditions in @xmath61 the values @xmath67 .
note that @xmath55 and @xmath1 are normalised to @xmath69 and the redshift interval chosen is @xmath107.,title="fig : " ] in we plot the solution for @xmath55 .
note that we normalise @xmath55 and @xmath1 to @xmath69 . in the left panel we fix @xmath104 , i.e. we are actually considering a pure @xmath6 theory , and vary @xmath30 . in the right panel , on the other hand
, we consider positive and negative values of @xmath108 .
as one may note , @xmath108 has a poor influence on the evolution of @xmath55 .
we could expect this from inspection of eq . .
indeed , @xmath108 only enters the source term on the right hand side and therefore , when @xmath42 grows to unity , its impact on the evolution of @xmath55 is weak . on the other hand , larger values of @xmath108
may have a relevant effect at early times , determining the slope of @xmath55 . .
note , in the left panel , that the evolution of @xmath55 starts from below for all the cases.,title="fig : " ] .
note , in the left panel , that the evolution of @xmath55 starts from below for all the cases.,title="fig : " ] in and we display the future evolution of @xmath55 , both as function of @xmath1 or of @xmath44 . for the same reason stated above , the effect of @xmath108 is not relevant .
as function of @xmath44.,title="fig : " ] as function of @xmath44.,title="fig : " ] finally , in we plot the solution for @xmath42 .
again , its evolution appears to be independent of @xmath108 . as a function of the redshift .
left panel : @xmath104 , i.e. we are considering a pure @xmath6 theory , and @xmath66 ( black , dashed red , dash - dotted blue and dotted green , respectively ) .
right panel : @xmath105 and @xmath106 ( black , dashed red , dash - dotted blue and dotted green , respectively ) .
note that the curves are superposed .
we have chosen as initial conditions in @xmath61 the values @xmath67.,title="fig : " ] as a function of the redshift . left panel : @xmath104 , i.e. we are considering a pure @xmath6 theory , and @xmath66 ( black , dashed red , dash - dotted blue and dotted green , respectively ) .
right panel : @xmath105 and @xmath106 ( black , dashed red , dash - dotted blue and dotted green , respectively ) .
note that the curves are superposed .
we have chosen as initial conditions in @xmath61 the values @xmath67.,title="fig : " ]
in this work we have investigated a description of holographic dark energy in terms of suitably reconstructed @xmath109 gravity theories .
the latter have been recently introduced as modifications of einstein s theory possessing some interesting solutions * which are * relevant in cosmology and astrophysics @xcite .
we have * considered two special * types of models : @xmath110 , i.e. a correction to the einstein - hilbert action depending on the matter content , and @xmath111 , i.e. a simple @xmath2-linear correction to the class of @xmath6 theories .
since we have assumed the matter content to be a pressureless perfect fluid , then @xmath51 , i.e. the corrections assumed are directly dependent on the energy density of the universe content .
we * have * constructed differential equations for the function @xmath55 under investigation and numerically solved * them * , physically specifying the required initial conditions .
our simple analysis shows that holographic dark energy models are contained in the larger class of @xmath0 theories , at least considering a given background evolution of the universe
. it would be interesting to investigate how the evolution of matter perturbations would change , depending on the description of dark energy .
we expect , in principle , different results when using holographic dark energy or its @xmath0 reconstruction and therefore there is possibility for discriminating between the two descriptions . for example , it would be interesting to adapt the recently proposed scheme for perturbations in @xmath6 theories @xcite to the broader @xmath0 class .
we leave this as a future work .
mjsh and ofp thank professor s. d. odintsov for useful comments and also cnpq ( brazil ) for partial financial support .
99 t. padmanabhan , `` cosmological constant : the weight of the vacuum , '' phys .
* 380 * ( 2003 ) 235 - 320 .
[ hep - th/0212290 ] .
j. e. peebles , b. ratra , `` the cosmological constant and dark energy , '' rev .
