article
stringlengths 0
745k
| abstract
stringlengths 0
92.7k
|
|---|---|
_ systolic complexes _ were introduced for the first time by chepoi @xcite under the name _ bridged complexes_. they were defined there as flag simplicial completions of _ bridged graphs _
@xcite the latter appear naturally in many contexts related to metric properties of graphs and , in particular , convexity .
systolic complexes were rediscovered by januszkiewicz- ' swiatkowski @xcite and haglund @xcite , independently , in the context of geometric group theory .
namely , _ systolic groups _ ,
that is , groups acting geometrically on systolic complexes have been used to construct important new examples of highly dimensional gromov hyperbolic groups
see e.g. @xcite . despite systolic groups and complexes
can have arbitrarily high dimension , in many aspects they behave like two - dimensional objects .
this phenomenon was studied thoroughly in @xcite ( see also appendix ) , and the intuitive picture is that systolic complexes and groups asymptotically do not contain essential spheres " .
such behavior is typical for objects of dimension at most two , but systolic complexes and groups , and related classes ( see e.g. @xcite ) are the first highly dimensional counterparts .
this implies , in particular , that systolic groups are very distinct from many classical groups , e.g. from uniform lattices in @xmath1 , for @xmath2 see @xcite . in the current article we present an algebraic counterpart of the asymptotic asphericity phenomenon
mentioned above . recall that a theorem of bieri ( * ? ? ?
* theorem b ) says that a finitely presented normal subgroup of a group of cohomological dimension at most two is either free or of finite index .
surprisingly , such strong restrictions on normal subgroups can hold also in the case of objects of arbitrarily high dimension , as the two following main results show .
a group is _ simphatic _ if it acts geometrically on a simply connected simplicially hereditarily aspherical ( simphatic ) complex ( see subsection [ s : simphatic ] for the precise definition ) .
[ t:0 ] a finitely presented normal subgroup of a simphatic group is either of finite index or virtually free . the following corollary is an immediate application of theorem [ t:0 ] to specific classes of simphatic groups .
[ c:0 ] finitely presented normal subgroups of groups acting geometrically on either : 1 .
systolic complexes , or 2 .
weakly systolic complexes with @xmath3 links , or 3 . simply connected nonpositively curved @xmath0complexes , or 4 .
simply connected graphical small cancellation complexes are of finite index or virtually free .
due to our best knowledge , even in the case of small cancellation groups the result is new .
goes beyond the case of ( graphical ) small cancellation groups as understood usually .
] however , the main point is that it holds also for higher dimensional groups .
the case ( a ) establishes a conjecture by dani wise ( * ? ? ?
* conjecture 11.8 ) ( originally concerning only torsion - free groups ) and answers some questions from ( * ? ? ?
* question 8.9(2 ) ) .
the case ( b ) deals with a subclass of weakly systolic groups studied in @xcite , containing systolic groups and exhibiting the same asphericity phenomena . in particular , the corollary extends some results from @xcite . for other classes of simphatic groups to which theorem [ t:0 ] applies see e.g. ( * ? ? ?
* examples 4.4 ) .
theorem [ t:0 ] is a consequence of the following more general result .
( @xmath4 denotes the @xmath5th homotopy pro - group at infinity see subsection [ s : prohom ] . )
[ t:2 ] let @xmath6 , and let @xmath7 be an exact sequence of infinite groups .
assume that @xmath8 and @xmath9 have type @xmath10 , @xmath11 and @xmath12
. then @xmath13 and @xmath14 .
this theorem , although quite technical in nature , is of its own interest and allows to conclude many other results around normal subgroups of groups with some asphericity properties .
we present below only three consequences of theorem [ t:2 ] .
( the notion of asymptotic hereditary asphericity is discussed in subsection [ s : aha ] . ) [ c:1 ] let @xmath8 be a finitely presented normal subgroup of an asymptotically hereditarily aspherical ( aha ) group @xmath9 of finite virtual cohomological dimension . then @xmath8 is either of finite index or virtually free .
theorem [ t:2 ] implies also similar restrictions on quotients .
[ c:1a ] let @xmath7 be an exact sequence of infinite groups with finitely presented kernel @xmath8 .
assume that @xmath9 is a hyperbolic aha group of finite virtual cohomological dimension .
then @xmath8 is virtually non - abelian free , and @xmath15 is virtually free . questions concerning aha group extensions appear e.g. in ( * ? ? ?
* question 8.9(2 ) ) and some partial results are presented in @xcite .
note that all the new examples of highly dimensional hyperbolic aha groups constructed in @xcite are of finite virtual cohomological dimension as in corollary [ c:1a ] .
it is not known in general whether hyperbolic systolic groups have finite virtual cohomological dimension .
nevertheless , the following result applies to all of them , and beyond .
[ c:3 ] let @xmath7 be an exact sequence of infinite groups with finitely presented kernel @xmath8 .
assume that @xmath9 is a simphatic hyperbolic group .
then @xmath8 is virtually non - abelian free , and @xmath15 is virtually free . finally , let us mention that a corollary of theorem [ t:2 ] is crucial for a construction of some optimal classifying spaces for systolic groups , which will be presented in a forthcoming paper .
the proof of theorem [ t:2 ] relies on the analysis of the homotopy type at infinity ( proper homotopy type ) of the group @xmath9 ( denoted @xmath16 in figure [ f : idea ] ) , and we proceed by a contradiction .
we restrict here considerations to the case @xmath17 this is enough for proving theorem [ t:0 ] .
suppose that @xmath18 . since @xmath19 are infinite , @xmath9 has one end , and furthermore @xmath19 are not one - ended , by @xmath11 ( see section [ s : pfb ] ) .
therefore , by theorems of stallings and of dunwoody the @xmath20homotopy type at infinity of @xmath19 ( @xmath21 and @xmath22 in figure [ f : idea ] ) is the type of a disjoint union of some points and some non - simply connected pieces ( see subsections [ s:1red ] & [ s : kpi1s ] ) . in the case
@xmath17 : the homotopy type at infinity @xmath16 of @xmath9 .
( some notations come from section [ s : pfb ] . ) ] furthermore , the homotopy type at infinity of @xmath9 is the join of the types of @xmath8 and @xmath15 ( see subsections [ s : yzmn ] & [ s : finale ] ) .
however , such join contains nontrivial @xmath0spheres appearing in suspensions of non - simply connected factors coming from @xmath8 ( the join @xmath23 in figure [ f : idea ] ) .
therefore , @xmath24 contradiction . for theorem [ t:0 ] , one observes that @xmath25 only when there are no one - ended factors in the splitting of @xmath8 .
this relies on the fact that finitely presented one - ended subgroups of a simphatic @xmath9 are not simply connected at infinity , since they are again simphatic .
in the next section [ s : prel ] we present some preliminary notions and results concerning e.g. : weakly systolic complexes and groups ( subsection [ s : sys ] ) , proper homotopy type ( subsection [ s : prohom ] ) , asymptotic hereditary asphericity ( subsection [ s : aha ] ) , and simphatic groups ( subsection [ s : simphatic ] ) . in section [ s : pfb ]
we prove the main theorem [ t:2 ] . in section
[ s : pfa ] we present the proofs of theorem [ t:0 ] and corollary [ c:0 ] . in section [ s : fin ] we prove corollaries [ c:1 ] , [ c:1a ] & [ c:3 ] .
few remarks are given in section [ s : finrem ] .
in appendix we present the topological @xmath0dimensional quasi - helly property of systolic complexes
throughout this article we work usually ( if not stated otherwise ) with regular cw complexes , called further simply _ complexes _ ( see e.g. @xcite for some notation ) .
accordingly , all _ maps _ are cellular maps .
we call a complex _ @xmath9complex _ if @xmath9 acts on it by cellular automorphisms .
the action is _ proper _ , or the complex is a _ proper @xmath9complex _ if stabilizers of cells are finite .
an @xmath5skeleton of a complex @xmath26 is denoted by @xmath27 .
a @xmath9complex @xmath26 has the @xmath5skeleton _ finite mod @xmath9 _ if the quotient @xmath28 is a finite complex .
recall that a group has _
type @xmath29 _ if there is a _ free _ ( that is , with trivial stabilizers of cells ) contractible @xmath9complex with the @xmath5skeleton finite mod @xmath9
. a group @xmath9 acts _ geometrically _ on @xmath26 , if @xmath26 is a finite dimensional proper contractible @xmath9complex with every skeleton finite mod @xmath9 . in particular , @xmath26 is then uniformly locally finite .
we usually work with the path metric on @xmath30 defined by lengths of paths in @xmath31 connecting two given vertices .
it follows that if a finitely generated @xmath9 acts geometrically on @xmath26 , then @xmath30 is quasi - isometric to @xmath9 equipped with a word metric coming from a finite generating set . for a subcomplex @xmath32 of a complex @xmath26 , by @xmath33
we denote the smallest subcomplex of @xmath26 containing @xmath34 ( set - theoretic difference ) . in this subsection
we restrict our studies to simplicial complexes .
we follow mostly notation from e.g. @xcite .
a simplicial complex @xmath26 is _ flag _ if it is determined by its @xmath20skeleton @xmath31 , that is , if every set of vertices pairwise connected by edges spans a simplex in @xmath26 .
a subcomplex @xmath32 of a simplicial complex @xmath26 is _ full _ if it is determined by its vertices , that is , if any set of vertices in @xmath32 contained in a simplex of @xmath26 , span a simplex in @xmath32 .
the _ link _ of a simplex @xmath35 of @xmath26 is the simplicial complex being the union of all simplices @xmath36 of @xmath26 disjoint from @xmath35 , but spanning together with @xmath35 a simplex of @xmath26 .
a @xmath37cycle is a triangulation of the circle @xmath38 consisting of @xmath37 edges and @xmath37 vertices . for @xmath39 , a flag simplicial complex
@xmath26 is _
( respectively , _
locally @xmath37large _ ) if there are no full @xmath40cycles ( as subcomplexes ) in @xmath26 ( respectively , in any link of @xmath26 ) , for @xmath41 .
a complex @xmath26 is _ @xmath37systolic _
( respectively , _ systolic _ ) if it is simply connected and locally @xmath37large ( respectively , locally @xmath42large )
see @xcite .
@xmath37wheel _ is a graph consisting of a @xmath37cycle and a vertex adjacent to all vertices of the cycle .
@xmath37wheel with a pendant triangle _ is a graph consitsting of a @xmath37wheel and a vertex adjacent to two vertices of the @xmath37cycle .
a complex @xmath26 satisfies the _
@xmath3 property _ if it is locally @xmath43large and every @xmath43wheel with a pendant triangle is contained in a @xmath20ball around some vertex of @xmath26 .
a flag simplicial complex is _ weakly systolic _ if it is simply connected and satisfies the @xmath3 property
see @xcite .
a weakly systolic complex in which all links satisfy the @xmath3 property is called a _
weakly systolic complex with @xmath3 links_. a group is _ systolic _
( respectively , _
@xmath37systolic _ , _ weakly systolic _ ) if it acts geometrically on a systolic ( respectively , @xmath37systolic , weakly systolic ) complex .
the basics about proper homotopy theory can be found in @xcite .
recall that a cellular map @xmath44 between complexes is _ proper _ if @xmath45 is a finite subcomplex for any finite subcomplex @xmath46 and any @xmath5 ( see e.g. ( * ? ? ?
* chapter 10 ) , where the name _ cw - proper _ is used ) .
two maps @xmath47 are _ properly homotopic _ , denoted @xmath48 , if there exists a proper map @xmath49 \to y$ ] with @xmath50 and @xmath51 , for all @xmath52 .
a proper map @xmath53 is a _ proper homotopy equivalence _
( respectively , _ proper @xmath5equivalence _ ) if there exists a _ proper homotopy inverse _
( respectively , _ proper @xmath5inverse _ ) @xmath54 with @xmath55 , and @xmath56 ( respectively , @xmath57 , and @xmath58 ) .
in such case we say that @xmath26 and @xmath32 have the same _ proper homotopy type _
( respectively , _ proper @xmath5type _ ) . for @xmath59
, we say that the _
@xmath5homotopy pro - group at infinity vanishes _ for @xmath26 , denoted @xmath60 , if the following holds . for every finite subcomplex @xmath61 there exists a subcomplex @xmath62 containing @xmath63 such that : for any map @xmath64 of the @xmath5sphere @xmath65 , there exists its extension @xmath66 .
if @xmath60 does not hold , then we write @xmath67 .
the condition @xmath68 means that @xmath26 is _ one - ended _ , and @xmath69 means _ simple connectedness at infinity_. the next proposition shows that for a group acting in a nice way on a complex , vanishing of homotopy pro - groups at infinity does not depend on the complex .
it is a version of ( * ? ? ?
* proposition 2.10 ) whose proof is analogous to the one in @xcite .
( we may omit the requirement of ` rigidity ' here , since an action induced on the subdivision of a regular cw complex is rigid . )
compare also ( * ? ? ?
* theorems 17.2.3 & 17.2.4 ) and ( * ? ? ?
* theorem 7.2 ) , and proofs there .
[ p : conninf ] let @xmath70 be two @xmath5connected proper @xmath9complexes with @xmath71skeleta finite mod @xmath9
. then @xmath60 iff @xmath72 .
therefore , we say that @xmath12 ( respectively , @xmath73 ) if @xmath60 ( respectively , @xmath67 ) , for some ( and hence , any ) @xmath9complex as in proposition [ p : conninf ] . similarly ( cf .
e.g. ( * ? ? ?
* theorem 7.2 ) ) , vanishing of the homotopy pro - groups at infinity is an invariant of the coarse equivalence , and hence of quasi - isometry .
however , directly from the definition above it follows that this is also an invariant of the proper homotopy type .
more precisely , if there is a proper @xmath71equivalence @xmath53 , then @xmath60 iff @xmath72 cf .
* chapter 17 ) .
this invariance is crucial in our approach see subsection [ s : kpi1s ] .
the notion of asymptotic hereditary asphericity ( aha ) was introduced in @xcite .
we follow here mostly notations from @xcite . given an integer @xmath74 , for subsets @xmath75 of a metric space @xmath76 we say that @xmath77 _ is @xmath78aspherical in @xmath79 _ if every simplicial map @xmath80 ( where @xmath81 denotes the rips complex with constant @xmath82 ) , where @xmath83 is a triangulation of the @xmath5sphere @xmath65 , has a simplicial extension @xmath84 , for some triangulation @xmath85 of the @xmath71ball @xmath86 such that @xmath87 . a metric space
@xmath26 is _ asymptotically hereditarily aspherical _
, shortly _ aha _ , if for every @xmath88 there exists @xmath89 such that every subset @xmath90 is @xmath78aspherical in itself , for every @xmath6 .
a finitely generated group is aha if it is aha as a metric space for some ( and hence any ) word metric coming from a finite generating set .
a subgroup of an aha group is aha ( * ? ? ?
* corollary 3.4 ) .
a finitely presented aha group @xmath9 has type @xmath91 ( * ? ? ?
* theorem c ) , and @xmath12 for all @xmath6 ( * ? ? ?
* theorem d ) .
if moreover , the virtual cohomological dimension of @xmath9 is finite , or if @xmath9 acts geometrically on a contractible complex of finite dimension , then @xmath11 ( * ? ? ?
* theorem 7.6 ) . for examples of aha groups
see the next subsection . in this subsection
we introduce the notions of simphatic complexes and simphatic groups .
a flag simplicial complex is _ simplicially hereditarily aspherical _ , shortly _
simphatic _ , if every its full subcomplex is aspherical .
a group acting geometrically on a simply connected simphatic complex is called _
let us note few simple facts about simphatic groups : 1 .
a simply connected simphatic complex is contractible , by whitehead s theorem ( see e.g. ( * ? ? ?
* theorem 4.5 ) or ( * ? ? ?
* proposition 4.1.4 ) ) .
simphatic groups are aha , by ( * ? ? ?
* corollary 3.2 ) .
weakly systolic complexes with @xmath3 links are simphatic by ( * ? ? ?
* proposition 8.2 ) .
in particular , systolic complexes are simphatic , and systolic groups are simphatic ( see @xcite ) .
4 . there are simphatic groups that are not weakly systolic e.g. some baumslag - solitar groups . for other examples of simphatic groups
see e.g. @xcite .
full subcomplexes and covers of simphatic complexes are again simphatic .
therefore , by a theorem of hanlon - martnez - pedroza ( * ? ? ?
* theorem 1.1 ) , finitely presented subgroups of simphatic groups are again simphatic .
we can not resist to formulate here the famous whitehead s asphericity question in the following way : are all aspherical @xmath0complexes simphatic ?
in this section we present the proof of theorem [ t:2 ] from introduction . recall that @xmath7 is an exact sequence of infinite groups .
since @xmath92 have type @xmath10 , the image @xmath15 has type @xmath10 as well see e.g. ( * ? ? ?
* theorem 7.2.21 ) .
since @xmath19 are infinite , we have that @xmath9 has one end , that is , @xmath93 see e.g. ( * ? ? ?
* corollary 16.8.5 ) . by the result of houghton @xcite and jackson @xcite
( see also ( * ? ? ?
* corollary 16.8.5 ) ) , from the fact that @xmath11 we conclude that neither @xmath8 nor @xmath15 has one end . for the rest of the proof
we suppose that @xmath94 and we will show that this leads to a contradiction .
the other case under the assumption @xmath95 is essentially the same .
this will conclude the proof .
since the kernel @xmath8 has more then one end , by the result of stallings @xcite , it decomposes into a free product with amalgamation over a finite group , or as an hnn extension over a finite group .
if the factors are not one - ended we may repeat decomposing further . a theorem of dunwoody @xcite
says that finitely presented groups are accessible , that is , such a procedure ends after finitely many steps . as a result , by the classical bass - serre theory
( see e.g. ( * ? ? ?
* chapter 6.2 ) ) , the group @xmath8 is the fundamental group of a finite graph of groups @xmath96 with vertex groups having at most one end , and finite edge groups . let @xmath97 be the collection of the vertex groups , and let @xmath98 be the set of edge groups of @xmath96 .
similarly @xmath15 is the fundamental group of a graph of groups @xmath99 , with vertex groups @xmath100 .
since @xmath94 , at least one group from @xmath97 , say @xmath101 , is one - ended and with @xmath102 . since @xmath19 have type @xmath10 , all the factors @xmath103 have also type @xmath10 . for
all @xmath40 , let @xmath104 ( respectively , @xmath105 ) be a contractible free @xmath106complex ( respectively , @xmath107complex ) with the @xmath71skeleton finite mod @xmath106 ( respectively , @xmath107 ) . by a general technique
( see e.g. ( * ? ? ?
* chapter 6.2 ) , and ( * ? ? ?
* chapter 1.b ) ) one may compose all @xmath104 s to obtain a tree of complexes @xmath32 being a contractible proper @xmath8complex with the @xmath71skeleton finite mod @xmath8 .
moreover , subject to modifying slightly each @xmath104 ( by e.g. coning - off @xmath108invariant subcomplexes ) , the complex @xmath32 may be arranged in such a way that the following condition is satisfied : @xmath32 is a union of copies of @xmath104 s , and any two copies have at most one vertex in common .
similarly , there is a contractible proper @xmath15complex with the @xmath71skeleton finite mod @xmath15 , which is a union of copies of @xmath105 s and any two copies have at most one vertex in common .
since @xmath32 and @xmath109 have @xmath71skeleta finite mod @xmath8 and @xmath15 , and since @xmath9 has type @xmath10 , there is a contractible proper @xmath9complex @xmath26 with the following properties ( see e.g. ( * ? ? ?
* proposition 17.3.4 ) ) : the @xmath71skeleton of @xmath26 is finite mod @xmath9 , and there is a proper @xmath71equivalence between @xmath26 and @xmath110 .
therefore , to obtain a contradiction under our assumptions , it is enough to show that @xmath111 : by the proper @xmath71equivalence , it implies @xmath67 and hence , by proposition [ p : conninf ] , @xmath73 contradiction . in the next subsection we study the topology at infinity of @xmath110 .
our goal is to show that @xmath111 . to do this
we need to find subcomplexes @xmath112 and @xmath113 with finite @xmath71skeleta , and having the following property : for any @xmath114 with finite @xmath71skeleton and containing @xmath115 , there is a map @xmath116 that is not homotopically trivial within @xmath117 . in the current subsection
we define @xmath63 and @xmath118 , and we study the homotopy type of @xmath117 .
pick a copy of @xmath119 in @xmath32 , and a copy of @xmath120 in @xmath109 .
we further denote them by @xmath119 and @xmath120 , respectively
this should not lead to confusion . since @xmath102 , by proposition [ p : conninf ] , @xmath121 . therefore , there exists a subcomplex @xmath122 with finite @xmath71skeleton and such that the following holds : for every @xmath123 with finite @xmath71skeleton and containing @xmath63 , there exists a sphere @xmath124 not homotopically trivial within @xmath125 .
adding path - connected components of @xmath125 with finite @xmath71skeleta to @xmath63 we may further assume that @xmath125 is path - connected . similarly , we choose a subcomplex @xmath126 with finite @xmath71skeleton and such that @xmath127 is path - connected .
we claim that @xmath128 are as required above , and in the rest of this subsection we study the homotopy type of @xmath129 .
since the @xmath71skeleton of @xmath32 ( respectively , @xmath109 ) is locally finite , and by our construction of @xmath32 ( respectively , @xmath109 ) , the set @xmath130 ( respectively , @xmath131 ) is a discrete set of vertices .
since the @xmath71skeleton of @xmath63 ( respectively , @xmath118 ) is finite , the set @xmath132 ( respectively , @xmath133 ) is an empty or finite set @xmath134 ( respectively , @xmath135 ) . by possibly enlarging @xmath118 within @xmath120
we may assume that @xmath136 , that is , @xmath137 . for @xmath138 ( respectively , @xmath139 ) let @xmath140 ( respectively , @xmath141 ) denote the closure in @xmath32 ( respectively , in @xmath109 ) of the connected component of @xmath142 ( respectively , @xmath143 ) , not containing @xmath119 ( respectively , @xmath120 ) .
observe that such closure is the union of the component and @xmath144 ( respectively , @xmath145 ) see figure [ f : eg0 ] ( left ) . to @xmath146.,scaledwidth=100.0% ] since @xmath140 ( respectively , @xmath141 ) is contractible , for @xmath147 ( respectively , @xmath148 ) we may contract @xmath140 to @xmath149 ( respectively , @xmath141 to @xmath150 ) not changing the homotopy type of @xmath117 .
therefore , let @xmath151 ( respectively , @xmath152 ) denote @xmath32 ( respectively , @xmath109 ) with @xmath140 ( respectively , @xmath141 ) , contracted to @xmath149 ( respectively @xmath150 ) , for all @xmath147 ( respectively , @xmath148 ) see figure [ f : eg0 ] ( right ) .
we have : @xmath153 and @xmath154 furthermore , by formula ( [ e:10 ] ) , and since @xmath122 ( respectively , @xmath126 ) and @xmath155 ( respectively , @xmath156 ) , we have @xmath157 where @xmath158 and @xmath159 intersect along @xmath160 , etc . again , we may contract subcomplexes of the type @xmath161 to @xmath162 , @xmath163 to @xmath164 , and @xmath165 to @xmath166 , obtaining the following homotopy equivalence see figure [
f : eg1 ] . @xmath167 where ( see figure [ f : eg1 ] ) : @xmath168=\\ = ( y_1 - m)\times ( \{b_1\}\sqcup \ldots \sqcup \{b_s\}),\\ ( y_1\times
z_1 - m\times n ) \cap [ ( \{a_1\}\sqcup \ldots \sqcup \{a_m\})\times z_1]=\\ = ( \{a_1\}\sqcup \ldots \sqcup \{a_m\})\times ( z_1 - n),\\ { \mathrm}{and}\ : ( \{a_i\ } \times z_1 ) \cap ( y_1 \times \ { b_j \ } ) = \ { a_i \ } \times \{b_j \}. \end{split}\end{aligned}\ ] ] .
( only @xmath169 and @xmath170 showed among @xmath171 s and @xmath172s.),scaledwidth=100.0% ] since @xmath173 are contractible , the subcomplexes @xmath174 and @xmath175 can be contracted to points within @xmath176 .
( this may by done by sliding " @xmath174 to @xmath177 , for some @xmath178 first , and then contracting @xmath177 to a point within @xmath179 . ) therefore , by ( [ e:13 ] ) and ( [ e:14 ] ) we obtain the following homotopy equivalence
see figure [ f : eg2 ] : @xmath180 where @xmath181 ( respectively , @xmath182 ) is the complex obtained from @xmath119 ( respectively , @xmath120 ) , by contracting the subcomplex @xmath125 ( respectively , @xmath183 ) to a point .
the number of copies of @xmath181 in ( [ e:15 ] ) is @xmath184 , and the number of copies of @xmath182 is @xmath185 . for all @xmath186 and @xmath187 , the point @xmath171 in the @xmath188th copy of @xmath181 ( that is , in the complex obtained after modifying @xmath189 )
is identified with the point @xmath172 in the @xmath40th copy of @xmath182 see figure [ f : eg2 ] .
we denote by @xmath190 and @xmath191 the copies of @xmath181 corresponding to , respectively , @xmath170 and @xmath192 . .
( only @xmath169 and @xmath170 showed among @xmath171 s and @xmath172s.),scaledwidth=100.0% ] therefore , the space @xmath146 and thus also @xmath193 have the homotopy type @xmath194 of the wedge of @xmath190 , @xmath191 and some union @xmath195 of @xmath176 , and ( possibly ) other copies of @xmath181 and @xmath182 . let a subcomplex @xmath114 with finite @xmath71skeleton , and containing @xmath115 be given .
we will see that there exists an @xmath5sphere @xmath196 that is not homotopically trivial within @xmath117 .
let @xmath197 and @xmath198 be subcomplexes with finite @xmath71skeleta such that @xmath199 . fix a vertex @xmath200 .
by our choice of @xmath63 , there exists a sphere @xmath201 that is not contractible within @xmath125 . using @xmath202
, we will construct a sphere @xmath203 .
then we will show that this sphere is homotopically nontrivial within @xmath117 . for @xmath204
, choose a path @xmath205 connecting @xmath206 with a vertex @xmath207 .
let @xmath208 \to ( y_1 - m')\times z$ ] be a map constructed by use of @xmath205 , that is , @xmath209 . combining @xmath210 and the contractions of the spheres @xmath211 within @xmath212 for @xmath204
, we obtain the sphere @xmath213 .
the equator of this sphere is mapped by @xmath202 to @xmath214 . by the homotopy equivalences from ( [ e:11 ] ) , ( [ e:12 ] ) , ( [ e:13 ] ) , ( [ e:14 ] ) , ( [ e:15 ] ) , and ( [ e:15a ] ) we have that @xmath215 ( seen as a map to @xmath193 )
is homotopically equivalent to a map @xmath216 mapping one hemisphere to @xmath190 , the other to @xmath191 , and the equator to the wedge point .
since @xmath119 is contractible , the homotopy pair exact sequence for @xmath217 gives us the following isomorphism @xmath218 therefore , the nontrivial class @xmath219 \in \pi_{n-1}(y_1 -m)$ ] corresponds to the class of @xmath220 in @xmath221 , for @xmath204 .
since @xmath119 is contractible , and @xmath125 is path - connected we have ( see e.g. ( * ? ? ?
* proposition 4.28 ) ) that the homomorphism @xmath222 induced by the quotient map is an isomorphism .
therefore , the map @xmath223 from ( [ e:16a ] ) is homotopically non - trivial .
since the inclusion @xmath224 induces an injection on @xmath225 , we have that @xmath226 is homotopically nontrivial .
because @xmath77 was chosen arbitrarily , it follows that @xmath227 is nontrivial .
since @xmath110 is properly @xmath71equivalent to @xmath26 , we have that @xmath67 and thus , by proposition [ p : conninf ] , @xmath73 contradiction .
this concludes the proof of theorem [ t:2].@xmath228
in this section we prove theorem [ t:0 ] and corollary [ c:0 ] from introduction .
_ proof of theorem [ t:0 ] : _ assume that @xmath8 is not of finite index .
then we have an exact sequence @xmath7 of infinite groups . a simply connected simphatic complex @xmath26 on which @xmath9 acts geometrically is aha by ( * ? ? ?
* corollary 3.2 ) .
thus @xmath229 , by ( * ? ? ?
* theorem 7.4 ) . since @xmath26 is also finite dimensional and contractible ( by whitehead s theorem ) , by ( * ? ? ?
* theorem 7.6 ) , we have that @xmath9 has type @xmath91 and @xmath11 . as a subgroup of an aha group @xmath8 is itself aha , and thus of type @xmath91 as well , by ( * ? ? ?
* theorem 6.1 ) .
therefore , the assumptions of theorem [ t:2 ] ( for the case @xmath17 ) are fulfilled . as in the proof of theorem [ t:2 ] ( see subsection [ s:1red ] ) we have that @xmath8 is the fundamental group of a graph of groups @xmath96 with vertex groups @xmath97 having at most one end each , and with finite edge groups .
we have to show that none of @xmath106 has one end then @xmath8 is the fundamental group of a graph of finite groups , that is , @xmath8 is virtually free .
note that a full subcomplex or a covering of a simphatic complex is again simphatic .
theorefore , by a theorem of hanlon - martnez - pedroza ( * ? ? ?
* theorem 1.1 ) , each group @xmath106 acts geometrically on a simply connected simphatic complex .
if @xmath106 is one - ended then , again by ( * ? ? ?
* theorem 7.6 ) , we have @xmath230 and thus @xmath18 contradiction . hence , all @xmath106 are finite , and the proof is completed.@xmath228 _ proof of corollary [ c:0 ] : _ it is clear that a subcomplex of a nonpositively curved ( that is , locally cat(0 ) ) @xmath0complex is itself nonpositively curved and two - dimensional , and hence aspherical .
thus some its simplicial subdivision is simphatic .
the simphatic property for weakly systolic complexes with @xmath3 links follows from ( * ? ? ?
* proposition 8.2 ) . in particular , systolic complexes are simphatic ( see also @xcite ) .
small cancellation graphical small cancellation complexes are ( up to some simplicial subdivision ) obviously simphatic
see e.g. ( * ? ? ?
* example 4.4 ) ( or use a theorem of wise ( * ? ? ?
* sections 6 & 12 ) saying that groups acting geometrically on such complexes act geometrically on systolic complexes ) . therefore , corollary [ c:0 ] follows directly from theorem [ t:0 ] .
@xmath228 for other examples of simphatic complexes extending the list in corollary [ c:0 ] ( that is , to which theorem [ t:0 ] applies ) see e.g. ( * ? ? ?
* section 1.1 ) and ( * ? ? ?
* section 4 ) .
_ proof of corollary [ c:1 ] : _ assume that @xmath8 is not of finite index . as in the proof of theorem [ t:0 ] it is enough to show that none of the factors @xmath97 in the splitting of @xmath8 is one - ended .
a finitely generated subgroup of an aha group is aha itself ( * ? ? ?
* corollary 3.4 ) , thus @xmath8 and all @xmath106 s are aha .
furthermore , since every @xmath106 has finite virtual cohomological dimension ( see e.g. ( * ? ? ? * chapter viii.11 ) ) , by ( * ? ? ?
* theorem 7.6 ) , none of them is one - ended .
it implies that @xmath8 is virtually free .
@xmath228 _ proof of corollary [ c:1a ] : _ the kernel @xmath8 is virtually free by corollary [ c:1 ] .
note that virtually cyclic subgroups of hyperbolic groups are quasiconvex see ( * ? ? ?
* theorem 8.1.d ) and ( * ? ? ?
* proposition 3.2 ) .
furthermore , it is shown in @xcite that a normal quasiconvex subgroup of a hyperbolic group is either finite or of finite index .
therefore , a virtually cyclic normal subgroup of a hyperbolic group is either finite or of finite index .
it follows that @xmath8 is virtually non - abelian free . by the same argument as for @xmath8 ,
the quotient @xmath15 is the fundamental group of a graph of groups @xmath99 with finite edge groups , and vertex groups @xmath100 of at most one end ( see the proof of theorem [ t:2 ] , subsection [ s:1red ] ) .
again , it is enough to show that none of @xmath107 is one - ended .
suppose that this is not so say , @xmath231 is one - ended . by ( * ? ? ?
* theorem a ) , the quotient @xmath15 , and thus also @xmath231 are hyperbolic .
hence , @xmath231 acts geometrically on a finite dimensional contractible complex
some rips complex of @xmath231 . by (
* proposition 2.9 ) ( as in the proof of ( * ? ? ? * theorem 3.2 ) ) there is @xmath232 with @xmath233 .
therefore @xmath234 contradiction with theorem [ t:2 ] since @xmath235 . @xmath228 _ proof of corollary [ c:3 ] : _ the proof is nearly the same as the one of corollary [ c:1a ] .
the only point is that instead of the finiteness of the virtual cohomological dimension we use the fact that @xmath106 s , being simphatic , act geometrically on contractible finite dimensional complexes . @xmath228
in theorem [ t:0 ] and corollaries [ c:1 ] , [ c:1a ] & [ c:3 ] we use some asphericity conditions in order to apply theorem [ t:2 ] . however , such assumptions could be weakened by restricting to vanishing only some higher homotopy groups .
in particular , there are versions of the statements using the restricted " versions of aha or simphaticity .
one could use e.g.the conditions on filling radius constant for some spherical cycles , like the @xmath236 condition from ( * ? ? ?
* section 5 ) .
we do not elaborate those issues not seeing any immediate applications . for other related comments and questions
see e.g. @xcite .
in this appendix we present yet another asphericity - like property of systolic complexes . besides various asymptotic asphericity features used extensively through the article like e.g. aha or vanishing of higher homotopy pro - groups at infinity @xcite
the new property reflects the ` two - dimensional - like ' nature of systolic complexes , implying the normal subgroup theorem studied above .
the result theorem a below is a helly - type theorem , and we believe that it will be of wide use for studying systolic ( and , in general , simphatic ) complexes and groups .
recall the helly property of the euclidean plane : if , for a given family @xmath237 of convex subsets , all intersections @xmath238 are nonempty , then @xmath239 . in @xcite a quasification of such property
is proved for systolic complexes : it is shown there that the intersection of neighborhoods of corresponding convex subcomplexes has to be non - empty .
our theorem a below extends this result : we do not assume that the subcomplexes @xmath240 are convex we make assumptions on homotopy types of them and of their intersections ( thus we call our version topological " ) .
this is a significant difference : the product ( with the product metric ) of two complexes satisfying the convex " version has the same convex " helly property , while for our topological " version the corresponding helly dimension " may increase by taking products . note also that the topological @xmath0dimensional quasi - helly property presented below is closely related to the _ @xmath241thin tetrahedra property _ announced in ( * ? ? ?
* introduction ) ( there , only geodesic triangles are considered ) .
@xmath242 any simplicial map @xmath243 from a triangulation @xmath83 of the @xmath0sphere can be extended to a simplicial map @xmath244 from a triangulation @xmath79 of the @xmath245disc , such that the boundary of @xmath79 is @xmath83 and @xmath79 has no new ( internal ) vertices , that is @xmath246 .
let @xmath247 be simply connected subcomplexes of a systolic complex @xmath26 .
assume that for all @xmath248 , the intersections @xmath249 are connected , and @xmath250 .
then there exists a simplex @xmath251 such that @xmath252 .
first , we will construct a simplicial @xmath0sphere , @xmath254 . for all @xmath255 , pick a vertex @xmath256 , where @xmath257 .
for all @xmath258 , let @xmath259 , where @xmath257 , be a path connecting @xmath260 and @xmath261 . for @xmath257 , the cycle consisting of @xmath262 , @xmath263 , and @xmath264 is contained in @xmath240 .
therefore , there is a simplicial filling @xmath265 of this cycle inside @xmath240 , where @xmath266 is a triangulation of the @xmath0disc .
combining such fillings , we get a simplicial @xmath0sphere @xmath254 .
the sphere @xmath83 has a natural structure of the boundary of the tetrahedron : vertices are some pre - images @xmath267 of @xmath268 s ; edges are some pre - images @xmath269 of @xmath270 s ; triangles are the discs @xmath271 s . however ,
for our purposes we consider a dual " structure defined as follows ( see figure [ f : helly ] ) . for every @xmath40 , pick a vertex @xmath273 inside the interior of @xmath271 subdivide @xmath271 if necessary ( and redefine @xmath215 appropriately ) .
for every @xmath274 , find a simple path @xmath275 in @xmath83 connecting @xmath273 and @xmath276 and contained in @xmath277 .
subject to subdividing @xmath83 , we may assume that the intersection @xmath278 is either empty or one end - vertex .
this defines a new structure of the boundary of the tetrahedron : vertices are @xmath273 s ; edges are @xmath275 s ; triangles are connected components of the rest . by @xcite ( see ( * ? ? ?
* theorem 2.4 ) ) , there exists a simplicial extension @xmath279 of @xmath215 with the following properties : @xmath79 is a triangulation of the @xmath245disc , with @xmath83 being a triangulation of the boundary of @xmath79 , and without new ( that is , internal ) vertices .
above is satisfied . ]
now we define a _ coloring _ of vertices of @xmath79 , that is a map @xmath280 : observe that the coloring is well - defined ( though not unique ) , and that it is sperner s coloring of the triangulation of the tetrahedron with vertices @xmath289 .
therefore , by sperner s lemma @xcite , there exists a @xmath245simplex in @xmath79 with vertices @xmath290 such that @xmath291 , hence @xmath292 , for all @xmath293 .
since @xmath294 is simplicial , it follows that @xmath295 are contained in a simplex , and thus the proof is finished .
|
a group is simphatic if it acts geometrically on a simply connected simplicially hereditarily aspherical ( simphatic ) complex .
we show that finitely presented normal subgroups of simphatic groups are either : * finite , or * of finite index , or * virtually free .
this result applies , in particular , to normal subgroups of systolic groups .
we prove similar strong restrictions on group extensions for other classes of asymptotically aspherical groups .
the proof relies on studying homotopy types at infinity of groups in question .
in appendix we present the topological @xmath0dimensional quasi - helly property of systolic complexes .
|
humanity has , after several decades of effort , finally achieved the direct detection of gravitational waves @xcite .
given that we expect to detect gravitational wave ( gw ) events from astrophysical systems located at cosmological distances even the first two events , gw150914 and gw151226 , were produced at non - trivial redshifts @xmath2 it is desirable to examine from first principles how gws propagate over cosmological spacetimes , by solving einstein s equations perturbatively about the background spatially flat friedmann
lematre robertson walker ( flrw ) universe we reside in .
such a strategy is to be contrasted against the asymptotically flat spacetime calculations from which gw observables are usually extracted . to be very scrupulous when predicting gw events at cosmological distances , as already pointed out in @xcite , one has to match the ` far zone ' predictions of these asymptotically flat calculations onto a near / intermediate zone cosmological one , before evolving the wave solutions out to infinity
furthermore , one may also wonder if these gws journeying over a size - able fraction of our observable universe could experience the expansion of space itself , as well as interact with the intervening inhomogeneities thereby picking up additional features that could in turn provide us with novel probes of dark energy and dark matter . in this paper
, we will continue our effort initiated in @xcite to understand the causal structure of gravitational ( and electromagnetic ) signals in cosmological geometries of all spacetime dimensions greater or equal to four .
we will do so by reducing the problem to @xmath3 and @xmath4 dimensions , so that the portion of waves that travel strictly on the null - cone can be cleanly separated from the part that does so less than unit speed ( also known as the tail " ) . specifically
, we will work in a @xmath5dimensional background cosmology sourced by a perfect fluid with a constant equation - of - state ( eos ) @xmath0 , so that the resulting geometry is @xmath6 ^ 2 \eta_{\mu\nu } , \qquad\qquad \eta_{\mu\nu } = \text{diag}[1,-1,\dots,-1 ] , \end{aligned}\ ] ] with cartesian - like coordinates @xmath7 .
the scale factor describing the size of the universe is , in turn , @xmath8 \equiv \left ( \frac{\eta}{\eta_0 } \right)^ { \frac{2}{q_w } } , \qquad\qquad q_w \equiv ( d-3 ) + ( d-1)w .\end{aligned}\ ] ] because we wish to consider expanding universes , whenever @xmath9 ( i.e. , @xmath10 ) , @xmath11 and whenever @xmath12 ( i.e. , @xmath13 ) , @xmath14 the @xmath15 case will not be considered .
we will then place an isolated astrophysical system in such a geometry , and seek the metric perturbations it engenders , by solving the resulting linearized einstein s equations in terms of appropriate green s functions convolved against components of its energy - momentum - shear - stress . in other words
, we shall solve for the @xmath16 in the perturbed metric @xmath17 ^ 2 \left ( \eta_{\mu\nu } + \chi_{\mu\nu } \right ) \end{aligned}\ ] ] due to the presence of this astrophysical system .
since we are working only to first order in metric perturbations , it is physically fruitful to employ the well - developed gauge - invariant cosmological linear perturbation theory because once the solutions are non - trivial , they can not be rendered trivial simply by transforming to a nearby " coordinate system and therefore ought to play key roles in characterizing physical observables such as the strain measured by a gw detector . in @xcite , where we stumbled upon a tail - induced gw memory effect ( to be explained shortly ) in de sitter / dark energy and matter dominated universes , only the 4-dimensional ( 4d )
gauge - invariant tensor mode was analyzed in detail . in the follow - up work of @xcite
, we solved the linear metric perturbations in all de sitter backgrounds higher than @xmath18 dimensions , but did not employ gauge - invariant variables .
more recently , tolish and wald @xcite have extended the 4d gw memory analysis to the full set of gauge - invariant scalar , vector and tensor metric variables . in this paper , while we specialize to a constant@xmath19 cosmology from the outset , we shall tackle not only the gauge - invariant scalar - vector - tensor perturbations but also provide analytic solutions for the relevant green s functions in all dimensions higher than three .
one key gw observable is the fractional displacement between a pair of test masses that comprises part of a hypothetical experiment .
it is convenient to assume that these test masses are freely falling and co - moving in some given spacetime , so their spatial coordinates @xmath20 and @xmath21 are time independent .
as we demonstrate in appendix , if we denote the proper spatial distance between the masses at a fixed time as @xmath22 $ ] , and if we began our gw experiment at time @xmath23 , then the fractional displacement @xmath24 \equiv ( l[\eta]-l[\eta'])/l[\eta']$ ] is given by the following integral of the total change in the spatial metric expressed in synchronous gauge : @xmath25 = - \frac{\widehat{n}^i \widehat{n}^j}{2 } \int_{0}^{1 } \delta\chi_{ij } { \text{d}}\lambda + \mathcal{o}\left [ ( \chi_{\mu\nu})^2,(\eta-\eta ' ) \dot{a}[\eta']/a[\eta ' ] \right],\end{aligned}\ ] ] where here and throughout the rest of the paper an overdot always denotes a derivative with respect to the time coordinate being used ; @xmath26 is the unit radial vector pointing from one mass to the other ; and @xmath27 - \chi_{ij}\left [ \eta ' , \vec{y}_0 + \lambda \left(\vec{z}_0-\vec{y}_0\right ) \right ] .\end{aligned}\ ] ] ( the argument @xmath28 describes the euclidean straight line joining @xmath29 . ) suppose the primary wave train of some gw event has just passed our detector .
gw memory is the phenomenon where @xmath30 does not return to its original value at @xmath31 $ ] , but instead to a ` dc shift ' such that @xmath32 , a constant matrix , in eq . .
in such a scenario , eq . then informs us that the fractional distortion becomes permanent : @xmath33 by placing pairs of test masses with different orientations @xmath34 , and by comparing the pair - separations before and after the passage of the gw signal , we may thus detect whether such a permanent fractional distortion of space itself has occurred .
the gw memory effect is not the only reason motivating us to solve the perturbed cosmological system at hand . since @xcite
we have been motivated by the possibility that future - generation gw detectors such as lisa may hear gws emanating from the strong field regime of super - massive black holes residing within the central core of galaxies .
while the unperturbed kerr black hole ( bh ) does not radiate , compact objects that orbit and plunge into its horizon do stir up vibrations of spacetime that carry energy - momentum to infinity .
there is some hope that , when the compact object is much less massive than the bh itself , this extreme - mass - ratio - inspiral " ( emri ) system can be tackled analytically by expanding in powers of @xmath35 , where @xmath36 and @xmath37 are respectively the compact body and bh masses . in the strong gravity region near the black hole horizon , and at least within the lorenz gauge condition , compact bodies exert a tail - induced self - force that needs to be modeled properly for gw detectors to successfully track emri systems over @xmath38 $ ] orbital cycles @xcite .
additionally , gws from compact binary systems , modeled by the well - established weak field post - newtonian / minkowskian ( pn / pm ) framework , are also known to scatter off their own spacetime curvature and develop tails .
therefore , in this new era of gw astronomy , the tail phenomenon in 4d curved spacetimes is no longer a mere intellectual curiosity ; but understanding it properly could lead to observable consequences . to this end , we hope it will prove instructive to first understand thoroughly the gravitational and electromagnetic green s functions in the simpler context of cosmology .
it is worth pointing out that such an indirect detection of the tail effect through its impact on the compact bodies dynamics , which in turn would alter the resulting emri / pn / pm gw - forms is a test of our understanding of classical field theory in curved spacetime . for ,
since the pioneering work of hadamard @xcite , it is known that green s functions in 4d curved spacetimes , at least whenever the observer at @xmath39 and the spacetime - point - source at @xmath40 can be joined by a unique geodesic , take the generic form @xmath41 = u[x , x ' ] \delta_+[\sigma_{x , x ' } ] + v[x , x ' ] \theta_+[\sigma_{x , x ' } ] , \end{aligned}\ ] ] where the dirac @xmath42-function term describes the strictly - null signals and the ( step- ) @xmath43-function term the tail signals .
( here , @xmath44 is proportional to the square of the geodesic distance between @xmath39 and @xmath40 . )
the key insight is that : while understanding the tail function @xmath45 requires the full wave equation to be solved , the null - cone parts @xmath46 and @xmath47 only require the geodesics of the background spacetime to be known .
this lends itself to the physical interpretation that classical wave tails arise from the long wavelength modes , because it is these low frequency waves that are sensitive to the background spacetime curvature .
we begin , in
, with a review of and commentary on the extant literature on the gw memory effect in cosmology .
following that , in , we describe in detail the cosmological setup we are considering in this paper , including the background metric and fluid solutions upon which we are going to carry out perturbation theory .
the technical core of the paper lies in , where we will carry out first order gauge - invariant perturbation theory by first computing the equations - of - motion ( eom ) of the metric variables before presenting their analytic solutions in terms of convolution of appropriate green s functions against the astrophysical source .
potential gw memory effects , in both the scalar and tensor sectors , will be pointed out . in , we carry out an analogous gauge - invariant analysis for electromagnetic fields in cosmological spacetimes , and work out potential electric memory effects .
we summarize our findings and discuss future directions in . in appendix
, einstein s equations are linearized about flat @xmath5-dimensional spacetime and solved in both the gauge - invariant and de donder gauge formalism .
the latter leads us to the linard - wiechert metric perturbation due to a point mass in motion , and its pure tail nature in odd dimensions is highlighted . in
, a similar line of pursuit is carried out for electromagnetism and linearized yang - mills theory culminating in the comparison between the linear gravitational memory due to point masses scattering off each other on unbound trajectories and the linear yang - mills memory due to point color charges doing the same . for completeness
we also solve maxwell s equations in a generalized lorenz gauge in constant@xmath19 spatially flat cosmologies of all dimensions @xmath48 . in , we re - derive the fractional displacement formula in eq . , and attempt to carefully lay out the assumptions that lead up to it . in
, the solution to the primary partial differential equation encountered in this paper is reviewed , and a caution is issued regarding its global incompleteness whenever @xmath49 ( cf .
. finally ,
in , we explain from an algebraic perspective why @xmath50 can be viewed as a dimension - raising operator in flat spacetime .
in this section we shall attempt to provide , in chronological order , a brief review of and commentary on the extant literature on linear gw memory in a spatially flat flrw cosmology .
* chu ( 2015 ) * while gw memory around a flat geometric background has been studied since the works of zeldovich and polnarev , braginsky and grishchuk @xcite , braginsky and thorne @xcite , christodoulou @xcite , thorne @xcite , etc .
as far as we are aware
it was our recent work in @xcite that first examined it in the cosmological context .
in particular , we uncovered a novel contribution to linear memory arising from gws propagating inside the null cone .
this latter
tail " phenomenon finds no counterpart in linear massless fields propagating in asymptotically flat 4d spacetimes .
specifically , in 4d de sitter spacetime , because the green s function tail of the spin@xmath51 graviton ( i.e. , the transverse - traceless portion of the spatial metric ) is a spacetime constant , we pointed out that this lead to a tail - induced linear memory that does not decay with increasing distance nor increasing time from the source(s) world - tube in spacetime . moreover , our work in @xcite showed that the light cone portion of the spin@xmath51 gws takes on the following universal form regardless of the details of cosmic evolution ( cf .
( 28 ) of @xcite ) : @xmath52 = 4 { g_{\rm n}}\int_{\mathbb{r}^3 } { \text{d}}^3 \vec{x } ' a[\eta_r]^3 \frac{\pi_{ij}^{\text{(t)}}[\eta_r,\vec{x}']}{a[\eta ] |\vec{x}-\vec{x}'| } .\end{aligned}\ ] ] here , we have employed weinberg s notation @xcite , where @xmath53 denotes the transverse - traceless ( tt ) part of @xmath30 ; @xmath54 is the ( negative ) tt part of the gw source s shear - stress tensor written in a co - moving orthonormal frame ; and @xmath55 is the retarded time .
we went on to argue that the @xmath56 redshift in eq .
is the dominant effect cosmic expansion has on gravitons propagating in a spatially flat flrw universe ; because the gw tail amplitude is suppressed relative to its null - cone counterpart of eq . by the duration of the gw source to the age of the universe , as well as
the observer - source spatial distance to the size of the observable universe .
* bieri - garfinkle - yau ( 2015 ) * a few months after @xcite was posted on the arxiv , bieri , garfinkle and yau @xcite tried to employ the techniques bieri and garfinkle developed in @xcite to study the gw memory effect in de sitter spacetime .
the bulk of @xcite consisted of solving the linearized weyl tensor sourced by a null matter stress - energy tensor .
following that , they integrated a certain leading piece of the electric portion of the weyl tensor to obtain the memory effect . because the weyl tensor vanishes in any conformally flat spacetime such as de sitter , this means the weyl tensor is gauge - invariant at first order in perturbations .
bieri , garfinkle and yau @xcite had probably hoped that this would thus provide a gauge - invariant formalism to investigate linear gw memory in cosmology .
however , they did not justify why the minkowski background and zero cosmological constant formalism of @xcite could be carried over to cosmological backgrounds .
in particular , why the weyl tensor alone is sufficient to determine gw memory was not explained . the standard treatment of gw memory appeals to the geodesic deviation equation . along a timelike geodesic
whose tangent vector we shall denote as @xmath57 , the displacement vector @xmath58 joining it to an infinitesimally close - by geodesic obeys @xmath59 where @xmath60 is the riemann tensor.\beta } + \gamma^\alpha_{\phantom{\alpha}\sigma[\mu } \gamma^\sigma_{\phantom{\sigma}\nu]\beta}$ ] , ricci tensor is @xmath61 , ricci scalar is @xmath62 , and the christoffel symbols are @xmath63 .
einstein summation convention is in force ; latin / english alphabets run over the spatial indices @xmath64 through @xmath65 while greek ones run from @xmath66 ( time ) through @xmath65 .
symmetrization of indices are denoted as , for e.g. , @xmath67 ; and anti - symmetrization as @xmath68 } \equiv t_{\alpha\beta}-t_{\beta\alpha}$ ] . ] to arrive at eq .
one makes use of @xmath69 \epsilon^\alpha = r^\alpha_{\phantom{\alpha}\beta \mu\nu } \epsilon^\beta , \qquad\qquad u^\sigma \nabla_\sigma u^\mu = 0 \qquad\qquad \text { and } \qquad\qquad [ u,\xi ] = 0 .\end{aligned}\ ] ] integrating eq
. allows one to solve for the fractional change in the spatial geodesic distance between a given pair of timelike geodesics .
gw memory occurs when this fractional change becomes permanent after the passage of the primary gw train .
now , the riemann tensor @xmath70 itself can be decomposed into its trace - free ( i.e. , the weyl tensor @xmath71 ) plus trace - parts , namely @xmath72 \beta } - r_{\beta [ \mu } g
_ { \nu ] \alpha } \right ) - \frac{1}{(d-1)(d-2 ) } g_{\alpha [ \mu } g_{\nu ] \beta } \mathcal{r } .\end{aligned}\ ] ] in eq .
, if we employ einstein s equation with a non - zero cosmological constant @xmath73 namely , @xmath74 to re - express the ricci - tensors @xmath75 and ricci - scalar @xmath76 into terms involving only the cosmological constant @xmath73 and the energy - momentum - shear - stress tensor of matter @xmath77 , the riemann tensor itself will be translated into @xmath78\beta } \nonumber \\ & \qquad\qquad + \frac{8 \pi { g_{\rm n}}}{d-2 } \left ( t^{\alpha}_{\phantom{\alpha}[\mu } g_{\nu]\beta } - t_{\beta[\mu } \delta_{\nu]}^{\alpha } - \delta^\alpha_{[\mu } g_{\nu]\beta } \frac{2 t}{d-1 } \right ) , \qquad\qquad t \equiv g^{\rho\sigma } t_{\rho\sigma } .\end{aligned}\ ] ] by inserting the form of the riemann tensor in eq . into the geodesic deviation eq .
, we see that even after taking into account the dynamics of general relativity tidal forces are not encoded solely within the weyl tensor in generic geometries . in the 4d
de sitter spacetime bieri - garfinkle - yau @xcite was considering , while @xmath79 as long as the gw detectors are assumed to be far away from the gw source(s ) , the cosmological constant @xmath73 is crucial for its very existence and therefore can not be zero .
in generic spatially flat flrw cosmologies , say a matter or radiation dominated universe , the evolution of the scale factor is driven by a non - trivial background perfect fluid , so that even if one neglects @xmath73 , the @xmath77 terms in eq . can not be discarded .
it is possible that the cosmological constant @xmath73 and matter @xmath77 terms in eq .
are sub - dominant relative to the weyl tensor term in the cosmological scenarios of interest but this we advocate , ought to be justified more carefully before the results in @xcite,@xcite can be meaningfully applied / extended to cosmological gw memory .
* kehagias - riotto ( 2016 ) * earlier this year , kehagias and riotto @xcite extended the strominger and zhiboedov @xcite work in 4d asymptotically flat spacetimes , to draw a connection between the bondi - van der burg - metzner - sachs ( bms ) symmetries at null infinity in decelerating but expanding spatially flat flrw universes to gw memories in the same cosmologies .
they used the retarded green s functions derived by iliopoulos et al .
@xcite to first compute , in a matter dominated universe , the tt tensor mode produced by @xmath80 point masses that collide at a spacetime point before moving apart on separate trajectories ; finding ( like we did in @xcite ) a light cone part that , apart from the @xmath56 redshift in eq . , is very similar to its minkowski counterpart as well as a space - independent tail signal .
then , they went on to integrate a variant of the geodesic deviation equation to find the fractional distortion between two test masses after the passage of a non - trivial gw bondi news function " .
the latter could , in turn , be gotten via a bms transformation of the perturbed spatially flat flrw metric near null infinity . *
chu ( 2016 ) * in @xcite the theoretical observation in @xcite , that the gauge - invariant tensor gw - tail of 4d de sitter ( ds ) is a spacetime constant , was extended to all higher even dimensional ds@xmath81 .
building on the work by ashtekar , bonga , and kesavan @xcite as well as date and hoque @xcite , we expressed the tensor waveform in terms of the mass and pressure quadrupole moments of its isolated astrophysical source , allowing us to measure the magnitude of the tail - induced memory effect directly in terms of the latter s net change over the course of the gw generation process . on a technical level , we also confirmed that conformal re - scaling , dimension reduction , and nariai s ansatz allow the @xmath5-dimensional green s functions of the full set of linear de sitter graviton fields , i.e. , not just the @xmath53 , to be derived explicitly . the result in eq .
was reiterated in @xcite to argue that all known asymptotically 4d minkowskian gw
memory effects should carry over to the asymptotically spatially flat flrw case , with the additional feature that these on - the - null - cone spin@xmath51 gws will be further diluted as the universe expands , due to the scale factor in the @xmath82|\vec{x}-\vec{x}'|)$ ] fall - off .
furthermore , in a matter dominated spatially flat flrw spacetime , the spin@xmath51 graviton tail is constant in space but decays in time over cosmological timescales ; this leads to a corresponding linear gw memory that is approximately spacetime constant over timescales of human experiments . _
remark _ this is an appropriate place to highlight that the portion of the gw memory effect that propagates on the null - cone , really arises from a non - zero difference between the asymptotic matter configuration after and before the time period of active gw production . for concreteness ,
let @xmath83 denote the cosmic time when gw production peaked . referring to eq . , and assuming the gw detector is located far enough from the system such that @xmath84 with the spatial coordinate axes centered within the astrophysical system at @xmath83 we then see that its tt contribution to the gw memory formulas of equations and is @xmath85 \approx \frac{4 { g_{\rm n}}}{a[\eta ] |\vec{x}| } \delta \int_{\mathbb{r}^3 } { \text{d}}^3 \vec{x } ' a^3 \pi_{ij}^{\text{(t ) } } , \end{aligned}\ ] ] where the @xmath86 terms refer to the net change in the spatial - total of ( the tt part of ) the shear - stress of the gw source : @xmath87 ^ 3 \pi_{ij}^{\text{(t)}}[\eta_r,\vec{x } ' ] - \lim_{\eta_r \ll \eta_\star } \int_{\mathbb{r}^3 } { \text{d}}^3 \vec{x } ' a[\eta_\star]^3 \pi_{ij}^{\text{(t)}}[\eta_r,\vec{x } ' ] .\end{aligned}\ ] ] here and throughout the rest of the paper by @xmath88 we really mean some retarded time well after peak gw production , but not long enough for the scale factor to have changed appreciably , consistent with the timescales of human experiments ; likewise , @xmath89 means some retarded time well before the gw event itself , but not long enough that the scale factor itself can still be approximated as @xmath90 $ ] .
* tolish - wald ( 2016 ) * most recently , tolish and wald @xcite examined potential gw memory effects in a generic spatially flat flrw spacetime due to a setup similar to the one in kehagias and riotto @xcite .
specifically , they considered @xmath91 incoming geodesic point particles that meet at a single spacetime event , followed by @xmath92 outgoing geodesic point particles emerging from that same event .
( their @xmath91 is not necessarily equal @xmath92 . ) after examining the partial differential equations for the first order gauge - invariant scalar , vector and tensor metric variables , they proceeded to argue that only the ( tt ) tensor metric perturbation would give rise to a derivative of a dirac @xmath42-function in the riemann curvature tensor linearized about eq . .
this , they asserted , shows that only the tensor mode exhibits gw memory in their setup .
they also argued that gw tails do not yield memory effects in their setup .
this presence of @xmath93 in the linearized riemann tensor as criterion for gw memory was based on their experience in minkowski spacetime @xcite,@xcite brought over to the cosmological context without further quantitative justification .
while this @xmath93-criterion is understandably mathematically convenient , we feel it obscures the physics behind gw memory that propagates on the null / acoustic - cone : as emphasized above , it is due to the difference between the gw source configuration in the asymptotic future versus its configuration in the asymptotic past ; whereas , the detailed process of active gw production is relevant only through its impact on the asymptotic future trajectories of the astrophysical system . in any case , we will provide evidence in
below , that despite the tolish - wald claim
scalar gws do in fact yield
acoustic - cone " memory effects in the very setup they considered .
throughout this paper , we will consider an isolated astrophysical system moving about in a background spatially flat friedmann - lematre - robertson - walker ( flrw ) universe driven by a perfect fluid of constant equation - of - state @xmath0 .
we wish to examine the gw signatures produced by this astrophysical system as detected by a hypothetical experiment located at distances large compared to the system s characteristic size .
the total action of such a system is @xmath94 the gravitational dynamics is governed by the einstein - hilbert action @xmath95 with @xmath96 being newton s constant in @xmath97 spacetime dimensions , @xmath98 is the square root of the absolute value of the determinant of the spacetime metric @xmath99 , and @xmath76 is the ricci scalar of the same geometry .
derivatives and the action is dimensionless , @xmath96 itself must be of dimensions [ length]@xmath100 or equivalently [ 1/mass]@xmath100 .
we have , evidently , also set @xmath101 .
the levi - civita symbol @xmath102 is zero whenever there are repeated indices , but otherwise returns the sign of the permutation bringing @xmath103 to @xmath104 ; in particular , @xmath105 .
the covariant levi - civita tensor is defined as @xmath106 . in what follows
, the square " of spatial vectors always means its euclidean dot product with itself ; for e.g. , @xmath107 , where @xmath108 is the kronecker delta . + even though it is increasingly irresponsible to do so due to the mounting astrophysical / cosmological observations that point to its existence we have set the cosmological constant to zero in this paper . for , having too many scales in a given problem makes analytic solutions difficult to come by . ] the perfect fluid dynamics is described by a lagrangian density that coincides with ( negative of ) its energy density @xmath109 , @xmath110 , \qquad\qquad n^2 \equiv n^\mu n_\mu .
\end{aligned}\]]the prime denotes a derivative with respect to the argument ; while the entropy current @xmath111 is built out of @xmath112scalar fields @xmath113 that labels the elements of the fluid itself in its co - moving frame
. if @xmath114 is the covariant levi - civita tensor , @xmath115 by construction , @xmath116 are identities . that the lagrangian density in eq
. governs the dynamics of a perfect fluid can be seen by obtaining the expected form of its stress energy tensor , which we shall denote as @xmath117 . by varying @xmath118 with respect to the metric ,
i.e. , replace @xmath119 , @xmath117 is then the coefficient of @xmath120 : @xmath121 - \rho_{\text{f}}'\left [ n^2/2\right ] n^2 \right ) + \rho_{\text{f}}'\left[n^2/2\right ] n^\alpha n^\beta , \end{aligned}\ ] ] with the prime on @xmath109 denoting a derivative with respect to its argument . in the co - moving frame , and provided @xmath111 is timelike , @xmath122 and @xmath123 , \qquad\qquad \,^{\text{(f)}}t^{\widehat{0}\widehat{i } } = 0 , \qquad\qquad \,^{\text{(f)}}t^{\widehat{i}\widehat{j } } = \delta^{ij } \left ( \rho_{\text{f}}'\left [ n^2/2\right ] n^2 - \rho_{\text{f}}\left[n^2/2\right ] \right ) ; \end{aligned}\ ] ] allowing us to identify the fluid pressure density @xmath124 as @xmath125 n^2 - \rho_{\text{f}}\left[n^2/2\right ] .
\end{aligned}\ ] ] here , because pressure depends only on the energy density and none other thermodynamic variables , we are neglecting other possible conserved quantities that might exist in a more general fluid .
if the perfect fluid has a constant pressure to energy density ratio @xmath0 , we will shortly explain why the energy / mass density then takes the form @xmath126 = \frac{(d-2)(d-1)}{4\pi{g_{\rm n}}(q_w \eta_0)^2 } \left(n^2\right)^{\frac{w+1}{2 } } .
\end{aligned}\ ] ] the stress tensor of the fluid in eq . , in turn , simplifies to @xmath127 .
\end{aligned}\ ] ] when @xmath128 , notice from equations and that the perfect fluid stress - tensor takes the form of the cosmological constant term in einstein s equations .
specifically , @xmath129 , with @xmath130 .
this means our linearized solutions for @xmath128 below should really be viewed as the first - order gauge - invariant form of the linearized de sitter solutions we obtained in @xcite by fixing a gauge .
finally , we will be largely agnostic about the details of the astrophysical system , as encoded within @xmath131 .
* equations - of - motion ( eom ) * by varying the total action in eq . with respect to the metric , one obtains einstein s equations for the cosmological system : @xmath132 - 8\pi{g_{\rm n}}\,^{(\text{f})}t_{\mu\nu } = 8\pi{g_{\rm n}}\,^{(\text{a})}t_{\mu\nu } , \end{aligned}\ ] ] where @xmath133 is einstein s tensor , @xmath134 is the energy - momentum - shear - stress tensor of the fluid in eq .
and @xmath135 is that of the astrophysical system described by @xmath131 .
the fluid dynamics is gotten by varying its action in eq . with respect to each and every scalar @xmath136 .
this yields the @xmath112 equations , with @xmath137 : @xmath138
n_\mu \epsilon^{\mu \alpha_1 \alpha_2 \dots \alpha_{\text{i}-1 } \sigma \alpha_{\text{i}+1 } \dots \alpha_{d-1 } } \partial_{\alpha_1 } \varphi^1 \dots \partial_{\alpha_{\text{i}-1 } } \varphi^{\text{i}-1 } \partial_{\alpha_{\text{i}+1 } } \varphi^{\text{i}+1 } \dots \partial_{\alpha_{d-1 } } \varphi^{d-1 } \right ) = 0,\end{aligned}\ ] ] where the @xmath139 is the levi - civita symbol with no dependence on the metric and @xmath140 . in principle , the astrophysical system @xmath131 obeys its own equations - of - motion as well .
* background solutions : constant eos @xmath0 * since we are primarily interested in how gws generated by isolated astrophysical systems propagate over cosmological distances , we shall begin by ignoring the astrophysical contribution in eq . and solving the background " einstein - perfect fluid equations @xmath141 - 8\pi{g_{\rm n}}\,^{(\text{f})}t_{\mu\nu}[{\bar{g } } ] = 0 .\end{aligned}\ ] ] because we are seeking analytic solutions to gw systems , in this paper , we shall simplify the cosmological perturbation theory below by studying a perfect fluid with a constant equation - of - state ( eos ) @xmath0 .
. tells us @xmath142}{\rho_{\text{f}}\left[n^2/2\right ] } \frac{n^2}{2 } = \frac{p_\text{f}}{\rho_\text{f } } \equiv w \qquad\qquad \text{(constant ) } .\end{aligned}\ ] ] the general solution of this first order ordinary differential equation for @xmath109 in terms of @xmath143 is @xmath144
= \rho_0 \left(\frac{n^2}{2}\right)^ { \frac{w+1}{2 } } , \qquad\qquad ( \rho_0 \text { constant } ) .\end{aligned}\ ] ] on the other hand , eq . tells us the ratio of the diagonal space - space component of einstein s tensor to its @xmath145 component must also equal @xmath0 . if we insert the spatially flat , homogeneous and isotropic geometry of eq .
( but for a generic scale factor @xmath146 $ ] ) into the background einstein - fluid eq .
, @xmath147 the general solution is @xmath146 = ( ( \eta-\eta_\odot)/\eta_0)^{2/q_w}$ ] , where @xmath148 corresponds to the cosmic time where the universe has zero size if @xmath49 or infinite size if @xmath149 . for technical convenience , we will set it to zero , @xmath150 .
this recovers the cosmic evolution in eq . .
through a direct calculation , one may verify the background solution @xmath151 of the fluid eq . to be @xmath152 for , when eq .
holds , the entropy current eq . evaluated on equations is @xmath153 and @xmath154 evaluated on the same gives @xmath155 with equations , and , the quantity inside the derivative @xmath156 in eq . then becomes independent of spatial coordinates and proportional to @xmath157 because of the fully anti - symmetric nature of the levi - civita symbol .
is thus satisfied . plugging eq . into eq . then allows us to fix the @xmath158 there through the background einstein s equation ; namely , setting the einstein tensor built from equations and to be equal to @xmath159 times the stress tensor of eq . in eq
then hands us eq .
with the background metric in eq . and fluid solution in eq
. understood , we will now introduce the astrophysical system @xmath135 in eq . as a first order perturbation .
expand the spacetime metric about its background one in eq .
, as in eq . ; and expand the fluid scalar fields away from their equilibrium values in eq .
, @xmath160 in this section , we will proceed to tackle the linearized gravitational system @xmath161 with @xmath162 and @xmath163 representing the parts of einstein s tensor and stress energy of the perfect fluid that are linear in the metric @xmath164 and fluid perturbation @xmath165 variables . the fluid eq . will be similarly linearized about its background one evaluated on eq . .
as already advertised in the introduction , we shall employ the first order gauge - invariant formalism commonly used in the physically important case of 4d cosmology . for our purposes ,
gauge - invariance means the quantities we are computing do not alter their form under an infinitesimal change in the coordinate system .
this ensures that , whatever physical effect we associate with some metric / fluid variable once it is found to be non - zero in one coordinate system , it can not be made trivial merely by switching to a nearby " one . on a more technical level ,
let us make the key observations that , since the left hand side of einstein s equations in eq
. is zero at the background level because we are treating the astrophysical system as a first order perturbation that means its first order counterpart @xmath166 in eq .
must be gauge - invariant .
( this will be elaborated upon below . ) likewise , the left hand side of the fluid eom in eq .
is , too , zero at the background level @xmath167 and must also be gauge - invariant at linear order in @xmath168 and @xmath169 . by identifying appropriate metric @xmath170 and fluid @xmath171 perturbation variables that are gauge - invariant , we are guaranteed that the linearized einstein and fluid equations will be expressed solely in terms of them thus providing us with not only physical equations to solve from the outset , but also a check of the intermediate calculations . hence , before we embark on perturbation theory proper , let us take a detour to lay out the scalar - vector(-tensor ) decomposition that underlies the gauge - invariant formalism ; followed by delineating the gauge - invariant metric and fluid perturbation variables that we will be solving for
. we will spend and doing so , before tackling the first order eom and their gw solutions in sections and . in order to exploit the background so@xmath172 symmetry of the cosmological einstein - fluid system , the first step to obtaining a gauge - invariant formulation of the equations governing the metric perturbations @xmath168 and that of the fluid @xmath169 is to perform an irreducible scalar - vector - tensor decomposition of the former and a similar scalar - vector one of the latter .
* scalar - vector decomposition * we will first tackle the scalar - vector decomposition .
any @xmath112-tuple of spacetime functions @xmath173 \vert \text{i } = 1,2,3,\dots , d-1 \}$ ] can be decomposed as @xmath174 = \partial_{\text{i } } \mathcal{e } + \mathcal{e}_{\text{i } } , \end{aligned}\ ] ] where @xmath175 is divergence - less @xmath176 .
this scalar - vector decomposition of eq .
holds whenever the spatial fourier transform of @xmath177 , namely @xmath178 \equiv \int_{\mathbb{r}^{d-1 } } \frac{{\text{d}}^{d-1}\vec{k}}{(2\pi)^{d-1 } } \widetilde{\mathcal{f}}_\text{i}[\eta,\vec{k } ] e^{ik_{\text{i } } x^\text{i } } , \end{aligned}\ ] ] exists as @xmath179 .
] and is non - singular in the long - distance limit @xmath180 . for , in fourier space ,
we may write @xmath181 = i k_\text{i } \frac{k_\text{j}}{i \vec{k}^2 } \widetilde{\mathcal{f}}_\text{j}[\eta,\vec{k } ] + \left ( \delta_{\text{i}\text{j } } - \frac{k_\text{i } k_\text{j}}{\vec{k}^2 } \right ) \widetilde{\mathcal{f}}_\text{j}[\eta,\vec{k } ] , \qquad\qquad \vec{k}^2 \equiv \delta^{\text{i}\text{j } } k_\text{i } k_\text{j } , \qquad\qquad \vec{k } \neq \vec{0 } .\end{aligned}\ ] ] notice that partial spatial derivatives are replaced with momentum vectors in fourier space , i.e. , @xmath182 , so that the spatial divergence turns into a dot product , " namely @xmath183 .
these allow us to read off the gradient piece " of eq . in fourier space as the first term ( from the left ) of eq .
: @xmath184 = \frac{k_\text{j}}{i \vec{k}^2 } \widetilde{\mathcal{f}}_\text{j}[\eta,\vec{k } ] ; \end{aligned}\ ] ] whereas the divergence - less piece " of eq . in fourier space is the second term ( from the left ) of eq .
@xmath185 & = p_{\text{i}\text{j } }
\widetilde{\mathcal{f}}_\text{j}[\eta,\vec{k } ] , \\
\label{transverse_projector } p_{\text{i}\text{j } } & \equiv \delta_{\text{i}\text{j } } - \frac{k_\text{i } k_\text{j}}{\vec{k}^2 } , \end{aligned}\ ] ] because the symmetric projector , which obeys @xmath186 and @xmath187 , is transverse to the momentum vector @xmath188 : @xmath189 . in what follows , it is also very useful to remember that , any relation of the form @xmath190 + \mathcal{u}_\text{i}[\eta,\vec{x } ] = 0 , \end{aligned}\ ] ] with @xmath191 , implies that @xmath192 = \mathcal{u}_\text{i}[\eta,\vec{x } ] = 0 , \end{aligned}\ ] ] as long as @xmath193 $ ] and @xmath194 $ ] admit spatial fourier transforms that are well - behaved in the ir " limit , @xmath180 .
this follows from first taking the divergence of eq . in
fourier space : @xmath195 = 0 , \end{aligned}\ ] ] which yields , for @xmath196 , @xmath197=0 $ ] .
this in turn means eq . in fourier space
now reads @xmath198 = 0 $ ] .
* scalar - vector - tensor decomposition * the scalar - vector - tensor decomposition of any @xmath199 symmetric matrix of spacetime functions @xmath200 proceeds in a similar spirit as the scalar - vector one above .
it is given by @xmath201 = \mathcal{e}_{\text{i}\text{j } } + \partial_{\{\text{i } } \mathcal{e}_{\text{j}\ } } + \frac{\mathcal{e}}{d-1 } \delta_{\text{i}\text{j } } + \left ( \partial_{\text{i } } \partial_{\text{j } } - \frac{\delta_{\text{i}\text{j}}}{d-1 } \vec{\nabla}^2 \right ) \mathcal{w } , \end{aligned}\ ] ] where @xmath202 and the following constraints are in force , @xmath203 the scalar - vector - tensor decomposition of eq .
holds whenever the spatial fourier transform of @xmath204 exists and is non - singular in the long - distance limit @xmath180 . for , in fourier space , we may write @xmath205 & = p_{\text{i}\text{j } \text{a}\text{b } } \widetilde{\mathcal{f}}_{\text{a}\text{b}}[\eta,\vec{k } ] + i k_{\{\text{i } } \left ( p_{\text{j}\}\text{k } } \frac{\widetilde{\mathcal{f}}_{\text{k}\text{l } } k_{\text{l}}}{i \vec{k}^2 } \right ) + \frac{\delta^{\text{k}\text{l } } \widetilde{\mathcal{f}}_{\text{k}\text{l}}}{d-1 } \delta_{\text{i}\text{j } } \nonumber\\ & \qquad\qquad - \left ( k_{\text{i } } k_{\text{j } } - \frac{\delta_{\text{i}\text{j}}}{d-1 } \vec{k}^2 \right ) \left ( -\frac{d-1}{d-2 } \frac{1}{\vec{k}^4 } \left\ { \widetilde{\mathcal{f}}_{\text{k}\text{l } } \cdot k_{\text{k } } k_{\text{l } } - \frac{\vec{k}^2}{d-1 } \left ( \widetilde{\mathcal{f}}_{\text{k}\text{l } } \delta^{\text{k}\text{l } } \right ) \right\ } \right ) , \end{aligned}\ ] ] whenever @xmath206 . the divergence - less projector @xmath207 is the one in eq . , while its higher rank counterpart reads @xmath208 & \equiv \frac{\delta_{\text{i}\ { \text{m } } \delta_{\text{n}\ } \text{j}}}{2 } - \frac{\delta_{\text{i}\text{j } } \delta_{\text{m}\text{n}}}{d-2 } + \frac{\delta_{\text{i}\text{j } } k_\text{m } k_\text{n } + \delta_{\text{m}\text{n } } k_\text{i } k_\text{j}}{(d-2 ) \vec{k}^2 } \nonumber\\ & \qquad\qquad - \frac{1}{2 \vec{k}^2 } \left ( \delta_{\text{i } \{\text{m } } k_{\text{n}\ } } k_\text{j } + \delta_{\text{j } \{\text{m } } k_{\text{n}\ } } k_\text{i } \right ) + \frac{d-3}{d-2 } \frac{k_\text{i } k_\text{j } k_\text{m } k_\text{n}}{\vec{k}^4 } .\end{aligned}\ ] ] it enjoys the following symmetric , transverse , and trace - free properties , @xmath209 and the projector property @xmath210 that eq . is an identity can be verified through a direct calculation .
taking into account the properties of the projectors @xmath211 and @xmath212 , comparison of the fourier - space scalar - vector - tensor decomposition in eq . to the real space one of eq
. then lets us read off : @xmath213 this is an appropriate place to highlight , except for the trace term @xmath214 , these scalar - vector and scalar - vector - tensor variables are non - local functions of the original tensor components because of the powers of momentum appearing in the denominator .
for instance , by performing the inverse fourier transform of eq . , we may witness that the longitudinal piece @xmath215 $ ] of @xmath177 requires knowing the latter everywhere in space , as the former is computed from a convolution of the latter against the euclidean laplacian s green s function : @xmath216 & = -\partial_\text{j } \int_{\mathbb{r}^{d-1}}{\text{d}}^{d-1}\vec{x } ' \mathcal{g}^{\text{(e)}}[\vec{x}-\vec{x } ' ] \mathcal{f}_\text{j}[\eta,\vec{x } ' ] , \\ \mathcal{g}^{\text{(e)}}[\vec{x}-\vec{x } ' ] & \equiv \int_{\mathbb{r}^{d-1 } } \frac{{\text{d}}^{d-1}\vec{k}}{(2\pi)^{d-1 } } \frac{e^{i k_{\text{l } } ( x^{\text{l}}-x'^{\text{l}})}}{\vec{k}^2 } .\end{aligned}\ ] ] similarly , the inverse fourier transform of eq .
yields the non - local tidal - shear " term @xmath217 as the following convolution : @xmath218 & = \frac{d-1}{d-2 } \left ( \partial_{\text{i } } \partial_{\text{j } } - \frac{\delta_{\text{i}\text{j}}}{d-1 } \vec{\nabla}^2
\right ) \int_{\mathbb{r}^{d-1}}{\text{d}}^{d-1}\vec{x } ' \left(\frac{1}{\vec{\nabla}^4}\right)[\vec{x}-\vec{x } ' ] \mathcal{f}_{\text{i}\text{j}}[\eta,\vec{x } ' ] , \\ \left(\frac{1}{\vec{\nabla}^4}\right)[\vec{x}-\vec{x } ' ] & \equiv \int_{\mathbb{r}^{d-1 } } \frac{{\text{d}}^{d-1}\vec{k}}{(2\pi)^{d-1 } } \frac{e^{i k_{\text{l } } ( x^{\text{l}}-x'^{\text{l}})}}{\vec{k}^4 } .\end{aligned}\ ] ] just like the scalar - vector decomposition , it is useful to keep in mind for applications below , that these scalar - vector - tensor components are independent in the following sense .
any relation of the form @xmath219 with @xmath220 and @xmath221 , implies that @xmath222 as long as these fields admit spatial fourier transforms that are well - behaved in the ir " limit , @xmath180 .
. follows from first taking the trace of eq . to deduce @xmath223 ; then , taking a single and double divergence in fourier space , @xmath224 as long as @xmath225 , @xmath226 from the second line ; taking this into account on the first line then tells us @xmath227 too .
consider an equation of the form @xmath228 , where @xmath229 and @xmath230 are both symmetric @xmath199 matrices of spacetime functions .
the above result implies , for all finite wavelength modes , we may perform a scalar - vector - tensor decomposition of @xmath229 and @xmath230 , and proceed to equate their @xmath231 " transverse - traceless ( tt ) , @xmath232 " divergence - free ,
@xmath233 " trace , and the @xmath217 " tidal - shear parts ( cf .
; and obtain four independent / irreducible sets of equations from @xmath228 .
similar remarks apply if we had instead @xmath234 : we may equate their @xmath233 " gradient / longitudinal parts and @xmath175 " divergence - less ones ( cf .
this strategy will in fact be employed repeatedly in the decomposition of the metric and fluid variables in and that of the linearized einstein s and perfect fluid equations in . *
* suppose we are provided with the transformation expressing some coordinate system @xmath235 in terms of another @xmath236 ; i.e. , suppose @xmath237 $ ] is given .
the components of the spacetime metric @xmath238 $ ] in @xmath239 is then related to @xmath240 $ ] in @xmath236 via the transformation @xmath241 = g_{\alpha\beta}\left[x[x']\right ]
\frac{\partial x^\alpha[x']}{\partial x'^\mu } \frac{\partial x^\beta[x']}{\partial x'^\nu } .\end{aligned}\ ] ] we are going to consider infinitesimal coordinate transformations of the @xmath97-dimensional perturbed cosmological metric in eq . . by infinitesimal transformations ,
we mean one brought about by a vector @xmath242 through @xmath243 , \qquad\qquad x ' \equiv ( \eta',\vec{x } ' ) , \end{aligned}\ ] ] where @xmath244 is small " in the same sense that the metric perturbations we are about to consider are small @xcite . at first order in @xmath16 and @xmath244 , according to eq .
, one would find that the components of eq . in the @xmath40-coordinate system
are @xmath245\right ] & \left ( \eta_{\alpha\beta } + \chi_{\alpha\beta}\left [ x[x ' ] \right ] \right ) \frac{\partial ( x'^\alpha + \xi^\alpha[x'])}{\partial x'^\mu } \frac{\partial ( x'^\beta + \xi^\beta[x'])}{\partial x'^\nu } \nonumber\\ & = a[\eta ' ] \left ( \eta_{\mu\nu } + \chi_{\mu\nu}[x ' ] + \partial _ { \ { \mu ' } \xi_{\nu\}}[x ' ] + 2 \xi^0[x ' ] \frac{\dot{a}[\eta']}{a[\eta ' ] } \eta_{\mu\nu } + \mathcal{o}\left [ ( \chi_{\alpha\beta})^2 , ( \xi_\alpha)^2 , \chi_{\alpha\beta } \xi_\sigma \right]\right ) , \end{aligned}\ ] ] where we are moving indices with the flat metric , @xmath246 , and @xmath247 . because coordinates in curved spacetimes are mere labels and have no intrinsic physical / geometric meaning until the metric itself is specified , it is customary to drop the primes on the ` new ' coordinate system on the right - hand - side of eq . and instead phrase the ensuing transformations as replacement rules . in the same vein ,
tells us the effect of the infinitesimal coordinate re - definition @xmath248 upon the metric can be , at first order , entirely attributed to the following transformation of the perturbations : @xmath249 \to a[\eta]\right ) .\end{aligned}\ ] ] if one has an expression strictly first order in @xmath16 that arose from a generally covariant field theoretic calculation , to implement an infinitesimal coordinate transformation of the form in eq .
, one only needs to replace every occurrence of @xmath16 with the right hand side of eq . .
in particular , the coordinates themselves do not need to be altered because any such change would only affect higher order perturbations .
more generally , for any rank@xmath250 perturbed tensor @xmath251 we may follow the same reasoning above , and attribute all the infinitesimal coordinate transformations upon it to its perturbation @xmath252 ; i.e. , @xmath253 whenever we implement @xmath248 .
this immediately teaches us that , whenever the background " tensor @xmath254 is zero , the first order perturbation of the same tensor is gauge invariant : @xmath255 we have already invoked this fact above , when introducing the first order perturbation theory of the einstein and fluid equations ; and we will do so again when discussing the linearized weyl tensor about a perturbed spatially flat flrw spacetime in appendix below .
. is also teaching us that , to a significant extent , the values of @xmath16 can be modified at will by choosing an appropriate vector @xmath244 , hence rendering the former s physical / geometric meaning obscure .
this arbitrariness of @xmath16 is the motivation for introducing first order gauge - invariant variables @xcite we now turn our attention towards its @xmath97-dimensional generalization . _
gauge - invariant perturbation variables _ the background cosmological metric ( and the minkowski limit , @xmath256 ) enjoys a global so@xmath172 invariance under @xmath257 , for all spacetime - independent matrices @xmath258 satisfying @xmath259 .
this is a guide to seeking the first order gauge - invariant metric perturbation variables and the pdes that govern them . by performing an irreducible scalar - vector - tensor ( svt ) decomposition of the metric perturbations as well as an analogous one for @xmath260
, we shall see that the svt decomposition of the linearized einstein - fluid equation system itself can be organized into a gauge - invariant form .
the scalar - vector decomposition of the generator - of - coordinate - transformations is @xmath261 whereas the svt one of the metric perturbation @xmath16 is @xmath262 where these variables subject to the following constraints : @xmath263 applying the irreducible decomposition of eq . to eq .
, under @xmath264 and @xmath265 , we may gather @xmath266 and the transverse - traceless graviton is gauge - invariant @xmath267 a direct calculation will then verify that the following variables are first order gauge invariant ones .
the @xmath268 scalar ones are @xmath269 in eq .
relative to that in @xcite ; @xmath270 in @xcite is @xmath271 here.]whereas the vector and tensor modes are , respectively , @xmath272 we will recall , as already remarked in @xcite , that the tensor mode @xmath53 , which obeys a minimally coupled massless wave equation , only exists when @xmath48 . * scalar fields for eft of fluids * we move on to understand how to parametrize , in a gauge - invariant manner , the first order fluid perturbations examined in this paper . in a @xmath97-dimensional spacetime there are @xmath112 scalar fields @xmath113 labeling the fluid elements in its co - moving frame . because they are scalar fields , under a generic coordinate transformation @xmath273 $ ] , @xmath274 = \varphi^\text{i}\left [ x[x ' ] \right ] .\end{aligned}\ ] ] earlier
, we have shown that the @xmath167 solves the background fluid equations - of - motion and , with an appropriate choice of lagrangian to govern their dynamics , produces a cosmology with a constant equation - of - state @xmath0 .
just as for the gravitational side of the problem , we shall study the perturbations about this background @xmath151 , namely @xmath275 ( cf .
eq . ) , and understand how the perturbations @xmath276 change under an infinitesimal coordinate transformation , @xmath248 .
eq . informs us that , to first order in @xmath244 , @xmath277 & \to x^\text{i}+\xi^\text{i } + \pi^\text{i}[x+\xi ] \nonumber\\ & = x^\text{i } + \pi^\text{i}[x ] + \xi^\text{i}[x ] + \mathcal{o}\left [ \pi^\text{i } \xi^\alpha , ( \xi^\alpha)^2 \right ] .\end{aligned}\ ] ] building on the above discussion for metric perturbations for any strictly first order expression involving @xmath165 which descended from a generally covariant set of equations we see that implementing the infinitesimal coordinate transformation @xmath248 reduces to keeping the coordinates themselves fixed and making the replacement @xmath278 \to \pi^\text{i}[x ] + \xi^\text{i}[x ] .\end{aligned}\ ] ] if we further perform a scalar - vector decomposition of @xmath276 , @xmath279 remembering @xmath280 , the irreducible pieces transform as @xmath281 a short calculation would verify that the following are first order gauge - invariant fluid perturbation variables : @xmath282 where the @xmath283 and @xmath284 are the metric perturbations in eq . .
let us begin this section by collecting the various ingredients necessary to construct the first order gauge - invariant equations for the metric @xmath285 and fluid @xmath286 perturbation variables .
* svt decomposition of astrophysical source * firstly , we shall suppose that the astrophysical system at hand has an energy - momentum - shear - stress tensor @xmath135 whose scalar - vector - tensor decomposition is given by @xmath287 because of the conservation of the stress tensor , these variables are not all independent .
we have @xmath288 handing us @xmath289 while the longitudinal and transverse parts of @xmath290 return @xmath291 _ non - conservation of mass _ in the cosmological geometry of eq . , the @xmath97-beins corresponding to the local lorentz frame of a co - moving observer are given by @xmath292.-beins we record that the @xmath54 in eq .
should be related to @xmath293 in eq . through : @xmath294 .
] in such an orthonormal frame , the mass @xmath295 of the system can be defined through the proper spatial volume integral of @xmath296 : @xmath297 \equiv \int_{\mathbb{r}^{d-1 } } { \text{d}}^{d-1}\vec{x } ' a[\eta]^{d-3 } \rho[\eta,\vec{x } ' ] .\end{aligned}\ ] ] similarly , the pressure @xmath298 in the @xmath299th direction can be defined in terms of the proper spatial volume integral of @xmath300 , with no summation over @xmath299 implied .
this in turn leads us to @xmath301 = \int_{\mathbb{r}^{d-1 } } { \text{d}}^{d-1}\vec{x } ' a[\eta]^{d-3 } \sigma[\eta,\vec{x } ' ] .\end{aligned}\ ] ] by virtue of being in a cosmological environment , the mass @xmath295 of the system is not , in fact , conserved .
taking a time derivative of eq . , and making use of the conservation law in eq .
, @xmath302 converting the laplacian term to a surface integral at spatial infinity , we see that it is trivial .
we may therefore gather , through the relation in eq . ,
@xmath303 = - \frac{\dot{a}[\eta]}{a[\eta ] } \sum_{i=1}^{d-1 } \,^{\text{(a)}}p_i[\eta ] .\end{aligned}\ ] ] * first order perturbations of @xmath304 * we will also need the perfect fluid stress tensor in eq .
developed up to first order in @xmath164 and @xmath165 .
useful intermediate results are @xmath305 and @xmath306 the desired @xmath307$]-accurate fluid stress tensor is @xmath308 in equations through , the @xmath309 " represent terms that scale as @xmath310 $ ] or higher . * linearized einstein s equations * we are now ready to enumerate the eom for the gauge - invariant metric perturbations @xmath311 , @xmath271 , @xmath312 and @xmath53 .
we used xact @xcite to work out the linearized einstein tensor @xmath313 and subtracted from it the first order piece of the fluid stress tensor in eq . to compute the left - hand - side of eq .
, @xmath166 . by first decomposing @xmath16 and @xmath276
according to equations and respectively , the end result for @xmath166 is gauge - invariant if we group together the appropriate terms according to the definitions in equations , , and .
as already remarked earlier , this is both a check of the calculation itself , as well as the starting point for the ensuing scalar - vector - tensor decomposition of the linearized einstein s equation . only appears in @xmath271 and @xmath314 only in @xmath311 .
we may therefore use eq . to replace all occurrences of @xmath315 with its equivalent expression in terms of @xmath271 , @xmath316 and @xmath284 ; eq . to replace all @xmath314 s in terms of @xmath311 , @xmath284 , and @xmath316 ; eq . to replace all @xmath317 s in terms of @xmath312 and @xmath318 .
one should find all the remaining @xmath316 , @xmath284 and @xmath318 to cancel .
it may also be useful , at this point , to refer to the decompositions of equations and and recall the discussion starting just before eq . through the end of .
furthermore we highlight that , if one first computes the following eom in fourier - space , the formulas in equations , , and may prove handy for extracting the relevant scalar - vector(-tensor ) components . ]
_ scalars _ the scalar eom for @xmath311 and @xmath271 arise from the @xmath145 component of eq . , with @xmath319 ( cf .
eq . ) , @xmath320 equating the gradient portions of the @xmath321 components on both sides of eq . , with @xmath322 given in eq .
, @xmath323 equating the spatial trace portions of the @xmath324 components on both sides of eq . , with @xmath325 given in eq . , @xmath326 as well as equating the tidal - shear part of the @xmath324 components on both sides of eq .
, again with @xmath325 given in eq .
, @xmath327 because @xmath271 can be immediately obtained from @xmath311 through eq .
, we will focus on solving @xmath311 throughout the rest of this paper .
_ vector _ the vector perturbations come from equating the divergence - free part of the @xmath321 components on both sides of eq . , with @xmath322 given in eq . ,
@xmath328 and equating the divergence - less vector portion of the @xmath324 components on both sides of eq . ,
with @xmath325 given in eq .
, @xmath329 _ tensor _ finally , the tensor equation arise from equating the transverse - traceless parts of the @xmath324 components on both sides of eq . , with @xmath325 given in eq .
, @xmath330 the @xmath331 is the scalar spacetime laplacian ; with the @xmath332 in eq . , @xmath333 .
we also remind the reader that @xmath334 ( cf .
eq . ) .
* linearized fluid eom * inserting into the fluid eom in eq .
the metric and fluid expansions of equations and , followed by their decompositions in equations and , a direct calculation then hand us the linearized fluid equations : @xmath335 a^{2(1-d ) } \left\ { \partial_{\text{i } } \left ( \partial_\text{k } \pi^\text{k } + \frac{\delta^{mn } \chi_{mn}}{2 } \right ) - ( 1-d ) \frac{\dot{a}}{a } \left ( \chi_{0\text{i } } + \dot{\pi}^\text{i } \right ) \right\ } \\ & + \rho_\text{f}'[a^{2(1-d)}/2 ] \left\ { \partial_{\text{i } } \left ( \partial_{\text{k } } \pi^{\text{k } } + \frac{\delta^{mn } \chi_{mn}}{2 } \right ) - ( 1-d ) \frac{\dot{a}}{a } \left ( \chi_{0\text{i } } + \dot{\pi}^\text{i } \right ) + \frac{1}{2 } \partial_{\text{i } } \chi_{00 } - \frac{1}{a } \partial_0 \left\ { a \left ( \chi_{0\text{i } } + \dot{\pi}^\text{i } \right ) \right\ } \right\ } = 0 .
\nonumber \end{aligned}\ ] ] we then proceed to group terms according to the definitions laid out in equations , , and . for all finite wavelength modes , the gradient " part of eq .
provides us with an equation for @xmath336 : @xmath337 a^{2(1-d ) } & \left\ { -\vec{\nabla}^2 \xi - ( d-1 ) \frac{\dot{a}}{a } \dot{\xi } + \frac{d-1}{2 } \phi \right\ } \nonumber\\ & \qquad\qquad + \rho_\text{f}'[a^{2(1-d)}/2 ] \left\ { -\vec{\nabla}^2 \xi - ( d-2 ) \frac{\dot{a}}{a } \dot{\xi } + \ddot{\xi } + \frac{d-1}{2 } \phi + \frac{1}{2}\psi \right\ } = 0 .\end{aligned}\ ] ] the divergence - free portion of eq .
yields the equation for @xmath338 : @xmath339 a^{2(1-d ) } & \left\ { ( d-1 ) \frac{\dot{a}}{a } \left ( v_\text{i } - \dot{\xi}_\text{i } \right ) \right\ } \nonumber\\ & \qquad\qquad + \rho_\text{f}'[a^{2(1-d)}/2 ] \left\ { ( d-2 ) \frac{\dot{a}}{a } \left ( v_\text{i } - \dot{\xi}_\text{i } \right ) - \left ( \dot{v}_\text{i } - \ddot{\xi}_\text{i } \right ) \right\ } = 0 .\end{aligned}\ ] ] equations and hold for any @xmath109 evaluated on the background fluid solution @xmath167 . if we specialize to the case of interest , namely with @xmath109 given in eq .
, we find equations and becoming respectively @xmath340 and @xmath341 as discussed in , the @xmath128 case corresponds to einstein s equations with a positive cosmological constant .
the background metric solution is de sitter spacetime and , in particular , there is really no background cosmological fluid driving its evolution this is reflected by the fact that , when we put @xmath128 in equations and , these fluid eom become a pair of irrelevant @xmath342 identities .
with the eom laid out , we will turn to describing their solutions for the remaining of this section . * tensor mode solutions * the tensor mode wave equation can be re - expressed as @xmath343 with @xmath344 .
the solution to the inverse of the wave operator taking the form @xmath345 in eq . , and throughout the rest of this paper , is reviewed in appendix ; we will simply report the results in the main text . here
, the retarded solution is @xmath346^{\frac{d-2}{2 } } d_{ij}[\eta,\vec{x } ] & = -16\pi{g_{\rm n}}\int_{\mathbb{r}^{d-1 } } { \text{d}}^{d-1}\vec{x } ' \int_{\eta_\circ}^\eta { \text{d}}\eta ' a[\eta']^{\frac{d-2}{2 } } \mathcal{g}^{\text{(t)}}[\eta,\eta';{\bar{\sigma } } ] \sigma_{ij}[\eta',\vec{x } ' ] , \end{aligned}\ ] ] where the symmetric green s functions are @xmath347 & = \left ( \frac{1}{2\pi } \frac{\partial}{\partial { \bar{\sigma } } } \right)^{\frac{d-2}{2 } } \left(\frac{\theta[{\bar{\sigma}}]}{2 } p_{-\frac{d-2}{q_w}}[1+s ] \right ) , \\ \label{firstorderpteqns_tensor_gsoln_oddd } \mathcal{g}^{\text{(t)}}_{\text{odd $ d \geq 5$}}[\eta,\eta';{\bar{\sigma } } ] & = \left ( \frac{1}{2\pi } \frac{\partial}{\partial { \bar{\sigma } } } \right)^{\frac{d-3}{2 } } \left(\frac{\theta[{\bar{\sigma}}]}{4\pi } \frac { \left(s + \sqrt{s(s+2 ) } + 1\right)^{\frac{(w-1)(d-1)}{q_w } } + 1}{\sqrt{{\bar{\sigma}}(s+2 ) } \left(s + \sqrt{s(s+2 ) } + 1\right)^{\frac{(w-1)(d-1)}{2q_w } } } \right ) , \\ { \bar{\sigma}}&\equiv \frac{1}{2}\left((\eta-\eta')^2-(\vec{x}-\vec{x}')^2\right ) , \qquad\qquad s \equiv \frac{{\bar{\sigma}}}{\eta\eta ' } .\end{aligned}\ ] ] the @xmath348 occurring in eq . is the legendre function , and will show up repeatedly in the even dimensional solutions in this paper .
let us highlight the 4d case @xmath349 .
its light cone part is @xmath350 & = -4 { g_{\rm n}}\theta_p \int_{\mathbb{r}^{3 } } { \text{d}}^{3}\vec{x } ' \frac{a[\eta_r]}{a[\eta ] } \frac{\sigma_{ij}[\eta_r,\vec{x}']}{|\vec{x}-\vec{x}'| } , \end{aligned}\ ] ] with retarded time @xmath351 ; and the symbol @xmath352 , as explained in appendix , reminds us that the solution is non - zero whenever @xmath49 only within the particle horizon of the gw source .
the tail portion is @xmath353 & = -4 { g_{\rm n}}\int_{\mathbb{r}^{3 } } { \text{d}}^{3}\vec{x } ' \int_{\eta_\circ}^{\eta_r } { \text{d}}\eta ' \frac{a[\eta']}{a[\eta ] \cdot \eta\eta ' } p'_{-\frac{2}{1 + 3w}}\left [ 1 + \frac{(\eta-\eta')^2 - ( \vec{x}-\vec{x}')^2}{2\eta\eta ' } \right ] \sigma_{ij}[\eta',\vec{x } ' ] .\end{aligned}\ ] ] _ null waves _ before before moving on to the vector and scalar solutions , let us compare the cosmological solution of eq . and its minkowski version , i.e. , the tt part of eq . below .
the scale factors in the former , multiplying @xmath53 on the left - hand - side tells us , to leading order , the portion of tensor waves propagating strictly on the null cone including their associated memories
is expected to redshift as @xmath354^{(d-2)/2}$ ] relative to their minkowskian cousins , at least in all relevant even dimensions @xmath355 . * vector solutions * according to eq .
, if there were no perturbations to begin with in the asymptotic past , then @xmath356 for all time .
to see this , we integrate once to obtain @xmath357 - \dot{\xi}_i[\eta,\vec{x } ] = \left(\frac{a[\eta]}{a[\eta']}\right)^{(d-1)w-1 } ( v_i[\eta',\vec{x } ] - \dot{\xi}_i[\eta',\vec{x } ] ) , \end{aligned}\ ] ] for some fixed initial time @xmath23 .
if this initial time @xmath23 is taken to be in the far past , so that we may assume no metric perturbations has been generated , this means @xmath358 = \dot{\xi}_i[\eta',\vec{x } ] = 0 $ ] and therefore @xmath356 .
using this in eq . now hands us @xmath359 which may be readily solved via the green s function of the laplacian in @xmath112-space , @xmath360 = -16\pi{g_{\rm n}}\int_{\mathbb{r}^{d-1 } } { \text{d}}^{d-1}\vec{x } ' \frac{\gamma\left[\frac{d-3}{2}\right ] \cdot \sigma_i[\eta,\vec{x}']}{4\pi^ { \frac{d-1}{2 } } |\vec{x}-\vec{x}'|^{d-3 } } .\end{aligned}\ ] ] our assumption that @xmath312 tends to zero in the distant past therefore amounts to assuming that @xmath361 itself is negligible in the distant past . as a check of the consistency of eq .
, we may apply the operator @xmath362 on both sides , followed by using eq .
, @xmath363 which is simply the conservation law of eq . .
alternatively , we may also integrate eq . once to obtain @xmath364 in terms of the @xmath365 component of @xmath135 .
@xmath366 = 16\pi{g_{\rm n}}\int_{\eta'}^{\eta } \left(\frac{a[\eta'']}{a[\eta]}\right)^{d-2 } \partial_{\{i } \sigma _ { j\ } } [ \eta'',\vec{x } ] { \text{d}}\eta '' + \left(\frac{a[\eta']}{a[\eta]}\right)^{d-2 } \partial_{\{i } v _ { j\ } } [ \eta',\vec{x}].\end{aligned}\ ] ] * scalar solution in de sitter ( @xmath128 ) * in de sitter spacetime , or equivalently when @xmath128 , the fluid equations become irrelevant .
we may now set @xmath128 in equations and , and eliminate @xmath271 to arrive at @xmath367 the solution is @xmath368 = -\frac{16\pi{g_{\rm n}}}{d-2 } \int_{\mathbb{r}^{d-1 } } { \text{d}}^{d-1}\vec{x } ' \frac{\gamma\left[\frac{d-3}{2}\right]}{4\pi^ { \frac{d-1}{2 } } |\vec{x}-\vec{x}'|^{d-3 } } \left ( \rho[\eta,\vec{x } ' ] + h a[\eta ] ( d-1 ) \sigma[\eta,\vec{x } ' ] \right ) , \end{aligned}\ ] ] where we have introduced a constant hubble parameter @xmath369 and expressed the de sitter scale factor as @xmath146 \equiv -1/(h\eta)$ ] .
* scalar solution in matter dominated universe ( @xmath370 ) * by putting @xmath370 in eq . , and invoking eq . to eliminate @xmath271
, we immediately obtain an ordinary differential equation in time for @xmath311 , sourced by @xmath47 and @xmath371 : @xmath372 if we assume the @xmath311 and @xmath373 were zero in the distant past , then we may integrate twice to obtain its solution : @xmath374 = 16\pi{g_{\rm n}}\int_{\eta_\circ}^{\eta } \frac{{\text{d}}\eta'''}{a[\eta''']^{2d-5 } } \int_{\eta_\circ}^{\eta ' '' } { \text{d}}\eta '' a[\eta'']^{2d-5 } \left ( \frac{\sigma[\eta'',\vec{x}]}{(d-1)(d-2 ) } + \frac{\vec{\nabla}^2 \upsilon[\eta'',\vec{x}]}{d-1 } + \frac{2 \dot{\upsilon}[\eta'',\vec{x}]}{(d-3 ) \eta '' } \right ) .\end{aligned}\ ] ] * scalar solution for @xmath1 * the reason why the @xmath128 and @xmath370 cases need to be treated separately is because the character of the differential equations involved becomes different at these 2 discrete points on the eos parameter line .
for instance , setting @xmath370 in eq .
sets to zero its laplacian term . as such , when @xmath128 or @xmath370 , the following manipulations will not be available to reduce the scalar equations to a single wave equation for @xmath311 .
we first use the fluid eq . to replace the @xmath375 in eq . with an expression involving @xmath376 , @xmath377 , @xmath311 and @xmath271 .
then , to get rid of the remaining @xmath376 and @xmath377 in eq .
, we first solve @xmath377 in terms of @xmath373 , @xmath271 and @xmath378 using eq . ; by taking one time derivative of this latter result , @xmath376 can also be replaced in terms of @xmath373 , @xmath379 , @xmath271 , @xmath380 , @xmath378 and @xmath381 .
the conservation law of eq . can be used to remove the @xmath381 term .
finally , we replace all @xmath271 in terms of @xmath311 using eq . .
after all these steps , one should arrive at @xmath382 on the left hand side of eq .
recovers that of eq .
( 7.57 ) in mukhanov s textbook @xcite .
( mukhanov sets @xmath383.)]the left hand side of eq .
is proportional to the scalar spacetime laplacian @xmath384 acting on @xmath311 , but in the geometry @xmath385 in detail , @xmath386 _ physical range of eos _ for @xmath387 , the speed of the @xmath311-wavefront is @xmath388 .
is commonly called the sound speed squared " . ] therefore , in order for the scalar - hydrodynamical system to not admit superluminal signals , its eos must be constrained to be @xmath389 .
we now re - write the scalar mode wave equation as @xmath390 with @xmath391 .
the @xmath311 solution is @xmath392^ { \frac{1}{2}\left ( 2d-5+(d-1)w \right ) } \phi[\eta,\vec{x } ] & = \frac{16\pi{g_{\rm n}}}{w^{\frac{d-1}{2 } } } \int_{\mathbb{r}^{d-1 } } { \text{d}}^{d-1}\vec{x } ' \int_{\eta_\circ}^\eta{\text{d}}\eta ' a[\eta']^ { \frac{1}{2}\left ( 2d-5+(d-1)w \right ) } \mathcal{g}^{(\phi)}[\eta,\eta';{\bar{\sigma}}_w ] \\ & \qquad\qquad\times \left ( - w \frac{\rho[\eta',\vec{x}']}{d-2 } + \frac{\sigma[\eta',\vec{x}']}{(d-2)(d-1 ) } + \frac{2 \dot{\upsilon}[\eta',\vec{x}']}{q_w \eta ' } + \frac{\vec{\nabla}^2_{\vec{x } ' } \upsilon[\eta',\vec{x}']}{d-1 } \right ) , \nonumber\end{aligned}\ ] ] where the symmetric green s functions are @xmath393 & = \left ( \frac{1}{2\pi } \frac{\partial}{\partial{\bar{\sigma}}_w } \right)^ { \frac{d-2}{2 } } \left ( \frac{\theta[{\bar{\sigma}}_w]}{2 } p _
{ -\frac{2d-5 + ( d-1)w}{q_w } } \left[1+s_w\right ] \right)\\ \label{firstorderpteqns_phig_w>0_oddd } \mathcal{g}^{(\phi)}_{\text{odd $ d \geq 5$}}[\eta,\eta';{\bar{\sigma}}_w ] & = \left ( \frac{1}{2\pi } \frac{\partial}{\partial{\bar{\sigma}}_w } \right)^{\frac{d-3}{2 } } \left(\frac{\theta[{\bar{\sigma}}_w]}{4\pi } \frac { \left(s_w + \sqrt{s_w(s_w+2 ) } + 1\right)^{\frac{3d-7 + ( d-1)w}{q_w } } + 1}{\sqrt{{\bar{\sigma}}_w(s_w+2 ) } \left(s_w + \sqrt{s_w(s_w+2 ) } + 1\right)^{\frac{3d-7 + ( d-1)w}{2q_w } } } \right ) , \\ { \bar{\sigma}}_w & \equiv \frac{1}{2}\left ( ( \eta-\eta')^2 - \frac{(\vec{x}-\vec{x}')^2}{w } \right ) , \qquad\qquad s_w \equiv \frac{{\bar{\sigma}}_w}{\eta\eta ' } .\end{aligned}\ ] ]
the presence of @xmath394 in eq . reminds us that the @xmath370 case considered separately above is disconnected from its @xmath395 counterpart .
we also remark that , unlike the tensor mode above , the scalar mode radiation can be sourced by 3 irreducible components @xmath396 of the energy - momentum - shear - stress of the astrophysical system .
we again highlight the 4d case by writing down its retarded green s function : @xmath397 & = \frac{1}{4\pi } \bigg ( \sqrt{w } \frac{\delta[\eta-\eta'-|\vec{x}-\vec{x}'|/\sqrt{w}]}{|\vec{x}-\vec{x}'| } \\ & \qquad\qquad + \frac{\theta[\eta-\eta'-|\vec{x}-\vec{x}'|/\sqrt{w}]}{\eta\eta ' } p ' _ { -\frac{3(1+w)}{1 + 3w } } \left[1+\frac{(\eta-\eta')^2-(\vec{x}-\vec{x}')^2/w}{2\eta\eta ' } \right ] \bigg ) \nonumber .\end{aligned}\ ] ] as already pointed out by tolish and wald @xcite , it is possible to generate cherenkov @xmath311-radiation if there are astrophysical sources moving faster than @xmath398 with respect to the cosmic rest frame . _
negative eos _
when @xmath399 , the scalar mode equation becomes a poisson - type one because the coefficient of the spatial laplacian term is now positive .
if one wishes to quantize the @xmath311 fluctuations as is the case during the primordial epoch of the universe , when the initial seeds of inhomogeneities were believed to be laid down this would lead to pathologies because half of the @xmath311 s mode functions , which solve the homogeneous equation @xmath400 would grow unbounded in amplitude with increasing time . specifically , in fourier space , where @xmath401 e^{i\vec{k}\cdot\vec{x}}$ ] , one would find that @xmath402 $ ] is a linear combination involving hankel functions : @xmath403 $ ] and @xmath404 $ ] , with @xmath405 . obeying @xmath406 , in some @xmath112-euclidean space , but insist that @xmath407 is ` time ' . by performing a fourier transform with respect to the rest of the cartesian
coordinates @xmath408 , namely @xmath409 \to \widetilde{\phi}[t ] e^{i\vec{k}_\perp \cdot \vec{x}_\perp}$ ] , one would find that @xmath410 $ ] is a linear combination of @xmath411 $ ] .
] at the purely classical level , even though we believe the relevant @xmath399 green s functions for eq .
exists , the latter s poisson - type nature means the ensuing solutions would depend on the sources @xmath396 in the future of the observer s spacetime event .
altogether , these reasons compel us to restrict our attention to the eos range @xmath1 . * fluid solutions * when @xmath128 , or equivalently , about de sitter spacetime , the fluid equations are irrelevant . when @xmath370 , eq .
allows us to integrate twice with respect to time and solve the longitudinal @xmath336 portion of the fluid perturbation in terms of the scalar metric perturbation @xmath271 and its value @xmath412 $ ] at some initial time @xmath23 .
@xmath413 = -\frac{1}{2 } \int_{\eta'}^{\eta } \frac{{\text{d}}\eta'''}{a[\eta ' '' ] } \int_{\eta'}^{\eta ' '' } a[\eta'']\psi[\eta'',\vec{x } ] + \xi[\eta',\vec{x}].\end{aligned}\ ] ] for @xmath1 , the longitudinal @xmath336 portion of the fluid obeys a wave equation sourced by the scalar metric perturbations @xmath311 and @xmath271 .
we re - express eq .
as @xmath414 where @xmath391 .
we will not consider the ranges @xmath415 and @xmath416 for the same reasons when discussing the wave solutions for @xmath311 above .
the retarded solutions are @xmath417^ { \frac{1}{2}\left ( 1-(d-1)w \right ) } \xi[\eta , \vec{x } ] = - \int_{\mathbb{r}^{d-1 } } { \text{d}}^{d-1}\vec{x } ' \int_{\eta_\circ}^{\eta } { \text{d}}\eta ' & a[\eta']^ { \frac{1}{2}\left ( 1-(d-1)w \right ) } \mathcal{g}^{(\xi)}[\eta,\eta';{\bar{\sigma } } ] \nonumber\\ & \times \left ( \frac{1}{2 } ( d-1 ) w \phi[\eta',\vec{x } ' ] + \frac{1}{2}\psi[\eta',\vec{x } ' ] \right ) , \end{aligned}\ ] ] with the following symmetric green s functions : @xmath418 & = \left ( \frac{1}{2\pi } \frac{\partial}{\partial{\bar{\sigma}}_w } \right)^ { \frac{d-2}{2 } } \left ( \frac{\theta[{\bar{\sigma}}_w]}{2 } p_{\frac{(d-1)w-1}{q_w}}[1+s_w ] \right ) , \\
\mathcal{g}^{(\xi)}_{\text{odd $ d \geq 5$}}[\eta,\eta';{\bar{\sigma } } ] & = \left ( \frac{1}{2\pi } \frac{\partial}{\partial{\bar{\sigma}}_w } \right)^ { \frac{d-3}{2 } } \left ( \frac{\theta[{\bar{\sigma}}_w]}{4\pi } \frac{1 + \left ( s_w + \sqrt{s_w(s_w+2 ) } + 1 \right)^ { \frac{(d-5 ) + 3 ( d-1 ) w}{q_w } } } { \sqrt{{\bar{\sigma}}_w(s_w+2 ) } \left ( s_w + \sqrt{s_w(s_w+2 ) } + 1 \right)^ { \frac{(d-5 ) + 3 ( d-1 ) w}{2q_w } } } \right ) , \\ { \bar{\sigma}}_w & \equiv \frac{(\eta-\eta')^2 - ( \vec{x}-\vec{x}')^2/w}{2 } , \qquad\qquad s_w \equiv \frac{{\bar{\sigma}}_w}{\eta\eta ' } , \qquad\qquad 0 < w \leq 1 .\end{aligned}\ ] ] finally , for the solution of the transverse portion @xmath419 of the fluid perturbation we recall in the discussion above , that we wish to implement the initial conditions @xmath420 = \xi_i[\eta_\circ,\vec{x } ] = 0 $ ] .
this implies from eq . that @xmath421 for all times . therefore , for any @xmath0 , the transverse @xmath419 portion of the fluid is @xmath422 = \int_{\eta'}^{\eta } v_i[\eta',\vec{x } ] { \text{d}}\eta ' + \xi_i[\eta',\vec{x } ] , \end{aligned}\ ] ] where @xmath423 $ ] is its value at some arbitrary initial time @xmath23 .
gravitational memory is usually defined as the permanent fractional displacement experienced by a pair of co - moving test masses after the passage of a primary gw train emitted by a distant astrophysical source .
we will begin with the observation that the synchronous gauge is the appropriate choice of coordinates to perform such a gw memory calculation .
this is because , the geodesic equations obeyed by co - moving observers are satisfied automatically , and what remains is to compute the metric perturbations that occur in the fractional displacement formula of eq . .
we will do so by converting the first order gauge - invariant results of the previous section to the synchronous gauge , and proceed to work out the gw signatures including potential memory effects from generic isolated astrophysical sources .
* synchronous gauge * the synchronous gauge refers to the choice of coordinates such that all metric perturbations occur solely within the spatial metric namely , @xmath424 ^ 2 \left ( ( { \text{d}}\eta)^2 - \left(\delta_{ij } - \chi_{ij}\right ) { \text{d}}x^i { \text{d}}x^j \right ) .\end{aligned}\ ] ] as can be verified by a direct calculation , the timelike trajectory with zero spatial velocity @xmath425 with @xmath21 constant ( i.e. , it is co - moving ) satisfies the geodesic equation @xmath426 to all orders in @xmath30 , with proper time @xmath427 defined through @xmath428{\text{d}}\eta$ ] . in other words : in the synchronous gauge metric , if a geodesic worldline has zero spatial proper velocity at a given instant in time , it is co - moving for all time .
this allows us to identify , from eq .
, that the induced metric on constant
time surfaces measures geodesic spatial distances between any pair of such co - moving geodesic test masses : @xmath429 ^ 2 \left ( \delta_{ij } - \chi_{ij } \right ) { \text{d}}x^i { \text{d}}x^j .\end{aligned}\ ] ] as we shall show in appendix , assuming the hypothetical gw experiment does not take place over cosmological timescales , this leads us to the fractional displacement formula of eq . .
now , for eq . to be useful , we need to understand how to extract the synchronous gauge metric perturbations from the solutions to the first order gauge invariant metric perturbation variables we obtained in the previous section . by referring to eq .
, we see synchronous gauge means @xmath430 starting with eq . , @xmath284 can be determined from @xmath271 by integrating @xmath431 : @xmath432 & = \int_{\eta'}^{\eta}\frac{{\text{d}}\eta'''}{a[\eta ' '' ] } \int_{\eta'}^{\eta ' '' } a[\eta '' ] \psi[\eta'',\vec{x } ] { \text{d}}\eta '' + \dot{k}[\eta',\vec{x } ] \int_{\eta'}^{\eta } \frac{a[\eta']}{a[\eta '' ] } { \text{d}}\eta '' + k[\eta',\vec{x } ] , \end{aligned}\ ] ] for some arbitrary but fixed initial time @xmath23 . employing this solution for @xmath284 , eq .
can be used to solve for the trace of the spatial metric perturbation : @xmath433}{d-1 } = \phi[\eta,\vec{x } ] + \frac{\vec{\nabla}^2 k[\eta,\vec{x}]}{d-1 } + \frac{\dot{a}[\eta]}{a[\eta ] } \left ( \int_{\eta'}^{\eta } \frac{a[\eta'']}{a[\eta ] } \psi[\eta'',\vec{x } ] { \text{d}}\eta '' + \frac{a[\eta']}{a[\eta ] } \dot{k}[\eta',\vec{x } ] \right ) .\end{aligned}\ ] ] this implies the spatial - trace plus tidal - shear " part of the spatial metric perturbation in eq . is @xmath434 + \frac{\dot{a}[\eta]}{a[\eta ] } \left ( \int_{\eta'}^{\eta } \frac{a[\eta'']}{a[\eta ] } \psi[\eta'',\vec{x } ] { \text{d}}\eta '' + \frac{a[\eta']}{a[\eta ] } \dot{k}[\eta',\vec{x } ]
\right ) \right ) \nonumber \\ & + \partial_i \partial_j \left\ { \int_{\eta'}^{\eta}\frac{{\text{d}}\eta'''}{a[\eta ' '' ] } \int_{\eta'}^{\eta ' '' } { \text{d}}\eta '' a[\eta '' ] \psi[\eta'',\vec{x } ] + \dot{k}[\eta',\vec{x } ] \int_{\eta'}^{\eta } \frac{a[\eta']}{a[\eta '' ] } { \text{d}}\eta '' + k[\eta',\vec{x } ] \right\ } .\end{aligned}\ ] ] finally , eq . tells us the portion of the spatial metric involving the divergence - free vector is @xmath435 & = - \int_{\eta'}^{\eta } \partial _ { \ { i } v _ { j \ } } [ \eta'',\vec{x } ] { \text{d}}\eta '' + \partial _ { \ { i } d _ { j \ } } [ \eta',\vec{x } ] .\end{aligned}\ ] ] we may now gather , by inserting equations and into eq . :
given @xmath53 , @xmath312 , @xmath311 and @xmath271 , the synchronous gauge metric perturbation is @xmath436 & = d_{ij}[\eta,\vec{x } ] - \int_{\eta'}^{\eta } \partial _ { \ { i } v _ { j \ } } [ \eta'',\vec{x } ] { \text{d}}\eta '' + \partial _ { \ { i } d _ { j \ } } [ \eta',\vec{x } ] \\ & + \delta_{ij } \left\ { \phi[\eta,\vec{x } ] + \frac{\dot{a}[\eta]}{a[\eta ] } \left ( \int_{\eta'}^{\eta } \frac{a[\eta'']}{a[\eta ] } \psi[\eta'',\vec{x } ] { \text{d}}\eta '' + \frac{a[\eta']}{a[\eta ] } \dot{k}[\eta',\vec{x } ] \right ) \right\ } \nonumber \\ & + \int_{\eta'}^{\eta}\frac{{\text{d}}\eta'''}{a[\eta ' '' ] } \int_{\eta'}^{\eta ' '' } { \text{d}}\eta '' a[\eta '' ] \partial_i \partial_j \psi[\eta'',\vec{x } ] + \partial_i \partial_j \dot{k}[\eta',\vec{x } ] \int_{\eta'}^{\eta } \frac{a[\eta']}{a[\eta '' ] } { \text{d}}\eta '' + \partial_i \partial_j k[\eta',\vec{x } ] .
\nonumber\end{aligned}\ ] ] , constitutes a constructive proof of the existence of the synchronous gauge itself.]the presence of the arbitrary initial metric perturbation @xmath437 $ ] and its velocity @xmath438 $ ] indicates the synchronous gauge does not completely eliminate the freedom to alter the coordinate system .
however , to recover a manifestly spatially flat flrw cosmology when there are no gw sources
i.e. , when all the @xmath439 are zero we shall take the liberty to choose @xmath440 = \dot{k}[\eta',\vec{x } ] = 0 .\end{aligned}\ ] ] furthermore , since it was arbitrary up till this point , let us now define @xmath23 to be the beginning of our hypothetical gw experiment , which we further suppose measures the @xmath24 $ ] in eq . .
at this point , with eq
. applied to eq .
, the total change in the spatial metric @xmath441 \equiv \chi_{ij}^{\text{(syn.)}}[\eta,\vec{x } ] - \chi_{ij}^{\text{(syn.)}}[\eta',\vec{x}]$ ] , required to calculate eq . , takes the expression @xmath442 = \delta d_{ij}[\eta,\vec{x } ] - \int_{\eta'}^{\eta } \partial _ { \ { i } v _ { j \ } } [ \eta'',\vec{x } ] { \text{d}}\eta '' \\ & + \delta_{ij } \left\ { \delta \phi[\eta,\vec{x } ] + \frac{\dot{a}[\eta]}{a[\eta ] } \int_{\eta'}^{\eta } \frac{a[\eta'']}{a[\eta ] } \psi[\eta'',\vec{x } ] { \text{d}}\eta '' \right\ } + \int_{\eta'}^{\eta}\frac{{\text{d}}\eta'''}{a[\eta ' '' ] } \int_{\eta'}^{\eta ' '' } { \text{d}}\eta '' a[\eta '' ] \partial_i \partial_j \psi[\eta'',\vec{x } ] , \nonumber\end{aligned}\ ] ] where we have defined @xmath443 & \equiv d_{ij}[\eta,\vec{x } ] - d_{ij}[\eta',\vec{x } ] , \\
\delta \phi[\eta \geq \eta',\vec{x } ] & \equiv \phi[\eta,\vec{x } ] - \phi[\eta',\vec{x } ] .\end{aligned}\ ] ] if we assume that @xmath312 was zero in the distant past , or if we note that the initial condition @xmath444 $ ] term in eq . there , which is just an arbitrary ` initial time ' , is not to be confused with the @xmath23 here , which is the starting time of our gw experiment
. ] decays as @xmath354^{d-2}$ ] in any case , we see that the time - integral involving @xmath445 in eq .
has to vanish rapidly outside the astrophysical system because the remaining term in eq .
is the time - integral of the irreducible @xmath446 component of its stress - energy tensor @xmath135 .
we will thus drop this @xmath445 term in eq . .
next , if we infer from eq .
that @xmath447 , then the single - integral term on the second line in eq . should scale relative to @xmath448 as , very roughly , ( timescale of gw experiment)/(age of universe ) due to the presence of the @xmath449 pre - factor ; similarly , the double integral on the same line containing @xmath450 must scale relative to @xmath448 , again roughly , as ( ( timescale of gw experiement)/(observer - source spatial distance))@xmath451 . for distant astrophysical sources and for human experimental timescales , we may therefore estimate these integral terms to be negligible . to sum
, these considerations have reduced the contributions of eq . to fractional distortion
@xmath24 $ ]
i.e. , equations and to the following two terms : @xmath452 & = - \frac{\widehat{n}^i \widehat{n}^j}{2 } \int_{0}^{1 } \left(\delta d_{ij}[\vec{w } ] + \delta_{ij } \delta \phi[\vec{w } ] \right ) { \text{d}}\lambda , \\ \vec{w}[\lambda ] & \equiv \vec{y}_0 + \lambda(\vec{z}_0-\vec{y}_0 ) .
\nonumber\end{aligned}\ ] ] at least in the long wavelength limit where both @xmath53 and @xmath311 can be treated as approximately constant between the two test masses at a fixed time while @xmath53 has a non - trivial tensorial structure that yields a different @xmath453 depending on the orientation of the pair of test masses @xmath454 , notice the @xmath311-waves produce uniform - compressional patterns with no direction dependence .
as we will now proceed to further argue , for the de sitter @xmath455 and matter dominated @xmath456 cases , the @xmath311 contribution may in fact be discarded . @xmath457 = - \frac{\widehat{n}^i \widehat{n}^j}{2 }
\int_{0}^{1 } \delta d_{ij}[\vec{y}_0 + \lambda(\vec{z}_0-\vec{y}_0 ) ] { \text{d}}\lambda \end{aligned}\ ] ] * gws and their memories in de sitter ( @xmath128 ) * the poisson - type solution for @xmath311 in eq
. suggests it may not fluctuate as rapidly as the wave solutions of its tensor mode counterpart in eq . , and therefore @xmath448 is sub - dominant relative to @xmath458 . to check this intuition
we apply the operator @xmath459 on both sides of eq . , and use the conservation laws in equations and to reveal @xmath460 far away from the source , we may first take a spatial gradient on both sides to conclude that @xmath461 plus terms that are suppressed by @xmath462(age of universe ) .
but we recognize @xmath463 to be an irreducible component of @xmath322 , which we assume is zero at the gw detector .
thus , we instead estimate @xmath464 to scale as the gradient of the rightmost term of eq .
, namely @xmath465 .
applying crude dimensional analysis , we now have @xmath466(age of universe ) .
this leads us to interpret eq . to mean that @xmath311 varies appreciably only over cosmological timescales , and
may be dropped in the formula of eq . .
_ tail - induced memory _ we have already observed in @xcite that the tail of the tt tensor mode solution in eq .
, for all even dimensional de sitter spacetimes higher than two , is a spacetime constant . through eq .
, this in turn leads a spacetime constant permanent fractional distortion of the space around the gw detector .
let @xmath467 be the constant hubble parameter for de sitter spacetime , so that @xmath146 = -1/(h\eta)$ ] .
then , eq . receives the following contribution over the duration @xmath468 $ ] of active gw generation : @xmath469 & = -16\pi { g_{\rm n}}\frac{h^{d-2}}{(2\pi)^{\frac{d-2}{2 } } } \frac{(d-2)!}{2^{\frac{d}{2 } } \left(\frac{d-2}{2}\right ) ! } \int_{\mathbb{r}^{d-1 } } { \text{d}}^{d-1}\vec{x } ' \int_{\eta_\text{i}}^{\eta_\text{f } } { \text{d}}\eta ' a[\eta']^{d-2 } \sigma_{ij}[\eta',\vec{x } ' ] .\end{aligned}\ ] ] this has no analog in even - dimensional @xmath48 flat spacetime because the memory effect there falls off as @xmath470 .
* gws and their memories in matter dominated universes ( @xmath370 ) * the right hand side of eq .
that governs @xmath311 when @xmath370 is likely to be dominated by the @xmath471 term far away from the gw source itself .
this is because @xmath47 is an irreducible component of the spatially - localized @xmath472 and the @xmath473 term is suppressed by @xmath474(age of universe ) . if we further invoke the conservation law in eq . to replace @xmath471 with the time derivative term involving @xmath378 on its left - hand - side by taking a spatial gradient on both sides of eq .
followed by crude dimensional analysis
we then estimate that @xmath464 scales as @xmath475 plus terms that are suppressed by the age of the universe , i.e. , terms containing @xmath449 .
this indicates @xmath476 evolves significantly only over cosmological timescales , since @xmath463 is an irreducible component of @xmath477 that can be assumed to vanish at the gw detector .
therefore , it appears likely that the only gw contribution to the fractional distortion @xmath24 $ ] arises from the tt tensor mode . _ tail - induced memory _ we have already observed in @xcite and @xcite that the tt tensor mode solution in eq .
, for @xmath478-dimensional matter dominated universes , has a tail that decays with increasing time but is otherwise space - independent .
over timescales of human gw experiments , this leads to a tail - induced memory effect very similar to the de sitter case just mentioned above , with no analog in flat spacetime . the contribution to eq .
is @xmath479 & = -4 { g_{\rm n}}\int_{\mathbb{r}^{3 } } { \text{d}}^{3}\vec{x } ' \int_{\eta_\text{i}}^{\eta_\text{f } } { \text{d}}\eta ' \frac{a[\eta']}{a[\eta ] \cdot \eta\eta ' } \sigma_{ij}[\eta',\vec{x } ' ] .\end{aligned}\ ] ] * gws and their memories for @xmath1 * we now turn to the scenario where both @xmath311 and @xmath53 admit wave solutions .
_ tails _ we begin by seeking , for all @xmath355 , whether there is a space - independent tail induced gw memory effect similar to the de sitter and matter dominated cases just discussed in cosmologies where the dominant pressure - to - energy density ratio is positive and non - trivial , i.e. , @xmath1 .
now , the tail of the tensor mode green s function is ( cf .
@xmath480 = \frac{\theta[{\bar{\sigma}}]}{2 } \left ( \frac{1}{2\pi\eta\eta ' } \right)^{\frac{d-2}{2 } } \left(\frac{\partial}{\partial s}\right)^{\frac{d-2}{2 } } \,_2f_1\left [ \frac{d-2}{q_w } , 1-\frac{d-2}{q_w};1;-\frac{s}{2}\right ] ; \end{aligned}\ ] ] and that of the @xmath311 green s function is ( cf .
@xmath481 = \frac{\theta[{\bar{\sigma}}_w]}{2 } \left ( \frac{1}{2\pi \eta\eta'}\right)^ { \frac{d-2}{2 } } & \left ( \frac{\partial}{\partial s_w } \right)^ { \frac{d-2}{2 } } \\ & \times \,_2f_1\left [ \frac{2d-5 + ( d-1)w}{q_w } , 1-\frac{2d-5 + ( d-1)w}{q_w};1;-\frac{s_w}{2}\right ] .
\nonumber\end{aligned}\ ] ] we have converted @xmath482 s to @xmath483 s , and used the hypergeometric - function equivalent expression for the legendre function , namely @xmath484=\,_2f_1[-\nu,1+\nu;1;-s/2]$ ] ( see eq .
8.702 of ) .
the reason for doing so is the following . to obtain a space - independent answer from the @xmath485 derivatives of @xmath486 in equations and , the @xmath486 needs to be a polynomial in @xmath487 ( or @xmath488 ) of degree @xmath485 . to have a trivial tail
, we would need the @xmath486 to be a polynomial of degree @xmath66 through @xmath489 .
in fact , @xmath490 $ ] is a polynomial in @xmath491 of non - negative integer degree @xmath492 whenever either @xmath493 or @xmath494 . to obtain a space - independent tail for the tensor waves , therefore , @xmath495 and to achieve the same for the scalar waves ,
@xmath496 some algebra would inform us that , no such tensor mode solution exists for all relevant even dimensions ; whereas no such scalar solution can be found for even @xmath497 .
however , in @xmath478-dimensional radiation dominated universes
i.e. , with @xmath498 cosmological scalar gw tails do not decay with distance from the astrophysical source : @xmath499 = \frac{4 { g_{\rm n}}}{\eta_0 ^ 2 a[\eta]^3 } \cdot 3^{\frac{3}{2 } } \int_{\mathbb{r}^{3 } } & { \text{d}}^{3}\vec{x } ' \int_{0}^{\eta_r-0^+ } { \text{d}}\eta ' a[\eta ' ] \\ & \times \left ( \frac{\sigma[\eta',\vec{x } ' ] - \rho[\eta',\vec{x}']}{6 } + \frac{\dot{\upsilon}[\eta',\vec{x}']}{\eta_0 a[\eta ' ] } + \frac{\vec{\nabla}^2_{\vec{x } ' } \upsilon[\eta',\vec{x}']}{3 } \right ) , \nonumber\end{aligned}\ ] ] where the retarded time is @xmath500 and @xmath501 .
( as a reminder , the scale factor of a radiation dominated 4d spatially flat universe is @xmath502 = \eta/\eta_0 $ ] . )
if the sources were actively producing @xmath311-waves over the time interval @xmath503 $ ] , the associated contribution to the memory effect in eq .
after the primary wave train has passed the observer at @xmath504 is simply @xmath505 = \frac{4 { g_{\rm n}}}{\eta_0 ^ 2 a[\eta]^3 } \cdot 3^{\frac{3}{2 } } \theta_p \int_{\mathbb{r}^{3 } } & { \text{d}}^{3}\vec{x } ' \int_{\eta_\text{i}}^{\eta_\text{f } } { \text{d}}\eta ' a[\eta ' ] \\ & \times \left ( \frac{\sigma[\eta',\vec{x } ' ] - \rho[\eta',\vec{x}']}{6 } + \frac{\dot{\upsilon}[\eta',\vec{x}']}{\eta_0 a[\eta ' ] } + \frac{\vec{\nabla}^2_{\vec{x } ' } \upsilon[\eta',\vec{x}']}{3 } \right ) .
\nonumber\end{aligned}\]]-signal to destructively interfere over the time interval @xmath503 $ ] .
such destructive interference does occur for the tensor mode in even dimensional de sitter higher than two @xcite whenever both the pressure and mass quadrupole moments of the astrophysical system take the same values at @xmath506 and at @xmath507 ; i.e. , whenever they revert to their original values.]just like the tensor gws in 4d matter dominated universes , these scalar gws would likely yield a tail induced gw memory effect that is roughly spacetime constant over timescales of human experiments . finally , we find there are no even dimensional @xmath53gw tails when @xmath508 ; in 4d this corresponds to the radiation dominated universe , where we have already noted the absence of tensor tails in @xcite . in the same vein ,
there are no @xmath311gw tails when @xmath508 for even @xmath497 .
_ acoustic cone _ for @xmath1 , i.e. , without specializing to radiation domination , the scalar wavefront in the same 4d cosmology which propagate strictly at speed @xmath388 take the form @xmath509^ { \frac{3}{2}\left ( 1+w \right ) } \phi^{\text{(direct)}}_{\text{4d},0 <
w \leq 1}[\eta,\vec{x } ] = \frac{4{g_{\rm n}}}{w } \theta_p \int_{\mathbb{r}^{3 } } { \text{d}}^{3}\vec{x } ' \frac{a[\eta_r]^ { \frac{3}{2}\left ( 1+w \right ) } } { |\vec{x}-\vec{x}'| } & \bigg ( \frac{\sigma[\eta_r,\vec{x } ' ] - 3w \rho[\eta_r,\vec{x}']}{6 } \\ & + \frac{2 \dot{\upsilon}[\eta',\vec{x}']}{(1 + 3w)\eta_r } + \frac{\vec{\nabla}^2_{\vec{x } ' } \upsilon[\eta_r,\vec{x}']}{3 } \bigg ) , \nonumber\end{aligned}\ ] ] where the retarded time is @xmath510 .
as already alluded to for the tensor gws , this portion of the scalar gws that travels strictly on the acoustic - cone should produce a non - zero memory effect whenever the matter components involved have different configurations well after and well before the duration of dominant gw production , whose peak cosmic time we will denote as @xmath83 .
the scalar gw memory effect at the detector location , after the primary gw train has gone by , contributes to eq . in the following manner : @xmath511 \approx \frac{4{g_{\rm n}}\cdot \theta_p}{a[\eta]^ { \frac{3}{2}\left ( 1+w \right ) } w |\vec{x}| } \delta \int_{\mathbb{r}^{3 } } { \text{d}}^{3}\vec{x } '
a^ { \frac{3}{2}\left ( 1+w \right ) } \left ( \frac{\sigma - 3w \rho}{6 } + \frac{2 \dot{\upsilon}}{(1 + 3w)\eta_\star } + \frac{\vec{\nabla}^2_{\vec{x } ' } \upsilon}{3 } \right ) , \end{aligned}\ ] ] where we have centered the spatial coordinate system within the central region of the astrophysical system at @xmath83 ; taken the far zone @xmath84 limit ; and defined @xmath512^ { \frac{3}{2}\left ( 1+w \right ) } \left ( \frac{\sigma[\eta_r,\vec{x } ' ] - 3w \rho[\eta_r,\vec{x}']}{6 } + \frac{2 \dot{\upsilon}[\eta_r,\vec{x}']}{(1 + 3w)\eta_\star } + \frac{\vec{\nabla}^2_{\vec{x } ' } \upsilon[\eta_r,\vec{x}']}{3 } \right ) \nonumber\\ & \qquad\qquad\qquad\qquad - \lim_{\eta_r \ll \eta_\star } \int_{\mathbb{r}^{3 } } { \text{d}}^{3}\vec{x } ' a[\eta_\star]^ { \frac{3}{2}\left ( 1+w \right ) } \left ( \frac{\sigma[\eta_r,\vec{x } ' ] - 3w \rho[\eta_r,\vec{x}']}{6 } + \frac{2 \dot{\upsilon}[\eta_r,\vec{x}']}{(1 + 3w)\eta_\star } + \frac{\vec{\nabla}^2_{\vec{x } ' } \upsilon[\eta_r,\vec{x}']}{3 } \right ) .
\nonumber\end{aligned}\ ] ] let us now identify at least two explicit ways this on - the - acoustic - cone gw memory can occur . for technical simplicity we shall focus on the local " terms in eq .
involving @xmath513 . in the @xmath514 term
; the presence of the @xmath515 in the @xmath473 term ; and the absence of any explicit @xmath0-dependence in the @xmath471 term when taken together , suggest that the gw memory effects ( due to @xmath513 ) we are about to elucidate are unlikely to be canceled if the @xmath473 and @xmath471 terms were included . ] in a 4d spatially flat flrw universe , the stress energy tensor of a point mass @xmath516 sweeping out a worldline @xmath517 $ ] , with proper time @xmath427 , is given by @xmath518 = m \int{\text{d}}\tau \dot{y}_\mu \dot{y}_\nu \delta^{(4)}[x - y[\tau ] ] , \qquad\qquad \dot{y}^\mu \equiv \frac{{\text{d}}y^\mu}{{\text{d}}\tau } ; \end{aligned}\ ] ] where we are moving indices with the flat metric , so @xmath519 . the @xmath520 ( recall equations and ) contribution to the scalar gw signal in eq .
is therefore @xmath521^ { \frac{3}{2}\left ( 1+w \right ) } \phi^{\text{(direct)}}_{\text{4d},0 < w
\leq 1}[\eta,\vec{x } ] & = \frac{4{g_{\rm n}}m}{w } \theta_p \frac{a[\eta_r]^ { \frac{1}{2}\left ( 5 + 3w \right ) } } { |\vec{x}-\vec{y}[\eta_r]| } \sqrt{1-\left({\text{d}}\vec{y}[\eta_r]/{\text{d}}\eta\right)^2 } \\ & \qquad\qquad \times \left\ { \frac{1}{6 a[\eta_r]^2 } \left ( \frac{1 - 3w}{1-\left({\text{d}}\vec{y}[\eta_r]/{\text{d}}\eta\right)^2 } \left(\frac{{\text{d}}\vec{y}[\eta_r]}{{\text{d}}\eta}\right)^2 - 3w \right ) + \dots \right\ } , \nonumber\end{aligned}\ ] ] where we have used the proper time relationship @xmath522 to replace all occurrences of @xmath523 with @xmath524 .
now consider an arbitrary ( @xmath525 ) number of compact bodies , which we will approximate as point masses , scattering off each other on unbound trajectories .
let the i - th body have mass @xmath526 and velocity @xmath527 . if we now superpose the signals in eq .
for each mass ; place the origin of the spatial coordinate system within the central region of their interaction region , so that @xmath528 in the far zone where the gw detector lies ; and also assume the scale factor @xmath529 does not change much over the duration of the experiment ; the on - the - acoustic - cone scalar gw memory contribution to eq .
resulting from such a scattering configuration would be @xmath530 & \approx \frac{2{g_{\rm n}}(1 - 3w)}{3w \cdot a[\eta_\star]|\vec{x}| } \theta_p \left(\frac{a[\eta_\star]}{a[\eta]}\right)^ { \frac{3}{2}\left ( 1+w \right ) } \\ & \qquad\qquad \times \delta \sum_{\text{i}=1}^{n } \frac{m_\text{i}}{\sqrt{1-\left({\text{d}}\vec{y}_\text{i}/{\text{d}}\eta\right)^2 } } \left(\frac{{\text{d}}\vec{y}_\text{i}}{{\text{d}}\eta}\right)^2 + \dots , \nonumber\end{aligned}\ ] ] where @xmath83 is the time of their closest approach .
by @xmath86 on the right hand side , we mean the difference between the point masses configurations in the asymptotic future and their asymptotic past , namely @xmath531 in the non - relativistic and elastic - collision limit , the sum of kinetic energies is a constant and eq .
becomes negligible .
away from this limit , though in the relativistic case or whenever the @xmath91-body scattering process involves significant gw emission and/or other non - gravitational dissipation , the scalar acoustic - cone " gw memory effect in equations and would be non - trivial .
, a wave equation for @xmath311 ( and by eq . , for @xmath271 as well ) their @xmath311 and @xmath271 were instead sourced , through poisson / elliptic - type equations ( their eqs .
( 35 ) and ( 36 ) ) , by an additional gauge - invariant scalar @xmath532 , which does obey a wave equation ( their eq .
( 45 ) ) . as such , while we were able to write down the explicit scalar gw solutions , and went on to readily identify the associated memories ; tolish and wald had to resort to indirect arguments in their appendix ( a ) .
we believe if they had access to the same wave equations for @xmath311 and @xmath271 that we did , and stepped through the arguments in their ( iv )
, their criterion for gw memory would have been found satisfied : the presence of a derivative of the dirac @xmath42-function in the linearized riemann tensor resulting from the scalar wavefront , i.e. , acoustic - cone " propagation . ] for the second example of acoustic - cone " 4d scalar gw memory , consider a gw - generation process where the astrophysical system undergoes a large change in its internal ( effective ) equation - of - state @xmath533 perhaps due to nuclear / particle physics mechanisms over the time interval @xmath534 $ ] .
( here , @xmath535 is the system s pressure density . ) for simplicity , let us assume @xmath536 .
if we again place the origin of the spatial coordinate system in the core region of the astrophysical system , and assume the observer is situated in the far zone @xmath84 , eq . informs us that eq .
will receive the contribution @xmath537 \approx \frac{2 { g_{\rm n}}}{w |\vec{x}| } \left(\frac{a[\eta_\text{f}]}{a[\eta]}\right)^ { \frac{3}{2}\left ( 1+w \right ) } \theta_p \int_{\mathbb{r}^{3 } } { \text{d}}^{3}\vec{x } ' \left ( w_a[\eta_\text{f},\vec{x } ' ] \rho[\eta_\text{f},\vec{x } ' ] - w_a[\eta_\text{i},\vec{x } ' ] \rho[\eta_\text{i},\vec{x } ' ] \dots \right ) .\end{aligned}\ ] ] we have made the approximation that the duration of the gw - process @xmath534 $ ] does not span cosmological timescales , so that @xmath538 \approx a[\eta_\text{i}-|\vec{x}|]$ ] . for estimating the size of such an effect , we record here that eq .
can be re - phrased into the heuristic statement @xmath539 where @xmath540 is the total mass of the astrophysical system , @xmath541 is its spatial distance to the observer , @xmath542 is the net change in the system s effective eos , and @xmath0 is the eos of the background cosmological fluid .
we remind the reader , the 4d null - cone @xmath53-memory was not discussed in this section , because we already did so around eq . .
* summary * we summarize the 4d wave properties of the first order gauge - invariant metric perturbation variables in the following table .
gw tensor @xmath53 null - cone and scalar @xmath543 acoustic - cone propagation will produce memory whenever the relevant components of the astrophysical source ( cf . equations and )
i.e. , @xmath293 for the tensor , and @xmath544 for the scalar do not have the same configuration in the asymptotic future as they did in the asymptotic past .
whenever there are space - independent tails , we believe it is likely there will be an associated gw memory effect , unless destructive interference occurs . we have also provided physical arguments to persuade the reader
, it is only the gauge - invariant variables admitting wave solutions that contribute to the fractional distortion @xmath453 ( cf .
measured by our gw experiment .
[ cols="<,^,^,^",options="header " , ]
in @xcite , bieri and garfinkle proposed an electric analog to the gw memory effect . here
, we will examine it in the cosmological context . according to the lorentz force law , in the geometry of eq .
, the proper spatial acceleration of an electric charge @xmath545 of mass @xmath36 is governed by @xmath546 where @xmath427 denotes proper time , and we have neglected all self - force terms .
the non - zero christoffel symbol terms of the left hand side of eq .
are @xmath547 .
we shall , however , actually neglect them in what follows because they are all proportional to @xmath449 , which means as long as our electric memory experiment does not take place over cosmological timescales , they must play a highly sub - dominant role .
the faraday tensor @xmath548}$ ] contains the electric @xmath549 and the magnetic fields @xmath550 whenever the electric term in eq
. dominates over the magnetic one as is often the case in non - relativistic systems and if we further choose the hamiltonian / temporal gauge @xcite @xmath551 then integrating both sides of eq . with respect to proper time , from some initial cosmic time @xmath23 to the present @xmath552 , gives us @xmath553_{\tau''=\tau_\text{i}}^{\tau''=\tau } = \frac{q}{m } \int_{\eta'}^{\eta } \frac{{\text{d}}\eta''}{a[\eta'']^2 } \left ( \partial_0 a_i + \dots \right ) , \end{aligned}\ ] ] where @xmath554 and @xmath427 are the proper times corresponding respectively to @xmath23 and @xmath552 . as already alluded to ,
if we watch our test charge move about in a cosmological spacetime , but only over timescales of human experiments , @xmath555 $ ] will not change very much .
we shall choose to evaluate it at @xmath552 and pull it out of the integral in eq . .
this tells us , within the hamiltonian / temporal gauge of eq .
, that the velocity kick received by our test electric charge due the passage of some electromagnetic signal , is proportional to the change in the spatial part of the vector potential : @xmath556_{\eta'}^{\eta } \approx \frac{q}{m a[\eta]^2 } \left [ a_i[\eta'',\vec{y}[\eta '' ] ] \right]_{\eta''=\eta'}^{\eta''=\eta } .\end{aligned}\ ] ] this is the bieri - garfinkle electric memory effect generalized to generic spatially flat flrw cosmologies : it is the permanent change in the proper spatial velocity of test charges , and is to be contrasted against the gravitational case involving permanent fractional displacement of pairs of test masses .
we will now proceed to calculate this electric memory generated by some externally prescribed current @xmath557 , by first solving maxwell s equations : @xmath558 we will do so within a gauge - invariant formalism , analogous to our analysis for the linearized einstein - fluid equations above . following that , we will translate our results to the hamiltonian / temporal gauge , and extract the change in the vector potential . for completeness , in appendix
, we will also solve the same equations in a generalized lorenz gauge .
* gauge - invariant formalism * the vector potential @xmath559 leaves the faraday tensor @xmath560 invariant under the gauge transformation @xmath561 for any arbitrary function @xmath562 .
let us perform a scalar - vector decomposition @xmath563 and note that under the gauge transformation in eq .
, @xmath564 while the transverse portion of the spatial vector potential is already gauge - invariant .
@xmath565 therefore , we may form the following gauge - invariant scalar @xmath566 in terms of these variables , the faraday tensor reads @xmath567 } .\end{aligned}\ ] ] the scalar - vector decomposition of the electromagnetic current is @xmath568 similar in spirit to the conservation of the matter stress tensor , the divergence - free property of this electromagnetic current , @xmath569 , tells us the electric charge density @xmath514 and the gradient portion of the spatial current are related : @xmath570 maxwell s eq . with the gauge - invariant variables employed through eq . now
take the following form .
the zeroth component is @xmath571 while the gradient and the divergence - free portions of the spatial components are , respectively , @xmath572 eq . is actually redundant ; it can be obtained by acting on both sides of eq . with @xmath573 ,
followed by utilizing eq . .
* solutions * the gauge - invariant scalar of eq .
can be solved readily using the green s function of the laplacian in @xmath112-euclidean space .
@xmath574 = a[\eta]^4 \int_{\mathbb{r}^{d-1 } } { \text{d}}^{d-1}\vec{x } ' \frac{\gamma\left [ \frac{d-3}{2 } \right ] \cdot \rho[\eta,\vec{x}']}{4\pi^{\frac{d-1}{2 } } |\vec{x}-\vec{x}'|^{d-3 } } .\end{aligned}\ ] ] the gauge - invariant vector mode of eq . can be solved by first re - expressing its wave equation as @xmath575 if we specialize to the constant@xmath19 cosmologies we have been considering in this paper , i.e. , equations and , we find this is reduced to @xmath576 the retarded solution is @xmath577^{\frac{d-4}{2 } } \alpha_i[\eta,\vec{x } ] & = \int_{\mathbb{r}^{d-1 } } { \text{d}}^{d-1}\vec{x } ' \int_{\eta_\circ}^{\eta } { \text{d}}\eta ' a[\eta']^ { \frac{d+4}{2 } } \mathcal{g}^{(\gamma)}[\eta,\eta';{\bar{\sigma } } ] \gamma_i\left [ \eta',\vec{x}'\right ] , \end{aligned}\ ] ] where the symmetric green s functions are @xmath578 & = \left(\frac{1}{2\pi } \frac{\partial}{\partial{\bar{\sigma}}}\right)^ { \frac{d-2}{2 } } \left(\frac{\theta[{\bar{\sigma}}]}{2 } p_{-\frac{d-4}{q_w}}\left [ 1+s \right]\right ) , \\ \label{maxwell_radiationg_oddd } \mathcal{g}^{(\gamma)}_{\text{odd $ d\geq 5$}}[\eta,\eta';{\bar{\sigma } } ] & = \left(\frac{1}{2\pi } \frac{\partial}{\partial{\bar{\sigma}}}\right)^ { \frac{d-3}{2 } } \left(\frac{\theta[{\bar{\sigma}}]}{4\pi } \frac { \left(s+\sqrt{s(s+2)}+1\right)^ { \frac{5-d + ( d-1)w}{q_w } } + 1 } { \sqrt{{\bar{\sigma}}(s+2 ) } \left(s+\sqrt{s(s+2)}+1\right)^ { \frac{5-d + ( d-1)w}{2q_w } } } \right ) , \\ { \bar{\sigma}}&= \frac{1}{2}\left ( ( \eta-\eta')^2 - ( \vec{x}-\vec{x}')^2 \right ) , \qquad\qquad s \equiv \frac{{\bar{\sigma}}}{\eta\eta ' } .\end{aligned}\ ] ] _ ( 3 + 1)-dimensions _
let us note that , apart from a re - scaling @xmath579 , the @xmath580 solutions are identical to the electromagnetic ones in 4d minkowski spacetime ; this is due to the conformal symmetry of the system at hand , and could already be seen by setting @xmath580 in eq . .
@xmath581 & = a[\eta]^4 \int_{\mathbb{r}^{3 } } { \text{d}}^{3}\vec{x } ' \frac{\rho[\eta,\vec{x}']}{4\pi |\vec{x}-\vec{x}'| } , \\
\alpha_{i_4}[\eta,\vec{x } ] & = \theta_p \int_{\mathbb{r}^{3 } } { \text{d}}^{3}\vec{x } ' a[\eta_r]^{4 } \frac{\gamma_i\left[\eta_r , \vec{x}'\right]}{4\pi |\vec{x}-\vec{x}'| } , \end{aligned}\ ] ] where @xmath351 is the retarded time . * cosmological electric memory * imposing the hamiltonian / temporal gauge in eq . on the scalar gauge - invariant variable in eq . , @xmath582 = -\int_{\eta'}^{\eta } \partial_i \phi[\eta'',\vec{x } ] { \text{d}}\eta '' + \partial_i \alpha[\eta',\vec{x } ]
; \end{aligned}\ ] ] which in turn means the change in the vector potential that gives rise to the electric memory effect of eq . receives contributions from the gauge - invariant vector and scalar in the following way : @xmath583 & \equiv a_i[\eta,\vec{x } ] - a_i[\eta',\vec{x } ] \nonumber\\ \label{maxwell_memory_v1 } & = \delta \alpha_i[\eta,\vec{x } ] - \int_{\eta'}^{\eta } \partial_i \phi[\eta'',\vec{x } ] { \text{d}}\eta '' , \end{aligned}\ ] ] where @xmath584 \equiv \alpha_i[\eta,\vec{x } ] - \alpha_i[\eta',\vec{x } ] .\end{aligned}\ ] ] by integrating the spatial derivative of eq . with respect to time , @xmath585 & = -\frac{1}{a[\eta]^{d-4 } } \int_{\eta'}^{\eta } a[\eta'']^d \partial_i \gamma[\eta'',\vec{x } ]
{ \text{d}}\eta '' + \left(\frac{a[\eta']}{a[\eta]}\right)^{d-4 } \partial_i \phi[\eta',\vec{x } ] .\end{aligned}\ ] ] assuming the electromagnetic current is spatially localized , so its irreducible component @xmath586 can be assumed to be zero at the test charge location , we conclude that @xmath587 $ ] far away from the source @xmath557 depends on time only through the overall prefactor @xmath354^{d-4}$ ] on the rightmost term of eq . .
referring to the @xmath311 solution in eq .
, we see that eq .
provides a coulomb - potential contribution to the electric memory formula of eq .
that scales as @xmath588(timescale of experiment)/(observer - source spatial distance ) ; whereas the @xmath589 is the radiative contribution to electric memory .
_ space - independent electromagnetic tails _ in even dimensions @xmath48 , by converting the @xmath589-tail function @xmath590 $ ] in eq . into @xmath486 s , and
following similar reasoning carried out for the gravitational case in the previous section we find that there are non - trivial space - independent electromagnetic tails for @xmath591 whenever @xmath592 explicitly , the wave tails of the gauge - invariant vector read @xmath417^{\frac{d-4}{2 } } \alpha_i^{\text{(tail)}}[\eta,\vec{x } ] & = \frac{(d-2)!}{\left(\frac{d-2}{2}\right ) ! } \int_{\mathbb{r}^{d-1 } } { \text{d}}^{d-1}\vec{x } ' \int_{\eta_\circ}^{\eta_r - 0^+ } { \text{d}}\eta ' \frac{a[\eta']^ { \frac{d+4}{2 } } } { 2(4\pi \eta\eta')^{\frac{d-2}{2 } } } \gamma_i\left [ \eta',\vec{x}'\right ] , \end{aligned}\ ] ] which indicates if @xmath534 $ ] corresponds to the active period of electromagnetic wave production
the associated electric memory registered at the observer , after the primary wave train has passed , is @xmath593 & \approx a[\eta]^{-\frac{d-4}{2 } } \frac{(d-2)!}{\left(\frac{d-2}{2}\right ) ! } \int_{\mathbb{r}^{d-1 } } { \text{d}}^{d-1}\vec{x } ' \int_{\eta_\text{i}}^{\eta_\text{f } } { \text{d}}\eta ' \frac{a[\eta']^ { \frac{d+4}{2 } } } { 2(4\pi \eta\eta')^{\frac{d-2}{2 } } } \gamma_i\left [ \eta',\vec{x}'\right ] .\end{aligned}\ ] ] _ zero - tail conditions _ it turns out there are no electromagnetic tails in even @xmath591 when @xmath594 where @xmath492 runs from @xmath66 through @xmath489 . and @xmath595 because of potential pathologies arising in first order perturbation theory when @xmath596 , @xmath597 or @xmath598 . here in this section , because we do not need to consider gravitational perturbations , we choose to remain agnostic about the underlying matter responsible for the constant@xmath19 background cosmology . ]
in the background spatially flat flrw cosmologies of dimensions @xmath48 described by equations and , we have solved the metric perturbations @xmath16 in eq . due to an isolated astrophysical system whose scalar - vector - tensor decomposition we defined in equations and .
we have done so using the gauge - invariant metric variables in equations , , and , which in turn are based on the scalar - vector - tensor decomposition of @xmath16 in eq . .
except in the de sitter case ( where @xmath128 ) , it is also necessary to simultaneously consider the perturbations of the background perfect fluid that drives the evolution of the cosmology itself ; we defined its deviation from equilibrium in eq . and
further re - phrased it in gauge - invariant terms through a scalar - vector decomposition in equations and .
the linearized einstein s equations governing the scalar perturbations can be found in equations through ; the vector ones in equations and ; and the sole tensor field in eq . .
the fluid eom involving the scalars and vectors are , respectively , equations and .
we find that , to avoid superluminality , acausal dependence on the future configuration of the gw source(s ) and potential quantum instabilities of the gauge - invariant metric scalar fields , the eos of the cosmic fluid should obey @xmath595 or @xmath128 .
( the latter case is de sitter spacetime , where there really is no background cosmological fluid . ) with this in mind , one can find the metric tensor perturbation wave - solution in eq . ; and the metric vector perturbation solution in eq .
expressed in a coulomb - type form , or in eq . as a time - integral involving a certain shear part of the matter source .
the scalar metric perturbations in de sitter spacetime yields a coulomb - type solution in eq . . in matter dominated universes , @xmath370 , they can be expressed as a double time integral involving the gw source in eq . .
finally , it is when @xmath1 that they admit wave solutions like their tensor counterpart
unlike the tensor mode whose wavefront travel at unit speed , however , these scalar ones do so at speed @xmath599 , which means cherenkov radiation is possible . in fig . , we summarize the causal structure of these gw solutions , including their potential memories . in the main text .
the gws are heard by a distant detector ( left world line ) .
the black dashed lines emanating from the worldline of the gw detector are either the past acoustic or null cones of events @xmath532 , @xmath600 and @xmath601 .
the bottom pair of light gray dashed lines emanating from the right world line is either the forward acoustic or null cone of the starting point of the gw - generation process ; the top pair is that of the ending point .
_ tensors : _ the gauge - invariant @xmath53 admits wave solutions whose wavefronts move at unit speed .
the light gray shaded region consists of @xmath53-waves propagating both on and inside the null cone ; and darker gray region is filled with only its wave tails .
there are no tensor tails in 4d radiation dominated universes ; while in ds@xmath81 as well as 4d matter dominated ones the tensor tail amplitude does not change with increasing spatial distance from the gw source .
this latter feature may lead to a novel contribution to the memory effect . _
scalars : _ within the eos interval @xmath1 , the wave solutions of the gauge - invariant scalar modes @xmath311 and @xmath271 have wavefronts propagating at speed @xmath388 . the light gray region would be filled with scalar waves traveling both on and within their acoustic cones ; and the darker gray region with only their wave tails , whose detailed properties depend in principle on the entire history of the source ( i.e. , the dashed - dotted segment ) . in a 4d radiation dominated universe ,
the scalar gw tail becomes independent of spatial coordinates ; and might thus bring about memory .
_ acoustic / null cone gws : _ we have highlighted , in equations and , that the portion of gw memory transmitted on the acoustic / null cone arises because the gw source configuration does not return to its original state after the main gw event .
specifically , if the @xmath293 part of the astrophysical system at @xmath602 is not the same as that it were at @xmath603 , the gw detector operational from @xmath532 through @xmath601 would experience a non - trivial tensor null - cone memory effect .
similarly , if @xmath604 have changed significantly from @xmath603 to @xmath602 , there will be a non - zero scalar acoustic - cone memory effect.,width=384 ] we then grouped the maxwell u@xmath605 gauge potential into gauge - invariant variables in equations and .
electromagnetism in the cosmologies of equations and is then encapsulated in equations through whereas their solutions are to be found in equations and .
we then identified a potential electric memory effect in eq .
due to the space - independent nature of the part of the gauge - invariant transverse vector that travels inside the light cone at specific values of @xmath0 given by eq . .
in this work , we have re - confirmed that , conformal re - scaling , dimension - reduction , nariai s ansatz @xcite ( and the euclidean green s function of the @xmath112-spatial laplacian ) lead us to analytic solutions to the full set of partial differential equations arising from general relativity linearized about the cosmological spacetimes of equations and for all @xmath48 .
this extends the results of @xcite , which did so for the specific case of de sitter spacetime @xmath455 . to investigate the cosmological generalization of christodoulou s nonlinear gw memory effect @xcite and to extend the de sitter quadrupole radiation formula derived by ashtekhar et al .
@xcite to a more general eos @xmath0 , we would need to go beyond the linear order in perturbation theory .
it is not clear to us that the gauge - invariant formalism used in this paper is the most computationally efficient means of doing so the gauge - fixed one of @xcite , @xcite , and @xcite might be worth re - visiting for this purpose .
moreover , while we have tried to employ physically motivated arguments to persuade the reader that the fractional distortion measured by a gw detector only receives contributions from the gauge - invariant tensor mode when @xmath128 or @xmath370 ( cf . ) ; and only from the net change in the tensor and scalar modes when @xmath1 ( cf .
in this age of precision cosmology " it may be physically important to further sharpen these arguments to either corroborate their robustness or to find cases where they fail .
finally , with the position - space green s functions of both the graviton and photon fully worked out in constant@xmath19 cosmologies in all relevant dimensions , we may now embark on a program to understand quantitatively the associated tail - induced self - force problem first uncovered by dewitt and brehme @xcite .
when @xmath1 , we have revealed through equations and that astrophysical systems could potentially lose some of their mass through scalar gravitational radiation .
( a scalar field mechanism for such mass loss has been previously noted by burko , harte and poisson @xcite . )
it would be instructive to find concrete setups where this phenomenon can be illustrated in detail .
one may also wonder if this phenomenon has any impact on the evolution of the large scale structure of our universe . finding a specific example to study its gw emission could also aid in understanding the procedure alluded to in the introduction of this paper how the matching of the cosmological waveforms in the near / intermediate zone to the far zone of the minkowskian one is actually carried out .
in particular , we note that , while the tt tensor metric perturbation is always the only gauge - invariant variable that admits wave solutions in minkowski spacetime , in @xmath606-cosmologies both the gauge - invariant scalar and tensor modes do so . how does this mismatch in the number of radiative degrees - of - freedom play into the gw matching procedure , and in turn ,
affect the physical features of the gws themselves ?
finally , in the toy cosmology model " of having a constant eos lying in the interval @xmath1 , we have identified at least two explicit ways acoustic - cone gw scalar memory can potentially occur : the kehagias - riotto / tolish - wald @xmath80-body scattering @xcite,@xcite and a major change in the internal eos of the astrophysical system .
this prompts us to wonder if similar acoustic - cone memories can indeed occur within the universe we reside in , where the scalar perturbations would obey a wave equation like that in eq . but with a sound speed @xmath607 that is no longer directly tied to @xmath0 .
while exact analytic solutions may no longer be available , jwkb methods could perhaps be exploited to obtain approximate expressions for the acoustic - cone part of the relevant position - space green s functions .
i wish to acknowledge nishant agarwal , thorsten battefeld , and audrey mithani for useful discussions on cosmological perturbation theory ; pasha bolokhov for probing questions ; as well as alexander tolish and robert wald for e - mail exchanges regarding their work in @xcite .
i am grateful to simon caron - huot for answering my question , what s the analog of a point mass in yang - mills theory ? "
( see @xcite ) ; and to gary gibbons for highlighting the important property that kerr - schild metrics linearize einstein s equations .
i have also benefited greatly from the numerous conversations i had with zvi bern and his group on the topic of gr @xmath608 ym@xmath451 , " during my week - long spring 2014 visit to ucla . during the final stages of this work
, i spent many hours at the tavern on the hill and
i wish to express my appreciation towards its excellent hospitable staff .
finally , the bulk of this work would not be possible without the software mathematica @xcite and the tensor package xact @xcite .
in this section we will review both the gauge - invariant and de donder gauge paths to obtain solutions of einstein s equations @xmath610 , with the cosmological constant set to zero , linearized about a minkowski spacetime background .
that is , we shall take the full spacetime geometry to be @xmath611 and expand the einstein to linear order in @xmath16 , namely @xmath612 , \\
\label{linearizedgravity_minkowski_einsteintensor_unbarred } ( g \vert 1)_{\mu\nu } & = \frac{1}{2 } \left ( - \partial^2 \chi_{\mu\nu } + \eta_{\mu\nu } \left ( \partial^2 \chi - \partial^\sigma \partial^\rho \chi_{\sigma\rho } \right ) + \partial_\sigma \partial _ { \ { \mu } \chi^\sigma_{\phantom{\sigma } \nu \ } } - \partial_\mu \partial_\nu \chi \right ) .\end{aligned}\ ] ] throughout this section , all indices will be moved with the flat metric @xmath613 .
if we defined the barred graviton as @xmath614 the linearized einstein tensor becomes @xmath615 .\end{aligned}\ ] ] evaluating the energy - momentum - shear - stress tensor of matter @xmath616 about the flat background , @xmath617 $ ] , we thus wish to solve @xmath618 we begin with the gauge - invariant formalism to solve the linearized einstein s eq .
.-dimensional generalization of flanagan and hughes @xcite 2.2 . ]
this will be the @xmath256 limit of what we have achieved in the main text , but we will attempt to keep the discussion fairly self - contained .
the infinitesimal coordinate transformation @xmath619 induces a change in the graviton field @xmath16 of eq . as @xmath620 * gauge - invariant variables * to obtain a gauge - invariant formulation for the dynamics of @xmath16 , we start with the following irreducible scalar - vector - tensor decomposition of the metric perturbation @xmath16 as well as an analogous one for @xmath621 .
that is , we do @xmath622 and @xmath623 where these variables obey the following constraints : @xmath624 the gauge transformation of the graviton in eq . , in terms of the variables in equations and
, reads @xmath625 while the transverse - traceless graviton is already gauge - invariant @xmath626 one may verify via a direct calculation that the following are gauge - invariant variables formed from the graviton components .
the @xmath268 scalar ones are @xmath627 whereas the vector and tensor modes are , respectively , @xmath628 ( as already explained in @xcite , there are tt tensor modes only when @xmath48 . ) we will need to perform a similar scalar - vector - tensor decomposition for the stress - energy tensor on the right hand side of eq . .
@xmath629 where these variables obey the following constraints : @xmath630 furthermore , these matter variables are not all independent because consistency of einstein s equations require the stress - energy tensor to be divergence - less , @xmath631
. a direct calculation would translate the latter to @xmath632 the interpretation is that , once @xmath514 , @xmath361 , @xmath47 and @xmath293 are specified , the @xmath378 , @xmath633 and @xmath371 can be determined by an appropriate convolution involving the euclidean green s function @xmath634 .
* linearized einstein s equations * the riemann and einstein tensor of the weak field metric in eq .
must be gauge - invariant at first order in the perturbation @xmath16 , because they are zero when @xmath635 . employing the variables of eq . in the linearized einstein tensor @xmath313 ( see eq . ) , and grouping terms in the spirit of the scalar - vector - tensor decomposition in eq .
a brute force calculation would then yield @xmath313 in terms of gauge - invariant metric perturbations defined in equations and : @xmath636 imposing einstein s equations @xmath637 , with the svt decomposition of matter in eq . , returns the equations @xmath638 because the einstein tensor itself obeys the bianchi identity
, one may expect not all these equations are independent .
we will now demonstrate that they can be reduced to the set : @xmath639 we may start by differentiating the first equation from the left in eq . once with respect to time , and employ @xmath640 from eq . to eliminate @xmath641 in eq . .
similarly , taking a time derivative of the first equation from the left in eq .
followed by replacing @xmath642 from gets rid of @xmath643 in eq . .
finally , replacing the @xmath644 in eq . with @xmath645 ( eq . ) and using the rightmost equation of eq . , yields @xmath646 , which is the time derivative of the second equation in eq . . in equations and ,
the scalar and vector perturbations obey poisson s equations ; only the tensor perturbation @xmath647 admits wave solutions : @xmath648 & = -\frac{16\pi{g_{\rm n}}}{d-2 } \int_{\mathbb{r}^{d-1 } } { \text{d}}^{d-1}\vec{x } ' \frac{\gamma\left [ \frac{d-3}{2 } \right ] \cdot \rho[t,\vec{x}']}{4 \pi^{\frac{d-1}{2 } } \vert \vec{x}-\vec{x } ' \vert^{d-3 } } , \\
\label{linearizedgravity_minkowski_einsteinsolution_vector } \overline{v}_i[t,\vec{x } ] & = -16\pi{g_{\rm n}}\int_{\mathbb{r}^{d-1 } } { \text{d}}^{d-1}\vec{x } ' \frac{\gamma\left [ \frac{d-3}{2 } \right ] \cdot \sigma_i[t,\vec{x}']}{4 \pi^{\frac{d-1}{2 } } \vert \vec{x}-\vec{x } ' \vert^{d-3 } } , \\
\label{linearizedgravity_minkowski_einsteinsolution_tensor } \overline{d}_{ij}[t,\vec{x } ] & = -16\pi{g_{\rm n}}\int_{-\infty}^{t } { \text{d}}t ' \int_{\mathbb{r}^{d-1 } } { \text{d}}^{d-1 } \vec{x } ' \overline{\mathcal{g}}_d[{\bar{\sigma } } ] \sigma_{ij}[t',\vec{x } ' ] .\end{aligned}\ ] ] where the symmetric green s functions @xmath649 in eq .
are @xmath650 & = \frac{1}{(2\pi)^{\frac{d-4}{2 } } } \left(\frac{\partial}{\partial { \bar{\sigma}}}\right)^ { \frac{d-4}{2 } } \left ( \frac{\delta[{\bar{\sigma}}]}{4\pi } \right ) , \\ \label{masslessscalarg_oddd } \overline{\mathcal{g}}_{\text{odd } d \geq 5}[{\bar{\sigma } } ] & = \frac{1}{(2\pi)^{\frac{d-3}{2 } } } \left(\frac{\partial}{\partial { \bar{\sigma}}}\right)^ { \frac{d-3}{2 } } \left ( \frac{\theta[{\bar{\sigma}}]}{2\pi \sqrt{2{\bar{\sigma } } } } \right ) , \\ { \bar{\sigma}}&\equiv \frac{1}{2 } \ { ( t - t')^2 - ( \vec{x}-\vec{x}')^2 \ } .\end{aligned}\ ] ] they obey @xmath651 = \partial_{x'}^2 \overline{\mathcal{g}}_d[{\bar{\sigma } } ] = 2 \delta^{(d)}[x - x ' ] .\end{aligned}\ ] ] _ coulomb vs. waves _ it is instructive to compare the time derivative of @xmath652 in eq . with that of @xmath647 in eq . .
for the former , let us employ @xmath640 in eq .
: @xmath653 & = -\frac{16\pi{g_{\rm n}}}{d-2 } \partial_i \int_{\mathbb{r}^{d-1 } } { \text{d}}^{d-1}\vec{x } ' \frac{\gamma\left [ \frac{d-3}{2 } \right ] \cdot \partial_{i ' } \sigma[t,\vec{x}']}{4
\pi^{\frac{d-1}{2 } } \vert \vec{x}-\vec{x } ' \vert^{d-3 } } , \end{aligned}\ ] ] where @xmath654 , @xmath655 ; we have integrated - by - parts one of the spatial derivatives and used the translation - symmetry of the @xmath656 to pull it out of the integral .
we see that , in the far zone where the observer is much farther from the source than the latter s characteristic size
each time derivative of @xmath657 scales as @xmath658(observer - source spatial distance ) .
note too that the total mass @xmath516 of the system @xmath659 \end{aligned}\ ] ] is conserved in general relativity linearized about minkowski spacetime .
this is because , if we differentiate both sides of eq .
with respect to time , and make the replacement @xmath660 ( eq . ) within the integrand on the right - hand - side , the resulting volume integral can be converted to a surface one involving the flux of the @xmath463 portion of @xmath661 , which is zero if the surface is at spatial infinity .
this in turn tells us the far zone scalar profile in eq .
, i.e. , the portion that scales as @xmath658(observer - source spatial distance)@xmath662 , is strictly time independent . for the time derivative of the wave solution in eq . ,
because they have the same form , we may refer to those in equation pairs and below to deduce , each time derivative of @xmath647 scales as @xmath658(characteristic timescale of @xmath293 ) .
as expected , in the far zone , @xmath647 varies as rapidly as the gw source itself . altogether
, these considerations suggest that the coulomb - type solution of @xmath652 is not radiative it does not carry energy - momentum to infinity but the @xmath647 wave solution in eq .
components of the graviton pseudo stress tensor . ]
as we will further argue below , this means it is the tensor mode alone that contributes appreciably to the tidal squeezing and stretching of the space around a distant gw detector . * gw detector in flat spacetime * in a background minkowski spacetime , the fractional distortion @xmath453 measured by a gw detector due to a astrophysical system located at a large distance , would need to be derived in a similar fashion as we did the cosmological equations and .
in particular , we would need the synchronous gauge metric perturbations produced by this gw source . if we are able to compute the gauge - invariant metric perturbations through equations , , , as well as @xmath663 in eq . ; then eq .
would supply the necessary result if we set @xmath256 , followed by replacing @xmath664 and all unbarred metric variables with barred ones .
stepping through similar estimates performed in the discussion right after eq .
, one would infer due to coulomb - type solutions of @xmath665 the integral terms involving @xmath666 and @xmath667 would very likely scale respectively as ( timescale of gw experiment)/(observer - source spatial distance ) and ( ( timescale of gw experiment)/(observer - source spatial distance))@xmath451 relative to the @xmath668 term .
furthermore , from the _ coulomb vs. waves _ " discussion above , one would expect @xmath668 itself to be highly sub - dominant to @xmath669 because the characteristic timescale governing changes in amplitude is the observer - source spatial distance for @xmath652 while it is the gw source(s) intrinsic frequencies for @xmath647 . to sum : the fractional distortion measured by a gw detector in a flat spacetime is similar to its counterpart ( eq . ) in a matter dominated or de sitter cosmology it receives contributions primarily from the tensor mode alone .
@xmath670 = - \frac{\widehat{n}^i \widehat{n}^j}{2 } \int_{0}^{1 } \delta \overline{d}_{ij}[\vec{y}_0 + \lambda(\vec{z}_0-\vec{y}_0 ) ] { \text{d}}\lambda \end{aligned}\ ] ] we now turn to solving eq . by imposing the de donder gauge @xmath671 (
this is analogous to imposing the lorenz gauge condition @xmath672 in electromagnetism , where @xmath673 is the @xmath674 gauge potential . )
eq . applied to eq . will reduce the linearized einstein s eq . to @xmath675
this converts the problem at hand to a @xmath676 matrix of minimally coupled massless scalar wave equations .
the retarded solutions are @xmath677 = -16 \pi { g_{\rm n}}\int_{-\infty}^t { \text{d}}t ' \int_{\mathbb{r}^{d-1 } } { \text{d}}^{d-1}\vec{x } ' \overline{\mathcal{g}}_d[{\bar{\sigma } } ] \overline{t}_{\mu\nu}[t',\vec{x } ' ] , \end{aligned}\ ] ] where the symmetric green s functions @xmath678\}$ ] can be found in equations and .
* linard wiechert graviton fields * as a simple application of these formulas , we shall work out the ( linard wiechert ) graviton field sourced by a point mass in motion in all dimensions higher than @xmath18 .
we will witness that the linear graviton field of a point mass propagates strictly on the null cone in even dimensions @xmath48 and exclusively within the null cone ( i.e. , it is pure tail ) in odd dimensions @xmath679 . following that , we generalize to all even dimensions @xmath48 the known 4-dimensional gw memory engendered by 2 ( or more ) compact bodies scattering off each other due to their mutual gravity .
the stress energy tensor of a point mass @xmath516 , with timelike trajectory @xmath680 $ ] and proper time @xmath427 , moving about in some geometry @xmath99 is given by @xmath681 = m \int_{-\infty}^{+\infty } { \text{d}}\tau \dot{y}^\mu[\tau ] \dot{y}^\nu[\tau ] \frac{\delta^{(d)}[x - y[\tau]]}{\sqrt{|g[x]| } } , \qquad\qquad \dot{y}^\mu \equiv \frac{{\text{d}}y^\mu[\tau]}{{\text{d}}\tau } .\end{aligned}\ ] ] in relation to coordinate time @xmath682 through the differential equation @xmath683 ; in particular , @xmath684.]to obtain the linear linard
wiechert graviton field @xmath685 in minkowski spacetime , we first set @xmath686 , so that @xmath687|}=1 $ ] . after integrating over spacetime @xmath688 in eq .
, there is a sole remaining integral , with respect to @xmath427 .
we then notice that all the @xmath482 derivatives in equations and can be expressed as @xmath689 \cdot ( y[\tau]-x ) .\end{aligned}\ ] ] ( the @xmath690 " is the minkowski dot product ; for example , @xmath691 . ) at this juncture , @xmath692 & = -16 \pi { g_{\rm n}}m \int_{t'=-\infty}^{t'=t } { \text{d}}\tau \frac{\dot{y}^\mu[\tau ] \dot{y}^\nu[\tau]}{(2\pi)^{\frac{d-4}{2 } } } \left(\frac{1}{\dot{{\bar{\sigma } } } } \frac{{\text{d}}}{{\text{d}}\tau}\right)^ { \frac{d-4}{2 } } \left ( \frac{\delta[{\bar{\sigma}}]}{4\pi } \right ) , \\
\label{linearizedgravity_minkowski_einsteinequation_dedonderlw_oddd_1 } \text{odd $ d \geq 5 $ : } \overline{\chi}^{\mu\nu}[t,\vec{x } ] & = -16 \pi { g_{\rm n}}m \int_{t'=-\infty}^{t'=t } { \text{d}}\tau \frac{\dot{y}^\mu[\tau ] \dot{y}^\nu[\tau]}{(2\pi)^{\frac{d-3}{2 } } } \left(\frac{1}{\dot{{\bar{\sigma } } } } \frac{{\text{d}}}{{\text{d}}\tau}\right)^ { \frac{d-3}{2 } } \left ( \frac{\theta[{\bar{\sigma}}]}{2\pi \sqrt{2{\bar{\sigma } } } } \right ) .\end{aligned}\ ] ] to compute the energy - momentum - stress - shear of these gws , we would need their gradients .
@xmath693 & = -16 \pi { g_{\rm n}}m \int_{t'=-\infty}^{t'=t } { \text{d}}\tau \frac{(x - y)_\alpha \dot{y}^\mu[\tau ] \dot{y}^\nu[\tau]}{(2\pi)^{\frac{d-4}{2 } } } \left(\frac{1}{\dot{{\bar{\sigma } } } } \frac{{\text{d}}}{{\text{d}}\tau}\right)^{\frac{d-2}{2 } } \left ( \frac{\delta[{\bar{\sigma}}]}{4\pi } \right ) , \\ \label{linearizedgravity_minkowski_einsteinequation_dedonderlw_oddd_1grad } \text{odd $ d \geq 5 $ : } \partial_\alpha \overline{\chi}^{\mu\nu}[t,\vec{x } ] & = -16 \pi { g_{\rm n}}m \int_{t'=-\infty}^{t'=t } { \text{d}}\tau \frac{(x - y)_\alpha \dot{y}^\mu[\tau ] \dot{y}^\nu[\tau]}{(2\pi)^{\frac{d-3}{2 } } } \left(\frac{1}{\dot{{\bar{\sigma } } } } \frac{{\text{d}}}{{\text{d}}\tau}\right)^{\frac{d-1}{2 } } \left ( \frac{\theta[{\bar{\sigma}}]}{2\pi \sqrt{2{\bar{\sigma } } } } \right ) .\end{aligned}\ ] ] note that there is no need to differentiate the upper limit @xmath694 because the integrand evaluated at the upper limit lies outside the light cone , i.e. , @xmath695 , and is hence set to zero by the @xmath696 $ ] , @xmath697 $ ] and the latter s derivatives . beginning with the even dimensional case in equations and
, we shall integrate by parts ( ibps ) the @xmath698 s within the integrands , to act them on the trajectories @xmath699 without incurring any boundary terms .
for , suppose this can be done up to the @xmath700th ibp , then the @xmath80th ibp hands us , for even @xmath591 and @xmath701 , @xmath702 & = \left[16 \pi { g_{\rm n}}m ( -)^n \left ( \left(\frac{1}{\dot{{\bar{\sigma } } } } \frac{{\text{d}}}{{\text{d}}\tau}\right)^ { n-1 } \frac{\dot{y}^\mu[\tau ] \dot{y}^\nu[\tau]}{(2\pi)^{\frac{d-4}{2 } } \dot{{\bar{\sigma}}}}\right ) \left(\frac{1}{\dot{{\bar{\sigma } } } } \frac{{\text{d}}}{{\text{d}}\tau}\right)^ { \frac{d-4}{2 } - n } \left ( \frac{\delta[{\bar{\sigma}}]}{4\pi } \right ) \right]_{t'=\tau=-\infty}^{t'=t } \\ & - 16 \pi { g_{\rm n}}m ( -)^n \int_{t'=-\infty}^{t'=t } { \text{d}}\tau \frac{{\text{d}}}{{\text{d}}\tau } \left\ { \left(\frac{1}{\dot{{\bar{\sigma } } } } \frac{{\text{d}}}{{\text{d}}\tau}\right)^ { n-1 } \left(\frac{\dot{y}^\mu[\tau ] \dot{y}^\nu[\tau]}{(2\pi)^{\frac{d-4}{2 } } \dot{{\bar{\sigma } } } } \right ) \right\ } \left(\frac{1}{\dot{{\bar{\sigma } } } } \frac{{\text{d}}}{{\text{d}}\tau}\right)^ { \frac{d-4}{2 } - n } \left ( \frac{\delta[\tau - \tau_{\text{ret}}]}{4\pi |\dot{{\bar{\sigma}}}[\tau_{\text{ret}}]| } \right ) ; \nonumber\end{aligned}\ ] ] and for @xmath703 @xmath704 = \left [ 16 \pi { g_{\rm n}}m ( -)^n \left ( \left(\frac{1}{\dot{{\bar{\sigma } } } } \frac{{\text{d}}}{{\text{d}}\tau}\right)^{n-1 } \frac{(x - y)_\alpha \dot{y}^\mu[\tau ] \dot{y}^\nu[\tau]}{(2\pi)^{\frac{d-4}{2 } } \dot{{\bar{\sigma } } } } \right ) \left ( \frac{1}{\dot{{\bar{\sigma } } } } \frac{{\text{d}}}{{\text{d}}\tau } \right)^{\frac{d-2}{2 } - n } \left ( \frac{\delta[{\bar{\sigma}}]}{4\pi } \right ) \right]_{\tau=-\infty}^{t'=t } \nonumber\\ & - 16 \pi { g_{\rm n}}m ( -)^n \int_{t'=-\infty}^{t'=t } { \text{d}}\tau \frac{{\text{d}}}{{\text{d}}\tau } \left\ { \left(\frac{1}{\dot{{\bar{\sigma } } } } \frac{{\text{d}}}{{\text{d}}\tau}\right)^{n-1 } \frac{(x - y)_\alpha \dot{y}^\mu[\tau ] \dot{y}^\nu[\tau]}{(2\pi)^{\frac{d-4}{2 } } \dot{{\bar{\sigma } } } } \right\ } \left(\frac{1}{\dot{{\bar{\sigma } } } } \frac{{\text{d}}}{{\text{d}}\tau}\right)^{\frac{d-2}{2 } - n } \left ( \frac{\delta[\tau-\tau_{\text{ret}}]}{4\pi |\dot{{\bar{\sigma}}}[\tau_\text{ret}]| } \right )
. \end{aligned}\ ] ] the upper limit of the boundary term in equations and , as already noted , lies outside the light cone and can be set to zero immediately .
the lower limit @xmath705 amounts to taking @xmath706 the trajectory @xmath680 $ ] is timelike where again @xmath697 $ ] and its derivatives are zero .
note , we have also converted @xmath697 = \delta[\tau - \tau_{\text{ret}}]/|\dot{{\bar{\sigma}}}[\tau_{\text{ret}}]|$ ] within the remaining integral of eq . , where the retarded proper time @xmath707 is the value of @xmath427 in the past that would set @xmath708 .
this change - of - variables would allow us to integrate over the @xmath709 $ ] once all the @xmath427-derivatives have been ibps onto the @xmath699 s . moving on to perform similar ibps for the odd dimensional @xmath679 case ,
if we assume all boundary terms may be discarded up to the @xmath700th ibp , the @xmath80th ibp , for @xmath710 , would yield : @xmath711 & = \left [ 16 \pi { g_{\rm n}}m ( -)^n \left ( \left(\frac{1}{\dot{{\bar{\sigma } } } } \frac{{\text{d}}}{{\text{d}}\tau}\right)^ { n-1 } \frac{\dot{y}^\mu[\tau ] \dot{y}^\nu[\tau]}{(2\pi)^{\frac{d-3}{2 } } \dot{{\bar{\sigma}}}}\right ) \left(\frac{1}{\dot{{\bar{\sigma } } } } \frac{{\text{d}}}{{\text{d}}\tau}\right)^ { \frac{d-3}{2 } - n } \left ( \frac{\theta[{\bar{\sigma}}]}{2\pi \sqrt{2{\bar{\sigma } } } } \right ) \right]_{t'=\tau=-\infty}^{t'=t } \\ & - 16 \pi { g_{\rm n}}m ( -)^n \int_{t'=-\infty}^{t'=t } { \text{d}}\tau \frac{{\text{d}}}{{\text{d}}\tau } \left\ { \left(\frac{1}{\dot{{\bar{\sigma } } } } \frac{{\text{d}}}{{\text{d}}\tau}\right)^ { n-1 } \frac{\dot{y}^\mu[\tau ] \dot{y}^\nu[\tau]}{(2\pi)^{\frac{d-3}{2 } } \dot{{\bar{\sigma } } } } \right\ } \left(\frac{1}{\dot{{\bar{\sigma } } } } \frac{{\text{d}}}{{\text{d}}\tau}\right)^ { \frac{d-3}{2 } - n } \left ( \frac{\theta[{\bar{\sigma}}]}{2\pi \sqrt{2{\bar{\sigma } } } } \right ) \nonumber ; \end{aligned}\ ] ] and for @xmath712 , @xmath713 = \left [ 16 \pi { g_{\rm n}}m ( -)^n \left(\left(\frac{1}{\dot{{\bar{\sigma } } } } \frac{{\text{d}}}{{\text{d}}\tau}\right)^{n-1 } \frac{(x - y)_\alpha \dot{y}^\mu[\tau ] \dot{y}^\nu[\tau]}{(2\pi)^{\frac{d-3}{2 } } \dot{{\bar{\sigma}}}}\right ) \left(\frac{1}{\dot{{\bar{\sigma}}}}\frac{{\text{d}}}{{\text{d}}\tau}\right)^{\frac{d-1}{2}-n } \left ( \frac{\theta[{\bar{\sigma}}]}{2\pi \sqrt{2{\bar{\sigma } } } } \right ) \right]_{t'=\tau=-\infty}^{t'=t } \nonumber\\ & - 16 \pi { g_{\rm n}}m ( -)^n \int_{t'=-\infty}^{t'=t } { \text{d}}\tau \frac{{\text{d}}}{{\text{d}}\tau } \left\ { \left(\frac{1}{\dot{{\bar{\sigma } } } } \frac{{\text{d}}}{{\text{d}}\tau}\right)^{n-1 } \frac{(x - y)_\alpha \dot{y}^\mu[\tau ] \dot{y}^\nu[\tau]}{(2\pi)^{\frac{d-3}{2 } } \dot{{\bar{\sigma } } } } \right\ } \left(\frac{1}{\dot{{\bar{\sigma } } } } \frac{{\text{d}}}{{\text{d}}\tau}\right)^{\frac{d-1}{2}-n } \left ( \frac{\theta[{\bar{\sigma}}]}{2\pi \sqrt{2{\bar{\sigma } } } } \right )
. \end{aligned}\ ] ] since the upper limit gives us a strictly negative @xmath714 , it is again set to zero by the presence of @xmath696 $ ] , @xmath697 $ ] and the latter s derivatives . and
since the lower limit corresponds to @xmath706 , the only boundary term that requires analysis is the one that does not contain @xmath42-functions , namely @xmath715 \left(\left(\frac{1}{\dot{{\bar{\sigma } } } } \frac{{\text{d}}}{{\text{d}}\tau}\right)^ { n-1 } \frac{\dot{y}^\mu[\tau ] \dot{y}^\nu[\tau]}{(2\pi)^{\frac{d-3}{2 } } \dot{{\bar{\sigma}}}}\right ) \left(\frac{{\text{d}}}{{\text{d}}{\bar{\sigma}}}\right)^ { \frac{d-3}{2 }
- n } \left ( \frac{1}{2\pi \sqrt{2{\bar{\sigma } } } } \right ) \right]_{t'=\tau=-\infty } \end{aligned}\ ] ] and @xmath716 \left(\left(\frac{1}{\dot{{\bar{\sigma } } } } \frac{{\text{d}}}{{\text{d}}\tau}\right)^ { n-1 } \frac{(x - y)_\alpha \dot{y}^\mu[\tau ] \dot{y}^\nu[\tau]}{(2\pi)^{\frac{d-3}{2 } } \dot{{\bar{\sigma}}}}\right ) \left(\frac{{\text{d}}}{{\text{d}}{\bar{\sigma}}}\right)^ { \frac{d-1}{2}-n } \left ( \frac{1}{2\pi \sqrt{2{\bar{\sigma } } } } \right ) \right]_{t'=\tau=-\infty } .\end{aligned}\ ] ] since the @xmath717 and @xmath718 terms do tend to zero in this asymptotic past limit
if we assume all components of the time derivatives @xmath719 remain bounded for the entire trajectory
what needs to be examined are the remaining proper time derivatives , which take the generic forms @xmath720 \equiv \frac{1}{\dot{{\bar{\sigma}}}^{a_m } } \frac{{\text{d}}}{{\text{d}}\tau } \left ( \dots \left ( \frac{1}{\dot{{\bar{\sigma}}}^{a_2 } } \frac{{\text{d}}}{{\text{d}}\tau } \left ( \frac{1}{\dot{{\bar{\sigma}}}^{a_1 } } \frac{{\text{d}}}{{\text{d}}\tau } \left(\frac{1}{\dot{{\bar{\sigma}}}^{b}}\right ) \right ) \right ) \right ) \end{aligned}\ ] ] and @xmath721 \equiv \frac{1}{\dot{{\bar{\sigma}}}^{a_m } } \frac{{\text{d}}}{{\text{d}}\tau } \left ( \dots \left ( \frac{1}{\dot{{\bar{\sigma}}}^{a_2 } } \frac{{\text{d}}}{{\text{d}}\tau } \left ( \frac{1}{\dot{{\bar{\sigma}}}^{a_1 } } \frac{{\text{d}}}{{\text{d}}\tau } \left(\frac{(x - y[\tau])_\alpha}{\dot{{\bar{\sigma}}}^{b}}\right ) \right ) \right ) \right ) , \end{aligned}\ ] ] for positive integers @xmath722 . now , by mathematical induction on @xmath36 , one may show that @xmath723 = \sum_{\ell = 1+\sigma_m}^{m + \sigma_m } \frac{c_\ell^{(m)}}{\dot{{\bar{\sigma}}}^{\ell+b } } , \qquad\qquad \sigma_m \equiv \sum_{i=1}^m a_i , \end{aligned}\ ] ] where @xmath724 is built out of sums and/or products of the scalars @xmath725 and @xmath726 .
the infinite past behavior of @xmath727 in eq .
is therefore encoded in that of the terms in eq . .
by dimension analysis , because @xmath714 itself scales as ( length)@xmath728 , we must have @xmath729 \bigg ] = \frac{\tau^{\sigma_m-(m - b)}}{l^{2(\sigma_m+b ) } } .\end{aligned}\ ] ] the @xmath492th term in the sum in eq .
, on the other hand , goes as @xmath730 \tau^{\ell+b}/l^{2(\ell+b)}$ ] , which when set equal to the right - hand - side of eq . , implies the numerator itself scales as @xmath731 \sim \left(\frac{{\text{d}}}{{\text{d}}\tau}\right)^{\ell - \sigma_m + m } [ ( \text{length})^{2(\ell - \sigma_m ) } ] .\end{aligned}\ ] ]
let us then exploit the fact that @xmath732 is a timelike vector to deduce that @xmath733 corresponds to , in the instantaneous rest frame of the source , the ( negative ) time elapsed between emission and receipt , namely @xmath734 ; in particular , @xmath735 as @xmath705 .
likewise , @xmath736 , for any @xmath97-vector @xmath737 , must be equal to the rest frame elapsed time multiplied by @xmath738 . because we wish to study the asymptotic behavior of @xmath739 in eq .
, we therefore require the most dominant terms in @xmath724 , since we already know that the denominator goes as @xmath740 .
based on the dimensional analysis above , the largest term must come from the highest power of @xmath732 .
@xmath741 we have deduced that the dominant behavior of @xmath727 in eq .
is given by any one of the terms in eq .
, because each and every term there scales the same way ; it is : @xmath742 by assumption , @xmath743 is strictly positive , so the @xmath705 limit of these nested derivatives of @xmath744 goes to zero .
all boundary terms in eq .
are therefore zero .
a similar line of reasoning should lead one to deduce ( recall eq . )
@xmath745 = \lim_{\tau \to -\infty } & \frac{1}{\dot{{\bar{\sigma}}}^{a_m } } \frac{{\text{d}}}{{\text{d}}\tau } \left ( \dots \left ( \frac{1}{\dot{{\bar{\sigma}}}^{a_2 } } \frac{{\text{d}}}{{\text{d}}\tau } \left ( \frac{1}{\dot{{\bar{\sigma}}}^{a_1 } } \frac{{\text{d}}}{{\text{d}}\tau } \left(\frac{(x - y[\tau])_\alpha}{\dot{{\bar{\sigma}}}^{b}}\right ) \right ) \right ) \right ) \nonumber\\ & \sim \frac{1}{(\text{instantaneous - rest - frame elapsed time})^{\sigma_m + b - 1 } } .\end{aligned}\ ] ] by assumption , @xmath746 is strictly non - negative , so the @xmath705 limit of these nested derivatives of @xmath744 does not blow up .
all boundary terms in eq .
are therefore zero .
_ results _ we have arrived at the linear linard wiechert graviton field of a point mass @xmath516 moving about in a background minkowski spacetime ( i.e. , eq . with @xmath747 ) . for even spacetime dimensions @xmath48 , @xmath748 & = \left .
4 { g_{\rm n}}m ( -)^ { \frac{d-4}{2 } } \left ( \left ( \frac{1}{\dot{{\bar{\sigma } } } } \frac{{\text{d}}}{{\text{d}}\tau } \right)^ { \frac{d-4}{2 } } \frac{\dot{y}^\mu[\tau ] \dot{y}^\nu[\tau]}{(2\pi)^{\frac{d-4}{2 } } \dot{{\bar{\sigma } } } } \right ) \right\vert_{\tau=\tau_{\text{ret } } } , \\ \label{linearizedgravity_minkowski_einsteinequation_dedonderlw_evend_grad } \partial_\alpha \overline{\chi}^{\mu\nu}[t,\vec{x } ] & = \left . 4 { g_{\rm n}}m ( -)^ { \frac{d-2}{2 } } \left ( \left ( \frac{1}{\dot{{\bar{\sigma } } } } \frac{{\text{d}}}{{\text{d}}\tau } \right)^ { \frac{d-2}{2 } } \frac{(x - y[\tau])_\alpha \dot{y}^\mu[\tau ] \dot{y}^\nu[\tau]}{(2\pi)^{\frac{d-4}{2 } } \dot{{\bar{\sigma } } } } \right ) \right\vert_{\tau=\tau_{\text{ret } } } ; \end{aligned}\ ] ] and for odd spacetime dimensions @xmath679 , @xmath749 & = 8 { g_{\rm n}}m ( -)^ { \frac{d-5}{2 } } \int_{-\infty}^{\tau_{\text{ret}}-0^+ } \frac{{\text{d}}\tau}{\sqrt{2 { \bar{\sigma } } } } \frac{{\text{d}}}{{\text{d}}\tau } \left ( \left ( \frac{1}{\dot{{\bar{\sigma } } } } \frac{{\text{d}}}{{\text{d}}\tau } \right)^ { \frac{d-5}{2 } } \frac{\dot{y}^\mu[\tau ] \dot{y}^\nu[\tau]}{(2\pi)^{\frac{d-3}{2 } } \dot{{\bar{\sigma } } } } \right ) , \\ \label{linearizedgravity_minkowski_einsteinequation_dedonderlw_oddd_grad } \partial_\alpha \overline{\chi}^{\mu\nu}[t,\vec{x } ] & = 8 { g_{\rm n}}m ( -)^ { \frac{d-3}{2 } } \int_{-\infty}^{\tau_{\text{ret}}-0^+ } \frac{{\text{d}}\tau}{\sqrt{2 { \bar{\sigma } } } } \frac{{\text{d}}}{{\text{d}}\tau } \left ( \left ( \frac{1}{\dot{{\bar{\sigma } } } } \frac{{\text{d}}}{{\text{d}}\tau } \right)^ { \frac{d-3}{2 } } \frac{(x - y[\tau])_\alpha \dot{y}^\mu[\tau ] \dot{y}^\nu[\tau]}{(2\pi)^{\frac{d-3}{2 } } \dot{{\bar{\sigma } } } } \right ) ; \end{aligned}\ ] ] where we record that @xmath750/{\text{d}}\tau$ ] and @xmath751 & = \left(\frac{{\text{d}}\tau_{\text{ret}}}{{\text{d}}t}\right)^{-1 } \frac{{\text{d}}y}{{\text{d}}t } \cdot \left ( y - x \right ) \\ & = -\frac{1 - \vec{v } \cdot \widehat{n}}{\sqrt{1-({\text{d}}\vec{y}/{\text{d}}t)^2 } } |\vec{x}-\vec{y}| , \qquad\qquad \widehat{n } \equiv \frac{\vec{x}-\vec{y}}{| \vec{x}-\vec{y}| } , \qquad\qquad \vec{v } \equiv \frac{{\text{d}}\vec{y}}{{\text{d}}t } . \nonumber\end{aligned}\ ] ] we also remind the reader that the retarded proper time @xmath707 is defined as the solution to @xmath752 = t - y^0[\tau_{\text{ret } } ] = \left\vert \vec{x } - \vec{y}[\tau_{\text{ret } } ] \right\vert .\end{aligned}\ ] ] as already advertised , in even dimensions @xmath48 , the graviton field of eq . and its gradient in eq .
are the gravitational signals coming from the intersection between the past light cone of the observer at @xmath753 and the worldline of the source , i.e. , from the point at @xmath754 $ ] . while in odd dimensions @xmath679 , the graviton field of eq . and its gradient in eq .
are the cumulative gravitational signals arising from the entire past trajectory of the source @xmath755 $ ] lying strictly within the past light cone of the observer at @xmath39 .
that this odd @xmath679 case yields pure tail signals is worth reiterating because the corresponding symmetric green s function of eq .
does contain @xmath697 $ ] and its derivatives , and thus one might otherwise had expected some portion of the signal to travel on the null cone .
* newtonian potential of point mass * as a check of our results here , let us recover the expected @xmath756 newtonian potential of a point mass at rest
. in this static case , proper time is coordinate time and from eq .
: @xmath757 moreover , from eq .
, the retarded time is @xmath758 , which tells us @xmath759=-|\vec{x}-\vec{y}| .\end{aligned}\ ] ] with these results , we may now perform a change - of - variables @xmath760 , to convert equations and to @xmath761 & = - \frac{4 { g_{\rm n}}m}{|\vec{x}-\vec{y}|^{d-3 } } \delta^\mu_0 \delta^\nu_0 \frac{(d-5)!!}{(2\pi)^{\frac{d-4}{2 } } } , \\
\label{linearizedgravity_minkowski_einsteinequation_dedonderstaticlw_oddd_2 } \text{odd $ d \geq 5 $ : } \overline{\chi}^{\mu\nu}[t,\vec{x } ] & = - \frac{8 { g_{\rm n}}m}{|\vec{x}-\vec{y}|^{d-3 } } \delta^\mu_0 \delta^\nu_0 \frac{(d-4)!!}{(2\pi)^{\frac{d-3}{2 } } } \int_{-\infty}^{-1 } \frac{{\text{d}}\lambda}{\sqrt{\lambda^2 - 1 } \lambda^{d-3 } } ; \end{aligned}\ ] ] for odd integer @xmath80 , @xmath762 ; and @xmath763 .
to arrive at equations and , we have made use of the following identity , valid for arbitrary positive integer @xmath80 : @xmath764 the integral in eq .
can be tackled by changing variables @xmath765 and using the integral @xmath766 \gamma[\beta]}{\gamma[\alpha+\beta ] } , \qquad\qquad \text{re}[\alpha ] , \
\text{re}[\beta]>0.\end{aligned}\ ] ] here , the @xmath767 $ ] s are gamma functions . after employing relevant @xmath768-function identities
, the final result takes the same expression for both even and odd @xmath48 : @xmath769 = 16\pi { g_{\rm n}}m \delta^\mu_0 \delta^\nu_0 \mathcal{g}^{(\text{e})}_{d-1}[\vec{x}-\vec{y}],\end{aligned}\ ] ] where @xmath770 is the green s function of the euclidean laplacian in @xmath112-spatial dimensions , @xmath771 = \left(\frac{1}{\vec{\nabla}^2 } \right)[\vec{x}-\vec{y } ] = -\frac{\gamma\left [ \frac{d-3}{2 } \right]}{4 \pi^{\frac{d-1}{2 } } \vert \vec{x}-\vec{y } \vert^{d-3 } } .\end{aligned}\ ] ] a note about causal structure . even though the static nature of the newtonian potential guarantees it takes the same form for all relevant dimensions
i.e. , we could have obtained the same answer through gauss law applied to the linearized einstein s equation the integral in eq .
is a reminder that , because of the pure tail nature of the gravitational signal in odd dimensions @xmath772 , strictly speaking , newtonian s law of gravitation requires the point mass to be still for its entire past history . in even dimensions , on the other hand , the static potential of eq .
only requires the point mass to be still in the neighborhood of the retarded time @xmath773 .
* graviton fields of continuous sources * before moving on to the subject of gw memory , let us record here that , the graviton field generated by an isolated but continuous ( as opposed to the discrete point mass ) source in eq . can be derived in a manner similar to that of the linard
wiechert fields above .
@xmath774}{t - t ' } \right)\right\vert_{t - t'=|\vec{x}-\vec{x}'| } \\ \label{linearizedgravity_minkowski_einsteinequation_dedondercontinuoussource_oddd } \text{odd $ d \geq 5 $ : } \overline{\chi}_{\mu\nu } & = -8{g_{\rm n}}\int_{\mathbb{r}^{d-1 } } { \text{d}}^{d-1}\vec{x } ' \int_{-\infty}^{t-|\vec{x}-\vec{x}'|-0^+ } \frac{{\text{d}}t'}{\sqrt{2{\bar{\sigma } } } } \frac{\partial}{\partial t ' } \left\ { \left ( \frac{1}{2\pi(t - t ' ) } \frac{\partial}{\partial t ' } \right)^{\frac{d-5}{2 } } \left ( \frac{\overline{t}_{\mu\nu}[t',\vec{x}']}{2\pi(t - t ' ) } \right ) \right\ } \end{aligned}\ ] ] the gradients are , for even @xmath48 : @xmath775}{t - t ' } \right)\right\ } \right\vert_{t - t'=|\vec{x}-\vec{x}'| } ; \end{aligned}\ ] ] while for odd @xmath679 ,
@xmath776}{2\pi(t - t ' ) } \right ) \right\ } .\end{aligned}\ ] ] * gw memory from unbound orbits in even dimensions
@xmath48 * we now consider an arbitrary number ( @xmath525 ) of compact bodies , which we will model as point masses , gravitationally scattering off each other on unbound trajectories .
specifically , in the asymptotic past , they have some ( incoming ) constant velocities @xmath777 . denoting the spacetime trajectory of the @xmath778-th compact object as @xmath779
, we have @xmath780}{{\text{d}}t } \equiv v_{(\text{i } \vert 0)}^\mu = ( 1,\vec{v}_{(\text{i } \vert 0 ) } ) \qquad\qquad \text{(constant)}.\end{aligned}\ ] ] these @xmath91 bodies interact gravitationally , and then scatter off to infinity , such that in the asymptotic future they again approach some ( outgoing ) constant velocities @xmath781 , namely @xmath782}{{\text{d}}t } \equiv v_{(\text{i } \vert 1)}^\mu = ( 1,\vec{v}_{(\text{i } \vert 1 ) } ) \qquad\qquad \text{(constant)}.\end{aligned}\ ] ] note that , when @xmath783 approaches a constant , so does @xmath784 ; in this limit , taking a proper time derivative of eq . returns @xmath785 which in turn allows us to replace the @xmath786 in eq . with @xmath787 . employing eq . , we may invoke superposition to obtain the fields generated by the asymptotic trajectories of these @xmath91 bodies : @xmath788 & = \frac{(d-5)!!}{(2\pi)^{\frac{d-4}{2 } } } \sum_{\text{i}=1}^{n } \left .
4 { g_{\rm n}}m_{\text{i } } \frac{\dot{y}_{\text{i}}^\mu \dot{y}_{\text{i}}^\nu}{\dot{{\bar{\sigma}}}^{d-3 } } \right\vert_{\tau_{\text{i}}=\tau_{(\text{i}\vert\text{ret } ) } } \end{aligned}\ ] ] at this point , we have generalized the gw memory effect to all even dimensions @xmath48 . for simplicity we shall define @xmath789 to lie within the central region of their closest approach , and proceed to take the far zone limit @xmath790 while employing eq .
to convert proper - time derivatives to coordinate - time ones : @xmath791 where the contribution to the linear graviton field from the worldlines of the point masses @xmath792 in the asymptotic past is @xmath793 and that from the asymptotic future is @xmath794 by referring to eq . , the contribution to the gw memory effect in even dimensions @xmath48 is the tt part of eq . .
in section we solved maxwell s equations in the spatially flat flrw cosmology of equations and , using a gauge - invariant formalism . in this section
we will solve maxwell s and the linearized yang - mills(-wong s ) equations in the same geometry , but by fixing a gauge .
we will continue to move indices with the flat metric .
in particular , the zeroth component of eq . can be expressed as @xmath795 and the spatial ones as @xmath796 in minkowski spacetime , we set @xmath797 and @xmath798 ; choosing the lorenz gauge @xmath799 maxwell s equations and reduce to @xmath800 we may employ the green s functions in equations and to write down the retarded solutions : @xmath801 = \int_{-\infty}^{t } { \text{d}}t ' \int_{\mathbb{r}^{d-1 } } { \text{d}}^{d-1 } \vec{x } ' \overline{\mathcal{g}}_d \left [ { \bar{\sigma}}\right ] \mathcal{j}^\mu[t',\vec{x } ' ] .\end{aligned}\ ] ] * linard wiechert fields * we now specialize to the current of a point electric charge @xmath545 in motion , sweeping out a timelike trajectory @xmath802 , @xmath803 = q \int_{-\infty}^{+\infty } { \text{d}}\tau \dot{y}^\mu[\tau ] \frac{\delta^{(d)}[x - y[\tau]]}{\sqrt{|g[x]| } } .\end{aligned}\ ] ] this expression , with @xmath427 being the point charge s proper time , is valid in any curved geometry @xmath99 ; in this section @xmath686 . inserting eq . into eq . , and following a calculation very similar to the gravitational one above , one would arrive at the lorenz gauge linard
wiechert vector potential of a point charge @xmath545 moving about in a background minkowski spacetime : @xmath804 & = \left .
\frac{q}{4\pi } ( -)^ { \frac{d-2}{2 } } \left ( \left ( \frac{1}{\dot{{\bar{\sigma } } } } \frac{{\text{d}}}{{\text{d}}\tau } \right)^ { \frac{d-4}{2 } } \frac{\dot{y}^\mu[\tau]}{(2\pi)^{\frac{d-4}{2 } } \dot{{\bar{\sigma } } } } \right ) \right\vert_{\tau=\tau_{\text{ret } } } , \\ \label{em_minkowski_lorenzlw_oddd } \text{odd $ d \geq 5 $ : } a^{\mu}[t,\vec{x } ] & = \frac{q}{2\pi } ( -)^ { \frac{d-3}{2 } } \int_{-\infty}^{\tau_{\text{ret}}-0^+ } \frac{{\text{d}}\tau}{\sqrt{2 { \bar{\sigma } } } } \frac{{\text{d}}}{{\text{d}}\tau } \left ( \left ( \frac{1}{\dot{{\bar{\sigma } } } } \frac{{\text{d}}}{{\text{d}}\tau } \right)^ { \frac{d-5}{2 } } \frac{\dot{y}^\mu[\tau]}{(2\pi)^{\frac{d-3}{2 } } \dot{{\bar{\sigma } } } } \right ) ; \end{aligned}\ ] ] where @xmath805 evaluated at retarded proper time @xmath707 is given in eq . .
the electromagnetic fields of such a point charge , encoded within the faraday tensor @xmath806}$ ] , is given by @xmath807 & = \left .
\frac{q}{4\pi } ( -)^ { \frac{d}{2 } } \left ( \left ( \frac{1}{\dot{{\bar{\sigma } } } } \frac{{\text{d}}}{{\text{d}}\tau } \right)^ { \frac{d-2}{2 } } \frac{(x - y[\tau])^{[\mu } \dot{y}^{\nu]}[\tau]}{(2\pi)^{\frac{d-4}{2 } } \dot{{\bar{\sigma } } } } \right ) \right\vert_{\tau=\tau_{\text{ret } } } , \\ \label{em_minkowski_lorenzlw_oddd_grad } \text{odd $ d \geq 5 $ : } f^{\mu\nu}[t,\vec{x } ] & = \frac{q}{2\pi } ( -)^ { \frac{d-1}{2 } } \int_{-\infty}^{\tau_{\text{ret}}-0^+ } \frac{{\text{d}}\tau}{\sqrt{2 { \bar{\sigma } } } } \frac{{\text{d}}}{{\text{d}}\tau } \left ( \left ( \frac{1}{\dot{{\bar{\sigma } } } } \frac{{\text{d}}}{{\text{d}}\tau } \right)^ { \frac{d-3}{2 } } \frac{(x - y[\tau])^{[\mu } \dot{y}^{\nu]}[\tau]}{(2\pi)^{\frac{d-3}{2 } } \dot{{\bar{\sigma } } } } \right ) .\end{aligned}\ ] ] * electromagnetic fields of continuous sources * the electromagnetic fields generated by an isolated but continuous ( as opposed to the discrete point charge ) source @xmath557 can be derived in a manner similar to that of the linard wiechert gauge field solutions above .
@xmath808 & = \frac{q}{4\pi } \int_{\mathbb{r}^{d-1 } } { \text{d}}^{d-1}\vec{x } ' \left.\left ( \frac{1}{2\pi(t - t ' ) } \frac{\partial}{\partial t ' } \right)^{\frac{d-4}{2 } } \left ( \frac{\mathcal{j}_{\mu}[t',\vec{x}']}{t - t ' } \right)\right\vert_{t - t'=|\vec{x}-\vec{x}'| } \\ \label{em_minkowski_lorenzlw_continuoussource_oddd } \text{odd $ d \geq 5 $ : } a_{\mu}[t,\vec{x } ] & = \frac{q}{2\pi } \int_{\mathbb{r}^{d-1 } } { \text{d}}^{d-1}\vec{x } ' \int_{-\infty}^{t-|\vec{x}-\vec{x}'|-0^+ } \frac{{\text{d}}t'}{\sqrt{2{\bar{\sigma } } } } \frac{\partial}{\partial t ' } \left\ { \left ( \frac{1}{2\pi(t - t ' ) } \frac{\partial}{\partial t ' } \right)^{\frac{d-5}{2 } } \left ( \frac{\mathcal{j}_{\mu}[t',\vec{x}']}{2\pi(t - t ' ) } \right ) \right\ } \end{aligned}\ ] ] the faraday tensor @xmath809}$ ] is , for even @xmath48 , @xmath810 & = \frac{q}{2 } \int_{\mathbb{r}^{d-1 } } { \text{d}}^{d-1}\vec{x } ' \left . \left ( \frac{1}{2\pi(t - t ' ) } \frac{\partial}{\partial t ' } \right)^{\frac{d-2}{2 } } \left ( \frac{(x - x')_{[\alpha } \mathcal{j}_{\beta]}[t',\vec{x}']}{t - t ' } \right ) \right\vert_{t - t'=|\vec{x}-\vec{x}'| } ; \end{aligned}\ ] ] while for odd @xmath679 , it reads instead @xmath811 & = q \int_{\mathbb{r}^{d-1 } } { \text{d}}^{d-1}\vec{x } ' \int_{-\infty}^{t-|\vec{x}-\vec{x}'|-0^+ } \frac{{\text{d}}t'}{\sqrt{2{\bar{\sigma } } } } \frac{\partial}{\partial t ' } \left\ { \left ( \frac{1}{2\pi(t - t ' ) } \frac{\partial}{\partial t ' } \right)^{\frac{d-3}{2 } } \left ( \frac{(x - x')_{[\alpha } \mathcal{j}_{\beta]}[t',\vec{x}']}{2\pi(t - t ' ) } \right ) \right\ } .\end{aligned}\ ] ] * causal structure * with the faraday tensor , we may construct the energy - momentum - shear - stress of the electromagnetic fields @xmath812 .
it is @xmath813 by inspecting equations through we learn that , just like the linearized gravitation case , the u@xmath605 gauge potential and the electromagnetic fields engendered by a spatially localized system of charges in motion transmit information solely on the light cone in even dimensions @xmath48 and strictly within the null cone in odd dimensions @xmath679 .
this statement also applies to the radiative electromagnetic poynting vector , i.e , the @xmath321 component of eq .
, since @xmath812 is built solely out of the faraday tensor @xmath814 ( and the metric ) .
* linearized yang - mills - wong s equations * in yang - mills ( non - abelian ) gauge theory , the vector potential @xmath815 now acquires an additional color index @xmath529 , and the full field strength tensor is @xmath816 } + i g f^{abc }
a^b_{\phantom{b}\mu } a^c_{\phantom{c}\nu } , \end{aligned}\ ] ] where @xmath817 measures the strength of the gauge boson self - interactions and the @xmath818 are the structure constants of the gauge group of interest .
in particular , if @xmath819 are the generators of the group , normalized such that @xmath820 } = \delta^{ab}/2 $ ] , then the structure constants arise from the commutator @xmath821 = i f^{abc } t^c$ ] .
the analog of the electromagnetic linard wiechert fields , i.e. , the yang - mills gauge field produced by a point charge , is governed by wong s equations @xcite .
wong introduced an isospin vector @xmath822 to source the color degrees of freedom , and the resulting equations are @xmath823 where the colored point charge
current is @xmath824 \equiv
q \int_{-\infty}^{+\infty } { \text{d}}\tau i^a[\tau ] \dot{y}^\nu[\tau ] \frac{\delta^{(d)}[x - y[\tau]]}{\sqrt{|g[x]| } } .
\end{aligned}\ ] ] in eq .
, @xmath825 is the gauge - covariant derivative in curved spacetime . because @xmath826 is an identity
, consistency demands that the right hand side of wong s equations be divergence - less in the gauge - covariant sense : @xmath827 .
this in turn leads to the conclusion that the isospin vector must be parallel transported along the color charge s worldline , @xmath828 .
( this is a direct analogy to the gravitational situation , where the consistency of einstein s equations require that the stress - energy tensor be divergence - less , @xmath829 ; the geodesic equation is uncovered when this is applied to the point mass case of eq . . )
moreover , the length of @xmath822 is preserved along this same trajectory , namely @xmath830 .
while the form of wong s equation in eq . holds in any geometry , we shall be content with @xmath747 and its linearized version . specifically ,
if we impose the lorenz gauge @xmath831 then keeping only terms linear in @xmath815 in eq .
yields @xmath832 = q \int_{-\infty}^{+\infty } { \text{d}}\tau i^a[\tau ] \dot{y}^\mu[\tau ] \delta^{(d)}[x - y[\tau ] ] .\end{aligned}\ ] ] this linearized lorenz gauge yang - mills - wong s equations are almost identical to its u@xmath605 maxwell counterpart ( i.e. , eq . inserted into eq . ) .
in fact , by comparing the right hand side of eq . to eq .
, we see the solution to eq .
can be obtained from equations and by replacing @xmath833 .
this hands us the linearized linard
wiechert solution to eq . in the lorenz gauge
: @xmath834 & = \left .
\frac{q}{4\pi } ( -)^ { \frac{d-2}{2 } } \left ( \left ( \frac{1}{\dot{{\bar{\sigma } } } } \frac{{\text{d}}}{{\text{d}}\tau } \right)^ { \frac{d-4}{2 } } \frac{i^a[\tau ] \dot{y}^\mu[\tau]}{(2\pi)^{\frac{d-4}{2 } } \dot{{\bar{\sigma } } } } \right ) \right\vert_{\tau=\tau_{\text{ret } } } , \\ \label{ymw_minkowski_lorenzlw_oddd } \text{odd $ d \geq 5 $ : } a^{a\mu}[t,\vec{x } ] & = \frac{q}{2\pi } ( -)^ { \frac{d-3}{2 } } \int_{-\infty}^{\tau_{\text{ret}}-0^+ } \frac{{\text{d}}\tau}{\sqrt{2 { \bar{\sigma } } } } \frac{{\text{d}}}{{\text{d}}\tau } \left ( \left ( \frac{1}{\dot{{\bar{\sigma } } } } \frac{{\text{d}}}{{\text{d}}\tau } \right)^ { \frac{d-5}{2 } } \frac{i^a[\tau ] \dot{y}^\mu[\tau]}{(2\pi)^{\frac{d-3}{2 } } \dot{{\bar{\sigma } } } } \right ) ; \end{aligned}\ ] ] where , once again , @xmath805 evaluated at retarded proper time @xmath707 is given in eq . . _ comparison with linearized gravity _ by comparing the energy - momentum - shear - stress tensor of a point mass in eq . to that of the color charge current of wong s equations in eq .
, let us observe we can go from the latter to the former by replacing mass with charge and isospin vector with the @xmath97-proper - velocity : @xmath835 comparing the linear graviton field generated by a point mass in equations and to the linear non - abelian gauge field engendered by a point charge in equations and , we further observe that the latter can be obtained from the former by replacing @xmath836 we highlight here , this sort of replacement rules to obtain einstein gravity results from those of yang - mills ( ym ) theory finds a precise dictionary within the context of their @xmath80-particle scattering amplitudes .
bern , carrasco , and johansson showed in @xcite how , once the scattering amplitude of @xmath80 on - shell non - abelian gauge bosons are expressed in a particular form in fourier space , the corresponding @xmath80 graviton amplitude can be gotten by simply replacing the ym gauge coupling with the appropriate gravitational constant and color ( group theoretic ) factors with their kinematic counterparts ( composed of momentum and/or polarization tensor dot products ) .
graviton and the spin@xmath837 non - abelian gauge boson .
this is nowadays known as the klt relations " .
] this specific set of replacement rules , nowadays known as the bcj color - kinematics duality , is one reason for the slogan
gr @xmath608 ym@xmath451 " .
an outstanding question is whether this
double copy " relation between gr and ym extends to their classical solutions . to this end , recent work by luna et al .
@xcite , @xcite , and monteiro et al .
@xcite ( see also ridgway and wise @xcite ) have exploited the kerr - schild class of gravity solutions , which includes the physically important schwarzschild / kerr black hole geometries .
it takes the form of flat metric plus perturbation , " with the perturbation being proportional to two copies of the same null vector @xmath838 : @xmath839 ( the @xmath840 is not only null with respect to both @xmath613 and @xmath99 , it is also geodesic , i.e. , @xmath841 . ) the technical significance of the kerr - schild class of metrics in eq .
is that it converts the usually nonlinear einstein s equations to linear ones .
this , in turn , is what allowed @xcite , @xcite , and @xcite to readily identify the corresponding linear ym solutions .
our replacement of the color isospin vector with the spacetime velocity eq .
, in the yang - mills - wong s equation , is ( heuristically speaking ) yet another example of this double - copy relationship between classical solutions of gravity and yang - mills .
in fact , in the limit where our point charge is permanently at rest , it has to coincide with the schwarzschild - tangherlini ( gr ) @xmath842 colored - coulomb ( ym ) relationship identified in @xcite .
although @xcite , @xcite , and @xcite found exact relations between classical solutions of gr and ym , whereas the comparison we are making here holds only at the linearized level , we wish to suggest that in the spirit of
gr @xmath608 ym@xmath451 " it is not necessary to only search for relations between exact classical solutions of the two theories . for one ,
the gr @xmath843 ym@xmath451 correspondence is highly non - trivial because it involves the nonlinear self - interactions of the gravitons and gauge boson , while the classical solution relations revealed so far are strictly linear ones .
moreover , on the gravity side , the weak field perturbative metric generated by @xmath844 point masses of the form in eq .
, is in fact foundational to our understanding of astrophysical dynamics , such as that of our own solar system or the binary neutron stars / small black holes whose gws we hope to detect .
this weak field expansion is a power series in newton s constant @xmath96 , and is known in the relativity literature as the post - minkowskian ( pn ) program ; while its non - relativistic version is the post - newtonian ( pn ) one .
we thus pose the question : _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ is there a perturbative post - colored - coulomb program whose square " is that of the pn / pm one , that perhaps involves solving for the yang - mills gauge field engendered by @xmath844 point charges governed by wong s eq .
and its generalizations ? _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ to answer this question , it may be useful to first seek whether an efficient gauge for such calculations exist , such that when there is only one point mass or one point charge in the system , the equations linearize automatically and the correspondence in @xcite is recovered .
the search for such a scheme on the gravity side , motivated primarily by the desire to improve pn / pm calculations , has been initiated recently by harte @xcite and harte and vines @xcite .
* memory in even @xmath48 minkowski * we now turn to illustrate , in direct analogy to the gravitational case above , how linear yang - mills memory may arise from scattering of @xmath525 point charges on unbound trajectories .
we shall assume that some non - abelian version of the lorentz force law in eq .
holds , so that the formula for the electric memory effect in eq
. remains applicable .
specifically , we find from the solutions in equations and , @xmath845 where the field generated by the color charges configuration in the asymptotic past is @xmath846 and that in the asymptotic future reads @xmath847 here , in addition to the asymptotic spacetime trajectories in equations and , we also need to assume analogous ones for the isospin vectors @xmath848 = i^a_{(\text{j } \vert 0 ) } \qquad\qquad \text{(constant ) } , \\ \label{colorsource_out } \lim_{t \to + \infty } i^a_{\text{j}}[t ] = i^a_{(\text{j } \vert 1 ) } \qquad\qquad \text{(constant ) } .\end{aligned}\ ] ] the gauge - invariant vector contribution @xmath849 to eq . is simply the transverse portion of eq . .
( although we will not display them explicitly , analogous formulas for u@xmath605 electromagnetic memory may be obtained from equations , and by simply discarding the isospin vector @xmath822 . )
_ double copy property _ as already advertised , the gauge - invariant contributions to the memories of gr and ym , produced by @xmath91 point sources scattering off each other on unbound trajectories , enjoy a double copy property . amongst the replacement rules in equations and
that would translate between the linear gr one in equations , , and the ym one in equations , and , the replacement of the isospin vector @xmath822 with the spacetime velocity @xmath850 appears to be most closely related to the spirit of the bcj color - kinematics duality @xcite .
we close this subsection by wondering , does the gravitational spin memory " discovered in @xcite have a yang - mills
double copy " analog ? for the second part of this section
to complement the gauge - invariant formalism of
we turn to solving maxwell s equations and in the generalized lorenz gauge @xmath851 the zeroth component is now @xmath852 while the spatial component becomes @xmath853 to make progress we shall now take the scale factor to describe a cosmology with a constant equation of state @xmath0 , given by eq . .
maxwell s equations with the gauge in eq .
now read @xmath854^{\frac{d-4}{2 } } a^0[\eta,\vec{x}]\right ) & = a[\eta]^{\frac{d+4}{2 } } \mathcal{j}^0[\eta,\vec{x } ] , \\ \left ( \partial^2 + ( d-4 ) \frac{1+(d-1 ) w}{q_w^2 } \frac{1}{\eta^2 } \right ) \left ( a[\eta]^{\frac{d-4}{2 } } a^i[\eta,\vec{x}]\right ) \label{maxwell_lorenz_i_w } & = a[\eta]^{\frac{d+4}{2 } } \mathcal{j}^i[\eta,\vec{x } ] .\end{aligned}\ ] ] * solutions * the retarded solutions to the gauge potential subject to the generalized lorenz gauge of eq .
are given by @xmath417^{\frac{d-4}{2 } } a^0[\eta,\vec{x } ] & = \int_{\eta_\circ}^{\eta } { \text{d}}\eta ' \int_{\mathbb{r}^{d-1 } } { \text{d}}^{d-1 } \vec{x } ' a[\eta']^{\frac{d+4}{2 } } \mathcal{g}^{(\gamma \vert \text{time})}[\eta,\eta';{\bar{\sigma } } ] \mathcal{j}^0[\eta',\vec{x } ' ] , \\
a[\eta]^{\frac{d-4}{2 } } a^i[\eta,\vec{x } ] & = \int_{\eta_\circ}^{\eta } { \text{d}}\eta ' \int_{\mathbb{r}^{d-1 } } { \text{d}}^{d-1 } \vec{x } ' a[\eta']^{\frac{d+4}{2 } } \mathcal{g}^{(\gamma \vert \text{space})}[\eta,\eta';{\bar{\sigma } } ] \mathcal{j}^i[\eta',\vec{x } ' ] ; \end{aligned}\ ] ] where the symmetric green s functions in all relevant even dimensions are @xmath855 & = \frac{1}{(2\pi)^{\frac{d-2}{2 } } } \left(\frac{\partial}{\partial { \bar{\sigma}}}\right)^{\frac{d-2}{2 } } \left ( \frac{\theta[{\bar{\sigma}}]}{2 } p_{-\frac{2d-7 + ( d-1)w}{q_w}}[1+s ] \right ) , \\
\mathcal{g}^{(\gamma \vert \text{space})}_{\text{even $ d \geq 4$}}[\eta,\eta';{\bar{\sigma } } ] & = \frac{1}{(2\pi)^{\frac{d-2}{2 } } } \left(\frac{\partial}{\partial { \bar{\sigma}}}\right)^{\frac{d-2}{2 } } \left ( \frac{\theta[{\bar{\sigma}}]}{2 } p_{-\frac{d-4}{q_w}}[1+s ] \right ) ; \end{aligned}\ ] ] and their counterparts in all relevant odd dimensions are @xmath856 & = \left ( \frac{1}{2\pi } \frac{\partial}{\partial { \bar{\sigma}}}\right)^{\frac{d-3}{2 } } \left ( \frac{\theta[{\bar{\sigma}}]}{4\pi } \frac{1 + \left(s+\sqrt{s ( s+2)}+1\right)^{\frac{3 d - 11 + ( d - 1 ) w}{q_w}}}{\sqrt{{\bar{\sigma}}(s+2 ) } \left(s+\sqrt{s ( s+2)}+1\right)^{\frac{3 d - 11 + ( d - 1 ) w}{2 q_w } } } \right ) , \\
\mathcal{g}^{(\gamma \vert \text{space})}_{\text{odd $ d \geq 3$}}[\eta,\eta';{\bar{\sigma } } ] & = \left ( \frac{1}{2\pi } \frac{\partial}{\partial { \bar{\sigma}}}\right)^{\frac{d-3}{2 } } \left ( \frac{\theta[{\bar{\sigma}}]}{4\pi } \frac{1 + \left(s+\sqrt{s ( s+2)}+1\right)^{\frac{d - 5 - ( d - 1 ) w}{q_w}}}{\sqrt{{\bar{\sigma}}(s+2 ) } \left(s+\sqrt{s ( s+2)}+1\right)^{\frac{d - 5 - ( d - 1 ) w}{2 q_w } } } \right ) .\end{aligned}\ ] ]
* setup * in this section , we re - visit the derivation of eq .