* 75 * ( 2003 ) 559 - 606 .
[ astro - ph/0207347 ] .
v. sahni , `` dark matter and dark energy , '' lect .
notes phys .
* 653 * ( 2004 ) 141 - 180 .
[ astro - ph/0403324 ] .
g. bertone , d. hooper , j. silk , `` particle dark matter : evidence , candidates and constraints , '' phys .
* 405 * ( 2005 ) 279 - 390 .
[ hep - ph/0404175 ] .
v. sahni , a. starobinsky , `` reconstructing dark energy , '' int .
j. mod .
phys . * d15 * ( 2006 ) 2105 - 2132 .
[ astro - ph/0610026 ] .
e. j. copeland , m. sami , s. tsujikawa , `` dynamics of dark energy , '' int .
j. mod .
* d15 * ( 2006 ) 1753 - 1936 .
[ hep - th/0603057 ] .
j. martin , `` quintessence : a mini - review , '' mod
. phys .
* a23 * ( 2008 ) 1252 - 1265 .
[ arxiv:0803.4076 [ astro - ph ] ] .
j. frieman , m. turner , d. huterer , `` dark energy and the accelerating universe , '' ann .
astrophys .
* 46 * ( 2008 ) 385 - 432 .
[ arxiv:0803.0982 [ astro - ph ] ] .
r. r. caldwell , m. kamionkowski , `` the physics of cosmic acceleration , '' ann .
nucl . part .
* 59 * ( 2009 ) 397 - 429 .
[ arxiv:0903.0866 [ astro-ph.co ] ] .
m. li , x. -d .
li , s. wang , y. wang , `` dark energy , '' commun .
* 56 * ( 2011 ) 525 - 604 .
[ arxiv:1103.5870 [ astro-ph.co ] ] .
g. t hooft , `` dimensional reduction in quantum gravity , '' [ gr - qc/9310026 ] .
l. susskind , `` the world as a hologram , '' j. math .
* 36 * ( 1995 ) 6377 - 6396 .
[ hep - th/9409089 ] .
r. bousso , `` the holographic principle , '' rev .
* 74 * ( 2002 ) 825 - 874 .
[ hep - th/0203101 ] .
a. g. cohen , d. b. kaplan , a. e. nelson , `` effective field theory , black holes , and the cosmological constant , '' phys .
* 82 * ( 1999 ) 4971 - 4974 .
[ hep - th/9803132 ] .
m. li , `` a model of holographic dark energy , '' phys .
lett . * b603 * ( 2004 ) 1 .
[ hep - th/0403127 ] .
q. -g .
huang , m. li , `` the holographic dark energy in a non - flat universe , '' jcap * 0408 * ( 2004 ) 013 .
[ astro - ph/0404229 ] .
x. zhang , f. -q .
wu , `` constraints on holographic dark energy from type ia supernova observations , '' phys
. rev .
* d72 * ( 2005 ) 043524 .
[ astro - ph/0506310 ] .
x. zhang , f. -q .
wu , `` constraints on holographic dark energy from latest supernovae , galaxy clustering , and cosmic microwave background anisotropy observations , '' phys .
* d76 * ( 2007 ) 023502 .
[ astro - ph/0701405 ] .
s. i .
nojiri , s. d. odintsov , `` introduction to modified gravity and gravitational alternative for dark energy , '' econf * c0602061 * ( 2006 ) 06 .
[ hep - th/0601213 ] .
s. i .
nojiri , s. d. odintsov , `` unified cosmic history in modified gravity : from f(r ) theory to lorentz non - invariant models , '' phys .
* 505 * ( 2011 ) 59 - 144 .
[ arxiv:1011.0544 [ gr - qc ] ] .
l. amendola , d. polarski , s. tsujikawa , `` are f(r ) dark energy models cosmologically viable ? , '' phys
. rev .
* 98 * ( 2007 ) 131302 .