( 5 ) in @xcite , which describes the fractional distortion between two timelike worldlines in cosmological geometries .
specifically , we wish to derive an integral formula for the geodesic spatial distance , on a constant time @xmath112-surface , between two point test masses at rest ( aka co - moving " ) in some cosmological spacetime . to this end
, it is very useful to employ the the perturbed cosmological geometry in synchronous gauge in eq . . for , in this synchronous gauge coordinate system ,
co - moving observers do not change their spatial coordinates . denoting their trajectories as @xmath857 $ ] and @xmath858 $ ]
, we shall define their constant spatial components as @xmath859 = \vec{y}_0 \qquad \qquad \text { and } \qquad \qquad \vec{z}[s ] = \vec{z}_0 ; \end{aligned}\ ] ] whereas their time components satisfy the proper time condition @xmath860 , namely @xmath861}{{\text{d}}s } = \frac{{\text{d}}z^0[s]}{{\text{d}}s } = \frac{{\text{d}}\eta}{{\text{d}}s } = \frac{1}{a[\eta ] } .\end{aligned}\ ] ] as already stated in , one may check readily that the relevant geodesic equations are satisfied to all orders in @xmath30 : @xmath862 * spacelike geodesic distance * now , the geodesic spatial distance @xmath22 $ ] between the pair of test masses on a constant@xmath863 hypersurface can be expressed through the positive square root of the integral @xmath864 ^ 2 \equiv a[\eta]^2 \int_{0}^{1 } \left ( \delta_{ij } - \chi_{ij } \right ) \frac{{\text{d}}w^i}{{\text{d}}\lambda } \frac{{\text{d}}w^j}{{\text{d}}\lambda } { \text{d}}\lambda .\end{aligned}\ ] ] the @xmath865 , in turn , obeys the spatial geodesic equation @xmath866 subject to the boundary conditions @xmath867 = y^i[\eta ] = y^i_0 \qquad\qquad \text { and } \qquad\qquad w^i[\lambda=1 ] = z^i[\eta ] = z^i_0 .\end{aligned}\ ] ] because eq .
is the action whose variation with respect to @xmath865 yields the geodesic equation of eq .
, to first order in @xmath30 , we may simply replace @xmath865 with its solution if the spatial metric were conformally flat ( i.e. , if @xmath868 ) .
in other words , if in eq .
we replaced @xmath865 with a straight line joining @xmath869=\vec{y}_0 $ ] to @xmath870=\vec{z}_0 $ ] , @xmath871 \equiv y_0^i + \lambda ( z_0-y_0)^i , \end{aligned}\ ] ] the error incurred would scale higher than @xmath872 $ ] . to see this more explicitly ,
we vary @xmath873 in eq . , taking into account @xmath874 = \delta w^i[\lambda = 1 ] = 0 $ ] ( from eq . ):
@xmath875 ^ 2 = -2 a[\eta]^2 \int_{0}^{1 } \delta w^i \left ( \delta_{ij } - \chi_{ij } \right ) \frac{d^2 w^j}{{\text{d}}\lambda^2 } { \text{d}}\lambda + \mathcal{o}[(\delta\vec{w})^2 ] .\end{aligned}\ ] ] when the spatial geodesic equation in eq .
is satisfied by @xmath865 , the integrand vanishes at first order in @xmath30 . at this point , making the replacement in eq .
and noting its @xmath876-derivative returns @xmath877 = ( z_0-y_0)^i$ ] , reduces eq . to @xmath878 ^ 2 \approx a[\eta]^2 \left ( ( \vec{z}_0-\vec{y}_0)^2 - ( z_0-y_0)^i ( z_0-y_0)^j \int_{0}^{1 } \chi_{ij}\left [ \eta , \overline{w}^l[\lambda ] \right ] { \text{d}}\lambda \right ) .\end{aligned}\ ] ] (
the variational argument here has previously been used in pfenning and poisson @xcite and chu and starkman @xcite to derive an expression , similar in spirit to eq .
, for the geodesic distance between any two points in a weakly curved spacetime . )
* results * the @xmath872$]-accurate spatial geodesic distance between the pair of test masses at @xmath20 and @xmath21 , at constant @xmath552 , is now provided by the integral @xmath879 \approx a[\eta ] \left\vert \vec{z}_0-\vec{y_0 } \right\vert \left ( 1 - \frac{\widehat{n}^i \widehat{n}^j}{2 } \int_{0}^{1 } \chi_{ij}\left [ \eta , \vec{y}_0 + \lambda \left(\vec{z}_0-\vec{y}_0\right ) \right ] { \text{d}}\lambda \right ) , \qquad\qquad \widehat{n } \equiv \frac{\vec{z}_0-\vec{y}_0}{|\vec{z}_0-\vec{y}_0| } .\end{aligned}\ ] ] suppose we began our hypothetical experiment at some time @xmath23 . at some later time @xmath880 , the fractional distortion induced on the pair of test masses
is @xmath881 \equiv \frac{l[\eta]}{l[\eta ' ] } - 1 , \end{aligned}\ ] ] which according to eq .
translates to @xmath882 = \left ( \frac{a[\eta]}{a[\eta ' ] } - 1 \right ) - \frac{a[\eta]}{a[\eta ' ] } \frac{\widehat{n}^i \widehat{n}^j}{2 } \int_{0}^{1 } \delta \chi_{ij } { \text{d}}\lambda + \mathcal{o}[(\chi_{\mu\nu})^2 ] , \end{aligned}\ ] ] where the @xmath883 refers to the total change in the spatial metric perturbations over the duration @xmath884 $ ] ( cf .
. to summarize : eq . is valid up at first order in metric perturbations in any @xmath97-dimensional background spatially flat flrw cosmology .
however , over the timescales of human experiments ( i.e. , @xmath885 $ ] ) , the size of the universe will not change appreciably .
we may thus taylor expand the scale factor in @xmath146/a[\eta ' ] - 1 \approx ( \dot{a}[\eta']/a[\eta ' ] ) ( \eta-\eta ' ) \ll 1 $ ] and arrive at eq . .
* remark i * previously , in @xcite , because we did not explicitly invoke the synchronous gauge form of the metric in eq . , what we derived there was really a comparison of the proper spatial distances between the pair of test masses with and without the metric perturbations at a fixed time .
( we also , implicitly , set @xmath31=0 $ ] . )
it is the desired fractional distortion result in eq . if the spatial coordinates of the test masses do not change with time , as is the case in the synchronous gauge employed here .
* remark ii * we also comment on the misleading transverse - traceless gauge " ( tt - gauge ) often invoked to derive gw observables such as the proper distance formula in eq . .
the tt - gauge amounts to setting to zero all components of @xmath16 in eq .
, except the spatial ones @xmath886 obeying @xmath887 .
this is a legitimate choice of coordinates if spacetime were completely empty ; but as we have seen in both the perturbed cosmological and minkowski cases , the tt - gauge " really does not exist once there is matter including the very astrophysical source of gws responsible , in the first place , for the fluctuations in proper distances measured by our laser interferometers . for
, the presence of a gw source would yield non - zero scalar @xmath543 and vector @xmath888 metric perturbations ; and since they are gauge - invariant , it is not possible to coordinate transform them to zero . , @xmath657 ) and vector ( @xmath889 ) gauge invariant quantities on the spatial boundary of some local vacuum region of spacetime , which in turn imply they have to vanish throughout its interior
physically speaking , the boundary values of these scalar , vector and tensor perturbations really ought to be determined by the matter / gw sources external to the region considered , and one should not be free to choose them arbitrarily . ] why , then , is one able to get away with adopting this inconsistent tt - gauge " for making gw predictions ?
it is likely due to the dynamics of general relativity itself . to understand this , and as a complementary approach to the synge s world - function based one used to derive eq .
, we shall examine the geodesic deviation equation that encodes the tidal forces exerted upon a gw detector . if @xmath890 is tangent to the timelike geodesic swept out by one component of lisa , for instance ; and @xmath58 is the vector joining it to a nearby component ; from eq .
, @xmath891 if the riemann tensor were zero , @xmath58 is parallel transported along @xmath892 and there would be no tidal forces .
it is convenient to work with the synchronous gauge metric in eq . , because it allows us to make the exact statement @xmath893 . by a direct calculation
, one may then verify that this choice of gauge eliminates the @xmath894 component of eq . because both @xmath895 and @xmath896 are individually zero . _ minkowski background _
let us , for a start , assume the detector is operating nearly at rest in a background minkowski spacetime . since the background riemann tensor vanishes , its first order counterpart is gauge - invariant .
in fact , the spatial components of the right hand side of eq . can be re - cast in terms of the gauge - invariant variables , @xmath897 where @xmath898 without the solutions to linearized general relativity at hand , one may be tempted to surmise that all scalar , vector and tensor perturbations produced by a distant astrophysical system contribute to the tidal squeezing and stretching of a gw detector .
but if we do in fact assume general relativity holds , so that equations , , and through may be exploited , @xmath899 by referring to eq .
we recognize the first line to contain various irreducible spatial components of the matter stress - energy tensor .
if we assume the gw source is isolated and well separated from the observer , this first line vanishes identically at the gw detector s location . as for the second line , by recalling the _ coulomb vs. waves _ " discussion in , we may estimate the scalar - to - tensor amplitude to scale as ( ( characteristic timescale of gw source)/(observer - source spatial distance))@xmath451 .
hence , in the far zone where the gw detector is located , @xmath900 to reiterate : it is only after the dynamics of general relativity has been incorporated , eq .
reduces to the statement that the acceleration of the deviation vector in spatial directions is driven primarily by the tensor mode filling the space around a distant gw detector . _ cosmological background _ we move on to discuss the case where gws are journeying across a cosmological background
. we shall invoke the form of the riemann tensor in eq .
, and for technical simplicity focus only on the weyl contribution to tidal forces .
now , the weyl tensor @xmath901 , with one upper index , is conformally invariant .
that means , when linearized about a perturbed spatially flat flrw universe , it must be equal to its flat spacetime counterpart .
moreover , since it vanishes about the background homogeneous and isotropic universe , the linearized weyl tensor @xmath902 must itself be gauge - invariant .
a direct calculation then shows the components of weyl relevant for the @xmath299th component of the geodesic deviation eq .
are @xmath903 where the first term ( from the left ) involves the scalar spacetime laplacian , @xmath904 .
since @xmath905 actually equals its linearized counterpart about minkowski , all scale factors really cancel out . but
putting the weyl tensor in this form allows the eom for @xmath53 , @xmath312 , @xmath311 and @xmath271 specifically , equations , , and to be employed . if we further make the approximation that the gw experiment does not run over cosmological timescales , so that all the ensuing @xmath449 terms may be neglected , eq . could then be re - expressed as @xmath906 much like the minkowski case above , the first line vanishes identically at the gw - detector s location because it contains various @xmath135 components of the isolated astrophysical system .
the subtlety in the cosmological case at hand arises on the second line : even though @xmath53 admits wave solutions and therefore always plays a key role in exerting tidal forces on the gw detector ; @xmath311 does so only when @xmath1 . for
, it is within this range of eos that @xmath311 obeys a wave equation ( namely , eq . ) and the @xmath907s acting on it in eq .
can be treated on roughly the same footing as the time derivative @xmath908(characteristic timescale of gw source ) .
we close this section by suggesting that the
conformal fermi normal coordinates " recently developed in dai , pajer and schmidt @xcite ( which built upon earlier work by pajer , schmidt and zaldarriaga @xcite ) for cosmological large scale structure applications , could perhaps also be used to integrate the geodesic deviation equation to study gw observables and potential memory effects in our universe .
in this section , for the reader s reference , we will provide a summary of appendix ( b ) of @xcite , which delineates how to solve the primary partial differential equation ( pde ) encountered there and in this paper : @xmath909 = \mathcal{s}[\eta,\vec{y } ] , \qquad\qquad \partial^2 \equiv \eta^{\mu\nu } \partial_\mu \partial_\nu , \end{aligned}\ ] ] where the source @xmath910 is assumed to be some externally prescribed function of spacetime . in the main text , and depending on the context ,
the coordinates are either @xmath911 or @xmath912 ( for eos @xmath1 ) ; here , we will just use @xmath913 .
the @xmath914 in the @xmath915 term is a convenient choice ; for our purposes it is to be treated as an arbitrary real constant . the key object
we wish to compute is the symmetric green s function @xmath916 obeying @xmath917 \delta^{(d-1)}[\vec{y}-\vec{y } ' ] .\end{aligned}\ ] ] once @xmath916 is known , the solution to eq .
is @xmath918 = \int_{\mathbb{r}^{d-1 } } { \text{d}}^{d-1}\vec{y } ' \int_{\eta_\circ}^{\eta } { \text{d}}\eta ' \mathcal{g}\left [ \eta,\vec{y } ; \eta',\vec{y } ' \right ] \mathcal{s}[\eta',\vec{y } ' ] .\end{aligned}\ ] ] * dimension reduction & nariai s ansatz * the green s function equation in eq . may be tackled via dimension reduction because of the space - translation symmetry and parity invariance of the wave operator in eq . .
once the 2d @xmath919 and 3d @xmath920 are known , the higher dimensional green s functions can be obtained by repeated differentiation with respect to synge s world function in minkowski spacetime @xmath921 : @xmath922 & = \left(\frac{1}{2\pi } \frac{\partial}{\partial{\bar{\sigma}}}\right)^ { \frac{d-2}{2 } } \mathcal{g}_2[\eta,\eta';{\bar{\sigma } } ] , \\
\label{g_oddd } \mathcal{g}_{\text{odd $ d \geq 3$}}[\eta,\eta';{\bar{\sigma } } ] & = \left(\frac{1}{2\pi } \frac{\partial}{\partial{\bar{\sigma}}}\right)^ { \frac{d-3}{2 } } \mathcal{g}_3[\eta,\eta';{\bar{\sigma } } ] .\end{aligned}\ ] ] both the 2d and 3d cases can be gotten by exploiting nariai s ansatz @xcite in 2d , we postulate that the tail function depends on spacetime solely through a single object @xmath923 : @xmath924 = \frac{\theta[{\bar{\sigma}}]}{2 } g_2[s ] , \qquad\qquad s \equiv \frac{{\bar{\sigma}}}{\eta\eta ' } .\end{aligned}\ ] ] this @xmath925 $ ] obeys the homogeneous version of eq . , and nariai s ansatz turns the pde into the ordinary differential equation @xmath926 = \frac{s ( 2 + s ) g_2''[s ] + 2 ( 1 + s ) g_2'[s ] - \kappa ( 1 + \kappa ) g_2[s]}{\eta^2 } = 0 .\end{aligned}\ ] ] following that , one needs to impose the null cone boundary condition @xmath927=1 $ ] .
this identifies the unique solution to be the legendre function @xmath928 = p_\kappa[1+s ] .\end{aligned}\ ] ] in 3d , we postulate that @xmath929 = \frac{\theta[{\bar{\sigma}}]}{2\pi } \frac{\widehat{g}_3[s]}{\sqrt{\eta\eta ' } } .\end{aligned}\ ] ] this @xmath930/\sqrt{\eta\eta'}$ ] obeys the homogeneous version of eq . , and nariai s ansatz turns the pde into the ordinary differential equation @xmath931}{\sqrt{\eta\eta ' } }
= \frac{4 s ( 2 + s ) \widehat{g}_3''[s ] + 12 ( 1 + s ) \widehat{g}_3'[s ] + ( 3 - 4 \kappa ( 1 + \kappa ) ) \widehat{g}_3[s]}{4 \eta^2 \sqrt{\eta\eta ' } } = 0 .\end{aligned}\ ] ] following that , one needs to impose the null cone boundary condition @xmath932/\sqrt{\eta\eta'}=1/\sqrt{2{\bar{\sigma}}}$ ] .
the unique solution is @xmath933}{\sqrt{\eta\eta ' } } = \frac{\left ( s + \sqrt{s(s+2 ) } + 1 \right)^{1 + 2\kappa } + 1}{2 \sqrt{{\bar{\sigma}}(s+2 ) } \left ( s + \sqrt{s(s+2 ) } + 1 \right)^{\frac{1}{2}(1 + 2\kappa ) } } .\end{aligned}\ ] ] * results * inserting the results in equations , , , into equations and , the symmetric green s function solutions to equation are thus @xmath934 & = \left ( \frac{1}{2\pi } \frac{\partial}{\partial{\bar{\sigma}}}\right)^ { \frac{d-2}{2 } } \left(\frac{\theta[{\bar{\sigma}}]}{2 } p_\kappa[1+s]\right ) , \\
\label{gresults_oddd } \mathcal{g}_{\text{odd $ d \geq 3$}}[\eta,\eta';{\bar{\sigma } } ] & = \left ( \frac{1}{2\pi } \frac{\partial}{\partial{\bar{\sigma}}}\right)^ { \frac{d-3}{2 } } \left ( \frac{\theta[{\bar{\sigma}}]}{4\pi } \frac{\left ( s + \sqrt{s(s+2 ) } + 1 \right)^{1 + 2\kappa } + 1}{\sqrt{{\bar{\sigma}}(s+2 ) } \left ( s + \sqrt{s(s+2 ) } + 1 \right)^{\frac{1}{2}(1 + 2\kappa ) } } \right ) , \\ { \bar{\sigma}}&\equiv \frac{(\eta-\eta')^2-(\vec{y}-\vec{y}')^2}{2 } , \qquad\qquad s \equiv \frac{{\bar{\sigma}}}{\eta\eta ' } .\end{aligned}\ ] ] note that the retarded green s function @xmath935 propagates signals strictly in the future of the source , so to obtain it from its symmetric cousin , we simply multiply it by a @xmath43-function of elapsed time : @xmath936 = \theta[\eta-\eta ' ] \mathcal{g}\left [ \eta,\eta ' ; { \bar{\sigma}}\right ] .\end{aligned}\ ] ] * caution * in this paper , whenever @xmath10 in eq . , the wave solutions in eq .
are globally incomplete because of the presence of a particle horizon they vanish outside of the light ( or , acoustic ) cone of the gw source at the big bang @xmath937 .
the symbol @xmath352 appearing throughout the main text reminds us of this fact . specifically , whenever @xmath149 , @xmath938 .
whereas , when @xmath49 , we shall define @xmath939 whenever the past light ( acoustic ) cone of the observer at @xmath504 has a non - trivial intersection with the spacetime world tube of the gw source ; and @xmath940 otherwise .
this global incompleteness of the wave solutions may lead to the violation of gauss law in both linearized einstein s equations and maxwell s electromagnetism , since the relevant flux integrals would vanish outside the particle horizon at a given time @xmath552 .
we hope to return to this issue in future work ; for now , see higuchi and lee @xcite for a discussion .
* remark * it has not escaped our attention that the @xmath941 occurring throughout this paper can be viewed as a proper lorentz so@xmath942 invariant object in a @xmath943-dimensional minkowski spacetime .
( so@xmath942 is also isomorphic to the conformal group in @xmath65 spatial dimensions ; see , for instance , di francesco et al .
@xcite for a discussion . )
even though we are not working exclusively in de sitter spacetime ( @xmath128 ) , we may pretend for this purpose that our observer at @xmath944 and source at @xmath945 lie on a fictitious hyperboloid of size @xmath946 , for some @xmath467 , embedded in a @xmath943-dimensional minkowski spacetime . in this higher dimensional flat spacetime ,
let us define the @xmath97 cartesian coordinates @xmath947\}$ ] to be @xmath948 and make a similar definition for @xmath949\}$ ] .
( unlike the de sitter case , where @xmath552 is non - positive , the sign of @xmath552 here depends on the sign of @xmath950 . )
if @xmath951 $ ] is the metric in the ambient minkowski spacetime , we not only obtain the equation of the hyperboloid @xmath952 we also have the proper lorentz invariant object @xmath953
in a @xmath97-dimensional geometry that enjoys spatial translation symmetry and parity invariance , the green s function @xmath955 of its massive wave operator @xmath956 may be obtained from its counterpart @xmath268 dimensions lower , i.e. , from @xmath957 through the formula @xmath958
= -\frac{1}{2\pi r } \frac{\partial}{\partial r } g_{d-2 } \left [ \eta,\eta';r \equiv |\vec{x}-\vec{x}'|_{d-3 } \right ] , \end{aligned}\ ] ] where on the left hand side @xmath959 is the euclidean distance in @xmath112-space and on the right hand side @xmath960 is that in @xmath961-space . this can be understood
see , for instance , @xcite and @xcite by viewing the @xmath962-dimensional green s function @xmath957 as being sourced by an appropriately defined plane source " extending into two extra spatial dimensions . in this section
, we will specialize to minkowski spacetime @xmath963 where @xmath964 is the metric on the @xmath962-sphere and show how the recursion relation eq .
can also be understood from an algebraic perspective .
to begin , let us note that , by the poincre invariance of minkowski spacetime , we expect the green s function to depend on spacetime through the time elapsed @xmath965 and the euclidean distance @xmath656 , where in cartesian coordinates @xmath966 is the location of the observer and @xmath967 that of the spacetime point source .
again by poincre invariance , we may place the source at @xmath968 without loss of generality .
this means the green s functions now depend on spacetime coordinates as @xmath969 $ ] .
let us define , in the coordinate system of eq .
, @xmath970 if we apply the @xmath97-dimensional wave operator @xmath971 of the minkowski metric in eq . to any function
@xmath972 $ ] that depends only on @xmath682 and @xmath541 , we would find @xmath973 = \partial_t^2 f - \frac{1}{r^{d-2 } } \partial_r \left ( r^{d-2 } \partial_r f \right ) .\end{aligned}\ ] ] the key algebraic observation here is that , applying the @xmath97-dimensional wave operator followed by the @xmath974 defined in eq .
is the same as applying @xmath974 first followed by the @xmath975-dimensional wave operator : @xmath976 = \left(\partial^2_{d+2 } + m^2\right ) \mathcal{d}_r f[t , r ] .\end{aligned}\ ] ] furthermore , if we use the identity @xmath977 ) ' = 0 $ ] to replace @xmath978 \to -\delta[r]/r$ ] , we find that @xmath974 converts the @xmath97-dimensional @xmath42-function measure associated with a spacetime point source at the origin into one in @xmath975 dimensions , namely @xmath979 \delta[r]}{\omega_{d-2 } r^{d-2}}\right ) = \frac{\delta[t ] \delta[r]}{\omega_{d } r^{d } } , \end{aligned}\ ] ] where @xmath980 $ ] is the solid angle subtended by a unit radius sphere embedded in a @xmath112-dimensional euclidean space . at this point , suppose we had solved the @xmath962-dimensional green s function , which obeys @xmath981 = \frac{\delta[t]\delta[r]}{\omega_{d-4 } r^{d-4 } } , \end{aligned}\ ] ] we may proceed to apply @xmath974 on both sides of eq . .
utilizing eq . on the left hand side and eq . on the right
then leads us to @xmath982 = \frac{\delta[t]\delta[r]}{\omega_{d-2 } r^{d-2 } } .\end{aligned}\ ] ] once @xmath983 $ ] is known , @xmath984 $ ] would solve the @xmath97-dimensional massive green s function equation this is equivalent to eq . by the translation invariance of space .-
de sitter spacetime to that in @xmath97-de sitter ; see footnote 6 of @xcite . ] * questions * that the minimally coupled massless scalar green s function has non - zero tails in all odd dimensional minkowski spacetimes , suggests spherical symmetry is not compatible with strictly - null propagation in the latter , since the green s function is itself an outgoing spherical wave . could operator methods , like the one deployed here , be used to sharpen / quantify this incompatibility ? and could that lead , in turn , to insight into why is it that , while tails in 4d curved spacetimes are often associated with low frequency waves interacting with the curvature of the background geometry , no such interpretation appears to be available in the zero curvature of odd dimensional flat spacetimes ?
b. p. abbott _ et al .
_ [ ligo scientific and virgo collaborations ] , `` observation of gravitational waves from a binary black hole merger , '' phys .
* 116 * , no .
6 , 061102 ( 2016 ) doi:10.1103/physrevlett.116.061102 [ arxiv:1602.03837 [ gr - qc ] ] .
b. p. abbott _ et al . _ [ ligo scientific and virgo collaborations ] , `` gw151226 : observation of gravitational waves from a 22-solar - mass binary black hole coalescence , '' phys .
lett . * 116 * , no .
24 , 241103 ( 2016 ) doi:10.1103/physrevlett.116.241103 [ arxiv:1606.04855 [ gr - qc ] ] . y. z. chu , `` transverse traceless gravitational waves in a spatially flat flrw universe : causal structure from dimensional reduction , '' phys . rev .
d * 92 * , no .
12 , 124038 ( 2015 ) doi:10.1103/physrevd.92.124038 [ arxiv:1504.06337 [ gr - qc ] ] .
l. bieri and d. garfinkle , `` perturbative and gauge invariant treatment of gravitational wave memory , '' phys .
d * 89 * , no .
8 , 084039 ( 2014 ) doi:10.1103/physrevd.89.084039 [ arxiv:1312.6871 [ gr - qc ] ] .
a. ashtekar , b. bonga and a. kesavan , `` asymptotics with a positive cosmological constant : iii . the quadrupole formula , '' phys
. rev .
d * 92 * , no .
10 , 104032 ( 2015 ) doi:10.1103/physrevd.92.104032 [ arxiv:1510.05593 [ gr - qc ] ] .
j. iliopoulos , t. n. tomaras , n. c. tsamis and r. p. woodard , `` perturbative quantum gravity and newton s law on a flat robertson - walker background , '' nucl .
b * 534 * , 419 ( 1998 ) doi:10.1016/s0550 - 3213(98)00528 - 8 [ gr - qc/9801028 ] .
a. tolish , l. bieri , d. garfinkle and r. m. wald , `` examination of a simple example of gravitational wave memory , '' phys .
d * 90 * , no . 4 , 044060 ( 2014 ) doi:10.1103/physrevd.90.044060 [ arxiv:1405.6396 [ gr - qc ] ] . y. z. chu , `` a line source in minkowski for the de sitter spacetime scalar green s function : massive case , '' class .
grav . *
32 * , no . 13 , 135008 ( 2015 ) doi:10.1088/0264 - 9381/32/13/135008 [ arxiv:1310.2939 [ gr - qc ] ] .
s. dubovsky , l. hui , a. nicolis and d. t. son , `` effective field theory for hydrodynamics : thermodynamics , and the derivative expansion , '' phys .
d * 85 * , 085029 ( 2012 ) doi:10.1103/physrevd.85.085029 [ arxiv:1107.0731 [ hep - th ] ] .
z. bern , j. j. m. carrasco and h. johansson , `` new relations for gauge - theory amplitudes , '' phys .
d * 78 * , 085011 ( 2008 ) doi:10.1103/physrevd.78.085011 [ arxiv:0805.3993 [ hep - ph ] ] .
h. kawai , d. c. lewellen and s. h. h. tye , `` a relation between tree amplitudes of closed and open strings , '' nucl .
b * 269 * , 1 ( 1986 ) .
doi:10.1016/0550 - 3213(86)90362 - 7 a. luna , r. monteiro , i. nicholson , d. oconnell and c. d. white , `` the double copy : bremsstrahlung and accelerating black holes , '' jhep * 1606 * , 023 ( 2016 ) doi:10.1007/jhep06(2016)023 [ arxiv:1603.05737 [ hep - th ] ] .
a. luna , r. monteiro , d. oconnell and c. d. white , `` the classical double copy for taub nut spacetime , '' phys .
b * 750 * , 272 ( 2015 ) doi:10.1016/j.physletb.2015.09.021 [ arxiv:1507.01869 [ hep - th ] ] .
r. monteiro , d. oconnell and c. d. white , `` black holes and the double copy , '' jhep * 1412 * , 056 ( 2014 ) doi:10.1007/jhep12(2014)056 [ arxiv:1410.0239 [ hep - th ] ] .
a. k. ridgway and m. b. wise , `` static spherically symmetric kerr - schild metrics and implications for the classical double copy , '' phys .
d * 94 * , no . 4 , 044023 ( 2016 ) doi:10.1103/physrevd.94.044023 [ arxiv:1512.02243 [ hep - th ] ] .
a. i. harte and j. vines , `` generating exact solutions to einstein s equation using linearized approximations , '' phys . rev .
d * 94 * , no .
8 , 084009 ( 2016 ) doi:10.1103/physrevd.94.084009 [ arxiv:1608.04359 [ gr - qc ] ] .
e. pajer , f. schmidt and m. zaldarriaga , `` the observed squeezed limit of cosmological three - point functions , '' phys .
d * 88 * , no .
8 , 083502 ( 2013 ) doi:10.1103/physrevd.88.083502 [ arxiv:1305.0824 [ astro-ph.co ] ] .
y. z. chu and g. d. starkman , `` retarded green s functions in perturbed spacetimes for cosmology and gravitational physics , '' phys . rev .
d * 84 * , 124020 ( 2011 ) doi:10.1103/physrevd.84.124020 [ arxiv:1108.1825 [ astro-ph.co ] ] .
|
we extend recent theoretical results on the propagation of linear gravitational waves ( gws ) , including their associated memories , in spatially flat friedmann lematre robertson walker ( flrw ) universes , for all spacetime dimensions higher than 3 . by specializing to a cosmology driven by a perfect fluid with a constant equation - of - state @xmath0 conformal re - scaling , dimension - reduction and nariai s ansatz
may then be exploited to obtain analytic expressions for the graviton and photon green s functions , allowing their causal structure to be elucidated .
when @xmath1 , the gauge - invariant scalar mode admits wave solutions , and like its tensor counterpart , must therefore contribute to the tidal squeezing and stretching of the space around a gw detector .
we identify potential 4-dimensional ( 4d ) scalar gw memory effects that propagate on its acoustic - cone " .
in addition , scalar gws in 4d radiation dominated universes like tensor gws in 4d matter dominated ones appear to yield a tail - induced memory effect that does not decay with increasing spatial distance from the source .
we then solve electromagnetism in the same cosmologies , and point out a similar tail - induced electric memory effect .
finally , in even dimensional minkowski backgrounds higher than 2 , we make a brief but explicit comparison between the linear gw memory generated by point masses scattering off each other on unbound trajectories and the linear yang - mills memory generated by color point charges doing the same and point out how there is a
double copy " relation between the two .
|
the symbiotic stars are currently understood as interacting binary systems consisting of a cool giant and a hot compact star , which is in most cases a white dwarf .
typical orbital periods are between 1 and 3 years , but they can be significantly larger . the mass loss from the giant represents the primary condition for appearance of the symbiotic phenomenon .
a part of the material lost by the giant is transferred to the more compact companion via accretion from the stellar wind or roche - lobe overflow .
this process generates a very hot ( @xmath5k ) and luminous ( @xmath6 ) source of radiation . on the basis of the way in which the generated energy is being liberated , we distinguish two phases of symbiotic binary . _ quiescent phases _ during which the hot component releases its energy approximately at a constant rate and spectral distribution
generally , we observe a wave - like variation in their lcs as a function of the orbital phase . during _ active phases _ the hot component radiation changes significantly , which leads to a 2 - 3mag brightening of the object in the optical .
a common feature of active phases is a high - velocity mass ejection .
generally , the hot radiation ionizes a fraction of the neutral circumbinary material , which gives rise to a strong nebular emission .
this component of radiation is physically displaced from the hot star and its optically thick part can be very complex in its shape . in addition , its location and shaping in the binary depend on the level of the activity . as a result
we often observe unexpected variation in the lcs as , for example , flares , drops in brightness , the effect of eclipses and outbursts .
a very interesting feature of variability in this respect is the effect of eclipses , which is very sensitive to physical displacement and radiative contributions of individual components in the system . in the case
that a significant fraction of radiation at the wavelength under consideration comes from the region which is subject to eclipse , a minimum in the lc is well observable . in the opposite case ,
the eclipse effect is very faint . therefore the eclipse effect can be observed only at specific brightness phases , at which the radiative contribution from a pseudophotosphere in the optical rivals that from the nebula .
accordingly , to reveal the above mentioned peculiarities in the lcs of symbiotic stars , it requires a very careful long - term monitoring programme . in this paper
we present the recent observational results of such our programme obtained during the period december 2001 to december 2003 .
we note that this paper continues the work of skopal _
( 2002 , hereafter s+02 , and references therein ) .
the majority of the @xmath7 measurements were performed in the standard johnson system using single - channel photoelectric photometers mounted in the cassegrain foci of 0.6-m reflectors at the skalnat pleso ( hereafter sp in tables ) and star lesn observatories ( sl ) .
values in tables represent means of the whole observing cycle .
usually , a 1-hour cycle contained about 10 to 20 individual differences between the target and the comparison .
this approach reduced the _ inner _ uncertainty of such the means to @xmath8 , @xmath9 , @xmath9 and @xmath9mag in the @xmath2 , @xmath10 , @xmath11 and @xmath12 filter , respectively .
larger uncertainties ( about 0.1mag ) during some nights are marked in tables by :. further details about observation procedure are given in skopal _
some observations in the @xmath10 and @xmath11 bands were made with the 50/70/172 cm schmidt telescope of the national astronomical observatory rozhen , bulgaria ( r ) .
other details about utilities and treatment of observations as well as standard stars were already presented in s+02 .
the @xmath13 observations of rwhya and ivvir were carried out at the san pedro mrtir observatory , baja california , mexico ( m ) , in 2003 april .
also in this case further details are given in s+02 .
observations of qwsge were performed by one of us ( zv ) at his private station belua near povask bystrica ( pb ) with a newton 180/700 telescope equipped with a ccd camera based on the texas instruments chip tc 211 ( from 01/01/2002 the chip was changed to tc-237b ) .
a set of @xmath14 filters for a modified johnson - kron - cousins system was used .
all frames were dark subtracted and flat fielded .
transformation to the international system was made by measuring the standard stars in the star - cluster m67 .
magnitudes of comparison stars were obtained in the same way as described in hric_et al . _
influence of a few arcsec distant companion ( f0v star reddened with @xmath15 = 0.20 , @xmath2 = 13.84 , @xmath10 = 13.59 , @xmath11 = 13.14 , @xmath12 = 12.53 and @xmath16 = 11.52 , munari , buson ( 1991 ) ) has not been subtracted .
in addition , 2800 visual magnitude estimates of arpav were obtained during 1982.2 2003.9 by one of us ( aj ) with a private 12".5 f/5 reflector .
other details concerning observations of arpav can be found in skopal _
we measured egand ( hd4174 , bd+39167 ) with respect to hd4143 ( sao63173 , bd+372318 ) . to obtain magnitudes in @xmath10 and @xmath11 we used the standard star hd3914 ( @xmath11 = 7.00 , @xmath17 = 0.44 ) and conversion between both stars , hd4143 hd3914 = 4.640 , 2.722 and 1.563 in the @xmath2 , @xmath10 and @xmath11 bands , respectively ( hric _ et al .
_ 1991 ) .
the data are compiled in table 1 .
figure 1 shows recent observations in @xmath2 and @xmath10 . at
/ around jd2452600 ( november 2002 ) the primary minimum occurred in the @xmath2-lc .
it is relatively narrow in profile with respect to those previously observed ( e.g. at @xmath3jd2451160 , see fig . 1 ) .
a broad wave - like variation is less pronounced and , in addition , a 0.2 0.3mag brightening in @xmath2 was detected by the latest observations in the autumn of 2003 .
these changes mean that the nebular component of radiation decreased and , instead , a pseudophotosphere with a more significant contribution in the optical was created around the central star . in such a case
the light in @xmath2 should be rather of stellar nature .
this view should be confirmed by spectroscopic observations .
this research was supported by a grant of the slovak academy of sciences no .
2/4014/4 , by allocation of san pedro mrtir observing time , from the czech - mexican project me 402/2000 and by the research plan j13/98 : 113200004 investigation of the earth and the universe .
authors thank the referee , dr . sergei shugarov , for helpful comments .
|
we present new photometric observations of egand , zand , bfcyg , chcyg , cicyg , v1329cyg , txcvn , agdra , rwhya , agpeg , axper , ivvir and the peculiar m giant v934her , which were made in the standard johnson @xmath0 system .
qwsge was measured in the kron - cousin @xmath1 system and for arpav we present its new visual estimates .
the current issue gathers observations of these objects to december 2003 .
the main results can be summarized as follows : * egand * : the primary minimum in the @xmath2 light curve ( lc ) occurred at the end of 2002 .
a 0.2 0.3mag brightening in @xmath2 was detected in the autumn of 2003 .
* zand * : at around august 2002 we detected for the first time a minimum , which is due to eclipse of the active object by the red giant .
measurements from 2003.3 are close to those of a quiescent phase .
* bfcyg * : in february 2003 a short - term flare developed in the lc .
a difference in the depth of recent minima was detected .
* chcyg * : this star was in a quiescent phase at a rather bright state .
a shallow minimum occurred at @xmath3jd 2452730 , close to the position of the inferior conjunction of the giant in the inner binary of the triple - star model of chcyg .
* cicyg * : our observations cover the descending branch of a broad minimum .
* txcvn * : at / around the beginning of 2003 the star entered a bright stage containing a minimum at @xmath3jd 2452660 . *
agdra * : new observations revealed two eruptions , which peaked in october 2002 and 2003 at @xmath39.3 in @xmath2 . * arpav * : our new visual estimates showed a transient disappearance of a wave - like modulation in the star s brightness between the minima at epochs e = 66 and e = 68 and its reappearance . * agpeg * : our measurements from the end of 2001 showed rather complex profile of the lc . *
rwhya * : observations follow behaviour of the wave - like variability of quiet symbiotics . *
axper * : in may 2003 a 0.5mag flare was detected following a rapid decrease of the light to a minimum . * qwsge * : ccd observations in @xmath4 bands cover a period from 1994.5 to 2003.5 .
an increase in the star s brightness by about 1mag was observed in all passbands in 1997 .
less pronounced brightening was detected in 1999/2000 . * v934her * : our observations did not show any larger variation in the optical as a reaction to its x - ray activity .
|
likely , the large hadron collider or lhc will become operative by the beginning of the next century , offering a wide programme of exciting possibilities in particle physics , among which the origin of mass at the electroweak scale and the possible discovery of supersymmetric particles , the investigation of cp violation in b mesons and detailed studies of top quark physics . other interesting topics related to charm and beauty flavours will be covered as well benefitting of a foreseen huge statistics to be collected with the machine running even at @xmath2low " luminosity ( @xmath3 @xmath4 ) @xcite @xcite . following this line of research
, we shall focus in this paper on inclusive hadroproduction of heavy resonances , in particular charmonia @xmath0 and @xmath1 states , whose relevance in the study of perturbative and non - perturbative qcd will be briefly reviewed .
to first approximation , heavy - flavour hadroproduction can be treated in the framework of perturbative qcd due to the relatively large quark masses @xcite and consequently the partonic production cross sections can be convoluted with parton distribution functions ( pdf s ) of the colliding protons .
this scheme can indeed be viewed as well founded since the underlying parton interaction is hard , therefore providing a reasonable justification for such a factorization of the production process .
higher - order corrections to the matrix elements of the short - distance processes may be incorporated likely improving the accuracy of the predicted cross sections .
even complex effects such as multiple gluon emission can be treated analytically according to resummation techniques @xcite .
nevertheless , such an essentially analytic approach may eventually oversight the real complexity of the full hadronic collision and a certain amount of modifications should be added if a more realistic description is required .
for example , parton evolution either at the initial or final state together with fragmentation and recombination of the beam remnants could play an important rle leading to substantial modifications in the final state , even for inclusive processes . on the other hand ,
event generators based on a monte carlo framework represent an alternative philosophy of coping with the complexity of very high energy hadronic reactions by dividing the full problem in a set of components to be dealt separately in a simulation chain .
the hard partonic interaction , constituting the underlying or skeleton process of the simulation , is dressed up with soft processes leading to hundreds of final particles .
certainly , such an approach is plagued with various sources of uncertainty , especially related to the long - distance aspects of the strong interaction dynamics , associated to the assumptions and modelling of the generation .
however , the impressive amount of collected experimental data used in a heuristic way to _ tune _ those parameters of the event generator , gives reliability to this method to a considerable degree providing a powerful tool to interpret current data and make sensible predictions for future experiments . in
this work we re - examine prompt production of heavy resonances at high - energy hadron colliders by means of the event generator pythia @xcite .
undoubtedly , there is a widespread recognition that the study of heavy quarkonia has been playing an important rle over the past decades in the development of the theory of the strong interaction and such beneficial influence seems far from ending .
moreover , a precise understanding of prompt charmonia production offer several advantages from both experimental and theoretical points of view at the lhc . the fact that both @xmath0 and @xmath1 resonances can decay into a oppositely charged lepton pair with a respectable branching fraction ( especially the former ) , permits its discrimination in a huge hadronic background . actually , charmonia production can be viewed as interesting in its own right for precise tests of both perturbative and non - perturbative qcd , search for exotic states or as a probe of the parton content of the proton .
furthermore , completely different processes as for example the decay @xmath5 , a golden channel to investigate cp violation in b decays , require a precise estimate of sources of background coming from random combinations involving prompt @xmath0 s .
on the other hand , the experimental discovery at fermilab @xcite of an excess of inclusive production of prompt heavy quarkonia ( mainly for @xmath0 and @xmath6 resonances ) in antiproton - proton collisions triggered an intense theoretical activity beyond what was considered conventional wisdom until recently @xcite @xcite .
the discrepancy between the so - called colour - singlet model ( csm ) in hadroproduction @xcite and the experimental data amounts to more than an order of magnitude and can not be attributed to those theoretical uncertainties arising from the ambiguities on the choice of a particular parton distribution function , the heavy quark mass or different energy scales .
the colour - singlet model assumes that the heavy quark pair is produced in a colour - singlet state with the right quantum numbers already during the hard collision .
the production cross section involves a perturbative calculation where the proper spin and angular momentum states are projected out , and the non - perturbative component of the hadronization comes from a single parameter to be identified with the wave function of the resonance at the origin ( or its derivatives for non @xmath7-wave states ) . the fact that the latter can be derived from potential models fitting the charmonia spectrum or from their leptonic decay modes , gives its predictive power to the colour - singlet model .
however , as mentioned above , such predictions stand quite below the experimental yield at high-@xmath8 , so that some modifications have to be done .
in addition , it should be emphasized that the model is clearly incomplete since relativistic corrections due to the relative velocity of quarks in the bound state are ignored - expected to be sizeable especially for charmonia .
moreover , another hint of the incompleteness of the csm comes from the existence of infrared divergences in the production cross section for @xmath9-wave states , which can not be absorbed in its non - perturbative factor . in order to resolve all these difficulties ,
it has been recently argued that the heavy quark pair not necessarily has to be produced in a colour - singlet state at the short - distance partonic process itself @xcite .
alternatively , it can be produced in a colour - octet state evolving non - perturbatively into quarkonium in a specific final state with the correct quantum numbers according to some computable probabilities governed by the internal velocity of the heavy quark .
this mechanism , usually named as the colour - octet model ( com ) , can be cast into the rigorous framework of an effective non - relativistic theory for the strong interactions ( nrqcd ) deriving from first principles @xcite .
however , the weakness of the com lies in the fact that the non - perturbative parameters characterizing the long - distance hadronization process beyond the colour - singlet contribution ( i.e. the colour - octet matrix elements ) are so far almost free parameters to be adjusted from the fit to experimental data , though expected to be mutually consistent according to the nrqcd power scaling rules , or to resort to non - perturbative models @xcite . on the other hand ,
an attractive feature of the colour - octet hypothesis consists of the universality of the nrqcd matrix elements ( me ) entering in other charmonium production processes like photoproduction @xcite .
let us look below in some detail at the way hadronization is folded with the partonic description of hadrons , focusing for concreteness on the couple of related papers @xcite . in the first paper @xcite , cho and leibovich considered for charmonium production ( in addition to the csm ) the @xmath10 coloured intermediate state as a first approach , computing the matrix elements as products of perturbative parts for the short - distance partonic processes , and the colour - octet me concerning the long - distance hadronization .
finally , a convolution of concrete parton distribution functions and the differential cross - section for the @xmath11 production subprocess was performed , whereby the @xmath8 dependence of the charmonia production exclusively coming from the latter . in ref .
@xcite the same authors take into account further contributions from new coloured states ( for more details see the quoted references ) concluding finally that at high enough @xmath8 a two - parameter fit is actually required to explain the observed inclusive @xmath8 distribution of charmonia production at the fermilab tevatron .
notice , however , that such calculations were carried out based on a possibly oversimplified picture of the hadronic interaction . indeed , it is well - known for a long time that higher - order effects ( @xmath12 factors ) play an important rle in inclusive hadroproduction . in particular , beyond the primordial transverse momentum @xmath13 of partons in hadrons related to fermi motion relevant at small @xmath8 , initial - state radiation of gluons by the interacting partons
add up to yield an _ effective _ intrinsic transverse momentum which certainly has to be considered in hadroproduction at high @xmath8 @xcite .
as we shall see , if overlooked at all , the effect on the fit parameters ( and ultimately on the colour - octet me s ) amounts to a _
systematic _ overestimate .
in fact , one should distinguish the primordial transverse momentum of partons owing to their fermi motion ( a non - perturbative effect ) from the perturbative contribution dynamically generated via gluon radiation generally implemented in the event generators by means of a parton shower algorithm @xcite .
by _ effective _
@xmath13 we merge both effects under a common name , though the former is actually overshadowed by the latter at high @xmath8 .
of course , the pythia treatment of the effective @xmath13 is not guaranteed to be perfect but , nevertheless , should give a reasonable estimate of such effects .
in this work we have extended our earlier analysis of @xmath0 hadroproduction at colliders @xcite @xcite to the @xmath1 state , affording a more complete test upon the validity of the nrqcd power scaling rules relating different long - distance matrix elements .
the paper is organized as follows . in section 2
we describe the hard processes considered for charmonia hadroproduction and we discuss their implementation in the event generator pythia , described in more detail in the appendix at the end of the paper where we also present a code for a fast simulation of charmonia using pythia . in section 3
we perform new fits to tevatron data on prompt @xmath0 and @xmath1 direct production extracting nrqcd matrix elements from them .
we finally make some predictions for charmonia direct production at the lhc in section 4 .
originally , the event generator pythia @xcite produces direct @xmath0 s and higher @xmath14 resonances in hadron - hadron collisions according to the csm .
however , as already mentioned this model fails off to account for the charmonia production rate at the fermilab tevatron by more than an order of magnitude .
consequently , we have implemented a code in pythia in order to include the colour - octet mechanism for @xmath0 hadroproduction ( @xmath6 generation has been based upon the same set of production channels ) via the following @xmath15 partonic processes : * @xmath16 * @xmath17 * @xmath18 we have included in the simulation the @xmath19 , @xmath20 and @xmath21 ( @xmath22 ) states in @xmath10 charmonia hadroproduction according to the com @xcite .
the squared amplitudes , taken from refs .
@xcite are reproduced below .
firstly let us consider the @xmath19 intermediate state : @xmath23 ^ 2 } \nonumber \\ & { \times } & [ ( \hat{t}^2+\hat{t}\hat{u}+\hat{u}^2)^2-m^2(\hat{t}+\hat{u } ) ( 2\hat{t}^2+\hat{t}\hat{u}+2\hat{u}^2)+m^4(\hat{t}^2+\hat{t}\hat{u}+\hat{u}^2 ) ] \nonumber \\ & { \times } & < 0{\mid}o_8^{j/\psi}(^3s_1){\mid}0 > \\ \nonumber \\ \overline{\sum}{\mid}{\cal a}(gq{\rightarrow}{j/\psi}q){\mid}^2 & = & -\frac{16{\pi}^3{\alpha}_s^3}{27m^3}\ \frac{\hat{s}^2+\hat{u}^2 + 2m^2\hat{t } } { \hat{s}\hat{u}(\hat{t}-m^2)^2}[4(\hat{s}^2+\hat{u}^2)-\hat{t}\hat{u}]\
< 0{\mid}o_8^{j/\psi}(^3s_1){\mid}0 > \nonumber \\ & & \\
\overline{\sum}{\mid}{\cal a}(q\overline{q}{\rightarrow}{j/\psi}g){\mid}^2 & = & \frac{128{\pi}^3{\alpha}_s^3}{81m^3}\ \frac{\hat{t}^2+\hat{u}^2 + 2m^2\hat{s } } { \hat{t}\hat{u}(\hat{s}-m^2)^2}[4(\hat{t}^2+\hat{u}^2)-\hat{t}\hat{u}]\ < 0{\mid}o_8^{j/\psi}(^3s_1){\mid}0 > \nonumber \\ & & \end{aligned}\ ] ] where the barred summation symbol refers to average over initial and final spins and colours ; @xmath24 , @xmath25 and @xmath26 stand for the mandelstam variables of the short - distance subprocesses .
we shall set @xmath27 for both @xmath0 and @xmath1 resonances in accordance with the assumption on the non - relativistic nature of heavy quarkonium made in the theoretical calculation of these amplitudes @xcite , i.e. vanishing relative momentum of the quarks in the bound state .
the corresponding expressions for the @xmath20 contributions are : @xmath28 ^ 2 } \nonumber \\ & { \times } & [ \hat{s}^2(\hat{s}-m^2)^2+\hat{s}\hat{t}\hat{u}(m^2 - 2\hat{s})+ ( \hat{t}\hat{u})^2]\ < 0{\mid}o_8^{j/\psi}(^1s_0){\mid}0 > \nonumber \\ & & \\
\overline{\sum}{\mid}{\cal a}(gq{\rightarrow}{j/\psi}q){\mid}^2 & = & -\frac{40{\pi}^3{\alpha}_s^3}{9m}\ \frac{\hat{s}^2+\hat{u}^2 } { \hat{t}(\hat{t}-m^2)^2}\ < 0{\mid}o_8^{j/\psi}(^1s_0){\mid}0 > \\ \overline{\sum}{\mid}{\cal a}(q\overline{q}{\rightarrow}{j/\psi}q){\mid}^2 & = & \frac{320{\pi}^3{\alpha}_s^3}{27m}\ \frac{\hat{t}^2+\hat{u}^2 } { \hat{s}(\hat{s}-m^2)^2}\ < 0{\mid}o_8^{j/\psi}(^1s_0){\mid}0 > \end{aligned}\ ] ] with regard to the @xmath21 contributions , they display altogether a similar ( i.e. degenerate ) transverse momentum behaviour as the @xmath20 component for @xmath29 gev @xcite .
thus from a pragmatic point of view , the generation of charmonia via intermediate @xmath9-wave coloured states becomes superfluous although they must properly be taken into account in the computation of the overall cross section as explained in more detail in the appendix .
we find from our simulation that gluon - gluon scattering actually stands for the dominant process as expected , the gluon - quark scattering contributes appreciably @xmath30 , whereas the quark - antiquark scattering represents a tiny contribution .
therefore we shall omit hereafter any further reference to the @xmath31 production channel . using the same numerical values for the colour - octet matrix elements as reported in tables i and ii of ref .
@xcite , if initial- and final - state radiation are turned off , there is a good agreement between the theoretical curve and the experimental points for both @xmath0 and @xmath6 ( see fig .
1 ) , as should reasonably be expected .
however , if initial- and final - state radiation spectrum , the former enhancing the high @xmath8 tail whereas the latter softens the distribution .
indeed , in considering the process @xmath32 in pythia only the gluon evolves in the final state , though the energy ( and momentum ) of charmonium is modified as a consequence of the final - state machinery @xcite ] are switched on , the predicted curve stands well above the experimental data , in accordance with the expected @xmath13-kick caused by the _ effective _ intrinsic transverse momentum of partons @xcite .
accordingly , keeping radiation effects in the theoretical analysis it turns out that the values for the colour - octet me s have to be lowered by significant factors . for the sake of clarity
it is important to point out that the inclusion in the event generation of multi - gluon emission by itself almost does not change the overall cross - section for charmonia production obtained from pythia .
( there is however an indirect bias effect slightly increasing the cross section @xmath33 with radiation on relative to radiation off because of the kinematical cut on the charmonia pseudorapidity . )
the main effect remains in the smearing of the transverse momentum distribution leading to the enhancement of the high-@xmath8 tail .
in order to assess the importance of the effective intrinsic transverse momentum of partons we have made three different choices for the proton pdf : a ) : : the leading order cteq 2l ( by default in pythia 5.7 ) b ) : : the next to leading order mrsd0 ( the same as used in @xcite ) c ) : : the next to leading order grv 94 ho as mentioned before , the theoretical curves for the inclusive @xmath8 distribution of prompt @xmath0 and @xmath1 stand in all cases above tevatron experimental points if the set of parameters from @xcite are blindly employed in the pythia generation with initial - state radiation on . motivated by this systematic discrepancy
, we have performed new fits for prompt @xmath0 and @xmath1 direct production at the tevatron ( feed - down to @xmath0 from radiative decay of @xmath14 resonances was experimentally removed ) .
those colour - octet matrix elements extracted from our analysis on @xmath0 production are reported in table 1 . in fig .
2a we plot the resulting fit to tevatron data for the cteq pdf , showing separately the three different components taken into account in this work . as a figure of merit for the overall fit we found @xmath34 . in the case of @xmath6 production ,
the respective me s are reported in table 2 and the corresponding plot for the cteq pdf is shown in figure 2b , with @xmath35 . [
cols="<,^,^",options="header " , ] notice that all the me s are lowered systematically with respect to the values reported in ref .
@xcite for both @xmath0 and @xmath6 . ) through the interpolating formula ( 2.22 ) of ref .
however , final state radiation was incorporated in our analysis in contrast to ref .
@xcite , softening the @xmath8 inclusive distribution as mentioned in footnote @xmath363 .
the neglect of those higher order qcd effects should mainly affect the high-@xmath8 region , hence the @xmath10 matrix element which could be somewhat underestimated . ]
basically this is because initial - state radiation enhances the differential cross section for @xmath37 gev , so that the non - perturbative parameters have to be re - adjusted to restore the fit to the experimental points .
let us also remark that the combined @xmath20 , @xmath38 me s are lowered by larger factors than those for the @xmath19 contribution .
consequently , the power counting rules derived from nrqcd are more faithfully satisfied in the overall than in ref .
@xcite - all of them should scale like @xmath39 .
it is especially significant that the numerical values obtained in our analysis for the @xmath20+@xmath38 me s are smaller by an order of magnitude with respect to those found in ref .
@xcite in accordance with the common belief widely expressed in literature @xcite . in this sense , theoretical expectations from the com might be reconciled with hera data on @xmath0 photoproduction @xcite .
the contributions from the colour - octet @xmath19 and @xmath20 + @xmath21 states to the differential cross section for @xmath0 and @xmath6 production are shown separately in fig .
2 . notice that the @xmath20 and @xmath21 components dominate in both cases over the other production channels at moderate @xmath8 , whereas the @xmath19 channel takes over at large @xmath8 becoming asymptotically responsible of the full production rate .
let us remark the especially relative smaller numerical values of the long - distance @xmath20 , @xmath21 matrix elements regarding the grv pdf .
this is in a qualitative agreement with the fact that the mrsd0 and cteq2l parton distribution functions give smaller values for the gluon distribution function at small partonic @xmath40 than the grv pdf ( the gluon density of the latter is steeper at small @xmath40 ) .
hence , a gluon density with a steeper behaviour at small @xmath40 steepens the @xmath8-spectrum for the @xmath19 component amounting to a larger production rate at moderate @xmath8 , decreasing the @xmath20+@xmath21 contribution from the combined fit to the experimental inclusive distribution .
reliable calculations of inclusive production rates of charmonia are of considerable interest at lhc experiments in many respects @xcite , as briefly outlined in the intorduction .
therefore , an order - of - magnitude estimate of its production rate becomes very valuable at the present stage of the lhc experiments .
keeping in mind such a variety of interests , we have generated direct @xmath0 s and @xmath1s in proton - proton collisions at lhc energies ( center - of - mass energy = 14 tev ) by means of our @xmath41modified " version of pythia with the colour - octet me s as shown in tables 1 and 2 - i.e. after normalization to tevatron data . the different contributions to @xmath0 hadroproduction are depicted altogether in fig . 3 for the cteq pdf .
figure 3 also illustrates the @xmath8 inclusive distributions of direct prompt @xmath0 s at the lhc obtained for the three pdf s employed in our study . comparing them with the distribution obtained by sridhar @xcite , we find our predictions standing below his theoretical curve at large @xmath8 .
in fact this is not surprising since sridhar considered only the @xmath19 coloured intermediate state , whose me was assumed larger than the one used in this work .
the argument ends by noticing that at high enough @xmath8 the dominant production comes from the @xmath19 channel as the combined @xmath20 + @xmath21 contribution falls off faster than the former . in fig .
4 we plot our prediction for direct @xmath6 production at the lhc using the cteq pdf , together with the @xmath0 curve . from the figure it is clear that muon pair production from direct @xmath0 is larger by more than an order of magnitude from @xmath0 as compared to @xmath1 .
it should be mentioned that our prediction for the @xmath1 production rate differs substancially with respect to the corresponding theoretical curve by sridhar @xcite all along the @xmath8 range - by an order of magnitude even at moderate @xmath8 in contrast to the @xmath0 case .
in this paper we report on the extraction of long - distance colour - octet matrix elements in a monte carlo framework from charmonia production at hadron colliders , already studied in our previous works @xcite @xcite , extending our analysis to the @xmath6 resonance .
we explain in more detail the procedure followed in our investigation , also providing a code for a @xmath2fast " simulation of charmonium in hadron collisions based on the colour - octet model and compatible with the tevatron results .
all three possible lowest order short - distance processes contributing to @xmath0 hadroproduction have been taken into account , namely , @xmath42 , @xmath43 and @xmath44 , contributing at order @xmath15 to charmonia production including the colour - octet mechanism .
the same set of channels have been considered for @xmath6 hadroproduction .
we have investigated higher - order effects induced by initial - state radiation , concluding that the overlooking of initial radiation leads systematically to a significant overestimate of the fit parameters .
the new colour - octet matrix elements obtained for three different pdf s from tevatron data are shown in tables 1 and 2 , which have to be regarded as reasonable estimates due to the experimental and theoretical uncertainties .
in particular , the numerical values for the combined @xmath20 + @xmath38 matrix elements are smaller by an order of magnitude than those obtained in ref .
@xcite and might resolve the difficulties of the com found in @xmath0 photoproduction at hera @xcite .
we thank s. baranov , p. eerola , n. ellis and m. smizanska and the atlas b physics working group for useful comments and an encouraging attitude .
comments by e. kovacs , t. sjstrand and k. sridhar are also acknowledged .
we have modified the pythia routine pysigh ( isub=86 ) @xcite for the partonic process : @xmath45 including those contributions from the @xmath19 , @xmath20 and @xmath21 coloured states according to the expressions ( 1 ) and ( 4 ) .
the me s of the latter components are related through the heavy quark spin symmetry at leading order in @xmath46 : @xcite @xmath47 moreover , it should be stressed that the colour - octet me s associated to the @xmath20 and the @xmath21 channels can not be disentangled from tevatron experimental data since both are degenerate in the sense that they display a similar @xmath8 behaviour for @xmath29 gev @xcite .
all three @xmath9-wave contributions altogether amount to about three times the @xmath48 contribution when setting @xmath49 .
therefore one may write the following approximate equality : @xmath50\ + \sum_{j=0,1,2}\frac{d\sigma}{dp_t}[^3p_j^{(8)}]\ { \simeq}\ 3\ { \times } \frac{1}{<o_8(^1s_0)>}\ \frac{d\sigma}{dp_t}[^1s_0^{(8)}]\ { \times}\ \biggl(\ \frac{<o_8(^1s_0)>}{3}+\frac{<o_8(^3p_0)>}{m_c^2}\ \biggr)\ ] ] and only the @xmath20 component has to be simulated in practice .
on the other hand , we have generated with pythia an event sample via the gluon - quark subprocess @xmath51 according to the expressions ( 2 ) and ( 5 ) , @xmath2adapting " for this end the pysigh routine isub=112 .
we found that this production channel amounts to a fraction of about @xmath52 of the total yield of direct @xmath0 s , almost constant over the @xmath8 range under examination ( see fig .
consequently , one may reduce considerably cpu time in the event generation by a slight change on the @xmath32 routine , including @xmath53 scattering by means of a scaling factor . let us note however that all the fits and plots shown in this work were performed using the @xmath53 event sample obtained from pythia as explained above . the routine presented below permits a fast generation of direct @xmath0 s ( or @xmath1 with some few changes in the values of the parameters and variables such as parp(38 ) , sqm3 , ... ) in hadronic collisions at the lhc .
elseif(isub.eq.86 ) then c ... g + g - > j / psi + g. c c ... color - octet matrix elements for j / psi c ( cteq 2l ) c sqm3=(2.*1.48)**2 xo83s1=0.0033 xo81s3p=0.0048 c c color singlet c facqqg1=comfac*as**3*(5./9.)*parp(38)*sqrt(sqm3 ) * & ( ( ( sh*(sh - sqm3))**2+(th*(th - sqm3))**2+(uh*(uh - sqm3))**2)/ & ( ( th - sqm3)*(uh - sqm3))**2)/(sh - sqm3)**2 c c color octet ( 3-s-1 ) c facqqg2=-comfac*as**3*(1./18.)*3.1416/sqrt(sqm3 ) * & ( 27.*(sh*th+th*uh+uh*sh)-19.*sqm3**2)*((th**2+th*uh+uh**2)**2- & sqm3*(th+uh)*(2.*th**2+th*uh+2.*uh**2)+sqm3**2 * & ( th**2+th*uh+uh**2))*xo83s1/ & ( ( th - sqm3)*(uh - sqm3))**2/(sh - sqm3)**2 c c color octet ( 1-s-0)/3 + ( 3-p-0 ) ! ! ! c facqqg3=1000.*comfac*(5.*3.1416*as**3/(4.*sqrt(sqm3)))/sh * & ( sh**2*(sh - sqm3)**2+sh*th*uh*(sqm3 - 2.*sh)+(th*uh)**2)/(th*uh ) * & ( ( sh**2-sqm3*sh+sqm3**2)**2- & th*uh*(2.*th**2 + 3.*th*uh+2.*uh**2))*xo81s3p/ & ( ( th - sqm3)*(uh - sqm3))**2/(sh - sqm3)**2 facqqg3=facqqg3/1000 .
c c ... scaling factor for g + q - > j / psi + q c scalgq=1.25 c if(kfac(1,21)*kfac(2,21).ne.0 ) then nchn = nchn+1 isig(nchn,1)=21 isig(nchn,2)=21 isig(nchn,3)=1 sigh(nchn)=facqqg1 nchn = nchn+1 isig(nchn,1)=21 isig(nchn,2)=21 isig(nchn,3)=1 sigh(nchn)=facqqg2*scalgq nchn = nchn+1 isig(nchn,1)=21 isig(nchn,2)=21 isig(nchn,3)=1 sigh(nchn)=3.*facqqg3*scalgq endif .... a lower cut - off was used throughout in the event generation ( @xmath56 gev by default in pythia for processes singular at @xmath57 ) thereby avoiding the problematic low-@xmath8 region for charmonia hadroproduction .
the effect of such a cut on the experimentally observed @xmath8 range ( @xmath37 gev ) is rather small , representing a slight underestimation of the production rate in the event generation with radiation on as a consequence of the isr effect @xcite .
lastly , let us point out that the theoretical curves appearing in all our plots were obtained from histograms ( filled with the corresponding pythia ouput data ) fitted by means of the four - parameter @xmath8 function : @xmath58\ = \ { \alpha}_1\ \frac{p_t^{{\alpha}_2}}{({\alpha}_3+p_t^2)^{{\alpha}_4}}\ ] ] providing an excellent fit in the whole @xmath8 range examined .
technical proposal of the atlas collaboration , cern / lhcc 94 - 43 ( 1994 ) technical proposal of the cms collaboration , cern / lhcc 94 - 38 ( 1994 ) p. nason , s. dawson , r.k .
ellis , nucl .
phys . * b303 * ( 1988 ) 607 s. catani , nucl .
b ( proc . suppl . ) * 54a * ( 1997 ) 107 t. sjstrand , comp . phys . comm .
* 82 * ( 1994 ) 74 cdf collaboration , f. abe at al .
* 69 * ( 1992 ) 3704 ; d0 collaboration , v. papadimitriou et al .
, fermilab - conf-95/128-e m. cacciari , m. greco , m.l . mangano and a. petrelli ,
* b356 * ( 1995 ) 553 m. beneke , cenr - th/97 - 55 ( hep - ph/9703429 ) r. baier , r. rckl , z. phys . *
c19 * ( 1983 ) 251 g.a .
schuler , cern - th-7170 - 94 ( hep - ph/9403387 ) e. braaten and s. fleming , phys .
* 74 * ( 1995 ) 3327 g.a .
schuler , z. phys .
* c71 * ( 1996 ) 317 g.t .
bodwin , e. braaten , g.p .
lepage , phys . rev .
* d51 * ( 1995 ) 1125 g.a .
schuler , cern - th/97 - 12 ( hep - ph/9702230 ) m. cacciari and m. krmer , phys .
* 76 * ( 1996 ) 4128 p. cho and a.k .
leibovich , phys . rev .
* d53 * ( 1996 ) 150 p. cho and a.k .
leibovich , phys . rev .
* d53 * ( 1996 ) 6203 l. rossi , nucl .
b ( proc . suppl . )
* 50 * ( 1996 ) 172 ; j. huston et al .
, phys . rev .
* d51 * ( 1996 ) 6139 ; s. frixione , m.l .
mangano , p. nason , g. rodolfi , nucl .
b431 * ( 1994 ) 453 .
; a. breakstone et al . , z. phys .
* c52 * ( 1991 ) 551 ; w.m .
geist at al , phys .
* 197 * ( 1990 ) 263 ; m. della negra et al .
, nucl . phys .
* b127 * ( 1977 ) 1 t. sjstrand , phys . lett . * b157 * ( 1985 ) 321 m.a .
sanchis - lozano and b. cano , to appear in nucl .
( hep - ph/9611264 ) b. cano and m.a .
sanchis - lozano , ific/97 - 01 ( hep - ph/9701210 ) to appear in phys .
b h. plothow - besch , @xmath59pdflib : nucleon , pion and photon parton density functions and @xmath60 , users s manual - version 6.06 , w5051 pdflib ( 1995 ) cern - ppe j.p .
ma , um - p-96/74 ( hep - ph/9705445 ) ; p. ko , snutp 96 - 084 ( hep - ph/9608388 ) k. sridhar , mod .
a11 * ( 1996 ) 1555
|
we present the results of our analysis on charmonia ( @xmath0 and @xmath1 ) hadroproduction taking into account higher - order qcd effects induced by initial - state radiation in a monte carlo framework , with the colour - octet mechanism implemented in the event generation .
we find that those colour - octet matrix elements extracted so far from fermilab tevatron data for both @xmath0 and @xmath1 production have to be lowered significantly .
we finally make predictions for charmonia production at the lhc , presenting a simple code for a _ fast _ simulation with pythia based on the colour - octet model .
= -.5 cm = 0.cm = 0.cm = 22.cm = 15.6 cm // + : quarkonia production ; color - octet ; nrqcd ; lhc ; tevatron
|
in recent years , a considerable effort has been devoted to extend the chaotic communication applications to the field of secure communications .
the possibility of synchronization of two coupled chaotic systems was first shown by pecora and carrol @xcite and opened the possibility of using the signals generated by chaotic systems as carriers for analog and digital communications .
this discovery soon aroused great interest as a potential means for secure communications @xcite .
accordingly , a great number of cryptosystems based on chaos have been proposed @xcite , some of them fundamentally flawed by a lack of robustness and security @xcite .
projective synchronization ( ps ) is an interesting phenomena firstly described by mainieri and rehacek @xcite , it consists in the synchronization of two partially linear coupled chaotic systems , master and slave , in which the amplitude of the slave system is a scalar multiple , called scaling factor , of that of the master system in the phase space .
the original study was restricted to three - dimensional partially linear systems .
later xu and li @xcite showed that ps could be extended to general classes of chaotic systems without partial linearity , by means of the feedback control of the slave system ; they illustrated the applicability to lorenz , chua and the hyperchaotic rssler systems .
in a recent paper li and xu @xcite proposed a secure communication scheme based on ps chaotic masking .
the authors claimed that the unpredictability of the scaling factor of the ps can additionally enhance the security of communications .
furthermore , the authors proposed the use of an invertible function @xmath0 , in such a way that the transmitted ciphertext signal will be @xmath1 $ ] , where @xmath2 and @xmath3 are the variables of a three - dimensional chaotic system and @xmath4 is the plaintext ( which in their paper is assumed to be a sound signal ) .
they claimed that the security of the information can also be guaranteed because the function @xmath0 could be arbitrarily chosen .
they illustrated the feasibility of the scheme with two examples , based on the lorenz and the hyperchaotic rssler systems , respecitvely . in this article
we show that the proposed cryptosystem is insecure and how to break it in two different ways , by high - pass filtering and by generalized synchronization , for both examples based on the lorenz and the hyperchaotic rssler systems . in @xcite , the first example
is based on the following lorenz system : @xmath5 with parameter values @xmath6 .
the transmitted signals from the sender to the receiver end are the shared scalar variable @xmath3 and the ciphertext @xmath1 $ ] , where the function @xmath0 was specified as @xmath7=y_1+m(t)$ ] .
the second example is based on the hyperchaotic rssler system defined by the authors as : @xmath8 with parameter values @xmath9 . in this example , the transmitted signals from the sender to the receiver end are the shared scalar variable @xmath10 and the ciphertext @xmath11 $ ] , where @xmath0 was specified as @xmath12=y_1+m(t)$ ] .
although the authors seem to base the security of its communication system on the chaotic behavior of the output of a chaotic or hyperchaotic non - linear system , no analysis of security was included . instead , an unproved assertion saying that the security of information can be guaranteed " was given in the conclusion .
the first issue to be considered is the key of the system .
a cryptosystem can not exist without a key , otherwise , it might be considered as a coding system , but never regarded as a secure system . in @xcite it is not considered whether there should be a key in the proposed system , what it should consist of , what the available key space would be , and how it would be managed . when cryptanalyzing a cryptosystem , the general assumption made is that the cryptanalyst knows exactly the design and working of the cryptosystem under study , i.e. , he knows every detail about the ciphering algorithm , except the secret key .
this is an evident requirement in today s secure communications systems , usually referred to as kerchoff s principle @xcite . in @xcite
it was stated that the arbitrary selection of the function @xmath0 will warrant the information security . but according to the kerchoff s principle , the function @xmath0 must be publicly known and may not be considered part of the key . at most , its structure could contain some factors or constants whose values can play the role of secret key ; but , unfortunately , the authors of @xcite have not considered such possibility , nor which conditions the function @xmath0 might satisfy , nor how many usable functions there are , nor how much they can contribute to the system security .
much care must be exercised when selecting the function @xmath0 .
otherwise , choosing different functions might create different ciphertexts , that can be decrypted using the same algorithm though , as shown in sec .
[ sec : recovery ] .
it is supposed that chaotic masking is an adequate means for secure transmission , because chaotic systems present some properties as sensitive dependence on parameters and initial conditions , ergodicity , mixing , and dense periodic points .
these properties make them similar to pseudorandom noise @xcite , which has been used traditionally as a masking signal for cryptographic purposes .
a fundamental requirement of the pseudorandom noise used in cryptography is that its spectrum should be infinitely broad , flat and of much higher power density than the signal to be concealed . in other words
, the plaintext power spectrum should be effectively buried into the pseudorandom noise power spectrum .
the cryptosystem proposed in @xcite does not satisfy this condition . on the contrary ,
the spectrum of the signal generated by the lorenz oscillator is of narrow band , decaying very fast with increasing frequency , showing a power density much lower than the plaintext at plaintext frequencies . in @xcite
the sound of a water flow was used as the plaintext message @xmath4 , but no details are given about its waveform or power spectrum . from (
2 ) it can be appreciated that its amplitude is roughly @xmath13 . in our simulation
we have used , instead , a well defined plaintext signal @xmath14 , which corresponds to a pure tone sound of 16.352 hz , which is the lowest note generated by a musical instrument , the @xmath15 of a 32 ft pipe of a pipe - organ @xcite and with the same peak amplitude of ( * ? ? ?
2 ) , namely @xmath13 .
figures [ fig : lorenz](a ) and [ fig : rosler](a ) illustrate the logarithmic power spectra of the ciphertext when the lorenz attractor and the hyperchaotic rssler attractor are used as the chaotic system , respectively , with the same parameter values previously described .
it can be seen that in both examples the plaintext signal clearly emerges at 16.352 hz over the background noise created by the lorenz and hyperchaotic rssler oscillators , with a power of @xmath16 and @xmath17 , respectively , relative to the maximum power of the ciphertext spectrum , while the power density of the ciphertext , at neighboring frequencies , falls below @xmath18 and @xmath19 , respectively . to recover the plaintext
we did not use a chaotic receiver , instead the ciphertext was high - pass filtered .
the procedure is illustrated in fig .
[ fig : lorenz](b ) and [ fig : rosler](b ) .
the filter employed was a finite impulse response one . to avoid phase nonlinearities and distortion ,
it was constructed with a 512-coefficient hamming window , with a cutoff frequency of 13 hz .
the result is a good estimation of the plaintext after a short initial transient of approximately 0.3 seconds for the lorenz system .
note that this is the hardest case an attacker can face from the point of view of plaintext frequency , because for higher sound frequencies the spectrum of the background noise created by the lorenz oscillator is even lower . while for the hyperchaotic rssler system the result is a perfect recovery after a short initial transient of approximately 0.1 seconds .
the former attack method works only for plaintext frequencies higher than the 13 hz cut - off frequency of the high - pass filter employed . for very low plaintext frequencies the noise created by the chaotic oscillators effectively masks the plaintext , preventing its retrieval by direct high - pass filtering .
but plaintext signals of very low frequency may be still retrieved if we know what kind of non - linear time - varying system was used for encryption , but without the knowledge of its parameter and initial condition values . to show such a possibility we have implemented two cryptanalysis procedures based on generalized synchronization @xcite . to break the ps - based chaotic masking scheme under study , using the lorenz system , we use the following intruder receiver : @xmath20 where @xmath21 are the intruder receiver s lorenz system parameters and @xmath22 is the instantaneous value of the recovery error @xmath23 . the terms @xmath24 and @xmath25 work as feedback of the recovery error , in order to achieve generalized synchronization between sender and receiver ; @xmath26 and @xmath27 are two scalars that may accept a wide range of values , from 1 to more than 400 , and even one of them alone may not exist . the synchronism is achieved for any combination of @xmath26 and @xmath27 values , but the amplitudes of @xmath28 and @xmath29 do not match with those of @xmath30 and @xmath31 while @xmath32 or @xmath33 .
vs. @xmath30 , for @xmath34 ; ( b ) plot of the sender variable @xmath3 vs. the intruder receiver variable @xmath28 , for @xmath35 ;
( c ) plot of the sender variable @xmath30 vs. the intruder receiver variable @xmath28 , for @xmath35 ; ( d ) plot of the sender variable @xmath30 vs. the intruder receiver variable @xmath28 , when @xmath36 .
the initial conditions in all cases were : @xmath37 . ] by making @xmath38 eq .
( [ receiver ] ) is simplified , becoming independent of @xmath29 , hence not depending upon the adjustment of @xmath39 . in this way @xmath40 whenever @xmath41 , regardless of the value of @xmath39 .
also we have found experimentally that the best results for fast convergence of the synchronism and minimum recovery error are obtained when @xmath42 . with those settings our intruder receiver
is redefined as : @xmath43 figure [ fig : gssinchro ] illustrates the synchronization mechanism between sender and intruder receiver .
the values of the the sender parameters are @xmath34 and the initial conditions of the sender and receiver are : @xmath44 .
figure [ fig : gssinchro](a ) shows the first 8 seconds of the plot of the sender variables @xmath3 vs. @xmath30 .
figure [ fig : gssinchro](b ) shows the plot of the sender variable @xmath3 vs. the intruder receiver variable @xmath28 when @xmath35 ; comparing ( a ) and ( b ) it can be seen that both phase portraits are identical , after the short initial transient originated by the different initial conditions .
figure [ fig : gssinchro](c ) shows the plot of the sender variable @xmath30 vs. the intruder receiver variable @xmath28 , when @xmath35 ; it can be seen that the phase and amplitude of @xmath28 match exactly those of @xmath30 , after the initial transient , although @xmath33 .
finally , fig .
[ fig : gssinchro](d ) shows the plot of the sender variable @xmath30 vs. the intruder receiver variable @xmath28 , when @xmath45 and @xmath39 completely differ from @xmath46 and @xmath47 , revealing that both systems are synchronized , although their amplitudes and phases do not match exactly .
fig4 ( 85,43)@xmath48 ( 59,7)@xmath49 ( 16.82,11.762)(13.32,7.35 ) ( 19,16)@xmath50 ( 19,13)@xmath51 ( 85,54.5)@xmath52 ( 50,-1)@xmath39 we have estimated and recorded the logarithm of the mean value of the squared error @xmath53 , i.e. the error power , for the range of the intruder receiver system parameter values @xmath45 and @xmath54 that give raise to the chaotic behavior of the lorenz attractor , with the same transmitter system parameters of the numerical example presented in ( * ? ? ? * fig . 2 ) and the intruder receiver described by eqs .
( [ receiver2a ] ) and ( [ receiver2b ] ) .
the results are presented in fig .
[ fig : logmeansquarederror ] .
the mean of @xmath53 is computed along the first 1.5 seconds , after a delay of 0.5 seconds to let the initial transient finish .
it is clearly seen that the error grows monotonically with the mismatch between the transmitter and receiver parameters @xmath55 , and that the minimum error corresponds to the receiver system parameters values @xmath56 exactly matching the transmitter system parameters values @xmath57 .
the parameters value recovery procedure consists of the straightforward search for the minimum recovery error @xmath22 .
once the correct values @xmath58 are found , the term @xmath59 vanishes and the recovery error is just equal to the plaintext signal @xmath4 .
the search of the correct parameter values @xmath21 can be done in the following way : first , select an initial value for @xmath45 centered in its usable range ; second , vary the value of @xmath39 until a minimum error is reached ; third , keep this value and vary the value of @xmath45 until a new minimum error is reached ; four , check if the remaining error @xmath22 is a clean recognizable plaintext , if not repeat the second and third steps .
note that this method retrieves all at once the correct values of @xmath45 , @xmath39 , and the plaintext .
the procedure is illustrated in fig .
[ fig : gsretrieved ] , for a plaintext @xmath60 , whose frequency is so low that it can not be retrieved by the previously described direct high - pass filtering method .
when the initial parameter values are chosen at random as @xmath61 , the corresponding error reaches a peak value near 70 .
then the @xmath39 value is varied until a minimum of the error is found for @xmath62 , it can be seen that now the peak error value after the initial transient is reduced to about 15 . next , the @xmath45 value is varied until a new error minimum is reached for @xmath63 , now the error is reduced to the plaintext itself , plus some noise of reduced amplitude that may be easily removed by a high - pass filter , if necessary .
fig5 ( 100,88)@xmath50 ( 100,93)@xmath51 ( 100,72)@xmath64 ( 100,77)@xmath65 ( 100,52)@xmath66 ( 100,57)@xmath65 ( 100,33)@xmath66 ( 100,38)@xmath49 ( 100,18.5)@xmath49 ( 100,13.5)@xmath66 ( 100,67)@xmath67 ( 100,47)@xmath67 ( 100,28)@xmath67 ( 100,8.5)@xmath68 we have found that our method woks as well for the whole family of functions of the form @xmath69 , were @xmath70 is a scalar of any value .
the plaintext @xmath4 is correctly recovered for any ciphertext of the form @xmath71 $ ] .
this fact demonstrates that a great care must be exercised when selecting the function @xmath0 , because some changes in its factors and constants values may be useless to enlarge the key space , hence not improving the system security at all . in fig .
[ fig : gsretrieved](e ) a time story of the retrieved message for a function of the form @xmath72 is presented , when decoded by an intruder receiver adjusted to recover a function of the type @xmath73 .
it can be observed that the only difference with fig .
[ fig : gsretrieved](d ) is the magnitude and duration of the initial transient . to break the ps - based chaotic masking scheme under study ,
when the hyperchaotic rssler system is used to generate the masking signal , we follow a similar procedure to the one used in the preceding section .
now the intruder receiver is : @xmath74 were @xmath26 is a scalar .
we have found the best results with @xmath75 . here
the instantaneous value of the recovery error is defined as @xmath23 .
when the synchronism is reached it happens that @xmath76 , hence the error is @xmath77 , thus allowing the exact recovery of the plaintext .
figure [ fig : roslergs ] illustrates the plaintext recovered for various intruder receiver parameter values sets , being the plaintext @xmath78 , whose frequency is so low that it can not be retrieved by the previously described direct high - pass filtering method .
it can be observed that the intruder receiver is quite insensitive to the values of the parameters @xmath79 and the structure of the function @xmath0 . in fig .
[ fig : roslergs](b ) it is shown that when the parameter values of sender and intruder receiver do not match at all , the plaintext is still visible , although a residual interference is present , but its intensity is not big enough to preserve the confidentiality of the communications
. this interference can be easily removed , if desired , by trial and error in few steps .
the first parameter to be adjusted is @xmath80 , because it is the most influent on the shape of the retrieved waveform . in fig .
[ fig : roslergs](c ) it is shown the waveform for @xmath80 correctly adjusted while @xmath81 is kept unadjusted .
then the parameter @xmath81 must be adjusted until a clean recovered plaintext waveform is reached , as illustrated in fig .
[ fig : roslergs](d ) .
we have tested several different invertible functions @xmath0 as building blocks of the sender ciphertext , and it has been observed that quite different functions allow for the almost correct retrieving of the plaintext .
figure [ fig : roslergs](e ) dramatically illustrates this fact : when a function as complicated as @xmath82 is used for transmission , while maintaining an intruder receiver designed for decoding a function of the type @xmath73 , together with a total parameter mismatch between sender and receiver as @xmath83 , we can see that the retrieved waveform still retain enough plaintext information to completely compromise the security of the communication .
fig6 ( 98,88)@xmath84 ( 98,93)@xmath85 ( 98,70)@xmath86 ( 98,75)@xmath87 ( 98,50)@xmath86 ( 98,55)@xmath88 ( 98,30)@xmath89 ( 98,35)@xmath88 ( 98,16)@xmath90 ( 98,11)@xmath91 ( 19,18)@xmath82
the knowledge of the scaling factor @xmath92 is not required to retrieve the plaintext , if a high - pass filter attack or a generalized synchronization recovering procedure are implemented , as we did .
hence the scaling factor does not add any strength to the system security , as opposed to the claims by the authors of @xcite .
thanks to the plain transmission of the shared scalar variables @xmath3 , or @xmath10 , inherent to the ps scheme , the third lorenz system parameter @xmath93 and the hyperchaotic rssler parameters @xmath94 and @xmath95 need not be determined to recover the plaintext when using a generalized synchronization receiver , in opposition to other cryptosystems that make use the lorenz or hyperchaotic rssler systems as a masking signal . therefore an additional advantage is offered by the authors of @xcite to the opponent eavesdropper .
in summary , the proposed ps - based chaotic masking cryptosystem is rather weak , since it can be broken in two easy ways , the first one being ignorant of the transmitter precise structure and the second one knowing the transmitter structure but ignoring its parameter values .
the alleged security advantage of the cryptosystem based on the eavesdropper lack of knowledge of the scaling factor @xmath92 is incorrect , its knowledge is completely irrelevant to retrieve the plaintext .
the function @xmath0 does not clearly enhance the security , owing to the fact that a wide range of different functions generate ciphertexts equally breakable with the same intruder receiver with a unique parameter adjustment .
there is no mention about what the key is , nor which is the key space , a fundamental aspect in every secure communication system .
the total lack of security discourages the use of this communication scheme for secure applications , unless some modifications are made to essentially enhance its security . c. s. zhou , t. l. chen , extracting information masked by chaos and contaminated with noise : some considerations on the security of communication approaches using chaos , phys .
lett . a 234 ( 6 ) ( 1997 ) 429435 .
|
this paper describes the security weaknesses of a recently proposed secure communication method based on chaotic masking using projective synchronization of two chaotic systems .
we show that the system is insecure and how to break it in two different ways , by high - pass filtering and by generalized synchronization . , , , and
|
the distribution of neutral hydrogen in the universe is largely traced by the ly@xmath3 forest , ly - limit systems , damped ly@xmath3 systems and galaxies , but the relationship between these tracers remains unknown .
the predicted existence of a large number of individual dark matter halos in groups of galaxies by the cold dark matter ( cdm ) models of the formation of galaxies and clusters may provide a clue to the link between these features .
if the excess satellites / halos exist , do they trace the distribution of the group and explain the abundance and location of the absorbers ?
all - sky hi surveys have detected only one type of object which could be the abundance of dark matter halos scattered throughout the local group .
these are the high - velocity clouds ( hvcs ) . using a standard cdm model of the formation of the local group ( moore et al .
2001 ) , we compare the properties of the cold dark matter halos with those of the high - velocity clouds and compact high - velocity clouds ( chvcs ) to determine if these objects may solve the `` missing satellite '' problem
. if the halos trace the distribution of a cluster , we are also able to place strong constraints on the fraction of neutral hydrogen associated with the dark matter .
the high - velocity cloud properties come from the @xmath4 spatial resolution catalog of wakker ( priv comm ) .
the wakker catalog is adapted from the catalog of wakker & van woerden ( 1991 ) and uses data from the dwingeloo telescope in the north ( hulsbosch & wakker 1988 ) and the iar 30 m in the south ( morras et al . 2000 ) .
this hvc catalog was used because it is from a fairly homogeneous resolution and sensitivity all - sky hvc dataset and uses a single catalog method .
the data have a 5@xmath5 brightness temperature sensitivity of @xmath00.05 k when it is smoothed kinematically . in this catalog of 626 clouds ,
the large complexes such as complex c and the magellanic stream are classified as one object , leading to a distinct difference in the number of individual objects between this catalog and the southern hvc catalog of putman et al .
the putman et al .
catalog ( described below ) was designed to catalog smaller objects and only merges clouds if the brightest enclosing contour is @xmath6 t@xmath7 .
this catalog has 1956 clouds in it and the flux distribution has a significantly steeper slope .
see putman et al . for the distribution plots of this catalog .
we also investigated the properties of the compact high - velocity clouds ( chvcs ) .
chvcs were originally defined by braun & burton ( 1999 ) as isolated clouds with diameters less than 2 ( based on the 50% of peak n@xmath8 contour ) .
this definition was slightly revised in the higher resolution and sensitivity putman et al .
catalog of chvcs to use the 25% of the peak n@xmath8 contour .
the hipass ( hi parkes all - sky survey ; barnes et al . 2001 )
data used in the putman et al .
catalog provide a marked improvement in the detectability of chvcs , so this catalog has been combined with the predominantly northern catalog of braun & burton , which was extracted primarily from the leiden - dwingeloo survey ( lds ) data ( @xmath9k ) .
the total number of chvcs is 215 .
the putman et al .
catalog covers declinations less than + 2and uses hipass data reduced with the minmed5 method ( putman 2000 ) .
hipass has a spatial resolution of 15.5@xmath10 , a spectral resolution , after hanning smoothing , of 26.4 km s@xmath11 , and a 5@xmath5 rms brightness temperature sensitivity of approximately 0.04 k , which corresponds to a @xmath12 column density sensitivity of @xmath13 atoms @xmath14 for an hvc with a linewidth of 35 .
the hipass hvc data is complete to fluxes of 2@xmath154 jy , however the catalog may only be 100% complete to total fluxes of 25 jy ( see putman et al .
2001 ) .
the primary selection effect when cataloging hvcs and chvcs is the gap in the velocity distribution due to the presence of galactic hi .
clouds between @xmath16 = 90 are missing in all of the catalogs discussed above .
this can be translated to a missing @xmath17 population with the equation , @xmath18 = @xmath19 + 220sin(@xmath20)cos(@xmath21 ) - 62cos(@xmath20)cos(@xmath21 ) + 40sin(@xmath20)cos(@xmath21 ) - 35sin(@xmath21 ) ( braun & burton 1999 ) .
we analyze a binary pair of massive halos that form in a hierarchical universe dominated by cold dark matter .
the masses , separation and relative velocities of the binary halos are close to those observed for the milky way and andromeda .
the binary system was chosen from a large cosmological system such that a nearby massive cluster similar to virgo was present .
the simulation is described in detail in moore et al .
( 2001 ) . at @xmath22 ,
each halo contains @xmath23 particles and is resolved to a distance of 0.1% of the virial radius , @xmath24 kpc .
the peak circular velocities of the halos are 220 and 200 .
these two halos are infalling for the first time at @xmath25 and their outer edges are separated by 700 kpc . over 2000 smaller dark matter halos lie within the high - resolution region and mostly within the virial radii of the two large halos that represent the milky way and andromeda .
it is these low mass dark matter halos that constitute the `` missing satellite '' problem for cdm models the local group galaxies make up just 1 - 2% of the expected number of satellites .
figure 1 shows the spatial distributions of all of the local group dark matter halos , the high - velocity clouds ( without distance constraints ; see below ) , and the compact high - velocity clouds in galactic coordinates .
the dark matter halos show a clear bias in their distribution towards m31 and the local group anti - barycenter region .
this bias in the distribution is independent of the size and mass of the halos , however , in terms of distance , it only appears beyond @xmath0500 kpc of the milky way ( see figure 2 ) .
the distribution of cataloged hvcs shown in the middle panel of figure 1 , excludes those hvcs which have distance determinations .
direct distance determinations have been made to several of the larger high velocity complexes which place them within approximately 10 kpc of the galaxy ( e.g. complex a , m , and we ; wakker 2001 ) , and the stream and leading arm originated from the interaction of the magellanic clouds with the milky way and are at distances within 100 kpc .
we can also exclude those complexes which are bright in h@xmath3 emission ( e.g. complex l , gcp ( the smith cloud ) , and c ) , as this emission would not have been detected at distances greater than 100 kpc ( e.g. bland hawthorn & maloney 1999 ) , unless associated with an active star forming region .
there are excesses of hvcs in the same general directions as the halos , but the extreme overabundance in the direction of m31 is not present .
furthermore , the distance determination methods have not uniformly sampled the sky , and this bias in the hvc distribution may disappear with future observations . the chvcs shown in the bottom panel of figure 1 are clustered , but clearly show a different distribution to the local group halos .
rrc object & # & mean @xmath26 + hvcs ( without distance constraints ) & 469 & -49 + chvcs ( compact hvcs ) & 215 & -57 + all dark matter halos & 2135 & -66 + halos outside of 300 kpc & 1530 & -69 + halos within 500 kpc & 584 & -27 + halos with m@xmath27 & 624 & -63 + figure 2 shows the spatial distributions of only those local group halos at least 300 kpc from both m31 and the galaxy ( top ) , within 500 kpc of the galaxy ( middle ) , and those halos with dark matter masses @xmath28 .
the bias in the spatial distribution of the halos is not evident within 500 kpc of the galaxy .
the beginning of an asymmetric distribution is seen between 500 and 800 kpc in this simulation .
since the simulation places the center of m31 at 1100 kpc , rather than the observed distance of 700 kpc ( freedman & madore 1990 ) , the actual bias in the distribution could appear somewhat closer .
the asymmetric distribution of halos remains in both the @xmath29 300 kpc and @xmath30 cases , but the total number of objects is greatly decreased ( see table 1 ) .
the velocity distributions of all of the dark matter halos , hvcs without distance constraints , and chvcs are shown in terms of galactic longitude ( @xmath31 ) and local group standard of rest velocity ( @xmath26 ) in figure 3 .
their mean lgsr velocities are tabulated in table 1 .
there is no obvious position - velocity relationship between the populations of clouds and the local group halos .
a wide range of velocities about m31 is shown in the simulation , as is a clear bias for negative @xmath26 s ( i.e. infall ) .
the hvc and chvc velocity distributions are limited by confusion with galactic emission between @xmath32 + /-
this is evident in the hvc distribution shown in the middle panel of figure 3 , with the hvcs clustering about a sinusoidal gap in the distribution .
both the hvcs and chvcs show a preference for negative velocities in the local group reference frame .
the chvcs show a tendency towards more positive velocities at @xmath33 and negative velocities at @xmath34 , with the largest range of velocities about @xmath35 .
the panels of fig .
4 show the velocity distributions of the selected groups of halos shown in fig .
the velocity distribution of the local group halos within 500 kpc is fairly uniform , as is the spatial distribution , with a mean @xmath17 the same magnitude more positive than the hvcs ( @xmath36 ) as the other halo populations are more negative .
the population of halos with m@xmath37 appears as a thinned version of the velocity distribution of all of the halos .
those halos over 300 kpc from m31 and the galaxy have a similar thinned distribution , but there is less scatter and less extreme velocities in the direction of m31 .
though it is difficult to infer a mass distribution of the hvcs and chvcs due to their unknown distances ( m@xmath38 d@xmath39 ) , we can infer a flux distribution of the halos using their masses and distances and assuming a certain percentage of gas is associated with the dark matter . if the halos contain 1% of their dark matter mass in neutral hydrogen , their total hi fluxes are directly comparable to those of the hvcs and chvcs , as shown in figure 5 .
note that some of the plots have different values on both axes .
the projected flux distribution of all of the dark matter halos is shown in the bottom left panel of figure 5 by the solid points .
the slope is -1.5 , which is the same as that of the hvcs shown in the upper left panel , but the number of halos in each flux bin is much higher .
the upper right panel shows that both the total number of chvcs and the slope of the flux distribution ( @xmath40 ) is different from that of the halos .
the selected populations of halos discussed above are shown in the bottom two panels of figure 5 .
halos outside of 300 kpc from the galaxy and m31 are represented by the dashed line in the bottom left panel .
the total number of clouds is again somewhat higher than that of the hvcs and the slope is steeper , like the chvcs .
the flux distribution of those halos with m@xmath27 ( @xmath41 ; solid points in the bottom right panel ) and those halos within 500 kpc of the galaxy ( @xmath42 ; dashed line in the bottom right panel ) are the most comparable to the hvcs in terms of total number and slope . the size distribution of the halos is another property which can be compared to the hvcs if one assumes that the gas and dark matter are coupled .
these plots will be shown in a future paper .
figure 5 also shows that even if the local group halos are not hvcs or chvcs , they should have been detected by the recent large area hi surveys if they contain @xmath01% of their dark matter mass in neutral hydrogen .
for instance , hipass covers over half of the sky and the velocities of the local group galaxies with a completeness level of @xmath43 jy , and the wakker catalog reaches fluxes of @xmath44 jy ( wakker & van woerden 1991 ) .
therefore , considering the entire population of halos ( 2135 total ) , almost all of them would have been detected if they contain 1% or greater of their dark matter mass in . by applying cuts in halo distance or mass ,
the number of halos which should have been detected is in closer agreement with the number of hvcs without lowering m@xmath8 to below 1% m@xmath45 .
there does not appear to be a close correlation between the entire population of dark matter halos in this standard cold dark matter local group simulation and the high - velocity clouds and compact high - velocity clouds .
however , when cuts are made in distance or mass , the remaining halos show properties which could be argued are consistent with those of the hvcs without distance constraints .
the compact hvcs do not show similar properties to the dark matter halos .
the justification behind the halo cuts are summarized below . for greater than 300 kpc from m31 and the galaxy ,
it is possible that the neutral hydrogen gas would not survive outside of a particular distance from these galaxies .
this could be due to either ram pressure or tidal stripping effects ( e.g. quilis & moore 2001 ) , or ionizing radiation escaping from the galaxy ( e.g. bland - hawthorn & maloney 1999 ) .
the cut which requires the halos associated with neutral hydrogen gas be within 500 kpc , is based on the idea that the is only able to condense within a certain radius of the galaxy ( e.g. oort 1970 ) . finally , the choice of only those halos with dark matter masses above @xmath46 having neutral gas associated with them comes from cdm simulations which include reionization at @xmath47 ( e.g. gnedin 2000 ) .
the increase in temperature of the gas during reionization suppresses the formation of low mass satellites due to a reduction in gas mass via the expulsion of already accreted gas and the suppression of further accretion . the primary difference between these sub - populations of halos and
the entire population is a reduction in the total number , but there are also differences in the flux distributions and within 500 kpc there is certainly a more uniform spatial and kinematic distribution .
considering the distance determinations which have been made to some of the larger high velocity complexes and the search for hvcs in other groups which have not found extragalactic hi clouds without stars down to m@xmath8 levels of @xmath48 ( e.g. zwaan 2001 ) , the possibility of the hvcs being the dark matter halos within a certain radius of the milky way seems the most likely possibility .
if the nearby halos are represented by hvcs , are the further away halos represented by the lower column density ly@xmath3 absorbers which have been detected in other groups ( e.g. penton , shull & stocke 2000 ) ?
it is difficult to find these low column density systems within the local group due to the the damping wings of the galaxy s ly@xmath3 and to some extent ly@xmath49 absorption lines obscuring those velocities .
the detection of these systems through other absorption lines may be possible in some cases ( e.g. @xmath50 ) , and may have already been detected along some sightlines ( e.g. sembach et al .
evidence for a link between hvcs and these few absorption line detections of lower column density high velocity gas lies in their spatial and kinematic relationship .
therefore , if this gas is related to the ly@xmath3 absorbers , one might expect to detect higher column density systems akin to hvcs in the vicinity of some of the low - redshift ly@xmath3 absorber systems .
the flux distribution plots of the halos , presented in the bottom two panels of figure 5 , put strong constraints on the amount of neutral hydrogen the halos can have .
hipass and other hvc surveys cover the velocities of all of the local group galaxies down to flux levels of 2 - 5 jy .
figure 5 shows that the majority of the halos should have been detected if they contain at least 1% of their dark matter mass in . since
only @xmath51 local group galaxies have been detected in , there is an obvious need to look towards other objects .
the number of hvcs and chvcs is also too low to agree with the number of halos that should have been detected , unless one applies distance and/or mass cuts ( see figure 5 and table 1 ) .
no other candidate `` missing satellites '' have been detected in in the local group , so either the hvcs are a population of the missing satellites , or the majority of the dark matter halos contain @xmath52% of their dark matter mass in .
|
we present a comparison of the leftover satellites at z=0 in a cold dark matter dominated simulation of the formation of the local group to the distribution of observed neutral hydrogen high - velocity clouds and compact high - velocity clouds .
the @xmath02000 leftover satellites in the simulation have dark matter masses which range between 0.5 to 10 @xmath1 , sizes between 3 to 30 kpc , and distances between 100 kpc and 2 mpc .
the dark matter halos show a clear bias in their distribution towards m31 and to a lesser extent towards the local group anti - barycenter .
if the local group halos contain @xmath01% of their dark matter mass in neutral hydrogen they should have been easily detected by the current surveys .
the only objects detected with the potential to be the local group halos are the high - velocity clouds . here
the spatial , kinematic , and flux properties of the clouds and dark matter halos are compared .
several different subsets of halos which may be more likely to contain neutral hydrogen are investigated , and the hvcs are found to have some similar properties to those halos within 500 kpc of the galaxy and those halos with dark matter masses @xmath2 .
the compact high - velocity clouds do not show similar properties to the halos .
# 1_#1 _ # 1_#1 _ = # 1 1.25 in .125 in .25 in
|
in this note i perform a study of su(3 ) breaking for the decays @xmath1 and @xmath0 .
this system is interesting , because it allows for a determination of the angle @xmath2 from mixing - induced and direct cp asymmetries @xcite provided the su(3 ) symmetry breaking corrections to a certain double ratio of amplitudes are known .
the analysis of su(3 ) breaking is done in the theoretical framework of qcd factorization @xcite , which expresses the hadronic decay amplitudes in terms of fundamental constants , decay constants , form factors etc . in the heavy quark limit .
the following is a preliminary version of work in progress in collaboration with m. neubert .
we write the decay amplitudes as @xmath3 where @xmath4 .
the `` tree '' and `` penguin '' amplitudes are defined as the coefficients of the two terms with different ckm matrix elements .
they also contain subleading penguin , electroweak penguin and weak annihilation amplitudes . in the su(3 )
symmetry limit @xmath5 and @xmath6
. we will be interested in the deviations of the ratios @xmath7 from unity .
the ratio of tree amplitudes is the product of a factorizable and a non - factorizable term , @xmath8 .
the factorizable term is given by @xmath9 where we have included the light meson masses and phase space effects .
the @xmath10 form factor is taken from @xcite , but the assumed value for the @xmath11 form factor is only an `` educated guess '' . in the context of qcd sum rules
this form factor is expected to depend sensitively on the poorly known first gegenbauer moment of the kaon light - cone distribution amplitude , since the form factor is dominated by the soft spectator - quark overlap term .
the factorizable su(3 ) breaking correction cancels in the double ratio @xmath12 @xcite .
this ratio and @xmath13 deviate from unity due to non - factorizable effects that can be computed in qcd factorization assuming that @xmath14 .
the computation is a straightforward extension of the results of @xcite .
i refer to this paper and @xcite for all details concerning the method and the values of the hadronic parameters .
the parameters relevant to su(3 ) breaking are : * @xmath15 vs. @xmath16 in the normalization of scalar penguin terms ; * the ratio @xmath17 vs. @xmath18 in the normalization of the hard - scattering and weak annihilation terms ; * the gegenbauer moments @xmath19 , @xmath20 vs. @xmath21 , @xmath22 , and the first inverse moments of the @xmath23-meson distribution amplitudes , @xmath24 vs. @xmath25 ; * the parameters for soft power corrections from hard scattering and weak annihilation , @xmath26 and @xmath27 .
since these parameters are often not well - known , we will make assumptions on the su(3 ) breaking and exhibit the effect of these assumptions on the ratios @xmath13 and @xmath12 .
this should be distinguished from more ambitious approaches as discussed in @xcite , where the su(3 ) breaking in the input parameters is also computed
. for a given pair of related parameters denoted @xmath28 , for instance @xmath29 , we set all other parameters and @xmath30 to their standard values .
we then write @xmath31 and vary @xmath32 , which controls the amount of su(3 ) breaking , between @xmath33 and 0.3 . for the complex parameters @xmath27 and @xmath26 we vary the magnitude and phase simultaneously and independently by this amount .
we make two exceptions to this treatment . for the light meson gegenbauer moments , which are already small corrections to the su(3 )
symmetric asymptotic distribution amplitudes , we take @xmath34 , @xmath35 , and @xmath36 .
second , we use @xmath37 with @xmath38 taken from @xcite .
the resulting variations of the magnitudes and phases of @xmath13 and @xmath12 about their default values @xmath39 are shown in the upper part of table [ tab : su3 ] .
adding up the various uncertainties we obtain non - factorizable su(3 ) breaking effects of order @xmath40 for the ratio of tree amplitudes and of order @xmath41 for the double ratio of penguin and tree amplitudes .
since the strong phases of the ratios are very small , the su(3 ) breaking effect on the phases is also small in absolute magnitude .
these estimates have to be regarded with caution , since they rely on the default estimate of weak annihilation ( @xmath42 ) , in which the annihilation amplitude does not have a strong phase . to study the impact of weak annihilation in more detail we pick the four values @xmath43 corresponding to @xmath44 with phase 0 , 90 , 180 , 270 degrees for the pion mode and allow the corresponding parameters to vary by @xmath45 for the kaon mode .
all the values of the annihilation parameter chosen by this procedure lie ( almost ) within our standard error assumption @xmath46
. the lower part of table [ tab : su3 ] demonstrates that the su(3 ) breaking effect on the ratio of penguin amplitudes can be large , up to @xmath47 in magnitude and @xmath48 on the phase of @xmath12 .
we conclude that unless a better understanding of su(3 ) breaking effects in the weak annihilation amplitude is found the assumption @xmath49 should not be made .
.[tab : su3 ] su(3 ) breaking effect on the magnitude and phase of @xmath13 and @xmath12 for variations of a given input parameter as described in the text . [ cols="^,^,^,^,^ " , ]
to estimate the theoretical uncertainty in the determination of the angle @xmath2 due to the su(3 ) breaking effects , we assume that the mixing phase @xmath50 is @xmath51 , corresponding to @xmath52 , and define the time - dependent cp asymmetry through @xmath53 & = & - s_f \sin(\delta m_b\,t ) + c_f \cos(\delta m_b\,t ) \,.\end{aligned}\ ] ] we assume that @xmath54 and @xmath55 have been measured
. we can then determine @xmath56}{t_\pi}\ ] ] as a function of @xmath2 , differs from @xcite by a sign , so that @xmath57 is near zero in qcd factorization . ] and predict @xmath58 , the direct cp asymmetry in @xmath59 decay , for a given assumed value of the su(3 ) breaking ratio @xmath12 .
a measurement of @xmath58 then results in a determination of @xmath2 with a theoretical error due to the uncertainty of @xmath12 .
we dismiss solutions with @xmath60 as unphysical , since a large penguin amplitude is or will be excluded by branching fraction measurements .
we generically find two solutions for @xmath61 .
the second solution satisfies @xmath62 only in a small range around @xmath63 and will not be discussed further .
the other solution already constrains @xmath64 from @xmath62 alone .
the question is how a measurement of @xmath58 narrows this range further . in figure
[ fig : su3 ] we show @xmath58 as a function of @xmath2 for five values of @xmath12 : the central value according to ( [ rptdef ] ) , and for the most pessimistic error assumptions corresponding to @xmath65 , @xmath66 , @xmath67 and @xmath68 .
the figure also shows @xmath69 and @xmath70 . the cp asymmetry exhibits a resonance - like behaviour after @xmath69 has gone through its minimal value , since the denominator of the expression for @xmath58 goes through a minimum when @xmath71 becomes small and positive .
we find that the largest error is introduced by the uncertainty in the phase of @xmath12 , which in our analysis is entirely due to weak annihilation .
for instance , if @xmath72 is found , the strategy would determine two values of @xmath2 in the absence of theoretical ( and experimental ) errors . including the su(3 ) breaking error
, we obtain @xmath73 and @xmath74 .
since @xmath70 is expected to be positive , the second range is theoretically favored . in general , the theoretical error can become larger or smaller depending on @xmath58 as seen from the figure .
it will also depend on what the mixing - induced and direct cp asymmetries in @xmath75 will eventually turn out to be , but our result should be generic . we therefore conclude that the strategy to determine @xmath2 from the @xmath75 to @xmath76 system _ may _ suffer from considerable theoretical uncertainties , unless additional information is available .
this can come from two sources : ( 1 ) excluding the possibility of a weak annihilation amplitude with a large strong rescattering phase would already eliminate the largest uncertainty that we could identify in the context of qcd factorization .
this could be achieved experimentally by excluding a large direct cp asymmetry in @xmath77 and neglecting su(3 ) breaking effects in the estimate of the weak annihilation phase .
( 2 ) if the @xmath78 mixing phase is known , the mixing - induced cp asymmetry in @xmath59 decay provides a fourth observable , which could be used to determine ( or eliminate ) @xmath57 , leaving only the su(3 ) breaking error on @xmath69 . to make this work in practice ,
a significant experimental effort is required .
r. fleischer , phys .
b * 459 * ( 1999 ) 306 ; + r. fleischer and j. matias , phys .
d * 66 * ( 2002 ) 054009 .
m. beneke , g. buchalla , m. neubert and c. t. sachrajda , phys .
* 83 * , 1914 ( 1999 ) ; nucl .
b * 591 * , 313 ( 2000 ) .
m. beneke , g. buchalla , m. neubert and c. t. sachrajda , nucl .
b * 606 * ( 2001 ) 245 .
m. beneke and m. neubert , [ hep - ph/0308039 ] .
|
-4.5 cm pitha 03/04 , hep - ph/0308040 , 27 june 2003 3.8 cm i estimate the su(3 ) flavour symmetry breaking in the ratio of penguin - to - tree ratios of the decays @xmath0 and @xmath1 , given an assumption on the flavour symmetry breaking of the hadronic input parameters .
the decay amplitudes are calculated in qcd factorization .
implications for the determination of @xmath2 are discussed .
|
consider two sets of @xmath5 correlated random variables in @xmath6 $ ] , @xmath7 and @xmath8 ( @xmath9 ) , respectively distributed according to the following joint probability densities ( jpd ) : @xmath10 where @xmath11 and @xmath12 are known normalization constants and @xmath13 . in the following two subsection we will provide physical motivations for considering such sets , namely set @xmath14 corresponds to the distribution of schmidt eigenvalues for entangled random pure states in bipartite systems ( see subsection [ ent ] ) , while set @xmath15 corresponds to the distribution of transmission eigenvalues of an open cavity in the chaotic regime ( see subsection [ condcav ] ) .
consider now the following statistical quantities : * the cumulative distribution @xmath16 $ ] of the largest member of set @xmath14 , @xmath17 . by definition , this is given by the following @xmath5-fold integral : @xmath18^n}d\lambda_1\cdots d\lambda_n \mathcal{p}_1(\lambda_1,\ldots,\lambda_n)\ ] ] differentiating @xmath19 , one obtains the _ probability density _ of @xmath20 , p_n(x)=q_n(x)[densq ] * the probability density @xmath21 $ ] of the quantity @xmath22 , which is given by the following @xmath5-fold integral : . ]
@xmath23^n}dt_1\cdots dt_n \mathcal{p}_2(t_1,\ldots , t_n)\delta\left(y-\sum_{i=1}^n t_i\right)\ ] ] simple algebraic manipulations , summarized in appendix [ appa ] lead to the following relation between the two quantities above : @xmath24 the identity in eq .
is the main result of this paper will be discussed in detail later on . ] . notwithstanding its remarkable simplicity , eq
actually permits an exact evaluation of @xmath19 , the so far unavailable distribution of the largest schmidt eigenvalue for _ finite _ @xmath5 ( see subsection [ ent ] ) , in terms of @xmath25 ( the probability density of landauer conductance , see subsection [ condcav ] ) about which much more is known .
the plan of the paper is as follows . in the next two subsections ,
we give a rather detailed introduction to the physics of entangled random pure states in bipartite systems ( related to the set @xmath14 above ) and about landauer conductance in chaotic mesoscopic cavities supporting a finite number of electronic channels in the two attached leads ( related to the set @xmath15 above ) . in section [ mainresults ] we exploit the identity to derive analytically the cumulative distribution and the density of @xmath20 in a few illustrative cases .
these results are then compared with numerical simulations with excellent agreement .
eventually we present concluding remarks in section [ concl ] and technical developments in the three appendices .
entanglement of pure bipartite systems is one of the most active areas of research nowadays , due to possible applications to quantum information and quantum computation problems @xcite .
it is also probably the simplest setting where well - behaved entanglement quantifiers can be defined , such as the von - neumann or rnyi entropies of either subsystem @xcite , the so called concurrence for two - qubit systems @xcite or other entanglement monotones @xcite .
_ typical _ properties of such states are best addressed by considering _
random _ pure states ( see e.g. @xcite for an excellent review ) . more precisely , consider a bipartition of a @xmath26-dimensional hilbert space @xmath27 as @xmath28 , where we assume without loss of generality that @xmath29 . for example , @xmath30 may be taken as a given system ( say a set of spins ) living in an external environment ( e.g. , a heat bath ) @xmath31 .
a quantum state @xmath32 of the composite system can be expanded as a linear combination @xmath33 are two complete basis of @xmath34 and @xmath35 respectively .
the coefficients @xmath36 s of this expansion form the entries of a rectangular @xmath37 matrix @xmath38 .
we consider here _ entangled random pure _ states @xmath32 .
this means that : 1 .
@xmath32 _ can not _ be expressed as a direct product of two states belonging to the two subsystems @xmath30 and @xmath31 .
2 . the expansion coefficients @xmath36 are random variables drawn from a certain probability distribution .
3 . the density matrix of the composite system is simply given by @xmath39 with the constraint @xmath40=1 $ ] , or equivalently @xmath41 .
more precisely , the density matrix of @xmath32 can then be straightforwardly expressed as = _
i,_j , x_i , x_j,^ * |i^aj^a| |^b ^b| , [ dem2 ] where the roman indices @xmath42 and @xmath43 run from @xmath14 to @xmath1 and the greek indices @xmath44 and @xmath45 run from @xmath14 to @xmath2 .
the _ reduced _ density matrix @xmath46 $ ] is obtained by tracing out the environmental degrees of freedom ( i.e. those of the subsystem @xmath31 ) : _ a = tr_b[]=_=1^m ^b||^b .
[ rdm1 ] using the expansion in eq .
one gets _ a = _ i , j=1^n _
= 1^m x_i , x_j,^ * the entries of the @xmath47 matrix @xmath48 . in a similar way
, one could obtain the reduced density matrix @xmath49 $ ] of the subsystem @xmath31 in terms of the @xmath50 matrix @xmath51 .
the two matrices @xmath52 and @xmath53 share the same set of nonzero ( positive ) real eigenvalues @xmath54 . in the diagonal basis
, one can express @xmath55 as _ a= _ i=1^n _ i |^a_i ^a_i| [ diaga ] where @xmath56 s are the normalized eigenvectors of @xmath57 and similarly for @xmath58 .
the original composite state @xmath32 in this diagonal basis reads : |= _
i=1^n normalization condition @xmath41 , or equivalently @xmath40=1 $ ] , imposes the constraint on the sum of eigenvalues , @xmath59 . for _ random _ pure states , the expansion coefficients in eq
can be typically drawn from an unbiased ( so called _ hilbert - schmidt _ ) distribution ( real or complex ) @xmath60\propto \delta\left ( { \rm tr}(\mathcal{x } \mathcal{x}^{\dagger})-1\right)\ ] ] where the dyson index @xmath13 corresponds respectively to real and complex @xmath38 matrices ) or break ( @xmath3 ) time - reversal symmetry . ] .
the meaning of eq .
is clear : all normalized density matrices compatible with unitary invariance are sampled with equal probability , which corresponds to having minimal _ a priori _ information about the quantum state under consideration .
this in turn induces nontrivial correlations among the schmidt eigenvalues ( which are now real random variables between @xmath61 and @xmath14 whose sum is @xmath14 ) and makes the investigation of several statistical quantities about such states quite interesting . here
we present a quick summary of known results : * the joint probability density ( jpd ) of schmidt eigenvalues , derived by lloyd and pagels @xcite , which is precisely given by @xmath62 in eq .
, with @xmath63 and @xmath64 .
note that the delta function there guarantees that @xmath40=1 $ ] .
the normalization constant in this case reads @xcite : c_n,=(/2)(n - m+1)-1^()= * the average von neumann entropy for large @xmath0 ( computed by page @xcite for @xmath3 and extended in @xcite to the case @xmath4 ) ; * the average von neumann entropy for _ finite _ @xmath0 and @xmath3 , conjectured by page @xcite and independently proven by many researchers soon after @xcite also in a non - extensive setting @xcite ; * density of schmidt eigenvalues ( one - point function ) for finite @xmath65 , derived independently in @xcite and @xcite for @xmath3 and in @xcite for @xmath4 ; * universality of eigenvalue correlations for @xmath3 @xcite ; * average fidelity between quantum states @xcite and distribution of so - called @xmath66-concurrence @xcite for @xmath3 ; * distribution of so - called _ purity _ for small @xmath1 @xcite , and phase transitions in its laplace transform for large @xmath1 @xcite ; * full distribution of rnyi entropies ( including large deviation tails ) , computed in @xcite for large @xmath67 and all @xmath45s using a coulomb gas method . as a byproduct
, the authors also obtain in @xcite the average and variance of rnyi entropy valid for large @xmath67 , and the density of schmidt eigenvalues for all @xmath45 s and @xmath68 large ; * distribution of smallest eigenvalue ( related to so - called _ demmel condition number _ @xcite ) for @xmath13 and finite @xmath69 , derived independently in @xcite and @xcite . in the latter paper ,
a conjecture by znidaric @xcite was proven , @xmath70 exactly for all @xmath67 , while for @xmath4 @xmath71 for large @xmath67 , where the constant @xmath72 is precisely known @xcite . ]
( see also @xcite for an extension of these results to the case @xmath73 ) ; * distribution of largest eigenvalue for _ large _ @xmath67 and all @xmath45s @xcite , including small and large deviation laws .
typical fluctuations around the mean @xmath74 ( once properly scaled ) are found to follow the tracy - widom distribution ( see also @xcite for a related result ) .
no results seem to be available for the case of _ finite _ @xmath0 . given the current interest in the distribution of extreme schmidt eigenvalues ,
the reader may on the one hand wonder whether they really encode useful information , and on the other why the largest eigenvalue distribution for finite @xmath0 is much harder to obtain via the same strategy used for the smallest one @xcite . in order to answer the first question , first note that due to the constraint @xmath59 and the fact that all eigenvalues are nonnegative
, it follows that in eq . .
] @xmath75 and @xmath76 .
now consider the following limiting situations .
suppose that the largest eigenvalue @xmath77 takes its maximum allowed value @xmath14 .
then it follows immediately that all the remaining @xmath78 eigenvalues must be identically @xmath61 .
in this situation eq .
( [ sch1 ] ) tells us that @xmath32 is fully _ unentangled _ ( completely separable ) . on the other hand ,
if @xmath79 ( i.e. , it takes its lowest allowed value ) , all the eigenvalues must have the same value , @xmath80 for all @xmath42 . in this case , the pure state @xmath81 is _ maximally _ entangled , as this state maximizes the von neumann entropy @xmath82 .
in other words , the knowledge of the largest eigenvalue distribution really provides essential information about how entangled a random pure state is .
a discussion about the asymmetry in the treatment of smallest and largest schmidt eigenvalues is included in appendix [ appb ] .
consider a cavity of submicron dimensions etched in a semiconductor and connected to the external world by two leads supporting @xmath2 and @xmath1 electronic channels .
it is well established that the electrical current flowing through such a cavity when brought out of equilibrium by an applied external voltage presents time - dependent fluctuations which persist down to zero temperature @xcite and are thus associated with the granularity of the electron charge @xmath83 .
typical features observed in experiments include weak localization @xcite , universality in conductance fluctuations @xcite and constant fano factor @xcite .
the landauer - bttiker scattering approach @xcite is rather successful in describing the statistics of quantum transport : it amounts to relating the wave function coefficients of the incoming and outgoing electrons through the unitary scattering matrix @xmath84 ( @xmath85 , if @xmath86 ) : @xmath87 where the transmission ( @xmath88 ) and reflection @xmath89 blocks are submatrices encoding the transmission and reflection coefficients among different channels ) are respectively of size @xmath90 and @xmath91 , while @xmath89 are of size @xmath50 and @xmath47 . ] .
experimental observables can be extracted from the eigenvalues of the hermitian transport matrix @xmath92 : for example , the dimensionless conductance and the shot noise are given respectively by @xmath93 @xcite and @xmath94 $ ] @xcite .
random matrix theory ( rmt ) has been very successful in describing the statistics of universal fluctuations in such systems , when the corresponding classical dynamics is chaotic : the scattering matrix @xmath84 is drawn from a suitable ensemble of random matrices , with the overall constraint of unitarity @xcite . a maximum entropy approach ( under the assumption of ballistic point contacts
@xcite ) forces the probability distribution of @xmath84 to be uniform within the unitary group , i.e. @xmath84 belongs to one of dyson s circular ensembles @xcite . from the uniformity of the probability density of @xmath84 within the unitary group , the jpd of the transmission eigenvalues @xmath8 of the matrix @xmath95 , from which the statistics of interesting experimental quantities could be in principle derived ,
is precisely given by eq .
, with @xmath96 and @xmath97 @xcite . there
, the dyson index @xmath45 characterizes different symmetry classes ( @xmath13 according to the presence or absence of time - reversal symmetry ) .
the eigenvalues @xmath98 are thus correlated real random variables between @xmath61 and @xmath14 .
the normalization constant @xmath12 is explicitly known from the celebrated selberg s integral as : k_n,^^()=_j=0^n-1 from , in principle the statistics of many observables of interest can be calculated . in particular , we focus here on the dimensionless landauer conductance @xmath99 , which satisfies the bounds @xmath100 .
mean and variance of @xmath66 and other quantities have been known for a long time when @xmath0 are large @xcite , and recently also for finite @xmath0 @xcite .
stimulated by some recent experimental progresses @xcite , which made eventually possible to explore the _ full distribution _ of the conductance ( and not just its mean and variance ) , a lot of effort has been devoted to its theoretical characterization .
an explicit expression was first obtained for @xmath101 @xcite , while more results were available in the case of quasi one - dimensional wires @xcite and 3d insulators @xcite . for the shot noise , the full distribution was known only for @xmath102 @xcite .
very recently , sommers _
@xcite announced two formulas for the distribution of conductance and shot noise , valid at arbitrary number of open channels and for any @xmath45 , which are based on fourier expansions . in @xcite , a complete and systematic approach to the full statistics of conductance and shot noise
was brought forward .
this is based on symmetric function expansions , and is valid for @xmath13 and arbitrary @xmath103 .
in addition , the authors of @xcite were able to provide general formulae in terms of determinants or pfaffians for the full probability distribution of conductance and shot noise at quantized @xmath45 and general @xmath0 ( see below ) .
in @xcite , the integrable theory of quantum transport in chaotic cavities ( based on painlev transcendent ) for @xmath3 was formulated , and recursion formulae for the efficient computation of conductance and shot noise cumulants have been derived .
in two recent publications @xcite , the distributions of conductance , shot noise and integer moments were computed in the limit @xmath104 using a coulomb gas method , and long _ power - law _ tails were detected in the distributions , a result confirmed by extensive numerical simulations . in @xcite ,
exact results ( basically equivalent to the general formulae in @xcite ) for the laplace transform @xmath105 of the conductance distribution for any @xmath0 and all @xmath45s are formally given as follows : * : @xmath106_{j , k=0,\ldots , n-1}$ ] for @xmath1 even , and @xmath107_{j , k=0,\ldots , n-1}$ ] for @xmath1 odd , where @xmath108 $ ] is the pfaffian of the even - dimensional antisymmetric matrix @xmath109 @xcite and the proportionality constants are known explicitly .
the arrays @xmath110 and @xmath111 are given by : @xmath112 * : in this case , we have a representation in terms of a hankel determinant _ g(s)_j , k=0, ,n-1[hankel]where : @xmath113 that was first derived in @xcite .
explicit inversions of the laplace transforms above are always possible on a case - by - case basis , and a catalogue of such evaluations for few interesting cases is provided in ref . @xcite . in next section and in appendix
[ appc ] , we will combine such explicit formulas with the identity in eq . to illustrate the validity of our approach to the distribution of largest schmidt eigenvalue .
for illustrative purposes , we consider two applications of the main identity eq . .
a third one is discussed in great detail in appendix [ appc ] . 1 . : in this case , the probability density of landauer conductance has been derived explicitly in @xcite as : @xmath114\theta(3-y)\label{pptt}\end{aligned}\ ] ] where @xmath115 is heaviside step function . 2 .
: again , the probability density of landauer conductance in this case has been derived explicitly in @xcite as : @xmath116 & \mbox{for } 1\leq y\leq 3 \end{cases}\label{pptt1}\ ] ] note that : * since the conductance density has always a compact support @xmath117 $ ] ( where @xmath96 ) , it follows immediately from eq . that the cumulative distribution of the largest schmidt eigenvalue @xmath19 and its density @xmath118 have compact support @xmath119 , as expected .
* since the conductance density is known to be continuous but not everywhere analytic @xcite ( i.e. it displays critical points at which higher derivatives are discontinuous ) , the cumulative distribution and the density of the largest schmidt eigenvalue enjoy this property too . in fig .
[ cumfig ] we plot the cumulative distributions @xmath19 corresponding to the two cases above . the curves are obtained via eq . , where @xmath25 is respectively given by and .
the distributions are increasing functions of the argument @xmath120 , as they should , and such that @xmath121 .
differentiation of the analytical formulas provide the density of the largest eigenvalue , @xmath118 , which is plotted in figs .
[ densityfig1 ] and [ densityfig2 ] along with numerical simulations .
these are obtained as follows @xcite :
we generate @xmath122 _ real _ gaussian @xmath91 matrices @xmath38 .
2 . for each instance
we construct the wishart matrix @xmath123 .
we diagonalize @xmath52 and collect its @xmath1 real and non - negative eigenvalues @xmath124 .
we define a new set of variables @xmath125 as @xmath126 , for @xmath127 .
the set of variables @xmath128 is guaranteed to be sampled according to the measure ( [ p1 ] ) .
we construct a normalized histogram of @xmath129 .
the agreement between theory and simulations is excellent .
we also evaluated exactly the average of @xmath130 for the two cases above and found : @xmath131 however , a general formula for @xmath132 is still elusive ( unlike @xmath133 @xcite ) and may well attract further researches ( see also appendix [ appc ] ) . of the largest schmidt eigenvalue via eq .
for @xmath134 ( violet ) and @xmath135 ( blue ) .
this distribution is identically zero for @xmath136 and @xmath121 as it should . ]
we have presented an exact identity relating two statistical quantities which arise in different contexts : the cumulative distribution of the largest schmidt eigenvalue for entangled random pure states in bipartite systems of sizes @xmath137 and the probability density of landauer conductance in chaotic cavities supporting @xmath1 and @xmath2 electronic channels in the two external leads .
recent analytical results for the latter are exploited to derive ( so far unavailable ) exact formulas for the former quantity at _ finite _ @xmath0 ( while large @xmath0 results are already available @xcite ) , which is of interest in order to quantify the degree of entanglement of random pure states .
a detailed introduction to the physics involved has been provided , along with a precise discussion of the asymmetry in the treatment of the smallest and largest schmidt eigenvalue distributions .
a general formula for @xmath138 , the average of the largest eigenvalue , valid for arbitrary @xmath0 is unfortunately still lacking and certainly deserves further investigations .
+ + * acknowledgments : * it is a pleasure to thank satya n. majumdar , cline nadal , oriol bohigas and gernot akemann for collaborations on related projects and many interesting discussions , and boaz nadler for useful correspondence .
i am also grateful to dima savin for a careful reading of the manuscript and many helpful advices .
consider the cumulative distribution @xmath19 of @xmath17 ( eq . ):
@xmath139^n}d\lambda_1\cdots d\lambda_n \delta\left(\sum_{i=1}^n\lambda_i -1\right)\prod_{i=1}^n \lambda_i^{\alpha}\prod_{j < k}|\lambda_j-\lambda_k|^\beta\ ] ] a change of variables @xmath140 leads to : @xmath141^n}d t_1\cdots d t_n \delta\left(x\sum_{i=1}^n t_i -1\right)\prod_{i=1}^n t_i^{\alpha}\prod_{j < k}|t_j - t_k|^\beta\end{aligned}\ ] ] using the property of delta functions @xmath142 , we obtain straighforwardly eq . .
consider the cumulative distribution @xmath143 of the _ smallest _ member of set @xmath14 , @xmath144 ( i.e. the distribution of the smallest schmidt eigenvalue for entangled random pure states ) . by definition
it is given by : @xmath145^n}d\lambda_1\cdots d\lambda_n \delta\left(\sum_{i=1}^n\lambda_i -1\right)\prod_{i=1}^n \lambda_i^{\alpha}\prod_{j < k}|\lambda_j-\lambda_k|^\beta\ ] ] where the upper limit of integration can be safely extended up to @xmath146 in view of the unit norm constraint .
in the case @xmath147 and @xmath3 , which we focus on here for illustrative purposes , the evaluation of this multiple integral proceeds via the auxiliary function @xmath148 @xcite : @xmath149^n}d\lambda_1\cdots d\lambda_n \delta\left(\sum_{i=1}^n\lambda_i -t\right ) \prod_{j
< k}|\lambda_j-\lambda_k|^2\ ] ] such that @xmath150 .
next , one takes the laplace transform of @xmath148 : @xmath151^n}d\lambda_1\cdots d\lambda_n e^{-s\sum_{i=1}^n\lambda_i }
\prod_{j < k}|\lambda_j-\lambda_k|^2\ ] ] and in the r.h.s performs a linear shift @xmath152 ( this is the crucial technical step ) , to get : @xmath153^n}d y_1\cdots d y_n e^{-\sum_{i=1}^n y_i } \prod_{j < k}|y_j - y_k|^2}_{\propto\mathcal{z}_{\rm{wl}}}\ ] ] note that , thanks to the linear shift , the dependence on the laplace variable @xmath154 has been entirely transferred outside the @xmath5-fold integral : this is now proportional to the partition function @xmath155 of an associated wishart - laguerre ( wl ) ensemble of random covariance matrices @xcite of the form @xmath156 , where @xmath38 is a gaussian rectangular matrix with real or complex entries .
the joint distribution of the @xmath5 nonnegative eigenvalues of @xmath52 is known @xcite ^(wl)(_1,
,_n ) = _ n,^ ( ) e^-_i=1^n _ i _ i=1^n _ i^ _ j < k |_j-_k|^ [ wishart1 ] where @xmath157 is a known normalization constant
. therefore , the jpd of schmidt eigenvalues can be seen as a _ fixed - trace _ ( microcanonical ) version of the wishart - laguerre ( canonical ) ensemble .
now , in the case of the cumulative distribution of the _ largest _ eigenvalue @xmath19 , the integrals on the r.h.s .
run over @xmath158 $ ] instead of @xmath159 $ ] , making the aforementioned linear shift less useful .
one could keep pursuing the laplace - transform route ( introducing an auxiliary function @xmath160 ) with the change of variables @xmath161 obtaining : @xmath162^n}d y_1\cdots d y_n e^{-\sum_{i=1}^n y_i } \prod_{j < k}|y_j - y_k|^2}_{\propto e_{\rm{wl}}[(sx,\infty)]}\ ] ] but the integral on the r.h.s .
does not permit this time a friendly laplace - inversion ( see however next appendix ) .
indeed , this integral is readily recognized as proportional to @xmath163 $ ] , where @xmath164 $ ] is the _ gap probability _ for a wishart - laguerre ensemble , i.e. the probability that the interval @xmath165 on the real axis is free of eigenvalues .
this quantity is exactly known in terms of painlev v @xcite , with the consequence that an explicit laplace inversion formula is not available to date . in summary ,
the asymmetry in the treatment of the smallest and largest schmidt eigenvalues arises because the linear shift that works in the former case fails in the latter , and this fact calls for the alternative approach developed in this paper .
in this appendix we discuss in more detail a third application of the main identity eq . .
in the case @xmath3 and @xmath69 , the hankel determinant representation in eq . actually allows a more systematic ( and more easily automatized ) treatment of the cumulative distribution of the largest schmidt eigenvalue .
following the laplace transform route outlined in the previous appendix , one can easily write down the following equation : _ 0^dt q_n(x , t)e^-st / x = c_n,0^(2 ) x^n^2_n ! _
j , k=0 , , n-1 where we have used the hankel determinant representation for the @xmath5-fold integral on the r.h.s . provided in @xcite ( fully equivalent to ) , with @xmath166 .
a change of variable @xmath167 on the l.h.s . followed by a ( formal ) laplace inversion , leads to the final formula for @xmath19 in this case_{j , k=0,\ldots , n-1}\right](1/x)$ ] @xcite . ] : q_n(x)=c_n,0^(2 ) n ! x^n^2 - 1 ^ -1(1/x)[invlap ] where @xmath168(t)$ ] is the inverse laplace transform of @xmath169 with parameter @xmath170 . the r.h.s .
of can be systematically evaluated in @xmath171 and its derivative ( @xmath118 ) has been plotted in fig .
[ densityapp ] for @xmath172 .
also the average of the largest eigenvalue @xmath175 can be evaluated exactly in @xmath171 for a given @xmath5 , and we provide a few evaluations in the following table . unlike the average of the smallest eigenvalue , which in the same circumstances ( @xmath3 and @xmath69 ) has the attractively simple form @xmath176 exactly for all @xmath1 @xcite , the situation for the largest eigenvalue seems more complicated and it is much harder to even conjecture a possible formula for the sought average valid for all @xmath67 . we leave this as a challenging open problem , noticing _ en passant _ that the asymptotic value @xmath177 derived in @xcite is approached _ very slowly _ as @xmath1 increases @xcite .
this fact makes the investigation of _ finite _ ( small ) @xmath1 results very much called for .
s. hill and w. k. wootters , entanglement of a pair of quantum bits , _ phys .
* 78 * , 5022 ( 1997 ) ; w. k. wootters , entanglement of formation of an arbitrary state of two qubits , _ phys .
* 80 * , 2245 ( 1998 ) .
s. lloyd and h. pagels , complexity as thermodynamic depth , _ annals of physics _ * 188 * , 186 ( 1988 ) ; see also e. lubkin , entropy of an @xmath5-system from its correlation with a @xmath178-reservoir , _ j. math .
* 19 * , 1028 ( 1978 ) and m.j.w .
hall , random quantum correlations and density operator distributions , _ phys .
lett . a _ * 242 * , 123 ( 1998 ) .
s. k. foong and s. kanno , proof of page s conjecture on the average entropy of a subsystem , _ phys .
lett . _ * 72 * , 1148 ( 1994 ) ; j. snchez - ruiz , simple proof of page s conjecture on the average entropy of a subsystem , _ phys .
e _ * 52 * , 5653 ( 1995 ) ; s. sen , average entropy of a quantum subsystem , _ phys .
lett . _ * 77 * , 1 ( 1996 )
. h. kubotani , s. adachi , and m. toda , exact formula of the distribution of schmidt eigenvalues for dynamical formation of entanglement in quantum chaos , _ phys .
lett . _ * 100 * , 240501 ( 2008 ) ; s. adachi , m. toda , and h. kubotani , random matrix theory of singular values of rectangular complex matrices i : exact formula of one - body distribution function in fixed - trace ensemble , _ annals of physics _ * 324 * , 2278 ( 2009 ) .
liu and d .- s .
zhou , local statistical properties of schmidt eigenvalues of bipartite entanglement for a random pure state , _ international mathematics research notices _ , doi:10.1093/imrn / rnq091 ( 2010 ) .
o. giraud , purity distribution for bipartite random pure states , _ j. phys . a : math .
theor . _ * 40 * , f1053 ( 2007 ) ; distribution of bipartite entanglement for random pure states , _ j. phys . a : math .
_ * 40 * , 2793 ( 2007 ) .
p. facchi , u. marzolino , g. parisi , s. pascazio , and a. scardicchio , phase transitions of bipartite entanglement , _ phys .
lett . _ * 101 * , 050502 ( 2008 ) ; a. de pasquale , p. facchi , g. parisi , s. pascazio , and a. scardicchio , phase transitions and metastability in the distribution of the bipartite entanglement of a large quantum system , _ phys . rev .
a _ * 81 * , 052324 ( 2010 ) . c. nadal , s. n. majumdar , and m. vergassola , phase transitions in the distribution of bipartite entanglement of a random pure state , _ phys .
lett . _ * 104 * , 110501 ( 2010 ) ; statistical distribution of quantum entanglement for a random bipartite state , _ preprint _ [ arxiv:1006.4091 ] ( 2010 ) .
r. landauer , spatial variation of currents and fields due to localized scatterers in metallic conduction , _ ibm j. res .
dev . _ * 1 * , 223 ( 1957 ) ; electrical resistance of disordered one - dimensional lattices , _ phil . mag .
* 21 * , 863 ( 1970 ) ; d. s. fisher and p. a. lee , relation between conductivity and transmission matrix , _ phys . rev .
b _ * 23 * , 6851 ( 1981 ) . m. novaes , full counting statistics of chaotic cavities with many open channels , _ phys . rev . b _ * 75 * , 073304 ( 2007 ) ; statistics of quantum transport in chaotic cavities with broken time - reversal symmetry , _ phys . rev . b _ * 78 * , 035337 ( 2008 )
. d. v. savin , h .- j .
sommers , and w. wieczorek , nonlinear statistics of quantum transport in chaotic cavities , _ phys .
b _ * 77 * , 125332 ( 2008 ) ; j .-
luque and p. vivo , nonlinear random matrix statistics , symmetric functions and hyperdeterminants , _ j. phys . a : math .
_ * 43 * , 085213 ( 2010 ) .
s. hemmady , j. hart , x. zheng , t. m. antonsen , e. ott , and s. m. anlage , experimental test of universal conductance fluctuations by means of wave - chaotic microwave cavities , _ phys .
b _ * 74 * , 195326 ( 2006 ) . k. a. muttalib and p. wlfle , one - sided log - normal distribution of conductances for a disordered quantum wire , _ phys
_ * 83 * , 3013 ( 1999 ) ; a. cresti , r. farchioni , and g. grosso , conductance distributions at the metal - insulator crossover in quasi 1-d pseudorandom wires , _ eur .
j. b _ * 46 * , 133 ( 2005 ) ; k. a. muttalib , p. wlfle , a. garca - martn , and v. a. gopar , nonanalyticity in the distribution of conductances in quasi - one - dimensional wires , _ europhys .
lett . _ * 61 * , 95 ( 2003 ) ;
l. s. froufe - prez , p. garca - mochales , p. a. serena , p. a. mello , and j. j. senz , conductance distributions in quasi - one - dimensional disordered wires , _ phys . rev .
lett . _ * 89 * , 246403 ( 2002 ) ; v. a. gopar , k. a. muttalib , and p. wlfle , conductance distribution in disordered quantum wires : crossover between the metallic and insulating regimes , _ phys .
b _ * 66 * , 174204 ( 2002 ) .
k. a. muttalib , p. marko , and p. wlfle , conductance distribution in strongly disordered mesoscopic systems in three dimensions , _ phys .
b _ * 72 * , 125317 ( 2005 ) ; p. marko , dimension dependence of the conductance distribution in the nonmetallic regimes , _ phys . rev .
b _ * 65 * , 104207 ( 2002 ) . v. al .
osipov and e. kanzieper , integrable theory of quantum transport in chaotic cavities , _ phys .
* 101 * , 176804 ( 2008 ) ; statistics of thermal to shot noise crossover in chaotic cavities , _
j. phys . a : math .
* 42 * , 475101 ( 2009 ) .
p. vivo , s. n. majumdar , and o. bohigas , distributions of conductance and shot noise and associated phase transitions , _ phys .
_ * 101 * , 216809 ( 2008 ) ; probability distributions of linear statistics in chaotic cavities and associated phase transitions , _ phys . rev . b _ * 81 * , 104202 ( 2010 ) .
g. akemann , g. m. cicuta , l. molinari , and g. vernizzi , compact support probability distributions in random matrix theory , _ phys .
e _ * 59 * , 1489 ( 1999 ) ; non - universality of compact support probability distributions in random matrix theory , _ phys .
e _ * 60 * , 5287 ( 1999 ) .
c. a. tracy and h. widom , fredholm determinants , differential equations and matrix models , _ commun .
_ * 163 * , 33 ( 1994 ) ; p. j. forrester and n. s. witte , application of the @xmath179-function theory of painlev equations to random matrices : pv , piii , the lue , jue , and cue , _ commun .
pure appl .
_ * 55 * , 679 ( 2002 ) .
|
a strategy to evaluate the distribution of the largest schmidt eigenvalue for entangled random pure states of bipartite systems is proposed .
we point out that the multiple integral defining the sought quantity for a bipartition of sizes @xmath0 is formally identical ( upon simple algebraic manipulations ) to the one providing the probability density of landauer conductance in open chaotic cavities supporting @xmath1 and @xmath2 electronic channels in the two leads .
known results about the latter can then be straightforwardly employed in the former problem for both systems with broken ( @xmath3 ) and preserved ( @xmath4 ) time reversal symmetry .
the analytical results , yielding a continuous but not everywhere analytic distribution , are in excellent agreement with numerical simulations .
|
the active asteroids are solar system bodies which have asteroid - like orbits but which also show transient , comet - like activity .
evidence amassed over the past decade shows that this activity results from a remarkably broad range of physical processes , including impact , thermal fracture , rotational instabilities and ice sublimation .
identification of these processes in the main - belt represents a scientific watershed , by revealing diverse asteroid processes that were previously unobserved ( hsieh and jewitt 2006 , jewitt 2012 , jewitt et al .
2015 ) .
324p / la sagra ( formerly 2010 r2 and hereafter `` 324p '' ) is one of four active asteroids for which mass - loss has been reported on different orbits ( the others are 133p / elst - pizarro , 238p / read and 313p / gibbs ; jewitt et al .
mass loss near successive perihelia is a natural indicator of a thermal process , presumably the sublimation of near - surface ice .
dynamical models show that the capture of ice - rich jupiter family comets from the kuiper belt , while possible , is highly inefficient given the current architecture of the solar system ( fernandez et al .
2002 , levison et al .
2006 , hsieh and haghighipour 2016 ) .
therefore , it is likely that at least some of the icy objects in the main - belt are primordial residents .
moreover , the observed rates of mass loss can not be sustained for billion - year timescales given the small sizes of most active asteroids .
it is therefore likely that these objects spend most of the time between triggering events ( impacts ? ) in an inactive state and that the currently active asteroids represent a much larger population of ice - containing but inert bodies in the asteroid belt .
they are the tip of the iceberg . while no spectroscopic detection of the gaseous products of sublimation has yet been reported ( except in 1 ceres , kuppers et al .
2014 ) , the limits to gas production ( @xmath161 kg s@xmath3 ) are consistent with the rates of activity estimated from dust . active asteroid 324p has semimajor axis , @xmath17 = 3.096 au , eccentricity @xmath18 = 0.154 , and inclination @xmath19 = 21.4 , leading to an asteroid - like tisserand parameter measured relative to jupiter , @xmath20 = 3.099 .
the object was discovered in an active state on ut 2010 september 14.9 ( nomen et al .
2010 ) , three months after perihelion on ut 2010 june 26 .
revived activity was again reported in the summer of 2015 ( hsieh and sheppard 2015 ) , just prior to the next perihelion on ut 2015 november 30 .
the development of activity in the discovery epoch was described independently in papers by moreno et al .
( 2011 ) and hsieh et al .
( 2012 ) , while measurements from 2013 ( when the object was in an apparently inactive state ) were described by hsieh ( 2014 ) .
thermal infrared detections were reported by bauer et al .
in this paper we report new observations from the hubble space telescope taken to investigate 324p at the highest available angular resolution .
observations using the hubble space telescope were secured under target - of - opportunity program go 14263 ( two orbits ) and continued with so - called mid - cycle time under go 14458 ( four orbits ) .
a journal of observations is provided in table ( [ geometry ] ) .
the earliest observations ( ut 2015 september 28 ) were obtained from a vantage point 9.5 below the orbital plane of 324p , affording a view of the distribution of dust particles across the orbital plane .
the last observations , on ut 2015 december 18 , were scheduled to coincide with the passage of the earth through the orbital plane ( in practice , the out - of - plane angle was -0.6 ) . from this nearly in - plane geometry ,
the images provide , instead , a complementary measure of the distribution of dust in a direction perpendicular to the orbit plane .
all observations employed the broadband f350lp filter , which has effective wavelength 6230 and full - width - at - half - maximum ( fwhm ) of 4758 when observing a sun - like spectrum .
this filter provides maximum sensitivity to faint and low surface brightness targets , and is ideally suited to the observation of weak cometary activity .
the images have a scale of 0.04 pixel@xmath3 , corresponding to 65 km pixel@xmath3 and 90 km pixel@xmath3 on the first and last dates of observation , respectively .
figure ( [ composite ] ) shows , for each date of observation , the median of five , 420 s integration images taken within a single orbit , along with arrows to mark the projected anti - solar direction ( yellow ) and the negative heliocentric velocity vector ( green ) .
inspection of figure ( [ composite ] ) shows that the tail of 324p maintains a position angle that closely follows the changing antisolar direction .
the position angle of the projected anti - solar vector changes considerably ( from @xmath1287 to @xmath1260 ) between the september and december observations , while the position angle of the projected orbit remains relatively constant near 238 ( table [ geometry ] ) .
the swing of the tail is as expected for recently - released particles that are small enough to be strongly accelerated by solar radiation pressure .
in contrast , we observe in figure ( [ composite ] ) no clear indication for the presence of dust particles along the direction of the projected orbit , suggesting that the abundance of large , radiation pressure - insensitive particles is small . to explore these qualitative inferences further , we computed the trajectories followed by dust particles ejected from the nucleus of 324p .
the latter trajectories are functions of the dust ejection velocity , @xmath21 , and of @xmath22 , the ratio of the radiation pressure induced acceleration to the local solar gravitational acceleration , @xmath23 .
ratio @xmath22 depends on the particle morphology ( e.g. compact sphere vs. fluffy aggregate ) and on the composition ( bohren and huffman 1983 ) . for compact dielectric spheres ,
@xmath22 is proportional to the inverse particle radius ( provided @xmath24 1 ) and , to a useful level of approximation , we write @xmath25 , where @xmath26 is the particle radius expressed in microns .
solid lines in figure ( [ synsyns ] ) show syndynes , which are the loci of positions of particles of a given @xmath22 released from the nucleus with negligible velocity at different times ( c.f .
finson and probstein 1968 ) . dashed lines in the figure show synchrones , which mark the loci of positions of particles having different @xmath22 but which are all released from the nucleus at a given time .
the direction and slight counter - clockwise curvature of the tail are best matched by syndynes with @xmath27 0.005 to 0.02 , corresponding to particle radii @xmath0 50 to 200 @xmath1 m .
we take @xmath0 100 @xmath1 m as the nominal grain size in 324p .
while evidence for larger particles ( smaller @xmath22 ) is distinctly absent , we note that such particles are always difficult to detect in comets .
this is because they contain a small fraction of the scattering cross - section and because their small acceleration by solar radiation pressure may limit their abundance far from the nucleus .
evidently , 324p is an extended source .
photometric measurements employing fixed - angle apertures sample a larger volume of coma at larger geocentric distances , potentially making the object appear artificially bright as it receeds from earth . to examine the magnitude of this `` aperture effect ''
, we determined apparent magnitudes in two ways .
first , in table ( [ photometry ] ) we list measurements within circular apertures of fixed angular radii 0.2 , 1.0 , 4.0 and 6.0 .
the background for these measurements was determined from the median signal within a concentric annulus having inner and outer radii 6.0 and 12.0 , respectively .
second , in table ( [ absolute ] ) we list measurements within circular apertures having fixed linear radii ( projected to the distance of the object ) of 460 km , 2300 km , 9200 km and 13,800 km . the fixed linear radius measurements employed background subtraction from a concentric annulus extending from 13,800 km to 27,600 km . in both tables ( [ photometry ] ) and ( [ absolute ] ) , the photometric errors increase with aperture radius because of the growing importance of sky background uncertainties in the larger aperture measurements .
comparison of tables ( [ photometry ] ) and ( [ absolute ] ) shows that the aperture effect is small , because 324p is very centrally - condensed .
nevertheless , we use only the fixed linear - aperture measurements from table ( [ absolute ] ) in the following discussion .
the apparent magnitudes , @xmath28 , were converted to absolute values by correcting to unit heliocentric , @xmath29 , and geocentric , @xmath30 , distance and to zero phase angle , @xmath31 , using the inverse square law @xmath32 here , @xmath33 is the phase function , equal to the ratio of the scattered light at phase angle @xmath31 to that at @xmath31 = 0 .
we assumed the phase function formalism of bowell et al .
( 1989 ) with parameter @xmath34 = 0.15 , as appropriate for a c - type object and @xmath34 = 0.25 , for an s - type spectrum .
the phase function of 324p is unmeasured , introducing uncertainty into the value of @xmath35 over and above that due to measurement errors . at the largest phase angles of the present observations ( @xmath31 = 22 , table [ geometry ] ) ,
the difference between assumed c - type and s - type phase corrections is @xmath120.1 magnitudes , giving an estimate of the magnitude of the phase correction uncertainty .
absolute magnitudes using @xmath36 are given in tables ( [ photometry ] ) and table ( [ absolute ] ) together with their statistical uncertainties .
the absolute magnitudes are related to the effective scattering cross - section of the material within the photometry aperture , @xmath37 ( km@xmath4 ) , by @xmath38 } \label{area}\ ] ] where @xmath39 is the geometric albedo of 324p and @xmath40 is the apparent magnitude of the sun , both at wavelength @xmath41 .
we assume @xmath42 = -26.77 .
bauer et al .
( 2012 ) obtained a 3@xmath43 upper limit to the visual geometric albedo @xmath44 0.06 ; we assume @xmath45 = 0.05 .
the resulting scattering cross - sections are listed in table ( [ photometry ] ) , computed assuming @xmath46 = 0.05 with small adjustments for @xmath47 and @xmath48 as indicated by the broadband colors .
uncertainties on @xmath37 reflect the larger ( @xmath49 magnitude ) systematic uncertainties estimated from the difference between the c - type and s - type phase functions .
figure ( [ hvplot ] ) shows the absolute magnitudes , @xmath35 , as a function of time , measured as day - of - year in 2015 , for four apertures of fixed linear radius .
the absolute magnitudes generally brighten with time , showing that material is being actively released from the nucleus into the dust tail over a @xmath122 month interval around perihelion ( day of year = 334 ) .
the largest ( 13,800 km ) aperture photometry brightens from @xmath35 = 17.00 to 15.75 , a factor of 3 in cross - section ( table [ absolute ] ) .
the brightening corresponds to the release of a mass of dust @xmath50 where @xmath10 is the mass density of the dust grains , @xmath51 is their mean radius and @xmath52 is the change in the scattering cross - section .
we take @xmath51 = 100 @xmath1 m , as suggested by the syndynes in figure ( [ synsyns ] ) , @xmath52 = [email protected] km@xmath4 between october 08 and december 18 from table ( [ absolute ] ) and assume @xmath10 = 1000 kg m@xmath11 to obtain a mass ejection @xmath54 = [email protected]@xmath710@xmath55 kg .
if ejected steadily over this 71 day interval , the average dust ejection rate would be @xmath2 0.2 kg s@xmath3 .
we solved the energy balance equation for an exposed , perfectly absorbing water ice surface located at the sub - solar point on 324p . at @xmath56 = 2.62 au , we find that ice would sublimate , in equilibrium with sunlight , at the specific rate @xmath57 = 4.3@xmath710@xmath9 kg m@xmath58 s@xmath3 . the area of exposed ice needed to supply dust at the rate @xmath59 is given by @xmath60 where @xmath61 is the ratio of the dust to gas mass production rates .
measurements of jupiter family comets generally show @xmath62 .
for example , a detailed investigation of 67p / churyumov - gerasimenko by fulle et al .
( 2016 ) gave @xmath63 5 to 10 while values 10 @xmath64 30 were obtained in comet 2p / encke by reach et al .
we conservatively adopt @xmath61 = 5 to find @xmath65 = 930 m@xmath4 ( only @xmath120.02% of the surface of a spherical nucleus of radius 550 m ) corresponding to a circular patch as small as @xmath66 @xmath1217 m in radius .
this is a lower limit to @xmath65 in the sense that we have assumed sublimation at the maximum possible rate by placing the ice patch at the subsolar point and assuming that it is perfectly absorbing .
ice that is reflective , or located away from the subsolar point or buried beneath a thin , insulating dust mantle would sublimate less rapidly and require a larger area to supply the dust mass loss rate . indeed , the much brighter coma observed in 2010 would require a larger area of exposed ice , @xmath67 19,000 m@xmath4 ( about 0.5% of the surface area , or @xmath68 80 m ) , if produced by sublimation in steady - state . in either case , it is evident
that sublimation from a very small fraction of the nucleus surface can supply the dust release in 324p near perihelion .
we checked to confirm that the drag from a gas flow of strength @xmath57 is more than sufficient to launch spherical particles of 100 @xmath1 m size against the gravity of a nucleus estimated at only 0.55 km in radius ( hsieh 2014 ) . indeed , neglecting adhesive contact forces , particles up to @xmath120.5 m could be launched by gas drag .
we use the surface brightness profile of the dust in the direction perpendicular to the projected orbit plane to set a constraint on the dust ejection velocity . for this purpose , we rotated the composite image taken on ut 2015 december 18 about the nucleus in order to bring the projected orbit to the horizontal , and then computed vertical profiles averaged along segments of the tail .
results from the profiles are shown in figure ( [ vertprofile ] ) , where three quantities are plotted .
gray circles show the location of peak dust brightness together with horizontal bars to indicate the width of each segment of the tail used to average the data .
the distances to the half - power points of each profile to the north and south sides of the tail are shown as green and red circles , respectively .
error bars on the latter grow with distance from the nucleus because of the growing effect of sky noise and background sky brightness errors ( caused by contamination from faint , trailed background sources ) as the tail surface brightness rapidly declines . in all cases ,
the dust is more extended to the south of the tail axis than to the north .
figure ( [ synsyns ] ) shows that this is a projection effect , caused by viewing the dust sheet from a vantage point slightly below the plane ( by -0.6 ; c.f .
table [ geometry ] ) .
we take the extension of the dust to the north of the axis as the most meaningful measure of the out - of - plane distribution .
this extension is about 0.05 ( linear distance @xmath69100 km ) at 2 ( @xmath704500 km ) from the nucleus , far below the typical resolution afforded by ground - based telescopes .
dust motion perpendicular to the orbit plane is unaffected by radiation pressure on timescales short compared to the orbital period ; the perpendicular distance travelled in time @xmath71 is just @xmath72 , where @xmath73 is the perpendicular velocity .
dust motion parallel to the orbit is accelerated by radiation pressure , such that the distance travelled after time @xmath71 is @xmath74 . here
, @xmath75 = 0.006 m s@xmath58 is the gravitational attraction to the sun at @xmath56 = 1 au , and @xmath56 is the heliocentric distance expressed in au .
eliminating @xmath71 gives @xmath76 calculations ( silsbee and draine 2016 ) show that silicate particles have @xmath24 1 , regardless of their size or aggregate structure ( small , metal particles can have @xmath77 1 but these are of dubious relevance in the active asteroid population ) . with @xmath24 1 , @xmath56 = 2.621 au and @xmath78 = 100 km at @xmath79 = 4500
km , equation ( [ speed ] ) gives @xmath5 1 m s@xmath3 , which we take as an upper limit to the dust ejection velocity . in the classical comet model ( whipple 1950 ) the terminal dust velocity resulting from gas drag
varies approximately as @xmath80 .
micron - sized grains are dynamically well - coupled to the gas , with terminal speeds near the sound speed ( which we take as @xmath81 450 m s@xmath3 at @xmath56 = 2.621 au ) .
we therefore expect that , in the classical model , the optically dominant @xmath12100 @xmath1 m sized particles should have @xmath82 45 m s@xmath3 , considerably larger than the upper limit to @xmath83 set by the width of the dust tail using equation ( [ speed ] ) .
similarly low dust speeds have been reported in other active asteroids ( jewitt et al .
2014 ) and explained as a consequence of sublimation from a small area source .
a small sublimating area limits the path length over which expanding gas can accelerate entrained dust and so leads to lower terminal velocities . from a source of horizontal dimension @xmath84 ,
the terminal velocity is given by equation ( a5 ) of jewitt et al .
( 2014 ) as @xmath85 here , @xmath86 is a dimensionless drag coefficient , @xmath87 is the gravitational constant , @xmath10 is the density of the ejected grain , @xmath88 is the density of the nucleus , @xmath89 is its radius and the other symbols are as defined above .
we assume @xmath10 = @xmath88 = 1000 kg m@xmath11 and take @xmath17 = 100 @xmath1 m , @xmath84 = 80 m to find @xmath90 3 m s@xmath3 .
this is an order of magnitude smaller than the speed estimated from @xmath80 scaling and more comparable to , but slightly larger than , the measured upper limit @xmath5 1 m s@xmath3 . with our measurements
we can not reject the possibility that the dust is ejected slowly by a process other than gas drag , but such an alternative explanation would struggle to account for the recurrence of activity at two successive perihelia .
photometry of the nucleus was obtained using small apertures fixed in linear radius ( 460 km ) projected to the distance of 324p ( table [ absolute ] ) .
the faintest absolute magnitude , @xmath35 = [email protected] on ut 2015
september 28 , is brighter than reported in data from the previous orbit .
hsieh ( 2014 ) found absolute red magnitude @xmath91 = [email protected] to [email protected] ( depending on the adopted phase angle correction ) which , assuming a sun - like color index v - r = 0.35 ( holmberg et al . 2006 ) , corresponds to @xmath35 = [email protected] to [email protected] .
the nucleus on ut september 28 is thus brighter by a magnitude or more , presumably indicating ongoing dust release in the recent data .
for this reason , hsieh s ( 2014 ) determination of the effective radius of the nucleus , @xmath89 = [email protected] km ( assumed geometric albedo 0.05 ) , remains the most trustworthy .
three contiguous hst orbits on ut 2015 december 18 ( table [ geometry ] ) were secured specifically to provide a timebase sufficient to assess short - term variations in the scattered light . to measure these images
, we removed cosmic ray artifacts by hand using the following procedure .
first , we computed the median image from the 5 images taken within each orbit of hst .
the use of the median effectively eliminates cosmic rays that are abundant in the individual images .
next , we subtracted this median image from each of the individual images in the coresponding orbit so as to remove the `` steady '' signals in the data , leaving only cosmic rays , noise and the residuals of passing background objects .
near - nucleus cosmic rays were then removed by interpolation of the brightness in surrounding pixels .
the final step was to add back the median image from each orbit , to produce a set of cosmic ray cleaned data suitable for nucleus photometry .
this procedure worked well in all but a few images for which cosmic ray and trailed field galaxy contamination was too severe to be removed .
we rejected such images from further consideration .
the nucleus photometry was measured using projected circular apertures 0.2 in radius .
the background was determined using the median signal computed within a surrounding annulus having inner and outer radii 0.2 and 0.4 , respectively .
this background annulus includes a contribution from the dust component of 324p but a comparison of measurements in table ( [ nucleus_phot ] )
with photometry obtained using a larger sky annulus ( table [ photometry ] ) shows that the dust has only a small influence on the photometry .
the results from table ( [ nucleus_phot ] ) are plotted in figure ( [ lightcurves ] ) , where brightness variations up to @xmath120.1 magnitude are seen to exist at each epoch of observation .
the variations are large compared to the uncertainties of measurement , which we estimate to be @xmath530.01 magnitudes , and thus must be considered real . the december 18 data have the longest timebase and clearly show brightness variations on timescales that are consistent with the effects of nucleus rotation , while single - orbit visits on september 28 , october 08 and december 07 provide only brief snapshots .
it is impossible to uniquely link the photometric variations across all four dates of observation because of the long intervals between observations and the resulting aliasing in the data .
however , we note that the rising brightness recorded on ut 2015 october 08 ( upper right panel of figure [ lightcurves ] ) very likely corresponds to the rising section of the lightcurve on december 18 ( lower right panel ) , meaning that the rotation period must be a sub - multiple of the interval between these dates . given the nucleus absolute magnitude , @xmath35 = [email protected] as described above , is so much fainter than the values plotted in figure ( [ lightcurves ] ) it is obvious that dust must contribute to the total cross - section .
for example , on ut 2015 september 28 about half the cross - section in the 460 km radius aperture ( c.f .
table [ absolute ] ) can be attributed to a 0.55 km radius spherical nucleus while , by ut 2015 december 18 , this fraction had fallen to @xmath121/3rd .
however , it is unlikely that near nucleus dust , even though it dominates the cross - section , is responsible for the observed photometric variability . to see this
we note that dust traveling at the maximum speed allowed by the tail width , namely @xmath121 m s@xmath3 would take 4.5@xmath710@xmath14 s ( @xmath125 day ) to cross the 460 km radius aperture and , therefore , would contribute a quasi - constant background to the daily photometry .
in particular , the decrease in brightness between @xmath1260 minutes and @xmath12120 minutes in the december 18 panel of the figure can not be explained by dust escaping the photometry aperture unless the dust speed is @xmath92130 m s@xmath3 , which is inconsistent with the dust tail width .
the variability in the figure is more naturally interpreted as a nucleus rotational lightcurve diluted by near - nucleus coma .
the @xmath120.1 magnitude range of variation is a lower limit to the actual range of the rotating bare nucleus because of the adjacent dust .
while the sampling of the hst data is too limited to be able to define the nucleus rotation period , we can nevertheless use the data to place physically important limits on the rotation .
crucially , the lightcurve on december 18 does not repeat , showing that the rotation period must be longer than the 3.8 hr interval between the first and last images in this sequence .
thus , we are confident that the rotation period of the nucleus of 324p is longer than the critical 2.2 hour `` rotational barrier '' period shown by suspected rubble - pile asteroids ( harris 1996 , pravec et al . 2002 ) . in turn , this provides no reason to think that mass loss from 324p might be strongly influenced by rotational instability , unless the nucleus is a highly - elongated , prolate body or one having extremely low density .
future time - resolved photometry of 324p when in an inactive state is needed to establish the rotational state .
the present absolute magnitudes are compared with published measurements as a function of the mean anomaly , @xmath93 , in figure ( [ mags ] ) . in making this plot ,
all photometry has been corrected to the v filter assuming v - r = 0.35 .
the figure shows that our measurements are intermediate between the weakly active state recorded earlier in 2015 ( hsieh and sheppard 2015 ) and the bright state in which 324p was discovered in 2010 ( hsieh et al .
( 2012 ) and gives the impression that the absolute brightness is a nearly monotonic function of the mean anomaly angle .
the persistence of activity from pre - perihelion @xmath94 310 to post - perihelion @xmath94 100 corresponds to a @xmath12660 day interval during which the heliocentric distance changes from 2.7 au inbound to 3.1 au outbound .
however , whether this is a real , temporal progression that repeats from orbit to orbit can not be known until more observations are acquired .
unfortunately , the elongation of 324p fell below 50 immediately following the observations reported here , terminating measurements from hst .
observations over the next year will be of great interest in showing whether or not the photometric behavior of 324p is repeatable from orbit to orbit .
the dust properties in the 2010 active phase were modeled by moreno et al .
they assumed @xmath95 and fitted the isophotes of the dust to find a grain speed @xmath21 = 0.2 m s@xmath3 for @xmath17 = 10@xmath58 m , with an uncertainty of order a factor of two . scaling to 100 @xmath1 m particle size
would give @xmath82 2 m s@xmath3 , compared with our upper limit @xmath5 1 m s@xmath3 .
the mass loss rates estimated by moreno et al . peaked near 4 kg s@xmath3 , which is larger by about a factor of 20 than the @xmath120.2 kg s@xmath3 deduced from our photometry in 2015 .
however , 324p was intrinsically much brighter when observed by moreno et al .
for example , figure ( [ mags ] ) shows that the average absolute magnitude in our data ( @xmath96 16.4 ) is fainter than the peak brightness recorded in 2010 ( @xmath96 13.0 ) by 3.4 magnitudes , corresponding to a factor @xmath1223 . using similar arguments , hsieh and sheppard ( 2015 ) found @xmath2 0.1 kg s@xmath3 near @xmath93 = 300 .
we estimated the total mass loss from 324p as follows .
we convert the instantaneous absolute magnitude to production rate using @xmath97 and then integrate with respect to time . from observations in the discovery epoch ( i.e. mean anomalies 12 @xmath98 54 , shown as yellow circles in figure [ mags ] ) we obtain @xmath99 = 2@xmath100 kg .
integrating over the full range of observations ( 280 @xmath101 100 ) gives @xmath102 4 @xmath100 kg , which provides an order of magnitude estimate of the mass lost from the nucleus in one orbit .
the mass of a 550 m radius sphere having the density @xmath10 = 10@xmath103 kg m@xmath11 is @xmath10410@xmath105 kg . if continued , mass loss would deplete all the mass in the nucleus of 324p in a time @xmath106 , where @xmath107 5 yr is the average keplerian orbital period .
substituting , we find @xmath108 10@xmath14 yr . to survive for time , @xmath109 , would require a duty cycle ( the fraction of time over which the body is active ) given by @xmath110 .
hsieh et al . ( 2012 ) used numerical integrations to show that the orbit of 324p is stable on timescales @xmath111 yr , while the collisional lifetime of a @xmath121 km diameter asteroid is @xmath112 4@xmath710@xmath113 yr ( c.f .
figure 14 of bottke et al .
2005 ) . substituting these timescales
we obtain @xmath114 . a somewhat smaller number , @xmath115 2@xmath710@xmath9 , was obtained from statistical arguments based on observations of four active asteroids , summarized in jewitt et al .
( 2015 ) . in steady - state , the existence of @xmath116 known
, repetitively active ( ice - containing ) objects corresponds to a dormant population , @xmath117 .
we set @xmath116 = 4 ( active asteroids 133p , 238p , 313p and 324p ) and substitute @xmath15 to find @xmath118
. these are almost certainly under - estimates of the dormant object population , both because we are sensitive only to objects in which the ice is close enough to the physical surface to be occasionally exposed to sunlight ( e.g. haghighipour et al .
2016 ) and because the surveys used to identify active asteroids are incomplete ( waszczak et al .
2013 , hsieh et al .
we obtained hubble space telescope images of active asteroid 324p on four occasions between ut 2015 september 28 and december 18 .
we find that : 1 .
high resolution images of 324p show a point - like , but still active , nucleus and a radiation - pressure swept dust tail consisting of @xmath12100 @xmath1 m sized particles .
an increase by a factor @xmath123 in the near - nucleus dust cross - section indicates the continuing ejection of dust of mass @[email protected]@xmath710@xmath55 kg .
the average dust production rate in this period , @xmath2 0.2 kg s@xmath3 , could be supplied by the sublimation of water ice covering as little as @xmath120 930 m@xmath4 ( about 0.2% ) of the nucleus surface .
the total mass lost in the orbit following the discovery of 324p is estimated at @xmath124@xmath710@xmath8 kg .
observations taken from near the orbital plane of 324p indicate that the dust ejection velocity is @xmath1211 m s@xmath3 ( for 100 @xmath1 m sized particles ) .
this low speed is compatible with a model of gas drag acceleration from a sublimating ice patch that is small compared to the nucleus .
the bare nucleus of 324p was not observed in our data .
however , we detect short - term photometric variations in the near - nucleus region that are likely caused by nucleus rotation .
if so interpreted , these variations rule - out periods @xmath1213.8 hr and suggest that rotational instability is unlikely to play a leading role in the loss of mass from this object .
the small size of the nucleus limits the mass - loss lifetime of 324p to about 10@xmath14 yr , which is short compared both to the reported ( 10@xmath113 yr ) dynamical lifetime and to the estimated collisional lifetime ( 4@xmath122 yr ) .
unless the time spent by 324p in orbit is severely over - estimated , the persistence of this object implies a small duty cycle ( @xmath15 ) and a corresponding population ( @xmath118 ) of dormant asteroidal counterparts containing near - surface ice .
we thank the anonymous referee for a prompt review . based on observations made with the nasa / esa hubble space telescope , obtained at the space telescope science institute , which is operated by the association of universities for research in astronomy , inc .
, under nasa contract nas 5 - 26555 .
these observations are associated with go programs 14263 and 14458 .
2015 sep 28 23:13 - 23:49 & 271 & -63 & 344.6 & 2.632 & 2.213 & 21.7 & 86.6 & 238.1 & -9.5 + 2015 oct 08 09:02 - 09:39 & 281 & -53 & 346.9 & 2.629 & 2.326 & 22.2 & 83.7 & 237.9 & 8.8 + 2015 dec 07 03:41 - 04:18 & 341 & 7 & 1.8 & 2.620 & 3.021 & 18.4 & 64.5 & 238.1 & -1.9 + 2015 dec
18 00:32 - 04:20 & 352 & 18 & 4.5 & 2.621 & 3.128 & 16.9 & 60.4 & 238.4 & -0.6 + september 28 & 0.2 & [email protected] & [email protected] & [email protected] + september 28 & 1.0 & [email protected] & [email protected] & [email protected] + september 28 & 4.0 & [email protected] & [email protected] & [email protected] + september 28 & 6.0 & [email protected] & [email protected] & [email protected] + october 08 & 0.2 & [email protected] & [email protected] & [email protected] + october 08 & 1.0 & [email protected] & [email protected] & [email protected] + october 08 & 4.0 & [email protected] & [email protected] & [email protected] + october 08 & 6.0 & [email protected] & [email protected] & [email protected] + december 07 & 0.2 & [email protected] & [email protected] & [email protected] + december 07 & 1.0 & [email protected] & [email protected] & [email protected] + december 07 & 4.0 & [email protected] & [email protected] & [email protected] + december 07 & 6.0 & [email protected] & [email protected] & [email protected] + december 18 & 0.2 & [email protected] & [email protected] & [email protected] + december 18 & 1.0 & [email protected] & [email protected] & [email protected] + december 18 & 4.0 & [email protected] & [email protected] & [email protected] + december 18 & 6.0 & [email protected] & [email protected] & [email protected]
+ september 28 & v & [email protected] & [email protected] & [email protected] & + september 28 & h@xmath123 & [email protected] & [email protected] & [email protected] & + september 28 & @xmath37 & [email protected] & 4.6 @xmath530.5 & [email protected] &
+ + october 08 & v & [email protected] & [email protected] & [email protected] & [email protected] + october 08 & h@xmath123 & [email protected] & [email protected] & [email protected] & [email protected] + october 08 & @xmath37 & [email protected] & [email protected] & [email protected] & [email protected] + + december 07 & v & [email protected] & [email protected] & [email protected] & [email protected] + december 07 & h@xmath123 & [email protected] & [email protected] & [email protected] & [email protected] + december 07 & @xmath37 & [email protected] & [email protected] & [email protected] & [email protected] + + december 18 & v & [email protected] & [email protected] & [email protected] & [email protected] + december 18 & h@xmath123 & [email protected] & [email protected] & [email protected] & [email protected] + december 18 & @xmath37 & [email protected] & [email protected] & [email protected] & [email protected] + + sep 28 & 23:13 & 22.66 & 17.79 & 2.12 + & 23:22 & 22.62 & 17.75 & 2.12 + & 23:31 & 22.63 & 17.76 & 2.18 + & 23:40 & & & + & 23:57 & 22.61 & 17.74 & 2.22 + oct 08 & 09:09 & 22.77 & 17.77 & 2.14 + & 09:18 & 22.75 & 17.75 & 2.18 + & 09:27 & 22.72 & 17.72 & 2.24 + & 09:37 & 22.70 & 17.70 & 2.28 + & 09:46 & 22.68 & 17.68 & 2.33 + dec 07 & 03:41 & 22.85 & 17.42 & 2.98 + & 03:50 & 22.82 & 17.38 & 3.07 + & 03:59 & 22.83 & 17.39 & 3.04 + & 04:09 & 22.82 & 17.39 & 3.07 + & 04:18 & & & + dec 18 & 00:32 & 22.78 & 17.32 & 3.26 + & 00:41 & 22.80 & 17.34 & 3.20 + & 00:50 & 22.75 & 17.29 & 3.35 + & 01:00 & 22.72 & 17.26 & 3.45 + & 01:09 & 22.71 & 17.25 & 3.48 + & 02:07 & 22.73 & 17.27 & 3.41 + & 02:16 & 22.74 & 17.28 & 3.38 + & 02:25 & 22.76 & 17.30 & 3.32 + & 02:35 & 22.76 & 17.30 & 3.32 + & 02:44 & 22.74 & 17.28 & 3.38 + & 03:42 & & & + & 03:51 & & & + & 04:00 & 22.76 & 17.30 & 3.32 + & 04:10 & 22.76 & 17.30 & 3.32 + & 04:19 & 22.75 & 17.29 & 3.35 + = 2@xmath710@xmath58 , 1@xmath710@xmath58 , 5@xmath710@xmath11 , 2@xmath710@xmath11 , 1@xmath710@xmath11 , 5@xmath710@xmath124 , 2@xmath710@xmath124 , 1@xmath710@xmath124 and 1@xmath710@xmath125 measured counterclockwise about the nucleus . for clarity , we label only the syndynes having @xmath126 = -2 , -3 and -4 .
synchrones are shown as straight dashed lines , with the synchrone of 2015 july 01 marked with a blue diamond .
synchrones clockwise from this date ( measured with the nucleus as center ) are plotted at 10-day intervals , while those counter - clockwise from 2015 july 01 have 100-day intervals .
solid blue and green lines mark the projected anti - solar and negative orbital velocity vectors , respectively .
[ synsyns],height=302 ]
|
hubble space telescope observations of active asteroid 324p / la sagra near perihelion show continued mass loss consistent with the sublimation of near - surface ice .
isophotes of the coma measured from a vantage point below the orbital plane are best matched by steady emission of particles having a nominal size @xmath0 100 @xmath1 m .
the inferred rate of mass loss , @xmath20.2 kg s@xmath3 , can be supplied by sublimation of water ice in thermal equilibrium with sunlight from an area as small as 930 m@xmath4 , corresponding to about 0.2% of the nucleus surface .
observations taken from a vantage point only 0.6 from the orbital plane of 324p set a limit to the velocity of ejection of dust in the direction perpendicular to the plane , @xmath5 1 m s@xmath3 .
short - term photometric variations of the near - nucleus region , if related to rotation of the underlying nucleus , rule out periods @xmath6 3.8 hr and suggest that rotation probably does not play a central role in driving the observed mass loss .
we estimate that , in the previous orbit , 324p lost about 4@xmath710@xmath8 kg in dust particles , corresponding to 6@xmath710@xmath9 of the mass of a 550 m spherical nucleus of assumed density @xmath10 = 1000 kg m@xmath11 . if continued , mass loss at this rate would limit the lifetime of 324p to @xmath121.6@xmath710@xmath13 orbits ( about 10@xmath14 yr ) . to survive for the 100 myr to 400 myr timescales corresponding , respectively , to dynamical and collisional stability
requires a duty cycle @xmath15 .
unless its time in orbit is over - estimated by many orders of magnitude , 324p is revealed as a briefly - active member of a vast population of otherwise dormant ice - containing asteroids .
|
observationally , power spectrum of the intensity fluctuation in our galaxy suggest existence of scale invariant structures in the density over length scales ranging as wide as sub parsec to a few hundred parsec .
these structures are understood in terms of compressible fluid turbulence in the interstellar medium ( ism ) . in present theoretical understanding of ism dynamics
, compressible fluid turbulence plays an important role in the ism evolution , energy transfer , star formation etc .
possible source of energy into the turbulence cascade is however debated , though is mostly ascribed to the supernova shocks as a large scale energy input .
different techniques have been developed to measure the velocity spectrum of the turbulence and hence infer the energy involved in the process .
techniques , originally developed to get the velocity structures in the galaxy , include statistics of centroid of velocities @xcite , velocity coordinate spectrum or vcs @xcite , velocity channel analysis or vca ( @xcite , henceforth lp00 ) , spectral correlation function @xcite etc . we particularly bring attention of our reader here to vca , which has been applied for observations in our galaxy as well as nearby dwarf galaxies like large and small magellanic clouds .
it was found that the velocity fluctuations also follow a power law .
interested reader may have a look at the article lp00 for a complete description of vca , here we outline the basic principle behind this analysis .
vca aims to extract the velocity power spectrum by comparing the power spectrum of intensity averaged over the entire velocity range of observation with the same averaged over a relatively small velocity range , smaller than the expected turbulence velocity dispersion .
differential rotation of our galaxy allows us to have a direct mapping of the velocity values in the position - position - velocity data cube with the line of sight distance to the observing cloud .
this in turn let us estimate the three dimensional power spectrum of the density fluctuations from the galaxy . however , turbulence velocity fluctuations change the velocity to distance mapping and hence also modifies the intensity power spectrum .
this is precisely what vca explores .
@xcite have used a visibility based power spectrum estimator to measure the intensity fluctuation power spectrum of the nearby dwarf galaxy ddo 210 .
they infer that the density power spectrum has a slope of @xmath0 over a length scale range of @xmath1 to @xmath2 pc .
they used the vca with their position - position - velocity data cube and inferred an upper limit to the slope of the velocity power spectrum .
@xcite has extended these study to several external dwarf and spiral galaxies and have estimated the density power spectrum .
recently , @xcite has estimated the power spectrum of the @xmath3 spiral galaxies from things sample and found that the power spectrum of column density follow a power law over the length scales ranging @xmath4 pc to @xmath5 kpc considering the entire sample .
slope of the power spectra for most of these galaxies was found to be within @xmath6 to @xmath7 .
generating mechanism of these large scale structures are yet to be understood .
measuring statistics of the velocity fluctuations would help us understand the dynamical phenomena responsible for these structures .
we note here two main difference between the position - position - velocity data cubes of the emission observation in the galaxy and external spiral galaxies .
line of sight to the observations for the emission from our galaxy is mostly along the plane of the disk , while , the external galaxies are mostly for the face on galaxies and the line of sight is perpendicular to the disk . for observations in our galaxy , different velocity slices of the data cube
can be considered to be at different distances but at the same angular direction in the sky , whereas , for the external galaxies , different velocity slices of the data cube originates from different parts of the galaxy .
this suggests that it would not be wise to directly use the results obtained in lp00 while inferring observations from external galaxies .
as there exist no direct position to velocity mapping for the external spiral galaxies , an analytical investigation on the effect of the turbulence velocity is not straight forward , we refer to the numerical methods here . in this letter
we perform numerical simulation to access how the intensity power spectrum is modified with the velocity fluctuations for the spiral galaxies .
section ( 2 ) gives a brief outline to our approach and the section ( 3 ) describes the numerical investigation we have performed . results and discussions are discussed in section ( 4 ) .
we conclude in section ( 5 ) .
we adopt a coordinate system centred at the cloud in concern ( or the external galaxy ) with the line of sight direction aligned to the @xmath8 axis , such that @xmath9 where @xmath10 is a two dimensional vector in the sky plane . at small optical depth limit
, the specific intensity of radiation @xcite with rest frequency @xmath11 originated from a gas at @xmath12 having temperature @xmath13 is given as @xmath14 where @xmath15 , @xmath16 is the frequency of observation , @xmath17 and @xmath18 is the line shape function : @xmath19.\ ] ] here @xmath20 is the line of sight component of the velocity of the gas and @xmath21 : boltzmann constant , @xmath22 : mass of hydrogen atom ] is the thermal velocity dispersion . in practice , the observed specific intensity is always averaged over a velocity width @xmath23 around @xmath24 , hence @xmath25 clearly , @xmath26 where , @xmath27 is the column density . in practice , as the emission from the galaxy falls of to zero beyond a certain velocity , say @xmath28 , it is sufficient to carry the integration in the above equation over the range @xmath29 to @xmath30 .
here we consider that the galaxy has no overall motion .
compressible fluid turbulence in the ism of the galaxies induce scale invariant fluctuations in the density as well as velocity .
power spectrum of the column density fluctuation is given as @xmath31 where the averaging is performed over all possible values of @xmath32 and all directions assuming homogeneity and isotropy in the random fluctuations .
we define the power spectrum of the observed intensity fluctuation as @xmath33 here we have assumed the homogeneity and isotropy of the intensity field and that the intensity power spectrum is independent of the centroid of the velocity @xmath24 .
clearly , @xmath34 this has been used extensively in literature to estimate the column density power spectrum of our galaxy , external dwarf @xcite and spiral galaxies @xcite .
the power spectra is found to follow power laws indicating turbulence to be operational , hence @xmath35 the line of sight component of velocity @xmath20 has component from the systematic rotation of the galaxy @xmath36 , as well as form the random motion of the cloud because of turbulence @xmath37 , i.e , @xmath38 .
power spectrum of the turbulent velocity component is also expected to follow power law @xmath39 lp00 has investigated the nature of @xmath40 in detail in order to estimate the modification of the power spectrum by turbulence . they show that for observed gas in our galaxy , for @xmath41 , @xmath42 while as we expect , @xmath43 hence , by estimating the power spectra in two different limits above one can infer the velocity power spectrum slope .
this method , usually known as velocity channel analysis , has been used to estimate the velocity fluctuation power spectrum of in our galaxy and nearby dwarf galaxies .
it is important to realise that the particular direct linear mapping between @xmath44 and @xmath8 was exploited in vcs , is rather different when we consider emission from an external galaxy .
considering tilted ring model , in later case , @xmath36 depends on the galacto - centric radius , position and inclination angles .
moreover , at a given @xmath24 with @xmath45 , only a part of the galaxy s disk is visible . in this letter
we attempt to see how the turbulent velocity modifies the power spectrum for the external spiral galaxies and investigate if a similar procedure as vcs can be adopted to estimate the velocity fluctuation spectrum . in order to simulate the emission from the external galaxies we adopt the following simplifications .
note that in this work we are not interested to simulate all aspect of the emission from the external galaxies , rather we are interested in investigating the modification in the power spectrum due to the turbulent velocity , which justifies these simplifications . * in case of a spiral galaxy the average profile @xmath46 varies with the galacto - centric radius as well as in vertical direction .
this leads to modification in the power spectrum as discussed in @xcite . here
, we consider @xmath46 to be independent of @xmath12 .
it is to be noted that this simplification also means that we are assuming the galaxy s disk to be thick .
we shall discuss the effect of this in the conclusions section . * systematic rotation of the galaxy @xmath36 , depends on the inclination and position angle as well as galacto - centric radius . to simplify matter we assume here that the position angle and inclination angle @xmath47 of the galaxy do not change with galacto - centric radius and adopt a flat rotation curve with tangential velocity @xmath48 .
in such a case , we can write @xmath49 * ism is known to be in pressure equilibrium ( see @xcite and references therein ) and more than one temperature gas coexists in it .
this means , in principle , we need to consider different temperatures at different part of the galaxy and hence a varying @xmath50 .
this would give rise to an additional fluctuation in the observed specific intensity .
here we assume that the gas across the galaxy is at a constant temperature and we adopt the temperature that of the cold gas .
we divide the simulation volume into @xmath51 individual grids ( cubes ) which represents individual cloud with associated @xmath52 and @xmath44 .
as discussed in the previous section , we have kept @xmath53 constant everywhere across the simulation volume .
it can be shown that for a thick cube the power spectrum of the three dimensional density distribution has the same slope of that of the projected density , like the column density here ( @xcite , lp 00 ) .
hence , we generate @xmath54 and @xmath55 such that they follow gaussian distribution with power spectrum of slope @xmath56 and @xmath57 respectively . @xcite have estimated the amplitude of the fluctuations for six galaxies of the things sample @xcite .
they found that the amplitude of the column density fluctuations are approximately @xmath58 that of the mean column density for the galaxies .
hence , here we consider @xmath59,\ ] ] with @xmath60 and @xmath61 , the random component because of turbulence .
@xcite has estimated the velocity dispersion for the galaxies in the things from moment - ii maps and found that the turbulence velocity dispersion @xmath62 vary in the range @xmath63 to @xmath64 km sec@xmath65 .
on the other hand , flattening velocity of the rotation curve for the same galaxies has the range @xmath66 to @xmath67 km sec@xmath65 @xcite .
here we adopt @xmath68 , with @xmath69 . considering the gas at a temperature of @xmath70 k , we adopt the thermal velocity dispersion to be @xmath71 , where @xmath72 .
note that the actual value of @xmath48 is unimportant here . in order to see the effect of the density and velocity fluctuations in @xmath40
, we consider @xmath73 , @xmath74 and all combinations of these . in literature
the velocity and density fluctuations because of turbulence are assumed to be uncorrelated .
here we consider two cases , either @xmath52 and @xmath55 are completely uncorrelated or completely correlated .
galaxies with higher inclination angles can have scale mixing in the projected direction .
rotational velocity effects also manifest more at higher inclination angle because of the @xmath75 factor in eqn .
we choose the inclination angle to be @xmath76 .
figure ( 1 ) shows different aspects of the simulated galaxy for @xmath77 , @xmath78 , @xmath79 and for the case when @xmath52 and @xmath55 are uncorrelated .
column density map is shown in figure ( 1a ) , while the line of sight velocity @xmath44 is shown in figure ( 1b ) .
figure ( 1c ) shows @xmath80 for a certain value of @xmath24 with @xmath81 . in this case
only a part of the galaxy is visible and the area over which @xmath40 can be estimated is restricted .
the integrated line profile of the galaxy is shown in figure ( 1d ) . assuming the centre of the galaxy to be at the centre of the simulation volume and the inclination angle to be @xmath82
, we generated the specific intensity given in eqn .
( 1 ) with velocity resolution same as thermal velocity dispersion . range of @xmath24 is chosen such that all the emission from the model galaxy is included .
we estimate @xmath40 defined in eqn .
( 7 ) for ( a ) @xmath23 covering the entire emission , i.e , @xmath83 and ( b ) @xmath81 .
results are discussed in the next section .
we first discuss the results for the case when we assume that the @xmath52 and @xmath55 are uncorrelated . figure ( 2 ) shows the power spectra @xmath40 with @xmath83 . here
thedot - dash " lines corresponds to the power spectra for @xmath78 while the dash - dash " lines are for @xmath84 .
power spectra corresponding to @xmath77 are shown in blue and @xmath85 are shown in black .
the solid lines corresponds to power law with slopes @xmath86 and @xmath6 respectively
. all curves are shifted in vertical direction arbitrarily for clarity .
it is clear that the power spectra of the intensity with @xmath83 has the same slope that of the @xmath52 , irrespective of the slope of the velocity power spectra . as in our simulation
we have considered a thick disk for the galaxy , we expect the slope of the power spectrum of column density to be same with that of @xmath52 .
hence , nature of the power spectra in figure ( 1 ) is quite expected and is just a verification of the eqn ( 8) .
we estimate the power spectra @xmath40 for all four combinations of @xmath56 and @xmath57 and @xmath87 with @xmath81 , where we expect to see the effect of turbulence velocity @xmath55 in the intensity power spectra .
figure ( 3 ) shows the corresponding power spectra as in figure ( 1 ) with @xmath81 .
before we interpret these curves , we need to realise that here emission is coming from only a part of the galaxy s disk , as shown in figure ( 1c ) .
in such a case , the observed intensity power spectra would have effect of the shape of the window where the emission is coming from .
this is precisely why all four curves in figure ( 3 ) have similar nature for @xmath88 and we can only expect to see the effect of @xmath56 or @xmath57 beyond that .
interestingly , for @xmath89 the power spectra is independent of the values of @xmath56 and is different for different @xmath57
. this can be the effect of velocity modification , i.e , effect of the line of sight component of the turbulent velocity @xmath55 on the intensity power spectrum .
as this is independent of @xmath56 , the nature of the velocity modification is different than what is expected from the result of lp00 ( see eqn .
( 10 , 11 ) ) . to investigate how @xmath40 for @xmath89 changes with different values of @xmath57
, we performed the same simulation with @xmath77 and @xmath79 , for values of @xmath57 ranging @xmath90 to @xmath91 . for each case
, we fit the power spectra at @xmath89 with a power law of the form @xmath92 and note the best fit values . since in simulation we have not added any contribution from the observational uncertainties , we only use the sample variance generated noise to do this fit .
results are shown in figure ( 4 ) , where we have plotted @xmath57 in x axis with @xmath93 with errors form the fit in y axis .
we use a second order polynomial to empirically fit the values of @xmath93 against @xmath57 , i.e , @xmath94 with @xmath95 and @xmath96 values @xmath97 and @xmath98 respectively .
next we consider the case when the fluctuations in @xmath52 and that in @xmath55 are correlated , in fact we use the same set of gaussian random variables to represent them .
hence , in this case , we only consider variation of @xmath56 since @xmath52 and @xmath55 fluctuations are scaled up version of the same original gaussian variables . as expected the power spectrum with @xmath83
has the slope of the column density power spectrum and the corresponding plot is exactly similar to figure ( 2 ) except from minute differences arising because of statistical fluctuations .
we estimated the power spectrum of intensity with @xmath99 .
these power spectra also show trends as in figure ( 3 ) , for @xmath88 it is dominated by the windowing effect and for larger @xmath100 , it is a power law with exactly similar variation of the slope of the spectra as in figure(4 ) .
we do not show these plots here to avoid repetition .
to summarise , for both the cases @xmath52 and @xmath55 uncorrelated and perfectly correlated the power spectrum @xmath40 with @xmath83 , we always reproduce power law with the same slope as @xmath52 power spectrum and with @xmath81 , at larger @xmath100 the power spectrum has a certain slope that only correlates with the slope of the velocity spectrum .
in this letter we investigate how the intensity fluctuation power spectrum is related to the number density and the line of sight component of the turbulent velocity for external galaxies .
we found that for scale invariant fluctuations in both density and velocity , when the emission is integrated over all velocity range , the intensity fluctuation power spectra follow a power law that has the same slope of the power spectra of the number density fluctuations .
we consider a thick disk for the galaxy in this case , in case of thin disk , the intensity spectra would have slope shallower by order unity ( @xcite , lp00 ) . when the emission is integrated over a velocity range smaller than the turbulent velocity dispersion , due to the galactic rotation , only a part of the galaxy is visible .
effect of the density or the velocity fluctuation in the intensity power spectra can be inferred only for higher values of @xmath100 and for a relatively narrow range of @xmath100 values .
we found that the slope of the spectra at these range follow approximately a power law with slope @xmath93 related nonlinearly only to the slope of the velocity power spectrum @xmath57 and independent of the power spectrum of the density .
this is a different result compared to lp00 , where it is expected that @xmath101 .
note that , this result is based on a power law fit to the power spectrum for a shallower range of @xmath100 .
nevertheless , it clearly demonstrates that velocity modification of the power spectrum for external galaxies is quite different from that in our galaxy . in our simulations
, we did several simplifications .
first we ignored the overall profile of the galaxy .
however , as it is shown in @xcite , effect of this profile is to modify the power spectra at lower @xmath100 , which considering the galaxy spanning over the simulation volume would happen at @xmath102 and is not the regime of interest for the velocity modification .
similarly , including a realistic rotation curve would have only change the window over which the power spectrum can be investigated for velocity modification . given these
, our results also stand out for real galaxies .
effect of varying thermal velocity dispersion over the galaxy , on the other hand , is more complicated and needs to be investigated in detail separately .
finally , we discuss the feasibility to use the relation we obtain between @xmath93 and @xmath57 in figure ( 4 ) for a real observation . considering the galaxy spread over our entire simulation volume ,
the dynamic range in @xmath100 from simulation is approximately same as that in the things observation .
@xcite has estimated the power spectra of 18 nearby spiral galaxies from thigns sample .
given the baseline coverage of observation , they could estimate the power spectra till @xmath103 the largest baseline .
this is because at higher baselines the baseline coverage is restricted and signal to noise is insufficient to estimate the power spectrum with statistical significance .
we use @xmath104 for performing our simulation . as the largest @xmath100 in simulation
is @xmath105 , @xmath103 the available baseline rang of things corresponds to @xmath106 in our simulation , making these observation inefficient to probe the velocity modification this way .
we choose an inclination angle of @xmath107 for our simulation , any higher inclination angle would result rather even smaller range in @xmath100 over which the values of @xmath93 can be estimated .
we conclude that with present telescopes to use vca techniques for external galaxies , we need high integration time .
an alternate method to estimate the turbulent velocity spectra of the external galaxies would be more useful .
we aim to investigate along this direction in future .
pd acknowledges useful discussion with somnath bharadwaj , jayaram n. chengalur , nirupam roy and nissim kanekar .
this work is supported by the dst inspire faculty fellowship award [ ifa-13 ph 54 ] and performed at indian institute of science education and research , bhopal .
pd is thankful to narendra nath patra , sushma kurapati and preetish kumar mishra for reading the earlier version of the draft and providing valuable comments .
d. , lazarian a. , 2009 , in revista mexicana de astronomia y astrofisica conference series vol .
36 of revista mexicana de astronomia y astrofisica , vol .
27 , line - of - sight statistical methods for turbulent medium : vcs for emission and absorption lines .
|
we use numerical simulations to investigate effect of turbulent velocity on the power spectrum of intensity from external galaxies when ( a ) all emission is considered , ( b ) emission with velocity range smaller than the turbulent velocity dispersion is considered .
we found that for case ( a ) the intensity fluctuation depends directly only on the power spectrum of the column density , whereas for case ( b ) it depends only on the turbulent velocity fluctuation .
we discuss the implications of this result in real observations of fluctuations . physical data and process : turbulence - galaxy : disc - galaxies : ism
|
measuring the cosmological redshifts of gamma - ray bursts of all types ( grbs , xrfs , short - hard bursts ) is crucial for our understanding of these events .
the measure of redshifts tells us the energy output of the source , it allows the measure of the burst intrinsic parameters ( duration , peak energy etc ... ) , and determines the position of the burster in the history of the universe .
unfortunately , the obtention of spectroscopic redshifts requires a succession of non trivial observing steps : reliable localization of the prompt emission in x - rays or gamma - rays , quick distribution of the alert to the ground , identification of the afterglow at optical , radio or x - ray wavelenths , and measure of the redshift of the host galaxy in absorption ( when the afterglow is bright ) or in emission ( when the afterglow has faded ) .
this complex sequence of events explains why in february 2005 , 8 years after the discovery of the first afterglow , the web page of jochen greiner contains only 40 spectroscopic redshifts among 263 localized grbs ( of which about 100 have an optical , x - ray or radio afterglow ) .
the statistics is similar for hete-2 grbs @xcite , which has localized 75 events , from which 25 afterglows have been found , and 14 redshifts measured .
the fact that we are currently measuring the redshift of one out of six localized grbs implies strong biases in the observed redshift distribution .
for instance grbs at high redshift are very difficult to detect in the optical range , and the lack of grbs with z @xmath0 5 in the current sample may be the consequence of selection effects .
the intrinsic faintness of xrfs makes them difficult to detect beyond z=1 , again biasing their observed redshift distribution .
it is precisely this small fraction of grbs with a redshift and the likely existence of biases in the observed redshift distribution which constitute the rationale for the construction of redshift indicators. in order to complement spectroscopic redshifts usefully , redshift indicators must provide redshift estimates for most grbs detected at high energies .
the construction of such indicators is considered in section [ building ] .
section [ issues ] addresses some open issues in the construction of redshift indicators .
possible applications of these indicators are presented in section [ appli ] .
gamma - ray bursts have a broad dispersion in peak luminosity ( by a factor 10@xmath1 at least ) , which prevents using them as standard candles straightforwardly . after the the measure of a dozen grb redshifts ,
much work has been devoted to the search of luminosity estimators with an intrinsic dispersion smaller than the peak luminosity .
this work brought two complementary results : * e@xmath2 , the energy radiated by grbs when the beaming factor of the emission is taken into account , the isotropic - equivalent radiated energy .
the identification of a jet - break allows the determination of the opening angle of the jet , giving the beaming factor and the energy radiated , e@xmath2 .
] , appears to be significantly less scattered than the peak luminosity ( l@xmath3 ) or the energy radiated ( e@xmath4 ) measured assuming an isotropic emission @xcite .
the standard deviation of the logarithm of e@xmath2 is 0.35 dex @xcite , providing a measure of luminosity which approaches a standard candle . * many temporal and spectral grb properties
are strongly correlated with the radiated energy ( either e@xmath4 or e@xmath2 ) .
these correlations permit to use these temporal or spectral properties as luminosity indicators , and as redshift indicators by assuming a cosmology .
even if the theoretical justification of these correlations is not fully understood yet , they are currently providing interesting luminosity indicators with typical standard deviations of the order of 0.1 dex @xcite .
aditional details on these luminosity indicators are given in section [ lumi ] below .
the construction of reliable luminosity indicators has started only recently , with the measure of the first grb redshifts , and significant improvements are to be expected in the coming years . in order to facilitate the comparison of existing and future luminosity indicators , we suggest the use a common measure of their quality .
we propose to use @xmath5 , the dispersion of log(d@xmath6/d@xmath7 ) , where d@xmath6 is the luminosity distance predicted by the luminosity indicator , and d@xmath7 is the luminosity distance which is measured . using this measure , the pseudo - redshifts proposed by atteia @xcite have @xmath8 dex , for 24 grbs with a redshift , and @xmath9 dex , for the subsample of 13 hete grbs . from a general point of view , good redshift indicators require combinations of grb parameters which have a small intrinsic scatter and , as possible , a strong dependence on redshift . concerning light curves , the two most publicized correlations are the lag - luminosity correlation @xcite , and the variability - luminosity correlation @xcite .
the construction of luminosity indicators based on these relations requires light curves with a large number of counts , which can only be provided by instruments having large effective area .
luminosity indicators based on grb spectra demand the measure of the three parameters of the band function usually adopted to fit grb spectra @xcite .
they are based on the so - called e@xmath10-e@xmath4 relation @xcite , on the e@xmath10-l@xmath4 relation @xcite , or on the e@xmath10-e@xmath2 relation @xcite designates the energy of the maximum of the @xmath11 spectrum of the prompt emission .
e@xmath4 and e@xmath2 have been defined in the footnote in section [ candles ] . ] .
these indicators require instruments having a broad spectral range ( at least two decades ) , and , in the case of the e@xmath10-e@xmath2 relation , a good sampling of the afterglow light curve .
these studies have shown that useful luminosity indicators can be constructed for grbs , even if they depend on the details of each grb detector ( effective area , spectral range , time and energy resolution ) , and must be adapted for each mission .
recently , the consistency of redshift indicators obtained from the light curves and from the spectra has been demonstrated @xcite . despite the rapid progress of the last years ,
some open issues remain , which are presented in the next section .
the importance of the empirical relations used in the definition of luminosity indicators , and the fact that they are based on a small number of grbs with redshifts , raised questions on their overall validity and on their applicability to the entire grb population .
various authors attempted to check the validity of the e@xmath10-e@xmath4 , and of the e@xmath10-e@xmath2 relations by verifying if , for every grb in the batse sample , one can find a redshift which puts it along the correlations observed for the grbs with a redshift .
the conclusions of these efforts are diverse : while lloyd et al .
@xcite actually _ predicted _ the e@xmath10-e@xmath4 relation , nakar & piran @xcite , and band & preece @xcite found that the majority of batse grbs can not follow this relation , but this conclusion was later contradicted by ghirlanda et al .
@xcite , and bosnjak et al .
these contradictory results may reflect the difficulty of properly accounting for the uncertainties in the measure of individual e@xmath10 for all grbs , or the existence of a class of outliers ( see below ) , treated differently in these studies .
luminosity indicators are not well calibrated , essentially due to the small number of grbs with a spectroscopic redshift , and to the restricted range of redshifts measured ( z=0.1 to 4.5 ) .
the number of grbs available for calibration is further reduced by the fact that many grbs have their light curves , spectra or afterglows known with insufficient accuracy to compute luminosity indicators reliably . with these restrictions in mind ,
the accuracy of present - day luminosity indicators is not too bad since the best of them have a standard deviation smaller than 0.1 dex .
the association of grb 980425 with the energetic supernova 1998bw at z=0.0085 revealed a sub - energetic burst , whose radiated energy is one thousand times smaller than other grbs with a redshift .
as a consequence this burst does nt fit the empirical relations used to define luminosity indicators .
the nature of the link between grb 980425 and the bulk of the classical grb population has not been elucidated yet , and we should wait the detection of additional events of this type to clarify the nature of the outliers .
recently , grb 031203 has also been considered has a potential outlier to the e@xmath10- e@xmath4 and to the lag - luminosity relations @xcite .
the case of this burst is , however , much less clear than the case of grb 980425 because its e@xmath10 has not been measured .
[ durete_fluence ] illustrates the contradiction raised by this burst : from the point of view of its prompt gamma - ray emission , grb 031203 appears fully comparable with the grbs detected with hete-2 ; on the other hand , this burst , the only integral grb with a redshift , does not follow the e@xmath10- e@xmath4 relation , while all the 14 hete grbs with a redshift follow this relation . in our opinion ,
the situation of grb 031203 with respect to the empirical e@xmath10- e@xmath4 , and lag - luminosity relations remains unclear .
[ durete_fluence ] -e@xmath4 relation , while the 14 grbs with a redshift detected with hete follow this relation . , title="fig:",width=359 ] x - ray flashes ( xrfs ) are soft grbs with low e@xmath10 , typically e@xmath10@xmath12 kev .
if they follow the e@xmath10-e@xmath4 relation , they must be intrinsically faint and can only be seen at low redshifts .
the redshifts of two xrfs have been measured to date : xrf 020903 , at z=0.25 , and xrf 040701 , at z=0.21 .
these measures put these two bursts on the e@xmath10-e@xmath4 relation , and suggest that luminosity indicators for grbs could also be applicable to xrfs .
the situation is however more complex than this simple picture since grbs at high redshift ( z=5 - 10 ) will also appear as x - ray flashes . due to the absence of grbs with a measured redshift greater than 5 , it is not clear if luminosity indicators can make the difference between xrfs from nearby intrinsically soft grbs and classical hard grbs at high redshift .
another puzzle is connected with the lack of supernova light in several xrfs @xcite .
since all nearby grbs seem to show sn light @xcite , this observation can be interpreted as an evidence that these xrfs lie at redshifts significantly larger than the values predicted by the redshift indicators ( but other interpretations are also possible @xcite ) . until we know more about the distances of xrfs , it is difficult to affirm with certainty that grb redshift indicators are also valid for xrfs . the question of redshift indicators for short - hard grbs is not adressed here since we have not yet measured the distance of a short - hard grb .
furthermore , the recent observation of a very bright gamma - ray flare from the galactic magnetar sgr 1806 - 20 , suggests that a significant fraction of the short - hard bursts could be due to extragalactic magnetars @xcite .
this could make short - hard grbs fundamentally different from long grbs for which luminosity indicators have been shown to work .
we briefly present below some ( not all ) of the studies which have used luminosity - redshift indicators in the last years .
the availability of redshift indicators exclusively based on the prompt emission opens the possibility to quickly distribute an evaluation of the distance of newly detected grbs .
this is especially important for grbs at high redshifts ( z@xmath05 ) which can only be identified with a rapid and specific follow - up @xcite . in order to facilitate the follow - up strategy of ground observers ,
the hete team has set up a program which performs the spectral analysis of grbs automatically , as soon as the data become available ( usually in the minutes following the burst ) , and evaluates a redshift indicator based on the e@xmath10- e@xmath4 relation .
grbs offer an unabsorbed view on the history of the star formation rate ( sfr ) .
the distribution of the pseudo - redshifts of hundreds of batse grbs has been computed by various authors ( see @xcite and ref . therein ) .
these studies show that grbs offer an interesting alternative to constrain the sfr beyond z @xmath13 , but we should keep in mind that redshift indicators have not yet been calibrated in this range of redshifts . the availability of accurate luminosity indicators for grbs can be used to measure luminosity distances independent of the redshift , and to derive the cosmological parameters . following @xcite ,
various authors assessed the constraints that grbs impose on the cosmological parameters . whether or not current data are sufficient in number and accuracy to constrain the cosmological parameters is the subject of an on - going debate @xcite , but there is a strong consensus on the fact that grbs will become a valuable tool in this field in the coming years .
luminosity and redshift indicators for grbs will certainly play a increasing role in the future . the growing number of redshifts is expected to bring several improvements to this field : a more accurate calibration of existing luminosity indicators , the construction of new indicators , a better understanding of systematic effects ( e.g. luminosity evolution ) and of the issues discussed in section [ issues ] .
the progress in this field may be fast in the swift era thanks to the small error boxes provided by the xrt , which permit efficient afterglow searches , and raise the hope that a significant fraction of swift grbs will have spectroscopic redshifts .
a recent good news is that the luminosity indicator based on the variability of the light curve , which was developped for batse bursts @xcite , seems to be directly applicable to swift grbs @xcite .
we would like to conclude with strong incentive to observers to quickly follow - up at x - ray and nir wavelengths , grbs with high pseudo - redshifts .
the identification of grbs at high redshifts ( z @xmath14 ) is a difficult task which requires deep , prompt observations with powerful instruments .
this can not be done for just any grb , and the luminosity indicators discussed in this paper could be a decisive tool to ensure the success of future searches for high - redshift grbs .
|
the measure of the distances and luminosities of gamma - ray bursts ( grbs ) led to the discovery that many grb properties are strongly correlated with their intrinsic luminosity , leading to the construction of reliable luminosity indicators .
these grb luminosity indicators have quickly found applications , like the construction of pseudo - redshifts , or the measure of luminosity distances , which can be computed independently of the measure of the redshift . in this contribution
i discuss various issues connected with the construction of luminosity - redshift indicators for gamma - ray bursts .
[ 1999/12/01 v1.4c il nuovo cimento ]
|
it can be shown using the standard calculus optimization technique that of all rectangles with a given perimeter , the square has the largest area .
( it also follows that of all rectangles with a given area , the square has the smallest perimeter . )
there is also an easy geometric way to prove this .
the picture below shows two rectangles : one of them , @xmath0 , is a square , and the other , @xmath1 , is not .
they have the same perimeter ( @xmath2 ) , but the square has a larger area since the areas of the shaded parts are equal , but the square also contains a region that the other rectangle does not .
the above area inequality can also be seen algebraically : @xmath3 as proved on p. 50 in @xcite , even among all quadrilaterals with a given perimeter , the square has the largest area .
we can ask a similar question in 3d : of all rectangular prisms with the given volume , which one has the smallest surface area ? or , equivalently , of all rectangular prisms with the given surface area , which one has the largest volume ?
the answer to both of these questions is a cube .
notice that these problems are too general for a single variable calculus course : there are too many degrees of freedom .
namely , there are three unknowns ( three dimensions ) , but only one constraint ( the given surface area or the given volume ) .
so typically in the first semester calculus class we only consider the 2-variable problem of optimizing a square prism .
this would actually be sufficient for our work below , however , we ca nt resist the temptation to show our reader the following geometric argument that works for all rectangular prisms .
suppose we have any rectangular prism that is not a cube .
choose any face that is not a square and make it the base .
consider the square that has , say , the same area .
it will have a smaller perimeter .
construct a prism with the same height as the original one : it will have the same volume but a smaller surface area !
( it will have the same volume because both the area of its base and its height are the same as those of the original prism ; and a smaller surface area because the areas of the base and the top are the same but the area of the sides is smaller since the perimeter of the base is smaller . )
thus a prism that is not a cube can not have the smallest possible surface area among all rectangular prisms with a fixed volume .
now we are ready to consider some more interesting problems .
some version of the following rectangular field optimization problem can be found in almost every calculus textbook ( see e.g. exercise 7 on p. 337 in @xcite ; example 1 on p. 258 in @xcite looks different at first , but actually is an equivalent optimization problem ) . [ cols= " < , < " , ] the reader is invited to check that the optimal dimensions are @xmath4{500/\pi}$ ] cm and @xmath5{500/\pi}$ ] cm .
@p8.9cm@p3cm@ notice that in the optimal can @xmath6 , and the same relationship holds for an optimal square prism box if we let @xmath7 be the distance from the center of the base to any side ( note that @xmath7 is the radius of the inscribed circle , called the inradius of the square ; the line segment from the center of a regular polygon to the middle of any side is also called its apothem ) . & is this a coincidence
? we will show that it is not , but rather these two problems ( optimizing the can and optimizing the square prism ) are equivalent .
let us see how the volume and the surface area of the can and the square prism are related : @xmath8
so @xmath9 now , @xmath10 notice that the expressions for both surface areas contain twice the area of the base , and @xmath11 .
if only the other two terms in these expressions were related in the same way , namely , if @xmath12 was @xmath13 times larger than @xmath14 , that would mean that @xmath15 was also @xmath13 times larger than @xmath16 , which , in turn , would mean that the two problems are equivalent .
so here is our question : is @xmath17 ? it is easy to see that these quotients are @xmath18 and @xmath19 , so indeed they are equal .
but is that a coincidence , or is there something deeper here ?
why @xmath17 ?
let us rewrite the above equation as follows : @xmath20
. now look again at the above equation in terms of @xmath7 : @xmath21 . on each side ,
the denominator is the derivative of the numerator !
this property is briefly illustrated below , while its formal proof for any regular polygon is given in @xcite and a more general case is explored in @xcite .
imagine increasing the radius of a circle by @xmath22 ( see picture below ) .
its area is increased by @xmath23 .
that s why the derivative of the area is @xmath24 .
the same is true for the square !
now , both areas ( of the circle and of the square ) are quadratic power functions of @xmath7 .
for any two quadratic power functions , say , @xmath25 and @xmath26 , their derivatives are @xmath27 and @xmath28 respectively , so the ratios of the function to its derivative would be @xmath29 .
it does not matter what the coefficient @xmath30 or @xmath31 is , since it appears in both the function and its derivative and gets canceled in the ratio .
so the above observation is not at all a coincidence .
moreover , as we will see in the next section , this applies to any prism that has a regular polygon as the base !
suppose we want to construct a box in the shape of some other prism whose base is a regular polygon
. what would be the optimal shape then ?
for example , given the desired volume , what ratio of the height @xmath32 to the inradius @xmath7 would minimize the surface area ( and thus minimize the amount of material needed ) ? as we know , area grows proportionally to the square of the length .
therefore , the area of any regular polygon is a square function of the inradius , i.e. of the form @xmath33 where @xmath34 is some constant .
the perimeter of the polygon is then @xmath35 .
this means that both the volume and the surface area of any such prism differ from those of a square prism by a constant multiple ( namely , @xmath36 ) , thus the problem of optimizing any such box is equivalent to the problem of optimizing a square prism .
so , in all cases the optimal shape is the one for which @xmath6 !
the following is a modification of problem 9 on p. 979 in @xcite . of all ellipses inscribed in a semi - circle of radius 1 , find the one with the largest possible area .
hint : if the semicircle is given by the equation @xmath37 , @xmath38 , the ellipse should have equation of the form @xmath39 . &
let us dig a bit deeper and calculate the points of intersection of the ellipse with the semicircle .
they are @xmath42 .
notice that these are exactly vertices of the largest - area rectangle inscribed in a semi - circle ( see e.g. example 5 on p. 336 in @xcite ; see also p. 130
in @xcite for a generalization of this problem ) : are the problems of optimizing the ellipse and optimizing the rectangle inscribed in a semi - circle equivalent ?
it certainly appears so , but what do the inscribed ellipse and the inscribed rectangle have in common ?
try to figure it out
! * acknowledgments . *
the author would like to thank her colleagues at fresno state university for numerous insightful conversations about calculus problems as well as for their encouragement to write this paper , and the anonymous reviewer for many good suggestions on its improvement
. 1 w. briggs , l. cochran , _ calculus : early transcendentals _ , 2nd ed . , pearson education , inc . ,
boston , 2015 .
m. dorff and l. hall , solids in @xmath43 whose area is the derivative of the volume , _ college math . j. _ , * 34 * no . 5 ( 2003 ) , 350358 j. miller , differentiating area and volume , _ two - year college math j. , _ * 9 * no . 1 ( 1978 )
, 4748 i. niven , _ maxima and mimima without calculus _
, dolciani mathematical expositions no .
6 , maa , 1981 r. silverman , _ essential calculus with applications _ , dover publications , inc .
, new york , 1989 j. stewart , _ calculus , early transcendentals _ , 5th ed .
, thomson learning , inc . , belmont , 2013
|
in the first semester calculus course we learn how to solve optimization problems such as maximizing the volume of a box or a can given its surface area , i.e. the amount of material , or minimizing the surface area given the desired volume .
whether you are a student or a teacher , have you ever wished you knew the answer to a problem after one glance at it , without doing long calculations ?
the patterns shown and explained below will enable you to do it for a few `` standard '' problems , but more excitingly , they will illustrate some very beautiful problem solving techniques and give insights into how different areas of mathematics are connected .
|
neutron stars contain matter in one of the densest forms found in the universe .
their central density ranges from a few time the density of normal nuclear matter to about one order of magnitude higher , depending on the star s mass and the equation of state ( eos ) .
neutron stars therefore provide us with a powerful tool for exploring the properties of such dense matter . in the last decades ,
this tool was applied to , among other topics , the determination of the eos of dense , charge neutral , @xmath3-equilibrated matter by means of comparing the theoretical predicted properties with observations of neutron stars .
this was attempted by studying , for example , the maximum stable star mass @xcite , the minimum rotation period @xcite , or the thermal behaviour ( tsuruta 1966 , schaab et al .
1996 , page 1997 ; see balberg , lichtenstadt & cook 1998 for a recent review ) . recently , strohmayer et al .
and van der klis et al .
discovered with the rossi x - ray timing explorer ( rxte ) kilohertz quasi - periodic oscillations ( qpos ) in the x - ray brightness of low - mass x - ray binaries ( lmxbs , see van der klis 1997 for a recent review ) . in subsequent observations ,
three qpos were often detected simultaneously in a given source .
the frequency separation between both qpos is almost constant , although the frequencies of the two qpos themselves vary by several hundred hertz . up to now
, the only exceptions are the atoll sources 4u1608 - 52 and 4u1735 - 44 and the z - source scorpius x-1 in which the frequency separation varies with the luminosity by roughly @xmath4% @xcite .
psaltis et al . found that the qpo data of the other lmxbs are consistent both with being constant and with varying by a similar fraction as in the two sources quoted before .
the frequency of oscillations during type i x - ray bursts detected in some sources are consistent with the frequency separation of the two qpos or with its first overtone . only in the source 4u1636 - 536
the averaged frequency separation , @xmath5 hz , and the half of the frequency in type i bursts , @xmath6 hz , differ by approximately 15% @xcite . in the power spectrum of bursts of
the same source miller found a second signal at 290 hz @xmath7 .
these observations strongly favour the beat frequency model where the frequency separation between the two qpos originates from the stellar spin , whereas the higher qpo is produced by accreting gas in a stable , nearly circular orbit around the neutron star .
though , it has to be clarified how the slightly varying frequency separation and the small deviation of the frequency separation from the burst frequency can be incorporated into this model . beside the two kilohertz qpos and the burst oscillations ,
also qpos with frequencies of a few tens of hertz were detected in some sources .
their frequencies correlate with the high frequency kilohertz qpos .
these low frequency qpos were interpreted by stella & vietri to originate from the lense - thirring precession of the accreting disc due to the frame dragging effect of the rapidly spinning neutron star @xcite . both , the identification of the high frequency kilohertz qpos with the orbital frequency of a stable circular orbit and the low frequency qpos with the frame dragging frequency of the same orbit allow to constrain the mass of the neutron star and also the eos of neutron star matter @xcite . if the evidence for the detection of qpos of the innermost stable circular orbit in the sources 4u1608 - 52 , 4u1636 - 536 @xcite , and 4u1820 - 30 @xcite can be confirmed by future observations , the constraints are rather severe , allowing only a few stiff eoss .
the frequency separation of the two kilohertz qpos show that the neutron star in lmxbs are rapidly rotating with periods ranging from 2.5 ms to 4 ms .
it can therefore be expected that the geometry of the neutron star and its exterior space time is non - spherical .
since the innermost stable orbit is located at only a few kilometres above the star s surface , the deviation from the kerr space time are large and should not be neglected . in order to compare theoretical neutron star models with qpo observations ,
a completely general relativistic calculation of the rotating neutron star structure and space time geometry is therefore necessary . in order to study the impact and the discrimination power of the qpo data in greater detail , we select a broad collection of modern eoss , which were obtained utilising numerous assumptions about the dynamics and composition of super dense matter . to mention several ,
these are : the many - body technique used to determine the eos ; the model for the nucleon - nucleon interaction ; description of electrically charge neutral neutron star matter in terms of either only neutrons and protons in generalised chemical equilibrium ( @xmath3 equilibrium ) with electrons and muons , or nucleons , hyperons and more massive baryon states in @xmath3 equilibrium with leptons ; hyperon coupling strengths in matter ; inclusion of meson ( @xmath8 , @xmath9 ) condensation ; treatment of the transition of confined hadronic matter into quark matter ; and assumptions about the true ground state of strongly interacting matter ( i.e. , absolute stability of strange quark matter relative to baryon matter ) .
the paper is organised as follows . in sect .
[ sec : equations ] we summarise the equations which govern the space time structure and compare the approximate values of the orbital frequencies , the radius of the innermost stable orbit , and the lense - thirring precession frequencies with the respective values from the exact numerical solution of einstein s equations . the physics of the eoss is discussed in sect .
[ sec : eos ] .
the high frequency kilohertz qpos and their interpretation in combination with their compatibility with the different eoss are discussed in sect .
[ sec : constr1 ] .
the implications of the identification of the low frequency qpos with lense - thirring precession are presented in sect .
[ sec : constr2 ] .
we summarise our results , the constraints to the neutron star masses , and the conclusions concerning the neutron star eos in sect .
[ sec : concl ] .
the stationary , axis - symmetric , and asymptotic flat metric in quasi isotropic coordinates reads @xmath10 where the metric coefficients @xmath11 are functions of @xmath12 and @xmath13 only .
the metric coefficients are determined by the einstein equation ( @xmath14 ) @xmath15 and the energy - momentum conservation @xmath16 where @xmath17 is the stress - energy tensor of an ideal fluid with the 4-velocity @xmath18 the energy density @xmath19 , and the pressure @xmath20 .
the lorentz factor @xmath21 is given by @xmath22 , hence @xmath23 @xmath24 is the angular velocity of the fluid with respect to an observer at infinity .
the proper velocity @xmath25 of the fluid with respect to the local eulerian observer @xmath26 @xcite is given by the equation @xmath27 note that if the fluid were at rest with respect to the observer @xmath26 , @xmath28 , it would not necessarily be at rest for an inertial observer at infinity , since @xmath29 .
this phenomena of _ dragging of the inertial frame _ was first studied by lense & thirring and turns out to be important for the investigation of the lense - thirring precession of the accreting disc ( see sect .
[ sec : constr2 ] ) .
the four non - trivial einstein equations together with the energy - momentum conservation are solved via a finite difference scheme @xcite .
we follow bonazzola et al . in compactifying the outer space to a finite region by using the transformation @xmath30 .
the boundary condition of approximate flatness can then be exactly fulfilled .
the neutron star model is uniquely determined by fixing one of the parameters : central density , gravitational mass , baryon number , kepler orbiting frequency at the star s equator or at the marginally stable radius , as well as one of the parameters : angular velocity , angular momentum , or stability parameter @xmath31 .
the models of maximum mass and/or maximum rotation velocity can also be calculated .
since @xmath32 and @xmath33 are killing vectors , the components @xmath34 and @xmath35 of the 4-momentum of a freely falling particle are constants of motion and can be identified with the negative of the specific energy @xmath36 ( in units of the rest mass @xmath37 ) and the specific angular momentum @xmath38 , respectively . for a particle motion confined to the equatorial plane the geodesic equation @xmath39 yields : @xmath40 where @xmath41 is the effective potential for the particle motion @xcite . a circular orbit , @xmath42 , is given by the expression for the 4-velocity of the fluid inside the star .
then , the specific energy @xmath36 and angular momentum @xmath38 can be expressed in terms of the lorentz factor @xmath21 ( eq . [ eq : gamma ] ) and the proper velocity @xmath25 ( eq . [ eq : proper.vel ] ) : @xmath43 the circular orbit exists if @xmath44 , thus represents the partial differential @xmath45 . ]
@xmath46 the proper velocity @xmath25 of a particle corotating with the star is given by @xmath47 the circular orbit is stable , if @xmath48 with ( see also cook , shapiro & teukolksky 1992 and datta , thampan & bombacci 1998 ) @xmath49 the zero of @xmath50 determines the radius @xmath51 of the innermost stable or marginally stable circular orbit .
the kepler frequency @xmath52 of a circular orbit as measured by a distant observer is given by the proper velocity @xmath25 through the expression @xmath53 and the lense - thirring precession frequency by @xmath54 the respective values at the innermost stable orbit with radius @xmath51 are denoted by @xmath55 and @xmath56 , respectively .
the circumferential radius @xmath57 of the object is linked to the equatorial coordinate radius @xmath12 by the expression @xmath58 though we shall use the exact expressions for @xmath52 , @xmath59 , @xmath55 , and @xmath60 in our investigation , we give also , for the purpose of comparison , the approximate expressions valid to first order of the dimensionless angular momentum @xmath61 , where @xmath62 and @xmath63 denotes the angular momentum and gravitational mass , respectively .
the approximate expressions have the advantage to constrain the mass and the angular momentum independently of the eos .
to first order in @xmath64 the corresponding expression are @xcite : @xmath65
the eos of neutron star matter is the basic input quantity to the structure equations discussed in sect .
[ sec : equations ] .
its knowledge over a wide range of densities is necessary .
whereas the density at the star s surface corresponds to the density of iron , the density in the centre of a very massive star can reach 15 times the density of normal nuclear matter . since neutron star matter in chemical equilibrium ( i.e. generalized @xmath3-equilibrium ) is highly isospin - asymmetric and
may carry net - strangeness its properties can not be explored in laboratory experiments .
therefore , one is left with models for the eos which depend on theoretically motivated assumptions and/or speculations about the behaviour of super dense matter .
one main source of uncertainty is the competition between non - relativistic versus relativistic models .
though both treatments give satisfactory results for normal nuclear densities , they provide quite different results if one extrapolates to higher densities ( see , e.g , huber et al . 1998 ) .
moreover , one encounters strong differences in the high density regime depending on the underlying dynamics , the many - body approximation and the assumptions about the composition .
the simplest approach describes neutron star matter by pure neutron matter . since
pure neutron matter is certainly not the true ground state of neutron star matter , it will quickly decay by means of the weak interaction into chemically equilibrated neutron star matter , whose fundamental constituents besides neutrons are protons , hyperons and possibly more massive baryons .
even a transition to quark matter ( so - called hybrid stars ) and the occurrence of pion- or kaon - condensates is possible .
the cross section of a neutron star can be split roughly into three distinct regimes @xcite .
the first one is the star s outer crust , which consists of a lattice of atomic nuclei and a fermi liquid of relativistic , degenerate electrons .
the inner crust extends from neutron drip density , @xmath66 , to a transition density of about @xmath67 @xcite . beyond this transition density @xmath68
one enters the star s third regime , that is , its core where all atomic nuclei have dissolved into their constituents , protons and neutrons .
furthermore , as outlined just above , due to the high fermi pressure the core will contain hyperons , eventually more massive baryon resonances , and possibly a gas of free up , down and strange quarks .
the eos of the outer and inner crust has been studied in several investigations and is rather well known .
we shall adopt for these regions the models derived by haensel & pichon and by negele & vautherin , respectively .
the models for the eos of the star s core are discussed in detail in schaab et al . .
an overview of the collection of eoss used in this paper for the core region is given in tab .
[ tab : eos ] .
we have chosen a representative collection of different models in order to check which eoss are compatible with the qpos and their theoretical interpretation . [
cols="^,^,^,^,^",options="header " , ] stella & vietri suggested that a third oscillation with a frequency @xmath69 around @xmath0 hz is also produced at the inner border of the accretion disk by lense - thirring precession .
the observed frequencies @xmath69 ( s. tab .
[ tab : qpo ] ) can again be compared with theoretical neutron star models .
the neutron star models are constructed for a given spin frequency @xmath70 and a fixed , unfortunately unknown , star mass @xmath63 .
the obtained monotone relations @xmath71 and @xmath72 are transformed into the relation @xmath73 .
since the lense - thirring precession frequency @xmath74 is , in first approximation , proportional to @xmath70 the relation @xmath75 as function of @xmath52 is shown in fig .
[ fig : rot.qpo ] for all eoss .
the observations are shown as circles .
the curves depend only weakly on the neutron star mass .
one can therefore draw conclusions from the comparison of the theoretical curves with the observations , although the star masses are unknown . since both @xmath75 and @xmath52 do not depend on @xmath70 in first approximation , the calculation of the theoretical curves for one specific spin frequency
, @xmath76 hz ( this corresponds to the spin frequency of the neutron star in 4u1728 - 34 ) , is sufficient . as can be seen in fig .