[ astro - ph/0603703 ] .
l. amendola , r. gannouji , d. polarski , s. tsujikawa , `` conditions for the cosmological viability of f(r ) dark energy models , '' phys .
rev . * d75 * ( 2007 ) 083504 .
[ gr - qc/0612180 ] .
s. capozziello , m. francaviglia , `` extended theories of gravity and their cosmological and astrophysical applications , '' gen .
* 40 * ( 2008 ) 357 - 420 .
[ arxiv:0706.1146 [ astro - ph ] ] .
t. p. sotiriou , v. faraoni , `` f(r ) theories of gravity , '' rev .
* 82 * ( 2010 ) 451 - 497 .
[ arxiv:0805.1726 [ gr - qc ] ] .
a. de felice , s. tsujikawa , `` f(r ) theories , '' living rev .
* 13 * ( 2010 ) 3 .
[ arxiv:1002.4928 [ gr - qc ] ] .
i. l. buchbinder , s. d. odintsov , i. l. shapiro , `` effective action in quantum gravity , '' bristol , uk : iop ( 1992 ) 413 p. t. harko , f. s. n. lobo , s. nojiri , s. d. odintsov , `` @xmath0 gravity , '' phys .
* d84 * ( 2011 ) 024020 .
[ arxiv:1104.2669 [ gr - qc ] ] .
m. j. s. houndjo , `` reconstruction of f(r , t ) gravity describing matter dominated and accelerated phases , '' [ arxiv:1107.3887 [ astro-ph.co ] ] .
s. capozziello , v. f. cardone , a. troisi , `` reconciling dark energy models with f(r ) theories , '' phys . rev .
* d71 * ( 2005 ) 043503 .
[ astro - ph/0501426 ] .
s. i .
nojiri , s. d. odintsov , `` modified f(r ) gravity consistent with realistic cosmology : from matter dominated epoch to dark energy universe , '' phys .
* d74 * ( 2006 ) 086005 .
[ hep - th/0608008 ] .
s. i .
nojiri , s. d. odintsov , h. stefancic , `` on the way from matter - dominated era to dark energy universe , '' phys .
* d74 * ( 2006 ) 086009 .
[ hep - th/0608168 ] .
s. i .
nojiri , s. d. odintsov , `` modified gravity and its reconstruction from the universe expansion history , '' j. phys .
* 66 * ( 2007 ) 012005 .
[ hep - th/0611071 ] .
x. wu , z. -h .
zhu , `` reconstructing f(r ) theory according to holographic dark energy , '' phys .
* b660 * ( 2008 ) 293 - 298 .
[ arxiv:0712.3603 [ astro - ph ] ] .
d. pavon , w. zimdahl , `` holographic dark energy and cosmic coincidence , '' phys .
* b628 * ( 2005 ) 206 - 210 .
[ gr - qc/0505020 ] .
s. del campo , j. .c .
fabris , r. herrera , w. zimdahl , `` on holographic dark - energy models , '' phys .
* d83 * ( 2011 ) 123006 .
[ arxiv:1103.3441 [ astro-ph.co ] ] .
n. j. poplawski , `` a lagrangian description of interacting dark energy , '' [ gr - qc/0608031 ] .
d. bertacca , n. bartolo , s. matarrese , `` a new approach to cosmological perturbations in f(r ) models , '' [ arxiv:1109.2082 [ astro-ph.co ] ] . | * we consider cosmological scenarios based on @xmath0 theories of gravity ( @xmath1 is the ricci scalar and @xmath2 is the trace of the energy - momentum tensor ) and numerically reconstruct the function @xmath0 which is able to reproduce the same expansion history generated , in the standard general relativity theory , by dark matter and holographic dark energy .
we consider two special @xmath0 models : in the first instance , we investigate the modification @xmath3 , i.e. the usual einstein - hilbert term plus a @xmath4 correction . in the second instance
, we consider a @xmath5 theory , i.e. a @xmath2 correction to the renown @xmath6 theory of gravity . * |
Subsets and Splits