[ fig : rot.qpo ] , the observations of ks1731 - 260 and 4u1735 - 44 agree with the theoretical curves of group 1 , which contains the stiffest eoss of our sample .
this is in agreement with the results of stella & vietri .
however , the other observations , including the observations of the z - sources which are not shown , lie above all theoretical curves .
the observed frequency @xmath69 is thus higher than the lense - thirring precession frequency even for the stiffest eoss . if one assume that the observed frequencies correspond to the first overtone of the lense - thirring precession , the atoll - source observations are in the range of the theoretical curves . in the case of the z - sources ,
the detected frequencies @xmath77 are not only too large in most sources , but additionally the slope of the relation @xmath78 of the z - source observations is much higher than the slope of @xmath79 for the theoretically determined models . even the assumption , that the first overtone of the lense - thirring precession frequency @xmath74 is detected @xcite can not explain these discrepancies .
in this paper , we derived models of rapidly rotating neutron stars and strange stars by solving the general relativistic structure equations for a broad collection of modern eoss .
we compared the space time geometry of these models with recently discovered qpos in the x - ray brightness of lmxbs .
if one follows the general beat - frequency interpretation of the kilohertz - qpos , i.e. that the higher frequency qpo originates at a stable circular orbit , one can constrain the mass of the neutron star to a range which depends on the eos .
this mass range is for all sources and for all eoss consistent with a canonical mass , @xmath80 . as it was stated by miller et al . and thampan et al .
the exact lower and upper limits of the neutron star s mass can only be determined by using fully relativistic models of rapidly rotating neutron stars .
the exact limits differ from the approximations with @xmath81 by roughly 10 % . as it was shown by kaaret et al .
and zhang et al . , the observation of a maximum frequency @xmath82 of the high frequency qpo in the sources 4u1820 - 30 , 4u1608 - 52 , and 4u1636 - 536 favour the interpretation that this qpo originates at the innermost stable orbit
. if this interpretation is correct , the mass of the neutron star can be exactly ( within the observational errors ) determined for a given eos .
the approximately obtained mass , @xmath83 , of the source 4u1820 - 30 is larger than the maximum mass of most of the considered eoss .
this conclusion is even strengthened if the observed maximum frequency @xmath82 is compared with the exact neutron star models .
the only allowed eoss of our broad collection are then the uv@xmath84+uvii and rhf@xmath85 , which both describe neutron star matter as consisting of nucleons and leptons only .
the derived masses of the three sources lie in the narrow range between @xmath86 and @xmath87 .
this result is also of some importance for the nature of the object left in the supernova 1987a . during the first ten seconds after the supernova ,
neutrinos were detected @xcite .
this means that a protoneutron star was formed in the supernova .
the fact that up to now no pulsar emission could be detected was interpreted by bethe & brown that the protoneutron star collapsed to a black hole when the star became transparent to neutrinos after roughly 10 s. the estimated value of the baryonic mass @xmath88 @xcite gives thus an upper limit to the maximum gravitational mass of a neutron star : @xmath89 .
this limit is in clear contradiction to the derived mass of , e.g. 4u1820 - 30 .
it is generally believed that neutron stars are born with a mass about @xmath90 .
neutron stars in lmxbs would therefore accrete @xmath91 during their lifetime .
it seems reasonable to assume that some of the neutron stars in lmxbs accrete enough matter to reach the maximally stable mass ( @xmath92 for uv@xmath84+uvii , @xmath93 for rhf@xmath85 ) .
the neutron star would then collapse to a black hole .
the interpretation of the qpo with frequencies
@xmath69 about 10 hz as lense - thirring precession frequency of the accretion disk seems not to be consistent with the theoretical star models , unless one assumes that the first overtone of the precession is observed . in the case of z - sources however , the necessary ratio of the frame dragging frequency and spin frequency , and thus the moment of inertia , of the models if too small compared to the observed frequency @xmath69 or half of it .
our general conclusion is that the observations of kilohertz qpos in lmxbs provide us another powerful tool for probing the interior of neutron stars .
compared to the other tools like observations of the maximum mass @xcite , the limiting spin period @xcite , and cooling simulations @xcite , the derived constraints , especially in the sonic - point - interpretation , are rather strong .
the lower limit @xmath94 is only consistent with two eoss , uv@xmath84+uvii and rhf@xmath85 , which are relatively stiff at high densities .
their stiff behaviour at high densities seems to be only possible if the neutron star matter consists of neutrons , protons , and leptons only . at the most
, a small admixture of hyperons may be allowed .
however , one has to admit that such a composition somehow contradicts our conception of neutron star matter , since fieldtheoretical models lead more or less inevitably to a more complex composition at high density .
if the given interpretation is correct , one has therefore to reinvestigate the problem of super dense neutron star matter by dropping , for instance , the standard assumptions about the coupling of the hyperons in such regimes .
this specific conclusion depends however on the interpretation of the maximum frequency of the kilohertz qpo within the sonic - point - model . as a kind of beat - frequency model , this model predicts a constant frequency separation @xmath95 .
recently two further examples , where this constancy is not observed , has been discovered : 4u1608 - 52 and 4u1735 - 44 @xcite .
moreover , the frequency separation in the atoll source 4u1636 - 536 seems not to be consistent with the half of the frequency of the qpo in type i bursts @xcite .
though the beat - frequency models are the most hopeful candidates in explaining the qpo - phenomenology it has to be awaited how the variation of the frequency separation and deviation from the qpo frequency in bursts can be incorporated to these models .
one of us ( ch . s. ) gratefully acknowledges the bavarian state for financial support . we would like to thank norman glendenning and jrgen schaffner - bielich for providing us tables of their eoss .
mndez m. , van der klis m. , van paradijs j. , lewin w.h.g .
, vaughen b.a . ,
kuulkers e. , zhang w. , lamb f.k .
, psaltis d. , 1998 , in _ accretion processes in astrophysical systems : some like it hot _ , aip conference proceedings 431 weber f. , glendenning n.k . , 1993 , in feng d.h . , he g.z
_ proceedings of the nankai summer school , `` astrophysics and neutrino physics '' , tanjin , china , june 17 - 27 , 1991 _ , world scientific , singapore , pp . 64183 weber f. , kettner c. , weigel m.k . , glendenning n.k . , 1995 , in vassiliadis g. , panagiotou a. , kumar s. , madsen j. , eds
. , _ proceedings of the international symposium on strangeness and quark matter _ , world scientific , pp .
308 weber f. , schaab c. , weigel m.k .
, glendenning n.k . , 1997 , in giovannelli f. , mannocci g. , eds . ,
_ vulcano workshop 1996 frontier objects in astrophysics and particle physics , may 27 june 1 , vulcano _ , italian physical society , bologna ( italia ) , p. 87 weber f. , schaab c. , weigel m.k . , glendenning n.k . , 1997 , in stcker h. , gallmann a. , hamilton j.h
. , eds . , _ procceedings of the international conference on nuclear physics at the turn of the millenium : structure of vacuum and elementary matter , march 10 - 16 , 1996 , wilderness , south africa _ , world scientific , singapore , p. 322
yu w. , zhang s.n .
, harmon b.a . , paciesas w.s . , robinson c.r . , grindlay j.e . ,
bloser p. , barret d. , ford e.c .
, tavani m. , kaaret p. , 1997
, in dermer c.d .
, strickman m.s .
, kurfess j.d .
, eds . , _ proceedings of the fourth compton symposium _ , aip conference proceedings 410 , p. 734
|
recently discovered quasi periodic oscillations in the x - ray brightness of low mass x - ray binaries are used to derive constraints on the mass of the neutron star component and the equation of state of neutron star matter .
the observations are compared with models of rapidly rotating neutron stars which are calculated by means of an exact numerical method in full relativity . for the equations of state we select a broad collection of models representing different assumptions about the many - body structure and the complexity of the composition of super dense matter .
the mass constraints differ from their values in the approximate treatment by @xmath0 % . under the assumption that the maximum frequency of the quasi periodic oscillations originates from the innermost stable orbit the mass of the neutron star is in the range : @xmath1 .
especially the quasi periodic oscillation in the atoll - source 4u1820 - 30 is only consistent with equations of state which are rather stiff at high densities which is explainable , so far , only with pure nucleonic / leptonic composition .
this interpretation contradicts the hypothesis that the protoneutron star formed in sn 1987a collapsed to a black hole , since this would demand a maximum neutron star mass below @xmath2 .
the recently suggested identification of quasi periodic oscillations with frequencies around 10 hz with the lense - thirring precession of the accretion disk is found to be inconsistent with the models studied in this work , unless it is assumed that the first overtone of the precession is observed .
stars : neutron equation of state gravitation stars : rotation accretion , accretion discs x - ray : stars
|
super kamiokande high statistics data @xcite indicate that the observed deficit in the @xmath7-like atmospheric events is due to the neutrinos arriving at the detector at large zenith angles , strongly suggestive of the @xmath1 oscillation hypothesis .
similarly , their data on the zenith angle dependence and recoil energy spectrum of solar neutrinos @xcite in combination with the results from homestake @xcite , sage @xcite , and gallex+gno @xcite experiments , have put on a firm observational basis the long standing problem of solar neutrinos , strongly indicating the need for @xmath0 conversions .
in addition to the solar and atmospheric neutrino results from underground experiments , there is also the indication for neutrino oscillations in the @xmath8 channel by the lsnd experiment @xcite .
all these results can be accommodated in a single oscillation framework only if there are at least three different scales of neutrino mass - squared differences .
the simplest case in which this condition is satisfied requires the existence of a fourth light neutrino , which must be _ sterile _ ( _ i.e. _ having interactions with standard model particles much weaker than the sm weak interaction ) in order not to affect the invisible @xmath9 decay width , precisely measured at lep @xcite .
there are various analyses in the literature comparing the oscillation channels into active or sterile neutrinos for both solar @xcite and atmospheric data @xcite performed in the framework of two neutrino oscillations ( where oscillations occur into either an active or a sterile state ) .
in particular the super kamiokande collaboration has claimed the oscillation into sterile neutrinos to be disfavoured both for solar @xcite and atmospheric neutrinos @xcite .
however , when considered in the framework of four neutrino mixing , oscillations into pure active or pure sterile states are only limiting cases of the most general possibility of oscillations into an admixture of active and sterile neutrinos @xcite .
such a more general framework of four neutrino mixing has also been analysed in recent studies , which have put constraints on the active sterile admixture either in the context of solar @xcite or atmospheric @xcite neutrino oscillations , respectively .
in this paper we present a _ combined _ analysis of the neutrino oscillation solutions of both the solar and the atmospheric neutrino problem , in the framework of four neutrino mixing .
we include in our analysis the most recent solar neutrino rates of homestake @xcite , sage @xcite , gallex and gno @xcite , as well as the recent 1258 day super kamiokande data sample @xcite , including the recoil electron energy spectrum for both day and night periods .
as for atmospheric neutrinos we include in our analysis all the contained event from the latest 79.5 kton - yr super
kamiokande data set @xcite , as well as the upward - going neutrino - induced muon fluxes from both super kamiokande and the macro detector @xcite .
the constraints arising from relevant reactor ( mainly from bugey @xcite ) and accelerator ( cdhsw @xcite and ccfr @xcite ) experiments are also imposed .
the outline of the paper is the following . in sec .
[ formalism ] we summarize the main expressions for the neutrino oscillation formulas that we use in the analysis of solar and atmospheric neutrino data which take into account matter effects both in the propagation in the sun and in the earth .
sections [ solar ] and [ atmos ] contain the results for the analysis of the four neutrino oscillation parameters for solar and atmospheric data respectively .
the results for the full combined analysis are described in sec . [
combined ] , with particular emphasis on the active sterile oscillation admixture which better allows for a simultaneous description of the solar and atmospheric neutrino data .
our results show that , against what may be naively expected , although the super kamiokande data disfavours both the pure @xmath4 atmospheric channel and the pure @xmath6 solar channel , the result from the combined analysis still favours any of these close to pure active and sterile oscillations and disfavours oscillations into a near maximal active sterile admixture .
there are six possible four neutrino schemes that can accommodate the results from solar and atmospheric neutrino experiments as well as the lsnd evidence .
they can be divided in two classes : @xmath10 and @xmath11 . in the @xmath10 schemes
there is a group of three neutrino masses separated from an isolated one by a gap of the order of 1 ev@xmath12 which is responsible for the short - baseline oscillations observed in the lsnd experiment . in @xmath11 schemes
there are two pairs of close masses separated by the lsnd gap .
@xmath10 schemes are disfavoured by experimental data with respect to the @xmath11 schemes @xcite , but they are still marginally allowed @xcite .
therefore we concentrate in the @xmath11 schemes shown in fig .
[ fourab ] . for the phenomenology of neutrino oscillations
these two schemes are equivalent up to the relabeling of the mass eigenstates ( or equivalently of the mixing angles ) .
thus in what follows we will consider the scheme b where the mass spectrum presents the hierarchy : @xmath13 ( we use the common notation @xmath14 ) . in this four - neutrino mixing scheme ,
the flavour neutrino fields @xmath15 ( @xmath16 ) are related to the mass eigenstates fields @xmath17 by @xmath18 where @xmath19 is a @xmath20 unitary mixing matrix . neglecting possible cp phases ,
the matrix @xmath19 can be written as a product of six rotations , @xmath21 , @xmath22 , @xmath23 , @xmath24 , @xmath25 and @xmath26 where @xmath27 where @xmath28 and @xmath29 .
the order of the product of the matrices corresponds to a specific parametrization of the mixing matrix @xmath19 . in order to study oscillations of the solar and atmospheric neutrinos which include the matter effects in the sun and/or in the earth
, it is convenient to use the following parametrization @xcite : @xmath30 this general form can be further simplified by taking into account the negative results from the reactor experiments , in particular the bugey one @xcite which implies that @xmath31 in the range of @xmath32 relevant for the lsnd experiment .
this leads to an upper bound on the projection of the @xmath0 over the 34 states : @xmath33 so that the two angles @xmath34 and @xmath35 must be small and their contribution to solar and atmospheric neutrino transitions is negligible .
therefore , in what follows we will set these two angles to zero , although one must keep in mind that at least one of them must be non - vanishing in order to account for the lsnd observation . in the approximation
@xmath36 the @xmath19 matrix takes the form : @xmath37 in the rest of this section we will discuss the consequences of the mixing structure of eq .
( [ umatrix ] ) on the relevant transition probabilities for solar and atmospheric neutrino oscillations .
the transition probabilities that take into account matter effects have been derived in ref .
some improvement concerning the calculation of the regeneration of solar @xmath0 s in the earth was presented in ref .
@xcite . here
we summarize the discussion on the flavour composition of the relevant states for both solar and atmospheric neutrino oscillations . in the scheme here considered ,
solar neutrino oscillations are generated by the mass - squared difference between @xmath38 and @xmath39 .
it is clear from eq .
( [ umatrix ] ) that the survival of solar @xmath0 s mainly depends on the mixing angle @xmath40 , whereas the mixing angles @xmath2 and @xmath3 determine the relative amount of transitions into sterile @xmath41 or active @xmath1 and @xmath42 .
let us remind the reader that @xmath1 and @xmath42 can not be distinguished in solar neutrino experiments , since their matter potential and their interaction in the detectors are due only to neutral - current weak interactions and therefore they are equal .
thus , instead of @xmath1 and @xmath42 , one can consider the linear combinations @xmath43 with @xmath44 the mixing of @xmath45 and @xmath46 with @xmath39 and @xmath38 is given by @xmath47 so that the solar neutrino oscillations occur only between the states @xmath48 with mixing angle @xmath40 , and @xmath49 gives the size of the projection of the sterile neutrino onto the state in which the solar @xmath0 oscillates to .
therefore , solar neutrino oscillations depend only on @xmath40 and the product @xmath50 .
we distinguish the following limiting cases of eq .
( [ nusol ] ) : * if @xmath51 then @xmath52 , @xmath53 , @xmath54 , corresponding to the limit of pure two - generation @xmath55 transitions ; * if @xmath56 then @xmath57 , @xmath58 and @xmath59 and we have the limit of pure two - generation @xmath6 transitions .
since the mixing of @xmath0 with @xmath39 and @xmath38 is described by a two - generation equation with mixing angle @xmath40 , the mixing of @xmath41 with @xmath39 and @xmath38 is given by the same formula multiplied by @xmath50 , and the mixing of @xmath45 with @xmath39 and @xmath38 is again described by the same equation times @xmath60 , it is clear that , in the general case of simultaneous @xmath6 and @xmath55 oscillations , the corresponding probabilities are given by @xmath61 where @xmath62 takes the standard two neutrino oscillation form @xcite for @xmath63 and @xmath40 , found by numerically solving the evolution equation in the sun and the earth matter with the modified matter potential @xmath64 with @xmath65 and @xmath66 . here
@xmath67 and @xmath68 are , respectively , the number densities of electrons and neutrons in the medium , @xmath69 is the neutrino energy and @xmath70 is the fermi constant .
these expressions satisfy the relation of probability conservation @xmath71 .
the analysis of the solar neutrino data in the four neutrino mixing schemes is therefore equivalent to the two neutrino analysis but taking into account that the parameter space is now three
dimensional @xmath72 .
although originally this derivation was performed in the framework of the @xmath11 schemes @xcite , it is equally valid for the @xmath10 ones @xcite
. let us finally comment on the range of variation of the mixing parameters . choosing the convention that @xmath73 , from eqs .
( [ pes ] ) and ( [ pea ] ) it is clear that for solar neutrino oscillations the full physical space is covered by choosing the mixing angles in the ranges : @xmath74 actually , since the dependence on @xmath2 and @xmath3 enters only through the combination @xmath75 , for what concerns the solar neutrino phenomenology one of these mixing angles could be fixed to zero .
however , as we will see in the next section , they enter independently in the relevant probabilities for atmospheric neutrinos .
atmospheric neutrino oscillations are generated by the mass - squared difference between @xmath76 and @xmath77 . from eq .
( [ umatrix ] ) one sees that the atmospheric @xmath0 s are not affected by the four neutrino oscillations in the approximation @xmath78 and neglecting the effect of @xmath63 in the range of atmospheric neutrino energies .
conversely the survival probability of atmospheric @xmath1 s mainly depends on the mixing angle @xmath79 .
the mixing angles @xmath2 and @xmath3 determine the relative amount of @xmath1 transitions into sterile @xmath41 or active @xmath42 . more explicitly eq .
( [ umatrix ] ) implies that the atmospheric neutrino oscillations , _
i.e. _ oscillations with the mass difference @xmath80 and mixing angle @xmath79 , occur between the states @xmath81 we distinguish the following limiting cases of eq .
( [ nuatm ] ) : * if @xmath82 then @xmath83 .
the atmospheric @xmath84 state oscillates into a state @xmath85 .
this is the limit studied in ref .
we will denote this case as `` restricted '' .
in particular : * * in the case @xmath86 we have @xmath87 , corresponding to the limit of pure two generation @xmath5 transitions ; * * in the case @xmath88 and @xmath89 we have @xmath90 , corresponding to the limit of pure two generation @xmath4 transitions ; * if @xmath91 there are not atmospheric neutrino oscillations as the projection of @xmath1 over the relevant states cancels out ( @xmath92 ) . thus the mixing angle
@xmath2 determines the size of the projection of the @xmath1 over the `` atmospheric '' neutrino oscillating states , @xmath93 as a consequence , one expects @xmath94 to be small in order to explain the atmospheric neutrino deficit and , as we will see in sec . [ atmos ] , this is the case .
furthermore , the negative results from the cdhs @xcite and ccfr @xcite experiments on searches for @xmath1disappearance also constrain such projection to be @xmath95 at 90% cl for @xmath96 ev@xmath12 .
approximate expressions for the relevant @xmath1 survival probability including matter effects in the earth can be found in ref .
@xcite . for what concerns the work presented here
, the relevant probabilities can be calculated by numerically integrating the evolution equation in the earth with the modified matter potential @xmath97 so that for pure atmospheric @xmath98 oscillations ( @xmath99 ) @xmath100 , while for @xmath101 ( @xmath102 , @xmath103 ) @xmath104 , as expected .
it is this modification of the earth matter potential what gives to the atmospheric neutrino data the capability to discriminate between the active and sterile oscillation solution . in particular ,
higher sensitivity to this potential effect is achievable for the higher energy part of the atmospheric neutrino flux which lead to the upward going muon data .
the main effect of the presence of this potential is that for pure active oscillations the angular distribution of upgoing muons is expected to be steeper at larger arrival angles , while a flattening is expected due to the matter effects for sterile oscillations @xcite . from eqs .
( [ nuatm ] ) and ( [ aatm ] ) we see that , unlike in the case of solar neutrinos , the angles @xmath2 and @xmath3 enter independently in the atmospheric oscillations .
thus the analysis of the atmospheric neutrino data in the four neutrino mixing schemes is equivalent to the two neutrino analysis but taking into account that the parameter space is now four
dimensional @xmath105 . concerning the range of variation of the mixing angles , the full parameter space relevant for atmospheric neutrino oscillation can be covered by choosing @xmath2 and @xmath3 in the ranges indicated in eq .
( [ rangessol ] ) , but , in general , the mixing angle @xmath79 must be allowed to take values in the interval @xmath106 ( in the convention @xmath107 ) . in the limiting case @xmath108 , there is an additional symmetry in the relevant probabilities so that the full parameter space can be spanned by @xmath109 , as expected since in this case we have the effective two neutrino oscillation @xmath110 .
we first describe the results of the analysis of the solar neutrino data in terms of @xmath0 oscillations .
we determine the allowed range of oscillation parameters using the total event rates of the chlorine @xcite , gallium @xcite and super
kamiokande @xcite ( corresponding to the 1258 days data sample ) experiments . for the gallium experiments
we have used the weighted average of the results from gallex+gno and sage detectors .
we have also included the super kamiokande electron recoil energy spectrum measured separately during the day and night periods .
this will be referred in the following as the day
night spectra data which contains @xmath111 data bins .
the analysis includes the latest standard solar model fluxes , bp00 model @xcite , with updated distributions for neutrino production points and solar matter density . for details on the statistical analysis applied to the different observables we refer to ref .
as discussed in the previous section , the analysis of the solar neutrino data in these four neutrino mixing schemes is equivalent to the two neutrino analysis but taking into account that the parameter space is now three
dimensional @xmath112 , so we have a total of 37 degrees of freedom ( d.o.f . ) .
we first present the results of the allowed regions in the three
parameter space for the global combination of solar observables .
notice that since the parameter space is three dimensional the allowed regions for a given cl are defined as the set of points satisfying the condition @xmath113 where , for instance , @xmath114 , @xmath115 , and @xmath116 for cl = 90% , 95% and 99% respectively , and @xmath117 is the global minimum in the three dimensional space . in fig . [ solfour ]
we plot the sections of such a volume in the plane ( @xmath118 ) for different values of the active sterile admixture @xmath119 .
the global minimum used in the construction of the regions lies in the lma region and for pure @xmath0active oscillations @xmath120 which for 37 d.o.f .
( 3 rates @xmath121 38 spectrum points @xmath122 1 spectrum free normalization @xmath122 3 parameters ) corresponds to a goodness of the fit ( gof ) of 46% . as seen in fig .
[ solfour ] the sma region is always a valid solution for any value of @xmath123 at 99% cl .
this is expected as in the two neutrino oscillation picture this solution holds both for pure @xmath0active and pure @xmath0sterile oscillations .
notice , however , that the statistical analysis is different : in the two neutrino picture the pure @xmath0active and @xmath0sterile cases are analysed separately , whereas in the four neutrino picture they are taken into account simultaneously in a consistent scheme . on the other hand , both the lma and low solutions disappear for larger values of the active sterile admixture .
let us comment here that , unlike for the case of atmospheric neutrinos where the discrimination between active and sterile oscillations arises from the difference in the matter potentials , for solar neutrinos the main source of difference is due to the lack of neutral current contribution to the water cerenkov experiments for the sterile case . unlike active neutrinos which lead to events in the super kamiokande detector by interacting via nc with the electrons , sterile neutrinos
do not contribute to the super kamiokande event rates .
therefore a larger survival probability for @xmath124b neutrinos is needed to accommodate the measured rate . as a consequence a larger contribution from @xmath124b neutrinos to the chlorine and gallium experiments is expected , so that the small measured rate in chlorine can only be accommodated if no @xmath125be neutrinos are present in the flux .
this is only possible in the sma solution region , since in the lma and low regions the suppression of @xmath125be neutrinos is not enough .
notice also that the sma region for oscillations into sterile neutrinos is slightly shifted downwards as compared to the active case .
this is due to the small modification of the neutrino survival probability induced by the different matter potentials .
the matter potential for sterile neutrinos is smaller than for active neutrinos due to the negative nc contribution proportional to the neutron abundance .
for this reason the resonant condition for sterile neutrinos is achieved at lower @xmath126 . in fig .
[ chi2sol ] we plot the difference @xmath127 between the local minimum of @xmath128 for each solution and the global minimum in the three dimensional space , as a function of the active sterile admixture @xmath129 . from the figure we find that : * solar neutrino data favours pure @xmath130 oscillations but sizable active
sterile admixtures are allowed ; * the dependence of @xmath127 on the active - sterile admixture is very gentle in the sma region while it is much more pronounced for the lma solution ; * the three dimensional regions disappear at 90% ( 99% ) cl for the following value of @xmath123 ?
= ? @xmath131 * at 95% cl the three dimensional lma region is allowed for maximal active sterile mixing @xmath132 while at 99% cl all solutions are possible for maximal admixture .
there is a subtlety on the meaning of these limiting values of the mixing @xmath123 .
the values quoted in eq .
( [ limsolar ] ) correspond to the mixings at which the corresponding three
dimensional region disappears at a given cl . strictly speaking
this is not the same as the allowed range of @xmath123 at a given cl when the parameters @xmath133 are left free to vary , either in the full parameter space or within a given solution .
such limits would be obtained by imposing the condition @xmath134 where @xmath135 is minimized in the parameter space 12 ( or in a given region of such parameter space ) and @xmath117 is minimized in the same region of the space 12 and in the mixing @xmath123 . in what follows we label as `` constrained '' the results of analyses when the parameters @xmath63 and @xmath40 are varied within a given solution while `` unconstrained '' refers to the case where we allow the variation of @xmath63 and @xmath40 in the full plane . for the unconstrained fit the corresponding curve in fig .
[ chi2sol ] is simply the lower envolvent of the curves from the constrained analysis , and it follows the lma curve for @xmath136 and the sma one for @xmath137 .
in our statistical analysis of the atmospheric neutrino events we use all the samples of super kamiokande data : @xmath138-like and @xmath7-like data samples of sub- and multi - gev @xcite , each given as a 5-bin zenith - angle distribution , and upgoing muon data including the stopping ( 5 bins in zenith - angle ) and through going ( 10 angular bins ) muon fluxes . we have also included the latest macro @xcite upgoing muon samples , with 10 angular bins which is also sensitive to the active sterile admixture .
so we have a total 45 independent inputs . for details on the statistical analysis applied to the different observables we refer to ref .
@xcite . as discussed in the previous section the analysis of the atmospheric neutrino data in these four
neutrino mixing schemes is equivalent to the two neutrino analysis but taking into account that the parameter space is now four
dimensional @xmath139 .
alternatively we present the results in the equivalent parameter space @xmath140 where @xmath141 , @xmath142 , and @xmath143 for cl = 90% , 95% and 99% respectively , and @xmath144 is the global minimum in the four dimensional space . in fig .
[ atmfour ] we plot the sections of such a volume in the plane ( @xmath145 ) for different values of the mixings @xmath2 and @xmath3 , which we parametrize by values of the projections @xmath146 and @xmath147 . as discussed in sec .
[ formalism ] for arbitrary values of @xmath148 and @xmath149 the full parameter space is covered for @xmath79 in the range given in eq .
( [ rangesatm ] ) . in fig .
[ atmfour ] we display this full parameter space by showing the region in ( @xmath150 ) with `` positive '' @xmath151 for @xmath152 and `` negative '' @xmath151 for @xmath153 . as seen from the figure the parameter space is composed of two separated regions , each of them around maximal mixing @xmath154 .
the global minimum used in the construction of the regions lies almost in the pure atmospheric @xmath1@xmath42 oscillations . at the best fit point : @xmath155 which for @xmath156 d.o.f
. corresponds to a gof of 92% .
a careful reader may find this gof surprisingly high .
let us comment here that there has been a reduction of the @xmath157 with respect to previous analysis @xcite due to a better agreement of the 79.5 sk electron distribution with the non - oscillation expectation ( same effect is found in ref .
since the scenarios here discussed do not affect the atmospheric @xmath0 predictions , this implies an overall improvement on the gof of any these four neutrino oscillation schemes . from fig .
[ atmfour ] we see that the region becomes considerably smaller for increasing values of the mixing angle @xmath2 , which determines the size of the projection of the @xmath1 over the `` atmospheric '' neutrino oscillating states , and for increasing values of the mixing angle @xmath3 , which determines the active sterile admixture in which the `` almost@xmath1 '' oscillates to . therefore from the analysis of the atmospheric neutrino data we obtain an upper bound on both mixings which , in particular , implies a lower bound on the combination @xmath158.a the allowed region in the parameter space @xmath159 . following the discussion below eq .
( [ limsolar ] ) we obtain the allowed two dimensional region by imposing the condition @xmath160 where @xmath161 is minimized with respect to @xmath162 and @xmath79 and @xmath163 is the global minimum in the four dimensional parameter space .
the diagonal cut in fig .
[ chi2atm].a gives the unitarity bound @xmath164 where we have used that in eq .
( [ umatrix ] ) @xmath165 . as seen from fig .
[ chi2atm].a , the analysis of the atmospheric data imposes a severe bound on the @xmath1 projection .
for instance , the two dimensional region extends only till @xmath166 at 90% ( 99% ) cl , which , as previously mentioned ( see eq .
( [ cdhs ] ) ) , is perfectly consistent with the bounds from the @xmath1disappearance accelerator experiments cdhdw @xcite and ccfr @xcite given in eq .
( [ cdhs ] ) . in order to determine the impact of the small but possible deviation from zero of @xmath167 we also study the `` restricted '' case of ref.@xcite in which we impose @xmath168 in fig .
[ atmfourr ] we plot the sections of this volume in the plane @xmath169 for different values of @xmath149 . as discussed in sec .
[ formalism ] for the `` restricted '' case the full parameter space is covered with @xmath170 .
this is graphically displayed in fig .
[ atmfourr ] , where the symmetry @xmath171 is clearly evident .
the constrains on the active sterile admixture in the two cases are displayed in fig .
[ chi2atm].b , where we plot the dependence of @xmath172 on the active sterile admixture @xmath173 , both in the general case ( _ i.e. _ when @xmath174 is minimized with respect to all other parameters in the problem ) , and in the `` restricted '' case ( for which the @xmath1 projection has been fixed @xmath175 ) .
the horizontal lines are the values of @xmath176 for cl = 90% and 99% . from this figure
we obtain the 90% ( 99% ) cl lower bounds on @xmath177 from the analysis of the atmospheric neutrino data : @xmath178 as seen in the figure @xmath179 has a monotonically growing behaviour as @xmath180 decreases for @xmath1810.6 , while the dependence flattens below that value with the presence of a shallow secondary minimum around @xmath1820.3 .
we have traced the presence of this secondary minimum to the combined effect of the behaviour of @xmath157 for the different sets of upgoing muon data . for all sets of upgoing muon data
the best fit is very close to the global minimum in eq .
( [ minatm ] ) .
macro data at the moment leads to a monotonically growing @xmath157 as the sterile admixture grows while super
kamiokande upgoing data on both stopping and through - going muons presents a flattening of their @xmath157 for @xmath1830.6 .
this lost of sensitivity of super kamiokande at larger sterile admixtures arises from the fact that their latest upgoing muon data presents a flattening of the muon suppression in the angular bins for arrival directions around @xmath184 which is compatible with large sterile admixtures .
the same effect has been found in ref .
when combining the @xmath157 for macro and super kamiokande data samples and taking into account the strong correlation between their respective theoretical errors , we find the appearance of this secondary shallow minimum . from the figure we see that oscillations into a pure sterile state are excluded at 99.97% cl as compared to pure active oscillations .
one must notice , however , that taken as an independent scenario , @xmath185 oscillations give a @xmath186 ( for @xmath187 d.o.f . ) which implies a gof of 35% and therefore can not be ruled out in the bases of its absolute quality . in summary , we see that the analysis of the atmospheric data implies that : * one of the two neutrino states in the 34 oscillating pair must be a close to pure @xmath1 ; * this close to
pure @xmath1 oscillates into a state which is preferably composed by @xmath42 , although a 48% admixture of sterile in this state is still allowed at 99% cl ; * when imposing the `` restricted '' condition that one of the states is a pure @xmath1 , the bound on the sterile admixture is tightened approximately by a factor 2 ( from 48% to 26% at 99% cl ) .
from the previous sections we have learned that the analysis of the solar data favours the scenario in which the solar oscillations in the plane 12 are @xmath0 oscillations into an active neutrino , and from that analysis we found an upper limit on the projection of the @xmath41 on the 12 states . on the other hand
, the atmospheric neutrino analysis prefers that the oscillations of the 34 states occur between a close to
pure @xmath1 and an active ( @xmath42 ) neutrino , so giving an upper bound on the projection of the @xmath41 over the 34 states or equivalently a lower bound on its projection over the 12 states .
the open question is then what is the best scenario for the active sterile admixture once these two bounds are put together . to address this question in full generality
we have studied the behaviour of the global @xmath188 function defined as @xmath189 with the mixing angles @xmath2 and @xmath3 determining the active sterile admixture for the following scenarios : * atm+sol@xmath190 : in this case we minimize @xmath188 with respect to @xmath63 , @xmath40 , @xmath162 and @xmath79 , allowing @xmath191 and @xmath40 to change from lma to sma as the active sterile admixture controlled by @xmath2 and @xmath3 change .
the @xmath1 projection @xmath192 is free to vary in the range allowed by the analysis ; * atm@xmath193+sol@xmath190 : in this case we fix @xmath194 , so @xmath195 .
we minimize @xmath188 with respect to @xmath63 , @xmath40 , @xmath162 and @xmath79 , allowing @xmath63 and @xmath40 to change from lma to sma as the mixing angle @xmath3 changes ; * atm+sol@xmath196 : in this case we minimize @xmath188 with respect to @xmath63 , @xmath40 , @xmath162 and @xmath79 with @xmath191 and @xmath40 constrained to be in the lma ( sma ) region .
the @xmath1 projection @xmath197 is free to vary in the range allowed by the analysis ; * atm@xmath193+sol@xmath198 : in this case we fix @xmath194 , so @xmath195 .
we minimize @xmath188 with respect to @xmath63 , @xmath40 , @xmath162 and @xmath79 , with @xmath63 and @xmath40 constrained to be in the lma ( sma ) region . in table
[ table_comb ] we give the values @xmath199 of the best fit point and its associated gof in the six dimensional ( five dimensional for the `` restricted '' case ) space in each of these cases together with the best fit value of the mixings @xmath200 and @xmath201 . in fig .
[ chi2_combined ] we plot the shift in @xmath157 from the corresponding global minimum in each of the scenarios as the mixing @xmath202 varies . from the table and the figure
we find the following behaviours : * there are two favourite configurations which we will denote as near pure sterile solar neutrino oscillations plus near pure active atmospheric neutrino oscillations ( npss+npaa ) , and close to active solar neutrino oscillations plus close to sterile atmospheric neutrino oscillations ( cas+csa ) , respectively .
npss+npaa oscillations are characterized by the minima @xmath203 where the exact position of the minimum depends on the allowed admixture @xmath204 .
cas+csa oscillations are characterized by the minima @xmath205 in both the cases the minima are not very deep in @xmath157 ( this is particularly the case for cas+csa ) ; * in none of the cases maximal active
sterile ( mas ) admixture @xmath206 is favoured . the scenarios with the solar solution either within the sma region ( fig .
[ chi2_combined].c ) or left unconstrained ( fig . [
chi2_combined].a ) prefer the npss+npaa scenario .
they give a better fit to the combined analysis with a gof of 66%68% . in these cases
the dependence of @xmath188 with the active sterile admixture is more pronounced and it is dominated by the dependence of the atmospheric neutrino analysis , although there is a secondary minimum at the cas+csa configuration which for the atm+sol@xmath190 case is acceptable at 92% cl . in all these scenarios
mas admixture is ruled out at more than 99% cl .
qualitatively this behaviour arises from the fact that in these scenarios increasing the sterile admixture in the atmospheric pair is strongly disfavoured while it is still acceptable for the solar pair since the sma solution is valid for pure @xmath207 solar oscillations . on the other hand , the scenarios with the solar solution within the lma region ( fig . [
chi2_combined].b ) prefer the cas+csa scenario , although the dependence of @xmath188 with the active sterile admixture is not very strong and it presents a secondary minimum near the complementary situation of close to sterile solar plus close to active atmospheric ( css+caa ) scenario , which is acceptable at 90% cl .
they give the worst fit to the combined analysis but they still present an acceptable gof : 47.5% ( 56% ) for the `` restricted '' ( unrestricted ) case . in these scenarios
mas admixture is ruled out at 99% cl for the `` restricted '' case and it is acceptable at about 95% cl for the unrestricted one .
this behaviour arises from the fact that in these scenarios increasing the sterile admixture both in the atmospheric pair and in the solar case is strongly disfavoured .
it is precisely in this case that one would expect the combined solution to lie near mas admixture .
however , the results show that it is the solar dependence ( which dominates in the combination ) what leads to the slightly better description for the cas+csa scenario with a secondary minimum at the complementary css+caa scenario . summarizing , in this paper we have performed a combined analysis of the neutrino oscillation solutions of the solar and atmospheric neutrino problem in the framework of four neutrino mixing .
such scenarios allow for simultaneous transitions of solar @xmath0 as well as atmospheric @xmath1 into active and sterile neutrinos .
our results show that although the solar and atmospheric data analyzed independently disfavour the pure @xmath4 atmospheric channel and the pure @xmath6 solar channel , the result from the combined analysis still favours close to pure active and sterile oscillations either in the solar or in the atmospheric sector ( depending on the selected solar solution ) and disfavours oscillations into a near maximal active sterile admixture .
99 y. fukuda _
et al . _ ,
* b433 * , 9 ( 1998 ) ; phys .
b436 * , 33 ( 1998 ) . y. fukuda _
b467 * , 185 ( 1999 ) ; phys .
lett . * 82 * , 2644 ( 1999 ) . super
kamiokande presentations in winter conferences : c. mcgrew in _ neutrino telescopes 2001 _ , 9th international workshop on neutrino telescopes ( venice , italy , march 2001 ) , to appear ; t. toshito in
_ moriond 2001 _ , xxxvi rencontres de moriond on electroweak interactions and unified theories ( les arcs , france , march 2001 ) , to appear .
kamiokande collaboration , y. fukuda _ et al .
. lett . * 81 * , 1158 ( 1998 ) ; erratum * 81 * , 4279 ( 1998 ) ; * 82 * , 1810 ( 1999 ) ; * 82 * , 2430 ( 1999 ) ; y. suzuki , nucl .
b ( proc . suppl . ) * 77 * , 35 ( 1999 )
. s. fukuda _ et al .
_ [ super kamiokande collaboration ] , hep - ex/0103032 .
s. fukuda _ et al .
_ [ super kamiokande collaboration ] , hep - ex/0103033 .
s. fukuda _ et al .
_ [ super kamiokande collaboration ] , phys . rev . lett . *
85 * , 3999 ( 2000 ) .
b. t. cleveland _
et al . _ ,
j. * 496 * , 505 ( 1998 ) ; r. davis , prog .
phys . * 32 * , 13 ( 1994 ) ; k. lande , talk at xix international conference on neutrino physics and astrophysics , sudbury , canada , june 2000 ( _ http://nu2000.sno.laurentian.ca_ ) ; sage collaboration , j. n. abdurashitov _
_ , phys . rev . *
c60 * , 055801 ( 1999 ) ; v. gavrin , talk at xix international conference on neutrino physics and astrophysics , sudbury , canada , june 2000 ( _ http://nu2000.sno.laurentian.ca_ ) .
gallex collaboration , w. hampel _ et al .
_ , phys . lett . *
b447 * , 127 ( 1999 ) .
e. belloti , talk at xix international conference on neutrino physics and astrophysics , sudbury , canada , june 2000 ( _ http://nu2000.sno.laurentian.ca_ ) . c. athanassopoulos , phys .
lett . * 75 * 2650 ( 1995 ) ; phys .
* 77 * 3082 ( 1996 ) ; phys .
* 81 * 1774 ( 1998 ) .
peltoniemi , d. tommasini and j.w.f .
valle , phys . lett.*b298 * , 383 ( 1993 ) ; e.j .
et al . _ ,
lett.*b357 * , 608 ( 1995 ) ; s.c .
et al . _ ,
b430 * , 296 ( 1998 ) ; b. brahmachari and r.n .
mohapatra , phys . lett . *
b437 * , 100 ( 1998 ) ; s. mohanty , d.p .
roy and u. sarkar , phys .
* b445 * , 185 ( 1998 ) ; j.t . peltoniemi and j.w.f .
valle , nucl .
b406 * , 409 ( 1993 ) ; q.y .
liu and a.yu .
smirnov , nucl .
b524 * , 505 ( 1998 ) ; d.o .
caldwell and r.n .
mohapatra , phys . rev . *
d48 * , 3259 ( 1993 ) ; e. ma and p. roy , phys . rev . *
d52 * , r4780 ( 1995 ) ; a.yu . smirnov and m. tanimoto ,
phys . rev .
* d55 * , 1665 ( 1997 ) ; n. gaur _ et al . _ ,
* d58 * , 071301 ( 1998 ) ; e.j .
chun , c.w .
kim and u.w .
lee , phys . rev . *
d58 * , 093003 ( 1998 ) ; k. benakli and a.yu .
smirnov , phys .
79 * , 4314 ( 1997 ) ; y. chikira , n. haba and y. mimura , eur .
j. c * 16 * , 701 ( 2000 ) ; c. liu and j. song , phys . rev . * d60 * , 036002 ( 1999 ) ; w. grimus , r. pfeiffer and t. schwetz , eur .
j. c * 13 * , 125 ( 2000 ) .
gomez - cadenas and m.c .
gonzalez - garcia , z. phys .
* c71 * , 443 ( 1996 ) ; n. okada and o. yasuda , int . j. mod . phys . * a12 * , 3669 ( 1997 ) ; s. goswami , phys . rev . * d55 * , 2931 ( 1997 ) ; s. m. bilenky , c. giunti , and w. grimus , phys
* d57 * , 1920 ( 1998 ) ; s. m. bilenky , c. giunti , and w. grimus , phys .
* d58 * , 033001 ( 1998 ) ; s.m .
bilenky , c. giunti , w. grimus and t. schwetz , astropart . phys . * 11 * , 413 ( 1999 ) ; v. barger , y.b .
dai , k. whisnant and b.l .
young , phys .
d * 59 * , 113010 ( 1999 ) ; v. barger , t.j .
weiler and k. whisnant , phys .
b * 427 * , 97 ( 1998 ) ; c. giunti , phys .
d * 61 * , 036002 ( 2000 ) ; c. giunti , phys .
b * 467 * , 83 ( 1999 ) ; s.m .
et al._,phys .
b * 465 * , 193 ( 1999 ) ; c. giunti , jhep * 0001 * , 032 ( 2000 ) ; a. ibarra and i. navarro , jhep * 0002 * , 031 ( 2000 ) .
bilenky , c. giunti and w. grimus , eur .
j. c * 1 * , 247 ( 1998 ) , proc . of _
neutrino 96 _ , helsinki , june 1996 , edited by k. enqvist _
et al . _ , p. 174
, world scientific , 1997 , hep - ph/9609343 ; s. m. bilenky , c. giunti , w. grimus , and t. schwetz , phys . rev .
* d60 * , 073007 ( 1999 ) .
v. barger , s. pakvasa , t. j. weiler , and k. whisnant , phys .
rev . * d58 * , 093016 ( 1998 ) .
d. dooling , c. giunti , k. kang and c.w .
kim , phys .
d * 61 * , 073011 ( 2000 ) .
n. fornengo , m. c. gonzalez - garcia and j. w. f. valle , nucl .
* b580 * ( 2000 ) 58 ; m.c .
gonzalez - garcia , h. nunokawa , o.l .
peres and j. w. f. valle , nucl .
b543 * , 3 ( 1999 ) and m. c. gonzalez - garcia .
r. foot , r.r .
volkas and o. yasuda , phys . rev .
* d58 * , 013006 ( 1998 ) ; o. yasuda , phys . rev . *
d58 * , 091301 ( 1998 ) ; e.kh .
akhmedov , a. dighe , p. lipari and a.yu .
smirnov , nucl .
b542 * , 3 ( 1999 ) .
p. lipari and m. lusignoli , phys .
d * 58 * , 73005 ( 1998 ) ; q. y. liu and a. yu .
smirnov , nucl .
b * 524 * , 505 ( 1998 ) ; q. y. liu , s. p. mikheyev , and a. yu .
smirnov , phys .
b * 440 * , 319 ( 1998 ) .
m. c. gonzalez - garcia , p. c. de holanda , c. pea - garay and j. w. f. valle , nucl .
* b573 * , 3 ( 2000 ) ; j.n .
bahcall , p.i .
krastev , a.yu .
smirnov , phys .
rev . * d58 * , 096016 ( 1998 ) .
m. c. gonzalez - garcia , c. pea - garay , nucl .
suppl . *
91 * , 80 ( 2000 ) . c. giunti , m. c. gonzalez - garcia , c. pea - garay , phys
. rev . * d62 * , 013005 ( 2000 ) ; m. c. gonzalez - garcia , c. pea - garay , phys .
d * 63 * , 073013 ( 2001 ) .
fogli , e. lisi , a. marrone , phys .
d * 63 * , 053008 ( 2001 ) ; hep - ph/0105139 .
fogli , e. lisi , a. marrone , hep - ph/0105139 .
o. yasuda , hep - ph/0006319 .
m. spurio [ macro collaboration ] , hep - ex/0101019 ; b. barish [ macro collaboration ] , talk at the xix international conference on neutrino physics and astrophysics , sudbury , canada , june 2000 ( _ http://nu2000.sno.laurentian.ca_ )
. b. achkar _ et al .
_ , nucl .
phys . * b424 * , 503 ( 1995 ) . f. dydak _ et al . _ , phys .
b * 134 * , 281 ( 1984 ) .
i. e. stockdale _ et al .
_ , phys .
lett . * 52 * , 1384 ( 1984 ) .
v. barger , b. kayser , j. learned , t. weiler , k. whisnant , phys .
* b489 * , 345 ( 2000 ) ; o. l. g. peres , a. yu .
smirnov , nucl .
b * 599 * , 3 ( 2001 ) ; w. grimus and t. schwetz , hep - ph/0102252 . c. giunti and m. laveder , jhep * 0102 * , 001 ( 2001 ) j. n. bahcall , s. basu and m. h. pinsonneault , astro - ph/0010346 .
see also , http://www.sns.ias.edu/@xmath208bahcall m. c. gonzalez - garcia , m. maltoni , c. pena - garay and j. w. valle , phys .
d * 63 * , 033005 ( 2001 ) .@xmath199 of the best fit point and its associated gof in the six dimensional ( five dimensional for the `` restricted '' case ) space in each of the cases discussed in the text together with the best fit value of the mixings @xmath209 and @xmath210 . [ cols="<,^,^,^,^",options="header " , ]
|
we present an analysis of the neutrino oscillation solutions of the solar and atmospheric neutrino problems in the framework of four neutrino mixing where a sterile neutrino is added to the three standard ones and the mass spectra presents two separated doublets .
such scenarios allow for simultaneous transitions of solar @xmath0 , as well as atmospheric @xmath1 , into active and sterile neutrinos controlled by the additional mixing angles @xmath2 and @xmath3 , and they contain as limiting cases the pure solar @xmath0active and @xmath0sterile neutrino oscillations , and the pure atmospheric @xmath4 and @xmath5 oscillations , respectively .
we evaluate the allowed active sterile admixture in both solar and atmospheric oscillations from the combined analysis .
our results show that , although the super kamiokande data disfavour both the pure @xmath4 atmospheric channel and the pure @xmath6 solar channel , the result from the combined analysis still favours close to pure active and sterile oscillations and disfavours oscillations into a near maximal active sterile admixture .
|
the problem of maximum likelihood ( ml ) estimation consists of finding a solution of the form @xmath0 where @xmath1 is an observed sample of a random variable @xmath2 defined on a sample space @xmath3 and @xmath4 is the log - likelihood function defined by @xmath5 defined on the parameter space @xmath6 , and @xmath7 denotes the density of @xmath2 at @xmath1 parametrized by the vector parameter @xmath8 .
the expectation maximization ( em ) algorithm is an iterative procedure which is widely used for solving ml estimation problems . the em algorithm was first proposed by dempster , laird and rubin and has seen the number of its potential applications increase substantially since its appearance .
the book of mclachlan and krishnan gives a comprehensive overview of the theoretical properties of the method and its applicability .
the convergence of the sequence of em iterates towards a maximizer of the likelihood function was claimed in the original paper but it was later noticed that the proof contained a flaw .
a careful convergence analysis was finally given by wu @xcite based on zangwill s general theory @xcite ; see also .
zangwill s theory applies to general iterative schemes and the main task when using it is to verify that the assumptions of zangwill s theorems are satisfied . since the appearance of wu s paper , convergence of the em algorithm
is often taken for granted in many cases where the necessary assumptions were sometimes not carefully justified .
as an example , an often neglected issue is the behavior of em iterates when they approach the boundary of the domain of definition of the functions involved .
a different example is the following .
it is natural to try and establish that em iterates actually converge to a single point @xmath9 , which involves proving uniqueness of the cluster point .
wu s approach , reported in is based on the assumption that the euclidean distance between two successive iterates tends to zero .
however such an assumption is in fact very hard to verify in most cases and should not be deduced solely from experimental observations .
the goal of the present paper is to propose an analysis of em iterates and their generalizations in the framework of kullback proximal point algorithms .
we focus on the geometric conditions that are provable in practice and the concrete difficulties concerning convergence towards boundaries and cluster point uniqueness .
the approach adopted here was first proposed in in which it was shown that the em algorithm could be recast as a proximal point algorithm . a proximal scheme for maximizing the function @xmath4 using the distance - like function @xmath10 is an iterative procedure of the form @xmath11 where @xmath12 is a sequence of positive real numbers often called relaxation parameters .
proximal point methods were introduced by martinet @xcite and rockafellar @xcite in the context of convex minimization .
the proximal point representation of the em algorithm is obtained by setting @xmath13 and @xmath14 to the kullback distance between some well specified conditional densities of a complete data vector .
the general case of @xmath15 was called the kullback proximal point algorithm ( kpp ) .
this approach was further developed in where convergence was studied in the twice differentiable case with the assumption that the limit point lies in the interior of the domain .
the main novelty of was to prove that relaxation of the kullback - type penalty could ensure superlinear convergence which was confirmed by experiment for a poisson linear inverse problem .
this paper is an extension of these previous works that addresses the problem of convergence under general conditions .
the main results of this paper are the following .
firstly , we prove that all the cluster points of the kullback proximal sequence which lie in the interior of the domain are stationary points of the likelihood function @xmath16 under very mild assumptions that are easily verified in practice . secondly ,
taking into account finer properties of @xmath10 , we prove that every cluster point on the boundary of the domain satisfies the karush - kuhn - tucker necessary conditions for optimality under nonnegativity constraints .
to illustrate our results , we apply the kullback - proximal algorithm to an estimation problem in animal carcinogenicity introduced in in which an interesting nonconvex constraint is handled . in this case
, the m - step can not be obtained in closed form .
however , the kullback - proximal algorithm can be analyzed and implemented .
numerical experiments are provided which demonstrate the ability of the method to significantly accelerate the convergence of standard em .
the paper is organized as follows . in section
[ kull ] , we review the kullback proximal point interpretation of em .
then , in section [ conv ] we study the properties of interior cluster points .
we prove that such cluster points are in fact global maximizers of a certain penalized likelihood function .
this allows us to justify using a relaxation parameter @xmath17 when @xmath17 is sufficiently small to permit avoiding saddle points .
section [ conv2 ] pursues the analysis in the case where the cluster point lies on a boundary of the domain of @xmath10 .
in this section , we review the em algorithm and the kullback proximal interpretation discussed in .
the em procedure is an iterative method which produces a sequence @xmath18 such that each @xmath19 maximizes a local approximation of the likelihood function in the neighborhood of @xmath20 .
this point of view will become clear in the proximal point framework of the next subsection . in the traditional approach
, one assumes that some data are hidden from the observer .
a frequent example of hidden data is the class to which each sample belongs in the case of mixtures estimation .
another example is when the observed data are projection of an unkown object as for image reconstruction problems in tomography .
one would prefer to consider the likelihood of the complete data instead of the ordinary likelihood .
since some parts of the data are hidden , the so called complete likelihood can not be computed and therefore must be approximated . for this purpose , we will need some appropriate notations and assumptions which we now describe .
the observed data are assumed to be i.i.d .
samples from a unique random vector @xmath2 taking values on a data space @xmath3 .
imagine that we have at our disposal more informative data than just samples from @xmath2 .
suppose that the more informative data are samples from a random variable @xmath21 taking values on a space @xmath22 with density @xmath23 also parametrized by @xmath8 .
we will say that the data @xmath21 is more informative than the actual data @xmath2 in the sense that @xmath2 is a compression of @xmath21 , i.e. there exists a non - invertible transformation @xmath24 such that @xmath25 .
if one had access to the data @xmath21 it would therefore be advantageous to replace the ml estimation problem ( [ ml ] ) by @xmath26 with @xmath27 . since @xmath28 the density @xmath29 of @xmath2 is related to the density @xmath30 of @xmath21 through @xmath31 for an appropriate measure @xmath32 on @xmath22 . in this setting ,
the data @xmath1 are called _ incomplete data _ whereas the data @xmath33 are called _
complete data_. of course the complete data @xmath33 corresponding to a given observed sample @xmath1 are unknown . therefore , the complete data likelihood function @xmath34 can only be estimated . given the observed data @xmath1 and a previous estimate of @xmath8 denoted @xmath35 , the following minimum mean square error estimator ( mmse ) of the quantity @xmath34 is natural @xmath36,\ ] ] where , for any integrable function @xmath37 on @xmath22 , we have defined the conditional expectation @xmath38=\int_{h^{-1}(\{y\ } ) } f(x ) k(x| y;\bar{\theta } ) d\mu(x)\ ] ] and @xmath39 is the conditional density function given @xmath1 @xmath40 having described the notions of complete data and complete likelihood and its local estimation we now turn to the em algorithm .
the idea is relatively simple : a legitimate way to proceed is to require that iterate @xmath19 be a maximizer of the local estimator of the complete likelihood conditionally on @xmath1 and @xmath20 .
hence , the em algorithm generates a sequence of approximations to the solution ( [ mlx ] ) starting from an initial guess @xmath41 of @xmath42 and is defined by @xmath43 \text{\hspace{1cm}\bf e step}\nonumber\ ] ] @xmath44 consider the general problem of maximizing a concave function @xmath45 .
the original proximal point algorithm introduced by martinet @xcite is an iterative procedure which can be written @xmath46 the quadratic penalty @xmath47 is relaxed using a sequence of positive parameters @xmath48 . in @xcite , rockafellar showed that superlinear convergence of this method is obtained when the sequence @xmath48 converges towards zero .
it was proved in that the em algorithm is a particular example in the class of proximal point algorithms using kullback leibler types of penalties .
one proceeds as follows .
assume that the family of conditional densities @xmath49 is regular in the sense of ibragimov and khasminskii , in particular @xmath50 and @xmath51 are mutually absolutely continuous for any @xmath8 and @xmath35 in @xmath52 .
then the radon - nikodym derivative @xmath53 exists for all @xmath54 and we can define the following kullback leibler divergence : @xmath55.\ ] ] we are now able to define the kullback - proximal algorithm . for this purpose , let us define @xmath56 as the domain of @xmath16 , @xmath57 the domain of @xmath58 and @xmath59 the domain of @xmath60 .
let @xmath12 be a sequence of positive real numbers .
then , the kullback - proximal algorithm is defined by @xmath61 the main result on which the present paper relies is that em algorithm is a special case of ( [ kullprox ] ) , i.e. it is a penalized ml estimator with proximal penalty @xmath14 .
[ equiem ] the em algorithm is a special instance of the kullback - proximal algorithm with @xmath62 , for all @xmath63 .
the previous definition of the kullback proximal algorithm may appear overly general to the reader familiar with the usual practical interpretation of the em algorithm .
however , we found that such a framework has at least the three following benefits : * to our opinion , the convergence proof of our em is more natural , * the kullback proximal framework may easily incorporate additional constraints , a feature that may be of crucial importance as demonstrated in the example of section [ prob2 ] below , * the relaxation sequence @xmath12 allows one to weight the penalization term and its convergence to zero implies quadratic convergence in certain examples .
the first of these three arguments is also supported by our simplified treatment of the componentwise em procedure proposed in and the remarkable recent results of @xcite based on a special proximal entropic representation of em for getting precise estimates on the convergence speed of em algorithms , however , with much more restrictive assumptions than the ones of the present paper .
although our results are obtained under mild assumptions concerning the relaxation sequence @xmath12 including the case @xmath64 , several precautions should be taken when implementing the method . however , one of the key features of em - like procedures is to allow easy handling of positivity or more complex constraints , such as the ones discussed in the example of section [ prob2 ] . in such cases
the function @xmath10 behaves like a barrier whose value increases to infinity as the iterates approach the boundary of the constraint set .
hence , the sequence @xmath12 ought to be positive in order to exploit this important computational feature . on the other hand ,
as proved under twice differentiability assumptions in when the cluster set reduces to a unique nondegenerate maximizer in the interior of the domain of the log - likelihood and @xmath65 converges to zero , quadratic convergence is obtained .
this nice behavior is not satisfied in the plain em case where @xmath13 for all @xmath63 . as a drawback ,
one problem in decreasing the @xmath65 s too quickly is possible numerical ill conditioning .
the problem of choosing the relaxation sequence is still largely open .
we have found however that for most `` reasonable '' sequences , our method was at least as fast as the standard em .
finally , we would like to end our presentation of kpp - em by noting that closed form iterations may not be available in the case @xmath66 .
if this is the case , solving ( [ kullprox ] ) becomes a subproblem which will require iterative algorithms . in some interesting examples ,
e.g. the case presented in section [ prob2 ] . in this case
, the standard em iterations are not available in closed form in the first place and kpp - em provides faster convergence while preserving monotonicity and constraint satisfaction .
the notation @xmath67 will be used to denote the norm on any previously defined space without more precision .
the space on which it is the norm should be obvious from the context .
for any bivariate function @xmath68 , @xmath69 will denote the gradient with respect to the first variable . in the remainder of this paper
we will make the following assumptions .
[ ass1 ] ( i ) @xmath16 is differentiable on @xmath56 and @xmath4 tends to @xmath70 whenever @xmath71 tends to @xmath72 .
+ ( ii ) the projection of @xmath59 onto the first coordinate is a subset of @xmath56 .
+ ( iii ) @xmath73 is a convergent nonnegative sequence of real numbers whose limit is denoted by @xmath74 .
we will also impose the following assumptions on the distance - like function @xmath10 .
[ ass2 ] ( i ) there exists a finite dimensional euclidean space @xmath75 , a differentiable mapping @xmath76 and a functional @xmath77 such that @xmath78 where @xmath79 denotes the domain of @xmath80 .
+ ( ii ) for any @xmath81 there exists @xmath82 such that @xmath83 .
moreover , we assume that @xmath84 for any bounded set @xmath85 . + for all @xmath86 in @xmath87 , we will also require that + ( iii ) ( positivity ) @xmath88 , + ( iv ) ( identifiability ) @xmath89 , + ( v ) ( continuity ) @xmath80 is continuous at @xmath90 + and for all @xmath91 belonging to the projection of @xmath87 onto its second coordinate , + ( vi ) ( differentiability ) the function @xmath92 is differentiable at @xmath91 .
assumptions [ ass1](i ) and ( ii ) on @xmath16 are standard and are easily checked in practical examples , e.g. they are satisfied for the poisson and additive mixture models . notice that the domain @xmath59 is now implicitly defined by the knowledge of @xmath56 and @xmath87 .
moreover @xmath10 is continuous on @xmath59 .
the importance of requiring that @xmath10 has the prescribed shape comes from the fact that @xmath10 might not satisfy assumption [ ass2](iv ) in general .
therefore assumption [ ass2 ] ( iv ) reflects the requirement that @xmath10 should at least satisfy the identifiability property up to a possibly injective transformation . in both examples discussed above
, this property is an easy consequence of the well known fact that @xmath93 implies @xmath94 for positive real numbers @xmath95 and @xmath96 .
the growth , continuity and differentiability properties [ ass2 ] ( ii ) , ( v ) and ( vi ) are , in any case , nonrestrictive . for the sake of notational convenience , the regularized objective function with relaxation parameter @xmath17 will be denoted @xmath97 finally we make the following general assumption . [ ass3 ] the kullback proximal iteration ( [ kullprox ] ) is well defined , i.e. there exists at least one maximizer of @xmath98 at each iteration @xmath99 . in the em case ,
i.e. @xmath100 , this last assumption is equivalent to the computability of m - steps .
a sufficient condition for this assumption to hold would be , for instance , that @xmath101 be sup - compact , i.e. the level sets @xmath102 be compact for all @xmath103 , @xmath104 and @xmath105 .
however , this assumption is not usually satisfied since the distance - like function is not defined on the boundary of its domain . in practice
it suffices to solve the equation @xmath106 , to prove that the solution is unique .
then assumption [ ass1](i ) is sufficient to conclude that we actually have a maximizer . using assumptions [ ass1 ] , we easily deduce monotonicity of the likelihood values and boundedness of the proximal sequence .
the first two lemmas are proved , for instance , in .
we start with the following monotonicity result .
[ truit ] for any iteration @xmath107 , the sequence @xmath18 satisfies
@xmath108 from the previous lemma , we easily obtain the boundedness of the sequence .
[ boundu ] the sequence @xmath18 is bounded .
the next lemma will also be useful .
[ yal ] assume that there exists a subsequence @xmath109 belonging to a compact set @xmath110 included in @xmath56 .
then , @xmath111 * proof*. since @xmath16 is continuous over @xmath110 , @xmath112 and @xmath113 is therefore bounded from above .
moreover , lemma [ truit ] implies that the sequence @xmath114 is monotone nondecreasing .
therefore , the whole sequence @xmath114 is bounded from above and convergent .
this implies that @xmath115 . applying lemma [ truit ] again , we obtain the desired result . @xmath116
the convergence analysis of kullback proximal algorithms is split into two parts , the first part being the subject of this section .
we prove that if the accumulation points @xmath9 of the kullback proximal sequence satisfy @xmath117 they are stationary points of the log - likelihood function @xmath16 .
it is also straightforward to show that the same analysis applies to the case of penalized likelihood estimation .
we start with the following useful lemma .
[ asymreg ] let @xmath118 and @xmath119 be two bounded sequences in @xmath87 satisfying @xmath120 assume that every couple @xmath121 of accumulation points of these two sequences lies in @xmath87 .
then , @xmath122 * proof*. first , one easily obtains that @xmath123 is bounded ( use a contradiction argument and assumption [ ass2 ] ( ii ) ) .
assume that there exits a subsequence @xmath124 such that @xmath125 for some @xmath126 and for all large @xmath99 .
since @xmath124 is bounded , one can extract a convergent subsequence .
thus we may assume without any loss of generality that @xmath127 is convergent with limit @xmath128 .
using the triangle inequality , we have @xmath129 .
since @xmath124 converges to @xmath130 , there exists a integer @xmath131 such that @xmath132 implies @xmath133 .
thus for @xmath132 we have @xmath134 .
now recall that @xmath135 is bounded and extract a convergent subsequence @xmath136 with limit denoted by @xmath137 .
then , using the same arguments as above , we obtain @xmath138 .
finally , recall that @xmath139 .
we thus have @xmath140 , and , due to the fact that the sequences are bounded and @xmath141 is continuous in both variables , we have @xmath142 .
thus assumption [ ass2 ] ( iv ) implies that @xmath143 and we obtain a contradiction .
hence , @xmath144 as claimed .
@xmath116 the main results of this section are the following .
first , we prove that under the assumptions [ ass1 ] , [ ass2 ] and [ ass3 ] , any cluster point @xmath9 is a global maximizer of @xmath145 .
we then use this general result to prove that such cluster points are stationary points of the log - likelihood function .
this result motivates a natural assumption under which @xmath9 is in fact a local maximizer of @xmath16 .
in addition we show that if the sequence @xmath146 converges to zero , i.e. @xmath147 , then @xmath9 is a global maximizer of log - likelihood .
finally , we discuss some simple conditions under which the algorithm converges , i.e. has only one cluster point .
the following theorem states a result which describes the stationary points of the proximal point algorithm as global maximizers of the asymptotic penalized function . [ the ] assume that @xmath148 .
let @xmath9 be any accumulation point of @xmath18 .
assume that @xmath149 .
then , @xmath9 is a global maximizer of the penalized function @xmath150 over the projection of @xmath59 onto its first coordinate , i.e. @xmath151 for all @xmath8 such that @xmath152 .
an informal argument is as follows .
assume that @xmath153 .
from the definition of the proximal iterations , we have @xmath154 for all subsequence @xmath109 converging to @xmath9 and for all @xmath155 .
now , assume we can prove that @xmath156 also converges to @xmath9 , we obtain by taking the limit and using continuity , that @xmath157 which is the required result .
there are two major difficulties when one tries to transform this sketch into a rigorous argument . the first one is related to the fact that @xmath16 and @xmath10 are only defined on domains which may not to be closed . secondly
, proving that @xmath156 converges to @xmath9 is not an easy task .
this issue will be discussed in more detail in the next section .
the following proof overcomes both difficulties .
* proof*. without loss of generality , we may reduce the analysis to the case where @xmath158 for a certain @xmath17 .
the fact that @xmath9 is a cluster point implies that there is a subsequence of @xmath159 converging to @xmath9 . for @xmath99
sufficiently large , we may assume that the terms @xmath160 belong to a compact neighborhood @xmath161 of @xmath162 included in @xmath59 . recall that @xmath163 for all @xmath8 such that @xmath164 and a fortiori for @xmath165 .
therefore , @xmath166 let us have a precise look at the `` long term '' behavior of @xmath10 .
first , since @xmath167 for all @xmath99 sufficiently large , lemma [ yal ] says that @xmath168 thus , for any @xmath126 , there exits an integer @xmath169 such that @xmath170 for all @xmath171 .
moreover , lemma [ asymreg ] and continuity of @xmath91 allows to conclude that @xmath172 since @xmath80 is continuous , for all @xmath126 and for all @xmath99 sufficienlty large we have @xmath173 on the other hand , @xmath174 is continuous in both variables on @xmath161 , due to assumptions [ ass1](i ) and [ ass2](i ) . by continuity in the first and second arguments of @xmath175 , for any @xmath176 there exists @xmath177 such that for all @xmath178 @xmath179 using ( [ eq ] ) , since @xmath16 is continuous , we obtain existence of @xmath180 such that for all @xmath181 @xmath182 combining equations ( [ eq1 ] ) and ( [ eq2 ] ) with ( [ eq0 ] ) , we obtain @xmath183 now , since @xmath184 , there exists an integer @xmath185 such that @xmath186 for all @xmath187 .
therefore for all @xmath188 , we obtain @xmath189 since @xmath10 is continuous and @xmath190 is bounded , there exists a real constant @xmath131 such that @xmath191 , for all @xmath192 .
thus , for all @xmath99 sufficiently large @xmath193 finally , recall that no assumption was made on @xmath8 , and that @xmath161 is any compact neighborhood of @xmath9 .
thus , using the assumption [ ass1](i ) , which asserts that @xmath4 tends to @xmath70 as @xmath71 tends to @xmath72 , we may deduce that ( [ maj ] ) holds for any @xmath8 such that @xmath194 and , letting @xmath195 tend to zero , we see that @xmath9 maximizes @xmath196 for over all @xmath8 such that @xmath197 belongs to @xmath59 as claimed .
@xmath116 using this theorem , we may now deduce that certain accumulation points on the strict interior of the parameter s space are stationary points of the log - likelihood function .
[ stat ] [ corolun ] assume that @xmath148 .
let @xmath9 be any accumulation point of @xmath18 .
assume that @xmath198 .
then , if @xmath16 is differentiable on @xmath56 , @xmath9 is a stationary point of @xmath4 .
moreover , if @xmath16 is concave , then @xmath9 is a global maximizer of @xmath16 .
* proof*. since under the required assumptions @xmath16 is differentiable and @xmath199 is differentiable at @xmath9 , theorem [ the ] states that @xmath200 since @xmath201 is minimum at @xmath9 , @xmath202 and we thus obtain that @xmath9 is a stationary point of @xmath16 .
this implies that @xmath9 is a global maximizer in the case where @xmath16 is concave .
@xmath116 .
theorem [ the ] seems to be much stronger than the previous corollary .
the fact that accumulation points of the proximal sequence may not be global maximizers of the likelihood is now easily seen to be a consequence of fact that the kullback distance - like function @xmath10 perturbs the shape of the likelihood function when @xmath8 is far from @xmath9 .
this perturbation does not have serious consequence in the concave case . on the other hand
, one may wonder whether @xmath9 can not be proved to be at least a local maximizer instead of a mere stationary point .
the answer is given in the following corollary .
let @xmath9 be an accumulation point of @xmath203 such that @xmath204 .
in addition , assume that @xmath16 and @xmath201 are twice differentiable in a neighborhood of @xmath9 and that the hessian matrix @xmath205 at @xmath9 is not the null matrix .
then , if @xmath74 is sufficiently small , @xmath9 is a local maximizer of @xmath16 over @xmath56 .
* proof*. assume that @xmath9 is not a local maximizer .
since @xmath206 is not the null matrix , for @xmath74 sufficiently small , there is a direction @xmath207 in the tangent space to @xmath56 for which the function @xmath208 has positive second derivative for @xmath91 sufficiently small .
this contradicts the fact that @xmath9 is a global maximizer of @xmath150 and the proof is completed .
@xmath116 the next theorem establishes global optimality of accumulation points in the case where the relaxation sequence converges to zero .
let @xmath9 be any accumulation point of @xmath203 .
assume that @xmath149 .
then , without assuming differentiability of either @xmath16 or of @xmath10 , if @xmath73 converges to zero , @xmath9 is a global maximizer of @xmath16 over the projection of @xmath59 along the first coordinate .
* proof*. let @xmath190 be a convergent subsequence of @xmath209 with limit denoted @xmath9 .
we may assume that for @xmath99 sufficiently large , @xmath160 belongs to a compact neighborhood @xmath161 of @xmath9 .
by continuity of @xmath16 , for any @xmath126 , there exists @xmath210 such that for all @xmath132 , @xmath211 on the other hand , the proximal iteration ( [ proxdef ] ) implies that @xmath212 for all @xmath213 .
fix @xmath213 .
thus , for all @xmath132 , @xmath214 since @xmath10 is a nonnegative function and @xmath12 is a nonnegative sequence , we obtain @xmath215 recall that @xmath209 is bounded due to lemma [ boundu ] .
thus , since @xmath10 is continuous , there exists a constant @xmath110 such that @xmath216 for all @xmath99 .
therefore , for @xmath99 greater than @xmath131 , @xmath217 passing to the limit , and recalling that @xmath73 tends to zero , we obtain that @xmath218 using the same argument as at the end of the proof of theorem [ the ] , this latter equation holds for any @xmath8 such that @xmath197 belongs to @xmath59 , which concludes the proof upon letting @xmath195 tend to zero .
@xmath116 one question remains open in the analysis of the previous section : does the sequence generated by the kullback proximal point converge ?
in other words : are there multiple cluster points ? in wu s
paper @xcite , the answer takes the following form . if the euclidean distance between two successive iterates tends to zero , a well known result states that the set of accumulation points is a continuum ( see for instance ( *
* theorem 28.1 ) ) and therefore , it is connected . therefore ,
if the set of stationary points of @xmath16 is a countable set , the iterates must converge .
[ corle ] let @xmath219 denote the set of accumulation points of the sequence @xmath18 .
assume that @xmath220 and that @xmath4 is strictly concave in an open neighborhood @xmath221 of an accumulation point @xmath9 of @xmath203 and that @xmath162 is in @xmath222 .
then , for any relaxation sequence @xmath12 , the sequence @xmath18 converges to a local maximizer of @xmath4 .
* proof*. we obtained in corollary [ stat ] that every accumulation point @xmath9 of @xmath18 in @xmath223 and such that @xmath224 is a stationary point of @xmath4 . since @xmath4 is strictly concave over @xmath225 , the set of stationary points of @xmath16 belonging to @xmath225 reduces to singleton .
thus @xmath9 is the unique stationary point in @xmath221 of @xmath16 , and _ a fortiori _ , the unique accumulation point of @xmath18 belonging to @xmath221 . to complete the proof , it remains to show that there is no accumulation point in the exterior of @xmath221 .
for that purpose , consider an open ball @xmath226 of center @xmath9 and radius @xmath195 included in @xmath221 . then
, @xmath227 is the unique accumulation point in @xmath226 .
moreover , any accumulation point @xmath228 , lying in the exterior of @xmath221 must satisfy @xmath229 , and we obtain a contradiction with the fact that @xmath219 is connected . thus every accumulation point lies in @xmath221 , from which we conclude that @xmath9 is the only accumulation point of @xmath203 or , in other words , that @xmath18 converges towards @xmath9 .
finally , notice that the strict concavity of @xmath4 over @xmath221 implies that @xmath9 is a local maximizer .
@xmath116 before concluding this section ,
let us make two general remarks .
* proving _ a priori _ that the set of stationary points of @xmath16 is discrete may be a hard task in specific examples . * in general
, it is not known whether @xmath220 holds .
in fact , lemma [ asymreg ] could be a first step in this direction . indeed if we could prove in any application that the mapping @xmath91 is injective , the desired result would follow immediately .
however , injectivity of @xmath91 does not hold in many of the standard examples ; in the case of gaussian mixtures , see for instance .
thus we are now able to clearly understand why the assumption that @xmath220 is not easily deduced from general arguments .
this problem has been overcome in where it is shown that @xmath91 is componentwise injective and thus performing a componentwise em algorithm is a good alternative to the standard em .
the goal of this section is to extend the previous results to the case where some cluster points lie on the boundary of the region where computation of proximal steps is well defined .
such cluster points have rarely been analyzed in the statistical literature and the strategy developed for the interior case can not be applied without further study of the kullback distance - like function .
notice further that entropic - type penalization terms in proximal algorithms have been the subject of an intensive research effort in the mathematical programming community with the goal of handling positivity constraints ; see @xcite and the references therein for instance .
the analysis proposed here applies to the more general kullback distance - like functions @xmath10 that occur in em .
our goal is to show that such cluster points satisfy the well known karush - kuhn - tucker conditions of nonlinear programming which extend the stationarity condition @xmath230 to the case where @xmath8 is subject to constraints .
as before , it is straightforward to extend the proposed analysis to the case of penalized likelihood estimation . in the sequel , the distance - like function will be assumed to have the following additional properties .
[ ass4 ] the kullback distance - like function @xmath10 is of the form @xmath231 where for all @xmath232 and @xmath233 , @xmath234 is continuously differentiable on its domain of definition , @xmath235 is a function from @xmath3 to @xmath236 , the set of positive real numbers , and the function @xmath237 is a non negative convex continuously differentiable function defined for positive real numbers only and such that @xmath238 if and only if @xmath239 . if @xmath240 and @xmath241 for all @xmath232 and all @xmath233 , the function @xmath10 is the well known @xmath237 divergence defined by csiszr in @xcite . assumption [ ass4 ] is satisfied in most standard examples ( for instance gaussian mixtures and poisson inverse problems ) with the choice @xmath242 .
the main property that will be needed in the sequel is that under assumption [ ass4 ] , the function @xmath10 satisfies the same property as the one given in lemma [ asymreg ] above , even on the boundary of its domain @xmath59 .
this is the result of proposition [ nondegphi ] below .
we begin with one elementary lemma .
[ convone ] under assumptions [ ass4 ] , the function @xmath237 is decreasing on @xmath243 , is increasing on @xmath244 and @xmath245 converges to @xmath72 when @xmath246 converges to @xmath72 .
we have @xmath247 if and only if @xmath248 .
* proof*. the first statement is obvious . for the second statement , the `` if '' part is trivial , so we only prove the `` only if '' part .
first notice that the sequence @xmath249 must be bounded .
indeed , the level set @xmath250 is bounded for all @xmath251 and contains the sequence @xmath252 for @xmath131 sufficiently large .
thus , the bolzano - weierstass theorem applies .
let @xmath253 be an accumulation point of @xmath249 .
since @xmath237 is continuous , we get that @xmath254 and thus we obtain @xmath255 . from this
, we deduce that the sequence has only one cluster point , which is equal to 1
. therefore , @xmath248 . @xmath116 using these lemmas , we are now in position to state and prove the main property of @xmath10 .
[ nondegphi ] the following statements hold .
\(i ) for any sequence @xmath18 in @xmath236 and any bounded sequence @xmath256 in @xmath236 , the fact that @xmath257 implies @xmath258 for all @xmath232,@xmath233 such that @xmath259 .
\(ii ) if one coordinate of one of the two sequences @xmath18 and @xmath256 tends to infinity , so does the other s same coordinate . *
proof*. fix @xmath232 in @xmath260 and @xmath233 in @xmath261 and assume that @xmath259 .
\(i ) we first assume that @xmath262 is bounded away from zero . since @xmath263 , then @xmath264 and lemma [ convone ] implies that @xmath265 .
thus , @xmath266 and since @xmath91 is continuous , @xmath267 is bounded .
this implies that @xmath268 .
next , consider the case of a subsequence @xmath269 which tends towards zero . for contradiction ,
assume the existence of a subsequence @xmath270 which remains bounded away from zero , i.e. there exists @xmath271 such that @xmath272 for @xmath99 sufficiently large .
thus , for @xmath99 sufficiently large we get @xmath273 and due to the fact that @xmath237 is increasing on @xmath244 , we obtain @xmath274 on the other hand , lemma [ convone ] says that for any @xmath275 , @xmath276 . since @xmath237 is convex , we get @xmath277 take @xmath278 in this last expression and combine with ( [ toto ] ) to obtain @xmath279 passing to the limit , we obtain @xmath280 which gives the required contradiction .
\(ii ) if @xmath281 then @xmath282 is a direct consequence of part ( i ) . indeed
, if @xmath267 remains bounded , part ( i ) says that @xmath258 , which contradicts divergence of @xmath283 .
now , consider the case where @xmath282 . then
, a contradiction is easily obtained if we assume that at least a subsequence @xmath284 stays bounded from above .
indeed , in such a case , we have @xmath285 and thus , @xmath286 for some @xmath287 since we know that @xmath237 is decreasing on @xmath243 and @xmath288 .
this implies that @xmath289 which is the required contradiction .
@xmath116 the main result of this section is the property that any cluster point @xmath9 such that @xmath162 lies on the boundary of @xmath59 satisfies the karush - kuhn - tucker necessary conditions for optimality on the domain of the log - likelihood function . in the context of assumptions [ ass4 ] , @xmath59 is the set @xmath290 we have the following theorem .
[ bord ] let @xmath9 be a cluster point of the kullback - proximal sequence .
assume that all the functions @xmath234 are differentiable at @xmath9 .
let @xmath291 be the set of all couples of indices @xmath292 such that the constraint @xmath293 is active at @xmath9 , i.e. @xmath294 . if @xmath9 lies in the interior of @xmath56 , then @xmath9 satisfies the karush - kuhn - tucker necessary conditions for optimality , i.e. there exists a family of reals @xmath295 , @xmath296 such that @xmath297 * proof*. let @xmath298 denote the bivariate function defined by @xmath299 let @xmath300 be a convergent subsequence of the proximal sequence with limit equal to @xmath9 . the first order optimality condition at iteration @xmath99 is given by @xmath301 we have @xmath302 for all @xmath232 and @xmath233 .
* claim a*. _ for all @xmath292 such that @xmath303 , we have @xmath304 _ * proof of claim a*. two cases may occur . in the first case , we have @xmath294 .
since the sequence @xmath305 is bounded due to lemma [ boundu ] , continuous differentiability of @xmath237 and the @xmath234 proves that @xmath306 is bounded from above .
thus , the desired conclusion follows . in the second case , @xmath307 and applying lemma [ yal ]
, we deduce that @xmath308 tends to zero .
hence , @xmath309 , which implies that @xmath310 . from this and assumptions
[ ass4 ] , we deduce that @xmath311 . since @xmath300 converges to @xmath9 and that @xmath307 , we obtain that the subsequence @xmath312 is bounded from above .
moreover , @xmath313 is also bounded by continuous differentiability of @xmath234 .
therefore , the fact that @xmath311 establishes claim a. @xmath116 using this claim , we just have to study the remaining right hand side terms in ( [ frst ] ) , namely the expression @xmath314 .
let @xmath315 be a subset of the active indices @xmath316 such that the family @xmath317 is linearly independent .
this linear independence is preserved under small perturbations , we may assume without loss of generality that the family @xmath318 is linearly independent for @xmath99 sufficiently large .
for such @xmath99 , we may rewrite equation ( [ frst ] ) as @xmath319 * claim b*. _ the sequence @xmath320 is bounded . _ * proof of claim b*. using the previous claim and the continuous differentiability of @xmath16 and @xmath234 , equation ( [ frstbis ] ) expresses that @xmath321 are proportional to the coordinates of the projection on the span of the @xmath322 of a vector converging towards @xmath323 . since @xmath322 , for @xmath324 , form a linearly independent family for @xmath99 sufficiently large , none of the coordinates can tend towards infinity .
we are now in position to finish the proof of the theorem .
take any cluster point @xmath325 of @xmath326 . using claim b , we know that @xmath327 lies in a compact set .
let @xmath328 be a cluster point of this sequence . passing to the limit
, we obtain from equation ( [ frst ] ) that @xmath329 for every cluster point @xmath74 of @xmath330 . for
all @xmath331 , set @xmath332 .
this equation is exactly the karuch - kuhn - tucker necessary condition for optimality .
@xmath116 if the family @xmath333 is linearly independent for @xmath99 sufficiently large , theorem [ bord ] holds and in addition the @xmath334 are nonnegative , which proves that @xmath9 satisfies the karush - kuhn - tucker conditions when it lies in the closure of @xmath335 .
the goal of this section is to illustrate the utility of the previous theory for a nonparametric survival analysis with competing risks proposed by ahn , kodell and moon in .
this problem can be described as follows .
consider a group of @xmath336 animals in an animal carcinogenecity experiment .
sacrifices are performed at certain prescribed times denoted by @xmath337 in order to study the presence of the tumor of interest .
let @xmath338 be the time to onset of tumor , @xmath339 the time to death from this tumor and @xmath340 be the time to death from a cause other than this tumor .
notice that @xmath338 , @xmath339 and @xmath340 are unobservable .
the quantities to be estimated are @xmath341 , @xmath342 and @xmath343 , the survival function of @xmath338 , @xmath339 and @xmath340 respectively . it is assumed that @xmath338 and @xmath339 are statistically independent of @xmath340 . a nonparametric approach to estimation of @xmath75 , @xmath344 and @xmath345 is proposed in : observed data @xmath346 are the number of deaths on every interval @xmath347 $ ] which can be classified into the following four categories , * death with tumor ( without knowing cause of death ) * death without tumor * sacrifice with tumor * sacrifice without tumor this gives rise to a multinomial model whose probability mass is parametrized by the values of @xmath75 , @xmath344 and @xmath345 at times @xmath348 .
more precisely , for each time interval @xmath347 $ ] denote by @xmath349 the number of deaths with tumor present , @xmath350 the number of deaths with tumor absent , @xmath351 the number of sacrifices with tumor present and @xmath352 the number of sacrifices with tumor absent .
let @xmath353 be the number of live animals in the population at @xmath354 , it is shown in that the corresponding log - likelihood is given by @xmath355 where @xmath356 is a constant @xmath357 , @xmath358 and @xmath359 , @xmath360 , @xmath361 and the parameter space is specified by the constraints @xmath362 where the last nonconvex constraint serves to impose monotonicity of @xmath75 .
note that monotonicity of @xmath344 and @xmath345 is a direct consequence of the constraints on the @xmath363 s and the @xmath364 s , respectively .
define the complete data @xmath365 as a measurement that indicates the cause of death in addition to the presence of absence of a tumor in the dead animals .
specifically , @xmath365 should fall into one of the following categories * death caused by tumor and death with incidental tumor * death without tumor * sacrifice with tumor * sacrifice without tumor to each time interval @xmath347 $ ] among those animals dying of natural causes , there correspond the numbers @xmath366 of deaths caused by tumor and the number @xmath367 of deaths with incidental tumor , neither of which are observable .
the associated complete log - likelihood function is given by @xmath368 now , we have to compute the expectation @xmath369 of the log - likelihood function of the complete data conditionally to the parameter @xmath35 .
the random variables @xmath366 and @xmath367 are binomial with parameter @xmath370 and @xmath371 where @xmath370 is the probability that the death was caused by the tumor conditioned on the presence of the tumor . conditioned on @xmath35 , we have @xmath372 ( see for details ) . from this
, we obtain that the conditional mean values of @xmath366 and @xmath367 are given by @xmath373=\lambda_j c_j \hspace{.4 cm } \text { and } \hspace{.4 cm } { \rm e}[a_{1j } \mid y;\bar{\theta}]=(1-\lambda_j ) c_j.\ ] ] therefore @xmath374 from this , we can easily compute the associated kullback distance - like function : @xmath375 with @xmath376 and @xmath237 is defined by @xmath377 .
it is straightforward to verify that assumptions [ ass1 ] , [ ass2 ] , [ ass3 ] and [ ass4 ] are satisfied .
the main computational problem in this example is to handle the difficult nonconvex constraints entering the definition of the parameter space @xmath378 .
the authors of and use the complex method proposed by box in @xcite to address this problem . however , the theoretical convergence properties of box s method are not known as reported in article mr0184734 in the math . reviews .
using our proximal point framework , we are able to easily incorporate the nonconvex constraints into the kullback distance - like function and obtain an efficient algorithm with satisfactory convergence properties . for this purpose ,
let @xmath379 be defined by @xmath380 where @xmath381 using this new function , the nonconvex constraints @xmath382 are satisfied for all proximal iterations and assumptions [ ass4 ] still hold .
we implemented the kullback proximal algorithm with different choices of relaxation sequence @xmath12 , @xmath383 .
the m - step of the em algorithm does not have a closed form solution , so that nothing is lost by setting @xmath65 to a constant not equal to one .
we attempted to supplement the kpp - em algorithm with the newton method and other built - in methods available in scilab but they were not even able to find local maximizers due to the explosive nature of the logarithms near zero , leading these routines to repetitive crashes . to overcome this difficulty
, we found it convenient to use the extremely simple simulated annealing random search procedure ; see @xcite for instance .
this random search approach avoids numerical difficulties encountered using standard optimization packages and easily handles nonconvex constraints .
convergence of this procedure is well established and recent studies such as @xcite confirm the good computational efficiency for convex functions optimization .
some of our results for the data of table 1 of are given in figures 1 to 4 . in the reported experiments , we chose three constant sequences with respective values @xmath384 .
we observed the following phenomena * 1 . * after one hundred iterations the increase in the likelihood function is less than @xmath385 except for the case @xmath386 ( figure [ fig4 ] ) where the algorithm had not converged . *
2 . * for @xmath386
we often obtained the best initial growth of the likelihood * 3 . * for @xmath387
we always obtained the highest likelihood when the number of iterations was limited to 50 ( see figure [ fig3 ] for the case mcl male al ) .
it was shown in that penalizing with a parameter sequence @xmath388 converging towards zero implies superlinear convergence in the case where the maximum likelihood estimator lies in the interior of the constraint set .
thus , our simulations results seem to confirm observation 3 .
the second observation was surprising to us but this phenomenon occured repeatedly in our experiments .
this behavior did not occur in our simulations for the poisson inverse problem in for instance . in conclusion
, this competing risks estimation problem is an interesting test for our kullback - proximal method which shows that the proposed framework can provide provably convergent methods for difficult constrained nonconvex estimation problems for which standard optimization algorithms can be hard to tune .
the relaxation parameter sequence @xmath389 also appeared crucial for this problem although the choice @xmath390 could not really be considered unsatisfactory in practice .
evolution of the log - likelihood versus iteration number : mcl female al case ]
the goal of this paper was the study of the asymptotic behavior of the em algorithm and its proximal generalizations .
we clarified the analysis by making use of the kullback - proximal theoretical framework .
two of our main contributions are the following .
firstly we showed that interior cluster points are stationary points of the likelihood function and are local maximizers for sufficiently small values of @xmath17 .
secondly , we showed that cluster points lying on the boundary satisfy the karush - kuhn - tucker conditions .
such cases were very seldom studied in the literature although constrained estimation is a topic of growing importance ; see for instance the special issue of the journal of statistical planning and inference @xcite which is devoted to the problem of estimation under constraints . on the negative side ,
the analysis from the kullback - proximal viewpoint allowed us to understand why uniqueness of the cluster point is hard to establish theoretically . on the positive side
, we were able to implement a new and efficient proximal point method for estimation in the difficult tumor lethality problem involving nonlinear inequality constraints .
h. ahn , h. moon and r.l .
kodell , `` attribution of tumour lethality and estimation of the time to onset of occult tumours in the absence of cause - of - death information '' .
c _ , vol .
2 , 157169 , 2000 .
h. moon , h. ahn , r. kodell and b. pearce `` a comparison of a mixture likelihood method and the em algorithm for an estimation problme in animal carcinogenicity studies , '' computational statistics and data analysis , 31 , no .
2 , pp . 227238 , 1999 .
|
in this paper , we analyze the celebrated em algorithm from the point of view of proximal point algorithms .
more precisely , we study a new type of generalization of the em procedure introduced in and called kullback - proximal algorithms .
the proximal framework allows us to prove new results concerning the cluster points .
an essential contribution is a detailed analysis of the case where some cluster points lie on the boundary of the parameter space .
|
the introduction of large - size nanoparticles of different shapes has emerged as the best practical way to improve the current performance of high - temperature superconductors . while in the first superconducting cables
the critical currents were limited by weak links , in the second - generation superconducting wires based on aligned yba@xmath0cu@xmath1o@xmath2 ( ybco ) films this problem has been mostly resolved and critical currents are determined by vortex pinning .
impressive progress has been made to enhance critical currents in these films using both isotropic @xcite and columnar @xcite inclusions . in spite of this progress ,
our understanding of strong pinning mechanisms is far from satisfactory .
theoretical estimates describing the pinning of vortex lines by an array of strong pins at low temperatures were elaborated by ovchinnikov and ivlev.@xcite this theory was applied to describe the behavior of the critical currents in the real ybco films in refs .. in particular , frequently observed power - law decay of the critical current as a function of the magnetic field with a power slightly larger than @xmath3 is naturally explained by this theory .
more recently , it was also argued that strong pins of unknown origin determine critical currents at low magnetic fields in several iron pnictide compounds.@xcite it is not clear , however , to what extent available qualitative estimates describe the real situation . due to the obvious importance of strong pinning by large - size inclusions for real superconducting materials , it is desirable to elaborate a quantitative theory describing pinning in such situations . moreover , the very important issue of pinning suppression by thermal fluctuations does not have any theoretical description in the strong - pinning regime .
in this paper we consider the pinning of vortex lines in a superconductor containing insulating inclusions with lateral sizes larger than the coherence length .
we focus on the pinning of individual vortex lines corresponding to small magnetic fields .
pinning occurs via the trapping of finite - size segments of a vortex line@xcite with a typical length @xmath4 , as illustrated in fig .
[ fig - trappedline ] .
the critical current is determined by the length of the trapped segment @xmath4 and the pin - breaking force @xmath5 , @xmath6 therefore , in the strong - pinning regime the critical - current problem is mostly reduced to evaluation of the trapped - segment length @xmath4 . in general , trapping of the vortex lines is a complicated dynamic process controlled by the competition between the pinning energy , line tension , and intervortex interactions .
different approaches may be used to evaluate the trapped - segment length .
one can assume that the pinning center always grabs a piece of the vortex line when it is energetically favorable .
this assumption implies that thermal fluctuations facilitate local equilibration . in this case
the parameters of the trapped configurations can be obtained from the energy - balance estimates .
we will call this type of trapping the equilibrium regime .
this energy - balance consideration determines trapped configurations of static vortex lines prepared by cooling in finite magnetic field
. however , such consideration is not applicable to the more typical dynamic scenario when moving lines are trapped after the driving force is slowly reduced down to the critical value . in this case
the line motion close to the critical force is a continuous trapping - detrapping process limited by local instabilities .
the driven vortex can be captured when the line either directly collides with a strong pin or passes sufficiently close to it . in the second case trapping
may occur due to the long - range pin - vortex interaction as a result of local instability.@xcite the line remains trapped until the force acting from the pin does not exceed the pin - breaking force . when the line finally stops , the pinned configuration is expected to be very anisotropic because the transverse pin - to - pin displacements which are determined by trapping events are much smaller than the longitudinal displacements limited by the pin - breaking criteria . to develop a quantitative picture and verify analytical estimates , we explore in this paper the pinning of the vortex lines by nanoparticles with extensive numerical simulations .
we study the dependence of the critical force on the density of pins , the statistical properties of trapped lines including average values , and the distributions of trapping length and pin - to - pin displacements .
we study long - range behavior of line displacements in the direction of the driving force and in the transverse direction .
we also study in detail the suppression of the apparent critical force by thermal fluctuations and the temperature dependence of the trapping parameters .
the paper is organized as follows . in section [ sec - inter ]
we describe parameters characterizing the interaction between a vortex line and large - size pinning centers . in section [ sec - analest ]
we present analytical estimates .
this includes the formulation of general conditions for stable trapped configurations in subsection [ subsec - cond ] , making estimates for parameters of trapped line in equilibrium , subsection [ subsec - equil ] , and in the case of dynamic trapping , subsection [ subsec - dyntrap ] . in the section [ sec - nummodel ]
we describe the model used in our numerical simulations . in the section [ sec - numresults ] we present our numerical results including the zero - temperature case in subsection [ subsec - zerot ] and the finite temperature in subsection [ subsec - finitet ] in section [ sec - discussion ] we discuss our results and make preliminary comparisons with experiments .
we consider first the essential parameters describing the interaction of vortex lines with large - size pinning centers .
the vortex pinning energy by an insulating spheroid inclusion with the axes @xmath7 and @xmath8 is given by@xcite @xmath9 with @xmath10.a very important parameter is the pin - breaking force , the maximum force with which the pinning center can attract the vortex line .
in contrast to small defects , the pin - breaking force for large - size defects is limited by the line tension of the vortices . with increasing external force , the tips of the vortex line slide along the surface of the insulating inclusion until they meet near the equator and reconnect leading to the depinning of the vortex . for the in - plane current in the anisotropic layered material , the upper estimate for such a line - tension - limited force can be obtained by considering the simple geometry of equally spaced pins aligned along the @xmath11 axis and neglecting the interaction between vortex tips at the pin surface . in this case , evaluating the external force at which the tips meet , we obtain the following estimate @xmath12 where @xmath13 is the anisotropy factor , @xmath14 , @xmath15 is the c - axis coherence length , and @xmath16 is the interlayer period of a layered superconductor .
this force only weakly depends on the size and shape of the pinning center .
the interaction of the vortex line with a remote pin is long ranged , due to the perturbation of the supercurrent flow around the vortex by the pin , @xmath17 where @xmath18 is the volume of the pinning center and @xmath19 are `` depolarization factors '' , which depend on the parameter @xmath20 . in particular , in the case @xmath21 which includes close - to - spherical inclusions , @xmath22 and @xmath23 .
recently , it was demonstrated that magnetic force microscopy can be effectively used not only for imaging but also for the manipulation of individual vortices @xcite .
this technique gives principal possibility to measure the interaction between a vortex line and an individual pinning center and extract the relevant interaction parameters described in this section .
consider a general vortex - line configuration trapped at the points @xmath24 . for simplicity , we assume that the forces from the pins are applied locally at the points @xmath25 . in between the trapped points , @xmath26 , the displacement obeys the following equation @xmath27 where @xmath28 is the driving force applied along the @xmath29 axis and @xmath30 is the line tension with @xmath31 being the logarithmic factor @xmath32.@xcite the displacement can be found as @xmath33 the force acting from the pin on the vortex line at @xmath25 is given by @xmath34\ ] ] and can be evaluated as @xmath35 the stability condition for the trapped line is given by @xmath36 while the critical state corresponds to the condition that at least one local force reaches the pin - breaking force @xmath37 one simple consequence of eqs .
( [ fn ] ) and ( [ stabcond ] ) is that for `` behind '' sites , @xmath38 , the line - tension force adds to the external force meaning that they , in average , have shorter trapping segments @xmath39 , @xmath40 .
consider the trapping of a single vortex line by strong - pinning centers with concentration @xmath41 and pinning energy @xmath42 , eq .
( [ pinenergy ] ) . assuming local equilibrium , the typical trapping length @xmath4 and transverse displacement @xmath43 are determined by the energy - balance condition
@xcite @xmath44 and by the condition that the average number of impurities in the trapping volume should be of the order of one , @xmath45 these equations give @xmath46 strictly speaking , the above conditions are obtained for zero current . assuming that the trapping length does not change much when current is applied , we obtain an estimate for the critical current for the equilibrium trapping @xmath47 it is expected to increase with the pin density as @xmath48 .
is much larger than the typical displacement in the perpendicular direction @xmath49 . _
lower figure _ is a visualization of the pinned line configuration obtained in simulations ( only a short section of the line is shown ) .
shadelike projections on the axis planes illustrate line displacements in the different directions .
short scale line wiggling is due to the thermal noise .
the definitions of the trapping parameters @xmath50 , @xmath51 , and @xmath49 are also illustrated.,width=307 ] below which the trapping becomes energetically favorable but separated by the energy barrier .
the lower right plot illustrates the energy profile near the trapping instability point . , width=230 ] the equilibrium estimates for the trapping parameters ( [ l - u2-eq ] ) are definitely valid for the line configurations prepared by cooling at fixed field and at zero transport current .
however , it is clear that they can not be applied to the vortex lines in the critical state at low temperatures when moving lines are trapped after the driving force drops below the critical value and equilibration does not take place .
the critical current in such a dynamic regime has been estimated in ref . for high fields when the intervortex interactions are essential .
these estimates can be directly generalized to the trapping of individual vortex lines at small fields.@xcite when the vortex line moves close to the pinning center , it may be trapped and the line remains trapped until the force acting from the pin on the vortex line does not exceed the pin - breaking force . in this regime
two typical trapping distances , in the direction of motion , @xmath52 , and in the transverse direction , @xmath53 , are very different and have very different origins , see fig .
[ fig - metastablesingle ] .
these distances and the trapped - segment length @xmath4 are connected by the geometric relation @xmath54 the longitudinal trapping distance , @xmath52 , is determined by the pin - breaking condition , @xmath55 this condition can be obtained from the @xmath29 component of eq .
( [ fn ] ) assuming that the two terms on the right - hand side are of the order of the pin - breaking force . the transverse displacement between the pins , @xmath53 , is determined by the trapping events .
the simplest assumption is that in most cases trapping occurs when the lines directly collide with the pins @xcite meaning that @xmath56 .
this immediately gives estimates for other trapping parameters , @xmath57 this corresponds to the following result for the critical current @xmath58 note that in this situation the estimate is somewhat similar to the result for the equilibrium case , eq .
( [ eqcritcur ] ) , and has the same dependence on the pin density , @xmath59 . the physical assumptions behind the two estimates , however , are completely different .
the above simple assumption , however , may underestimate @xmath53 .
due to the long - range pin - vortex interaction ( [ pinvortintera ] ) a pinning center may capture the vortex line even without direct collisions .
when the line passes sufficiently close to the pinning center it may be trapped by this center due to instability . to estimate the maximum transverse trapping distance @xmath53
, we consider the interaction energy of a segment of length @xmath4 with a pinning center located at a distance @xmath60,@xcite see fig .
[ fig - segmentinteraction ] , @xmath61 with @xmath62 $ ] .
this gives the interaction force
@xmath63 introducing the reduced variables @xmath64 we rewrite the energy and force as @xmath65 , \\
f(x ) & = -\varepsilon_{1}\frac{4r}{l}\left ( x-\frac{w}{(1-x)^{3}}\right ) .\end{aligned}\ ] ] the equilibrium points are determined by @xmath66 the instability point corresponds to the value of @xmath67 when the equilibrium points vanish which happens at @xmath68=3^{3}/4^{4}$ ] . therefore the condition for the instability can be written as @xmath69 which determines the maximum trapping distance @xmath53 as @xmath70 the vortex line will be trapped by the pinning centers located closer than this distance in the direction perpendicular to the driving force .
note that the numerical coefficient in this equation should not be taken too literally because it is only correct for the simplest geometry illustrated in fig .
[ fig - segmentinteraction ] . using this result
, we find from eqs .
( [ trapmetaeq2 ] ) and ( [ trapmetaeq1 ] ) @xmath71 ^{4/9},\ \label{ltmeta}\\ u_{l } & = \left [ \frac{f_{p}^{5/4}}{n_{p}\varepsilon_{1}\left ( \varepsilon_{0}v_{p}\right ) ^{1/4}}\right ] ^{4/9},\label{ulmeta}\\ u_{t } & = \left [ \frac{\varepsilon_{0}^{2}v_{p}^{2}}{n_{p}\varepsilon _ { 1}f_{p}}\right ] ^{1/9}.\label{utmeta}\end{aligned}\ ] ] this gives the following estimate for the critical current @xmath72 comparing this result with the critical current for the equilibrium regime ( [ eqcritcur ] ) , we can see that the two regimes are characterized by somewhat different dependences on the parameters .
however , dependences on the pin density occur to be close , the exponent in the power law , @xmath73 in this dynamic - trapping regime is somewhat smaller than the power @xmath3 for the equilibrium regime , @xmath74 . at finite temperatures the line moves at all driving forces but for small forces very slow motion occurs due to the rare thermally activated jumps ( creep regime ) .
the creep and flow regimes are separated by the effective critical force which can be evaluated using a velocity criterion .
such effective critical force is criterion dependent and thus differs from the sharply defined zero - temperature critical force of the depinning transition , but it has the advantage that it can be directly compared with experimental estimates .
we do not consider creep in this paper and our purpose is to evaluate the suppression of this critical force by the thermal noise .
the main mechanism of thermal suppression is reduction of the effective pin - breaking force . at finite temperatures
the trapped vortex segment has a finite lifetime on the pin even if the pinning force @xmath75 is smaller than the maximum pin - breaking force . to quantify this effect
, we can introduce the temperature - dependent force @xmath76 at which the trapped segments are typically released from pins . to evaluate this force
, we assume that the line motion in the crossover region consists of segment jumps , meaning that the average line velocity can be estimated as @xmath77 where @xmath78 $ ] is the typical time during which the vortex segment remains pinned and @xmath79 is the typical energy barrier for detrapping . when the force @xmath75 acting from the pin is only slightly smaller than the maximum pin - breaking force , the barrier behaves as in the single - particle case,@xcite @xmath80 , where @xmath81 and @xmath82 is the second derivative of the interaction force with respect to the line displacement at the maximum - force point .
the velocity criterion for the effective critical force can be written as @xmath83 where @xmath84 is the viscosity coefficient and @xmath85 . at low temperatures
this gives us the following relation for @xmath86 @xmath87 \approx\frac{u_{l}}{c_{f}f_{c}}.\ ] ] using the estimate @xmath88 , the geometric relation ( [ trapmetaeq2 ] ) and assuming for simplicity that @xmath56 , we obtain @xmath89 .
\label{renormfp}\ ] ] we expect that at low temperatures the effective critical force and trapping parameters can be roughly evaluated using the simple replacement @xmath90 .
to develop a quantitative understanding of the strong pinning by the array of inclusions , we numerically simulated motion of the vortex line described by the dynamic equation @xmath91 here @xmath28 is the driving force along the @xmath29 direction from the current , @xmath92 are the random pin coordinates , @xmath93 is the langevin thermal force @xmath94 and @xmath95 is the interacting force with a strong pin .
we model the interaction potential by the function @xmath96 where the cutoff function @xmath97 is introduced for numerical convenience , @xmath98 for @xmath99 and @xmath100 for @xmath101 with @xmath102 .
an important feature that was not taken into account in modeling before is the long @xmath103 tail in the interaction potential .
however , this model does not describe the line - tension limited pin - breaking force . in our model
the pin - breaking force from an isolated pin is given by @xmath104 which is achieved at @xmath105 .
the model in its original form has an unrealistic feature .
for improbable configurations when many pins are located at distances smaller than @xmath7 , the vortex interaction with such a cluster may increase without limit .
this , of course , does not happen in real superconductors . to bring our model
somewhat closer to reality , we renormalized the total pin - vortex interaction force as @xmath106 as @xmath107 so that the maximum force can not exceed @xmath108 .
this modification has only a minor influence on the interaction of the vortex line with an isolated pin . for numerical implementation of the model
, we use the reduced variables @xmath109 in which the equation takes the simpler form @xmath110 with @xmath111 in this dimensionless form , the equation depends only on the reduced temperature and reduced pin density @xmath112 .
the condition of the strong pinning regime is @xmath113 corresponding to @xmath114 . for the typical parameters @xmath115 , @xmath116
nm , and @xmath117nm@xmath118@xmath119 , the reduced pin density can be estimated as @xmath120 . in simulations
we mostly used @xmath121 . for an isolated pin
this gives the pin - breaking force @xmath122 .
we also used @xmath123 in the cutoff function @xmath124 .
we consider systems of size @xmath125 , where @xmath126 is the vortex - line length , @xmath127 is the size in the direction of line motion , and @xmath128 the size in the transverse direction .
one of our tasks is to compute the dependence of the critical force on the density of the pinning centers .
the calculation of the steady - state critical force is not as trivial problem as it may appear . for a finite - size system
there is always a metastable configuration giving the maximum depinning force .
assuming that such a configuration is always reachable from any initial condition at long times , this would be the critical force we seek . in a large system
this configuration is , however , determined by a rare , nontypical configuration of pins . as a consequence
, the maximum depinning force slowly grows with the increasing the system sizes @xmath127 and @xmath128 , meaning that it is not a self - averaging parameter for a fixed line length @xmath126 . to avoid this ,
a carefully chosen ( anisotropic ) thermodynamic limit in all directions was proposed for @xmath129 dimensions @xcite .
the maximum depinning force of a very large system is also very difficult to compute .
simulations of eq .
( [ redeq ] ) at zero temperature are not very suitable for this purpose , because when the external force is close to the critical value , the vortex line traps forever in the first metastable state it finds .
it is not clear how representative this state is and how close the corresponding critical force is to the maximum value . in addition , neither the maximum critical force nor depinning forces for a few accidental metastable states are very interesting quantities from a practical point of view .
the maximum critical force is essentially a property of an isolated vortex line .
a more interesting quantity is the typical pinning force for the finite density of the vortex lines . indeed , even at low densities when vortex - vortex interactions can be neglected , we expect a typical pinning force rather than an extreme non - self - averaging force value to determine the observable critical current . to evaluate a typical pinning force at zero temperature , instead of the fixed - force approach , we employ fixed - velocity simulations using approach suggested in ref . .
namely , we replaced the fixed external force @xmath130 in eq .
( [ redeq ] ) with a slowly moving parabolic potential , @xmath131,\ w(t)=w_0+vt\label{movingparab}\ ] ] such a potential forces the vortex line to move with the average velocity @xmath132 .
every time the line finds a metastable pinned state and stops , the dragging force starts to increase with time until it exceeds the critical force for this state and the line resumes motion .
this trick allows us to explore many metastable states and to avoid the extreme value statistics of the sample - dependent critical force .
the typical critical force is then evaluated as the average force acting on the vortex line in the critical configurations , @xmath133\rangle_z\rangle_t\label{fcmovingparab}\ ] ] where @xmath134 implies averaging over the local maxima of the instantaneous force . the spring constant @xmath135 and the drag velocity @xmath132 have to be taken sufficiently small so that they do not influence the calculation of the critical force .
we typically use @xmath136 and @xmath137 .
the spring constants satisfy @xmath138 in each case , assuring a proper thermodynamic limit for the critical force and associated critical configuration @xcite , and the velocities are small enough to assure a quasistatic stick - slip motion .
figure [ fig - displ - force - time](left ) illustrates the typical dependences of the force acting on the line on the displacement of its center of mass for different pin densities @xmath41 and line lengths @xmath139 . and line lengths @xmath139 ( the curves are marked by @xmath140 ) .
the vertical segments correspond to trapped states .
the typical critical forces , @xmath141 , are obtained by averaging over the local maxima of these curves .
_ right : _ examples of the displacement - time dependences used to evaluate average velocities at fixed temperature for the parameters shown in the plot and for different forces .
each force is represented by three curves corresponding to different realizations of the random potential .
the line motion becomes more and more uneven with decreasing force.,title="fig:",width=157 ] and line lengths @xmath139 ( the curves are marked by @xmath140 ) .
the vertical segments correspond to trapped states .
the typical critical forces , @xmath141 , are obtained by averaging over the local maxima of these curves .
_ right : _ examples of the displacement - time dependences used to evaluate average velocities at fixed temperature for the parameters shown in the plot and for different forces .
each force is represented by three curves corresponding to different realizations of the random potential .
the line motion becomes more and more uneven with decreasing force.,title="fig:",width=168 ] we also explore the velocity - force dependences at finite temperatures using the direct fixed - force simulations described by eq .
( [ redeq ] ) .
even though the term `` _ _ critical force _ _ '' is widely used in the experimental community , at finite temperatures the concept of the critical force does not have an exact meaning because the velocity is finite at all forces due to the thermal creep .
nevertheless , one can still introduce the characteristic force describing crossover between the flux - flow and flux - creep regimes using some average - velocity criterion . at low temperatures near such critical force
the line motion becomes very uneven , see fig .
[ fig - displ - force - time](right ) .
it spends considerable time in metastable traps waiting for a strong fluctuation which allows it to continue motion .
such line motion is illustrated by animation.@xcite as a consequence , a proper averaging over such events requires huge simulation times and/or averaging over many realizations of the random potential .
in addition to the critical force , we explore statistical parameters of trapping which allow us to understand better the pinning mechanism .
we evaluated the average length of a trapped segment @xmath142 and typical displacements along the motion direction @xmath52 and in the perpendicular direction @xmath53 .
we studied the distribution of these parameters and their evolution with temperature and force .
we also studied the long - range wandering of the line in the direction of motion and in the transverse direction . for a numerical solution the reduced equation ( [ redeq ] )
has to be discretized both in time and in the @xmath143 coordinate .
we typically used @xmath144 and @xmath145 for the discretization steps . to study finite - size effects ,
the equation was solved for different numbers of @xmath143-axis slices , @xmath146 , ranging from 512 to 2048 .
the finite - size effects in the critical force are weak.,width=288 ] . _
upper plot : _ the dependences of the average trapped - segment length @xmath50 . for comparison ,
we show the expected trapping length extracted from the critical force . _ lower plot : _ the pin - density dependences of the average pin - to - pin displacements in the direction of the force ( longitudinal ) and in the perpendicular direction ( transverse ) .
the longitudinal displacement , @xmath147 , has a considerable finite - size effect with respect to the line length @xmath139.,width=240 ] we systematically studied the behavior of the critical force and the properties of trapped line configurations within a pin - density range spanning two orders of magnitude , from @xmath148 to @xmath149 .
figure [ fig - fc - np ] presents the dependences of the critical force @xmath141 on the pin density @xmath150 for different systems sizes @xmath139 .
we found that @xmath141 increases with @xmath150 according to the power law @xmath151 .
the fit gives for the power index a value slightly smaller than 1/2 , @xmath152 and the coefficient @xmath153 .
in fact , the square - root dependence @xmath154 also provides a reasonable description of the data .
the power is , however , clearly larger than the value @xmath155 suggested by the dynamic - trapping estimates in the case when trapping occurs due to instabilities . for the used line lengths
@xmath156 a noticeable finite - size effect becomes visible only for small pin densities @xmath157 . to understand the statistical properties of trapped configurations , we plot in fig .
[ fig - trap - param - np ] the pin - density dependences of the average trapping parameters defined in fig .
[ fig - metastablesingle ] , the trapping length @xmath50 , and the pin - to - pin displacements along the direction of force , @xmath51 , and in the transverse direction , @xmath49 . in the plot @xmath158
we show the expected value of the typical trapped segment extracted from the value of the critical force @xmath159 .
we can see that the real trapping segments extracted from the configurations indeed closely follow the expected values .
we also see that the average trapping segments are systematically smaller than the segments which determine the critical force .
this is a natural behavior because one can expect that the critical force is determined by `` weak spots '' where the trapping segments are longer than average along the line .
the difference , however , is not very significant .
the trapping length shows a weak but unexpected size effect , it slightly _ decreases _ with increasing total line length .
we found that the product @xmath160 , which on general ground is expected to be of the order of unity is , in fact , slowly decreases with @xmath150 from @xmath161 to @xmath162 . from the plots of the pin - to - pin displacements
, we can see that the pinned configurations are strongly anisotropic , the average displacement along the direction of the driving force @xmath51 significantly exceeds the displacement in the transverse direction @xmath49 .
this is consistent with the dynamic - trapping picture described in sec .
[ subsec - dyntrap ] .
the difference grows with decreasing pin density @xmath150 .
in addition , the longitudinal displacement shows a significant size effect , it grows with an increasing system size @xmath139 .
further analysis shows that this is an indication of growing local stresses with increasing line length and suggests the destruction of the vortex lines by the pinning potential in the critical state for sufficiently large systems .
this behavior is not anticipated by the simple estimates .
it is interesting to note , however , that this size effect in @xmath51 does not lead to a significant size dependence of the critical force . on the other hand
, the average transverse displacement @xmath49 does not show any size effect .
it slowly grows with decreasing @xmath150 from 1.24 at @xmath163 to 2.54 at @xmath164 .
as @xmath49 remains comparable with the defect size , the regime in which the transverse trapping is determined by the long @xmath165 tail of the pin - vortex interaction is not quite realized .
this explains why the power index in the force@xmath166pin - density dependence is larger than suggested by the metastable - regime estimates which assume @xmath167 . to obtain further insight into the properties of trapped lines we show in fig .
[ fig - distrib ] examples of the distribution functions of trapping parameters for a fixed density @xmath168 and different system sizes .
one can see that these distributions are characterized by long exponential tails .
there is a noticeable probability to find segments with very large @xmath51 and @xmath50 .
note that the tails have opposite size effects for these parameters , the probability to find a large @xmath51 increases with @xmath139 while the probability to find a large @xmath50 _ decreases _ with @xmath139 .
the last trend is opposite to naive expectations . to characterize the long - range displacements of the line , we studied behavior of the structure factors , the fourier transforms of the displacement correlation functions , @xmath169 examples of these quantities are presented in fig .
[ fig - sqt0 ] ( left ) for two pin densities .
similar to local quantities , the long - range displacements are also strongly anisotropic .
such anisotropic scaling of the displacements is a general property of the driven lines in the critical regime independent of the pinning mechanism.@xcite for both components we clearly observe two regions of @xmath170 characterized by different power - law dependences @xmath171 . for the smallest @xmath170 s we found the roughness exponents @xmath172 and @xmath173 .
the value of @xmath174 for the transverse direction corresponds to the line displacements increasing as @xmath175 ^ 2\rangle\propto z^{2\zeta_t}$ ] at large @xmath143 . on the other hand
, the value of @xmath176 found for the longitudinal displacements implies that the assumed elastic approximation is not self - consistent in the thermodynamic limit , the average local stress @xmath177 increases with the line length @xmath126 as @xmath178 . in this case
the longitudinal line displacements grow quadratically @xmath175 ^ 2\rangle= c_l z^{2}$ ] with the coefficient increasing with the line length as @xmath179.@xcite this provides a natural explanation for the strong size dependence of the longitudinal pin - to - pin displacement @xmath51 in the lower plot of fig . [ fig - trap - param - np ] .
the negative size effects for the trapping lengths in the upper plot of the same figure can also be understood . growing local stress with increasing @xmath139 forces the line to travel longer distances in the longitudinal direction which increases the probability of finding a pin separated by a smaller distance in the @xmath143 direction .
the found exponents are slightly different from the values @xmath180 and @xmath181 obtained in ref . from the approximate functional renormalization group calculations .
however , a similar situation was found for elastic lines in a plane where the numerically computed index @xmath182 @xcite also exceed the predicted value @xmath183
we see that the transverse displacements somewhat reduce the exponent value in the 3d case .
since the small-@xmath170 exponents are expected to be universal , i.e. , independent of the pinning mechanism , our results suggest that the elastic description will also break down for the weak pinning case .
although we have considered the linear approximation for the elastic forces , this conclusion is expected to hold for the full non - quadratic energy of the deformed vortex line.@xcite at larger @xmath170 s we observe the regime where both components behave as @xmath184 giving the short - scale indices @xmath185 .
this corresponds to displacements induced by a short - range - correlated random force , and such behavior is actually similar to the static larkin regime for weak pinning .
however , the random forces in our case clearly have a very different origin .
the crossover between different regimes occurs at a wave vector that scales approximately as @xmath186 .
this allows us to approximately collapse the structure factors at different pin densities into a single curve using the scaling @xmath187 .
we found @xmath188 and @xmath189 .
these scaled dependencies are shown on the right - hand side of fig .
[ fig - sqt0 ] . scaling works better for the longitudinal structure factor .
this scaling is consistent with the identification of the trapping parameter @xmath158 as a geometric crossover length at @xmath190 , between a short - distance roughness regime with exponents @xmath185 and a large - distance universal regime with exponents @xmath191 and @xmath192 .
this behavior is again very similar to the crossover between the larkin regime and the random - manifold regime established for the case of weak collective pinning even though disorder is not weak in the present case .
we can also see from fig .
[ fig - sqt0](left ) that both structure factors increase with the pin density for all wave vectors .
this means that , in contrast to elemental pin - to - pin displacements @xmath193 plotted in fig .
[ fig - trap - param - np ] , the line displacements at fixed @xmath143 _ grow _ with increasing pin density .
this behavior can be easily understood .
the displacements at small distances are determined not only by the behavior of @xmath193 but also by the behavior of the trapping length @xmath142 .
all these parameters decrease with increasing pin density , meaning that smaller displacements occur on a smaller length scale .
the net increase of the line displacements at fixed @xmath143 coordinate with increasing @xmath150 is a consequence of a faster @xmath194 decrease than that of @xmath195 , as can be seen from fig .
[ fig - trap - param - np ] . in this section
we consider the influence of thermal noise on the dynamic response and configurations of vortex lines interacting with strong pins .
figure [ fig - ivtrapparamtn20nz2048 ] presents the temperature evolution of the velocity - force dependences and trapping parameters in the critical region for two very different pin densities , @xmath196 and @xmath197 .
we remind that we use the pinning energy of a single pin as the temperature unit . at finite temperatures
the critical force does not have an exact meaning , because the line velocity is finite at all forces due to the thermal creep .
nevertheless , we can introduce a typical force corresponding to the crossover between the flow and creep regimes , similar to the voltage criterion widely used in experiment .
we use the criterion @xmath198 for this force .
the first important observation is that , independently of criterion , the apparent critical force is quite strongly suppressed by the thermal noise .
for example , as we can see in fig .
[ fig - ivtrapparamtn20nz2048 ] , for the small pin density @xmath196 at the temperature only 5% of the pinning energy , the apparent critical force is already suppressed about fourfold .
as one can see from the trapping - parameter plots , the longitudinal pin - to - pin displacement @xmath51 in the critical region very rapidly decreases with the temperature while the transverse displacement @xmath49 slightly increases with the temperature . as a consequence , at some temperature they become of the same order .
the longitudinal displacement typically has a nonmonotonic dependence on the driving force and reaches a maximum at some force in the critical region ( the maximum - stress force ) .
the trapping length decreases with decreasing line velocity and its value at @xmath198 slightly increases with the temperature .
figure [ fig - fc - t](a ) shows the temperature dependences of the apparent critical forces for a wide range of pin densities .
we can see that these dependences are quite similar .
however , plots of the relative critical forces in fig .
[ fig - fc - t](b ) clearly show that thermal suppression weakens with increasing pin density .
this is consistent with the estimate for the temperature renormalization of the effective pin - breaking force , eq.([renormfp ] ) , due to the @xmath150 dependence under the logarithm . for illustration
, we also present the real - temperature scales on the top axes of the plots in fig .
[ fig - fc - t](a , b ) computed for spherical particles with @xmath199 nm and typical ybco parameters .
this scale , however , is very sensitive to the value of @xmath7 .
for example , the liquid nitrogen temperature @xmath200k corresponds to the reduced temperature @xmath201 in the plot . for particles with @xmath202 nm
the same real temperature would correspond to the much smaller value @xmath203 . as we found approximately @xmath204 at @xmath205 , according to the estimate of eq .
( [ crcurrmet_b ] ) we also expect the relation @xmath206 .
this means that , according to eq .
( [ renormfp ] ) , we expect the dependence @xmath207^{2/3}=1-(\beta_{dp } \tilde{t})^{2/3}$ ] with @xmath208 .
this is directly verified in fig.[fig - fc - t](c ) where we observe the approximate linear dependencies of @xmath207^{2/3}$ ] versus @xmath209 .
moreover , as shown in the inset , the coefficient @xmath210 found from the linear fits for different pin densities indeed has a logarithmic dependence on @xmath150 , @xmath211 .
these observations provide justification for our assumption that the reduction of the typical pin - breaking force is the main source of thermal suppression of the effective critical force .
we consider now the influence of thermal noise on the long - range behavior of the line displacements .
figures [ fig - sl - ul2-t](a , c ) present the evolution of the structure factor with increasing temperature for the pin density @xmath212 .
we can see that the slope of the small-@xmath170 dependences does not change indicating that the roughening indices @xmath213 and @xmath214 are temperature independent .
however , the coefficient is significantly reduced for the longitudinal displacements and is enlarged for the transverse displacements . correspondingly , as one can see from the line wandering plots shown in fig .
[ fig - sl - ul2-t](b , d ) , the components of displacements in the critical region have opposite tendencies : the longitudinal displacements decrease and the transverse displacements increase with increasing temperature .
as the longitudinal displacements dominate , the lines become more straight in the critical region .
another important observation is that the random - force regime @xmath215 at large @xmath170 is rapidly washed out by thermal noise for both components and is replaced by the isotropic fluctuational line wandering @xmath216 .
using the square - root fit of the pin - density dependence of critical force in fig .
[ fig - fc - np ] , we can restore the @xmath41 dependence of the critical force in real units @xmath217 this coincides with the estimate ( [ crcurrmet_b ] ) which is obtained assuming that the transverse trapping distance @xmath49 is on the order of the pin size @xmath7 . this does not contradict too much our numerical results , because our average values of @xmath49 only slightly exceed the pin size .
note that in our simulations the pin - breaking force @xmath218 is fixed by interaction with the pin while in real superconductors for large - size inclusions it is determined by the in - plane line energy , see eq .
( [ pinbreakfrc ] ) . substituting this estimate
, our result leads to the following estimate for the critical force @xmath219 where , assuming @xmath220 , we estimated @xmath221^{3/2}[\ln(l_t / s)]^{-1/2}$ ] . for preliminary comparison with experiment
we use results of the recent paper in which approximately spherical ( y - gd)@xmath222o@xmath223 particles with radii @xmath224 nm were embedded into the ybco films .
the typical concentration of particles was @xmath225@xmath119 . for estimates , we assume the temperature - dependent london penetration depth as @xmath226nm@xmath227 with @xmath228k and the anisotropy @xmath115 .
such parameters correspond to the reduced concentration @xmath229 . from the above zero - temperature result , we estimate @xmath230 and the critical current @xmath231 amp/@xmath232 which is close to the low - field experimental value at @xmath233k , @xmath234 amp/@xmath232 .
the reduced temperature can be evaluated as @xmath235 $ ] . at 55 k
this gives @xmath236 . according to fig .
[ fig - fc - t](b ) , at this temperature thermal suppression of the critical current is expected to be around 50% .
taking this factor and the temperature dependence of the parameters into account , we expect a critical current @xmath237 amp/@xmath232 , which again is quite close to the experimental value of @xmath238 amp/@xmath232 .
we can conclude that the our strong - pinning result for the critical current is in reasonable agreement with experiment .
our results show that the wandering of dynamically pinned lines is strongly anisotropic , displacements in the direction of motion are much larger than displacements in the transverse direction . in principle , experimentally this can be directly demonstrated using flux visualization techniques , such as a scanning squid or hall probes . with these techniques individual vortices
can only be resolved at small fields .
usually statically pinned vortices are visualized after cooling in fixed magnetic field .
nevertheless , it should also be possible to visualize shapes of the dynamically - pinned vortices near the boundary of the bean profile , which is formed when the magnetic field is applied after cooling of a superconducting sample in zero field .
we expect the vortex field profiles to be strongly elongated along the direction of motion .
the elongation length along the direction of motion is given by the longitudinal displacement @xmath51 at a distance on the order of the london penetration depth .
this length is expected to decrease with increasing temperature . in conclusion
, we developed a quantitative description of individual vortex lines pinned by an array of nanoparticles .
we found that the critical force grows roughly as the square root of the pin density .
this result is consistent with qualitative estimates assuming that for our pin - density range trapping events mostly occur as a result of direct collisions with pinning centers .
the apparent critical force is strongly suppressed by thermal noise .
the relative suppression reduces with increasing pin density .
the configurations of pinned lines are strongly anisotropic , displacements in the drive directions are much larger than those in the transverse direction .
the displacement anisotropy is rapidly reduced by thermal noise mostly due to the rapid reduction of the longitudinal displacements .
this leads to straightening of the lines in the critical region . analyzing the behavior of the structure factors at small wave vectors
, we found that the roughening index for the longitudinal displacements exceeds one .
this means that the local stresses in the critical region increase with the line length indicating a breakdown of the elastic description in the thermodynamic limit .
the behavior of the structure factor at large wave vectors for both directions is typical for the displacement induced by random force . at finite temperatures this random - force regime
is rapidly replaced by the thermal displacements .
the authors would like to thank j. r. thompson for very helpful discussions and for providing unpublished parameters of the films studied in ref . .
the authors also acknowledge useful discussions with d. k. christen , l. civale , and v. b. geshkenbein .
was supported by uchicago argonne , llc , operator of argonne national laboratory , a u.s .
department of energy office of science laboratory , operated under contract no .
de - ac02 - 06ch11357 .
this work was also supported by the center for emergent superconductivity , an energy frontier research center funded by the u.s .
department of energy , office of science , office of basic energy sciences under award number de - ac0298ch1088 . a.b.k acknowledges hospitality at argonne national laboratory and the support from cnea , conicet under grant no .
pip11220090100051 , and anpcyt under grant no .
pict2007886 .
j. gutirrez , a. llords , j. gzquez , m. gibert , n. rom , s. ricart , a. pomar , f. sandiumenge , n. mestres , t. puig , and x. obradors , nature materials * 6 * , 367 ( 2007 ) ; j. gutierrez , t. puig , and x. obradors , appl .
. lett . * 90 * , 162514 ( 2007 ) .
h. yamasaki , k. ohki , h. yamada , y. nakagawa and y. mawatari , supercond .
* 21 * , 125011 ( 2008 ) .
m. miura , b. maiorov , s. a. baily , n. haberkorn , j. o. willis , k. marken , t. izumi , y. shiohara , and l. civale , phys .
b * 83 * , 184519 ( 2011 )
. . polat , j. w. sinclair , y. l. zuev , j. r. thompson , d. k. christen , s. w. cook , d. kumar , yimin chen , and v. selvamanickam , phys .
b * 84 * , 024519 ( 2011 ) .
the concentration of ( y - gd)@xmath222o@xmath223 precipitates in the studied ( gd - y)ba@xmath222cu@xmath223o@xmath239 films was @xmath240@xmath119 and the low - field critical current for 2.8 @xmath241m - thick film was @xmath234 amp/@xmath242 at 5k and @xmath238 amp/@xmath242 at 55k ( private communication from j. r. thompson , c. cantoni , . polat , and d. k. christen ) .
j. l. macmanus - driscoll , s. r. foltyn , q. x. jia , h. wang , a. serquis , l. civale , b. maiorov , m. e. hawley , m. p. maley , and d. e. peterson , nature materials * 3 * , 439 ( 2004 ) ; b. maiorov , s. a. baily , h. zhou , o. ugurlu , j. a. kennison , p. c. dowden , t. g. holesinger , s. r. foltyn and l. civale , nature materials * 8 * , 398 ( 2009 ) . c. v. varanasi , p. n. barnes , and j. burke , supercond .
technol . *
20 * , 1071 ( 2007 ) ; c. v. varanasi , j. burke , l. brunke , h. wang , m. sumption , and p. n. barnes , j. appl . phys .
* 102 * , 063909 ( 2007 ) ; c.v .
varanasi , j. burke , and l. brunke , h. wang , j.h .
lee , and p.n .
barnes , j. mater .
res . , * 23 * , 3363 ( 2008 ) ; c. v. varanasi , j. burke , h. wang , j. h. lee , and p. n. barnes , appl . phys .
* 93 * , 092501 ( 2008 ) . c. j. van der beek , g. rizza , m. konczykowski , p. fertey , i. monnet , th .
klein , r. okazaki , m. ishikado , h. kito , a. iyo , h. eisaki , s. shamoto , m. e. tillman , s. l. budko , p. c. canfield , t. shibauchi , and y. matsuda , phys .
b * 81 * , 174517 ( 2010 ) ; c. j. van der beek , m. konczykowski , sh .
kasahara , t. terashima , r. okazaki , t. shibauchi , and y. matsuda , phys .
lett . * 105 * , 267002 ( 2010 ) .
e. w. j. straver , j. e. hoffman , o. m. auslaender , d. rugar , k. a. moler , appl .
* 93 * , 172514 ( 2008 ) ; o. m. auslaender , l. luan , e. w. j. straver , j. e. hoffman , n. c. koshnick , e. zeldov , d. a. bonn , r. liang , w. n. hardy , k. a. moler nature physics , * 5 * , 35(2009 ) ; c. reichhardt , nature physics * 5 * , 15 ( 2009 ) .
the cutoff lengths @xmath246 and @xmath247 depend on microscopic details .
typically , @xmath246 is determined by the length scale of line deformation in the @xmath143 direction , which in our case is given by the trapping length @xmath50 .
the lower bound for @xmath247 is given by the c - axis coherence length @xmath15 .
however , in layered superconductors more accurately @xmath247 can be estimated as the typical distance between the pancake vortices in the neighboring layers devided by the anisotropy factor @xmath13 . near depinning threshold , the pin - vortex potential as a function of the pinned - point displacement @xmath43 can be expanded as @xmath248 where @xmath249 and @xmath250 is the marginal equilibrium point . below threshold
this yields the barrier @xmath251(f_p - f)^{3/2}$ ] .
|
the pinning of vortex lines by an array of nanoparticles embedded inside superconductors has become the most efficient practical way to achieve high critical currents . in this situation
pinning occurs via trapping of the vortex - line segments and the critical current is determined by the typical length of the trapped segments . to verify analytical estimates and develop a quantitative description of strong pinning
, we numerically simulated isolated vortex lines driven through an array of nanoparticles .
we found that the critical force grows roughly as the square root of the pin density and it is strongly suppressed by thermal noise .
the configurations of pinned lines are strongly anisotropic , displacements in the drive directions are much larger than in the transverse direction .
moreover , we found that the roughening index for the longitudinal displacements exceeds one .
this indicates that the local stresses in the critical region increase with the total line length and the elastic description breaks down in the thermodynamic limit .
thermal noise reduces the anisotropy of displacements in the critical regions and straightens the lines .
|
spectroscopy is one of the most powerful methods in plasma diagnostics . in particular
, the analysis of the shape of spectral lines allows to determine the properties of the plasma such as temperature , density , and composition . for dense ,
strongly coupled plasmas , optical spectra have been investigated to infer the parameter values not only of laboratory , but also of astrophysical plasmas , see , e.g. refs .
@xcite and refs .
@xcite , respectively .
the shape of spectral lines is determined by different processes . besides the natural line width given by the finite lifetime of excited states ,
doppler broadening is related to the thermal motion of the emitters .
the influence of the surrounding medium becomes more important at increasing density and leads to pressure broadening .
a large effect is caused by charged particles in the plasma .
this stark broadening is determined by the distribution of ions as sources of the microfield , whereas the contribution of the electrons is usually treated in impact approximation @xcite .
the weisskopf radius can be used to distinguish between weak and strong collisions . however , also the neutral components in the surrounding medium contribute to line profile , i.e. as van der waals broadening .
a systematic approach to the shape of spectral lines in dense plasmas has been worked out on the basis of perturbation theory @xcite , and many - particle effects have been incorporated .
griem @xcite gives a review of different classical and semiclassical methods as well as their application to line shapes in different plasmas .
the _ unified _ theory @xcite was developed to describe the center as well as the wings of the lines . a quantum statistical approach based on the many - particle green functions method
was presented by gnter @xcite and has been developed systematically for charged perturbers .
special attention was payed to show the equivalence between this rigorous approach and the aforementioned classical and semiclassical concepts in the appropriate limits .
the quantum statistical approach has been applied to a broad range of systems , e.g. hydrogen plasmas @xcite , helium plasmas @xcite and plasmas of other elements containing h or he like ions . in most cases , a hot and dense plasma with a large fraction of free electrons
was considered . in such systems ,
the main contribution to the shift and width of a given spectral line , besides the natural line width and doppler broadening , comes from the ionic microfield , caused by ions surrounding the radiating atom or ion , collisions with free electrons , and interaction with collective excitations in the plasma ( plasmons ) .
theoretical results are in very good agreement with experiments carried out , e.g. by wilhein @xcite . in weakly ionized plasmas ,
the influence of neutral perturbers becomes dominant .
van der waals broadening has been investigated within second order perturbation theory .
based on the derivation of the van der waals potential between two hydrogen atoms in their ground state as given by london @xcite , a red shift of the spectral line follows as was shown by margenau @xcite .
a review of van der waals broadening is given by traving @xcite .
line broadening due to van der waals interaction has been investigated in detail , e.g. by walkup _
_ @xcite .
we show how to extend the quantum statistical approach , elaborated for the influence of charged particles on the line profile , to include also the interaction with neutral perturbers . in this way
, the extension of the van der waals broadening by accounting for many - particle effects such as dynamical screening or strong collisions is possible .
a unified description of the influence of the perturbing partially ionized plasma on the line profile of a radiating atom can be performed , treating charged particles and neutrals within the same formalism .
the theory will be compared to optical spectra obtained from hydrogen microdroplets exposed to strong femtosecond laser pulses .
so far , most of the experimental and theoretical work has concentrated on nanosized clusters @xcite .
studies on the light emission from these systems have focused on x rays and euv
radiation @xcite , motivated , e.g. by the search for novel light sources .
hydrogen as a target material has attracted considerable attention since nuclear fusion has been demonstrated as a result of coulomb explosion of dense deuterium clusters @xcite .
various efforts to produce micrometer sized hydrogen targets are reported , cf .
e.g. ref .
other studies on such large systems include , e.g. the measurement of x ray emission from kr clusters @xcite and from methanol microdroplets @xcite .
an expanding microplasma can be produced with interesting parameter values .
in particular , strongly coupled plasmas are obtained , where density and temperature are time - dependant .
time resolved experimental techniques have been demonstrated to measure transient plasma properties like the evolution of the plasma density @xcite . in the experiment on @xmath0 m sized droplets reported here
, lines of the hydrogen balmer series are observed as distinctive features in the optical emission spectrum .
the present paper will give a first interpretation of the recorded line shapes .
the analysis of the line profiles shows that van der waals broadening due to neutral perturbers is the predominant contribution .
values for the effective plasma parameters are deduced to interprete the observed line profiles , which differ for the different lines .
however , an equilibrium picture can not give an agreeable description of the measured balmer spectra .
a consistent dynamical picture of the expanding microplasma will be given by use of hydrodynamic simulations .
this enables us to interpret the observed spectra if considering the time stages where excited atomic states can exist . the work is organized as follows : in the second section we will briefly outline the method of thermodynamic green functions with emphasis on its application to the calculation of spectral line shapes .
the method will be applied to the case of a weakly ionized plasma , where the interaction among atoms is governed by the dipole - dipole term .
the margenau result appears in the case of ground state perturbers . in the third section , our results
will be compared to the balmer spectra obtained from laser produced hydrogen microplasmas .
parameter values for the effective temperature and density are inferred . in the fourth section , a hydrodynamical description of the expanding droplet after laser excitation is given .
a quantum statistical approach to the optical spectra of dense , strongly coupled systems can be given within linear response theory .
emission and absorption of radiation is related to the transverse dielectric function @xmath1 . in the optical region ,
the wavelength is large compared with atomic distances so that the long - wavelength limit @xmath2 can be considered . for the absorption coefficient we find @xmath3 @xmath4 is the refraction index , @xmath5 is the velocity of light in vacuum . for @xmath2 the transverse and longitudinal dielectric function @xmath6 become identical .
we consider in the following the longitudinal one .
the dielectric function for a charged particle system can be evaluated in a rigorous way using quantum statistical methods . according to @xmath7 ,
it is linked to the polarization function @xmath8 , for which a systematic perturbation expansion can be given . a green function approach to the polarization function
is outlined in app .
[ app : pi_2 ] . using feynman diagrams and partial summations ,
appropriate approximations can be found which account for different microscopic processes contributing to the behavior of the charged particle system . in particular , for partially ionized plasmas , a cluster decomposition @xmath9 can be performed to give the contribution of free carriers as well as of bound states in a systematic way @xcite .
@xmath10 collects the single - particle contribution . in lowest order , neglecting collisions with other particles , the well - known rpa polarization function @xmath11 is obtained . here ,
@xmath12^{-1}$ ] is the fermi distribution function of particles of species @xmath5 ( including spin ) with charge @xmath13 and mass @xmath14 , @xmath15 is the single - particle kinetic energy , and @xmath16 the chemical potential . within this lowest order approximation ,
single - particle excitations as well as collective plasmon modes are described .
the limit @xmath17 has to be performed after the summation over momenta .
the rpa contribution @xmath18 is improved if interactions with other particles beyond mean field are included , see app .
[ app : pi_2 ] .
medium effects , such as collisions , dynamical screening and exchange , enter the polarization function via the single - particle self - energies and the vertex function .
the vertex function describes the in - medium coupling of particles to the radiation field .
it has to be taken in the same approximation as the self - energy as dictated by consistency constraints , i.e. ward identities @xcite .
going beyond the mean - field ( hartree - fock ) approximation , collisions can be considered in lowest order born approximation .
higher orders lead to dynamical screening and t - matrix expressions for strong collisions .
this way , the continuum of optical spectra is described , in particular inverse bremsstrahlung @xcite .
however , line spectra are missing in @xmath10 .
the next term in the cluster decomposition is @xmath19 .
it is given by the convolution of two atomic ( two - particle ) propagators and describes also transitions between bound states , i.e. the line spectrum @xcite . in the lowest order , any interaction with further particles are neglected .
similar to @xmath20 given above , replacing single - particle propagators by atomic propagators one obtains @xmath21 the factor 4 accounts for spin degeneration , @xmath22^{-1}$ ] is the bose distribution function of electron - ion bound states at @xmath23 , with @xmath24 denoting the internal quantum numbers , @xmath25 the center of mass momentum , and @xmath26 .
the unperturbed matrix elements @xmath27 are given by @xmath28\\ & \simeq e\left [ \delta_{\alpha_1\alpha_2}- \int\!\mathrm{d}^3\vec r\,\psi_{\alpha_1}^*(\vec r)\,\mathrm{e}^{i\vec k\cdot\vec r}\,\psi_{\alpha_2}(\vec r)\right]~. \label{eqn : matrixelementm_r}\end{aligned}\ ] ] @xmath29 and @xmath30 are atomic wave - functions in relative momentum representation and coordinate representation , respectively .
( [ eqn : matrixelementm_r ] ) results for @xmath31 .
in contrast to a simple chemical picture where a mixture of free charge carriers and atoms is considered , scattering states are included , and double counting of diagrams has to be avoided . in this approximation , one obtains the unperturbed line - spectrum of isolated atoms including the doppler profile due to thermal motion .
as before , medium effects are described by taking the self - energy of the constituents and the vertex into account , which now are given on the two - particle level , see app .
[ app : pi_2 ] . in the simplest approximation ,
the dynamically screened interaction can be considered similar to the @xmath32 approximation for the single - particle propagator @xcite .
this dynamically screened interaction contains the polarization function , which once more can be decomposed into the contribution of single - particle states , two - particle states and higher cluster states . in this way
, we obtain the dynamically screened born approximation for collisions with free particles or composed clusters of the partially ionized plasma . in the case of dense and strongly ionized systems ,
the main contribution to the self - energy comes from the interaction of the radiating atom with charged particles , i.e. electrons and ions .
the influence of free particles is given by the rpa - polarization function , i.e. the first term in the cluster decomposition .
this interaction can be inserted into the self - energy , but the vertex - function has to be taken in the same approximation . reducing the screened interaction to its one - loop approximation , which amounts to the second born approximation with respect to the statically screened coulomb potential ,
the so - called impact approximation is obtained .
systematic improvements by taking higher order contributions into account lead to the dynamical screening of the coulomb potential in the case of weak collisions and the t - matrix approach in the case of strong collisions .
in particular , the contribution of collisions with electrons to the line profile in dense plasmas is treated this way , for a review see gnter @xcite .
the following expression for the line - profile @xmath33 as a function of the frequency displacement @xmath34 is obtained @xcite , @xmath35^ * \frac{\mathrm{e}^{-\beta \hbar(\omega_{if}+\delta\omega)}\left(\omega_{if } + \delta\omega \right)^4}{8\pi^3c^3 } \int\frac{\mathrm{d}^3 p}{(2\pi)^3}\,\mathrm{e}^{-\beta\frac { \hbar^2 p^2}{2m}}\,\int_0^\infty\mathrm{d}\bar\beta\,w_\rho(\bar\beta ) \\
\times \langle\left . \,i\,\right|\langle\left . \,f\,\right|\left
[ \hbar \delta\omega-\frac{\hbar^2 \vec p\cdot\vec k}{m}-\frac{\hbar^2 k^2}{2m}-\mathrm{re}\,\left\ { \sigma_i(\delta\omega,\bar\beta)-\sigma_f \right\}+i\mathrm{im}\,\left\ { \sigma_i(\delta\omega,\bar\beta)+\sigma_f \right\}+i\gamma^v \right]^{-1}\left|\,f'\,\right.\rangle\left|\,i'\,\right.\rangle~. \label{eqn : lineprofile_sigma}\end{gathered}\ ] ] here , @xmath36 denote the electronic contribution to the self - energy of the initial or final state , respectively , @xmath37 the vertex contribution . improvements of the born approximation by accounting for dynamical screening and strong collisions , have extensively been discussed in the literature @xcite . due to their large masses ,
ions are much slower than electrons . in the adiabatic limit
, the ions may be considered as a static distribution of charged particles during the process of emission .
considering different ionic configurations in the plasma , the distribution @xmath38 of the so - called microfield @xmath39 , @xmath40 , ( @xmath41 is the density of ions with charge z in the system ) is introduced , and the averaging leads to the stark profile as given in eq .
( [ eqn : lineprofile_sigma ] ) .
also in the case of ionic contributions , the systematic many - particle treatment leads to further improvements such as the dynamic microfield . in the next step
, one has to take into account the interaction between the radiator and neutral particles ( atoms ) , as well .
this can be performed along the same lines as for the interaction with charged particles , discussed before .
the bound state contribution is considered in the cluster decomposition of the polarization function , expanding the dynamically screened interaction which enters the self - energy , see app .
[ app : pi_2 ] .
in particular we will consider the two - particle self - energy describing the influence of the medium on the radiator , where the bound state component is taken into account in lowest approximation .
we consider the second order born approximation with respect to the unscreened coulomb potential to describe the interaction between neutrals .
the detailed calculation is given in the next section .
again , improvements can be obtained systematically . by calculating the dynamically screened interaction with the polarization function up to the second term in the cluster decomposition , the interaction with free and bound particles as well as with collective excitations could be included in a consistent way .
strong collisions can be described via the four - particle t - matrix .
in addition to the self - energy term , also the vertex is modified if bound states are included .
the calculation of the interaction @xmath42 between two atoms at distance @xmath43 in state @xmath44 and @xmath45 is carried out in app .
[ app : bound ] .
we obtain @xmath46 \psi_{\alpha_2}^*(\vec r_2)\psi_{\alpha_4}(\vec r_2 ) \right|^2~. \label{eqn : v_vdw_r}\end{aligned}\ ] ] here , @xmath47 denote quantum numbers of the intermediate states .
only elastic processes are considered where the internal quantum numbers for the initial state and the final state are identical . due to the ion s heavy mass
, we neglect the contribution of kinetic energy in the intermediate propagator in comparison with the excitation energies . besides @xmath43 which is the distance between the nuclei of the interacting atoms , @xmath48 and @xmath49 are vectors of position of the two electrons of atom 1 and atom 2 , respectively . this result ( [ eqn : v_vdw_r ] )
coincides with results from second order perturbation theory . at large distances
@xmath50 we can perform the dipole approximation using @xmath51 $ ] for @xmath52 .
we obtain the van der waals potential , @xmath53
\psi_{\alpha_2}^*(\vec r_2)\psi_{\alpha_4}(\vec r_2 ) \right|^2~. \label{eqn : v_vdw_r_dip}\end{aligned}\ ] ] for hydrogen - like atoms , the sum in eq . (
[ eqn : v_vdw_r_dip ] ) has been performed by eisenschitz and london @xcite .
their result for the interatomic potential between two hydrogen atoms in ground state ( @xmath54 ) is given by @xmath55 the constant @xmath56 was calculated as @xmath57 , where @xmath58 is the rydberg energy and @xmath59 is the bohr radius . for the interaction between a radiator in an excited state @xmath60 and a perturber in ground state , @xmath61 , the following expression for the interaction strength @xmath62 , can be derived , as shown in app .
[ app : interaction_strength ] , @xmath63~. \label{eqn : int_strength}\ ] ] note that there is no exchange contribution in eq .
( [ eqn : v_vdw_r ] ) , since the distance @xmath43 is assumed to be large compared with the bohr radius . at shorter distances ,
the higher order terms as well as exchange terms have to be taken into account which will diminuish the van der waals behavior at small distances and possibly , in the case of electrons in the triplet state , result in a repulsion due to the exchange contribution . within a general approach ,
the contribution of the dense medium to the spectral line shape is given by the self - energy and vertex contribution .
the first order of the interaction gives the hartree - fock mean field , in second order the impact approximation is obtained where collisions are described in born approximation . within this concept ,
the van der waals contribution to the self - energy is @xmath64 which in coordinate space reads @xmath65 here , @xmath66 is the density of pertubing atoms in state @xmath67 . in the following ,
we consider the perturbing atoms in their ground state @xmath68 @xcite and only the radiator in an excited state @xmath69 . as already discussed in connection with the ionic contribution to the spectral line shapes , because of their large masses the atomic motion is slow , and we can apply the microfield concept of strong interaction with a given field distribution . the intensity distribution of a given spectral line , due interaction between the radiator in initial and final states @xmath70 and @xmath71 , respectively and the perturber in its ground state @xmath72 is obtained by performing an averaging procedure over the distribution of perturbing atoms , @xmath73 here , @xmath74 is the probability density for having atom 1 in the volume element @xmath75 at @xmath76 , atom 2 in the volume element @xmath77 at @xmath78 , etc .
the evaluation of the sum ( [ eqn : profile_at - at ] ) in the case of statistically independent atoms , i.e. @xmath79 , is shown in app . [ app : vdw - profile ] .
the result for the dipole limit of the interaction eq .
( [ eqn : vdwpot_00 ] ) is @xcite , @xmath80^{-3/2}\mathrm{e}^{\pi\lambda_{if}^2/\delta\omega}~ , \label{eqn : profile_london}\ ] ] with @xmath81 .
this is known as the margenau profile .
note that @xmath82 is to be taken at negative values , i.e. the van der waals interaction shifts the spectral line to smaller energies ( red shift ) .
the maximum of the intensity distribution ( [ eqn : profile_london ] ) is located at @xmath83 with @xmath84 and @xmath85 the number density of ( ground state ) hydrogen atoms in the system .
the full width of half maximum ( fwhm ) of the line is found as @xmath86 , see also refs .
@xcite . the margenau profile eq .
( [ eqn : v_vdw_r_dip ] ) will be used in order to analyse the hydrogen balmer spectra obtained from laser produced microplasmas .
these experiments will be described in the following section .
hydrogen droplets are produced by expansion of pre - cooled h@xmath87 gas at temperatures between 17 and 22k through a 20@xmath0 m orifice into a vacuum chamber using a backing pressure of about 8bar . at these source conditions ,
h@xmath87 droplets of about @xmath88 m in diameter are emitted from the nozzle at a density of @xmath89 .
roughly 10 mm behind the nozzle the droplets are irradiated by intense ir laser pulses generated by a ti - sapphire laser system with a wavelength of @xmath90 and a pulse energy of @xmath91 .
the beam is focussed to a spot with a beam waist of @xmath92 .
the pulse width @xmath93 is tuned in the range of @xmath94 , corresponding to laser peak intensities between @xmath95 and @xmath96 . when the droplets are exposed to the strong laser field , optical emission from the hydrogen microplasmas is observed .
the light is analyzed by a 0.34 m grating spectrometer ( 200 lines / mm ) and monitored with a ccd camera , providing a spectral dispersion of 0.185 nm per pixel .
the entrance slit width was optimized for a high signal when integrating 90000 shots at 1 khz repetition rate .
the resulting spectral resolution of the spectrometer is @xmath97 .
two regions of the visible spectral range are analyzed in more detail , i.e. ( a ) @xmath98 nm , and ( b ) @xmath99 nm .
[ fig : h_spectra ] shows representative emission spectra from the laser excited hydrogen microplasmas .
elastic rayleigh scattering of the incident radiation produces large enhancement of the measured spectra towards the excitation wavelength @xmath100 . on top of the background signal , h - balmer spectral lines are clearly identified , i.e. h@xmath101 at @xmath102 , h@xmath103 at @xmath104 , and h@xmath105 at @xmath106 . in order to quantitatively analyze the balmer line signals , one has to subtract the background signal .
therefore , the neighbouring spectral ranges of each line , i.e. 640 .. 655 nm and 670 .. 685 nm for the h@xmath101 line , 460 .. 467 nm and 472 .. 480 nm for h@xmath103 as well as 430 .. 433 nm and 438 .. 445 nm for h@xmath105 are fitted by an exponential function , which is then subtracted from the data set . as an example , figs .
[ fig:1000fs_h_a_margenaufit ] ( a ) and ( b ) show the background subtracted spectra for a pulse width of @xmath107 fs ( dots ) .
the solid curves are the margenau profiles fitted to the data ; they will be discussed in more detail in sec .
[ subsec : vdw - broadening ] . by comparing integrated intensities from different spectral lines ,
one can determine the effective temperature in the system .
given the ratio of wavelength - integrated intensities @xmath108 of two spectral lines of the same atomic species and the same ionization stage at central wavelengths @xmath109 and @xmath110 , and assuming again local thermal equilibrium , the effective temperature is @xcite @xmath111}~. \label{eqn : teff}\ ] ] @xmath112 , @xmath113 are the angular momentum averaged einstein coefficients for the transition between levels corresponding to the spectral line 1 and 2 , respectively and @xmath114 and @xmath115 are the degeneracy factors of the excited level ( principal quantum number @xmath116 ) of each transition , i.e. @xmath117 . @xmath118 and @xmath119 are the energies of the excited level for each transition . tab . [
tab : teff ] lists effective temperatures for each pulse width .
as can be seen , @xmath120 is dependent on @xmath93 .
one obtains different results from the two ratios @xmath121 and @xmath122 . for small @xmath93
, @xmath123 rises with increasing @xmath93 . at @xmath124
, @xmath125 saturates near @xmath126 . on the other hand
, @xmath127 gives values around @xmath128 .
.effective line temperatures according to eq .
( [ eqn : teff ] ) versus pulse length .
the first set of figures was obtained from the short wavelength spectrum , shown in fig .
[ fig : h_spectra](a ) , the second set results from the data shown in fig . [ fig : h_spectra](b ) . [ cols="^,^,^,^,^,^,^,^,^,^,^,^",options="header " , ] [ tab : teff ] these effective temperatures have to be analyzed within a microscopic description of the expanding plasma source .
self - absorption should be considered so that the intensities occurring in eq .
( [ eqn : teff ] ) are modified when propagating through the microplasma .
furthermore , different values @xmath120 from the intensities of h@xmath101 , h@xmath103 and h@xmath105 indicates that the system is in a generic non - equilibrium state , which can not be described by a single temperature .
despite the measured spectra are time integrated , we have to assume a dynamic evolution of the plasma parameters temperature and density over the course of the expansion of the excited hydrogen droplets . within a hydrodynamical description given below in sec .
[ subsec : hydro ] , local thermal equilibrium with plasma parameters depending on space and time are introduced . during the expansion process
, radiation can only be produced if the upper level of the transition in question is existent in the system so that it can be occupied . from simple geometric and energetic arguments
it is clear , that @xmath129 levels are not present in hydrogen at liquid densities .
this point will be discussed in more detail in sec .
[ subsec : levels ] .
the continuum background stems from free - bound and free - free ( bremsstrahlung ) transitions .
the spectral behavior can also be used to estimate effective temperatures which should be analyzed within a dynamical expansion model . since
more experimental work is necessary to separate the continuum background radiation , we will consider only the line spectra in this work . to analyze the line profiles ,
one first has to consider the effect of the broadening due to the finite resolution of the experimental setup .
the true line profile is obtained after deconvolution with the detector function . in a separate experiment
the spectral resolution was measured as @xmath97 .
this is consistent with the steep increase of intensity on the blue wing of the unperturbed spectral line .
-line for @xmath107 fs pulse length .
the raw data ( cf .
inset ) have been summed over different vertical ranges .
the solid line marks the central region of the laser focus ( framed part in the inset ) , the dashed curve corresponds to the data from the peripheric region . ] on the long wavelength wing , all spectral lines show a significant broadenening . in fig .
[ fig : h_1000fs_2cuts ] the spectral line shape is analyzed as a function of the distance of the origin of radiation from the plasma s center .
this is achieved by integration of the ccd - data over different ranges in the vertical direction , cf .
the inset in fig .
[ fig : h_1000fs_2cuts ] .
the corresponding spectra for the central region of the laser focus ( framed part in the inset ) and for the peripheric region , about 100 @xmath0 m beside the focus are shown .
after background subtraction , the spectra are normalized to their maximum value . in this way
it can be seen , that the width of the spectral line depends significantly on the spatial origin of the radiation and decreases with increasing distance from the center .
this behaviour indicates that the asymmetric broadening of the spectral lines is related to the properties of the microplasma and should be described as a density effect .
pressure broadening of spectral lines is caused by charged as well as by neutral perturbers . as discussed in sec .
[ sec : theory_lineshapes ] , free electrons are treated in impact approximation leading to a symmetric lorentzian line profile . also the ionic microfield contributes to a voigt profile with symmetric broadening on both the red as well as the blue wing ( linear stark effect ) .
both effects are smaller than the spectral resolution of 1 nm , as can be seen in the spectra , e.g. fig . [ fig : h_1000fs_2cuts ] .
thus , we conclude that the influence of free charged particles is not clearly identified so that the free electron density is below @xmath130 which at @xmath131 = 1 ev gives for h@xmath101 the fwhm of 1 nm . the asymmetric red shift can be described by the interaction with neutral perturbers
. we will now compare the experimental data to the line - shape due to interaction with neutral perturbers , as outlined in sec .
[ sec : bound - bound_born ] . in fig .
[ fig:1000fs_h_a_margenaufit ] , the measured spectra in the vicinity of the h@xmath101 line are compared to the emission profile as given by margenau , eq .
( [ eqn : profile_london ] ) , convoluted with the detectors resolution function @xcite . in the case of the h@xmath101-line
, the best fit was obtained using @xmath132 for the density of neutrals . in the case of h@xmath103 ,
the density @xmath133 gives the minimum @xmath134 .
as for the effective temperatures , also the different values for the effective density contradict an equilibrium picture . a dynamical description of the microplasma s evolution is needed .
this task will be accomplished in the next section by means of hydrodynamical simulations .
in the previous section the effective temperature and density of the microplasma were determined by analysis of the experimental spectra . for both quantities ,
different values have been obtained , depending on the external parameters , e.g. the pulse width of the laser , and on the wavelengths considered in the analysis .
this strongly indicates that the plasma parameters are time - dependant quantities , which have to be modeled , e.g. by hydrodynamic simulations to improve the description of the microplasma and the interpretation of the experimental data .
hydrodynamic simulations are a versatile tool to infer the dynamics of a strongly coupled many - particle system . in the case of plasmas
, hydrocodes have been successfully applied to study these systems under the influence of strong external fields @xcite as well as the relaxation of a plasma in an excited state into equilibrium . here , we will study the hydrodynamic expansion of the excited , i.e. heated h microdroplet after times which are long compared to the pulse length of the laser .
simulations are performed using the hydrocode multi2002 @xcite . as initial conditions ,
we assume a homogenous density profile @xmath135 and also homogeneous temperature distribution . assuming 50 % absorption of the laser energy given above by the droplet @xcite and accounting for the ratio of the droplet s cross - section to the focal spotsize , i.e. ( 10 @xmath0m/40 @xmath0m)@xmath136 as well as for the energy needed to break up the molecular bounds , i.e. 2.2 ev per atom , we obtain @xmath137 as the initial temperature . and @xmath138 .
the broad solid and dashed curves give the conditions after 0.7 ns and 0.9 ns . at this time
, the central density has reached the density which has been obtained from the fit of the margenau profile to the h@xmath101 and the h@xmath103 spectral line , respectively for @xmath139 pulselength .
the corresponding temperatures at these times are close to the effective temperatures @xmath140 and @xmath141 obtained from the spectra , see tab . [ tab : teff ] .
, title="fig : " ] and @xmath138 .
the broad solid and dashed curves give the conditions after 0.7 ns and 0.9 ns . at this time
, the central density has reached the density which has been obtained from the fit of the margenau profile to the h@xmath101 and the h@xmath103 spectral line , respectively for @xmath139 pulselength .
the corresponding temperatures at these times are close to the effective temperatures @xmath140 and @xmath141 obtained from the spectra , see tab . [ tab : teff ] .
, title="fig : " ] fig .
[ fig : dens - temp_profile ] shows the number density and temperature profiles .
the uppermost curve corresponds to @xmath142 , while the last curve is taken at @xmath143 . both parameters @xmath144 and
@xmath145 show a fast decline in the center of the droplet .
the density decreases by a factor of 10 every 300 ps .
secondly , all profiles are nearly constant over several tens of micrometers starting from the center and decrease sharply when approaching the rim of the droplet . therefore , the central values @xmath146 and @xmath147 can be taken as representative values for the whole droplet .
the temporal evolution of the central number density and temperature is shown in fig .
[ fig : dens - temp_evolution ] . as a function of time obtained from saha equation ( bold solid curve ) .
the narrow horizontal line gives the ionization after 0.7 ns , i.e. at the time , when the central density has decreased to @xmath148 , which has been obtained as best - fit parameter with respect to the margenau profile of the h@xmath101 spectral line . ]
the hydrodynamical description gives the time evolution of the excited droplet .
the assumption of local thermal equilibrium may be justified on the time scale of ps for the thermalization of kinetic energies . if assuming also local ionization equilibrium , the composition during the time evolution can be calculated .
the temporal behaviour of the ionization in the droplet s center is plotted in fig .
[ fig : saha ] . for the initial conditions
@xmath138 and @xmath89 , we have an ionization degree @xmath149 .
after @xmath150 , the density and temperature have decreased so far , that @xmath151 drops below @xmath152 .
this corresponds to a concentration of free electrons of roughly @xmath153 . at this density , and
@xmath154 , the stark effect leads to a broadening of the h@xmath101 line of 1 nm , as was discussed in sec . [ subsec : vdw - broadening ] .
thus , at times larger than 0.7 ns , stark broadening does not give a notable contribution to the width of the spectral lines .
this analysis has shown , that the emission of balmer line radiation occurs at comparatively late times of the expansion , when the density has decreased to a certain level .
this is due to the fact , that in a dense system , excited levels are not well defined due to interaction with neighbouring atoms .
both wavefunctions and atomic potentials are disturbed . in the following section we will analyse the question
, which density has to be established in the system , so that excited atomic levels are defined and the corresponding radiative transition may occur . in a dense system ,
the potential energy of an electron in a given atom is modified by the medium @xcite .
thus , energy eigenvalues and wave functions are changed due to screening by free charge carriers . in particular , it is well - known that bound states merge with the continuum of scattering states at high densities so that the electrons are no longer bound to a special ion , but move relatively free in the plasma as denoted by a transition from dielectric to metallic behavior @xcite .
considering excited states , the dissolution occurs already at lower densities so that these excited states can not any longer be occupied to produce a line spectrum .
we focus here to the influence of bound states in the medium to analyze at which densities the excited states are dissolved into the continuum so that no line spectra are formed . in a first approximation , we calculate the modification of the potential in the schrdinger equation for the hydrogen atom due to the perturbing atoms in the surrounding medium .
only ground state atoms are considered , and the correlations between the electrons in the perturbing ground state with the radiating electron in the excited state are neglected . solving the poisson equation ,
the effective potential is obtained as superposition of the coulomb potential and the potential of the neighboured atoms , which leads to a lowering of the coulomb potential by the amount @xmath155 here , @xmath156 is the position of the perturbing atom .
the lowering of the coulomb potential has two effects : on the one hand , the energy level is shifted to lower energies , on the other hand , at a certain critical distance , the threshold of potential energy between neighboured atoms crosses the shifted electrons energy level , and the bound state is dissolved into the continuum .
this is the case , if the distance between neighbours comes into the range of the extension of the electrons wave - function .
we assume a closely packed configuration where the radiating atom is surrounded by 12 perturbing next neighbours .
the shift of the energy levels with principal quantum numbers @xmath157 = 3 and 4 has been evaluated in first order of perturbation theory and can be given in an analytical form .
results for the shift as a function of the interatomic distance are given in fig .
[ fig : potential_shifts](a ) . in fig
[ fig : potential_shifts](b ) we give the potential energy @xmath158 in the direction to a next neighbour for two different distances , together with the corresponding shift of the energy levels with @xmath157 = 3 and @xmath159 .
as can be seen , the bound state disappears at a critical distance @xmath160 where the threshold energy becomes lower than the binding energy , and the electron is no longer bound to the central ion , but escapes the effective potential and moves relatively freely within the cluster . from the critical distance @xmath160
, one can then infer the `` critical '' density @xmath161 that has to be established in the expanding system , before the upper level @xmath157 of the considered transition can exist , i.e. before the corresponding spectral line can occur . in this way
, we obtained for the second excited level , relevant for the h@xmath101 transition , a `` critical '' density of @xmath162 .
this value for @xmath163 is by a factor of two larger than the value @xmath164 obtained from the fit of the @xmath165 line to the line shape due to van der waals interaction , cf .
[ fig:1000fs_h_a_margenaufit](a ) .
it has to emphasized in this context , that these are exploratory calculations to estimate the region of density where well - defined excited energy levels of the radiating atom can exist .
a more detailed calculation should include electronic correlations between radiator and perturber which have been neglected in the calculation of the perturbing potential .
the account for correlation and exchange effects will modify the potential and the critical density . in addition , we considered only a perturbing atom at mean distance , neglecting any fluctuations in configurations . on the other hand ,
the experimental spectra are time integrated measurements and hence the density has to be interpreted as a mean value , averaged over the whole exposure time . at the moment
the radiation starts , the density might in fact be larger than the value inferred from the measured spectra . for @xmath166 ,
the critical density is obtained as @xmath167 , which also exceeds the fit parameter for the h@xmath103 line ( @xmath133 ) for the same reasons as given above for the case of the h@xmath101 line ( cf .
[ fig:1000fs_h_a_margenaufit](b ) ) . in conclusion ,
the experimental balmer spectra are not characteristic for the first stage of the laser excitation of the cluster , because the density of hydrogen in the condensed state is too high to form well - defined excited atomic levels with the corresponding principal quantum numbers .
such levels appear only during the process of expansion and may be used as a signal to infer the state of the microplasma at the corresponding time stage .
as discussed in the previous section , the observed balmer spectra can not be interpreted within an equilibrium picture of the laser produced microplasma .
we could not infer consistent values of plasma parameters for temperature and density .
a consistent description is only possible if the time evolution of the expanding microplasma is considered as a non - equilibrium process .
we follow the dynamical expansion of the laser excited hydrogen droplet as given by the hydrodynamical calculation .
first , the time evolution of the density is considered , with account to the effective densities derived from the measured line profiles .
thereby , one can determine the time at which the h@xmath101 line appears , i.e. where the inferred density @xmath164 is reached .
starting with the excitation temperature @xmath168 ev after the short - pulse laser excitation , this density is established after 0.7 ns .
the corresponding temperature obtained from the hydrocode multi is @xmath169 ev and is in reasonable agreement with the estimate for @xmath125 given in sec .
[ subsec : vdw - broadening ] .
thus , the h@xmath101 line profile reflects the state of the expanding microplasma after 0.7 ns . in the case of the h@xmath103 we observe the following : from the van der waals fit to the data , cf .
[ fig:1000fs_h_a_margenaufit](b ) , @xmath170 was obtained , which is reached in the hydrodynamic simulation after @xmath171 .
the temperature is @xmath172 at that moment , which is close to the value @xmath128 obtained for @xmath127 , cf .
[ tab : teff ] .
h@xmath103 thus gives us information about the state of the droplet at @xmath171 .
second , as for the empirically determined plasma parameters , the critical densities for h@xmath101 and h@xmath103 as discussed in sec .
[ subsec : levels ] , have to be compared to the results of the hydrodynamic simulation . the central density of the droplet decreases to the critical density for h@xmath101 to appear , @xmath173 after 0.45 ns .
after the same time , the temperature at the center of the droplet has decreased to 1.3 ev , which is in good agreement with the value 1 ev inferred from @xmath125 .
the critical density for h@xmath103 , @xmath174 is reached after 0.8 ns .
the central temperature is 0.5 ev at that moment , which again coincides well with the value 0.5 ev as obtained from @xmath127 , cf
[ tab : teff ] .
this analysis shows , that the droplet in fact undergoes a complex dynamical evolution .
using hydrocodes to model the droplets history , we can understand the experimental observations .
the initial temperature of @xmath175 ev is consistent with the observed temperatures and densities established at later times of the evolution .
the line radiation obviously stems from relatively late times . what happens before line radiation appears ?
at the initial conditions of @xmath176 and @xmath138 , the ionization of the droplet was calculated as @xmath177 , solving saha s equation , cf .
[ fig : saha ] .
the plasma emission is thus dominated by continuum radiation ( free - free and free - bound transitions ) .
however , since the detector integrates over the whole evolution of the plasma , only a nearly constant background remains in the spectrum from this early stage of the evolution .
as the ionization decreases , bound states begin to form and line radiation occurs .
however , for a given spectral line , both upper and lower level of the corresponding transition have to be well defined . to this end
, the density has to fall below a certain value , in order to allow for the excited energy level to appear underneath the effective potential in the dense medium . considering the time scales for the expansion of the cluster of the order @xmath178 according to the hydrodynamical calculations
, the local thermodynamic equilibrium can be assumed , and the life time of the excited states of the order @xmath179 is short compared with the evolution of the microplasma . in a more rigorous approach , the variation of temperature and density in space and time should be accounted for to synthesize the spectra .
we found that the visible spectra can be used to get signatures for a time - resolved picture of an expanding microplasma .
the line profiles allow for the determination of the microplasma s temperature and density during its hydrodynamic expansion .
we considered an energy deposition in the liquid hydrogen droplet which is sufficiently weak so that in the expanded phase most of the electrons are found in atomic bound states .
thus , the spectral line profiles are determined by van der waals broadening .
we presented a general quantum statistical approach to line profiles which allows for the unified description of charged and neutral perturbers , including density effects such as dynamical screening or strong collisions .
although the spectra are time integrated measurements , we can use the balmer lines as signatures for the expanding microplasma at definite values of density .
a consistent scenario for the emission of balmer lines has been given . to follow the time evolution of an excited droplet more directly , time resolved spectra have to be analyzed .
possibly , this can be achieved with pump - probe experiments and/or stimulated emission measurements .
this work has been supported by the deutsche forschungsgesellschaft ( dfg ) through the collaborative research center ( sfb ) 652 .
tl acknowledges financial support from the dfg under grant no .
la 1431/2 - 1 .
the general expression for the single state contribution @xmath180 to the polarization function has the form @xmath181 the full single - particle propagator contains the self - energy @xmath182 , @xmath183 the vertex @xmath184 describes the coupling to the electromagnetic field and can also be expressed in terms of an effective interaction kernel .
both quantities , self - energy @xmath185 and vertex @xmath184 , have to be approximated in a consistent way . for the self - energy one has the @xmath186 approximation , given by the diagram @xmath187 which is a self - consistent equation for the full propagator , the full vertex @xmath184 , as well as for the screened interaction potential @xmath188 , @xmath189 while the vertex function @xmath184 is the solution of the bethe - salpeter equation , given in diagrammatic form by @xmath190 solving eq .
( [ eqn : dse_full ] ) and eq .
( [ eqn : vertex_bse ] ) simultaneously is a formidable task .
considerable simplification of the problem is obtained by replacing the full vertex @xmath184 in eq .
( [ eqn : dse_full ] ) by the bare vertex @xmath191 .
this is the so - called @xmath32 approximation , @xmath192 the @xmath32 approximation is well studied in condensed matter physics .
it also describes bremsstrahlung an can be further improved accounting for plasma effects , see ref .
@xcite .
the two - particle contribution @xmath193 to the polarization function is described in analogy to eq .
( [ eqn : pi1_diagram ] ) , @xmath194{eps / feyn_pi_2_ggv.eps}}~ , \label{eqn : pi_2_diagram}\ ] ] where we have the full two - particle propagator @xmath195 with the two - particle self - energy @xmath196 , @xmath197 and the vertex follows from an equation in analogy to eq . ( [ eqn : vertex_bse ] ) @xmath198 for the polarization function which defines the screened interaction used in the two - particle self - energy eq .
( [ eqn : sigma_2_gw ] ) , we perform the cluster decomposition as outlined in sec .
[ sec : theory_lineshapes ] . in the impact - approximation
, we replace the full screened interaction by the first non - ideal term in the iteration of eq .
( [ eqn : vscreened_diagrams ] ) .
diagrammatically , the two - particle self - energy in impact - approximation reads @xmath199 note that one has to take into account that double counting has to be avoided so that scattering states in @xmath193 should not interfer with contributions from @xmath200 .
important are the account of bound states in @xmath193 .
the higher order corrections to the full vertex @xmath201 can be shown to effectively reduce the two - particle self - energy @xcite by the amount @xmath202 .
after analytic continuation @xmath203 to real frequencies , we obtain @xmath204^{-1}~. \label{eqn : pi_2_sigma}\end{gathered}\ ] ] note , that the dependance on @xmath205 is neglected in the self - energy of the lower energy level @xcite . neglecting the self - energy and the effective vertex in eq .
( [ eqn : pi_2_sigma ] ) , we arrive at eq .
( [ eqn : pi_20_sigma ] ) .
the interaction between two atoms in state @xmath44 and @xmath45 , moving with momentum @xmath206 and @xmath207 respectively , and carrying matsubara frequencies @xmath208 and @xmath209 , upon exchange of momentum @xmath210 is given by the following diagram : @xmath211^{-1}~ , \label{eqn : v_vdw_diagram } \end{gathered}\ ] ] with the unscreened coulomb propagator @xmath212 . due to the ion s heavy mass ,
we neglect the transfer momentum @xmath210 in the intermediate propagator as well as the kinetic energy @xmath213 . replacing the matsubara frequencies @xmath208 and @xmath209 by their on - shell values
, we can perform the summations over momenta in eq .
( [ eqn : matrixelementm_p ] ) and arrive at the familiar expression eq .
( [ eqn : v_vdw_r ] ) .
the expression eq . ( [ eqn : v_vdw_r_dip ] ) has to be evaluated .
we rewrite it as @xmath214 the primed sum indicates summation only over states , which assure a finite denominator . henceforth , we assume one atom in an excited state @xmath215 , and the second atom in its ground state , @xmath68 . the denominator is dominated by the term @xmath216 .
( [ eqn : deriv_strength010 ] ) becomes @xmath217 the diagonal matrix element of @xmath218 vanishes for atoms with no permanent electric dipole moment .
chosing the nucleus - nucleus axis parallel to the @xmath219-axis of the coordinate system , the operator @xmath218 reads @xmath220 since the expectation values of any coordinate @xmath221 or @xmath222 vanishes , e.g. @xmath223 , the mixed terms in @xmath224 vanish .
we obtain @xmath225~,\\ \intertext{and with } \langle r_i^2\rangle=\frac{1}{3}\langle r^2\rangle\\ c_6(\alpha_1,0)&=\left(\frac{e^2}{4\pi\epsilon_0}\right)^2 \frac{1}{\left ( 1 + 1/n_1 ^ 2 \right)\mathrm{ry } } \frac{2}{3}\langle0\!\left|\,\mathrm{r}^2\,\right|\!0\rangle\langle \alpha_1\!\left|\,\mathrm{r^2}\,\right|\!\alpha_1\rangle~. \label{eqn : deriv_strength030}\end{aligned}\ ] ] using @xmath226~ , \label{eqn : r2}\ ] ] we finally obtain eq .
( [ eqn : int_strength ] ) .
note , that for the case , that the first atom is also in its ground state , i.e. @xmath227 , @xmath228 is obtained .
the difference to the london - eisenschitz result 12.94 is due to the approximative treatment of the denominator in eq .
( [ eqn : deriv_strength010 ] ) in the calculation presented above .
for excited states , which are considered in this work , this contribution becomes negligible .
we start with the general form of expression ( [ eqn : v_vdw_r ] ) , i.e. the intensity distribution of a given spectral line , due to interaction between the radiator , which is in state @xmath229 before the transition and in state @xmath71 after the transition and the perturber @xmath230 in state @xmath231 .
it is obtained by performing an averaging procedure over the distribution of the @xmath232 perturbing atoms , @xmath233 here , @xmath74 is the probability density for having atom 1 in the volume element @xmath75 at @xmath76 , atom 2 in the volume element @xmath77 at @xmath78 , and so forth .
in can be expanded in a cluster decomposition as @xmath234 where the lowest non - ideal term is the pair distribution function , giving the probability to find a second particle at @xmath235 if there is one particle at @xmath236 .
the @xmath237-fold integral in eq . ( [ eqn : app : profile_at - at ] ) can be evaluated using the fourier representation of the delta function , @xmath238 with @xmath239 $ ] . in the case of statistically independant atoms , which are all in the same quantum state @xmath240 i.e. @xmath241 ,
one immediatly finds @xmath242 } \right]^{n}~. \label{eqn : profile_random}\ ] ] using the van der waals limit of the atom - atom potential , eq .
( [ eqn : vdwpot_00 ] ) , eq .
( [ eqn : profile_random ] ) turns to @xmath243^{n}~. \label{eqn : profile_random_dipole}\ ] ] the integration over @xmath244 can be rewritten in the form @xmath245\\ & = \frac{v}{4\pi}\left ( 1-\frac{4\pi\tilde v(\rho)}{v } \right)~ , \label{eqn : app : profile_integration_1}\end{aligned}\ ] ] with @xmath246 . in the thermodynamic limit , i.e. @xmath247 , while keeping constant the density of neutrals @xmath248 , one obtains @xmath249 for @xmath250 , @xmath251 is obtained , and finally the fourier integral eq .
( [ eqn : profile_random_dipole ] ) can be evaluated to give @xmath252 which corresponds to eq .
( [ eqn : profile_london ] ) .
|
balmer spectra are investigated which are obtained from hydrogen droplets irradiated by ultra - short intense laser pulses .
a unified quantum statistical description of bremsstrahlung , the stark broadening and the van der waals profile of hydrogen spectral lines is used , which allows to include many - particle effects . analyzing the line profiles , a low ionization degree of a dense plasma is inferred , where the main contribution to the spectral line shape originates from the interaction with the neutral components .
effective temperatures and densities of the radiating microplasma are deduced .
a dynamical description is given within plasma hydrodynamics , explaining the formation of excited atomic states in the expanding system and the occurrence of the observed balmer lines only below a critical density .
|
the energy evolution in the time dependent optical dipole trap can be expressed as follows @xcite : @xmath91 where @xmath92 is the truncation parameter with @xmath93 being the truncation energy and @xmath94 stands for the energy carried away by the evaporated atom in addition to @xmath95 . in an harmonic trap
, @xmath94 can be evaluated analytically @xcite : @xmath96 where @xmath97 is the incomplete @xmath52 function . the first term in the right hand of eq .
( [ eq : energyev ] ) describes evaporation cooling .
generally , evaporation in the optical trap is accompanied by weakening of the trapping confinement , thus causing additional adiabatic cooling which is accounted for by the second term in eq .
( [ eq : energyev ] ) . the depth of the optical dipole potential is reduced exponentially in time : @xmath98 with the time constant @xmath99 defined by the ratio of initial and final potential depths : @xmath100 accordingly , the temporal behaviour of the trap s oscillation frequency is : @xmath101 with @xmath102 being the geometric mean of the initial oscillation frequencies .
the atom number loss rate equation due to evaporation is @xcite : @xmath103 where @xmath104 .
@xmath105 is the two - body elastic collision rate . in derivation of eq .
( [ eq : atomnev ] ) , @xmath36 is a reference density defined as the central peak density in the limit of @xmath106 @xcite . in an harmonic trap @xmath36
is related to @xmath107 , @xmath34 and the total number of atoms @xmath7 : @xmath108 .
@xmath109 implies thermal averaging over maxwell - boltzmann distribution of the product of two - body collisional cross - section @xmath110 , defined by eq .
( 3 ) in the main text , and the relative velocity @xmath111 , where factor of @xmath84 stands for the reduced mass in the center - of - mass coordinate frame : @xmath112 where @xmath113 .
the evaporation efficiency @xmath114 contains the ratio between the effective volume for elastic collisions leading to evaporation @xmath115 and the single atom volume @xmath116 @xcite : @xmath117 taking all these factors into account , expanding eq .
( 3 ) of the main text to the lowest order in @xmath118 and assuming constant @xmath119 we can express eq .
( [ eq : atomnev ] ) in the final form : @xmath120\times\nonumber\\ \times\left(\alpha t\right ) \exp\left(-\frac{3t}{2\tau}\right ) n^2 , \label{eq : atomnevf}\end{aligned}\ ] ] where @xmath121 is the effective recombination length defined in the main text , and the rate constant of the atom number evolution @xmath122 and @xmath123 are : @xmath124 in an harmonic trap the average energy per atom is @xmath125 .
time derivative of this equation , combined with eqs .
( [ eq : energyev ] ) and ( [ eq : atomnev ] ) , leads to the rate equation for temperature : @xmath126 finally , integrating eq .
( [ eq : thermavg ] ) and substituting the expression of @xmath127 to eq .
( [ eq : tempev ] ) , we obtain : @xmath128l_e^2\times\nonumber\\ \times \left(\alpha t\right)^2 \left(\frac{\hbar\bar{\omega}_0}{k_b}\right ) \exp\left(-\frac{3t}{2\tau}\right ) n - \frac{t}{2\tau}\ ; , \label{eq : tempevf}\end{aligned}\ ] ] where @xmath129 is the rate constant of the temperature evolution : @xmath130 eqs . (
[ eq : atomnevf ] ) and ( [ eq : tempevf ] ) form a set of two coupled differential equations describing evaporation cooling dynamics .
their numerical solution is represented in fig . 2 of the main text .
three - body recombination rate is locally defined as @xmath131 .
the loss rate of atoms is obtained by integrating the three - body recombination rate over the sample s volume : @xmath132 . to obtain @xmath50 coefficients we fit the experimentally measured time evolutions of atom number and temperature with
two coupled rate equations following the model developed in ref .
@xcite : @xmath133 where @xmath134 and @xmath52 is the single - body loss rate coefficient due to residual collisions with thermal atoms in the vacuum . a typical set of measurements at a magnetic field value close to the scattering length zero - crossing
is shown in fig .
[ fig : ntevol ] .
combined fit to the rate equations reveals the value of @xmath135 . in
contrary , if only the atom number decay measurement is fitted keeping @xmath136 constant , @xmath137 .
thus , fitting the single equation of @xmath138 causes underestimation of the @xmath50 coefficient by @xmath139 as far as @xmath140 of atoms are lost .
we include this underestimation in the amplitude factor , discussed in the next section , and fit only the atom number loss measurements while keeping the loss fraction constant .
we note that the initial temperature of atoms is measured every time the loss measurement is performed .
we find it to be constant over the whole range of the magnetic field values presented in fig . 3 of the main text .
in addition , oscillation frequencies of the trap are regularly measured to carefully characterize atomic densities .
in the regime of @xmath53 , as mentioned in the main text , @xmath50 coefficient usually deviates from the overall @xmath54 scaling by a constant factor . to define contributions of the relevant experimentally measured quantities
to @xmath142 we recall that by definition given in the previous section , @xmath143 , where @xmath144 is the volume of the sample . in a cylindrically symmetric harmonic trap @xmath145 ) .
thus , @xmath146 . in this expression temperature is measured by the time - of - flight and thus the experimental uncertainties are buried in the calibration of the size of a camera pixel @xmath147 including magnification of the imaging system , @xmath148 .
the number of atoms is measured by the absorption shadow of the cloud on the ccd camera .
it is related to @xmath149 , where @xmath150 is the atomic absorption cross - section for weak imaging light of finite linewidth .
thus , the calibration error scales as @xmath151 and is corrected by @xmath142 .
the main systematic uncertainty is hidden in @xmath150 and thus in the absolute atom number counting .
our uncertainties in @xmath152 is @xmath153 dominated by the uncertainties in the oscillation frequencies . in case of evaporation cooling the finite temperature
is defined by @xmath154 . in the limit of @xmath9 ,
( [ eq : tempevf ] ) yields @xmath155 .
thus , @xmath156 so that its error scales as @xmath157 where @xmath158 implies an additional correction to @xmath142 due to uncertainties in the temperature measurement . the experimental error in temperature measurements is well tracked as being @xmath159 .
this allows for the estimation of @xmath158 according to @xmath160 . our fitting procedure in fig .
2 of the main text reveals @xmath161 yielding @xmath162 , in a good agreement with the above estimation .
|
we report on measurements of three - body recombination loss rates in an ultracold gas of @xmath0li atoms in the extremely nonuniversal regime where the two - body scattering length vanishes .
we show that the loss rate coefficient is well defined and can be described by two - body parameters only : the scattering length @xmath1 and the effective range @xmath2 .
we find the rate to be energy independent , and , by connecting our results with previously reported measurements in the universal limit , we cover the behavior of the three - body recombination rate in the whole range from weak to strong two - body interactions . we identify a nontrivial magnetic field value in the nonuniversal regime where the rate should be suppressed .
the few - body problem underlies fundamental processes in physics , yet it is notoriously difficult for finding analytic and numerical solutions @xcite .
it challenges our mind with the complexity of small and system - dependent molecular structures at the size of their interaction potentials . at the same time , in the regime of resonant pairwise interactions , it provides an elegant description of unusually large bound states possessing universal properties .
two important two - body length scales are involved in this description .
the first is the van der waals length @xmath3 , which is constant and connected to the radial range of the potential .
the second is the @xmath4-wave scattering length @xmath1 , which can be tuned magnetically via a feshbach resonance @xcite .
when @xmath5 , universal two - body states with size @xmath6 emerge , and a wealth of phenomena known as efimov physics is opened up in three- and , generally , @xmath7-body sectors .
the fundamental states belonging to the efimov effect ( efimov trimers ) depend log - periodically on @xmath8 @xcite .
sufficiency of two - body parameters in the efimov scenario indicates that two - body physics plays a decisive role in the universal few - body processes in ultracold gases .
the following questions arise as to how far this dominance of two - body physics extends when going to the opposite , nonuniversal limit , _
i.e. _ when @xmath9 .
what parameters govern the three - body processes and can we identify the influence of the short - range three - body forces in this limit ?
although recent theoretical studies probe the non - universal regime @xcite , their approaches can not be directly extended to the regions of extreme nonuniversality , the scattering length zero crossings .
three - body recombination induced loss of atoms has been extensively used as an efficient probe of efimov physics in trapped ultracold gases @xcite .
the process involves collisions of three particles resulting in a formation of a two - body bound state and a free atom .
the binding energy , usually much larger than the trap depth , is then released as kinetic energy of the colliding partners leading to losses . in this letter
, we investigate three - body recombination rates in the vicinity of two different zero crossings in a gas of ultracold @xmath0li atoms .
one zero crossing is associated with a broad feshbach resonance , while the other one is connected to a narrow resonance .
we observe a clear difference between the two regions with @xmath10 , and explain the magnetic field dependence of the three - body recombination loss rate coefficient in terms of an effective length parameter .
this parameter describes the two - body scattering phase shift for vanishing scattering length , and can be understood in terms of the finite range of the two - body potential , given by the van der waals length , and by the width of the feshbach resonance . assuming the dominance of two - body physics in three - body processes , we start by considering the effective range expansion of the scattering phase shift @xmath11 in its usual form @xcite : @xmath12 however , when @xmath9 the first term in the right hand of this equation diverges , and , to compensate this divergence , the second term diverges as well ( @xmath13 ) , which makes the above expression inconvenient in this limit .
a better form of the effective range expansion can be obtained by revamping eq .
( [ eq_cotandelta ] ) as @xmath14 where @xmath15 is the effective volume @xcite .
we show that although neither @xmath1 nor @xmath2 are good lengths at zero crossings , their combination in the form of the effective length @xmath16 remains finite , and captures the behavior of the three - body recombination loss rate coefficient remarkably well exactly when @xmath17 @xcite , where @xmath18 is the background scattering length and @xmath19 is inversely proportional to the width of the resonance .
however , for overlapping resonances , as is the case in @xmath0li , the analytic expression for the effective length is more involved and will be the subject of a future publication . instead , we utilize , here , numerical values for @xmath20 resulting from a coupled - channels calculation . ] .
once @xmath1 and @xmath20 are known , no further information on the short - range two - body or three - body potentials is needed . moreover , while the two - body collisional cross section becomes energy dependent at a zero crossing , the three - body recombination loss rate remains energy independent .
in addition we show that @xmath20 continues to be the dominant length in the inelastic three - body processes for larger scattering lengths up to a region where the universal three - body physics takes over and the leading length becomes @xmath1 . in the experiment
, we use an ultracold sample of @xmath0li atoms prepared in the @xmath21 state , where two feshbach resonances allow for the tunability of the scattering length @xmath1 .
the first resonance is located at @xmath22 g and is closed - channel dominated .
the second resonance is located at @xmath23 g , and is in between being closed- and open - channel dominated .
these resonances were located and characterized in ref .
@xcite by fitting the binding energies of weakly bound dimers obtained in rf - association spectroscopy with coupled channels calculations .
figures [ fig : scatteringlength](a ) and [ fig : scatteringlength](b ) show @xmath1 and @xmath2 dependence on the external magnetic field @xmath24 from our most up - to - date coupled channels analysis @xcite .
divergences of the scattering length correspond to the resonance positions . in ref .
@xcite , a log - periodic behavior of the three - body recombination loss rate coefficient has been demonstrated in the vicinity of the @xmath23 g resonance in the limit of @xmath5 . here ,
in contrast , we are interested in regions where the scattering length crosses zero . in fig .
[ fig : scatteringlength ] , two zero crossings are shown : at @xmath25 g and @xmath26 g . we ignore the third zero crossing at @xmath27 g ( not shown in the figure ) for its apparent similarity to the one at @xmath26 g from the point of view of the present study . as was noted earlier
, @xmath2 diverges at zero crossings which can be seen in fig .
[ fig : scatteringlength](a ) .
first , we consider two - body elastic processes in the vicinity of the @xmath25 g zero crossing . from eq .
( [ eq_tandelta ] ) we obtain the energy - dependent s - wave collisional cross section between two identical bosons @xmath28 accordingly , @xmath29 vanishes when @xmath30 , reflecting the emerging energy dependence of the two - body collisional cross section .
elastic collisions are crucial for evaporative cooling , therefore , the lack of cooling serves as the experimental strategy to find a vanishing cross section @xmath31 . in the experiment ,
@xmath0li atoms are trapped in a crossed beam optical dipole trap @xcite and we perform a fast cooling at the last stage of the evaporation cycle by lowering the power of the laser beams in 500 ms at different magnetic fields .
the initial temperature of atoms is measured to be @xmath32k and the final temperature , as a function of @xmath24 , is represented in fig .
[ fig:2bodyealstics ] . a maximum in temperature can be clearly identified at @xmath33 g which indicates the total failure of the evaporative cooling routine .
we model the temperature decrease during evaporation with a set of two coupled rate equations : for number of atoms @xmath7 and temperature @xmath34 @xcite .
the model includes time evolution of temperature due to evaporation cooling along with adiabatic cooling due to weakening of the optical dipole potential @xcite .
the latter cooling mechanism , being a single - body effect , is independent of @xmath29 and works equally well for all magnetic field values .
the former becomes inefficient when the elastic collision rate @xmath35 is minimized . in this expression
, @xmath36 is the peak density of atoms in the trap and @xmath37 implies thermal averaging over the maxwell - boltzmann distribution @xcite . the solid red line in fig .
[ fig:2bodyealstics ] represents the solution to our model with two fitting parameters : position of the scattering length s zero crossing and the amplitude factor ( discussed later ) that reflects our experimental uncertainties .
except for the last experimental point , the agreement with the data is remarkable which confirms that the two - body elastic collisional cross - section is minimized at the magnetic field of @xmath33 g .
the position of the scattering length s zero crossing is determined to be at @xmath38 g , where the error is dominated by the magnetic field calibration uncertainty .
this value agrees with the coupled channels analysis prediction of @xmath39 g ( see fig . [
fig : scatteringlength](a ) ) within less than 1@xmath40 of the experimental error and suits , even better , the position derived in the most recent coupled channels analysis of collisions in both isotopes of lithium atoms in ref .
@xcite .
the minimum in the elastic collisional cross section is shifted from the scattering length s zero - crossing position due to the energy dependence of @xmath29 . for comparison ,
the dashed line in the figure represents the same model excluding the energy dependence of the cross - section which disagrees with data .
the same experimental strategy can be applied to identify another @xmath31 position related to the lower magnetic field zero crossing . however
, weak dependence of @xmath1 on @xmath24 in this region allows for a significant tolerance in the uncertainty of the zero - crossing position . from the point of view of the present study , this is also indicated by a very weak magnetic field dependence of the three - body effective length @xmath41 ( explained later ) around @xmath26 g ( see fig.[fig : scatteringlength](b ) ) .
next , we consider the three - body inelastic processes ; however , first it is worth noting that the @xmath42 state is not the absolute ground state of the system , and thus , two - body inelastic dipole - dipole relaxation is allowed . however , as shown in ref .
@xcite it is extremely weak for all magnetic field values relevant here , and the dominant trap - loss mechanism is induced by three - body recombination .
after loading atoms from a magneto - optical trap into an optical dipole trap we perform evaporative cooling at the wing of the narrow resonance
. then we jump across the resonance to @xmath43 g within @xmath44ms and wait for @xmath45ms before we move within @xmath46ms to the target field where atom number and temperature time evolutions are measured .
the typical holding time at the final magnetic field varies from @xmath47 to @xmath48 sec where we lose @xmath49 of atoms .
the three - body recombination loss rate coefficient @xmath50 is extracted from the fit of atom number decay measurements with the solution to the three - body loss rate differential equation : @xmath51 , where @xmath52 is the single - body loss rate coefficient due to residual collisions with thermal atoms in vacuum @xcite .
three - body recombination is known to be accompanied by heating and @xmath50 should be extracted from a set of two coupled rate equations : atom number and temperature time evolutions @xcite .
however , in the regime of @xmath53 , even with this precaution @xmath50 usually deviates from the overall @xmath54 scaling by a constant factor . excluding log - periodic oscillations of @xmath50 due to efimov physics
, the deviation can be attributed mainly to the systematic uncertainties in the atom number calibration @xcite .
an elegant method to calibrate atom number is based on the equation of state measurements recently used to characterize the unitary regime where @xmath55 @xcite . here , to calibrate our data we use the amplitude factor for @xmath50 previously determined in our system in the universal regime @xcite .
analysis of the systematic uncertainties , presented in supplemental material , shows that the amplitude factor applied for @xmath50 can be related to the amplitude factor introduced earlier as a fitting parameter in fig .
[ fig:2bodyealstics ] @xcite .
a good agreement between these factors hints at their correct interpretation as being dominated by the experimental uncertainties in the atom number calibration .
note , also , that using a single rate equation for atom number decay to extract @xmath50 causes only a slight correction to the overall amplitude factor @xcite .
for @xmath56 the three - body recombination loss rate coefficient is commonly represented as @xmath57 .
the general @xmath54 power dependence is dictated by the resonantly enhanced two - body interactions , while @xmath58 reflects the discrete scaling invariance of @xmath50 related to efimov physics . for positive scattering length
, @xmath58 shows log - periodic behavior with the maximum value of @xmath59 . because of the relatively short lifetime of efimov trimers in @xmath0li ( large inelastic parameter @xmath60 )
, this value was measured to be somewhat smaller , @xmath61 @xcite .
the universal @xmath54 dependence is expected to break down when @xmath9 .
still @xmath50 can be formally represented here as @xmath62 with @xmath63 being a characteristic recombination length for the measured @xmath50 values and @xmath64 assumed to be a scattering length independent constant .
measurements of @xmath63 as a function of the magnetic field @xmath24 for two different temperatures are shown in fig .
[ fig:3bodyrecomb ] , setting @xmath65 , the universal limit s maximal value . above @xmath66 g
we reproduce the measurements from ref .
@xcite where @xmath63 follows @xmath1 . for lower values of the magnetic field ,
deviations from @xmath1 dependence become evident .
instead of decreasing with @xmath1 , @xmath63 saturates at @xmath67 where @xmath68 is the bohr radius and starts to increase again below @xmath69 g .
nothing dramatic happens at the point of the scattering length s zero - crossing where @xmath63 continues to increase smoothly . for comparison
, we note that the van der waals length of @xmath0li is @xmath70 . to analyze the data , we assume that the two - body physics plays a decisive role in three - body recombination loss rates , and that the relevant parameter can be directly extracted from the expansion of the two - body phase shift @xmath71 when @xmath9 , the dominant term in eq .
( [ eq_phaseshift ] ) is @xmath72 where the effective length is @xmath73 .
@xmath20 is represented in fig .
[ fig:3bodyrecomb ] as a blue dashed - dotted line ( indistinguishable from the pink line below @xmath74 g ) .
notably , it stays finite for vanishing @xmath1 and diverging @xmath2 .
moreover , in the vicinity of @xmath25 g it agrees remarkably well with the data with no adjustable parameters .
this agreement confirms our assumption that the two - body physics alone defines the behavior of @xmath50 . for larger magnetic field values
, @xmath20 starts to slightly deviate from the data .
however , according to eq .
( [ eq_phaseshift ] ) , for larger scattering length the two - body scattering volume should be appended by the @xmath75 term and the effective length then becomes @xmath76 in fig .
[ fig:3bodyrecomb ] @xmath41 is shown as a pink dashed line and it is in excellent agreement with the data up to @xmath77 g at which the first term in eq .
( [ eq_phaseshift ] ) becomes dominant .
note that for two - body elastic scattering in the @xmath9 limit , the effective volume @xmath78 is multiplied by @xmath79 ( see eq .
( [ eq_sigma2b ] ) ) to form the effective collisional length which is then explicitly energy dependent . on the contrary ,
@xmath20 and @xmath41 posses no explicit energy dependence . as the relevant length for @xmath50 can be constructed differently as compared to eq .
( [ eq_effectivelength ] ) , _
e.g. _ as the aforementioned effective collisional length , the question of energy dependence of @xmath50 remains open .
thus , we measure the three - body recombination loss rate coefficient at a different temperature , @xmath80k , and the result is shown in fig .
[ fig:3bodyrecomb ] ( red dots ) .
the two measurements are indistinguishable within the experimental errors providing evidence that @xmath50 is energy independent . as an additional argument
, we note that if @xmath41 is defined according to eq .
( [ eq_effectivelength ] ) , the measured recombination length @xmath63 agrees with it if @xmath65 in eq .
( [ eq_3bodyrate ] ) , the same value as in the universal limit where @xmath56 .
any attempt to define an energy - dependent effective recombination length , for instance as @xmath81 in analogy to the two - body collisional length , will require tuning of @xmath64 to a significantly different value .
finally , we consider the zerocrossing at @xmath26 g . if we assume , again , @xmath82 to be the relevant length there , @xmath50 can be evaluated from eq .
( [ eq_3bodyrate ] ) substituting @xmath63 with @xmath83 ( see fig . [
fig : scatteringlength](b ) ) .
this predicts the @xmath50 to be @xmath84 orders of magnitude smaller than the values measured at @xmath25 g which is well below our resolution limit for the highest achievable densities set by collisions with residual atoms from vacuum .
thus , from our data , we can only extract the upper limit for @xmath82 at @xmath26 g , @xmath85 , which is consistent with the predicted general trend of being smaller than the one at @xmath25 g . in conclusion ,
we show a well - behaved three - body recombination loss rate at vanishing two - body scattering length .
we identify that the only relevant parameters to define @xmath50 in this limit are @xmath1 and @xmath2 .
it is interesting to extend our studies to other atomic species .
feshbach resonances in @xmath0li are of intermediate character which is not the case for @xmath86cs atoms .
therefore , a similar study for @xmath86cs , @xmath87k , and @xmath88rb would be interesting to complete the description of @xmath50 at a zero crossing .
it is also interesting to note that @xmath89 at @xmath90 g ( see fig . [
fig : scatteringlength](b ) ) setting minimum in @xmath50 .
it would be an interesting region for a bose - einstein condensate as its lifetime is expected to be maximal there .
counter - intuitively , the minimum of three - body losses does not occur at zero scattering length . finally , we note that the recombination length , @xmath41 , may play an important role in the effective range corrections to the lowest energy level in the efimov spectrum .
note that the energy dependence of @xmath50 was reported in the vicinity of a narrow feshbach resonance in a two - component fermi gas @xcite
. however , direct three - body recombination in such a mixture is forbidden , and it is mediated through collisions of closed channel molecules with free atoms .
this mechanism might become relevant in bose gases for extremely narrow feshbach resonances .
_ note added : _ while finalizing this manuscript we became aware of a recently developed theory that supports our findings @xcite .
l. k. and o. m. acknowledge fruitful discussions with c. h. greene .
this research was supported by grants from the united states - israel binational science foundation ( bsf ) and the israel science foundation ( isf ) , jerusalem , israel . 29 natexlab#1#1bibnamefont # 1#1bibfnamefont # 1#1citenamefont # 1#1url # 1`#1`urlprefix[2]#2 [ 2][]#2 , _ _ ( , ) . , , , , * * , ( ) . , * * , ( ) . , , , * * , ( ) . , , , , * * , ( ) . , , , * * , ( ) . ,
* * , ( ) . , * * , ( ) . , * * , ( ) . , , , , , , * * , ( ) . , , , , , * * , ( ) . , , , , , , , , , , , * * , ( ) . , , , , , , , , , * * , ( ) . , , , ,
* * , ( ) . , , , , , .
, , , , , , , * * , ( ) . , _
_ ( , , ) . , , , * * , ( ) . , , , ,
, * * , ( ) . , , ,
, * * , ( ) . ,
* * , ( ) .
, , , * * , ( ) . , , , , * * , ( ) . , * * , ( ) . , , , , , * * , ( ) . , , , , * * , ( ) . , , , ,
* * , ( ) . , . ,
|
there is a growing consensus that a large fraction of core - collapse supernovae , possibly including the whole subset of type ib / c events , undergo significantly asymmetrical explosions . while clearly rotation and magnetic fields may play some role in this , instability associated with the stalled shock that forms at a radius of 100 - 300 km and lasts for a few hundred microseconds post bounce offers the intriguing possibility of generating asymmetrical explosions from a symmetrical progenitor , since such instability is usually dominated by low mode @xmath0 , 2 oscillations of the shock front .
the original speculation for the origin of such oscillations @xcite in terms of convection behind the shock , following @xcite has been revised in recent years .
@xcite demonstrate that post shock advection can stabilize convective instability ( @xcite only deals with a static case ) , and that convection alone can not produce the dominant @xmath0,2 modes . in adiabatic simulations of gas with polytropic index 4/3 , which as shown by @xcite is a good description of shocked inner regions of a core - collapse supernova ,
@xcite were the first to associate the instability with the shock itself .
@xcite attributed this to the growth of trapped sound waves , travelling essentially laterally around the central accretor .
this must be a purely shock driven instability , since convection is absent , and hence qualitatively different to the original speculations @xcite .
on the other hand , @xcite and @xcite attribute this shock - driven instability to an `` advective - acoustic '' cycle , originally introduced by @xcite in the context of black hole accretion flows .
vorticity perturbations generated at the distorted shock front advect inwards , and couple with outgoing sound waves at the inner boundary of the postshock accretion flow . upon reaching the outer shock , these sound waves further distort the shock front , leading to a positive feedback .
while the mechanisms of @xcite and @xcite appear to produce similar outcomes in non - rotating cases , indeed @xcite actually confirm the numerical values of frequencies and growth rates measured by @xcite , it is possible the presence of rotation might lead to substantial differences , which we discuss further below .
instabilities of the accretion shock also are observed in simulations with a more accurate treatment of supernova microphysics , with a realistic equation of state , neutrino transport and attendant nuclear reactions , and more physically motivated boundary conditions . @xcite and @xcite investigate the role of such instabilities in generating pulsar natal kicks , while @xcite study the ejection of metal clumps to the outer regions of supernova ejecta . @xcite
also see the standing accretion shock instability 100 - 300 ms postbounce .
the important feature of their simulation , though , is a core gravity wave at late times that transfers acoustic power to the shock to continue powering the explosion , the initial perturbation for which is possibly seeded by the accretion shock instability . @xcite
also favor the interpretation of @xcite of an advective - acoustic cycle , though in this and the other more realistic simulations cited above , the exact mechanism of instability is difficult to identify .
to avoid this problem , @xcite performed numerical simulations based on unperturbed spherically symmetrical shock accretion flows given by @xcite , a sort of `` middle ground '' between the adiabatic flows of @xcite and @xcite and the full supernova simulations of other workers .
@xcite clearly identify and measure growth rates for @xmath0 and 2 mode instabilities .
the real parts of the frequencies scale with @xmath2 , the integral between the inner radius of the accretion flow and the shock of the sum of the inverses of the sound speed and advection speed , suggesting that an advective - acoustic mechanism , as advanced by @xcite and @xcite , is at work .
however @xcite only perturb the radial velocity component of the initially spherical symmetrical unperturbed flow , which may unduly influence their results , compared to allowing perturbations in polar and azimuthal directions as well .
@xcite include rather more physics , and again based on the frequencies at which modes grow , conclude that an advective - acoustic cycle is at work . in this work we present another approach to the problem .
following the work of @xcite on planar shocks in the interstellar medium , we derive an approximate dispersion relation for oscillations of the accretion shock by adding the effect of a gravitational field and spherical geometry . in the case that the postshock advection is neglected , a quartic equation ( actually a quadratic in @xmath3 where @xmath4 is the frequency of oscillation ) results , becoming a quintic equation in @xmath5 with the inclusion of advection .
we find that in both cases for @xmath6 , for all except the highest values of the ratio of the shock radius to the radius of the inner boundary , modes with @xmath0 grow the fastest ( in fact for cases where the postshock radial velocity , @xmath7 , no other nonradial modes grow ) .
this suggests that for @xmath0 in @xmath8 gas in this regime , postshock advection is not crucial to the operation of the instability .
thus in general these results will support the conclusions of @xcite that the instability proceeds by the growth of trapped waves in this regime .
nevertheless , regions of @xmath9 and @xmath10 parameter space exist where we find modes that are only unstable in the presence of postshock advection , which we interpret as an advective - acoustic cycle , as suggested by @xcite and @xcite . in both cases ,
the frequencies of the growing modes are similar , suggesting that frequency alone is not a good discriminator of the mechanism of instability .
as will be seen below , this might be expected in a comparison of a perturbation advected radially between the shock and inner boundary , and a sound wave traveling essentially laterally around the shock .
we consider an unperturbed model in which spherically accreting plasma is decelerated at a spherically symmetrical shock , before accreting onto a protoneutron star .
the accretion shock is at radius @xmath11 , and we take an inner boundary at @xmath12 , maintained at constant pressure , where plasma cools and decouples from the postshock flow .
the postshock flow is modeled by the bernoulli equation , as given in appendix a. gas with polytropic index @xmath13 decelerates , and gas with @xmath14 accelerates away from the shock towards the inner boundary .
we shall be solely concerned with @xmath13 .
the dominant variations with radius are in pressure and density , as shown in figure 1 . to treat the perturbation , we follow in large part the methods and notation of @xcite who derive an approximate dispersion relation for application to shocks with arbitrary postshock structure .
in spherical coordinates , we write the hydrodynamic equations ; an equation of continuity and momentum equations in the @xmath15 , @xmath16 and @xmath17 directions as continuity : @xmath18=0&\end{aligned}\ ] ] radial force : @xmath19&\cr = -{\partial\left(p+\delta p\right)\over\partial r}+{\partial\over\partial r } \left[gm\left(\rho + \delta\rho\right)\over r\right]&\end{aligned}\ ] ] poloidal force : @xmath20=&\cr -{1\over r}{\partial\left(p+\delta p\right)\over\partial\theta } & \end{aligned}\ ] ] azimuthal force : @xmath21&\cr+ \left(\rho + \delta\rho\right)\left [ \left(u_{\phi}+v_{\phi}\over r\sin\theta\right){\partial\left(u_{\phi}+v_{\phi } \right ) \over\partial\phi}-{\left(u_{\phi}+v_{\phi}\right)\left(u_r+v_r\right ) \over r}\right]= -{1\over r\sin\theta } { \partial\left(p+\delta p\right)\over\partial\phi},&\end{aligned}\ ] ] where @xmath22 and @xmath23 are the initial and perturbed density respectively , @xmath24 and @xmath25 are the initial and perturbed radial velocities and @xmath26 and @xmath27 and the initial and perturbed azimuthal velocities . the initial poloidal velocity is zero , and its perturbed value is @xmath28 . @xmath29 and @xmath30 are initial and perturbed pressures , and @xmath31 is the gravitational constant times the enclosed mass which is dominated by that of the protoneutron star . writing @xmath32 , and assuming @xmath33 where @xmath34 we derive a linearized continuity equation , @xmath35 where @xmath36 where @xmath37 and @xmath38 is the density scale length , defined by @xmath39 .
the @xmath17 derivatives of terms in @xmath26 and @xmath27 have been taken to yield the @xmath40 dependence , but the @xmath16 derivative of the term in @xmath28 remains explicit .
below ( in section 3.2 ) we will assume @xmath41 rendering @xmath42 constant .
the radial force equation is linearized as follows .
the zero order equation gives @xmath43 if @xmath44 and @xmath45 , and @xmath46 ( @xmath7 ) or @xmath47 ( @xmath48 ) , then equation 6 evaluates to zero .
we will assume that this is approximately true for all @xmath49 .
( certainly @xmath50 throughout this range , whereas for @xmath14 , @xmath51 ) .
this helps ensure that for all cases where the postshock gas decelerates , the radial @xmath52 mode will be stable , in agreement with the conclusions of @xcite , and @xcite and @xcite , where we are assuming a neutrino luminosity below the critical value in these last two references . taking @xmath53 ( i.e. an approximately isothermal perturbation ; this gives a considerable simplification , see appendix b ) the equation to first order in small quantities becomes @xmath54 = \cr & \left[{2u_{\phi}v_{\phi}\over r}-{v_{\theta}\over r}{\partial u_r\over\partial \theta } + \delta\left[\left({gm\over r}-c_s^2\right)\lambda + { 2u_r^2+u_{\phi}^2\over r}+{u_r^2\over l } \right]\right]\exp\int i{\omega ^{\prime}\over u_r}dr .\cr\end{aligned}\ ] ] we note that @xmath55=\left[{-d\lambda /dr+i\left(\omega /u_r^2\right ) \partial u_r/\partial r\over \left(i\omega ^{\prime}/u_r+\lambda\right)^2}+1\right ] \exp\int\lambda+i\omega ^{\prime}/u_rdr.\ ] ] for a global mode we expect @xmath56 to be of order @xmath57 , ( in fact for @xmath8 , @xmath58 ; see equations 16 below ) .
if @xmath5 represents laterally propagating sound waves , @xmath5 is in the range @xmath59 , whereas @xmath60 for radially propagating vorticity perturbations . in either case , with @xmath61 from appendix a , @xmath62 and we may put @xmath55\simeq\exp\int\lambda+i\omega ^{\prime}/u_rdr.\ ] ] in equation 7 we neglect terms in @xmath63 , @xmath64 ( of order @xmath65 ) and integrate using equation 9 , further assuming @xmath66 , to get @xmath67 with @xmath68 and @xmath69 , equation 7 gives @xmath70 where we have substituted from equation 10 for @xmath25 to simplify the right hand side .
taking @xmath71 from appendix a , we derive @xmath72 expressions for @xmath28 and @xmath27 are easily derived in the limits @xmath73 or @xmath74 . staying with the nonrotating case for the time being ,
we give for @xmath68 : @xmath75 where the primes have been dropped from @xmath5 .
substituting our expression for @xmath25 , @xmath76 , @xmath28 and @xmath27 into the linearized continuity equation ( 5 ) we find @xmath77 where the @xmath40 dependence disappears ( as it should in spherical symmetry ) using properties of the spherical harmonics .
this can be solved for @xmath56 to give @xmath78 we write ( neglecting terms in @xmath63 ) @xmath79 as discussed by @xcite , a third solution exists , with @xmath80 everywhere . in that case @xmath81 , and @xmath28 and @xmath82 .
the linearized continuity equation ( 5 ) then gives @xmath83 taking @xmath84 we find @xmath85 the derivation of the dispersion relation from the boundary conditions in terms of these three solutions is carried out in appendix b , and given by equation b6 .
we use @xmath86 , with @xmath87 and @xmath88 evaluated at the outer shock and inner boundary respectively in terms of @xmath89 and @xmath90 , the advection velocities at these locations ( see appendix b for more details ) .
then in equation b6 we approximate @xmath91 and @xmath92\cr & + \left({1\over l}+{4\gamma -5\over\gamma -1}{1\over r}\right){i\omega\over 2\left(gm / r - c_s^2\right ) } \left[u_{rs}\left(2-q{1+\beta ^q\over 1-\beta ^q}+{1\over q}{1+\beta ^q\over 1-\beta ^q}\right ) u_{ri}\left(2+q{1+\beta ^q\over 1-\beta ^q}-{1\over q}{1-\beta ^q\over 1-\beta ^q}\right)\right]\end{aligned}\ ] ] where we have put @xmath93and neglect terms of @xmath94 and higher .
these then give @xmath95 { 1\over q}{1+\beta ^q\over 1-\beta ^q}\cr & -\omega ^2\left[{l\left(l+1\right)\over r^2}v_su_r + { gm / r - c_s^2\over al}\left({1\over l}+{5\over 2r}\right)\right]\cr & + \omega ^2\left[{gm / r - c_s^2\over 2}\left({1\over l}+{4\gamma -5\over\gamma -1}{1\over r}\right ) \left({1\over
l}+{5\over 2r}\right)\left(1+q{1+\beta ^q\over 1-\beta ^q}\right)+ { gm / r - c_s^2\over 2}\left({1\over l}+{4\gamma -5\over\gamma -1}{1\over r}\right ) { 1\over al}\left(1-q{1+\beta ^q\over
1-\beta ^q}\right)\right]\cr & + \omega\left[iu_{rs}\left({1\over l}+{5\over 2r}\right)l\left(l+1\right){c_s^2\over r^2 } + i{u_{ri}\over al}l\left(l+1\right){c_s^2\over r^2}+ i{u_{ri}\over alq}{1+\beta ^q\over 1-\beta ^q } l\left(l+1\right){v_su_{rs}\over r^2}\right]\cr & + \omega\left[-{i\over 2}\left({gm\over r}-c_s^2\right)\left({1\over l}+{4\gamma -5\over \gamma -1}{1\over r}\right)^2\left(u_{rs}\left({1\over l}+{5\over 2r}\right)\left(1-q{1+\beta ^q\over 1-\beta ^q}\right)+{u_{ri}\over l}\left(1+q{1+\beta ^q\over 1-\beta ^q}\right)\right)\right]\cr & + \omega\left[{{gm\over r}-c_s^2\over 2al}\left({1\over l}+{4\gamma -5\over \gamma -1}{1\over r}\right)\left({1\over l}+{5\over 2r}\right)\left(3iu_{rs}+{iu_{rs}\over q}{1+\beta ^q\over 1-\beta ^q}+iu_{ri}+{iu_{ri}\over q}{1+\beta ^q\over 1-\beta ^q}\right)\right]\cr & + { gm / r - c_s^2\over al}\left({1\over l}+{5\over 2r}\right ) { l\left(l+1\right)c_s^2\over r^2 } -{gm / r - c_s^2\over 2al}\left({1\over l}+{4\gamma -5\over\gamma -1}{1\over r}\right ) \left({1\over
l}+{5\over 2r}\right ) { l\left(l+1\right)v_su_{rs}\over r^2 } \left(1-q{1+\beta ^q\over 1-\beta ^q}\right)\cr & + { 1\over\omega}\left[{gm / r - c_s^2\over 2al}\left({1\over l}+{5\over 2r}\right)l\left(l+1\right){c_s^2\over r^2 } \left({1\over l}+{4\gamma -5\over\gamma -1}{1\over r}\right)\left(-iu_{rs}\left(1+q{1+\beta ^q\over 1-\beta ^q } \right)-iu_{ri}\left(1-q{1+\beta ^q\over 1-\beta ^q}\right)\right)\right]\cr & { 1\over\omega}\left[-{l\left(l+1\right)v_su_{rs}\over r^2 }
\left(i{u_{ri}\over 2al}\left(gm / r - c_s^2\right)\left({1\over l}+{4\gamma -5\over\gamma -1}{1\over r}\right)^2 \left(1+q{1+\beta ^q\over 1-\beta ^q}\right)+i{u_{ri}\over al}{l\left(l+1\right)c_s^2\over r^2}\right)\right]=0\end{aligned}\ ] ] if @xmath7 , equation 22 is very similar to equation 13a of @xcite , and is identical if we take @xmath96 , @xmath97 ( to agree with their sign convention ) , @xmath98 and @xmath99 ( i.e. @xmath100 ) . in their case
the limit @xmath101 gives a dispersion relation similar to the exact solution in the thin shock limit @xcite , and identical to this limit if terms describing the evolution of the shock are neglected from the exact solution .
they then proceed to argue that taking @xmath102 gives a suitable approximate dispersion relation away from the thin shock limit .
our case is particularly suitable since the standing accretion shock can be treated as an equilibrium state @xcite , i.e. it does not evolve in our approximation . just taking @xmath7 , equation 20 yields a quadratic equation in @xmath3 for which simple solutions exist .
@xcite find growing shock oscillations in the case of decelerating shocks , with a post shock rarefaction . in our case ,
the growing oscillations exist for density increasing with distance behind the shock .
this difference arises because the square of the sound speed , @xmath103 in our case , and @xmath104 . when @xmath7 ( except in terms in @xmath105 ) and taking @xmath106 , @xmath107 where @xmath108 when @xmath109 , this gives a pair of sound waves for given @xmath9 with @xmath110 taking the positive sign , and the negative sign gives a pair of gravity waves with @xmath111 where @xmath112 is a constant coming into our definition of the interior boundary condition , as explained in appendix b. for @xmath113 , we have four roots of the form @xmath114 and @xmath115 , and all roots have a combined character of acoustic and gravity waves . reinstating the terms in odd powers of @xmath5 with nonzero @xmath24 , the roots separate into pairs of what may be termed acoustic and gravity waves , although the difference in frequency between them is generally rather small and both pairs are of presumably mixed character , and one imaginary ( i.e. purely growing or damping ) mode . assuming a value of @xmath116 , we solve the quintic equation numerically .
results for each growing mode are then iterated until the value of @xmath116 produced by the frequency and wavenumber matches the input .
results of this procedure for growing modes are given in table 1 for @xmath8 , @xmath117 , and in table 2 for @xmath118 , @xmath119 , the values of @xmath112 being chosen to reproduce as far as possible the stability properties of the radial mode .
we compare in each case the results of the full quintic equation with a simpler case taking @xmath120 . with @xmath8 , taking @xmath121 ( from equation a2 neglecting @xmath122 ) , and without advection , we find growth for @xmath0 for all values of @xmath123 .
growth at @xmath0 at @xmath124 and at @xmath1 for all @xmath125 require non - zero advection terms in the dispersion relation , though the @xmath1 mode is stable as @xmath124 , even in the presence of advection .
the @xmath52 mode is marginally stable in the absence of advection .
with advection included , it is stable for the lowest @xmath125 , but otherwise unstable .
taken together , there is reasonable qualitative agreement with @xcite , but quantitatively , the model is less stable than it should be .
growth rates are too high , and the @xmath52 mode becomes unstable too quickly as @xmath125 increases .
better qualitative and quantitative agreement is found in the model with @xmath118 and @xmath119 given in table 2 . , and
also illustrated in figure 2 .
the stability of @xmath52 is improved , and advection becomes relatively more necessary to the instability of @xmath0 .
these results have a simple interpretation in terms of the mechanisms put forward by @xcite and @xcite .
where advection is not necessary to the instability , i.e. for the @xmath0 mode at low values of @xmath125 , the wave growth must be due to trapped sound waves similarly to the simulations of @xcite .
the inclusion of advection in these cases generally increases the growth rate , either due to a separate contribution of an `` advective - acoustic '' nature or because of the effect of postshock advection on the propagation of sound waves . in cases where no instability occurs
unless advection terms are present , the growth must be due to an advective - acoustic cycle similar to @xcite . in our model ,
these cases include all the @xmath1 modes , as well as the @xmath0 modes for large values of @xmath125 .
these results and interpretation add a new twist to the conclusions of @xcite .
these authors demonstrate that for @xmath126 the advective - acoustic cycle is the cause of the @xmath0 instability .
their analysis makes use of a wkb approximation that is not valid for lower @xmath125
. however they proceed to argue that since growth rates and frequencies vary smoothly as one goes to lower values of @xmath125 , the mechanism of instability should remain the same .
however our results suggests that even though the eigenvalues vary smoothly , the mechanism of instability changes to more closely resemble that suggested by @xcite .
a number of simplifying assumptions have been introduced into the analysis to make the algebra tractable . we have assumed an isothermal perturbation , used a greatly simplified interior boundary condition , and have dropped all terms in the advection velocity of order higher than @xmath127 .
nevertheless our model appears to capture the basic features of the numerical simulations of @xcite , _ and _ the analytic theory of @xcite and @xcite , in the appropriate regions of parameter space .
looking at the values of @xmath128 given by equation 15 , we find when @xmath129 , @xmath130 , @xmath131 and @xmath132 for @xmath133 .
consequently , @xmath134 gives a density perturbation @xmath135 that is greatest at the outer shock and considerably smaller at the inner boundary , while @xmath136 gives a density perturbation that varies much less with radius , but can be largest at the outer shock or at the inner boundary .
when @xmath52 , @xmath137 and @xmath138 . in this case the radial mode can have a large maximum of @xmath139 at the inner boundary . given the ad hoc nature of our inner boundary condition ( see appendix b ) , we might expect less accurate results for the radial mode than for nonradial modes , which does indeed seem to be the case .
the vorticity ( calculated from equations 13 ) increases inwards approximately as @xmath140 .
the third solution with @xmath141 everywhere must represent a purely vortical perturbation .
so far we have developed a dispersion relation from the equations of hydrodynamics , and found instabilities through the existence of complex roots .
we have not made any comment on the precise mechanism of the instability , other than to comment that sound / gravity waves appear to grow .
the vortical - acoustic cycle has been explained by @xcite .
a distortion of the accretion shock produces a vorticity perturbation that is advected inwards . upon reaching the inner boundary ,
the vorticity perturbation produces an upward moving sound wave , which reinforces the shock distortion when it arrives there , producing a net increase in the original perturbation . over many cycles , exponential growth ensues .
sound waves produced at the outer shock , in a realistic model , will not propagate towards the center due to the effects of refraction .
instead , they will propagate around the circumference of the shock .
@xcite speculate that sound waves produced at one position on the shock surface by a density inhomogeneity will propagate around until they meet on the opposite side , where their excess pressure produces a shock distortion that sends another pair of sound waves back again .
it is not immediately clear without detailed calculation that such a process should produce a growing ( as opposed to damping ) oscillation .
we speculate that growth must also be aided by the flow of plasma in through the accretion shock . in figure 3
we show flowlines for shocked plasma in an accretion shock with a 10% @xmath0 modulation of the shock radius . when the shock is shifted up by the instability , incoming plasma in the equatorial regions is diverted downwards by the now oblique shock .
this non - radial flow of shocked plasma enhances the oscillation and leads to further wave growth .
this aspect is similar to the thin shell overstability of decelerating shocks @xcite , ( although @xcite put forward a different view ) , though in these cases the pressure produced by gravitational confinement plays no role , and overstability exists for a postshock density gradient oppositely directed to our case .
thus both the advective - acoustic ( or vortical - acoustic ) instability and the growth due to trapped sound waves have their origins in perturbations of either vorticity or pressure in the post shock flow produced by the distorted shock front .
there is considerable interest in the effects of rotation on core - collapse explosions . in particular ,
in which direction with respect to the rotation axis should one expect the strongest growth ?
this would have obvious implications for the directions of pulsar kicks , assuming the such instabilities are the correct mechanism . in the case of an advective - acoustic instability
, we should expect the decrease in the postshock advection in the plane of rotation to give stronger instability and hence strongest growth along the axis of rotation . in the case of instability driven by trapped sound waves we can sketch out the effect of including a nonzero @xmath26 in the forgoing analysis .
we derive the extra terms to be included in the dispersion relation in the limits @xmath74 and @xmath142 .
we also neglect any distortion of the shock front by the rotation .
such considerations require more detailed physics , neutrino cooling near the inner boundary at a minimum @xcite , and ideally also neutrino luminosities and mass accretion rates ( see e.g. * ? ? ?
* ; * ? ? ?
* ) , and are beyond the scope of this paper .
this keeps @xmath143 , which should be true in any case if @xmath7 , but even in the absence of advection , a distortion of the shock front may change the boundary condition leading to equation b2 .
the perturbed velocities are @xmath144 with @xmath145 the linearized continuity equation becomes @xmath146 after dividing through by @xmath139 , with solutions for @xmath128
@xmath147 we again have @xmath148 if @xmath141 , so equation ( 14 ) still holds .
then , with unchanged boundary conditions , the dispersion relation is the same as before ( with @xmath7 and @xmath149 in equation b6 ) , but with the extra terms ( neglecting the change @xmath150 , which in any case is trivial for @xmath8 ) @xmath151 the new quintic equation is again solved numerically , with results given in table 3 for the @xmath152 component of the @xmath0 mode .
we generally find higher growth rates increasing with rotation rate for the @xmath152 ( @xmath153 remains unchanged ) , which would indicate higher growth in the plane of rotation .
this happens even in cases where in the absence of rotation , the advective - acoustic instability seems to be dominant .
we speculate that rotation increases the growth rate of trapped sound waves for nonzero @xmath40 , so much so that they can grow even in cases with zero growth in the absence of rotation . at higher rotation rates ,
the distortion of the shock front can not be neglected .
we expect the equilibrium shock position to be oblate , i.e. bulging out at the equator as demonstrated quantitatively by @xcite .
we emphasize again that our results are not applicable in this regime .
using heuristic arguments based on @xcite , and a few other approximations , we have derived an approximate dispersion relations for oscillations of a standing spherical accretion shock .
the pursuit of an analytic treatment further through the problem than has previously been attempted arguably allows greater physical insight , albeit at the expense of reduced quantitative accuracy .
while our model might not be considered much more than a `` toy model '' , it does capture most of the features of previous works of the instability of such shocks .
it offers a potential resolution to the mild disagreement between the works of @xcite and @xcite .
we find results consistent with both works , but in different regions of parameter space . for low values of @xmath125 ,
the growth of trapped waves dominates in the @xmath0 , and appears to be the most important mechanism .
the inclusion of post shock advection gives a small change , usually an increase , in the growth rate . at
higher @xmath9 , or @xmath125 , postshock advection appears to become crucial to the instability , which leads us to interpret it as an advective - acoustic cycle . in a treatment of @xmath8 gas in core collapse supernovae
, @xcite only find advective - acoustic instability for @xmath154 , with the growth rate increasing out to @xmath155 .
@xcite are only able to demonstrate the operation of the advective - acoustic cycle for @xmath156 , due to the nature of the approximations involved , but conjecture that it should also operate at lower @xmath125 .
our models agree with their demonstrations at large shock radii , but suggest that at smaller shock radius , the instability should change to resemble that observed by @xcite .
the precise mechanism of instability probably becomes more significant in the rotating case .
the advective - acoustic instability must grow preferentially along the rotation axis , where the postshock advection is not affected by the rotation .
by contrast , we conjecture in section 3.3 that rotation should enhance the growth of trapped sound / gravity waves principally in the plane of rotation , leading to stronger growth in directions perpendicular to the rotation axis
. such effects may have important consequences for pulsar natal kicks , if these derive from hydrodynamic kick mechanisms .
this work has been supported by the chandra gi program , the nasa ltsa and apra programs , and by basic research funds of the office of naval research .
i acknowledge a scientific discussion with ethan vishniac , and an extremely careful and constructive reading of the paper by an anonymous referee .
here we summarize the model for the postshock flow in spherical accretion used by @xcite .
the flow is given by the bernoulli equation @xmath157 and just behind the shock itself the jump conditions give the velocity , density and pressure as @xmath158 evaluating @xmath159 from the jump conditions and substituting gives @xmath160 where @xmath161 and @xmath162 . neglecting the first term ( i.e the kinetic energy ) , @xmath163 and hence @xmath164 .
these results are exact for @xmath165 . in this approximation ,
the density decreases monatonically with increasing radial distance , and @xmath166 defined above is @xmath167 .
we note that for @xmath99 , close to the shock , the density may increase with increasing radial distance , i.e. a postshock rarefaction exists . for @xmath46
, this region extends between @xmath168 , and disappears for @xmath169 .
we recapitulate and modify very slightly the approach of @xcite .
let @xmath170 be the amplitudes of the @xmath128 oscillations of the accretion shock . then at the accretion shock , @xmath171 , which comes from the radial variation of the shock ram pressure .
we follow @xcite and neglect the term in @xmath172 , which renders the perturbation isothermal . @xcite
also comment that the term in @xmath172 is smaller by a factor of a few than the term in @xmath173 in determining the shock ram pressure .
hence @xmath174 where @xmath175 , and the subscript @xmath176 indicates that this is evaluated at the outer shock . the interior boundary , and hence the boundary conditions here , are less well defined in our model .
we are assuming that the region where strong neutrino cooling sets in and causes the accreting matter to decouple from the hot postshock flow will give a boundary from which waves can reflect .
@xcite , considering a plane parallel shock , assume a constant pressure boundary condition , @xmath184 .
we assume something similar ; @xmath185 we do not take the local value of @xmath166 in this boundary condition , but take the same value as assumed above at the forward shock , modified by a constant @xmath112 which we vary to find the best match to the @xmath52 stability . in any case ,
@xmath186 as it must be , because the background density increases significantly over that predicted by the bernoulli model due to radiative cooling by neutrino emission .
we prefer to cast the interior boundary condition in terms of @xmath166 and @xmath15 evaluated at the outer shock , because of the algebraic simplification it produces , and the somewhat easier conditions required to keep the @xmath52 mode stable .
it is clear from the work of various authors ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) that this must be true .
these last two references , in particular , include considerably more physics , and are able to predict the shock radius , rather than just specifying it as we do , but they only treat radial stability .
consequently we have @xmath187\right\ } + b_-\left\{1-{1\over \omega ^2al}\left[\left({gm\over r}-c_s^2\right)\lambda _ { -i}^{\prime}\right]\right\}=0,\ ] ] where the @xmath188 are evaluated with @xmath189 at the inner boundary , but otherwise with @xmath166 and @xmath15 at the outer shock . @xmath191-\omega ^2{l\left(l+1\right)\over r^2 } v_su_r\left[1-\beta ^q\right ] -\omega ^2\left({gm\over r}-c_s^2\right){1\over al}\left[\lambda _ { + i}^{\prime } -\lambda _ { -i}^{\prime}\beta ^q\right]\cr & -\omega ^2\left({gm\over r}-c_s^2\right)\left({1\over l}+{5\over 2r}\right ) \left[\lambda ^{\prime}_{-s}-\lambda ^{\prime}_{+s}\beta ^q\right ] + { l\left(l+1\right)\over r^2}v_su_r\left({gm\over r}-c_s^2\right ) { 1\over al}\left[\lambda _ { + i}^{\prime}-\lambda _ { -i}^{\prime}\beta ^q\right]\cr & + \left({gm\over r}-c_s^2\right)^2 { 1\over al}\left({1\over l}+{5\over 2r}\right ) \left[\lambda _ { + i}^{\prime}\lambda _ { -s}^{\prime}-\lambda _ { + s}^{\prime}\lambda _ { -i}^{\prime}\beta ^q\right ] = 0\end{aligned}\ ] ] blondin , j. m. , mezzacappa , a. , & demarino , c. 2003 , apj , 584 , 971 blondin , j. m. , & mezzacappa , a. 2006 , , 642 , 401 burrows , a. , livne , e. , dessart , l. , ott , c. d. , & murphy , j. 2006 , , 640 , 878 burrows , a. , & goshy , j. 1993 , , 416 , l75 chandrasekhar , s. 1961 , hydrodynamic and hydromagnetic stability ( new york : dover ) foglizzo , t. 2002 , , 392 , 353 foglizzo , t. , scheck , l. , & janka , h .- th .
2006 , , 652 , 1436 foglizzo , t. , galletti , p. , scheck , l. , & janka , h .- th .
2007 , , 654 , 1006 galletti , p. , & foglizzo , t. 2005 , in proceedings of the sf2a meeting , 17 - 30 june 2005 , strasbourg , edps conference series in astronomy and astrophysics , ed .
f. casoli , t. contini , h. m. hameury & l. pagani , astro - ph/0509635 herant , m. , benz , w. , & colgate , s. 1992 , , 395 , 642 herant , m. 1995 , physics reports , 256 , 117 janka , h .- th . 2001 , , 368 , 527 kifonidis , k. , plewa , t. , scheck , l. , janka , h .- th . , & mller , e. 2005 , , 453 , 661 nakayama , k. 1992 , , 259 , 259 ohnishi , n. , kotake , k. , & yamada s. 2006 , , 641 , 1018 scheck , l. , plewa , t. , janka , h .- th . , kifonidis , k. , & mller , e. 2004 , , 92 , 011103 - 1 scheck , l. , kifonidis , k. , janka , h .- th .
, & mller , e. 2006 , , 457 , 963 velikovich , a. l. , zalesak , s. t. , metzler , n. & wouchuk , j. g. 2005 , phys . rev .
e , 72 , 046306 vishniac , e. t. , & ryu , d. 1989 , apj , 339 , 917 vishniac , e. t. 1983 , , 274 , 152 yamasaki , t. , & yamada , s. 2005 , , 623 , 1000 yamasaki , t. , & yamada , s. 2006 , , submitted , astro - ph/0606581 lrrrrrr & & + @xmath125 & @xmath52 & @xmath0 & @xmath1 & @xmath52 & @xmath0 & @xmath1 + 2 & @xmath196 & @xmath197 & & @xmath196 & @xmath198 & @xmath199 + & & @xmath200 & & @xmath201 & @xmath202 & @xmath203 + + 3 & @xmath204 & @xmath205 & & @xmath206 & @xmath207 & @xmath208 + & & @xmath209 & & @xmath210 & @xmath211 & @xmath212 + + 5 & @xmath213 & @xmath214 & & @xmath215 & @xmath216 & @xmath217 + & & @xmath218 & & @xmath219 & @xmath220 & @xmath221 + + 10 & @xmath222 & @xmath223 & & @xmath224 & @xmath225 & @xmath226 + & & @xmath227 & & @xmath228 & @xmath229 & @xmath230 + + @xmath231 & & & & @xmath216 & @xmath232 & + & & & & @xmath233 & @xmath229 & + lrrrrrr & & + @xmath125 & @xmath52 & @xmath0 & @xmath1 & @xmath52 & @xmath0 & @xmath1 + 2 & @xmath234 & @xmath235 & & @xmath236 & @xmath237 & @xmath238 + & & @xmath239 & & @xmath240 & @xmath241 & @xmath242 + + 3 & @xmath243 & @xmath244 & & @xmath245 & @xmath246 & @xmath247 + & & @xmath248 & & @xmath249 & @xmath250 & @xmath251 + + 5 & @xmath252 & & & @xmath253 & @xmath254 & @xmath255 + & & & & @xmath256 & @xmath257 & @xmath258 + + 10 & @xmath259 & & & @xmath260 & @xmath261 & @xmath262 + & & & & @xmath263 & @xmath264 & @xmath265 + + @xmath231 & & & & @xmath266 & @xmath267 & + & & & & @xmath268 & @xmath269 & + lrrrrrr & + @xmath125 & 0.0 & 0.05 & 0.1 & 0.2 & 0.3 & 0.4 + 2 & -0.490 & -0.498 & -0.502 & -0.511 & -0.520 & -0.527 + & @xmath241 & @xmath270 & @xmath200 & @xmath271 & @xmath272 & @xmath273 + + & 0.490 & 0.489 & 0.484 & 0.473 & 0.461 & + & @xmath241 & @xmath274 & @xmath275 & @xmath276 & @xmath277 & + + + 10 & 0.401 & 0.430 & 0.433 & 0.441 & 0.449 & 0.456 + & @xmath264 & @xmath278 & @xmath279 & @xmath271 & @xmath280 & @xmath281 + + & -0.401 & -0.425 & -0.430 & -0.464 & -0.483 & -0.495 + & @xmath264 & @xmath251 & @xmath282 & @xmath283 & @xmath284 & @xmath285 +
|
we explore an analytic model of the accretion shock in the post bounce phase of a core - collapse supernova explosion .
we find growing oscillations of the shock in the @xmath0 and @xmath1 modes , in agreement with a variety of existing numerical simulations . for modest values of the ratio of the outer accretion shock to that of the inner boundary to the shocked flow , the instability appears to derive from the growth of trapped sound waves , whereas at higher values , postshock advection clearly plays a role .
thus the model described here may relate to the different mechanisms of instability recently advocated by blondin & mezzacappa , and by foglizzo and collaborators .
|
surface effects are of fundamental importance in liquid crystal physics@xcite . chemical and topographical treatments induce a preferred molecular orientation that , due to elasticity , tends to align the material in the bulk .
patterned surfaces produce a wealth of phenomena , including multistable alignment@xcite and arbitrary control of the alignment orientation and associated anchoring energy@xcite .
highly confined liquid crystals , in channels , emulsion droplets , or polymer dispersed systems are also strongly influenced by the interaction of the lc material with the adjacent medium . as liquid crystals are deployed beyond display applications@xcite , the exploration of their behavior in complex geometries has become increasingly important .
almost ubiquitously , the shape of their enclosure is regarded as a rigid body and the equilibrium configuration of the liquid crystal found by minimizing an appropriate energy functional , a challenging enough task given the multiple competing physical effectselasticity , surfaces , applied and internal fieldsinvolved .
if the external environment is not rigid , the minimization problem also requires solving for the shape of the body .
it is this latter category that shall be discussed in this paper , which we refer to as shape - order problems since they require simultaneous solution of both shape and ordering .
one situation where the variable shape problem may occur is when droplets or shells of liquid crystalline phase are dispersed in an isotropic fluid . for thermotropic liquid crystals ,
the nematic - isotropic interfacial energy is much larger than the elastic energy , and such droplets remain approximately spherical .
topology restricts the number and type of defects on the surface@xcite , which can be used to template other objects @xcite .
for lyotropic materials , however , the effective energy of the phase boundary is weak and highly deformed droplets , known as _
tactoids , _ are observed .
observed in dilute vanadium pentoxide by zocher @xcite in 1925 and assemblies of viral particles bernal and fankuchen @xcite in 1941 , there has been a resurgence of interest in the last couple of decades and a rich variety of morphologies observed . in an applied electric field , paramagnetic tactoids
are deformed strongly as a flow field develops internally @xcite .
while most tactoids observed possess a characteristic eye shape , a rare triangular version has been seen using an actin system with crosslinkers @xcite .
non simply - connected shapes are commonly seen , with the isotropic inclusions themselves forming negative tactoids @xcite and examples with many inclusions and defects have been seen in chromonic liquid crystals @xcite . while for most tactoids the anchoring is parallel to the interface , elongation has also been seen in small droplets with perpendicular ( homeotropic ) anchoring @xcite .
tactoid structures have been shown to act as sensors via chirality amplification @xcite and can be used to guide motile bacteria @xcite .
they are also valuable architectural elements of self assembly , for example providing nucleation sites for growth of the smectic phase @xcite .
theoretical treatments of tactoids are more sparse due to the difficulty of the combined shape - order minimization problem .
chandrasekhar argued for a taxonomy of four possible shapes @xcite : elliptical , rectangular , a rounded rectangle and the eye shape observed in earlier experimental work .
kaznacheev proposed an elastic theory @xcite for the size - dependence of the equilibrium shape , determining that larger tactoids become more spherical in agreement with experiments @xcite .
prinsen argued that the aspect ratio of the tactoid depends primarily on the anchoring strength while the configuration of the director field depends on the ratio of anchoring to elastic constant @xcite . in a second paper , they showed that differences between elastic constants can lead to several different stable director configurations and predicted a phase diagram for the ground state @xcite .
simulation based approaches have been used to study tactoids .
bates used a monte carlo approach to show the tendency of droplets to elongate @xcite .
recently , a coarse grained simulation @xcite of @xmath0 droplets showed that the liquid crystal may spontaneously undergo a symmetry breaking transition to a chiral configuration in a tactoid , despite being composed of achiral materials . a second class of materials that undergo dramatic shape change
are liquid crystal elastomers ( lces ) , soft polymeric materials consisting of a cross - linked network and rod - like mesogens that spontaneously align .
these materials combine the properties of rubbery materials and liquid crystals @xcite .
their chief distinctive feature is the coupling between strain and orientation order : stretching a lce material can rotate the mesogens and vice versa @xcite .
they may be induced to undergo large deformations by external stimuli @xcite including electric fields @xcite , induced swelling by absorption of solvent molecules @xcite and light @xcite .
these fascinating properties suggest many applications requiring flexibility or a large response to a modest stimulus , such as soft actuators , artificial muscles , self - folding structures , smart materials , flexible displays , and flexible microgenerators @xcite . in light of the strong contemporary interest in these materials ,
some groups have begun to produce high quality numerical solvers to simulate their behavior@xcite . in light of the above applications in liquid crystal physics , as well as others in soft matter more broadly
, there is a need for generic tools to solve shape - order problems .
excellent tools for shape minimization problems , such as the _ _
surface evolver__@xcite , exist but are limited to systems without order and to problems involving low dimensional manifolds .
unfortunately , these extensions make the problem much more challenging because the numerical representationi.e .
the mesh for a finite element approachmust be co - evolved during the minimization along with the boundary . in this paper , we discuss some initial explorations and propose a regularization scheme to address these difficulties .
the paper is structured as follows : in section [ sec : model ] the numerical challenges in this class of problem are discussed and a regularization strategy proposed . in section [ sec : applications ] we apply this algorithm to two problems of increasing difficulty : in subsection [ sub : a - flexible - capacitor ] a flexible capacitor for which an analytical solution is available and in subsection [ sub : tactoids ] liquid crystal tactoids . conclusions and prospects for future work are presented in section [ sec : conclusion ] .
as an illustration of the challenges associated with shape minimization problems , consider perhaps the simplest possible example : minimize the length of a loop subject to a constant enclosed area of unity .
the problem may be solved in discrete form by defining linear elements between adjacent points @xmath1 with @xmath2 $ ] and imposing a periodic condition @xmath3 .
the program is then , @xmath4 by differentiating the target and constraint functionals , we obtain generalized forces @xmath5 and @xmath6 representing the generalized force on vertex @xmath7 exerted by the length functional and area constraint respectively . from an initial configuration @xmath8 ,
the program ( [ eq : program ] ) is solved by taking gradient descent steps , @xmath9 where the component of @xmath10 that tends to change the area is subtracted off .
the stepsize @xmath11 is determined by a line search , i.e. performing a 1d minimization of the energy functional with respect to @xmath11 .
after each step , the solution is rescaled to have unit area .
although simple , this algorithm is at the core of programs like _ surface evolver _ whose complexity arises from the wide variety of energy functionals available . to obtain a solution , the user performs a succession of relaxation steps from an initial guess to find a solution and is able to interactively refine the mesh .
minimization of the length of a loop at constant area . * a * intermediate steps at iterations @xmath12 , @xmath13 , @xmath14 , @xmath15 and @xmath16 . *
b * length of the line minus the analytical perimeter @xmath17 as a function of iteration .
* c * stepsize determined from linesearches as a function of iteration . ] if the initial guess is far from the solution , however , problems can occur . in fig .
[ fig : noreg ] we display a minimization starting from an initial elliptical loop of aspect ratio @xmath18 and unit area .
the solution converges toward the correct circular solution , but very slowly . in fig .
[ fig : noreg]a , note that mesh points that started at the ends of the ellipse become bunched together .
hence , even after @xmath16 iterations , only a relatively poor approximant to the circle has been found . in fig .
[ fig : noreg]b the quantity @xmath19 is reported on log - linear scales as a function of iteration number , where @xmath20 is the length of the loop and @xmath21 is the analytical perimeter of a circle with unit area . for a good initial guess , this quantity should converge exponentially , i.e. @xmath22 with some rate @xmath23 and hence would appear on this plot as a straight line ; the line observed shows that the convergence is subexponential .
the reason for this is apparent from a plot of the stepsize @xmath11 as a function of iteration , shown in fig .
[ fig : noreg]c , that shows that the algorithm is forced to take smaller and smaller steps as the points become bunched together .
bunching arises because the solution of the program ( [ eq : program ] ) possesses a null space with respect to the position of the vertices .
suppose all vertices other than @xmath1 are fixed , but the @xmath7th vertex is perturbed @xmath24 where @xmath25 is a unit vector .
expanding the length of a loop as a series in @xmath26 we obtain , @xmath27 the change in length can be cancelled to linear order if the direction of the perturbation is chosen to be perpendicular to the second factor in the dot product .
the factor in question is , by definition , the generalized force acting on the vertex @xmath7 . at equilibrium , this must balance the force due to the area constraint , which always points along the normal .
hence , for the solution of ( [ eq : program ] ) , the nullspace lies for each vertex locally along the tangent line .
the existence of the nullspace reveals that the problem is underconstrained , and this characteristic is generic for shape evolution problems .
for instance , if the interior of the loop were to be meshed as well , the position of these additional mesh points does not enter the program ( [ eq : program ] ) , and so they would not be moved .
moreover , it is an issue fundamentally associated with the discrete system as the continuous functional possesses a unique solution .
minimization of the length of a loop with regularization . * a * intermediate steps at iterations @xmath12 , @xmath28 , @xmath29 , @xmath30 and @xmath14 .
* b * length of the line minus the analytical perimeter @xmath17 as a function of iteration .
* c * value of the regularization functional as a function of iteration .
* d * stepsize @xmath11 and * e * regularization stepsize @xmath31 as determined from linesearches by iteration . ] to solve this problem , _ surface evolver _ permits the user to merge nearby mesh points as they become bunched , and to create new mesh points by splitting long elements .
this is effective , but means that the number of mesh points can not be controlled as easily and requires discrete changes to the mesh . here
, we propose a different solution that allows continuous regularization of the mesh .
the idea is to supplement the minimization problem with an auxiliary functional that captures our intuition of mesh quality . for the present problem ,
a suitable functional is , @xmath32\label{eq : regularization}\ ] ] where @xmath33 , i.e. the mean length of the two linear elements at point @xmath7 . this functional favors equally spaced mesh points . by differentiating the regularization functional
, a corresponding generalized force @xmath34 may be obtained .
it is essential that the regularization functional does not interfere with the solution of the program ( [ eq : program ] ) , so the regularization functional is minimized separately in the nullspace of the target functional , i.e ensuring that @xmath35 .
following each gradient descent step eq .
( [ eq : gradientdescent ] ) , a regularization step is then taken , @xmath36 where the regularization stepsize @xmath31 is found by performing a linesearch on the regularization functional ( [ eq : regularization ] ) . execution of this algorithm on the same elliptical initial configuration as above yields a plot displayed in figure [ fig : reg]a that has converged onto the correct circular solution with the mesh points are approximately evenly spaced as desired .
inspection of the length as a function of iteration @xmath19 in figure [ fig : reg]b shows that the algorithm converges exponentially , within around @xmath14 iterations , even from this poor initial guess .
while the new algorithm involves approximately twice the work per iteration , i.e. an additional force evaluation and line search , this is clearly a considerable improvement over the naive algorithm .
the value of the @xmath37 is shown in figure [ fig : reg]c indicating that the solution is not fully converged with respect to the regularization functional .
while the stepsizes @xmath11 and @xmath38displayed in figure [ fig : reg]d and [ fig : reg]e respectivelyare very different in value and rather noisy , it is apparent that they remain approximately constant with iteration , explaining the fast convergence .
the strategy developed for the very simple problem considered in the present section is readily usable to any combination of target functionals , constraints and regularization functionals .
in particular , it is also applicable to problems where the minimization must be performed not only on the shape of the domain but on field quantities defined upon it ; it is to such a problem that we turn in the following section .
we now examine a more complex problem , a cylindrical capacitor made from flexible electrodes shown schematically in the inset of fig .
[ fig : conformal]a . normal capacitors with rigid plates held at fixed potential difference experience a force that tends to separate the plates . here , the capacitor is initially circular in the absence of an electric field and tends to elongate in the vertical direction when a potential difference is applied to the terminals .
the device is filled with an incompressible fluid dielectric and so maintains a constant area .
the energy functional contains three terms , a line tension that minimizes droplet circumference , the electrostatic energy resulting from the electric field , and an anchoring energy used to enforce a boundary value @xmath39 without discontinuities on the electric potential , @xmath40 this is to be minimized on a 2d domain @xmath41 with respect to the electric potential @xmath42 and the shape of the boundary @xmath43 , subject to fixed area , @xmath44 the parameters @xmath45 , @xmath46 and @xmath47 represent the surface tension , anchoring and electrostatic energies respectively ; @xmath48 is a length of the order of the size of the droplet .
note that this problem is easily transformed into a liquid crystal problem by replacing the electric potential with the director angle @xmath49 , modifying the elastic term to include variable elastic constants and replacing @xmath50 with a spatially varying easy axis .
the energy ( [ eq : energy-1 ] ) can be made dimensionless by dividing through by @xmath51 , and making a change of variables @xmath52 , @xmath53 where @xmath54 and the primed variables correspond to the coordinate system scaled by @xmath48 .
there are two dimensionless parameters , @xmath55 , which represents the ratio of line tension to applied field , and @xmath56 , which represents the relative importance of anchoring and electric forces .
the problem is discretized using a conventional finite element approach : the domain is represented by an initially circular mesh with triangular elements .
the potential @xmath42 is represented on the vertices and a linear interpolation is used on the interior of the triangles , hence @xmath57 is constant in each triangular element . using this discretization ,
an estimate of the functional ( [ eq : energy - nodim ] ) can be calculated for a given configuration .
formulae for these expressions are available in standard finite element textbooks ; we do not supply them here . from these , it is possible to obtain the generalized force @xmath58 acting on each vertex of the mesh , and also a scalar generalized force @xmath59 acting on the field .
boundary vertices experience a contribution to the force from all terms in ( [ eq : energy - nodim ] ) ; the interior vertices only experience forces due to the electrostatic term . because of this , it is clear that the problem is underconstrained with respect to the interior vertices .
convergence to the conformal mapping solution for a flexible capacitor with @xmath60 and @xmath61 . * a * norm of the difference between simulation and conformal mapping results for the field ( brown ) and shape ( cyan ) for simulations with ( solid ) and without ( dashed ) the regularization routine as the simulation reaches convergence .
comparison plots of the * b * initial and * c * final states of the simulation .
the outer borders of the cm ( cyan ) and simulation ( brown ) solutions are shown , and points interior to both shapes are shaded by the difference between the field solutions at those points . ] to remedy this , two auxiliary functionals are introduced to regularize the problem .
the first , @xmath62^{2},\label{eq : reg1}\ ] ] prefers to maintain equal energy density among all triangles ; the second , @xmath63,\label{eq : reg2}\ ] ] favors triangles with equal angles . in ( [ eq : reg1 ] ) and ( [ eq : reg2 ] ) , @xmath64 is the interior angle of face @xmath7 at vertex @xmath65 and @xmath66 is the total number of triangles composing the mesh . using the algorithm described in the previous section , gradient descent
is performed by successive line searches on the target and auxiliary functionals . to validate our algorithm
, we take advantage the fact that an approximate analytical solution is available from conformal mapping , at least in the limit of small deformations .
we use a one - parameter area conserving conformal map @xmath67 of the form , @xmath68 to map the undeformed capacitor , i.e. the unit disk , parametrized by complex coordinates @xmath69 , to a target domain with coordinates @xmath70 .
the electric potential is also represented in the source domain with a complex function @xmath71 , @xmath72 with coefficients @xmath73 . by translating the energy functional ( [ eq : energy - nodim ] ) into complex form , inserting the map eq .
( [ eq : ourmap-1 ] ) and potential eq .
( [ eq : usoln-2 ] ) , and performing the integrations , we obtain an expression for the energy that is quadratic in @xmath74 and @xmath73 and readily minimized numerically .
full details of the derivation are provided in the appendix .
shape of a flexible capacitor with applied electric field .
deep cyan ( brown ) regions indicate a field value of @xmath75 ( @xmath76 ) .
all cases were started with a circular shape and a field that changes linearly along the vertical .
solutions below the dashed line converge on a `` pinched '' configuration that prefers to separate . ] for a test case with parameters @xmath60 and @xmath61 , we show in fig .
[ fig : conformal]a a comparison between the conformal mapping solution and the finite element solution as a function of iteration for the @xmath77 norm of the difference in the field , @xmath78 and the @xmath77 norm of the difference in shape , i.e. the area between the finite element boundary and the conformal mapping boundary .
both quantities converge to a fixed nonzero value because the conformal mapping solution is itself approximate .
additionally , we show these quantities with and without the regularization scheme active . note that without regularization , the method fails to converge due to bunching of the mesh points . in figure fig .
[ fig : conformal]b and fig .
[ fig : conformal]c , a local comparison between the conformal mapping and finite element solution is shown for the start and end configurations , showing a good agreement . in figure
[ fig : flexiblecapa ] we display configurations that result from these minimizations as a function of @xmath79 and @xmath80 including those that exhibit large deformations outside the validity of the conformal mapping solution . for @xmath81 ,
the capacitors tend to remain fairly circular ; as @xmath79 is increased the solution conforms more closely to the boundary conditions .
conversely , for @xmath82 significant deformations of the capacitor are allowed ; the shape elongates vertically at the expense of horizontal invaginations . beneath the dashed line in fig .
[ fig : flexiblecapa ] , the deformation is so extreme that the invaginations meet ; if allowed to , the capacitor would seperate into two disconnected domains .
shape of a nematic droplet .
all cases were started with a circular shape and a director field with virtualized defects on the horizontal axis .
depending on the anchoring strength , these defects may spread out to infinity or may move into the droplet . ]
we now adapt the simulation developed in the previous section to study nematic tactoids .
parametrizing the director as @xmath83 , the energy functional contains line tension , elastic and anchoring terms respectively , @xmath84da\\ & & + w\int_{\partial d}(\theta-\theta_{e})^{2}dl\end{aligned}\ ] ] where @xmath85 and @xmath86 are the splay and bend elastic constants respectively .
this is to be minimized subject to an area constraint as before , and we divide through by @xmath87 and make the change of variables @xmath52 in order to non - dimensionalize the energy .
when @xmath88 , the elastic energy density reduces to @xmath89 and the problem is almost exactly the same as that discussed in the previous section , with one important difference : the easy axis @xmath90 is not fixed on the boundary , but dynamically determined at each iteration relative to the local estimate of the tangent plane .
even so , we use the same regularization functionals and algorithm , only inserting a modified force functional for the elastic anisotropy . in figure
[ fig : shape - of - a ] , a set of equilibrium shapes is shown for @xmath88 as a function of @xmath91 and @xmath92 . in all cases , the minimization was started from a circular initial configuration , with virtual defects outside the tactoid .
stronger anchoring causes the nematic to better conform to the boundary condition , effectively moving the virtual defects closer to the surface of the droplet .
weaker line tension , on the other hand , permits more elongation of the droplet .
simulations of nematic droplets with weak line tension ( @xmath93 ) .
line segments indicate the director orientation at those points . ]
the effect of elastic anisotropy is shown in fig .
[ fig : simulations - of - nematic ] , which displays four equilibrium configurations as a function of @xmath94 and @xmath92 .
it is clear that significant shape deformation is only achieved with the introduction of elastic anisotropy .
stronger elastic constants lead to a fairly homogeneous field with excluded defects , while weak elastic constants allow the field the better follow the anchoring condition , resulting in a pair of boojum defects .
simulations of nematic droplets with weak line tension ( @xmath93 ) with interior holes .
the holes in the droplets form negative tactoids that align either * a * perpendicular or * b * parallel to the orientation of the outer tactoid depending on the ratio of @xmath46 to @xmath85 . ] it is straightforward to apply our technique to more complicated geometries , such as non - simply connected domains . in figure
[ fig : holes ] , we display two equilibrium configurations with negative tactoid holes on the interior . these were started from an initially annular configuration with an area constraint additionally applied to the hole .
solutions with the interior negative tactoid aligned perpendicular to ( fig .
[ fig : holes]a ) or parallel to ( fig .
[ fig : holes]b ) with the exterior were found .
we note that the array of achievable tactoid shapes fit within the classes defined in @xcite .
additionally , the shape trends displayed in fig .
[ fig : shape - of - a ] closely mimic those proposed by lishchuk @xcite .
unique to our results is the ability to predict tactoid shapes with more complex boundaries .
this allows for better agreement with experimentally produced cusped ellipsoids @xcite , as well as tactoids containing isotropic inclusions @xcite .
excitingly , experimental evidence of the latter tends to produce negative tactoids that align parallel to the exterior tactoid , in agreement with one of our classes .
the present work describes our initial exploration of a very challenging class of problems , where the equilibrium configuration of a system must be found by minimizing some functional with respect to internal fields as well as the shape of the domain .
gradient descent methods are effective for some subset of these problems , and form the basis of programs like _ surface evolver _ , but require management of mesh quality during the minimization .
some of the challenges were illustrated with a toy example of minimizing the length of a loop at constant area , showing that mesh points become bunched during the minimization and prevent convergence on the correct solution .
a regularization strategy was shown to be effective in overcoming these difficulties whereby the target functional was supplemented with an auxiliary functional to promote equally - spaced mesh points .
the same approach was used to develop a solver for the equilibrium configuration of a flexible capacitor , as well as that of a 2d nematic tactoid . exploring the space of shapes as a function of the surface tension and anisotropic elastic constants , we found structures similar to those experimentally observed , as well as some others predicted in the literature .
we are presently working to generalize our program to work with higher dimensional manifolds , arbitary combinations of fields defined upon them , a broader selection of energy functionals and dynamics problems . _
the authors wish to thank the tufts graduate school of arts and sciences for provision of a summer scholarship for adb .
both tja and adb are grateful for funding from a cottrell award from the research corporation for science advancement . _
the integrals in ( [ eq : energy-1 ] ) and ( [ eq : areaconstraint ] ) must be translated into complex notation using the map @xmath67 . to do this , we use familiar properties of complex functions , i.e that @xmath95 plays the role of a metric , so that lengths transform under the action of @xmath70 like @xmath96 and areas transform like @xmath97 . parametrizing the boundary of the unit circle as @xmath98 ,
the program in complex notation is to minimize , @xmath99 with respect to @xmath67 and the real function @xmath71 and subject to the constraint , @xmath100 where we used the property that the electrostatic energy is invariant under a conformal mapping .
the conformal map @xmath67 and the scalar potential @xmath42 are expanded as a power series in @xmath69 , @xmath101 and @xmath102 where positive powers only have been retained to ensure regularity at the origin @xmath103 and the imaginary part is chosen due to the symmetry of the boundary condition . substituting @xmath104 into ( [ eq : usoln ] )
, we obtain @xmath105 and turn to the issue of evaluating each term in the energy functional , as well as the area constraint , in order of difficulty .
the first term we consider is the electrostatic energy , which is after all invariant under the mapping . inserting the solution ( [ eq : usoln-1 ] ) into the second term in ( [ eq : complexenergy ] ) yields , @xmath106 next
, we turn to the area constraint . inserting ( [ eq : zsoln ] ) into ( [ eq : areaconstraintcomplex ] ) and integrating , @xmath107 at this point , we shall specialize the problem to only consider the first two odd terms of @xmath67 , @xmath108 where the first is the identity mapping and the second represents the first nonzero term of the deformation induced by the field .
this approximation is equivalent to assuming that the capacitor is only weakly deformed from the circular reference shape , which is true if the line tension dominates in the energy functional .
we therefore create a one parameter family of maps that preserves the area of the unit disk , @xmath109 the line tension term in ( [ eq : complexenergy ] ) is highly nonlinear . inserting the restricted map ( [ eq : ourmap ] ) and evaluating using @xmath110 on the boundary
, we obtain @xmath111 which may be evaluated in terms of the complete elliptic integral @xmath112 , the anchoring term also contains @xmath96 , and hence is also nonlinear .
we follow a similar strategy to the line tension , expanding this in powers of @xmath74 on the boundary @xmath110 and retaining only terms up to quadratic order , @xmath117 inserting this , together with the series solution for @xmath42 , eq . ( [ eq : usoln-1 ] ) , into the anchoring energy term in ( [ eq : complexenergy ] ) and performing the integration yields an expression of the form , @xmath118 where @xmath119 , @xmath120 and @xmath121 are scalar , vector and matrix quadratic functions of @xmath74 .
@xmath122 is a symmetric band - diagonal matrix with a checkerboard structure , @xmath123 if @xmath124 . for the sake of brevity
, we do not list these elements here ; they are available in a _ mathematica _
notebook provided as supplementary material .
all terms in the free energy ( [ eq : complexenergy ] ) have now been evaluated in terms of the parameter @xmath74 that defines the area - preserving conformal map ( [ eq : ourmap ] ) as well as the coefficients of the field ( [ eq : usoln-1 ] ) .
expressions for line tension , electric energy and anchoring are quoted in equations ( [ eq : energylt ] ) , ( [ eq : energyel ] ) and ( [ eq : energyanchoring ] ) respectively .
it may therefore be seen that the total free energy can be put into the form , @xmath125 where @xmath126 , @xmath127 and @xmath128 are quadratic functions of @xmath74 .
the scalar @xmath126 contains contributions from the anchoring and line tension terms ; @xmath127 contains only contributions from the anchoring term ; @xmath128 contains electrostatic ( on the diagonal ) and anchoring contributions .
the special structure arises because the series expansion in @xmath70 is low order , while the expansion in @xmath42 was carried out to arbitrary order and is valid for the situation where the shape deformation is small .
a higher order expansion in @xmath70 will yield @xmath126 , @xmath127 and @xmath128 that are also variables of higher order parameters @xmath129 .
furthermore , series expansion of @xmath96 limits the order of @xmath126 , @xmath127 and @xmath128 to being quadratic in @xmath74 ; this expansion could easily be extended further .
minimization of eq .
( [ eq : fquadratic ] ) with respect to @xmath74 and @xmath130 yields the following system of equations , @xmath131 where we note that @xmath7 is a free index in the second equation .
these equations are efficiently solved by the nested iteration scheme , @xmath132 for sufficiently small stepsize @xmath11 and a single initial seed value @xmath133 .
a _ mathematica _
notebook that implements this scheme is presented as supplementary material .
|
we consider a class of liquid crystal free - boundary problems for which both the equilibrium shape and internal configuration of a system must simultaneously be determined , for example in films with air- or fluid - liquid crystal interfaces and elastomers .
we develop a finite element algorithm to solve such problems with dynamic mesh control , achieved by supplementing the free energy with an auxiliary functional that promotes mesh quality and is minimized in the null space of the energy .
we apply this algorithm to a flexible capacitor , as well as to determine the shape of liquid crystal tactoids as a function of the surface tension and elastic constants .
these are compared with theoretical predictions and experimental observations of tactoids from the literature .
|
an important property of the clusters of a normal phase in superconductor consists in their capability to trap a magnetic flux . by virtue of their capacity to hold the vortices from moving under the action of the lorentz force , such clusters can act as effective pinning centers @xcite-@xcite .
this feature is used widely in the making new composite superconducting materials of high current - carrying capability @xcite , @xcite .
the morphological characteristics of clusters of a normal phase exert an appreciable effect on magnetic flux dynamics in superconductors , especially when the clusters have fractal boundaries @xcite-@xcite . in the present work the geometric probability properties of such fractal clusters
are considered in detail , and their influence on the dynamics of trapped magnetic flux and critical currents is analyzed .
the further consideration will be concerned with the superconductor containing inclusions of a normal phase , which are out of contact with one another .
let us assume that these inclusions are oriented in such a way that their extent along one of directions far exceeds other linear sizes .
the similar columnar defects are of most interest for creating the artificial pinning centers @xcite , @xcite-@xcite .
when such a superconducting structure is cooled below the critical temperature in the magnetic field along the direction of the longest size of these inclusions , the magnetic flux will be frozen in the normal phase clusters .
even after the external field has been turned off , the flux trapped in these clusters is kept unchanged due to the currents that are steadily circulating around them through the superconducting loops .
the distribution of the trapped magnetic flux resulting from such a magnetization in the field - cooling regime will be two - dimensional .
the similar distribution can readily be realized in the superconducting film where the normal phase inclusions are created during the growth process at the sites of defects on the boundary with the substrate in such a way that their orientation is normal to the surface of the film @xcite , @xcite , @xcite .
let us suppose that the film surface fraction covered by the normal phase is below the percolation threshold for the transfer of magnetic flux ( 50 % for 2d- percolation @xcite ) .
in this case the relative portion of superconducting phase exceeds the percolation threshold , so there is a superconducting percolation cluster in the plane of the film where a transport current can flow .
such a structure provides for effective pinning and thereby raises the critical current , because the magnetic flux is locked in finite clusters of a normal phase , and so the vortices can not leave them without crossing the surrounding superconducting space .
if the transport current is passed through the sample , the trapped magnetic flux remains unchanged as long as the vortices are still held in the normal phase clusters .
when the current is increased , the magnetic flux starts to break away from the clusters of pinning force weaker than the lorentz force created by the transport current . as this takes place , the vortices will first pass through the weak links , which connect the normal phase clusters between themselves .
such weak links form especially readily in high - temperature superconductors ( hts ) characterized by an extremely short coherence length .
various structural defects , which would simply cause some additional scattering at long coherence length , give rise to the weak links in hts .
there is a hierarchy of weak links over a wide range of scales in hts @xcite , @xcite-@xcite . at an atomic level
the weak links are formed by the structural atomic defects , primarily , by oxygen vacancies @xcite , @xcite . on a mesoscopic scale twin boundaries are mainly responsible for weak links existence @xcite-@xcite , @xcite . the twins can be spaced up to several nanometers apart , so even single crystal has fine substructure caused by twins . at last , on a macroscopic scale there are manifold structural defects which can form weak links : that may be grain or crystallite boundaries as well as barriers arising from the secondary degrading the non - stoichiometric crystal into the domains with a high and low content of oxygen @xcite-@xcite .
moreover , a magnetic field further reduces a coherence length @xcite , thus resulting in more easy weak links formation . in conventional low - temperature superconductors , which are characterized by a large coherence length
, the weak links can be formed due to the proximity effect in sites of minimum distance between the next normal phase clusters .
as soon as the transport current is turned on , this one is added to all the persistent currents , which maintain the magnetic flux trapped .
each of these currents is circulating through the superconducting loop around the normal phase cluster wherein the corresponding portion of the magnetic flux is trapped .
the loop contains weak links that join the adjacent normal phase clusters transversely to the path of the current . as the transport current
is increased , there will come a point when the overall current flowing through the weak link will exceed the critical value , so this link will turn into a resistive state .
as this takes place , the space distribution of the currents throughout the superconducting cluster is changed in such a way that the resistive subcircuit will be shunted by the superconducting paths where weak links are not damaged yet . magnetic field created by this re - distributed transport current acts via the lorentz force on the current circulating around the normal phase cluster . as a consequence
, the magnetic flux trapped therein will be forced out through the resistive weak link , which has become permeable to the vortices .
thus , whatever the microscopic nature of weak links may be , they form the channels for vortex transport where the pinning force is weaker than the lorentz force created by the transport current .
it appears that according to their configuration each normal phase cluster has its own value of the critical current of depinning , which contributes to the overall statistical distribution .
when a transport current is gradually increased , the vortices will break away first from clusters of small pinning force , and therefore , of small critical current .
thus the decrease in the trapped magnetic flux @xmath0 is proportional to the number of all the normal phase clusters of critical currents less than a preset value @xmath1 .
therefore , the relative decrease in the trapped flux can be expressed with the cumulative probability function @xmath2 for the distribution of the critical currents of clusters : @xmath3 the right - hand side of eq .
( [ probi1 ] ) is the probability that any @xmath4-th cluster has the critical current @xmath5 less than a given upper bound @xmath1 . on the other hand ,
the magnetic flux trapped into a single cluster is proportional to its area @xmath6 , so the decrease in the total trapped flux can be represented by the cumulative probability function @xmath7 for the distribution of the areas of the normal phase clusters , which is a measure of the number of the clusters of area smaller than a given value of @xmath6 : @xmath8 the distribution function @xmath7 of the cluster areas can be found by the geometric probability analysis of electron photomicrographs of superconducting films @xcite .
so , in the practically important case of thick ybco films containing columnar defects @xcite , @xcite the exponential distribution can be realized : @xmath9 where @xmath10 is the mean area of the cluster .
thus , in order to clear up how the transport current acts on the trapped magnetic flux , it is necessary to find out the relationship between the distribution of the critical currents of the clusters ( eq . ( [ probi1 ] ) ) and the distribution of their areas ( eq . ( [ proba2 ] ) ) .
the larger the cluster size , the more weak links rise from its perimeter bordering with the surrounding superconducting space , and therefore , the smaller is the critical current at which the magnetic flux breaks away from this cluster . on the basis of this simple geometric argument , let us suppose that this critical current @xmath1 is inversely proportional to the perimeter @xmath11 of the normal phase cluster : @xmath12 , because the larger the perimeter is , the higher is the probability to get a weak link there . at this point it is assumed that the weak link concentration per unit perimeter length is constant for all clusters and all the clusters of equal perimeter have the same pinning force .
also suppose that after the vortex enters the weak link , it will pass all the way between two adjacent normal phase clusters without being trapped . here , the magnetic flux is transferred by josephson vortices .
the josephson penetration depth is large enough in the considered materials , so the size of the region , where the vortex is localized , much exceeds the characteristic length of all possible structural defects that can occur along the transport channel .
thus the probability that such a vortex will be trapped in passing through a weak link under the action of the lorentz force is very small .
thus , to deal with the distribution function of eq .
( [ probi1 ] ) , the relation between perimeter and area of clusters should be studied . as has been first found in ref .
@xcite , the fractal nature of the normal phase clusters exerts an appreciable effect on the dynamics of a magnetic flux in superconductors . for fractal clusters a relation between perimeter and area
has the form : @xmath13 where @xmath14 is the fractal dimension of the cluster perimeter ( so - called coastline dimension ) @xcite .
the relation of eq .
( [ scaling4 ] ) is consistent with the generalized euclid theorem @xcite , @xcite , which states that the ratios of the corresponding measures are equal when reduced to the same dimension .
hence it follows that @xmath15 , which is valid both for euclidean ( @xmath16 ) , and for fractal clusters ( @xmath17 ) .
the fractal dimension value can be estimated by means of regression analysis of the sampling of the areas and perimeters of the normal phase clusters .
such geometric probability analysis was carried out by the procedure described in ref .
@xcite . for this purpose an electron photomicrograph of ybco film , which was similar to that published earlier in ref .
@xcite , has been scanned .
the perimeters and areas of clusters have been measured by covering their digitized pictures with a square grid of spacing @xmath18nm@xmath19 .
the results of the statistical treatment of these data are presented in table [ table1 ] .
the normal phase has occupied 20% of the total surface only , so the transport current can flow through the sufficiently dense percolation superconducting cluster .
the primary sampling has contained 528 normal phase clusters located on the scanned region of a total area of 200 @xmath20m@xmath19 .
the distribution of the cluster areas is fitted well to exponential cumulative probability function of eq .
( [ expa3 ] ) with the mean cluster area @xmath21 @xmath20m@xmath19 .
all the points of the primary sampling are marked by crosses on the plot 1 of fig .
[ figure1 ] , which shows the perimeter - area relation for the normal phase clusters .
figure [ figure1 ] also demonstrates such an important peculiarity of this relation : the scaling law of eq .
( [ scaling4 ] ) is valid in the range of almost three orders of magnitude in cluster area .
the scaling perimeter - area behavior means that there is no characteristic length scale between 0.1 @xmath20 m and 10 @xmath20 m in the linear size of the normal phase cluster .
whatever the shape and size of the clusters may be , all the points fall closely on the same straight line in logarithmic scale ; so that there are no apparent kinks or crossovers on the graph .
this enables the fractal dimension @xmath14 of the cluster perimeter to be estimated from the slope of the regression line of the form of eq .
( [ scaling4 ] ) , thus a least squares treatment of the perimeter - area data for the primary sampling gives the estimate of @xmath22 with correlation coefficient 0.929 .
this point that the found value of coastline fractal dimension differs appreciably from unity engages a great attention .
what this means is that the fractal properties of the cluster boundary are of prime importance here .
two straight lines ( 5 ) in fig .
[ figure1 ] bound the range of the slopes that the dependencies of the perimeter on the cluster area can have for any arbitrary fractal dimension .
the least slope corresponds to euclidean clusters @xmath23 , the most one relates to the clusters of the greatest possible coastline dimension , which is equal to the topological dimension of a smooth surface @xmath24 .
such a fractal dimension is inherent , for example , in peano curves , which fill the whole plane @xcite .
whatever the geometric morphological properties of clusters may be , the slope of their perimeter - area graphs will be always bounded by these two limiting lines .
when we deal with the geometric features of the normal phase clusters , we are considering the cross section of the extended columnar defects by the plane carrying a transport current . therefore , though normal phase clusters are self - affine fractals [ 30 , 31 ] , it is possible to examine their geometric probability properties in the planar section only , where the boundaries of the clusters are statistically self - similar .
next , using the relation of eq .
( [ scaling4 ] ) between the fractal perimeter and the area of the cluster , as well as our initial assumption that the critical current of each cluster is inversely proportional to its perimeter , we get the following expression for the critical current : @xmath25 , where @xmath26 is the form factor . in accordance with starting formulas of eq .
( [ probi1 ] ) and eq .
( [ proba2 ] ) , the exponential distribution of cluster areas of eq .
( [ expa3 ] ) gives rise to an exponential - hyperbolic distribution of critical currents : @xmath27 \label{exhyp5}\ ] ] where @xmath28 is the dimensionless transport current , and @xmath29 is the critical current of the resistive transition .
thus , the geometric probability properties of the normal phase clusters are responsible for main features of the critical current statistical distribution . in turn , given the distribution of eq .
( [ exhyp5 ] ) , the change in the trapped magnetic flux caused by the transport current can be found with the aid of eq .
( [ probi1 ] ) .
two of these graphs are displayed in fig .
[ figure2 ] both for the fractal clusters of coastline dimension @xmath30 found above ( curve ( 2 ) ) and for the euclidean ones ( curve ( 3 ) ) . in order to get the relationship between the dynamics of the trapped magnetic flux and geometric morphological properties of the superconducting structure ,
the sample empirical function @xmath31 of the distribution of the critical currents has been computed .
this function gives a statistical estimate of the cumulative probability function @xmath32 .
first , the empirical distribution function @xmath33 for the primary sampling of the areas of the normal phase clusters has been found .
the value @xmath34 was calculated for each order statistic as the relative number of clusters of area smaller than a given value @xmath6 .
next , the empirical distribution of the critical currents was computed for the same order statistics ( step line with open circles ( 1 ) in fig .
[ figure2 ] ) using the following transformations : @xmath35 figure [ figure2 ] shows that in the range of currents @xmath36 the empirical distribution function , which describes the morphological properties of the superconducting structure ( plot ( 1 ) ) , coincides ideally with the cumulative probability function ( curve ( 2 ) ) for the coastline fractal dimension of @xmath30 . starting with the value of the current @xmath37 the crossover is observed , resulting in that the empirical distribution function passes to the dependence for euclidean clusters @xmath23 .
this transition into the euclidean region is over at large transport currents , when the magnetic flux changes mainly for the breaking of the vortices away from the small clusters ( as the smaller clusters have the larger pinning force ) .
the observed crossover has its origin in the finite resolution capability of measuring the cluster geometrical sizes .
the distinctive feature of the topologically one - dimensional fractal curve is that its measured length @xmath11 depends on the measurement accuracy in such a way : @xmath38 , where @xmath39 is the yardstick size used to measure this length , @xmath40 is the hausdorff codimension for the euclidean 1d - space @xcite . in our case
such a fractal curve is represented by the boundary of the normal phase cluster .
that is why just the statistical distribution of the cluster areas , rather than their perimeters , is fundamental for finding the critical current distribution of eq .
( [ exhyp5 ] ) .
the topological dimension of perimeter is equal to unity and does not coincide with its hausdorff - besicovitch dimension , which strictly exceeds the unity .
therefore the perimeter length of a fractal cluster is not well defined , because its value diverges as the yardstick size is reduced infinitely . on the other hand ,
the topological dimension of the cluster area is the same as the hausdorff - besicovitch one ( both are equal to two ) .
thus , the area restricted by the fractal curve is well - defined finite quantity .
taking into account the effect of the measurement accuracy , the perimeter - area relationship of eq .
( [ scaling4 ] ) can be re - written as @xmath41 which holds true when the yardstick length @xmath39 is small enough to measure accurately all boundary of the smallest cluster in sampling .
when the resolution is deficient , the euclidean part of the perimeter length will dominate the fractal one , so there is no way to find the fractal dimension using the scaling relation of eq .
( [ perima6 ] ) .
it means that if the length of a fractal curve was measured too roughly with the very large yardstick , its fractal properties could not be detected , and therefore such a geometrical object would be manifested itself as euclidean one .
it is just the resolution deficiency of this kind occurs at the crossover point in fig .
[ figure2 ] . starting with the cluster area less than 0.023
@xmath20m@xmath19 ( corresponding to the currents of @xmath42 ) it is impossible to measure all `` skerries '' and `` fjords '' on the cluster coastlines , whereas all the clusters of area less than the size of the measuring cell ( @xmath43 @xmath20m@xmath19 that relates to the currents of @xmath44 ) , exhibit themselves as objects of euclidean boundaries @xmath45 .
this resolution deficiency can be also observed in fig .
[ figure1 ] : some crosses at its lower left corner are arranged discretely with the spacing equal to the limit of resolution ( 60 nm ) , because some marks for smallest clusters coincide for the finite resolution of the picture digitization procedure .
the coastline fractal dimension was found above by means of regression analysis of the whole primary sampling , where the very small clusters of sizes lying at the breaking point of the resolution limit were also included .
therefore , it is necessary to control how much the estimated value of the fractal dimension could be distorted by the presence of such small clusters in that sampling . for this purpose
the truncated sampling has been formed in such a way that only 380 clusters of the area greater than 0.0269 @xmath20m@xmath19 , for which the resolution deficiency is not appeared , have been selected from the primary sampling .
the corresponding points are plotted as open circles ( 2 ) in fig .
[ figure1 ] , whereas the results of statistical treatment of this truncated sampling are presented in the third column of table [ table1 ] .
the least squares estimation of these perimeter - area data gives the value of coastline fractal dimension @xmath46 with correlation coefficient 0.869 .
the slope of the regression line for the truncated sampling ( dotted line ( 4 ) in fig .
[ figure1 ] ) is slightly steeper than for the primary one ( solid line ( 3 ) ) .
it is natural , because the presence in the primary sampling of small clusters , which exhibit themselves as euclidean ones at the given resolution , leads to underrating the found magnitude of fractal dimension .
nevertheless , the values of fractal dimensions found for both samplings , virtually do not differ within the accuracy of the statistical estimation .
this is due to the high robustness of procedure of the fractal dimension estimation on a basis of the scaling relation of eq .
( [ scaling4 ] ) : all the points both for the primary sampling and for the truncated one fall on the same straight line , without any bends or breaks ( see fig . [ figure1 ] ) . at the same time
, it is necessary to note , that the empirical distribution function approach ( see fig .
[ figure2 ] ) provides the most data - sensitive technique of estimating the resolution capability required to study the fractal properties of clusters .
the found cumulative probability function of eq .
( [ exhyp5 ] ) allows us to fully describe the effect of the transport current on the trapped magnetic flux . using this function , the probability density @xmath47 for the critical current distribution
can be readily derived : @xmath48 \label{densi7}\ ] ] this function is normalized to unity over all possible positive values of critical current . the use of the exponential - hyperbolic critical current distribution of eq .
( [ exhyp5 ] ) allows us to avoid the inevitable uncertainty caused by truncation of non - physical negative values of depinning currents , as it takes place , for example , in the case of a normal distribution @xcite-@xcite . the exponential - hyperbolic distribution of eq .
( [ exhyp5 ] ) has such an important property : the function @xmath32 is extremely
flat in the vicinity of the co - ordinate origin .
it is easy to show that all its derivatives are equal to zero at the point of @xmath49 : @xmath50 therefore even the taylor series expansion in the vicinity of the origin converges to zero , instead of the quantity @xmath51 itself .
this mathematical feature has a clear physical meaning : so small transport current does not affect the trapped magnetic flux , because there are no pinning centers of such small critical currents in the overall statistical distribution , so that all the vortices are still too strongly pinned to be broken away .
as can be seen from fig .
[ figure2 ] , the change in the magnetic flux becomes appreciable after the transition into a resistive state only ( in the transport current range of @xmath52 ) . the relative change in the trapped flux @xmath53 , which can be calculated from eq .
( [ exhyp5 ] ) , also defines the density of vortices @xmath54 broken away from the pinning centers by the current @xmath55 : @xmath56 where @xmath57 is the magnetic field , @xmath58 is the magnetic flux quantum ( @xmath59 is planck s constant , @xmath60 is the velocity of light , @xmath61 is the electron charge ) .
figure [ figure2 ] also displays such a practically important property of superconducting structure containing fractal clusters of a normal phase : the fractality intensifies the magnetic flux trapping , hindering its breaking away from pinning centers , and thereby enhances the critical current which sample is capable to withstand , remaining in a superconducting state .
really , the transport current of a magnitude @xmath62 causes the 43% of the total trapped magnetic flux to break away from the usual euclidean clusters ( curve ( 3 ) ) , whereas this value is equal only to 25% for fractal normal phase clusters of coastline dimension @xmath30 ( curve ( 2 ) ) .
it is equivalent to pinning reinforcement on 73% in the latter case .
thus the pinning amplification due to the fractality can be characterized by the pinning gain factor @xmath63 which is equal to relative decrease in the fraction of magnetic flux broken away from fractal clusters of coastline dimension @xmath14 compared to the case of euclidean ones @xmath45 .
this quantity can be calculated from the following formula : @xmath64\ ] ] the characteristic dependencies of the pinning gain on a transport current as well as on fractal dimension are given in fig .
[ figure3 ] .
the highest amplification is reached when the cluster boundaries have the greatest possible fractality : @xmath65 , with the maximum of @xmath66 at transport current @xmath67 .
let us note that the pinning gain characterizes the properties of a superconductor in the range of the transport currents corresponding to a resistive state @xmath68 . at smaller current
the total trapped flux remains unchanged ( see fig . [ figure2 ] ) for lack of pinning centers of such small critical currents ,
so the breaking of the vortices away has not started yet . when a transport current becomes greater than the current of a resistive transition , some finite resistance appears , so that the passage of electric current is accompanied by the energy dissipation .
as for any hard superconductor ( type - ii , with pinning centers ) this dissipation does not mean the destruction of phase coherence yet .
some dissipation always accompanies any motion of a magnetic flux that can happen in a hard superconductor even at low transport current .
therefore the critical current in such materials can not be specified as the greatest non - dissipative current .
the superconducting state collapses only when a growth of dissipation becomes avalanche - like as a result of thermo - magnetic instability .
the principal reason of pinning enhancement due to the fractality of the normal phase clusters lies in the fundamental properties of the critical current distribution .
figure [ figure4 ] demonstrates the peculiarities of the fractal probability density specified by eq .
( [ densi7 ] ) . as in fig .
[ figure1 ] , the thin lines show the extreme cases of euclidean clusters @xmath45 and clusters of boundary with the maximum fractality @xmath24 .
as may be clearly seen from these graphs , the bell - shaped curve of the distribution broadens out , moving towards greater magnitudes of current as the fractal dimension increases .
this shift can be described by dependencies of average and mode of the critical current distribution on the fractal dimension , as it is shown in the inset of fig .
[ figure4 ] .
the mode of the distribution , which is equal to the value of the critical current that provides the maximum of the probability density of eq .
( [ densi7 ] ) , depends linearly on the fractal dimension : _ mode_@xmath69 .
the average critical current obeys much more strong superlinear law specified by euler gamma function : @xmath70 the mean value of the critical current for euclidean clusters is equal to @xmath71 , while for clusters of maximum fractality this value becomes infinite : @xmath72 .
figure [ figure4 ] clearly demonstrates that increasing the fractal dimension gives a growth of the contribution made by clusters of greater critical current to the overall distribution , resulting just in enhancement of the magnetic flux trapping .
in the resistive state the hard superconductor is adequately specified by its voltage - current ( _ v - i _ ) characteristic .
the critical current distribution of eq . ( [ densi7 ] ) allows us to find the electric field arising from the magnetic flux motion after the vortices have been broken away from the pinning centers . inasmuch as each normal phase cluster contributes to the total critical current distribution , the voltage across a superconductor @xmath73 is the response to the sum of effects made by the contribution from each cluster .
such a response can be expressed as a convolution integral : @xmath74 where @xmath75 is the flux flow resistance .
the similar approach is used universally to consider behavior of the clusters of pinned vortex filaments @xcite ; to analyze the critical scaling of _ v - i _ characteristics of superconductors @xcite ; that is to say , in all the cases where the distribution of the depinning currents occurs .
the present consideration will be primarily concentrated on the consequences of the fractal nature of the normal phase clusters specified by the distribution of eq .
( [ densi7 ] ) , so all the problems related to possible dependence of the flux flow resistance @xmath75 on a transport current will not be taken up here .
let us consider the simplest case , wherein all of the pinning centers would have an identical critical current @xmath76 so all the vortices would be broken away simultaneously at @xmath77 .
then , referring to eq .
( [ volt8 ] ) , their density would have the following form : @xmath78 where @xmath79 is dirac delta function , @xmath80 is heaviside step function .
thus , the trapped flux would change by 100% at once : @xmath81 .
let us note that in this case @xmath82 due to the convenient normalization chosen above : @xmath28 . in the model case of delta - shaped distribution of the critical currents the voltage across a superconductor in the flux flow regime , according to eq .
( [ volt9 ] ) , would obey the simple linear law : @xmath83 .
the corresponding _ v - i _ characteristic is shown in fig .
[ figure5 ] by the dotted line ( 1 ) . for fractal distribution of critical currents
the situation is quite different , because the vortices are being broken away now in a wide range of the transport currents .
after the substitution of the function of eq .
( [ densi7 ] ) in eq .
( [ volt9 ] ) , upon integration by parts , the voltage across a superconductor can be expressed with the cumulative probability function of eq .
( [ exhyp5 ] ) : @xmath84 integration of which gives : @xmath85 \left\ { i-\left ( \frac{2+d}{2}\right ) ^{\frac{2+d}{2}}u\left [ \frac{d}{2},\frac{d}{2},\left ( \frac{2+d}{2}\right ) ^{\frac{2}{d}+1}i^{-\frac{2}{d}}\right ] \right\ } \label{volthyp11}\ ] ] where @xmath86 is tricomi confluent hypergeometric function . in extreme cases for @xmath16 and for @xmath87 expression of eq .
( [ volthyp11 ] ) can be simplified ( see appendix ) : \(a ) euclidean clusters @xmath23 : @xmath88 \label{erf12}\ ] ] where erfc@xmath89 is the complementary error function .
\(b ) clusters of boundary with the maximum fractality @xmath90 : @xmath91 \label{ei13}\ ] ] where @xmath92 is the exponential integral function .
the _ v - i _ characteristics of a superconductor containing fractal normal phase clusters are presented in fig .
[ figure5 ] .
all the curves are virtually starting with the transport current value of @xmath93 that is agreed with the onset of the resistive state found above from cumulative probability function of eq .
( [ exhyp5 ] ) . when the current increases the trapped flux remains unchanged until the vortices start to break away from the pinning centers .
as long as the magnetic flux does not move , no electric field is arisen .
two thin lines ( 2 ) , calculated using the formulas of eq .
( [ erf12 ] ) and eq .
( [ ei13 ] ) , bound the region the _ v - i _ characteristics can fall within for any possible values of fractal dimension .
as an example , the curve ( 3 ) demonstrates the _ v - i _ characteristic of a superconductor containing fractal clusters of previously obtained coastline dimension @xmath30 .
this figure shows that the fractality reduces appreciably an electric field arising from the magnetic flux motion .
this effect is especially strong in the range of the currents @xmath94 , where the pinning enhancement also has a maximum ( see fig . [ figure3 ] ) .
both these effects have the same nature , inasmuch as their reason consists in the peculiarities of fractal distribution of critical currents of eq .
( [ densi7 ] ) . as is seen from fig .
[ figure4 ] , an increase of fractality causes a significant broadening of the tail of the distribution @xmath95 .
it means that more and more of small clusters , which can best trap the magnetic flux , are being involved in the game .
hence the density of vortices broken away from pinning centers by the lorentz force is reducing , so the smaller part of a magnetic flux can flow , creating the smaller electric field . in turn , the smaller the electric field is , the smaller is the energy dissipated when the transport current passes through the sample .
therefore , the decrease in heat - evolution , which could cause transition of a superconductor into a normal state , means that the current - carrying capability of the superconductor containing such fractal clusters is enhanced .
thus , fig .
[ figure5 ] , as well as fig .
[ figure3 ] , obviously demonstrates such a practically important result : the fractality of the boundary of the normal phase clusters , which act as the pinning centers , prevents the destruction of a superconductivity by a transport current , and therefore , causes the critical current to increase .
thus , the fractal properties of the normal phase clusters have an essential influence on the dynamics of the trapped magnetic flux . the crucial change of the critical current distribution caused by increasing of the coastline fractal dimension of the normal phase clusters forms the basis of this effect .
the most important result is that the fractality of cluster boundary strengthens the flux pinning and thereby hinders the destruction of superconductivity by the transport current , resulting in enhancement of the current - carrying capability of a superconductor .
this phenomenon provides the principally new possibilities for increasing the critical current value of composite superconductors by optimizing their geometric morphological properties .
in order to obtain the expressions of eq .
( [ volthyp11 ] ) - eq . ( [ ei13 ] ) for the voltage across a superconductor ,
it is necessary to integrate the cumulative probability function for the critical current distribution of eq .
( [ exhyp5 ] ) .
the substitution of exponential - hyperbolic distribution of eq .
( [ exhyp5 ] ) in eq .
( [ voltgen10 ] ) gives : @xmath96 using the change of a variable of the form @xmath97 , we can get the following expression : @xmath98 which , upon integration by parts , becomes : @xmath99 where @xmath100 is the incomplete gamma function .
this function can be represented as : @xmath101 where @xmath102 is tricomi confluent hypergeometric function ; and @xmath103is euler gamma function . thus with the help of eq .
( [ a3 ] ) , the expression for the voltage across a superconductor of eq .
( [ a2 ] ) can be written in its final form : @xmath104 \label{a4}\ ] ] this formula is similar to the expression of eq .
( [ volthyp11 ] ) .
the corresponding _ v - i _ characteristic of a superconductor calculated using this expression at @xmath30 is shown in fig .
[ figure5 ] by the curve ( 3 )
. the formulas of eqs .
( [ a5 ] ) , ( [ a6 ] ) describe dependencies of the voltage across a superconductor in a resistive state on the transport current for extreme values of the coastline fractal dimension .
two corresponding _ v - i _ curves are shown in fig .
[ figure5 ] by thin lines ( 2 ) . whatever the geometric morphological properties of the normal phase clusters may be
, the _ v - i _ characteristics of a superconductor will fall within the region bounded by those two limiting curves , as it is shown in fig .
[ figure5 ] ( like the curve ( 3 ) drawn for @xmath30 ) .
m. r. beasley , artificially - structured superconductors .
- in : percolation , localization and superconductivity ( ed . by a. m. goldman
end s. a. wolf , nato asi series , ser.b , vol.109 , plenum press , new york , 1984 ) , pp.115 - 143 .
r. b. laibowitz , r. f. voss , e. i. alessandrini , clustering in thin au films near the percolation threshold .
- in : percolation , localization and superconductivity ( ed . by a.
m. goldman end s. a. wolf , nato asi series , ser.b , vol.109 , plenum press , new york , 1984 ) , pp.145 - 160 .
|
the influence of geometry and morphology of superconducting structure on critical currents and magnetic flux trapping in percolative type - ii superconductor is considered .
the superconductor contains the clusters of a normal phase , which act as pinning centers .
it is found that such clusters have significant fractal properties .
the main features of these clusters are studied in detail : the cluster statistics is analyzed ; the fractal dimension of their boundary is estimated ; the distribution of critical currents is obtained , and its peculiarities are explored .
it is examined thoroughly how the finite resolution capacity of the cluster geometrical size measurement affects the estimated value of fractal dimension .
the effect of fractal properties of the normal phase clusters on the electric field arising from magnetic flux motion is investigated in the case of an exponential distribution of cluster areas .
the voltage - current characteristics of superconductors in the resistive state for an arbitrary fractal dimension are obtained .
it is revealed that the fractality of the boundaries of the normal phase clusters intensifies the magnetic flux trapping and thereby raises the critical current of a superconductor .
|
in this paper we describe the construction and performance of a straw tube chamber with a unique type of readout : rather than signals being read from anode wires , signals are read from pads placed against the straw tubes which pick up signals induced by the gas avalanches .
this readout scheme has three advantages : _ ( 1 ) _ fully - correlated @xmath0 and @xmath1 position information ; _ ( 2 ) _
the pad sizes can be chosen almost arbitrarily , for example , to give equal pad occupancy over a region in which particle flux is varying ; and _ ( 3 ) _
the high - voltage and front - end electronics are decoupled . for the gas avalanche signals to be picked up by the pads
, the straw tube material must be resistive such that the straw tube wall does not substantially shield the signal .
there are several issues which determine the feasibility of using such a detector under experimental conditions : the size of the pickup signal , the level of noise ( both capacitive and pickup ) , and the distribution of pickup signals among neighboring pads .
multiwire proportional chambers with resistive cathodes and pickup pad readout have been studied by battistoni _
et al._@xcite .
resistive straw tubes with pickup _ wires _ wrapped around the straws have been studied by bychkov _
et al._@xcite .
our final objective is to use such a chamber to trigger on tracks with large transverse momentum ( @xmath2 ) in the _ hera - b _ experiment at desy@xcite . to accomplish this
, we position three chamber stations within the _ hera - b _
spectrometer magnet and input the pad hit pattern to fast coincidence logic .
the logic flags hit patterns that are characteristic of tracks which bend little within the magnetic field and which lie at larger angles with respect to the beam . to minimize hit multiplicities and simplify the trigger logic , we desire that the pickup signal be as localized as possible , preferentially occurring on only one pad .
we have constructed two prototype chambers which we refer to as chamber a and chamber b. the chambers are similar and consist of two distinct parts : _ ( 1 ) _ the straw tubes , their endplates , and the high voltage distribution cards ; and _ ( 2 ) _ the pickup pads and their preamplifier cards .
the straw tubes have a diameter of 5.0 mm and are configured in two rows of sixteen straw tubes each . for chamber
a , the tubes have a length of 60 cm and the rows are offset with respect to each other by one straw tube radius .
this `` close - packed '' geometry is especially resistant to mechanical stresses . for chamber b ,
the tubes have a length of 40 cm and the rows are offset by a smaller amount , only 0.6 mm .
chamber b also has a thin line of conductive paint applied to the surface of every third straw tube .
this paint is held at ground potential , and its purpose is to limit the pickup signal to a single pad ( to be discussed later ) . in all other aspects the two chambers are identical .
the pickup pads for both chambers have dimensions @xmath3
@xmath4 and are etched on one side of a printed circuit board .
the arrangement of straw tubes and pad plane is shown in fig .
[ fig : prototype ] .
4.0 truein .2 cm the straw tubes are rolled@xcite out of two - layer material : the inner layer consists of carbon - loaded kapton with a resistivity of approximately 1 m@xmath5/square@xcite , and the outer layer consists of mylar .
the kapton layer serves as the cathode and is held at ground potential , while the mylar layer helps provide a gas seal .
the thickness of the kapton is 0.001 inch , and the thickness of the mylar is 0.0005 inch .
the straws are strung with gold - plated tungsten wire having a diameter of 20 @xmath6 m .
for most of our tests we used a gas mixture of 90% argon and 10% methane ( `` p10 '' ) .
the pad plane consists of a two - layer printed circuit board with copper pads on one side and traces on the other .
the traces connect to the pads via plated through - holes and carry signals to one end of the pad plane , where preamplifier cards are mounted . to minimize detector mass ,
the material used for the printed circuit board is kapton .
the thickness of the kapton is only 0.006 inches ( including the cu pads and traces ) , which corresponds to 0.28% of a radiation length .
the trace width is 0.006 inches .
the front - end readout cards are based upon the asd-8 amplifier - shaper - discriminator asic developed at the university of pennsylvania@xcite .
we have conducted a series of bench tests to measure the signal size , noise , and the relative sizes of pickup signals among neighboring pads .
the test setup used is shown in fig .
[ fig : test_setup ] . 2.5 truein .2 cm to measure signal size , we positioned an @xmath7fe source ( @xmath8 kev ) above a small hole drilled in the pad plane .
this hole is located directly above a straw tube as shown in fig .
[ fig : test_setup ] .
we measured the signal collected on both the anode wire and the pickup pad using an ortec multichannel analyzer ( mca ) system consisting of a 142pc charge amplifier , a 570 spectroscopy amplifier , and a 419 mca pc plug - in board .
the spectra show a significant signal from the pickup pad : the ratio of pad signal to anode wire signal is about 1/3 .
the size of the pickup signal is very sensitive to the size of the gap between the straw tubes and the pad plane .
we have measured this dependence using the same test setup ; the resulting data is shown in fig .
[ fig : sig_vs_gap ]
. we account for this data quantitatively by considering the geometry of the insert in fig .
[ fig : sig_vs_gap ] . for a single straw tube and an infinitely long pickup strip oriented parallel to it , the capacitance per unit length between the straw tube and pickup strip is@xcite : @xmath9 where @xmath10 is the permittivity of free space ( neglecting the effect of the thin mylar layer of the straw tube wall ) , @xmath11 is the width of the pickup strip , @xmath12 , and @xmath13 . in these expressions
@xmath14 is the distance between the center of the tube and the pickup strip , and @xmath15 is the radius of the tube .
if the potential on the straw tube is @xmath16 , then the signal on the pickup strip per unit length is @xmath17 .
it is due to the factor @xmath18 that the pickup signal depends strongly on the ratio @xmath19 .
the signal size predicted by eq .
( [ eqn : sig_vs_gap ] ) is superimposed on the data in fig .
[ fig : sig_vs_gap ] and shows very good agreement .
3.2 truein .2 cm the basic assumption of this model is that the straw tube wall is an equipotential surface ; i.e. , the charge induced on the tube wall has sufficient mobility to form an equipotential surface within a time scale much less than the response time of the readout electronics .
this time scale is expected to be of order @xmath20@xcite , where @xmath21 is the permittivity of kapton and @xmath22 is its bulk conductivity . since our film has a resistivity of 1 m@xmath5/square and a thickness @xmath23 of 0.001 inch , @xmath24-m)@xmath25 .
if we take the permittivity of kapton to be that of free space @xmath26 , then the time scale required is 225 ps .
this interval is relatively short : the fraction of signal charge collected in time @xmath27 is @xmath28 $ ] , where @xmath29 , @xmath30 is the radius of the anode wire , @xmath31 is the radius of the straw tube , @xmath32 is the mobility of the positive ions , and @xmath33 is the anode wire voltage . inserting values @xmath34
@xmath6 m , @xmath35 mm , @xmath36
mm@xmath37-sec ) , and @xmath38 v , we find that in 225 ps only 1.4% of the signal charge has been collected .
we have measured the distribution of pickup signals among neighboring pads using an @xmath7fe source .
the pad geometry studied is shown in fig .
[ fig : pad_geometry ] .
the @xmath7fe source is placed above pad # 7 , row # 5 .
there is a small hole drilled in the center of this pad to allow x - rays from the source to enter the straw tube below .
we trigger on signals from this pad and measure the signals picked up by neighboring pads .
the results are summarized in table [ tab : spread ] . from table [
tab : spread ] we observe that the signal distribution along the straw tube direction ( pad # 8 , row # 5 ) is narrow .
there are two mechanisms affecting the distribution in this direction : _
( a ) _ the direct signal - induction mechanism , and _ ( b ) _ signal propagation along the transmission line formed by the anode wire and the cathode tube .
the signal distribution due to @xmath39 can be calculated analytically using green s reciprocation theorem@xcite ; a solution is given in the long version of this paper@xcite .
the resulting charge distribution on the pads after 6 ns ( the shaping time of the asd-8 ) is relatively narrow : the standard deviation of a gaussian fit to the distribution is only @xmath401.4 mm .
this value is consistent with the value measured ( table [ tab : spread ] ) ; however , the measurement is dominated by noise and also receives contributions from wider angle photons that produce gas avalanches closer to the pad boundary .
thus , we interpret the values in table [ tab : spread ] as upper bounds . the signal distribution due to propagation via the transmission line formed by the anode wire and the cathode tube is also calculated in ref
. @xcite .
since the tube wall is made of a highly - resistive material , the signal propagates with strong attenuation .
the calculation predicts that for a straw tube resistivity of 1 m@xmath5/square , the signal is attenuated by a factor of 89 over a distance of 1 cm .
we thus conclude that for pad lengths @xmath41 cm , the signal distribution among pads due to this mechanism is also narrow .
3.4 truein .2 cm table [ tab : spread ] shows that the pickup signal on the pads transverse to the straw tube direction is 12% of that on the main pad .
this result has been successfully simulated using two models : _ ( 1 ) _ a simple `` capacitive coupling '' model , and _ ( 2 ) _ a finite element analysis ( fea ) simulation . the capacitive coupling model corresponding to chamber a is shown in fig . [ fig : ccmodel ] .
in this figure , @xmath42 is the capacitance between a straw tube and a pad , and @xmath43 is the capacitance between adjacent straw tubes .
note that one pad ( @xmath44 cm ) covers three straw tubes ( @xmath45 mm ) .
the gas avalanche is modeled as a current source injecting charge @xmath46 into a single node as indicated in the figure .
we apply kirchhoff s law to this circuit to obtain a set of linear , coupled equations .
we then solve these equations numerically to obtain the charge induced on each pad .
we have calculated the induced charge for two cases : @xmath47 and @xmath48 .
the first case corresponds to a gap size of approximately 0.1 mm , while the latter case corresponds to a gap size of only @xmath400.012 mm .
the result for @xmath48 is very close to the measured value : the ratio of the signal on an adjacent pad to that on the main pad is 0.13 , while the measured value is 0.12 ( see table [ tab : spread ] ) .
the result obtained with the fea simulation is identical .
a detailed description of the fea calculation is given in ref .
@xcite .
5.0 truein .2 cm our bench measurements and simulation results indicate that the distribution of pickup signals perpendicular to the direction of the straw tubes when fit to a gaussian function has a standard deviation of @xmath400.5 pad .
this spread is relatively large for triggering purposes . to prevent adjacent pads from recording hits
, we have developed a technique to limit the pickup signal to a single pad .
this technique is as follows : at the juncture between straw tubes which lie at pad boundaries , we apply a thin line of conductive paint as shown in fig .
[ fig : silver_paint ] .
this paint is held at ground potential via contact with the chamber endplates and provides a preferred location for electric field lines to end . in this manner the capacitive coupling between adjacent straw tubes is reduced .
this mechanism has been simulated by the fea program , and the resultant distribution of pickup signals indicates that the ratio of signal on an adjacent pad to that on the main pad is only 1/62 .
we have measured this ratio using the pad arrangement of fig .
[ fig : pad_geometry ] and obtained 1/38 .
this value is in rough agreement with that calculated and is significantly smaller than that measured without the conductive paint applied , @xmath401/8 ( table [ tab : spread ] )
. 2.4 truein .2 cm
we have tested prototype chamber a in a beamtest at brookhaven national laboratory ( bnl ) in order to measure the efficiency and noise of the chamber under experimental conditions .
the beamtest was conducted at the alternating gradient synchrotron ( ags ) using a secondary beam consisting mostly of pions with momentum in the range @xmath49 gev/@xmath50 . a plan view of the apparatus used is shown in fig .
[ fig : bnl_plan_view ] . immediately upstream of the prototype chamber was a conventional straw tube tracking chamber consisting of two @xmath0-measuring and two @xmath1-measuring planes .
a lecroy 4448 latch module was used to record hits from the pad chamber , while several lecroy 2277 tdc s were used to record hits from the straw tube tracker .
the tdc hit information allowed us to reconstruct tracks , project them to pads of the prototype chamber , and measure the efficiency of the chamber and the distribution of pickup signals among neighboring pads .
4.5 truein .2 cm the efficiency of the pad chamber was measured by reconstructing tracks in the tracker , projecting them to the pad chamber , and recording whether the pad(s ) to which a track projects has a signal .
the results are listed in table [ tab : beamtest_eff ] . for high voltage ( hv )
applied to both layers of straw tubes , the efficiency is at least 99.5% for all three discriminator threshold settings .
the table also shows that when high voltage is applied to ( at least ) the upper layer of straws , there is very little dependence of the efficiency on the discriminator threshold setting .
this indicates that even our highest threshold setting ( 1.75 v ) corresponds to a signal well below typical pulse - heights .
we measured the distribution of pickup signals among neighboring pads by projecting reconstructed tracks to the pad chamber and recording the total number of pads which had hits ; i.e. , including those pads adjacent to the pad to which the track projects .
we implement this procedure for two subsets of tracks : those which project to within 2 mm of the center of a pad , and those which project to within 2 mm of the center in the @xmath1-coordinate and to within 2 mm of the _ edge _ in the @xmath0-coordinate .
we expect the latter subset on average to have pickup signals distributed over a larger number of pads . for both subsets the number of contiguous pads which had hits
is shown in fig .
[ fig : beamtest_spread ] .
the mean number of pads with hits is 1.9 for tracks which project to the center of a pad , and 2.1 for tracks which project to the edge of a pad .
[ fig : beamtest_spread ] shows that even for the optimal case of a gas avalanche occurring in the center of a pad , the frequency of only one pad having a hit is only 39% . to increase this fraction
, we developed the technique of applying conductive paint to every third straw tube as discussed previously .
we gratefully acknowledge the strong support of the bnl ags testbeam group , in particular alan carroll and ed schwaner .
we are indebted to william sands and robert klemmer of the princeton experimental particles laboratory for their excellent work in constructing the prototype chambers , and to stanley chidzik of the electronics shop for producing the asd-8-based preamplifier cards . finally , we thank mitch newcomer , frank shoemaker , and kirk mcdonald for many useful discussions .
.[tab : spread ] the signals picked up by three neighboring pads when triggering on pad # 7 , row # 5 .
the first and second pads listed are located along the straw tube direction , and the first and third pads listed are located perpendicular to the straw tube direction . [ cols="<,^,^,^",options="header " , ] john d. kraus , _ electromagnetics , 4th ed . _
( mcgraw - hill , new york , 1992 ) , sec .
the capacitance to a strip of width @xmath11 is @xmath51 times the capacitance to the entire plane in which the strip lies .
|
we have developed a straw tube chamber with pickup - pad readout .
the mechanism for signal pickup , the size of the pickup signal , and the distribution of signals among neighboring pads are discussed .
we have tested a prototype chamber in a beamtest at brookhaven national laboratory and have measured chamber efficiencies in excess of 99% .
|
an attitude / orientation control system is necessary for all satellites with different satellites requiring varying degrees of pointing accuracy .
satellites with a high - gain directional antenna , a telescope , or an earth - imaging instrument as payloads require highly accurate ( few arcminutes or arcseconds ) pointing system . determining
the current pointing position is of primary importance in such applications and it is usually achieved using star sensors : a wide - fov camera with online image processing capabilities which reduces the image data to an attitude quaternion that describes the rotation of the sensor coordinate system with respect to the earth - centered inertial ( eci ) coordinate system .
this is achieved by applying multiple algorithms on the image in sequence @xcite .
we have developed a low - cost star sensor _
starsense _ to deploy on minisatellites and cubesats using off - the - shelf components , in which we have implemented a centroiding algorithm for finding centroids of stars seen in the image , a geometric voting algorithm for star pattern identification , and a quaternion estimator ( quest ) algorithm for quaternion estimation . to determine how each algorithm performs in terms of sky coverage , memory requirements , calculation time , and so on
, we simulate the algorithms along with idealized and real hardware parameter inputs using a specially developed testing software package . in this paper
, we present a complete software package intended to evaluate the performance of algorithms implemented on the _ starsense_. this software package can be applied to any star sensor by changing the hardware parameters , such as the focal length of the imaging setup , the field of view ( fov ) of the camera , angle measurement accuracy , distortion effects , and others .
it is written to run under matlab due to the simplicity of scripting and its excellent capability of visualizing the results .
the implementation on actual star sensor electronics hardware will be done in c , keeping in view its easy portability to other platforms . in section [ sec : hardware_info ] , we briefly describe the hardware of the star sensor .
section [ sec : software ] describes the architecture of the software package we used to evaluate the algorithms .
section [ sec : algorithms ] describes the algorithms in detail . in section [ sec : performance_est ] , we describe the testing methodology to analyze the performance parameters of these algorithms , and the final results from different tests .
_ starsense _ is essentially a wide - field imaging camera with high sensitivity .
the optics consists of a four - element tessar lens system designed to minimize chromatic aberration , coma , and distortion .
the lens system is designed with weight as a limiting constraint , and a condition of seeing a minimum of 3 stars in fov in any field of the sky .
the lens assembly is designed to sustain vibrations experienced during the satellite launch .
a baffle is used to prevent the stray - light from the sun and the earth entering the optical system , thereby contributing to the sky background .
in addition , we have used a thermal cut - down filter ( a hot mirror ) , which reflects the ir and uv wavelengths and only allows optical wavelengths from 450750 nm to pass through , to prevent the detector from getting heated by the direct sun .
the imaging sensor is the only electronic component directly open to the radiation environment and , therefore , its performance is prone to degradation with time . we have used a radiation - hardened cmos detector star-1000 @xcite , which is sensitive to low light conditions due to its large pixel size ( @xmath0 @xmath1m@xmath2 ) .
the optics has been designed specifically for this detector , so that the point spread function ( psf ) is spread over 4 pixels .
the electronic system of _ starsense _ is inspired by cubestar @xcite .
it is distributed over two separate printed circuit boards ( pcbs ) : detector pcb and image processor pcb .
the image sensor and its biasing circuits are soldered on the detector pcb , which is fixed inside the star sensor structure in such a way that the focal plane of the star sensor optics coincides with the image sensor plane .
the detector is connected to the image processor board through a flat plastic cable .
such an arrangement allows for the image processor board to be removed from the structure without disturbing the focal plane alignment of the detector pcb . the readout system for the image sensor
is implemented using a mil - grade spartan-6 field programmable gate array ( fpga ) @xcite on the image processor pcb .
the pcb also hosts sdram , required for the online processing of images , and a flash memory for nonvolatile storage of the bitstream and a star catalog .
figure [ fig : processor_board ] shows the _ starsense _ image processor board .
since the optics , electronics and the structure were fully designed in - house , their characteristic properties and modeling parameters are well known , and we have verified them using different calibration tests .
these parameters are used as inputs to algorithms discussed further in the paper .
the key parameters of the camera are described in table [ table : camera_param ] , and the key features of the electronics system are presented in table [ table : star_sensor_electronics ] . a detailed paper justifying the choice of the hardware and various design parameters is forthcoming . ._starsense _ camera parameters [ cols= " < , < " , ] [ table : sky_coverage ]
we have put together a complete software package to reduce star sensor images to quaternions and to evaluate the performance of the operational algorithms .
the package simulates centroiding , geometric voting and quest algorithms , and evaluates such performance parameters as attitude accuracy , calculation time , required memory , star catalog size , sky coverage , etc . , and estimated the errors introduced by each algorithm .
the testing is parametrized for different hardware parameters of the star sensor , such as the focal length of the imaging setup , fov of the camera , angle measurement accuracy , distortion effects , and others .
we conclude with the following remarks : * for our _ starsense _ , we find that a limiting magnitude of v=6.0 is optimal to get significant sky coverage with minimal calculation time . *
this software package is robust , fast , user - friendly in terms varying hardware parameters , and easily portable to various operating platforms . * due to the parametrized approach in package development , it can be applied to evaluate the performance of such algorithms in any star sensor
. the source codes of the software package can be obtained from a github repository @xcite .
part of this research has been supported by the department of science and technology ( government of india ) under grant ir / s2/pu-006/2012 .
kolomenkin , m. , pollak , s. , shimshoni , i. and lindenbaum , m. , 2008 , _ geometric voting algorithm for star trackers _ , in ieee transactions on aerospace and electronic systems , vol .
441 - 456 , april 2008 .
|
we have developed a low - cost off - the - shelf component star sensor ( _ starsense _ ) for use in minisatellites and cubesats to determine the attitude of a satellite in orbit .
_ starsense _ is an imaging camera with a limiting magnitude of 6.5 , which extracts information from star patterns it records in the images .
the star sensor implements a centroiding algorithm to find centroids of the stars in the image , a geometric voting algorithm for star pattern identification , and a quest algorithm for attitude quaternion calculation . here
, we describe the software package to evaluate the performance of these algorithms as a star sensor single operating system .
we simulate the ideal case where sky background and instrument errors are omitted , and a more realistic case where noise and camera parameters are added to the simulated images .
we evaluate such performance parameters of the algorithms as attitude accuracy , calculation time , required memory , star catalog size , sky coverage , etc . , and estimate the errors introduced by each algorithm .
this software package is written for use in matlab .
the testing is parametrized for different hardware parameters , such as the focal length of the imaging setup , the field of view ( fov ) of the camera , angle measurement accuracy , distortion effects , etc . , and
therefore , can be applied to evaluate the performance of such algorithms in any star sensor . for its hardware implementation on our
_ starsense _ , we are currently porting the codes in form of functions written in c. this is done keeping in view its easy implementation on any star sensor electronics hardware .
|
the origin of the observed baryon asymmetry is a fundamental question in modern cosmology .
forty years ago sakharov showed that three conditions must be satisfied to generate the asymmetry @xcite , and since that time various models have been proposed to explain it .
but we still do nt have any idea which , if any , is correct , because baryogenesis usually depends on unknown physics beyond the reach of current experiments and on the unknown early history of the universe beyond the reach of current observations .
a successful cosmological history must produce necessary relics such as the baryon asymmetry , but must also avoid or dilute unwanted cosmological relics such as gravitinos @xcite and moduli @xcite , which can destroy the successful predictions of big bang nucleosynthesis @xcite or over - dominate the universe depending on the scale of supersymmetry breaking .
thermal inflation @xcite provides a compelling solution to the unwanted relic problem .
it occurs after the usual high energy inflation , at an energy scale between the intermediate and electroweak scales , releasing enough entropy at a low enough energy scale to dilute the unwanted relics . however , the usual baryogenesis scenarios do not work well in the context of thermal inflation because of the large entropy production at a low energy scale .
thus a new model of baryogenesis that fits naturally with thermal inflation is desirable @xcite . in a previous paper @xcite
, we proposed a simple model based on the superpotential throughout this paper . ]
@xmath1 where @xmath2 is the flaton whose roll out from the origin ends thermal inflation and generates the minimal supersymmetric standard model ( mssm ) @xmath3-term .
the flaton must also have some unsuppressed couplings to the thermal bath , in order to hold it at the origin during the thermal inflation , but the extra fields involved in these couplings acquire intermediate scale masses after the thermal inflation and hence are only weakly constrained , so we do not include them explicitly .
the baryogenesis is an affleck - dine ( ad ) @xcite type of leptogenesis using @xmath4 , @xmath5 and @xmath6 , and triggered by the roll out of the flaton . in ref .
@xcite , we demonstrated the generation of a lepton asymmetry using a simple homogeneous numerical simulation . in the usual high energy ad scenarios , the amplitude of the ad field
would then be damped by the hubble expansion , bringing the field into the symmetric region of the potential near the origin , ensuring that the generated lepton number was conserved .
however , in our low energy ad scenario , the hubble damping is negligible .
instead , we suggested preheating and thermal friction might damp the amplitude of the homogeneous mode , but testing this idea was beyond the scope of our homogeneous numerical simulation . during preheating ,
energy is transferred from a homogeneous oscillating mode to inhomogeneous modes , resulting in reduction of the amplitude of the homogeneous mode .
well known mechanisms of preheating include parametric resonance @xcite and tachyonic instability @xcite .
usually , it is regarded as an intermediate process between the end of inflation and full reheating , i.e. thermalization .
however , it can be relevant to the evolution of any condensate , in particular ad condensates .
especially , it may be a crucial effect for low energy ad scenarios , providing an alternative to the usual hubble damping . in this paper
, we perform three dimensional lattice simulations to numerically investigate the effect of preheating on the amplitude of the ad fields , and hence on the conservation of the lepton number , in our model . in section [ model ] , we briefly review our model . in section [ lepto ]
, we describe our scenario for the generation and conservation of the lepton number asymmetry . in sections [ sim ] and [ results ] ,
we describe and present the results of our numerical simulations . in section [ dis ] ,
we discuss interesting points , limitations of our current work , and work to be done in future . while this work was in progress , and after some of our initial results were reported @xcite , another work on the same topic appeared @xcite .
their simulation was one dimensional , which restricts the number of modes available for preheating , and they chose different initial conditions and parameters , but this is not important at this stage of simulations of our model . ] .
this may explain why their results look somewhat different from ours , though they come to the same conclusion that the model seems to work . unfortunately , there are insufficient details of their simulation in their paper for us to reproduce their results .
we hope to clarify these issues after this paper is published .
our model is based on the superpotential of eq .
( [ w ] ) . the fourth term generates the mssm @xmath3-term when the flaton @xmath2 settles down to its vacuum expectation value of the order of @xmath7 to @xmath8 .
the fifth term generates light majorana masses for the neutrinos , and the sixth term stabilizes @xmath2 .
we assume @xmath9 to give the correct values for @xmath3 and the neutrino masses , assuming a flaton mass of the order of the electroweak scale .
initially , @xmath2 is held at origin and so the @xmath3-term vanishes .
therefore @xmath5 and @xmath6 lack the usual mssm @xmath3-term contribution to their mass - squareds .
this allows the possibility that some supersymmetric flat directions involving @xmath5 or @xmath6 , in particular @xmath10 or @xmath11 , may initially be unstable , although they are stable in our vacuum .
thus we consider that @xmath10 , @xmath11 and @xmath2 may have nonzero values , and assume that the other fields remain at the origin .
then , truncating to a single generation for simplicity , gauge fixing and imposing the d - term constraints , we can parameterize those flat directions as @xmath12 with the remaining @xmath13-term constraint @xmath14 and associated gauge constraint @xmath15 which are not easily integrated out .
the charges are @xmath16 , @xmath17 and @xmath18 , and @xmath19 is the gauge covariant derivative .
the gauge field @xmath20 is heavy and so not dynamical .
the potential is @xmath21 \nonumber \\ \label{v } & & { } + \left| \lambda_\mu \phi^2 h_d + \lambda_\nu l^2 h_u \right|^2 + \left| \lambda_\mu \phi^2 h_u \right|^2 + \left| \lambda_\nu l h_u^2 \right|^2 + \left| 2 \lambda_\mu \phi h_u h_d + \lambda_\phi \phi^3 \right|^2\end{aligned}\ ] ] @xmath22 is adjusted to give zero cosmological constant and all the other supersymmetry breaking parameters are of the order of the electroweak scale .
we denote a typical electroweak scale mass by @xmath23 .
we assume @xmath24 so that the thermal inflation ends by @xmath2 rolling away .
we assume @xmath25 so that @xmath10 can roll away from the origin before the end of thermal inflation , providing the initial condition for our ad leptogenesis .
this nontrivial assumption results in there being a deeper non - mssm vacuum @xcite so that our mssm vacuum is meta - stable .
however , as explained in @xcite , this is phenomenologically and cosmologically consistent . for simplicity , we neglect any field dependent renormalization of the supersymmetry breaking parameters . in this section
we give formulae for some physically important parameters of the potential .
note that @xmath26 and @xmath27 .
the magnitude of the flaton vacuum expectation value @xmath28 is given by @xmath29 the potential energy density at the origin is @xmath30 the flaton mass - squared eigenvalues at @xmath31 are @xmath32 where the subscripts @xmath33 and @xmath34 denote the radial and angular eigenvectors .
when @xmath35 , the magnitudes of @xmath36 and @xmath37 at the minima of their potential are given by @xmath38 when @xmath2 settles down to @xmath28 , the mssm @xmath3 and @xmath39 parameters are generated @xmath40 and the ad mass - squared eigenvalues at the ad origin become @xmath41 the mass - squared along the @xmath42 flat direction changes its sign at the critical value @xmath43 given by @xmath44 to estimate the potential energy available in the ad sector , we use @xmath45 \left| l_0 \right|^2
\\ \label{v2 } v_2 \enskip \equiv & v(\phi_0,h_{u0},0,l_0 ) & = \enskip \left| \mu \right|^2 \left| l_0 \right|^2 + v_1\end{aligned}\ ] ] where @xmath46 is the full potential .
initially , all the fields are held at the origin due to the finite temperature effective potential .
we assume @xmath47 leaves the origin first .
it will do so via a first order phase transition in which rare bubbles expand to hubble volume size before percolating , since the expansion rate @xmath48 is much less than the scale of the phase transition @xmath23 . as @xmath49 rolls away
, it gives an increasing mass to the fields it couples to .
these then decay and hence are eliminated from the thermal bath , but deposit their energy , including that of their extra mass gained from @xmath49 , back into the thermal bath .
this could lead to two possibilities for @xmath2 s dynamics , depending on @xmath2 s couplings and the detailed dynamics .
some fields are removed from the thermal bath , and if these fields couple to @xmath2 then @xmath2 s finite temperature effective potential could be reduced , possibly triggering @xmath2 s phase transition . in this first case ,
@xmath49 and @xmath2 roll away at the same time , though with @xmath49 in the lead if @xmath50 .
however , the temperature of the thermal bath is raised which would have the opposite effect , tending to prolong the thermal inflation . in this second case
, @xmath49 would have time to settle to its minimum , preheat and maybe even decay or be diluted by the remaining thermal inflation , before @xmath2 undergoes its own first order phase transition , again with rare bubbles expanding to hubble volume size before percolating . in this paper
, we consider the second case as it seems simpler .
the initial condition for the leptogenesis will then be @xmath2 s phase transition ending the thermal inflation , with @xmath49 in some initial state near its minimum . in this paper ,
we will only attempt to crudely mimick @xmath2 s first order phase transition , as described in section [ simini ] .
a bound on the initial state of @xmath49 can be estimated as follows .
after @xmath49 rolls away , the energy density in the ad sector is initially @xmath51 the ad sector will decay to the thermal bath at a rate @xmath52 and the rate of transfer of energy from the ad sector to the thermal bath will be @xmath53 therefore the energy transferred in the first hubble time @xmath54 will be much larger than @xmath55 , and so the decay of the ad sector will be able to maintain the thermal inflation .
the thermal inflation will end when @xmath56 , i.e. after @xmath57 thus , we can expect the initial fluctuations in @xmath49 to be fairly small and well preheated if not thermalized .
however , the leptogenesis seems not very sensitive to this initial state , see section [ results ] . when the temperature of the thermal bath drops to around the electroweak scale , it is @xmath2 s turn to roll away from the origin , ending thermal inflation . as @xmath2 rolls out
, @xmath58 acquires linear terms in its potential which force it to become non - zero . when @xmath2 becomes of the order of its vacuum expectation value , @xmath37 acquires an additional contribution to its mass - squared from the term @xmath59 , bringing it , and @xmath36 and @xmath58 , back in towards the origin . at the same time , the phase of @xmath49 is rotated by the term @xmath60 , generating the lepton asymmetry @xmath61 note that the nonzero @xmath58 is necessary to generate the lepton asymmetry , and is also essential to stabilize the dangerous field directions @xmath62 and @xmath63 which would lead one to the deeper non - mssm vacuum @xcite . in the usual high energy ad scenario ,
the generated asymmetry is conserved because the hubble expansion damps the kinetic energy of the ad field , bringing it into the symmetric region of the potential near the origin . however , in low energy ad scenarios the hubble damping is negligible and the asymmetry could be washed out by rescattering off the asymmetric part of the potential away from the origin .
fortunately , the amplitude of the ad fields can also be damped by the transfer of energy to gradient energy and other fields . in this paper
we focus on the former effect , which generally goes by the name of preheating @xcite . in our model
, the gradient energy of the flaton initially grows very rapidly as the flaton rolls away from the origin , due to the tachyonic potential in the radial direction and the divergence of trajectories in the angular direction .
this rapid growth flattens off as the flaton settles down to oscillating about its minimum , with modest and declining growth of gradient energy due to the non - quadratic terms in the potential around the minimum .
the formation and decay of flaton toplogical defects also has an effect .
the ad potential has a strong dependence on the flaton , so inhomogeneities in the flaton field are transferred to the ad fields , leading to rapid preheating of the ad fields .
in addition to this , preheating due to the couplings among the ad fields themselves is likely to be significant , but it is difficult to analyze analytically due to the complicated form of the ad potential .
the energy of the ad fields can also be transferred to other fields .
when an ad field passes near the origin , some fields which couple to it , such as quarks , may become sufficiently light that they can be thermally populated .
as the ad field moves away from the origin , the fields become heavy again , extracting energy from the ad field .
this energy is then transferred to the thermal bath , initially composed mostly of gluons , as the fields decay .
this results in a friction like effect , which we call ` thermal friction ' , on the motion of the ad fields when they pass near the origin .
the efficiency of thermal friction will depend on how close the ad field has to pass to the origin for the quark or other field to become light enough to be thermally populated , and also on how high the temperature is .
how close the ad field has to pass to the origin will depend on the couplings and the temperature of the thermal bath .
the initial temperature of the thermal bath may be relatively small , perhaps electroweak scale , but the mssm has a large hierarchy of yukawa couplings .
this should allow first generation quarks and leptons to be thermally populated when an ad field passes only moderately close to the origin .
as the temperature rises , quarks and leptons of the higher generations will also be able to be populated .
acting against these effects , the flaton dependence of the ad potential will tend to transfer energy from the flaton to the ad fields .
the initial flaton energy density @xmath64 is much larger than the initial ad energy density @xmath65 , so there is potentially a lot of energy to be transferred .
however , one would expect the energy flow to decrease as the amplitude of the flaton decreases due to preheating , energy transfer and eventually hubble damping .
the efficiency of these effects will be crucial in determining whether our lepton number asymmetry survives .
however , it is unclear how efficient any of these effects is and even which of them is dominant .
so , to start somewhere , in this paper we analyze the preheating effect numerically , which also includes the energy flow from the flaton to the ad fields .
once the amplitude of the ad fields has been damped sufficiently for the lepton violating terms @xmath66 and @xmath60 to become negligible , @xmath67 will be conserved .
we denote this initial time by @xmath68 .
the ad fields will complete their decay to the thermal bath while the temperature is still above the electroweak scale , and so sphaleron processes will convert our lepton asymmetry to a baryon asymmetry
. finally , the flaton decay completes , diluting the baryon asymmetry .
we denote this final time by @xmath69 .
the final baryon asymmetry is given by @xmath70 where @xmath71 and @xmath72 are energy density and temperature of the universe .
we use the factor @xmath73 to give a dimensionless measure of the efficiency of the lepton asymmetry generation and preservation .
it can be interpreted roughly as the ratio of lepton number density to the total ad number density when the ad fields start oscillating .
we expect @xmath74 and @xmath75 as described in section [ physpara ] and ref .
the last factor in eq .
( [ nb ] ) depends on the equation of state of the flaton .
if the preheating of the flaton only populates modes with @xmath76 , then the flaton would have @xmath77 .
the energy flow from the flaton sector to the ad sector could also affect this factor , but , although this flow is likely significant from the ad point of view , we guess it is not so significant from the flaton point of view . with these assumptions ,
one expects @xmath78 thus , we obtain @xmath79 compared with the observational result @xmath80
we performed a three dimensional lattice simulation with periodic boundary conditions using the algorithm described in the appendix .
the potential was that of eq .
( [ v ] ) with constraints eqs .
( [ d ] ) and ( [ j ] ) , and canonical gauge invariant kinetic terms . the constraint eq .
( [ d ] ) was modified to @xmath81 to cut off the singularity at @xmath82 .
the physical parameters and fields were rescaled by a typical electroweak scale mass @xmath23 and a typical flaton or ad field expectation value @xmath83 or @xmath84 , as described in ref .
@xcite .
the lattice volume @xmath85 , number of lattice points @xmath86 , time step @xmath87 , and @xmath13-term constraint singularity cutoff @xmath88 , were taken as @xmath89 we tested our results using different values of these numerical parameters .
limited computing power constrained us to @xmath90 .
the @xmath91-modes allowed by this lattice are @xmath92 with @xmath93 where @xmath94 with @xmath95 having degeneracy 1 and the rest degeneracy 2 .
this spans the range @xmath96 the physical parameters were taken as follows .
@xmath97 the mass - squareds were given with values up to three decimal places to avoid any accidental resonances and also to fit the mssm constraints described in ref .
@xmath98 @xmath99 @xmath100 @xmath101 note that the phases of eq .
( [ phaseredef ] ) are not physical and can be adjusted by field rotations .
the cp phases were taken as @xmath102 for the choice of parameters above , the formulae of section [ physpara ] become @xmath103 @xmath104 @xmath105 @xmath106 @xmath107 @xmath108 the real initial conditions for the leptogenesis includes the realization of the first order phase transitions of @xmath49 and @xmath2 , or at least the first order phase transition of @xmath2 and the state of @xmath49 at that time , as discussed in section [ leptoini ] . during the transitions , bubbles of non - zero field value nucleate ,
expand and percolate .
the typical size of a bubble just before percolation is expected to be slightly smaller than the hubble radius .
this introduces a scale much larger than the inverse electroweak scale typical of the rest of the dynamics , making it difficult to simulate both the phase transitions and the leptogenesis dynamics simultaneously .
we leave detailed analytic and numerical investigation of these initial conditions to a subsequent paper , and here just use a simple ansatz that covers a range of possibilities , hopefully without missing too much relevant physics .
we do not include a thermal bath as after the fields roll away from the origin , they decouple from the thermal bath , apart from effects like thermal friction which we plan to consider in a subsequent paper .
we set the initial conditions as @xmath109 with the initial conditions for @xmath37 determined by @xmath110 we choose the initial homogeneous displacement of @xmath2 to be either @xmath111 or @xmath112 .
@xmath111 corresponds roughly to a second order phase transition , and provides one extreme .
a finite value of @xmath113 is meant to correspond roughly to the displaced field inside a bubble .
@xmath13 is given by eq .
( [ d ] ) and @xmath114 leads via eq .
( [ j ] ) to @xmath115 @xmath116 , @xmath117 and @xmath118 are inhomogeneous fluctuations taken to be the combination of vacuum and thermal fluctuations at a temperature @xmath119 .
since , as discussed in section [ leptoini ] , in some cases @xmath117 may be quite large , we also test the effects of an initial spectrum of @xmath117 fluctuations with a thermal spectrum of temperature @xmath120 up to the lattice cutoff @xmath121 , roughly corresponding to a preheated state with @xmath122 .
our most important results are figs .
[ d4l ] and [ d0l ] showing the lepton number asymmetry plotted as a function of time . in figs .
[ d4l ] and [ d0l ] , the red and blue lines show that a lepton number asymmetry is generated , with sign depending on the sign of the cp violating phases .
the green line is the result of a simulation with cp conserving phases used to estimate the noise level due to the random initial conditions and finite size of the lattice . in the case of @xmath112 ,
the angle of initial displacement can contribute to the cp violation and hence affect the lepton asymmetry , as shown in fig .
[ d4lphase ] .
we average over this initial angle , corresponding to summing over bubbles with different initial flaton phases , to obtain the lepton asymmetry of fig .
[ d4l ] . a common feature of figs .
[ d4l ] and [ d0l ] is that , although some of the initially generated asymmetry is washed out , a significant fraction is conserved to late times .
the initial conditions , @xmath111 and @xmath112 , correspond to a range of initial @xmath113 so these results suggest that conservation can be achieved for a wide range of initial conditions
. moreover , fig .
[ d0lt ] shows that the lepton asymmetry is not sensitive to the state of @xmath49 at the time when the flaton starts to roll away .
this suggests that the fluctuations of the ad fields acquired from those of the flaton through their couplings dominate over any initial fluctuations of @xmath49 .
the lepton asymmetry declines slowly at late times .
this is due to the preheating becoming saturated and energy flow from the flaton sector , as will be discussed in sections [ preheating ] and [ energyflow ] . in figs .
[ d0l ] and [ d0lt ] , the @xmath123 lepton asymmetries are enhanced after a short period of settling down .
it is unclear what this is due to , but it does not seem to be harmful because it disappears leaving a large enough lepton number asymmetry .
the spatial distribution of the lepton number density shown in fig .
[ d0cpplnumsl ] shows a large ball type structure , which seems to be responsible for the strange behavior in fig .
we suspect it may be a quasi - stable q - ball @xcite , but it is unclear .
we postpone proper analysis of this to future work .
thus we have demonstrated the central aim of this paper , that preheating can damp the amplitude of the affleck - dine fields sufficiently to conserve the generated asymmetry .
we hope to extend this result to more realistic initial conditions in a future paper .
the dispersions and spectra of the fields give a more detailed picture of the dynamics leading to the conservation of the lepton asymmetry .
the field amplitudes in figs .
[ d4v ] and [ d0v ] ( top left of each figure ) show the ad fields settling down into the symmetric region of the potential near the origin , which leads to the conservation of the lepton number .
this is a consequence of very efficient preheating as shown in the graphs of the field dispersions ( left middle and bottom ) and mean squares about vacuum expectation value ( top right ) in the figures .
while the decreasing behavior around @xmath124 in the graphs of mean square about vacuum expectation value is due to energy flow from homogeneous modes to inhomogeneous modes , the similar behavior in the dispersion graphs is due to energy flow to higher @xmath91-modes . as is clear from figs .
[ d4l ] and [ d0l ] , and the graphs in figs .
[ d4v ] and [ d0v ] , the two cases of @xmath125 and @xmath126 have different initial transient behaviors .
this is due to the difference in the efficiency of the preheating of the flaton . for @xmath112 ,
the flaton has an initial homogeneous displacement larger than the fluctuations , so that it rolls away along a specific direction with relatively small fluctuations , leading to slower growth of gradient energy .
this results in the flaton oscillating several times before settling down to its minima , and causes the violent initial behavior of the ad fields due to their flaton - dependent mass - squared s whose sign also depends on the flaton field value .
however , for @xmath111 , the flaton rolls away from the origin with initial fluctuations in all directions .
this leads to rapid growth of gradient energy , resulting in the flaton settling down to its minima within a single oscillation .
note that the large initial flaton dispersion and its drop around @xmath124 in fig .
[ d0v ] is due to the formation and annihilation of flaton strings and walls , which can be seen in fig .
[ d0cppphasesl ] .
in contrast , fig .
[ d4cppphasesl ] shows that the @xmath112 case is initially dominated by a single domain , as would be expected due to its homogeneous displacement .
our model has a @xmath127 gauge symmetry identifying the four flaton vacua @xcite .
however , a discrete gauge symmetry is more difficult to simulate than a global symmetry , so in our simulation we used a @xmath127 global symmetry , and relied on the random initial conditions and finite size of the lattice to get the strings and walls to annihilate , in accord with the dynamics expected for the @xmath127 gauge symmetry .
after the initial transient stage , the amplitudes of the ad fields start to grow again .
this is because the energy flow from the flaton sector becomes greater than the rate of energy dispersion to higher @xmath91-modes due to preheating and thermalization .
this is the reason why the lepton asymmetry in figs .
[ d4l ] and [ d0l ] slowly decreases at late times .
the difference in the late time rate of increase of the ad amplitudes in figs .
[ d4v ] and [ d0v ] is due to difference in the rate of energy flow into the ad sector , which is caused by the difference in field amplitudes at the time when preheating is effectively shut down .
so the efficiency of preheating at the initial transient stage determines the late time behavior of the graphs .
this is shown well by the lattice cutoff dependence shown in fig .
[ d0cppvslattice ] .
the spectra in fig .
[ kspec ] show the amplification of @xmath91-modes through preheating , which damps the homogeneous motion of the flaton and ad fields . in the flaton sector
, we initially have tachyonic growth of low @xmath91-modes as the flaton rolls away from the origin .
after this initial stage of preheating due to instability , the flaton oscillates around its minimum . at this stage ,
nonlinear effects populate slightly higher @xmath91-modes of the flaton . at late times
, the flaton spectrum is cut off at high @xmath91 , which is consistent with theoretical expectations @xcite .
the preheating of the flaton sector is transferred to the ad sector through the flaton - dependent potential of the ad fields .
remarkably , ad @xmath91-modes higher than the physical cutoff in the flaton sector are also rapidly populated hitting the high @xmath91 lattice cutoff before @xmath124 with rapid energy flow from lower to higher @xmath91-modes .
as a result , the ad lattice modes are fully thermalized for most of the simulation .
once all the lattice modes have been thermalized , preheating is truncated artificially in the simulation , even though it may continue in reality .
this suggests that the reliability of the lattice results will be compromised at late times , as energy that physically should be flowing into the higher @xmath91 modes is artificially , numerically shunted into the modes that are present in the grid .
this effect should artifically weaken preheating and we believe it accounts for some of the late time decrease in the lepton asymmetry seen in figs .
[ d4l ] and [ d0l ] .
we have performed experiments with coarser lattices ( @xmath128 ) and shown that preheating in the ad sector becomes stronger as the uv resolution is increased , confirming our expectation that our limited resolution is artificially damping the effects of preheating .
( these experiments also show that the grid resolution is not affecting the flaton sector . )
our results thus present a lower bound on the effects of preheating .
thus in reality one should expect more preheating in the ad sector and hence better conserved lepton number than the simulation shows .
we postpone further study of this issue to future work . figs .
[ d4cppenergy ] to [ d0cppenergy ] show the components of the energy of the flaton and ad fields as a function of time with total energy densities defined as @xmath129 where @xmath130 denotes the lattice average and @xmath131 + \left| \lambda_\phi \phi^3 \right|^2 \\ v_\mathrm{ad}(\phi,\psi ) & \equiv & v - v_\mathrm{fl}\end{aligned}\ ] ] as @xmath2 rolls away from the origin , its potential energy @xmath22 is rapidly converted into kinetic and gradient energy . even at late times
, the flaton s potential energy is slowly being converted into kinetic and gradient energy , due to continued preheating .
the flaton dependence of the ad potential initially generates potential energy for the ad fields , which is then rapidly converted to kinetic and gradient energy . at late times
, the ad lattice modes have thermalized and the energy from the flaton sector flows into kinetic , gradient and potential energies of the ad fields .
the total energy flow from flaton to ad fields is shown in figs .
[ d4cppenergytot ] and [ d0cppenergytot ] .
the qualitative features of figs .
[ d4cppenergytot ] and [ d0cppenergytot ] can be understood from the energy flow rate into the ad sector @xmath132 which is obtained by use of the equation of motion of @xmath2 .
we can separate the energy flow into three stages . at the first stage ,
the ad potential is lifted up causing the ad fields to oscillate with large amplitude .
then , preheating spreads energy into higher @xmath91-modes , reducing the field amplitudes .
this results in decreasing energy flow rate . finally , the ad sector preheating is saturated due to the lattice cutoff , so energy from the flaton sector starts being accumulated in a fixed number of modes .
this causes increasing field amplitude and therefore increasing energy flow rate .
thus , the late time behavior of the energy flow depends on the efficiency of preheating , which in turn depends on the initial conditions and the lattice cutoff .
the dependence on the initial conditions explains the different rates of decline in figs .
[ d4l ] and [ d0l ] .
we expect a smaller energy flow for a finer lattice , as shown in fig .
[ d0cppvslattice ] , but this will be the case only until a physical cutoff of the ad preheating appears .
if the preheating becomes saturated due to a physical cutoff and the energy flow is still significant , then we may need the help of thermal friction which could provide another channel for energy to flow away .
we plan to study these issues in a future paper .
recently , we proposed a minimal extension of the mssm which naturally gives rise to thermal inflation followed by ad leptogenesis .
the expansion rate @xmath133 at the end of thermal inflation is too low for the usual ad hubble damping mechanism to conserve the generated lepton asymmetry , but we proposed that preheating and thermal friction may perform the same function .
however , those processes are very complicated and not easy to study analytically even in a simple case .
it is even difficult to know which of these processes will dominate since it depends on the details of the dynamics . in this paper
we numerically analyzed the effect of preheating on the conservation of the lepton number asymmetry in our model .
the lepton number asymmetry is generated as the flaton rolls away at the end of thermal inflation .
it will be conserved if the preheating of the ad sector is efficient enough that the homogeneous oscillation amplitudes of the ad fields are quickly reduced , bringing the fields into the symmetric region of the potential near the origin .
the preheating process starts in the flaton sector by the magnification of the initial fluctuations , that include vacuum and thermal contributions , as the flaton rolls away from the origin and oscillates about the minimum of its potential . since the flaton potential has negative curvature near the origin and somewhat larger angle - dependent curvature along angular directions far from the origin , very rapid tachyonic preheating is dominant initially .
when the flaton oscillates around its minimum preheating continues due to the anharmonic terms in its potential ( weak parametric resonance ) .
the ad fields in our model have flaton - dependent masses , so their mass squareds vary with time and even change sign .
this violently transfers inhomogeneities in the flaton sector to the ad sector .
the preheating amplifies high @xmath91-modes in the ad sector much more rapidly than in the flaton sector , so that the ad fields quickly thermalize over the range of momenta present in our lattice .
if we were able to simulate with higher resolution we would expect to see even more efficient preheating .
the flaton sector also feeds energy into the ad sector .
the energy flow rate depends on the amplitudes of the fields and the time derivative of the flaton .
this means that if the flaton and ad fields are efficiently preheated , resulting in small field values , the energy flow rate will be small . since the ad sector preheating is very rapid initially ,
the energy flow in the initial transient stage is not harmful in terms of lepton number conservation .
but the preheating is quickly saturated , causing the energy flowing from the flaton sector to accumulate in a fixed number of modes .
this causes an increase in the oscillation amplitude of the ad fields , which can wash out the asymmetry , as well as increase the energy flow rate . in this sense ,
the balance between the preheating in the ad sector and the energy flow from the flaton sector is crucial for the late time evolution of the lepton asymmetry .
there is a large structure in the spatial distribution of the lepton asymmetry , which may be a quasi - stable q - ball , but it is not harmful since it soon disappeared without much effect on the asymmetry . the origin of this object is not clear . including the renormalisation of the ad masses should make such objects less stable , assuming it is a type of q - ball .
the outcome of all these effects is , as the simulations show , that the lepton asymmetry in our model is reasonably well conserved at late times due to the efficient preheating , and is not affected much by the initial state of the ad fields .
the lepton asymmetry decreases slowly at late times but the rate is small enough for one to expect that sufficient lepton asymmetry will be preserved for long enough for other physical processes like thermal friction to start working and stop the decrease .
there are several limitations to this work , which may affect the conservation of the lepton asymmetry .
firstly , all of our results may depend crucially on the initial conditions .
we found that the initial state of the ad fields has little effect on the lepton asymmetry .
the initial state of the flaton , however , affects the preheating of the ad fields and the energy flow , and thus may affect the conservation of the lepton asymmetry .
a proper treatment of this issue will thus require simulating flaton bubbles .
once this has been done , it becomes possible to determine the regions of parameter space in which our model works .
the limited size and resolution of the lattice also affect the preheating efficiency .
since the lattice cutoff is reached early in our current simulation , it restricts the preheating artificially .
a larger and finer lattice would provide more available modes for preheating and thermalization such that it would weaken or even stop the decreasing behavior of the lepton asymmetry .
finally , we may have to consider thermal friction as another source of damping for the homogeneous mode , if it turns out that preheating is not efficient enough to conserve the lepton asymmetry by itself given realistic initial conditions for the flaton .
we postpone all of these issues to future work . in summary
, our simulation revealed that preheating is efficient and rapid enough to conserve the lepton asymmetry in our low energy ad leptogenesis model , replacing the hubble damping of the usual high energy ad scenario .
eds thanks masahide yamaguchi for collaboration on an earlier related project .
gnf , wip and eds thank the canadian institute for theoretical astrophysics ( cita ) for hosting them during the early stages of this project .
hbk , wip and eds thank the astrophysical research center for the structure and evolution of the cosmos ( arcsec ) and cita for providing computing resources .
hbk , wip and eds would like to acknowledge the support from the korea institute of science and technology information ( kisti ) under ` the strategic supercomputing support program ' .
the use of the computing system of the supercomputing center in kisti is also greatly appreciated .
gnf was supported by nsf grant phy-0456631 .
hbk , wip and eds were supported in part by arcsec funded by the korea science and engineering foundation ( kosef ) and the korean ministry of science , the korea research foundation grants krf pbrg 2002 - 070-c00022 and krf-2005 - 201-c00006 funded by the korean government ( moehrd ) , the kosef grant r01 - 2005 - 000 - 10404 - 0 , and brain korea 21 .
in this appendix , we describe our algorithm for numerical integration of our dynamics .
it is a gauge invariant adaptive leapfrog type algorithm with constraints .
it is derived from a discrete action , and so is a variational integrator @xcite , which means that it exactly conserves the constraints and charges , and has good long term energy conservation .
the adaptivity is particularly important to help deal with the singular configuration manifold induced by the @xmath13-term constraint of eq .
( [ d ] ) . the remaining gauge field in our model , corresponding to the @xmath13-term and gauge constraints of eqs .
( [ d ] ) and ( [ j ] ) , is heavy and so not dynamical , but , unlike the other gauge fields , can not be easily integrated out .
a gauge field is a connection relating variables at two different points in space - time , and so should be considered as a link variable living on a link joining two points .
we will follow the usual way of defining covariant derivatives on a lattice in terms of link variables .
let s denote nodes by @xmath134 and links by @xmath36 .
a link @xmath36 connects from the node @xmath135 to the node @xmath136 .
the links @xmath137 flow into , and the links @xmath138 flow out of , the node @xmath134 .
the length of a link @xmath36 is @xmath139 .
a link @xmath140 is in the time direction and a link @xmath141 is in the space direction .
( 40,30)(-10,-5 ) ( 0,0)(10,0)3 ( 0,10)(10,0)3 ( 0,20)(10,0)3 ( 20,5 ) ( 10,12.5 ) ( 10,15 ) ( 0.5,0)(10,0)2(1,0)9 ( 0.5,10)(10,0)2(1,0)9 ( 0.5,20)(10,0)2(1,0)9 ( 0,0.5)(0,1)9 ( 0,10.5)(0,1)9 ( 10,0.5)(0,1)9 ( 10,10.5)(0,1)1.5 ( 10,13)(0,1)1.5 ( 10,15.5)(0,1)4 ( 20,0.5)(0,1)4 ( 20,5.5)(0,1)4 ( 20,10.5)(0,1)9 ( 2.5,-5)@xmath152 ( -6,3.5)@xmath87 our discrete action is the sum of kinetic terms on time links , gradient terms on space links , and potential and constraint terms on nodes , with spacetime volume weightings @xmath153 appropriate to their domains @xmath154 where @xmath155 we choose a regular spacetime lattice with spatial lattice spacing @xmath156 and time step @xmath87 for the gradient and potential terms .
this regular lattice is denoted by the solid nodes and the links joining them in fig .
[ lattice ] .
extra nodes , the hollow nodes in fig .
[ lattice ] , are added adaptively to the regular lattice s time links , and we use this adaptive lattice for the kinetic and constraint terms .
this helps to follow the dynamics on our constraint manifold which is singular in the limit @xmath157 . varying our action of eq .
( [ action ] ) with respect to the lagrange multiplier @xmath160 gives the d - term constraint @xmath161 therefore @xmath162 varying with respect to the gauge field @xmath146 gives the gauge constraint @xmath163 combining eqs .
( [ dsdl ] ) and ( [ dsda ] ) gives @xmath164 for @xmath166 with @xmath167 eq .
( [ gij ] ) simplifies to @xmath168 choosing the positive solution consistent with the limit @xmath169 , we get @xmath170 note that in our case of @xmath171 and @xmath172 , eq . ( [ dcon ] ) gives @xmath173 we use eq .
( [ asol ] ) to determine @xmath174 , and from eq .
( [ gaugefix ] ) we have @xmath175 varying our action of eq .
( [ action ] ) with respect to @xmath176 gives @xmath177 where @xmath178 the hat indicates improper normalization due to the use of @xmath87 instead of @xmath179 , and @xmath180 solving for @xmath181 we get @xmath182 where @xmath183 therefore , defining @xmath184 and @xmath185 eq .
( [ dsdphi ] ) becomes @xmath186 eq .
( [ con ] ) gives @xmath187_{tn+ } \left ( \frac{n}{n^\dagger n } \right)_{n } \delta_{tn+}\ ] ] and eq .
( [ pphi ] ) gives @xmath188 combining these gives @xmath189 \nonumber \\ \label{pstep0 } & & { } - \frac{1}{2 } \left [ ( \mathcal{d}\phi)^\dagger q ( \mathcal{d}\phi ) \right]_{tn+ } \left ( \frac{n}{n^\dagger n } \right)_{n } \delta_{tn+}\end{aligned}\ ] ] implementing our gauge choice of eqs .
( [ gaugefix ] ) and ( [ gaugefixu ] ) and defining @xmath190 @xmath191\ ] ] @xmath192 and @xmath193 reduces eq .
( [ pstep0 ] ) to @xmath194 while eq .
( [ dl ] ) reduces to @xmath195 eqs .
( [ pstep ] ) and ( [ xstep ] ) are our basic algorithm , but we still need to determine @xmath196 . contracting eq .
( [ pstep ] ) around @xmath145 gives @xmath197 \tilde\delta_{n } + \frac{1}{4 } \pi_{tn+}^2 \left ( n^\dagger q n \right)_n \tilde\delta_{n}^2\ ] ] choosing the solution which is well behaved in the limit @xmath198 , we get @xmath199 ^ 2 - \pi_{n}^\dagger q \pi_{n } n_n^\dagger q n_n \tilde\delta_{n}^2 } \,}\ ] ] the lepton current is @xmath205 where @xmath206 , @xmath207 and @xmath208 . for our gauge choice of eqs .
( [ gaugefix ] ) and ( [ gaugefixu ] ) the lepton density reduces to @xmath209 our algorithm gives @xmath210 \delta_{\bullet n}\end{aligned}\ ] ] therefore our algorithm exactly conserves lepton number apart from the lepton number violating terms in the potential . our algorithm of eqs .
( [ pstep ] ) and ( [ xstep ] ) , with @xmath211 given by eq .
( [ pi ] ) , @xmath212 given by eq .
( [ cds ] ) , @xmath213 determined by eq .
( [ asol ] ) and @xmath196 given by eq .
( [ pisol ] ) , exactly conserves the @xmath13-term constraint @xmath214 the gauge constraint @xmath215 and lepton number apart from the lepton number violating terms in the potential , and has good long term energy conservation , as is expected for a variational integrator @xcite .
a. d. sakharov , `` violation of cp invariance , c asymmetry , and baryon asymmetry of the universe '' , pisma zh .
fiz . 5 ( 1967 ) 32 , [ jetp lett . 5 , 24 ( 1967 ) ]
m. y. khlopov and a. d. linde , `` is it easy to save the gravitino ? '' , phys .
b138 ( 1984 ) 265 ; j. ellis , j. e. kim and d. v. nanopoulos , `` cosmological gravitino regeneration and decay '' , phys .
b145 ( 1984 ) 181 ; m. kawasaki and t. moroi , `` gravitino production in the inflationary universe and the effects on big bang nucleosynthesis '' , prog .
( 1995 ) 879 , [ arxiv : hep - ph/9403364 ] ; m. yu .
khlopov , yu . l. levitan , e. v. sedelnikov and i. m. sobol , `` nonequilibrium cosmological nucleosynthesis of light elements : calculations by the monte carlo method '' , phys . at .
nuclei 57 ( 1994 ) 1393 ; m. kawasaki , k. kohri and t. moroi , `` hadronic decay of late - decaying particles and big - bang nucleosynthesis '' , [ arxiv : astro - ph/0402490 ] . g. d. coughlan , w. fischler , e. w. kolb , s. raby and g. g. ross , `` cosmological problems for the polonyi potential '' , phys .
b131 ( 1983 ) 59 ; j. r. ellis , d. v. nanopoulos and m. quiros , `` on the axion , dilaton , polonyi , gravitino and shadow matter problems in supergravity and superstring models '' , phys .
b174 ( 1986 ) 176 ; t. banks , d. b. kaplan and a. e. nelson , `` cosmological implications of dynamical supersymmetry breaking '' , phys .
d49 ( 1994 ) 779 , [ arxiv : hep - ph/9308292 ] ; b. de carlos , j. a. casas , f. quevedo and e. roulet , `` model - independent properties and cosmological implications of the dilaton and moduli sectors of 4-d strings '' , phys .
b318 ( 1993 ) 447 , [ arxiv : hep - ph/9308325 ] . e.w . kolb and m.s .
turner , `` the early universe '' , ( addison - wesley , 1990 ) .
d. h. lyth and e. d. stewart ,
`` cosmology with a tev mass gut higgs '' , phys .
lett . 75 ( 1995 ) 201 ,
[ arxiv : hep - ph/9502417 ] ; `` thermal inflation and the moduli problem '' , phys .
d53 ( 1996 ) 1784 , [ arxiv : hep - ph/951020 ] .
k. yamamoto , `` phase transition associated with intermediate gauge symmetry breaking in superstring models '' , phys .
lett . b168 ( 1986 ) 341 ; o. bertolami and g. g. ross , `` inflation as a cure for the cosmological problems of superstring models with intermediate scale breaking '' , phys .
b183 ( 1987 ) 163 ; j. r. ellis , k. enqvist , d. v. nanopoulos and k. a. olive , `` comments on intermediate scale models '' , phys .
b188 ( 1987 ) 415 .
early attempts at baryogenesis in the context of thermal inflation include : r. n. mohapatra and j. w. f. valle , `` late baryogenesis in superstring models '' , phys .
b186 ( 1987 ) 303 ; k. yamamoto , `` a model for baryogenesis in superstring unification '' , phys .
b194 ( 1987 ) 390 ; g. lazarides , c. panagiotakopoulos and q. shafi , `` baryogenesis in superstring motivated models '' , nucl .
b307 ( 1988 ) 937 .
e. d. stewart , m. kawasaki and t. yanagida , `` affleck - dine baryogenesis after thermal inflation '' , phys .
d54 ( 1996 ) 6032 - 6039 , [ arxiv : hep - ph/9603324 ] donghui jeong , kenji kadota , wan - il park and ewan d. stewart , `` modular cosmology , thermal inflation , baryogenesis and a prediction for particle accelerators '' , jhep 0411 : 046 , 2004 , [ arxiv : hep - ph/0406136 ] .
i. affleck and m. dine , `` a new mechanism for baryogenesis '' , nucl .
b249 ( 1985 ) 361 ; m. dine , l. randall and s. thomas , `` supersymmetry breaking in the early universe '' , phys .
75 ( 1995 ) 398 , [ arxiv : hep - ph/9503303 ] ; `` baryogenesis from flat directions of the supersymmetric standard model '' , nucl .
b458 ( 1996 ) 291 , [ arxiv : hep - ph/9507453 ] .
l. kofman , a. d. linde and a. a. starobinsky , `` reheating after inflation '' , phys .
73 ( 1994 ) 3195 - 3198 ,
[ arxiv : hep - th/9405187 ] . g. n. felder , j. garcia - bellido , p. b. greene , l. kofman , a. d. linde and i. tkachev , `` dynamics of symmetry breaking and tachyonic preheating '' , phys .
( 2001 ) 011601 ,
[ arxiv : hep - ph/0012142 ] . w. i. park , `` affleck - dine leptogenesis induced by the flaton of thermal inflation '' , summer institute 2006 ( international workshop on supersymmetry , electroweak symmetry breaking and particle cosmology ) , apctp , pohang , korea , aug 23 - 30 , 2006 .
m. kawasaki and k. nakayama , `` late - time affleck - dine baryogenesis after thermal inflation '' , [ arxiv : hep - ph/0608335 ] .
h. komatsu , `` new constraints on parameters in the minimal supersymmetric model '' , phys . lett .
b215 ( 1988 ) 323 ; j.a .
casas , a. lleyda and c. munoz , `` strong constraints on the parameter space of the mssm from charge and color breaking minima '' , nucl .
b471 ( 1996 ) 3 , [ arxiv : hep - ph/9507294 ] .
s. r. coleman , `` q balls '' , nucl . phys .
b262 ( 1985 ) 263 ; a. kusenko , `` solitons in the supersymmetric extensions of the standard model '' , phys .
b405 ( 1997 ) 108 , [ arxiv : hep - ph/9704273 ] . k. enqvist and j. mcdonald , `` q balls and baryogenesis in the mssm '' , phys .
b425 ( 1998 ) 309 , [ arxiv : hep - ph/9711514 ] .
k. enqvist and j. mcdonald , `` b - ball baryogenesis and the baryon to dark matter ratio '' , nucl .
b538 ( 1999 ) 321 , [ arxiv : hep - ph/9803380 ] .
a. lew , j. e. marsden , m. ortiz and m. west , `` an overview of variational integrators '' , in l. p. franca , editor , `` finite element methods : 1970s and beyond '' ( cimne , barcelona , spain , 2003 ) ; m. west , `` variational integrators '' , ph.d .
thesis ( california institute of technology , 2004 ) .
lepton number density , averaged over the lattice , as a function of the initial phase of the flaton , for @xmath112 .
the @xmath127 symmetry of our model means that the range of initial angles is @xmath216 . ]
lepton number density , averaged over the lattice , as a function of the initial phase of the flaton , for @xmath112 .
the @xmath127 symmetry of our model means that the range of initial angles is @xmath216 . ]
lepton number density , averaged over the lattice , as a function of time , for @xmath111 , @xmath123 and @xmath217 , with different initial effective temperatures for @xmath49 .
we set an initial thermal spectrum up to the lattice cutoff @xmath121 , so that an effective temperature of @xmath120 would correspond to a preheated state with fluctuation @xmath218 . ]
lepton number density , averaged over the lattice , as a function of time , for @xmath111 , @xmath123 and @xmath217 , with different initial effective temperatures for @xmath49 .
we set an initial thermal spectrum up to the lattice cutoff @xmath121 , so that an effective temperature of @xmath120 would correspond to a preheated state with fluctuation @xmath218 . ]
lepton number density as a function of the 2d lattice for time slices at @xmath219 , for @xmath112 , @xmath123 , initial flaton phase @xmath220 . , title="fig:",scaledwidth=50.0% ] lepton number density as a function of the 2d lattice for time slices at @xmath219 , for @xmath112 , @xmath123 , initial flaton phase @xmath220 .
, title="fig:",scaledwidth=50.0% ] lepton number density as a function of the 2d lattice for time slices at @xmath219 , for @xmath112 , @xmath123 , initial flaton phase @xmath220 . , title="fig:",scaledwidth=50.0% ] lepton number density as a function of the 2d lattice for time slices at @xmath219 , for @xmath112 , @xmath123 , initial flaton phase @xmath220 . , title="fig:",scaledwidth=50.0% ] lepton number density as a function of the 2d lattice for time slices at @xmath219 , for @xmath112 , @xmath123 , initial flaton phase @xmath220 . , title="fig:",scaledwidth=50.0% ] lepton number density as a function of the 2d lattice for time slices at @xmath219 , for @xmath112 , @xmath123 , initial flaton phase @xmath220 . , title="fig:",scaledwidth=50.0% ] lepton number density as a function of the 2d lattice for time slices at @xmath219 , for @xmath111 , @xmath123 . , title="fig:",scaledwidth=50.0% ] lepton number density as a function of the 2d lattice for time slices at @xmath219 , for @xmath111 , @xmath123 . , title="fig:",scaledwidth=50.0% ] lepton number density as a function of the 2d lattice for time slices at @xmath219 , for @xmath111 , @xmath123 .
, title="fig:",scaledwidth=50.0% ] lepton number density as a function of the 2d lattice for time slices at @xmath219 , for @xmath111 , @xmath123 . , title="fig:",scaledwidth=50.0% ] lepton number density as a function of the 2d lattice for time slices at @xmath219 , for @xmath111 , @xmath123 . , title="fig:",scaledwidth=50.0% ] lepton number density as a function of the 2d lattice for time slices at @xmath219 , for @xmath111 , @xmath123 .
, title="fig:",scaledwidth=50.0% ] dispersions and variances averaged over the lattice , as a function of time , for @xmath112 , @xmath221 , @xmath123 .
left top - field amplitudes in log scale ; left middle and bottom - dispersions in log and linear scales ; right top - mean square with respect to vev in log scale ; right middle and bottom - mean squares with respect to origin in log and linear scales .
, title="fig:",scaledwidth=50.0% ] dispersions and variances averaged over the lattice , as a function of time , for @xmath112 , @xmath221 , @xmath123 .
left top - field amplitudes in log scale ; left middle and bottom - dispersions in log and linear scales ; right top - mean square with respect to vev in log scale ; right middle and bottom - mean squares with respect to origin in log and linear scales .
, title="fig:",scaledwidth=50.0% ] dispersions and variances averaged over the lattice , as a function of time , for @xmath112 , @xmath221 , @xmath123 .
left top - field amplitudes in log scale ; left middle and bottom - dispersions in log and linear scales ; right top - mean square with respect to vev in log scale ; right middle and bottom - mean squares with respect to origin in log and linear scales . ,
title="fig:",scaledwidth=50.0% ] dispersions and variances averaged over the lattice , as a function of time , for @xmath112 , @xmath221 , @xmath123 .
left top - field amplitudes in log scale ; left middle and bottom - dispersions in log and linear scales ; right top - mean square with respect to vev in log scale ; right middle and bottom - mean squares with respect to origin in log and linear scales .
, title="fig:",scaledwidth=50.0% ] dispersions and variances averaged over the lattice , as a function of time , for @xmath112 , @xmath221 , @xmath123 .
left top - field amplitudes in log scale ; left middle and bottom - dispersions in log and linear scales ; right top - mean square with respect to vev in log scale ; right middle and bottom - mean squares with respect to origin in log and linear scales . ,
title="fig:",scaledwidth=50.0% ] dispersions and variances averaged over the lattice , as a function of time , for @xmath112 , @xmath221 , @xmath123 .
left top - field amplitudes in log scale ; left middle and bottom - dispersions in log and linear scales ; right top - mean square with respect to vev in log scale ; right middle and bottom - mean squares with respect to origin in log and linear scales . ,
title="fig:",scaledwidth=50.0% ] dispersions and variances averaged over the lattice , as a function of time , for @xmath111 , @xmath123 .
left top - field amplitudes in log scale ; left middle and bottom - dispersions in log and linear scales ; right top - mean square with respect to vev in log scale ; right middle and bottom - mean squares with respect to origin in log and linear scales .
, title="fig:",scaledwidth=50.0% ] dispersions and variances averaged over the lattice , as a function of time , for @xmath111 , @xmath123 .
left top - field amplitudes in log scale ; left middle and bottom - dispersions in log and linear scales ; right top - mean square with respect to vev in log scale ; right middle and bottom - mean squares with respect to origin in log and linear scales . ,
title="fig:",scaledwidth=50.0% ] dispersions and variances averaged over the lattice , as a function of time , for @xmath111 , @xmath123 .
left top - field amplitudes in log scale ; left middle and bottom - dispersions in log and linear scales ; right top - mean square with respect to vev in log scale ; right middle and bottom - mean squares with respect to origin in log and linear scales . ,
title="fig:",scaledwidth=50.0% ] dispersions and variances averaged over the lattice , as a function of time , for @xmath111 , @xmath123 .
left top - field amplitudes in log scale ; left middle and bottom - dispersions in log and linear scales ; right top - mean square with respect to vev in log scale ; right middle and bottom - mean squares with respect to origin in log and linear scales .
, title="fig:",scaledwidth=50.0% ] dispersions and variances averaged over the lattice , as a function of time , for @xmath111 , @xmath123 .
left top - field amplitudes in log scale ; left middle and bottom - dispersions in log and linear scales ; right top - mean square with respect to vev in log scale ; right middle and bottom - mean squares with respect to origin in log and linear scales . ,
title="fig:",scaledwidth=50.0% ] dispersions and variances averaged over the lattice , as a function of time , for @xmath111 , @xmath123 .
left top - field amplitudes in log scale ; left middle and bottom - dispersions in log and linear scales ; right top - mean square with respect to vev in log scale ; right middle and bottom - mean squares with respect to origin in log and linear scales . ,
title="fig:",scaledwidth=50.0% ] flaton phase as a function of the 2d lattice for time slices at @xmath219 , for @xmath112 , @xmath220 , @xmath123 .
, title="fig:",scaledwidth=50.0% ] flaton phase as a function of the 2d lattice for time slices at @xmath219 , for @xmath112 , @xmath220 , @xmath123 . , title="fig:",scaledwidth=50.0% ] flaton phase as a function of the 2d lattice for time slices at @xmath219 , for @xmath112 , @xmath220 , @xmath123 .
, title="fig:",scaledwidth=50.0% ] flaton phase as a function of the 2d lattice for time slices at @xmath219 , for @xmath112 , @xmath220 , @xmath123 . , title="fig:",scaledwidth=50.0% ] flaton phase as a function of the 2d lattice for time slices at @xmath219 , for @xmath112 , @xmath220 , @xmath123 . , title="fig:",scaledwidth=50.0% ] flaton phase as a function of the 2d lattice for time slices at @xmath219 , for @xmath112 , @xmath220 , @xmath123 . , title="fig:",scaledwidth=50.0% ] flaton phase as a function of the 2d lattice for time slices at @xmath219 , for @xmath111 , @xmath123 .
, title="fig:",scaledwidth=50.0% ] flaton phase as a function of the 2d lattice for time slices at @xmath219 , for @xmath111 , @xmath123 . , title="fig:",scaledwidth=50.0% ] flaton phase as a function of the 2d lattice for time slices at @xmath219 , for @xmath111 , @xmath123 . ,
title="fig:",scaledwidth=50.0% ] flaton phase as a function of the 2d lattice for time slices at @xmath219 , for @xmath111 , @xmath123 .
, title="fig:",scaledwidth=50.0% ] flaton phase as a function of the 2d lattice for time slices at @xmath219 , for @xmath111 , @xmath123 . , title="fig:",scaledwidth=50.0% ] flaton phase as a function of the 2d lattice for time slices at @xmath219 , for @xmath111 , @xmath123 .
, title="fig:",scaledwidth=50.0% ] kinetic energy spectra of flaton and ad fields , at @xmath222 for @xmath123 .
thermally normalized so that a thermal spectrum is flat and energy normalized so that the area under the graph is the total energy from top to bottom .
upper half : @xmath112 , @xmath220 .
lower half : @xmath111 .
, title="fig:",scaledwidth=50.0% ] kinetic energy spectra of flaton and ad fields , at @xmath222 for @xmath123 .
thermally normalized so that a thermal spectrum is flat and energy normalized so that the area under the graph is the total energy from top to bottom .
upper half : @xmath112 , @xmath220 .
lower half : @xmath111 .
, title="fig:",scaledwidth=50.0% ] kinetic energy spectra of flaton and ad fields , at @xmath222 for @xmath123 .
thermally normalized so that a thermal spectrum is flat and energy normalized so that the area under the graph is the total energy from top to bottom .
upper half : @xmath112 , @xmath220 .
lower half : @xmath111 .
, title="fig:",scaledwidth=50.0% ] kinetic energy spectra of flaton and ad fields , at @xmath222 for @xmath123 .
thermally normalized so that a thermal spectrum is flat and energy normalized so that the area under the graph is the total energy from top to bottom .
upper half : @xmath112 , @xmath220 .
lower half : @xmath111 .
, title="fig:",scaledwidth=50.0% ] kinetic energy spectra of flaton and ad fields , at @xmath222 for @xmath123 .
thermally normalized so that a thermal spectrum is flat and energy normalized so that the area under the graph is the total energy from top to bottom .
upper half : @xmath112 , @xmath220 .
lower half : @xmath111 .
, title="fig:",scaledwidth=50.0% ] kinetic energy spectra of flaton and ad fields , at @xmath222 for @xmath123 .
thermally normalized so that a thermal spectrum is flat and energy normalized so that the area under the graph is the total energy from top to bottom .
upper half : @xmath112 , @xmath220 .
lower half : @xmath111 .
, title="fig:",scaledwidth=50.0% ] kinetic energy spectra of flaton and ad fields , at @xmath222 for @xmath123 .
thermally normalized so that a thermal spectrum is flat and energy normalized so that the area under the graph is the total energy from top to bottom .
upper half : @xmath112 , @xmath220 .
lower half : @xmath111 .
, title="fig:",scaledwidth=50.0% ] kinetic energy spectra of flaton and ad fields , at @xmath222 for @xmath123 .
thermally normalized so that a thermal spectrum is flat and energy normalized so that the area under the graph is the total energy from top to bottom .
upper half : @xmath112 , @xmath220 .
lower half : @xmath111 .
, title="fig:",scaledwidth=50.0% ] energy densities of the flaton as a function of time , for @xmath112 , @xmath123 , @xmath220 . from top to bottom at @xmath223 : kinetic , gradient , potential .
, title="fig : " ] energy densities of the flaton as a function of time , for @xmath112 , @xmath123 , @xmath220 . from top to bottom at @xmath223 : kinetic , gradient , potential .
, title="fig : " ] energy densities of the flaton and ad fields as a function of time , for @xmath111 , @xmath123 . from top to bottom at @xmath223 : kinetic , gradient , potential .
, title="fig : " ] energy densities of the flaton and ad fields as a function of time , for @xmath111 , @xmath123 . from top to bottom at @xmath223 : kinetic , gradient , potential .
, title="fig : " ] dependence on lattice with @xmath224 and @xmath225 for @xmath126 , @xmath123 .
upper half : thermally normalized and energy normalized spectra in order at @xmath226 .
lower half : variances , total energies and lepton numbers densities , averaged over the lattice , as a function of time .
the graph of the lepton number densities is including the cases of @xmath227 and @xmath217 as well.,title="fig:",scaledwidth=50.0% ] dependence on lattice with @xmath224 and @xmath225 for @xmath126 , @xmath123 .
upper half : thermally normalized and energy normalized spectra in order at @xmath226 .
lower half : variances , total energies and lepton numbers densities , averaged over the lattice , as a function of time .
the graph of the lepton number densities is including the cases of @xmath227 and @xmath217 as well.,title="fig:",scaledwidth=50.0% ] dependence on lattice with @xmath224 and @xmath225 for @xmath126 , @xmath123 .
upper half : thermally normalized and energy normalized spectra in order at @xmath226 .
lower half : variances , total energies and lepton numbers densities , averaged over the lattice , as a function of time .
the graph of the lepton number densities is including the cases of @xmath227 and @xmath217 as well.,title="fig:",scaledwidth=50.0% ] dependence on lattice with @xmath224 and @xmath225 for @xmath126 , @xmath123 .
upper half : thermally normalized and energy normalized spectra in order at @xmath226 .
lower half : variances , total energies and lepton numbers densities , averaged over the lattice , as a function of time .
the graph of the lepton number densities is including the cases of @xmath227 and @xmath217 as well.,title="fig:",scaledwidth=50.0% ] dependence on lattice with @xmath224 and @xmath225 for @xmath126 , @xmath123 .
upper half : thermally normalized and energy normalized spectra in order at @xmath226 .
lower half : variances , total energies and lepton numbers densities , averaged over the lattice , as a function of time .
the graph of the lepton number densities is including the cases of @xmath227 and @xmath217 as well.,title="fig:",scaledwidth=50.0% ] dependence on lattice with @xmath224 and @xmath225 for @xmath126 , @xmath123 .
upper half : thermally normalized and energy normalized spectra in order at @xmath226 .
lower half : variances , total energies and lepton numbers densities , averaged over the lattice , as a function of time .
the graph of the lepton number densities is including the cases of @xmath227 and @xmath217 as well.,title="fig:",scaledwidth=50.0% ] dependence on lattice with @xmath224 and @xmath225 for @xmath126 , @xmath123 .
upper half : thermally normalized and energy normalized spectra in order at @xmath226 .
lower half : variances , total energies and lepton numbers densities , averaged over the lattice , as a function of time .
the graph of the lepton number densities is including the cases of @xmath227 and @xmath217 as well.,title="fig:",scaledwidth=50.0% ] dependence on lattice with @xmath224 and @xmath225 for @xmath126 , @xmath123 .
upper half : thermally normalized and energy normalized spectra in order at @xmath226 .
lower half : variances , total energies and lepton numbers densities , averaged over the lattice , as a function of time .
the graph of the lepton number densities is including the cases of @xmath227 and @xmath217 as well.,title="fig:",scaledwidth=50.0% ]
|
previously , we proposed a model of low energy affleck - dine leptogenesis in the context of thermal inflation . the lepton asymmetry is generated at the end of thermal inflation , which occurs at a relatively low energy scale with the hubble parameter somewhere in the range @xmath0 .
thus hubble damping will be ineffective in bringing the affleck - dine field into the lepton conserving region near the origin , leaving the possibility that the lepton number could be washed out .
previously , we suggested that preheating could damp the amplitude of the affleck - dine field allowing conservation of the lepton number . in this paper , we demonstrate numerically that preheating does efficiently damp the amplitude of the affleck - dine field and that the lepton number is conserved as the result .
in addition to demonstrating a crucial aspect of our model , it also opens the more general possibility of low energy affleck - dine baryogenesis .
|
dwarf spheroidal ( dsph ) galaxies are believed to play an important role in the formation and evolution of much more luminous galaxies ( e.g. * ? ? ?
dsphs are characterized by their low surface brightness , low total luminosity and spheroidal shapes which are consistent with their pressure - supported stellar kinematics ( e.g. * ? ? ?
knowledge of dsph masses is essential for comparison with cosmological simulations of galaxy formation .
good mass estimates help us to establish whether the paucity in the number of observed systems ( a few tens ) to the number of predicted satellite haloes ( several thousands ) represents a fundamental failure of our cosmological model ( e.g. * ? ? ?
* ) , or whether it is simply telling us that galaxy formation is inefficient on small scales ( e.g. * ? ? ?
recent studies have suggested that the dsph galaxies share a common mass within a certain radius @xcite .
if confirmed , this would be an important clue to the processes which regulate the formation of the lowest luminosity galaxies .
all mass estimates implicitly assume that contamination of the kinematic sample by foreground stars or unbound tidal tails is negligible , and that the system is in ( or close to ) virial equilibrium .
since the mass of an equilibrium stellar system is proportional to its velocity dispersion squared , an over - estimate of the velocity dispersion will result in an inflated mass estimate .
the assumption of virial equilibrium may be called into question if the system is tidally disrupting @xcite .
the recently discovered hercules dsph @xcite , with an ellipticity of @xmath12 @xcite , is an example of a galaxy in which assumptions of equilibrium may be incorrect . in @xcite
we showed that the mean velocity of the hercules dsph is embedded in the foreground dwarf star velocity distribution . in order to weed out the foreground dwarf stars we used the strmgren @xmath13 index .
this index is able to clearly identify red - giant branch ( rgb ) stars redder than the horizontal branch , enabling a separation of rgb stars in the dsph galaxy and foreground dwarf stars . by weeding out the foreground contaminants we found that the dispersion for hercules is reduced from @xmath14 to @xmath15 . in this letter
we explore the consequences of this finding . in section 2 ,
we derive a new mass for hercules and show that it is not consistent with a common mass scale for the dsphs . in section 3
, we investigate the possibility of a velocity gradient in the kinematic data of hercules . in section 4
, we discuss the relative importance of tides and dark matter in hercules .
section 5 summarises our conclusions .
we use the velocity dispersion , @xmath16 @xcite , to constrain the mass of the hercules dsph .
we assume that the system is in dynamical equilibrium , is spherically symmetric , has an isotropic velocity distribution , and a flat velocity dispersion profile . with these assumptions , the jeans equation for the mass distribution ( eq .
( 4.215 ) in * ? ? ?
* ) becomes : @xmath17 where @xmath18 is the three - dimensional distance from the centre of the galaxy , @xmath19 is the de - projected stellar density profile and @xmath20 is the enclosed mass .
we use the exponential profile from @xcite to describe the stellar density profile .
the de - projected exponential profile is given by @xcite as ( setting @xmath21 in their eq .
( 4 ) ) : @xmath22 where @xmath23 is the exponential scale radius , related to the half - light radius as @xmath24 , and @xmath25 is the central stellar density . solving eq .
( [ eq1 ] ) for @xmath20 using eq .
( [ eq2 ] ) yields @xmath26 using a half - light radius of @xmath27pc @xcite we solve eq .
( [ eq3 ] ) for @xmath28 pc , which corresponds to the outermost member in our kinematic sample ( table [ tab ] ) .
we estimate the error in the mass using @xmath29 monte carlo realisations of the half - light radius , distance and velocity dispersion of hercules drawn from within the individual error bars on each parameter . in this way
, we obtain @xmath30 ( fig .
[ fig1 ] ) .
the quoted errors are 1-@xmath31 limits from the monte carlo sampling . assuming a total luminosity of @xmath32 @xcite
, we find a median mass - to - light ratio of @xmath33 $ ] .
@xcite emphasise the importance of weeding out foreground , contaminating dwarf stars .
this is particularly vital for the hercules dsph as its systemic velocity coincides with the bulk motion of dwarf stars in the milky way disk in the direction of hercules . using the contaminated velocity outliers
also , the initial candidate selection for the velocities were chosen using colour - magnitude diagram cuts .
even so , a significant fraction of foreground stars remain in this sample and are detected only through the use of the strmgren photometry .
] velocity dispersion ( table [ tab ] ) we would obtain a mass of @xmath34 , almost a factor of four larger than the uncontaminated estimate .
@xcite speculate that the mass within a fixed radius , 300pc , is approximately the same ( @xmath35 ) in all the observed dsphs .
our revised velocity dispersion implies a mass within the inner 300 pc ( m@xmath36 ) of only @xmath37 .
this indicates that hercules falls considerably below this `` common mass scale '' for dsphs .
since 300 pc is also approximately the half - light radius , this implies that hercules also lies significantly off the @xcite enclosed half - light mass scaling relation . to confirm that our result is not sensitive to our choice of surface brightness profile , we have repeated our calculation using a plummer profile with scale radius @xmath38 ( as used by *
in this case we find m@xmath39 in agreement with the mass calculated above .
if instead we use our contaminated velocity dispersion and the exponential surface density profile , we obtain m@xmath40 which agrees with the value presented in @xcite
. however , we note that @xcite used a dispersion of @xmath41 ( taken from * ? ? ? * ) to obtain their mass estimate .
this is smaller than our contaminated value of @xmath42 , yet they obtain a similar median mass to us . if we use a dispersion of @xmath43 as they do , we obtain a mass @xmath44m@xmath3 .
this is just consistent with their determination within our mutual error bars .
we note that the median @xmath45 will depend on the details of the mass modelling procedure and so can be expected to differ between their study and ours ( matt walker priv .
comm . ) .
additionally , we have repeated our calculation of the mass within 300pc using a more recent estimate of the half - light radius , @xmath46pc , by @xcite .
this half - light radius is smaller and gives a mass that is @xmath47 times more massive than the mass calculated using the half - light radius from @xcite .
the presence of a velocity gradient in a dsph could either be indicative of an intrinsic rotation , or a sign of tidal interaction with the milky way . in this section ,
we test for possible velocity gradients in hercules . assuming that the rotation around the semi - minor axis ( in the ellipse that desribes the orientation of the dsph ) is more likely then around the semi - major axis we can define the `` semi - minor axis distance '' @xmath48 as the distance , perpendicular to the semi - minor axis , between the axis and the star . for each of the 18 rgb stars with radial velocities we calculate this distance for different position angles @xmath49 .
we then derive the gradient , @xmath50 , and zeropoint , @xmath51 , by solving the following equation with a least - squares fit for each position angle , @xmath52 where @xmath53 is the radial velocity measurement for each stars .
the least - squares fit to this function yields a @xmath54 value for each @xmath49 ( fig . [
fig2]a ) .
the distribution enclosed by @xmath55 corresponds to @xmath56 for a normal distribution @xcite .
we use this to obtain the error in the position angle which minimizes @xmath54 .
we find a position angle of @xmath57 degrees with a velocity gradient of @xmath5kms@xmath6kpc@xmath6 , and a zeropoint of @xmath58 .
we obtain a reduced @xmath54 of @xmath59 for our 18 stars with 2 degrees of freedom .
following @xcite , we estimate the significance of the velocity gradient using @xmath29 monte carlo realisations . in each realisation
, we sample the velocity and spatial distributions independently .
thus we scramble the correlation between velocity and position , while maintaining the original velocity distribution and spatial positions .
this is valid as long as the phase space distribution function of the stars is separable ( which implies that the velocity dispersion is independent of radius ) . we determine the significance of the velocity gradient by computing the fraction of realisations that fail to produce a @xmath54 as low as the one calculated from the real data .
we find a significance of the velocity gradient of 78 per cent ( @xmath60 ) .
in the previous section we found tentative evidence for a velocity gradient in the hercules dsph .
the peak - to - peak difference of @xmath61 within a radius of less than 1kpc could be interpreted as the effect of galactic tides @xcite .
interestingly , @xcite recently estimated the orbit of hercules based on the assumption that its elongation is tidally induced , and predicted a velocity gradient of at least @xmath62 which is consistent with our observed gradient .
additionally , in @xcite we found that the spatial distribution of the hercules stars is asymmetric at @xmath630.5 kpc .
there are three significant outlier stars to the south , but no corresponding members at this distance in either the northern or western fields ( fig .
[ fig3 ] ) .
these three stars are unambiguously identified as rgb stars from strmgren photometry .
we now consider the possibility that the velocity gradient and the positional outliers are evidence that hercules is being tidally disrupted , and use this information to obtain a second mass estimate for the system .
the tidal radius of a dsph depends on the potential of the host galaxy , the potential of the dsph , the orbit of the dsph within the host galaxy and the orbit of the stars within the dsph ( e.g. * ? ? ?
we parameterise the milky way potential using the default model in @xcite , analysed in the galactic plane ; and the hercules potential using a generalised hernquist profile @xcite : @xmath64 where @xmath65 is the total mass , @xmath66 the scalelength , and @xmath67 the central logarithmic cusp slope .
we consider the ranges @xmath68kpc and @xmath69 .
our results are not sensitive to these choices .
the orbit of the hercules dsph is unknown but it currently lies at a distance of @xmath70kpc from the galactic centre , and has a heliocentric velocity of @xmath71 , which implies a galactocentric radial velocity of @xmath72 ( see eq .
( 5 ) in * ? ? ?
using the above potential model for the milky way and setting the tangential velocity component for hercules to zero , this gives us a minimum apocentre for hercules of @xmath73kpc .
thus , we consider apocentre and pericentre ranges of @xmath74kpc and @xmath75kpc , respectively .
@xcite find that photometric features are typically seen beyond the retrograde tidal radius .
to proceed , we make the assumption that the outliers in fig .
[ fig3 ] indicate the location of the retrograde tidal radius of hercules , i.e. @xmath76pc . if we assume further that the tidal radius of hercules is set at the pericentre of its orbit , we can solve eq .
( 7 ) of @xcite to calculate the mass , @xmath77 , of hercules as a function of its orbital pericentre @xmath78 , for our assumed ranges of @xmath66 , @xmath67 and @xmath79 .
we calculate both its mass within @xmath10 ( @xmath80 ) and its mass within 433pc ( @xmath81 ) which can then be compared with our mass estimate based on the spherical jeans equation ( @xmath82 ) .
the results of this calculation are given in fig .
the horizontal light grey band marks @xmath82 , the dark grey band marks @xmath81 and the medium grey band marks @xmath80 .
the vertical solid and dashed lines mark the pericentres at which @xmath83 .
the width of the tidal mass bands is due to the unknowns : @xmath66 , @xmath67 , and @xmath79 .
however , to lowest order @xmath10 depends only on the mean density enclosed within it , and so these bands are narrow . for this reason
, we obtain an estimate of both the orbital pericentre of hercules ( @xmath84kpc ) and its mass within the tidal radius ( @xmath85m@xmath3 ) .
the primary source of error on both of these quantities is our assumed tidal radius @xmath10 .
empirically we derive scalings of : @xmath86 with @xmath87 over the range @xmath88$]kpc , and : @xmath89 with @xmath90 over the same range .
if we assume @xmath91 pc , we obtain a similar estimate of the pericentric distance as that obtained by @xcite .
we have shown that our hercules data are consistent with the presence of dark matter .
we now consider whether models without dark matter could also reproduce the data .
the most extreme scenario is that hercules is disintegrating and its velocity dispersion arises solely due to the motion of its unbound member stars .
if so , it will rapidly become unobservable ( i.e. reach a lower surface brightness than the detection limit of the sdss ) .
hercules has a surface brightness of @xmath92 mag arcsec@xmath93 @xcite .
if the unbound stars are moving away from hercules at a velocity equal to the velocity dispersion ( @xmath15 ) it would require only @xmath94 years for the hercules dsph to fall below the detection limit of sdss ( @xmath95 * ? ? ?
given this short timescale , it is very unlikely that we would observe hercules at this phase of its evolution . @xcite simulated the disruption of the uma ii dsph . using a model that does not distinguish between luminous and dark matter they simulate the surface brightness , velocity dispersion and the mean radial velocity for uma ii after 9 , 10 and 11gyr ( their fig .
9 ) . in the absence of dark matter ,
the equilibrium velocity dispersion ( i.e. @xmath96 ) for the hercules dsph would be @xmath97 , assuming a stellar mass of @xmath98 and a radius equal to the observed half - light radius ( @xmath99pc ; * ? ? ? * ) .
this is much lower than the measured velocity dispersion of @xmath100 .
we conclude that the hercules dsph can only be a dark matter free system if its velocity dispersion has been inflated significantly - i.e. if it is in the advanced stages of tidal disruption .
however , in the simulations of @xcite the evolutionary phase in which a tidal remnant exhibits both an inflated velocity dispersion and a velocity gradient , and remains centrally concentrated is very short . within @xmath101gyr
the system goes from bound , with almost no signs of tidal disturbance , to complete disruption .
( we note that @xcite performed simulations of dsph galaxies without dark matter , and found that it is possible to obtain long - lived remnants whose properties are remarkably similar to those of hercules in terms of velocity dispersion and velocity gradient ( see his fig .
however , the remnants simulated by @xcite are significantly more luminous than the hercules dsph . based on the simulations of @xcite
, it seems unlikely that lower luminosity , purely stellar remnants with the correct properties could survive for a significant time . )
it seems difficult to understand hercules without dark matter .
however , it could be dark matter dominated _ and _ experiencing significant tidal disturbance . in this case
, our simple jeans analysis may have over - estimated its mass and hercules might lie even further away from a possible common mass scale for the dsph galaxies .
we have calculated the mass of the hercules dsph using the new velocity dispersion for the system obtained by @xcite .
we find that the mass within the volume enclosed by our observed stars is @xmath102 , leading to a mass - to - light ratio of @xmath103 $ ] .
interestingly , the mass within 300pc is significantly lower than the `` common mass scale '' found by @xcite , suggesting that hercules does not share the halo properties seen in other dsphs .
we found tentative evidence for a velocity gradient of @xmath104kms@xmath6kpc@xmath6 , and evidence of an asymmetric extension in the light distribution at @xmath105kpc .
we explored the hypothesis that these features are due to tidal interactions with the milky way . assuming a tidal radius of @xmath106pc
, we show that a self - consistent model requires hercules to be on an orbit with pericentre @xmath107kpc , and with a mass within @xmath10 of @xmath108m@xmath3 .
we are very grateful to matt walker for useful discussions regarding the mass estimates in earlier works .
we thank pavel kroupa for stimulating discussions . d.a
thanks lennart lindegren at lund observatory for help with statistics . s.f . is a royal swedish academy of sciences research fellow supported by a grant from the knut and alice wallenberg foundation .
m.i.w is supported by a royal society university research fellowship . , j. & tremaine , s. 2008 , galactic dynamics : second edition ( galactic dynamics : second edition , by james binney and scott tremaine .
isbn 978 - 0 - 691 - 13026 - 2 ( hb )
. published by princeton university press , princeton , nj usa , 2008 . )
l l c c c c c c @xcite & 28 [ rgb stars , @xmath13 sel . ] & - & - & - & - & @xmath109 + this study & 32 [ only @xmath110 sel . ] & @xmath111 & @xmath112 & @xmath113 & @xmath114 & - + this study & 18 [ rgb stars with @xmath110 ] & @xmath116 & @xmath117 & @xmath118 & @xmath119 & - +
|
we present a new mass estimate for the hercules dwarf spheroidal galaxy ( dsph ) , based on the revised velocity dispersion obtained by @xcite .
the removal of a significant foreground contamination using newly acquired strmgren photometry has resulted in a reduced velocity dispersion . using this new velocity dispersion of @xmath0
, we find a mass of @xmath1 within the central 300pc , which is also the half - light radius , and a mass of @xmath2m@xmath3 within the reach of our data to 433pc , significantly lower than previous estimates .
we derive an overall mass - to - light ratio of @xmath4 $ ] .
our mass estimate calls into question recent claims of a common mass scale for dsph galaxies .
additionally , we find tentative evidence for a velocity gradient in our kinematic data of @xmath5kms@xmath6kpc@xmath6 , and evidence of an asymmetric extension in the light distribution at @xmath7kpc .
we explore the possibility that these features are due to tidal interactions with the milky way .
we show that there is a self - consistent model in which hercules has an assumed tidal radius of @xmath8pc , an orbital pericentre of @xmath9kpc , and a mass within @xmath10 of @xmath11m@xmath3 .
proper motions are required to test this model .
although we can not exclude models in which hercules contains no dark matter , we argue that hercules is more likely to be a dark matter dominated system which is currently experiencing some tidal disturbance of its outer parts .
|
high energy cosmic ray ( cr ) particles are shielded by earth s atmosphere and reveal their existence on the ground only by indirect effects such as ionization and showers of secondary charged particles covering areas up to many km@xmath1 for the highest energy particles .
in fact , in 1912 victor hess discovered crs by measuring ionization from a balloon @xcite , and in 1938 pierre auger proved the existence of extensive air showers ( eas ) caused by primary particles with energies above @xmath2ev by simultaneously observing the arrival of secondary particles in geiger counters many meters apart @xcite .
after almost 90 years of research , the origin of cosmic rays is still an open question , with a degree of uncertainty increasing with energy @xcite : only below 100 mev kinetic energy , where the solar wind shields protons coming from outside the solar system , the sun must give rise to the observed proton flux .
above that energy the cr spectrum exhibits little structure and is approximated by broken power laws @xmath3 : at the energy @xmath4ev called the `` knee '' , the flux of particles per area , time , solid angle , and energy steepens from a power law index @xmath5 to one of index @xmath6 .
the bulk of the crs up to at least that energy is believed to originate within the milky way galaxy , typically by shock acceleration in supernova remnants .
the spectrum continues with a further steepening to @xmath7 at @xmath8ev , sometimes called the `` second knee '' .
there are experimental indications that the chemical composition changes from light , mostly protons , at the knee to domination by iron and even heavier nuclei at the second knee @xcite .
this is in fact expected in any scenario where acceleration and propagation is due to magnetic fields whose effects only depend on rigidity , the ratio of charge to rest mass , @xmath9 .
this is true as long as energy losses and interaction effects , which in general depend on @xmath10 and @xmath11 separately , are small , as is the case in the galaxy , in contrast to extra - galactic cosmic ray propagation at ultra - high energy . above the so called `` ankle '' or `` dip '' at @xmath12ev , the spectrum flattens again to a power law of index @xmath13 .
this latter feature is often interpreted as a cross over from a steeper galactic component , which above the ankle can not be confined by the galactic magnetic field , to a harder component of extragalactic origin .
the dip at @xmath12ev could also be partially due to pair production by extra - galactic protons , especially if the extra - galactic component already starts to dominate below the ankle , for example , around the second - knee @xcite .
this latter possibility appears , however , less likely in light of a rather heavy composition up to the ankle suggested by several experiments @xcite . in any case , an eventual cross over to an extra - galactic component is also in line with experimental indications for a chemical composition becoming again lighter above the ankle , although a significant heavy component is not excluded and the inferred chemical composition above @xmath14ev is sensitive to the model of air shower interactions and consequently uncertain presently @xcite . in the following
we will restrict our discussion on ultra - high energy cosmic rays ( uhecrs ) above the ankle .
although statistically meaningful information about the uhecr energy spectrum and arrival direction distribution has been accumulated , no conclusive picture for the nature and distribution of the sources emerges naturally from the data .
there is on the one hand the approximate isotropic arrival direction distribution @xcite which indicates that we are observing a large number of weak or distant sources . on the other hand ,
there are also indications which point more towards a small number of local and therefore bright sources , especially at the highest energies : first , the agasa ground array claims statistically significant multi - plets of events from the same directions within a few degrees @xcite , although this is controversial @xcite and has not been seen so far by the fluorescence experiment hires @xcite .
the spectrum of this clustered component is @xmath15 and thus much harder than the total spectrum @xcite .
second , nucleons above @xmath16eev suffer heavy energy losses due to photo - pion production on the cosmic microwave background the greisen - zatsepin - kuzmin ( gzk ) effect @xcite which limits the distance to possible sources to less than @xmath17mpc @xcite .
for a uniform source distribution this would predict a `` gzk cutoff '' , a drop in the spectrum .
however , the existence of this `` cutoff '' is not established yet from the observations @xcite and may even depend on the part of the sky one is looking at : the cutoff could be mitigated in the northern hemisphere where more nearby accelerators related to the local supercluster can be expected . apart from the sugar array which was active from 1968 until 1979 in australia , all uhecr detectors completed up to the present were situated in the northern hemisphere .
nevertheless the situation is unclear even there : whereas a cut - off seems consistent with the few events above @xmath18ev recorded by the fluorescence detector hires @xcite , it is not compatible with the 8 events above @xmath18ev measured by the agasa ground array @xcite .
it can be remarked , however , that analysis of data based on a single fluorescence telescope , the so - called monocular mode in which most of the hires data were obtained , is complicated due to atmospheric conditions varying from event to event @xcite .
the solution of this problem may have to await more analysis and , in particular , the completion of the pierre auger project @xcite which will combine the two complementary detection techniques adopted by the aforementioned experiments and whose southern site is currently in construction in argentina .
this currently unclear experimental situation could easily be solved if it would be possible to follow the uhecr trajectories backwards to their sources . however , this may be complicated by the possible presence of extragalactic magnetic fields , which would deflect the particles during their travel .
furthermore , since the gzk - energy losses are of stochastic nature , even a detailed knowledge of the extragalactic magnetic fields would not necessarily allow to follow a uhecr trajectory backwards to its source since the energy and therefor the larmor radius of the particles have changed in an unknown way .
therefore it is not clear if charged particle astronomy with uhecrs is possible in principle or not . and even if possible , it remains unclear to which degree the angular resolution would be limited by magnetic deflection .
this topic will be discussed in sect .
the physics and astrophysics of uhecrs are also intimately linked with the emerging field of neutrino astronomy ( for reviews see refs .
@xcite ) as well as with the already established field of @xmath0ray astronomy ( for reviews see , e.g. , ref . @xcite ) .
indeed , all scenarios of uhecr origin , including the top - down models , are severely constrained by neutrino and @xmath0ray observations and limits . in turn , this linkage has important consequences for theoretical predictions of fluxes of extragalactic neutrinos above about a tev whose detection is a major goal of next - generation neutrino telescopes : if these neutrinos are produced as secondaries of protons accelerated in astrophysical sources and if these protons are not absorbed in the sources , but rather contribute to the uhecr flux observed , then the energy content in the neutrino flux can not be higher than the one in uhecrs , leading to the so called waxman - bahcall bound for transparent sources with soft acceleration spectra @xcite .
if one of these assumptions does not apply , such as for acceleration sources with injection spectra harder than @xmath19 and/or opaque to nucleons , or in the top - down scenarios where x particle decays produce much fewer nucleons than @xmath0rays and neutrinos , the waxman - bahcall bound does not apply , but the neutrino flux is still constrained by the observed diffuse @xmath0ray flux in the gev range .
this will be discussed in the following section .
electromagnetic ( em ) energy injected above the threshold for pair production on the cosmic microwave background ( cmb ) at @xmath20ev at redshift @xmath21 ( to a lesser extent also on the infrared / optical background , with lower threshold ) leads to an em cascade , an interplay between pair production followed by inverse compton scattering of the produced electrons .
this cascade continues until the photons fall below the pair production threshold at which point the universe becomes transparent for them . in todays universe
this happens within just a few mpc for injection up to the highest energies above @xmath18ev .
all em energy injected above @xmath22ev and at distances beyond a few mpc today is therefore recycled to lower energies where it gives rise to a characteristic cascade spectrum @xmath23 down to fractions of a gev @xcite .
the universe thus acts as a calorimeter where the total em energy injected above @xmath20ev is measured as a diffuse isotropic @xmath0ray flux in the gev regime .
this diffuse flux is not very sensitive to the somewhat uncertain infrared / optical background @xcite .
any observed diffuse @xmath0ray background acts as an upper limit on the total em injection . since in any scenario involving pion production the em energy fluence is comparable to the neutrino energy fluence , the constraint on em energy injection also constrains allowed neutrino fluxes .
this diffuse extragalactic gev @xmath0ray background can be extracted from the total @xmath0ray flux measured by egret by subtracting the galactic contribution . since publication of the original egret limit in 1995 @xcite , models for this high latitude galactic @xmath0ray foreground were improved significantly .
this allowed the authors of ref .
@xcite to reanalyze limits on the diffuse extragalactic background in the region 30 mev-10 gev and to lower it by a factor 1.5 - 1.8 in the region around 1 gev .
there are even lower estimates of the extragalactic diffuse @xmath0ray flux @xcite . in this article , however , we will use the more conservative limits from ref.@xcite .
the energy in the extra - galactic @xmath0ray background estimated in ref .
@xcite is slightly more than one hundred times the energy in uhecr above the gzk cutoff .
the range of such trans - gzk cosmic rays is about @xmath24mpc , roughly one hundredth the hubble radius , and only sources within that gzk range contribute to the trans - gzk cosmic rays .
therefore , any mechanism involving sources distributed roughly uniformly on scales of the gzk energy loss length @xmath24mpc and producing a comparable amount of energy in trans - gzk cosmic rays and photons above the pair production threshold can potentially explain this energy flux ratio .
the details depend on the exact redshift dependence of source activity and other parameters and in general have to be verified by numerically solving the relevant transport equations , see , e.g. , ref .
such mechanisms include shock acceleration in powerful objects such as active galactic nuclei @xcite .
on the other hand , any mechanism producing considerably _ more _ energy in the em channel above the pair production threshold than in trans - gzk cosmic rays tend to predict a ratio of the diffuse gev @xmath0ray flux to the trans - gzk cosmic ray flux too high to explain both fluxes at the same time . as a consequence , if normalized at or below the observational gev @xmath0ray background , such scenarios tend to explain at most a fraction of the observed trans - gzk cosmic ray flux .
such scenarios include particle physics mechanisms involving pion production by quark fragmentation , e.g. extra - galactic top - down mechanisms where uhecrs are produced by fragmenting quarks resulting from decay of superheavy relics @xcite .
most of these quarks would fragment into pions rather than nucleons such that more @xmath0rays ( and neutrinos ) than cosmic rays are produced .
overproduction of gev @xmath0rays can be avoided by assuming the sources in an extended galactic halo with a high @xmath25 overdensity compared to the average cosmological source density , which would also avoid the gzk cutoff @xcite .
these scenarios , however , start to be constrained by the anisotropy they predict because of the asymmetric position of the sun in the galactic halo for which there are no indications in present data @xcite .
scenarios based on quark fragmentation also become problematic in view of a possible heavy nucleus component and of upper limits on the photon fraction of the uhecr flux @xcite . as a specific example for scenarios involving quark fragmentation , we consider here the case of decaying z - bosons . in this `` z - burst mechanism '' z - bosons
are produced by uhe neutrinos interacting with the relic neutrino background @xcite .
if the relic neutrinos have a mass @xmath26 , z - bosons can be resonantly produced by uhe neutrinos of energy @xmath27 .
the required neutrino beams could be produced as secondaries of protons accelerated in high - redshift sources .
the fluxes predicted in these scenarios have recently been discussed in detail , for example , in refs .
@xcite . in fig .
[ fig1 ] we show an optimistic example taken from ref .
it is assumed that the relic neutrino background has no significant local overdensity .
furthermore , the sources are assumed to not emit any @xmath0rays , otherwise the z - burst model with acceleration sources overproduces the diffuse gev @xmath0ray background @xcite .
we note that no known astrophysical accelerator exists that meets the requirements of the z - burst model @xcite . up to @xmath28ev and for redshifts between 0 and 3 .
the sources are assumed to be exclusive neutrino emitters .
all neutrino masses were assumed equal with @xmath29 ev and we again assumed maximal mixing between all flavors . also shown are predicted and observed cosmic ray and @xmath0ray fluxes , the atmospheric neutrino flux @xcite , as well as existing upper limits on the diffuse neutrino fluxes from macro @xcite , amanda ii @xcite , baikal @xcite , agasa @xcite , the fly s eye @xcite and rice @xcite experiments , and the limits obtained with the goldstone radio telescope ( glue ) @xcite and the forte satellite @xcite , as indicated .
the cosmic ray data are from the agasa @xcite and hires @xcite experiments , and the new egret estimate of the extra - galactic diffuse @xmath0ray flux is shown to the left . from ref .
[ fig1 ] ] however , a combination of new constraints allows to rule out that the z - burst mechanism explains a dominant fraction of the observed uhecr flux , even for pure neutrino emitting sources : a combination of cosmological data including the wmap experiment limit the sum of the masses of active neutrinos to @xmath30ev @xcite .
solar and atmospheric neutrino oscillations indicate that individual neutrino masses are nearly degenerate on this scale @xcite , and thus the neutrino mass per flavor must satisfy @xmath31ev .
however , for such masses phase space constraints limit the possible over - density of neutrinos in our local group of galaxies to @xmath32 on a length scale of @xmath33mpc @xcite .
since this is considerably smaller than the relevant uhecr loss lengths , neutrino clustering will not significantly reduce the necessary uhe neutrino flux compared to the case of no clustering . for the maximal possible value of the neutrino mass @xmath34ev , the neutrino flux required for the z - burst mechanism to explain the uhecr flux is only in marginal conflict with the forte upper limit @xcite , and factor 2 higher than the new glue limit @xcite , as shown in fig .
[ fig1 ] . for all other cases
the conflict with both the glue and forte limits is considerably more severe .
also note that this argument does not depend on the shape of the low energy tail of the primary neutrino spectrum which could thus be even mono - energetic , as could occur in exclusive tree level decays of superheavy particles into neutrinos @xcite .
however , in addition this possibility has been ruled out by overproduction of gev @xmath0rays due to loop effects in these particle decays @xcite .
the possibility that the observed uhecr flux is explained by the z burst scenario involving normal astrophysical sources which produce both neutrinos and photons by pion production is already ruled out by the former egret limit : in this case the gev @xmath0ray flux level would have roughly the height of the peak of the neutrino flux multiplied with the squared energy in fig .
[ fig1 ] , thus a factor @xmath35 higher than the egret level . any further reduction in the estimated contribution of the true diffuse extra - galactic @xmath0ray background to the observed flux , therefore , leads to more severe constraints on the total em injection .
for example , future @xmath0ray detectors such as glast @xcite will test whether the diffuse extragalactic gev @xmath0ray background is truly diffuse or partly consists of discrete sources that could not be resolved by egret .
astrophysical discrete contributions such as from intergalactic shocks are in fact expected @xcite .
this could further improve the cascade limit to the point where even acceleration scenarios may become seriously constrained .
cosmic magnetic fields are inextricably linked with cosmic rays in several respects .
first , they play a central role in fermi shock acceleration .
second , large scale extra - galactic magnetic fields ( egmf ) can cause significant deflection of charged cosmic rays during propagation and thus obviously complicate the relation between observed uhecr distributions and their sources .
magnetic fields are omnipresent in the universe , but their true origin is still unclear @xcite . magnetic fields in galaxies are observed with typical strengths of a few micro gauss , but there are also some indications for fields correlated with larger structures such as galaxy clusters @xcite .
magnetic fields as strong as @xmath36 in sheets and filaments of the large scale galaxy distribution , such as in our local supercluster , are compatible with existing upper limits on faraday rotation @xcite .
it is also possible that fossil cocoons of former radio galaxies , so called radio ghosts , contribute significantly to the isotropization of uhecr arrival directions @xcite . to get an impression of typical deflection angles one can characterize the egmf by its r.m.s .
strength @xmath37 and a coherence length @xmath38 .
if we neglect energy loss processes for the moment , then the r.m.s .
deflection angle over a distance @xmath39 in such a field is @xmath40 @xcite , where the larmor radius of a particle of charge @xmath41 and energy @xmath42 is @xmath43 . in numbers
this reads @xmath44 for @xmath39 .
this expression makes it immediately obvious that fields of fractions of micro gauss lead to strong deflection even at the highest energies .
this goes along with a time delay @xmath45yr which can be millions of years .
a source visible in uhecrs today could therefore be optically invisible since many models involving , for example , active galaxies as uhecr accelerators , predict variability on shorter time scales .
quite a few simulations of the effect of extragalactic magnetic fields ( egmf ) on uhecrs exist in the literature , but usually idealizing assumptions concerning properties and distributions of sources or egmf or both are made : in refs .
@xcite sources and egmf follow a pancake profile mimicking the local supergalactic plane . in other studies
egmf have been approximated in a number of fashions : as negligible @xcite , as stochastic with uniform statistical properties @xcite , or as organized in spatial cells with a given coherence length and a strength depending as a power law on the local density @xcite .
only recently attempts have been made to simulate uhecr propagation in a realistically structured universe @xcite . for now , these simulations are limited to nucleons . in ref .
@xcite the magnetized extragalactic environment used for uhecr propagation is produced by a simulation of the large scale structure of the universe .
the simulation was carried out within a computational box of @xmath46mpc length on a side , with normalized hubble constant @xmath47 km s@xmath48 mpc@xmath49 = 0.67 , and using a comoving grid of 512@xmath50 zones and 256@xmath50 dark matter particles .
the egmf was initialized to zero at simulation start and subsequently its seeds were generated at cosmic shocks through the biermann battery mechanism @xcite . since cosmic shocks form primarily around collapsing structures including filaments , the above approach avoids generating egmf in cosmic voids . in ref .
@xcite constrained simulations of the local large scale structure were performed and the magnetic smoothed particle hydrodynamics technique was used to follow egmf evolution .
the egmf was seeded by a uniform seed field of maximal strength compatible with observed rotation measures in galaxy clusters .
the questions considered in these two works were somewhat different , however . in ref .
@xcite deflections of uhecr above @xmath51ev were computed as a function of the direction to their source which were assumed to be at cosmological distances .
this made sense , because ( i ) the constrained simulations gives a viable model of our local cosmic neighborhood within about 100 mpc , at least on scales beyond a few mpc and ( ii ) the deflections typically were found to be smaller than a few degrees .
concrete source distributions were not considered .
g. note that both panels correspond to the same cuts through the full large scale simulation box used.,title="fig:",scaledwidth=90.0% ] g. note that both panels correspond to the same cuts through the full large scale simulation box used.,title="fig:",scaledwidth=90.0% ] in contrast , ref .
@xcite was not concerned with concrete sky distributions or deflection maps because the simulation was unconstrained and thus only gave a typical large scale structure model and not our concrete local neighborhood .
instead , the question was asked which observer positions and source distributions and characteristics lead to uhecr distributions whose spherical multi - poles for @xmath52 and auto - correlation at angles @xmath53 are consistent with observations . as a result it was found that ( i ) the observed large scale uhecr isotropy requires the neighborhood within a few mpc of the observer is characterized by weak magnetic fields below @xmath54 g , and ( ii ) once that choice is made , current data do not strongly discriminate between uniform and structured source distributions and between negligible and considerable deflection .
nevertheless , current data moderately favor a scenario in which ( iii ) uhecr sources have a density @xmath55 and follow the matter distribution and ( iv ) magnetic fields are relatively pervasive within the large scale structure , including filaments , and with a strength of order of a @xmath56 g in galaxy clusters . a two - dimensional cut through the baryonic density and egmf environment of the observer in a typical such scenario
is shown in fig .
[ fig2 ] .
it was also studied in ref .
@xcite how future data of considerably increased statistics can be used to learn more about egmf and source characteristics .
in particular , low auto - correlations at degree scales imply magnetized sources quite independent of other source characteristics such as their density .
the latter can only be estimated from the auto - correlations halfway reliably if magnetic fields have negligible impact on propagation .
this is because if sources are immersed in considerable magnetic fields , their images are smeared out , which also smears out the auto - correlation function over several degrees . for a sufficiently high source density ,
individual images can thus overlap and sensitivity to source density is consequently lost .
the statistics expected from next generation experiments such as pierre auger @xcite and euso @xcite should be sufficient to test source magnetization by the auto - correlation function @xcite . with respect to the line of sight to the sources .
this is for a scenario from ref .
@xcite in good agreement with uhecr data , where the sources follow the baryon density and have average density @xmath57 , and the egmf included in the large scale structure simulation reaches several micro gauss in the most prominent galaxy cluster .
shown are the average ( middle , histogram ) and 1-@xmath58 variations ( upper and lower curves ) above @xmath51ev , over 24 realizations varying in the positions and luminosities @xmath59 of individual sources , the latter assumed to be distributed as @xmath60 with @xmath61 in arbitrary units . also given on top of the figure are average and variances of the distributions.,scaledwidth=98.0% ] interestingly , however , there is a considerable quantifiable difference in the typical deflection angles predicted by the two egmf scenarios in refs .
@xcite that can _ not _ be compensated by specific source distributions : even for homogeneous source distributions , the average deflection angle for uhecrs above @xmath51ev obtained in ref .
@xcite is much larger than in ref .
@xcite , as can be seen in fig .
in fact , even if the magnetic field strength is reduced by a factor 10 in the simulations of ref .
@xcite , the average deflection angle above @xmath51ev is still @xmath62 , only a factor @xmath63 smaller
. this non - linear behavior of deflection with field normalization is mostly due to the strongly non - homogeneous character of the egmf .
most of these differences are probably due the different numerical models for the magnetic fields .
although ref .
@xcite start with uniform seed fields , whereas in ref .
@xcite seed fields are injected at shocks , by itself , this difference should not influence the resulting egmf very much at late times , at least inside galaxy clusters @xcite .
it should be noted , however , that in the filaments , where the gas motions are more uniform , the simulated magnetic fields may depend to a certain extent on the initial seed fields although that is not trivial to quantify in general terms . in addition , numerical resolution may play an important role because it affects the amplification and the topological structure of the magnetic fields , both of which are important for the normalization procedure , see below .
the resolution in ref .
@xcite is constant and much better in filaments and voids but worse in the core of galaxy clusters than the ( variable ) resolution in ref .
if in both simulations the magnetic fields are normalized to ( or reproduce ) the same `` observed ''
values in the core of rich clusters then obviously their values in the filaments will be very different for the reasons outlined above .
this may partly explain why the contribution of filaments to uhecr deflection is more important in ref .
@xcite , although a more detailed analysis and comparison are required to settle the issue . in any case , the magnetic fields obtained in ref .
@xcite seem to be quite extended , as can be seen in fig .
[ fig4 ] : about 10% of the volume is filled with fields stronger than 10 nano gauss , and a fraction of @xmath64 is filled by fields above a micro gauss .
the different amounts of deflection obtained in the simulations of refs .
@xcite show that the distribution of egmf and their effects on uhecr propagation are currently rather uncertain .
finally we note that these studies should be extended to include heavy nuclei @xcite since there are indications that a fraction as large as 80% of iron nuclei may exist above @xmath65ev @xcite . as a consequence , even in the egmf scenario of ref .
@xcite deflections could be considerable and may not allow particle astronomy along many lines of sight : the distribution of deflection angles in ref .
@xcite shows that deflections of protons above @xmath51ev of @xmath66 cover a considerable fraction of the sky .
suppression of deflection along typical lines of sight by small filling factors of deflectors is thus unimportant in this case .
the deflection angle of any nucleus at a given energy passing through such areas will therefore be roughly proportional to its charge as long as energy loss lengths are larger than a few tens of mpc @xcite .
deflection angles of @xmath67 at @xmath68ev should thus be the rule for iron nuclei .
in contrast to the contribution of our galaxy to deflection which can be of comparable size but may be corrected for within sufficiently detailed models of the galactic field , the extra - galactic contribution would be stochastic .
statistical methods are therefore likely to be necessary to learn about uhecr source distributions and characteristics . in addition , should a substantial heavy composition be experimentally confirmed up to the highest energies , some sources would have to be surprisingly nearby , within a few mpc , otherwise only low mass spallation products would survive propagation @xcite . the clustered component of the uhecr spectrum may play a key role in this context and may be caused by discrete sources in directions with small deflection . since , apart from energy losses , cosmic rays of same rigidity @xmath9 are deflected similarly by cosmic magnetic fields , one may expect that the composition of the clustered component may become heavier with increasing energy . indeed , in ref .
@xcite it was speculated that the agasa clusters may be consistent with consecutive he , be - mg , and fe bumps .
we have reviewed two current issues in theoretical ultra - high energy cosmic ray research .
the first one concerns constraints on scenarios attempting to explain highest energy cosmic rays by extra - galactic sources producing not only cosmic rays but also photons : improved data analysis and , in the future , improved data for example from glast can considerably reduce estimates of the true extra - galactic gev @xmath0ray background which acts as a calorimeter of electromagnetic energy injected above @xmath20ev .
already current estimates imply that scenarios producing considerably more photons than hadrons , such as extra - galactic top - down scenarios and the z - burst mechanism , can not explain all of the highest energy cosmic ray flux
. a further reduced diffuse gev @xmath0ray background will start to constrain even normal acceleration scenarios . as for the second issue we pointed out that the influence of large scale cosmic magnetic fields on ultra - high energy cosmic ray propagation is currently hard to quantify and may not allow to do `` particle astronomy '' along most lines of sight , especially if a significant heavy nucleus component is present above @xmath65ev . in this case
extensive monte carlo simulations including nuclei and based on constrained large scale structure simulations will be necessary to fully exploit data from future instruments such as the pierre auger @xcite and euso projects @xcite .
the material presented here is based on collaborations with torsten enlin , francesco miniati and dmitry semikoz whom i would especially like to thank . thanks to francesco also for suggestions on the manuscript .
i also thank igor tkachev for discussions on differences between our uhecr simulations in a structured magnetized universe . for a general introduction on cosmic rays
see , e.g. , v. s. berezinsky , s. v. bulanov , v. a. dogiel , v. l. ginzburg , v. s. ptuskin , _ astrophysics of cosmic rays _ ( north - holland , amsterdam , 1990 ) ; t. k. gaisser , _ cosmic rays and particle physics _ , cambridge university press ( cambridge , 1998 ) .
see , e.g. , j. r. hoerandel , arxiv : astro - ph/0402356 .
v. s. berezinsky , s. i. grigorieva and b. i. hnatyk , arxiv : astro - ph/0403477 .
see , e.g. , a. a. watson , arxiv : astro - ph/0312475 , and references therein . m. teshima et al .
28th international cosmic ray conference , tsukuba , japan , * 1 * ( 2003 ) 437 .
w. s. burgett and m. r. omalley , phys .
d * 67 * , 092002 ( 2003 ) [ arxiv : hep - ph/0301001 ] . c. b. finley and s. westerhoff , arxiv : astro - ph/0309159 . c. b. finley et al .
28th international cosmic ray conference , tsukuba , japan , * 1 * ( 2003 ) 433 .
k. greisen , phys .
* 16 * , 748 ( 1966 ) ; g. t. zatsepin and v. a. kuzmin , jetp lett . * 4 * , 78 ( 1966 ) [ pisma zh
. eksp .
fiz . * 4 * , 114 ( 1966 ) ] .
f. w. stecker , phys .
* 21 * , 1016 ( 1968 ) .
d. r. bergman , proc .
28th international cosmic ray conference , tsukuba , japan , * 1 * ( 2003 ) 397 .
t. abu - zayyad _ et al .
_ [ high resolution fly s eye collaboration ] , arxiv : astro - ph/0208243 ; t. abu - zayyad _ et al . _ [ high resolution fly s eye collaboration ] , arxiv : astro - ph/0208301 . m. takeda _ et al .
_ , phys .
lett . * 81 * , 1163 ( 1998 ) [ arxiv : astro - ph/9807193 ] ; m. takeda _ et al .
_ , astrophys .
j. * 522 * , 225 ( 1999 ) [ arxiv : astro - ph/9902239 ] ; n. hayashida _ et al .
_ , arxiv : astro - ph/0008102 ; see also http : //www - akeno.icrr.u - tokyo.ac.jp / agasa/. for recent reviews see , e.g. , f. halzen and d. hooper , rept .
phys . *
65 * , 1025 ( 2002 ) [ arxiv : astro - ph/0204527 ] ; a. b. mcdonald , c. spiering , s. schonert , e. t. kearns and t. kajita , rev .
instrum . * 75 * , 293 ( 2004 ) [ arxiv : astro - ph/0311343 ] . for recent short reviews
see , e.g. , h. j. vlk , arxiv : astro - ph/0401122 ; h. j. vlk , arxiv : astro - ph/0312585 .
e. waxman and j. n. bahcall , phys .
d * 59 * , 023002 ( 1999 ) [ arxiv : hep - ph/9807282 ] ; j. n. bahcall and e. waxman , phys .
d * 64 * , 023002 ( 2001 ) [ arxiv : hep - ph/9902383 ] .
k. mannheim , r. j. protheroe and j. p. rachen , phys .
d * 63 * , 023003 ( 2001 ) [ arxiv : astro - ph/9812398 ] ; j. p. rachen , r. j. protheroe and k. mannheim , arxiv : astro - ph/9908031 . v. s. berezinsky and a. yu . smirnov , ap .
32 * , 461 ( 1975 ) .
p. s. coppi and f. a. aharonian , astrophys .
j. * 487 * , l9 ( 1997 ) [ arxiv : astro - ph/9610176 ] .
p. sreekumar _ et al .
_ , astrophys .
j. * 494 * , 523 ( 1998 ) [ astro - ph/9709257 ] .
a. w. strong , i. v. moskalenko and o. reimer , arxiv : astro - ph/0306345 .
u. keshet , e. waxman and a. loeb , arxiv : astro - ph/0306442 .
d. v. semikoz and g. sigl , arxiv : hep - ph/0309328 .
for a recent review see , e.g. , d. f. torres and l. a. anchordoqui , arxiv : astro - ph/0402371 . for a review
see , e.g. , p. bhattacharjee and g. sigl , phys . rept . 327 ( 2000 ) 109 .
v. berezinsky , m. kachelriess and a. vilenkin , phys .
lett . * 79 * , 4302 ( 1997 ) [ arxiv : astro - ph/9708217 ] .
m. kachelriess and d. v. semikoz , arxiv : astro - ph/0306282 ; h. b. kim and p. tinyakov , arxiv : astro - ph/0306413 . t. j. weiler , phys . rev .
lett . * 49 * , 234 ( 1982 ) .
astrophys .
j. * 285 * , 495 ( 1984 ) .
astropart .
phys . * 11 * , 303 ( 1999 ) [ hep - ph/9710431 ] .
d. fargion , b. mele and a. salis , rays , astrophys . j. * 517 * , 725 ( 1999 ) [ astro - ph/9710029 ] .
s. yoshida , g. sigl and s. j. lee , phys .
* 81 * , 5505 ( 1998 ) [ hep - ph/9808324 ] .
z. fodor , s. d. katz and a. ringwald , phys .
* 88 * , 171101 ( 2002 ) [ hep - ph/0105064 ] ; hep - ph/0105336 ; jhep * 0206 * , 046 ( 2002 ) [ arxiv : hep - ph/0203198 ] ; a. ringwald , hep - ph/0111112 .
see , e.g. , p. lipari , astropart .
* 1 * , 195 ( 1993 ) . for general information
see http://wsgs02.lngs.infn.it:8000/macro/ ; see also m. ambrosio _ et al .
_ [ macro collaboration ] , astro - ph/0203181 .
p. niessen [ the amanda collaboration ] , arxiv : astro - ph/0306209 ; for the energy range @xmath69ev we rescaled the amanda - b10 limit from http://www-rccn.icrr.u-tokyo.ac.jp/icrc2003/proceedings/pdf/326.pdf to amanda - ii exposure .
s. yoshida for the agasa collaboration , proc . of 27th icrc ( hamburg )
* 3 * , 1142 ( 2001 ) .
r. m. baltrusaitis _ et al .
_ , astrophys .
j. * 281 * , l9 ( 1984 ) ; phys .
d * 31 * , 2192 ( 1985 ) .
i. kravchenko _ et al .
_ , astropart .
phys . * 20 * , 195 ( 2003 ) [ arxiv : astro - ph/0206371 ] ; i. kravchenko , arxiv : astro - ph/0306408 .
p. w. gorham , k. m. liewer and c. j. naudet , astro - ph/9906504 ; p. w. gorham , c. l. hebert , k. m. liewer , c. j. naudet , d. saltzberg and d. williams , arxiv : astro - ph/0310232 . n. g. lehtinen , p. w. gorham , a. r. jacobson and r. a. roussel - dupre , arxiv : astro - ph/0309656 .
o. e. kalashev , v. a. kuzmin , d. v. semikoz and g. sigl , phys .
d * 65 * , 103003 ( 2002 ) [ arxiv : hep - ph/0112351 ] .
d. s. gorbunov , p. g. tinyakov and s. v. troitsky , astropart .
phys . * 18 * , 463 ( 2003 ) [ arxiv : astro - ph/0206385 ] .
s. hannestad , jcap * 0305 * , 004 ( 2003 ) [ arxiv : astro - ph/0303076 ] .
see , e.g. , m. maltoni , t. schwetz , m. a. tortola and j. w. valle , arxiv : hep - ph/0309130 .
s. singh and c. p. ma , phys .
d * 67 * , 023506 ( 2003 ) [ arxiv : astro - ph/0208419 ] .
g. gelmini and a. kusenko , phys .
lett . * 84 * , 1378 ( 2000 ) [ arxiv : hep - ph/9908276 ] . v. berezinsky , m. kachelriess and s. ostapchenko , phys .
lett . * 89 * , 171802 ( 2002 ) [ arxiv : hep - ph/0205218 ] . for general information
see http://www-glast.stanford.edu a. loeb and e. waxman , nature * 405 * , 156 ( 2000 ) [ arxiv : astro - ph/0003447 ] ; f. miniati , mon . not .
* 337 * , 199 ( 2002 ) [ arxiv : astro - ph/0203014 ] . for reviews see , e.g. , p. p. kronberg , rept .
phys . * 57 * , 325 ( 1994 ) ; d. grasso and h. r. rubinstein , phys .
rept . * 348 * , 163 ( 2001 ) [ arxiv : astro - ph/0009061 ] .
j. p. valle , fundamentals of cosmic physics * 19 * ( 1997 ) 1 ; j. l. han and r. wielebinski , arxiv : astro - ph/0209090 .
d. ryu , h. kang , and p. l. biermann , astron . astrophys .
* 335 * ( 1998 ) 19 .
p. blasi , s. burles and a. v. olinto , astrophys .
j. * 514 * , l79 ( 1999 ) [ arxiv : astro - ph/9812487 ] .
g. a. medina - tanco and t. a. ensslin , astropart .
* 16 * , 47 ( 2001 ) [ arxiv : astro - ph/0011454 ] .
e. waxman and j. miralda - escude , astrophys . j. * 472 * , l89 ( 1996 ) [ arxiv : astro - ph/9607059 ] .
g. sigl , m. lemoine and p. biermann , astropart .
* 10 * , 141 ( 1999 ) [ arxiv : astro - ph/9806283 ] . c. isola , m. lemoine and g. sigl , phys .
d * 65 * , 023004 ( 2002 ) [ arxiv : astro - ph/0104289 ] .
m. lemoine , g. sigl and p. biermann , arxiv : astro - ph/9903124 . t. stanev , arxiv : astro - ph/0303123 ; t. stanev , d. seckel and r. engel , phys .
d * 68 * , 103004 ( 2003 ) [ arxiv : astro - ph/0108338 ] ; t. stanev , r. engel , a. mucke , r. j. protheroe and j. p. rachen , phys .
d * 62 * , 093005 ( 2000 ) [ arxiv : astro - ph/0003484 ] . c. isola and g. sigl , phys .
d * 66 * , 083002 ( 2002 ) [ arxiv : astro - ph/0203273 ] .
p. sommers , astropart .
* 14 * , 271 ( 2001 ) [ arxiv : astro - ph/0004016 ] .
p. blasi and d. de marco , astropart .
phys . * 20 * , 559 ( 2004 ) [ arxiv : astro - ph/0307067 ] .
p. blasi and a. v. olinto , phys .
d * 59 * , 023001 ( 1999 ) [ arxiv : astro - ph/9806264 ] .
h. yoshiguchi , s. nagataki , s. tsubaki and k. sato , astrophys .
j. * 586 * , 1211 ( 2003 ) [ erratum - ibid . *
601 * , 592 ( 2004 ) ] [ arxiv : astro - ph/0210132 ] ; h. yoshiguchi , s. nagataki and k. sato , astrophys .
j. * 592 * , 311 ( 2003 ) [ arxiv : astro - ph/0302508 ] ; h. yoshiguchi , s. nagataki and k. sato , cosmic rays above 10**19-ev with modifications by the galactic magnetic astrophys .
j. * 596 * , 1044 ( 2003 ) [ arxiv : astro - ph/0307038 ] .
r. aloisio and v. berezinsky , arxiv : astro - ph/0403095 . g. medina tanco , `` cosmic magnetic fields from the perspective of ultra - high - energy cosmic rays propagation '' , lect . notes phys . * 576 * ( 2001 ) 155 .
g. sigl , f. miniati and t. a. ensslin , phys .
d * 68 * , 043002 ( 2003 ) [ arxiv : astro - ph/0302388 ] ; g. sigl , f. miniati and t. a. ensslin , arxiv : astro - ph/0309695 ; g. sigl , f. miniati and t. a. ensslin , arxiv : astro - ph/0401084 .
r. m. kulsrud , r. cen , j. p. ostriker and d. ryu , astrophys .
j. * 480 * ( 1997 ) 481 .
g. bertone , c. isola , m. lemoine and g. sigl , phys .
d * 66 * , 103003 ( 2002 ) [ arxiv : astro - ph/0209192 ] .
see , e.g. , l. n. epele and e. roulet , jhep * 9810 * , 009 ( 1998 ) [ arxiv : astro - ph/9808104 ] . t. yamamoto , k. mase , m. takeda , n. sakaki and m. teshima , astropart . phys . * 20 * , 405 ( 2004 ) [ arxiv : astro - ph/0312275 ] .
|
we address some current theoretical issues around ultra - high energy cosmic rays .
we recall that scenarios producing more @xmath0rays than cosmic rays up to high redshift can in general only provide a sub - dominant contribution to the ultra - high energy cosmic ray flux .
this includes extra - galactic top - down and the z - burst scenarios .
finally we discuss the influence of large scale cosmic magnetic fields on ultra - high energy cosmic ray propagation which is currently hard to quantify .
the views presented here represent the authors perspective .
|
pairing symmetry in the cuprate superconductors was a subject of intense debate for many years after the discovery of high - temperature superconductivity @xcite , primarily because the interpretation of many conventional techniques , such as quasiparticle tunneling , nmr , angle - resolved photoemission , and penetration depth measurements , was model - dependent .
nevertheless , these phase - insensitive techniques have produced a large body of evidence for @xmath0-wave pairing in the cuprates .
the recent development of phase - sensitive pairing symmetry test @xcite , has yielded compelling evidence for predominantly @xmath0-wave pairing symmetry in a number of optimally doped cuprates @xcite .
a question naturally arises : how universal is the @xmath5 pairing in cuprate superconductors ?
there are numerous theoretical studies suggesting the stability of the @xmath0-wave pair state depends on the details of band structure and the pairing potential @xcite .
there has also been considerable theoretical studies indicating @xcite that a pure @xmath5 pair state is not stable against the formation of time reversal symmetry breaking states such as @xmath6 or @xmath7 , at surfaces , interfaces , near impurities , or below a certain characteristic temperature .
on the experimental side , raman data on bi@xmath2sr@xmath2cacu@xmath2o@xmath8 and tl@xmath2ba@xmath2cuo@xmath8 systems as a function of oxygen content indicate that the order parameter has @xmath0-wave symmetry near optimal doping and isotropic pairing in the over - doped regime @xcite .
there are additional indirect experimental evidences for the fragility of the pure @xmath0-wave state @xcite .
the following will examine the universality issues based mainly on the results of phase sensitive pairing symmetry experiments .
we will conclude with a brief discussion on implications of @xmath0-wave superconductivity in the cuprates .
the sign and magnitude of the pair tunneling critical current i@xmath1 across a josephson junction made with at least one superconductor with unconventional pairing symmetry depends sensitively on the gap function symmetry and relative orientation of the junction electrodes .
the ground state of a superconducting loop with an odd number of negative critical currents ( termed a frustrated " or
@xmath9 " loop ) is doubly degenerate and shows spontaneous magnetization of one - half flux quantum ( @xmath10 - 7@xmath11 ) , provided that @xmath12 , where @xmath13 is the self - inductance of the ring and @xmath14 is the critical current of the weakest junction in the loop @xcite . by varying the loop geometry , the presence or absence of the half - integer flux quantum effect can be used for a definitive determination of the pairing symmetry .
the tricrystal pairing symmetry tests use a ring consisting of three crystals with controlled orientation ( see fig .
1 ) to define the direction of the pair wavefunction .
[ fig : trigeo ] the magnetic flux threading through the tricrystal ring is measured with a scanning squid microscope @xcite for different tricrystal configurations to differentiate between various pairing symmetries .
the @xmath15 effect is intrinsic in a frustrated geometry , and is observed in ring , disk , and blanket film samples at the tricrystal point .
the establishment of @xmath0-wave pairing in some of the cuprates has imposed well defined constraints on possible models of high - temperature superconductivity , but it does not specify the high - t@xmath1 mechanism . to gain further insight into the origin of high - temperature superconductivity , it is important to determine whether the @xmath5 pairing symmetry is universal .
one can investigate the universality issues from the following aspects .
phase - sensitive squid interferometry @xcite , tricrystal magnetometry @xcite , squid magnetometry @xcite , and single junction interferometry @xcite experiments have demonstrated @xmath0-wave pairing symmetry in ybco single crystals or @xmath16-axis oriented epitaxial films .
in addition , tricrystal magnetometry experiments have demonstrated @xmath0-wave pairing symmetry in various high - t@xmath1 hole - doped cuprate systems such as tl@xmath2ba@xmath2cuo@xmath17 , gdba@xmath2cu@xmath3o@xmath4 , and bi@xmath2sr@xmath2cacu@xmath2o@xmath8 .
( for a more complete tabulation of the phase - sensitive experiments see ref .
more recently , the pairing symmetry of the electron - doped cuprate superconductors nd@xmath18ce@xmath19cuo@xmath20 ( ncco ) and pr@xmath18ce@xmath19cuo@xmath20 ( pcco ) has been determined to also be predominantly @xmath0-wave by observing the half - flux quantum effect in @xmath16-axis thick blanket films ( @xmath211@xmath22 m thick ) epitaxially grown on tricrystal substrates with the configuration of fig .
samples with two other geometries , designed to be unfrustrated for a @xmath0-wave superconductor , do not show the half - flux quantum effect . shown in fig .
2a is a scanning squid microscope image of a ncco film [ fig : nccodif ] deposited on a frustrated tricrystal sto substrate , cooled to 4.2k in nominal zero field .
this image was obtained by subtracting an image with an externally applied field of -0.15 mg , which stabilized the half - quantum magnetic vortex with the flux pointing down , from an image taken with an applied field of 0.15 mg , resulting in a half - quantum magnetic vortex pointing up , and dividing by 2 .
this results in an image of the half - quantum magnetic vortex with effects from surface roughness , a smoothly varying magnetic background , and the mutual inductance between the squid and sample subtracted out .
figure 2b shows fitting of cross - sections through the data of fig .
2a to expressions for the magnetic signals expected from a josephson vortex centered at the tricrystal point , with a total flux of @xmath15 @xcite .
the height @xmath23 of the pickup loop was determined by fitting a bulk abrikosov vortex .
the solid lines in fig .
2b are fits to cross - sectional data for the horizontal and diagonal grain boundaries , assuming a total flux of @xmath15 for this vortex , using the josephson penetration depths @xmath24 for the two grain boundaries as the sole fitting parameters .
the dashed lines , assuming a flux of @xmath25 , are much worse fits .
these results indicate that the electron - doped superconductors , just as the hole - doped superconductors , have @xmath0-wave pairing symmetry .
it is of interest to find out whether @xmath0-wave pairing will change to @xmath26-wave , for example , as a function of temperature .
however , the phase - sensitive experiments described above were all performed at liquid helium temperature . with a recently developed variable sample temperature scanning squid microscope @xcite ,
the temperature dependence of the half - integer flux quantum effect in tricrystal ybco was studied .
it was found that the total flux of the josephson vortex at the tricrystal point remains constant ( i.e. @xmath15 ) from 0.5k through t@xmath1 ( @xmath2190k ) @xcite .
this finding means that the @xmath0-wave pair state dominates at all temperatures below t@xmath1 , and that there is little , if any , time - reversal symmetry breaking component in the order parameter , over the entire temperature range .
this is consistent with earlier studies , which found less than a few percent of time - reversal symmetry breaking component in phase - sensitive measurements at low temperatures @xcite .
is there any band structure ( doping ) effect on pairing symmetry ? in the competition between @xmath26-wave and @xmath0-wave channels for high - temperature superconductivity , is there any factor that can stabilize the @xmath0-wave pairing ? in the literature ,
there have been some theoretical studies dealing with these issues : for example , based on a next - near - neighbor pairing interaction , the symmetry of a bcs condensate is predicted to vary as a function of energy band filling and other band parameters @xcite .
the @xmath0-wave pairing channel is energetically favored in a wide range of doping centered around half - filling .
the stability of the @xmath0-wave pair state may be enhanced by the proximity of the fermi level to the van hove singularity in the 2d bands of the cuo@xmath2 plane @xcite .
a series of phase - sensitive pairing symmetry tests as a function of doping in a model system such as tl@xmath2ba@xmath2cuo@xmath17 , hgba@xmath2cuo@xmath27 , bi@xmath2sr@xmath2cacu@xmath2o@xmath8 , or y@xmath28ca@xmath29ba@xmath2cu@xmath3o@xmath4 could test for these effects .
phase - sensitive experiments have produced strong evidence for predominantly @xmath0-wave pairing symmetry in hole- and electron - doped cuprate superconductors .
these tests also indicate , within the experimental accuracy ( @xmath30 at low temperatures ) , the absence of pairing with broken time - reversal symmetry .
the universality of @xmath0-wave pairing symmetry in bulk high - temperature superconductors is thus well - established .
the possibility of pair states without time - reversal symmetry invariance ( e.g. @xmath6 or @xmath7 ) induced at surfaces , interfaces , or around impurities or defects is being actively investigated ( see ref . @xcite and references therein ) .
recent observations of a splitting of the zero - bias conductance peak in quasiparticle tunneling @xcite and spontaneous magnetization in @xmath16-axis ybco thin films @xcite represent possible evidence for such pair states .
dominant @xmath0-wave pairing symmetry in both the hole- and electron - doped cuprates is expected from single - band models of the cuo@xmath2 planes , in which the electron - hole symmetry is an intrinsic property ( see e.g. poole _ et al . _
@xcite ) . from another perspective , the predominance of @xmath5 pairing in cuprates underscores the important role of strong electron correlation in determining the superconducting gap symmetry and other properties .
it is experimentally and theoretically well established that a strong on - site coulomb repulsion is present in all cuprates .
such a strong electron correlation causes the universally - seen mott transition at half - filling .
the strong on - site coulomb repulsion rules out simple @xmath26-wave pairing .
the authors wish to thank m. bhushan , c.c .
chi , a. gupta , z.z .
ketchen , k.a .
moler , d.m .
newns , h. raffy , z.f .
ren , j.z .
sun , g. trafas , and j.h .
wang for invaluable assistance in the course of our tricrystal experiments .
|
phase - sensitive tests of pairing symmetry have provided strong evidence for predominantly @xmath0-wave pairing symmetry in both the hole- and electron - doped high - t@xmath1 cuprate superconductors .
temperature dependent measurements in yba@xmath2cu@xmath3o@xmath4 ( ybco ) indicate that the @xmath0-wave pairing dominates , with little if any imaginary component , at all temperatures from 0.5k through t@xmath1 . in this article
we review some of this evidence and discuss the implications of the universal @xmath0-wave pairing symmetry in the cuprates .
|
the profiles of the broad emission lines in the spectrum of active galactic nuclei ( agns ) are the main source of information about the motion of the high density gas in the broad line region ( blr ) .
detailed studies of such profiles have been the focus of intense investigation for many years ( see netzer 1990 for a review of older work and marziani et al .
1996 and richards et al .
2002 for more recent publications ) .
unfortunately , several rather different geometries can conspire to result in similar line profiles and today , there is no way to infer , directly , the global blr motion from line profile fitting .
a less ambitious goal is to use a measure of the observed line width , e.g. the line fwhm , or the line dispersion ( see peterson et al .
2004 for definitions ) as indicators of the mean emissivity - weighted velocity of the blr gas .
such measurements are crucial for deducing black hole ( bh ) mass ( @xmath1 ) in cases where the emissivity - weighted radius , , is measured directly from reverberation mapping ( rm ) experiments , or estimated from from l-@xmath6 relationships that are based on such studies ( see kaspi et al . 2000 ; kaspi et al 2005 ; vestergaard and peterson 2006 for reviews ) .
a typical expression of this type is @xmath7 ^{\gamma } \,\ , pc , \label{eq : old_rblr}\ ] ] where is the continuum luminosity ( @xmath8 ) at 5100and @xmath9 .
the constant @xmath10 depends on the line in question . for , @xmath11pc ( e , g , bentz et al .
2009 ) and for , @xmath12pc ( kaspi et al .
2007 after assuming @xmath13 ) . for a virialized blr ,
the above can be combined with a measure of the fwhm , or the line dispersion , to obtain the bh mass , @xmath14 where the constant @xmath15 is a geometrical correction factor of order unity that takes into accounts the ( unknown ) gas distribution and dynamics .
various possible values of @xmath15 have been computed by collin et al .
( 2006 ) for various possible geometries .
however , the only empirical way to determine @xmath15 is to compare the results of eqn .
[ eq : old_mbh ] with independent measurements of @xmath1 , like those available in the the case where the central bh resides in a bulge and @xmath1 can be estimates from the @xmath5 relationship ( e.g. tremaine et al . 2002 ) . such comparisons by onken et al ( 2004 ) and by woo et al .
( 2010 ) , using the rm data base , suggest @xmath16 .
in a recent paper , marconi et al .
( 2008 ; hereafter m08 ) investigated the role of radiation pressure force and its effect on the motion of the blr gas and the required modification to the bh mass estimate . according to m08
, radiation pressure plays an important role in affecting the cloud motion provided the column density ( ) of most blr clouds is smaller than about @xmath17 . according to m08 , in such a case
, there is a need to add a second term to eq .
[ eq : old_mbh ] .
this term depends on the source luminosity and .
the modified form suggested in m08 is @xmath18 where@xmath19 replaces @xmath15 in eqn .
[ eq : old_mbh ] and @xmath20 is a second constant . if @xmath21 and is measured in units of @xmath22 , @xmath23 .
according to m08 , failing to account for the second term results in the underestimation of @xmath1 . obviously , the inclusion of such a term results in @xmath24 .
m08 repeated the analysis of onken et al .
( 2004 ) and vestergaard and peterson ( 2006 ) , taking into account the new term and solving for @xmath19 and .
this resulted in @xmath25 and @xmath26 .
the m08 suggestion can be tested by comparing low redshift samples of type - i and type - ii agns since the estimate of @xmath1 in the latter does not involve the source luminosity and gas dynamics .
netzer ( 2009 ) carried out such a comparison and found that radiation pressure force plays only a marginal role in such sources .
the conclusion is that , in many agns , the mean column density of the blr clouds exceeds @xmath27 . in a later work , marconi
et al ( 2009 ; hereafter m09 ) argued that firm conclusions regarding the role of radiation pressure force are difficult to obtain since the column density in some blrs can be different than in others and there is no simple way to evaluate the overall effect of such a column density distribution .
the treatment of a certain type of cloud in all sources , or even in a single blr , is of course highly simplified and eqns .
[ eq : old_mbh ] and [ eq : m08_mbh ] must be treated as crude first approximations .
the critical and detailed evaluation of the role of radiation pressure force in `` real '' blrs is the subject of the present paper . in
2 we present our basic equations and in 3 we use them to calculate various expected broad emission line profiles and mass normalization factors , @xmath15 . 4 deals with the evaluation of present day @xmath1 estimates and suggests a new way to estimate @xmath1 and which is consistent with our calculations .
in this work we focus on the `` cloud model '' of the blr . the general framework of this model is explained in netzer ( 1990 ) and in kaspi and netzer ( 1999 ) and a major empirical justification is obtained from the recent x - ray detected single blobs , or clouds , moving in a region which is typical , in terms of velocity and dimension , of the blr ( risaliti et al .
2010 ; maiolino et al .
we do not consider the locally optimally - emitting cloud ( loc ) model ( baldwin et al .
1995 ; korista & goad 2000 ) where , at every location , there is a large range in cloud properties .
the dynamics of the blr gas in this model has never been treated and is far more complicated than the one considered here .
another possibility that has been discussed , extensively , is that wind from the inner disk plays an important role in feeding and driving the blr gas .
possible evidence for this scenario comes from radio observations ( e.g. vestergaard , wilkes , & barthel 2000 ; jarvis & mclure 2006 ) and spectropolarimetry ( smith et al .
2004 ; young et al .
theoretical considerations are discussed in bottorff et al .
( 1997 ) , murray & chiang ( 1997 ) , proga , stone , & kallman ( 2000 ) , everett ( 2003 ) , young et al .
( 2007 ) , and several other papers . while our calculations apply to any cloud , even those created and driven by such winds , the specific examples given below are more applicable to bound clouds where inward and outward motions are both allowed .
the basic equation of motion , ignoring drag force , is @xmath28 -\frac{1}{\rho } \frac{dp_g}{dr } \,\ , , \label{eq : a_r_modified_1}\ ] ] where @xmath29 is the force multiplier , @xmath30 is the bolometric luminosity , @xmath31 is the average number of nucleons per electron , and @xmath32 .
the force multiplier depends on the gas composition and its level of ionization .
an interesting case is a compton thin neutral cloud that absorbs all the ionizing radiation ( a compton thin `` block '' ) . in this case @xmath33 ) where @xmath34 is the hydrogen column density and @xmath35 is the fraction of the bolometric luminosity which is absorbed by the gas .
for such a `` block '' , @xmath36 but in general @xmath35 is radius dependent because of the changing column density and level of ionization of the gas ( see below ) .
ignoring thermal pressure we obtain @xmath37 \label{eq : a_r_modified_2}\ ] ] where @xmath38 .
thus , radiation pressure is the dominant force when @xmath39 the above expressions , including the one for the limiting @xmath40 , include only radial terms and assume a pure radially dependent radiation pressure force .
the calculations of real orbits , and the conditions for cloud escape , require their integration and will thus include the standard constants of motion ( energy and angular momentum ) .
obviously , the conditions for escape depend on the cloud azimuthal velocity , @xmath41 , and can differ substantially from what is obtained by using eqn .
[ eq : lim_gamma ] .
m08 and m09 derived similar expressions for the case of completely opaque clouds . according to them ,
radiation pressure dominates the cloud motion if @xmath42 where @xmath43 .
the two expressions provide the same limiting @xmath40 when @xmath44 a recent paper by ferland et al .
( 2009 ) , where mostly neutral , infalling clouds are considered , reaches basically identical conclusions .
blr clouds are likely to be confined .
the confining mechanism is not known but high temperature gas and magnetic confinement have been proposed .
the approach chosen here is consistent with the idea of magnetic confinement and some justifications for it are given in rees , netzer and ferland ( 1989 ) .
we adopt a simple model of numerous individual clouds that are moving under the combined influence of the bh gravity and radiation pressure force .
following netzer ( 1990 ) , and kaspi and netzer ( 1999 ) , we assume that clouds retain their mass as they move in or out and the gas density changes with radius in a way that depends on the radial changes of the confining pressure .
assume the external pressure and hence the gas density in individual clouds are proportional to the radial coordinate , @xmath45 .
a reasonable guess that agrees with observations is @xmath46 ( rees , netzer and ferland 1989 ) .
this results in a radial dependence of the ionization parameter ( the ratio of ionizing photon density to gas density ) , @xmath47 . for spherical clouds , @xmath48 , @xmath49 and @xmath50 , where @xmath51 is the cloud radius and
@xmath52 its geometrical cross section .
the line intensity contributed by a single cloud , @xmath53 , depends on its covering factor and the line emissivity @xmath54 which depends on the conditions in the gas , @xmath55 in the real calculations we ignore factors of order unity relating the mean cloud `` size '' and its mean column density since this is not known and require different type of calculations .
the above considerations suggest that the importance of radiation pressure increases with distance because of the dependence of @xmath34 on @xmath56 , i.e. @xmath57 since @xmath40 depends on the global accretion rate which has little to do with cloud properties , the more physical approach is to consider the case of a certain @xmath40 and follow the cloud motion .
the examples discussed below follow this approach . in this work we consider three types of clouds : 1 .
very large column density clouds where radiation pressure force is negligible at all distances . here
virial cloud motion is a good approximation ( the ferland et al 2009 infalling clouds belong to this category ) .
2 . cloud for which radiation pressure is very important somewhere inside the `` classical blr '' ( e.g. inside the rm radius ) .
such clouds will escape the system on dynamical time scales and their contribution to the line profiles is small except for times immediately after a large increase in .
clouds for which radiation pressure is non - negligible but is not strong enough to allow escape .
such clouds are the ones discussed by m08 albeit without the radial dependence of considered here .
this case is the one most relevant to real blrs and we discuss it in detail in the following section .
the modified equation of motion is obtained from eqn .
[ eq : a_r_modified_2 ] by including the radial dependence of @xmath34 .
we define @xmath58 to be the distance where @xmath59 .
this gives @xmath60 \,\ , .
\label{eq : a_r_modified_3}\ ] ] the column density dependent critical distance where radiation force dominates the cloud motion is , @xmath61 ^{3/2s } \,\ , .
\label{eq : r_min}\ ] ] for example , the case of @xmath62 and @xmath63 gives a critical radius of @xmath64 for radially moving clouds .
this dependence of @xmath56 on @xmath40 is the main motivation to suggest a new method for evaluating @xmath1and ( 4 ) . as explained ,
the critical radius should not be confused with the point of escape from the system .
non - radial velocity components ( @xmath41 ) , that reflect the energy and angular momentum of the system will act to reduce this radius ( see examples below ) .
the motion of blr clouds with the above properties involves an acceleration term of the form , @xmath65 where @xmath66 and @xmath67 the radial potential is , @xmath68 where @xmath69 is the radius where @xmath70 .
below we use this potential to calculate cloud orbits and line profiles . the energy and angular momentum terms that result from the above integration ,
are included in the calculation by fixing the initial conditions , @xmath56 and the two velocity components at this location .
we carried out a series of calculations under a variety of conditions considered to be typical of different blrs .
every model is calculated for assumed @xmath1 and @xmath40 .
this specify and thus the potential @xmath71 .
the additional model parameters are : 1 . the radial parameter @xmath72 .
2 . the cloud column density normalization factor @xmath58 .
3 . the initial radius @xmath73 and the initial velocity @xmath74 .
we assume that the orbits of clouds with very large @xmath34 are ellipses of given eccentricities .
@xmath73 is chosen to be the apogee of the orbit and @xmath75 ( given below in units of the keplerian velocity , @xmath76 ) is determined from these conditions . this a simple way to specify the angular momentum . in the examples below ,
we focus on those cases where the resulting fwhms are consistent with the observations of the broad and emission lines but give details for several others .
the initial ionization parameter , @xmath77 .
we note that the exact value of the gas density , @xmath78 , is less important . in the following we assume @xmath79 for all cases .
the three - dimensional distribution of orbits .
this is done in two steps .
first we calculate the motion of numerous identical clouds in a plane and then distribute many such planes in a spherical geometry specified by the inclinations of the planes to the line of sight .
the profiles given below are only those for a line of sight which is perpendicular to the central plane of motion ( if such a plane exists ) .
all calculations assume a large enough number of clouds such that the predicted profiles are smooth ( see bottorff and ferland 2000 for discussion and earlier references on this issue ) .
physical properties that _ are not _ included in the present calculations are non - isotropic central radiation field , non - isotropic line emission , the photoionization of gas with a range of density and metallicity , different inclinations of the line of sight to the central plane of motion , large cloud covering factors in a specific direction , and central obscuration , e.g. by the accretion disk .
several of those are likely to be important in real blrs but are beyond the scope of the present work .
[ fig : motion_plane ] illustrates the orbits of three @xmath80 , @xmath81 cm , @xmath82 and @xmath83 clouds moving under the influence of a @xmath84 bh .
the first is an ellipse typical of a cloud which is not affected by radiation pressure ( e.g. @xmath85 ) .
this is shown by a thick solid line .
formally speaking , such clouds are compton thick but this is of no practical implications since the only intention is to show a simple , gravity dominated orbit .
the second is a case where @xmath86 . here
, radiation pressure force is significant and acts to constantly changing the direction of motion of the cloud .
this results in a rotating planar orbit .
the third orbit ( dashed line ) follows the trajectory of a smaller column density cloud ( @xmath87 ) that escapes the system .
increasing @xmath40 will result in similar type orbits for the rotating orbit second cloud except that the angle between two successive revolutions will increase .
we calculated various line profiles for the case of @xmath1=@xmath84 , @xmath86 , @xmath88 and @xmath40 in the range of 0.05 ( negligible radiation pressure force ) to 0.735 ( just below escape ) .
the bolometric luminosity in each of those is obtained from the combination of @xmath1 and @xmath40 .
we assume @xmath63 at all radii and @xmath53 which takes into account _ only _ geometrical factors ( i.e. constant @xmath54 , see eqn . [ eq : emissivity ] ) and isotropic line emission . in terms of total line emission , this is a reasonable approximation for lines like that reprocess roughly a constant fraction of the ionizing continuum radiation .
obviously , a large optical depth in will result in line emission anisotropy which is not considered here .
it is not appropriate for lines like whose intensity is more sensitive to the level of ionization and the gas temperature . at this stage
we specifically avoid the use of a varying @xmath35 since the effect on the orbit can be significant even for small changes in this parameter ( see below ) .
the resulting profiles , assuming a complete spherical atmosphere ( the entire @xmath89 radians range relative to the central plane ) , are shown in fig .
[ fig : profiles_ledd ] . as expected , the profile becomes narrower with the increasing @xmath40 reflecting the fact that , as the luminosity increases , the cloud spend less and less time at small radii .
the top part of table 1 provides additional information about the calculations . for each profile
we give the fwhm in units of @xmath90 , the mean emissivity weighted radius , @xmath91 , and the mass correction factor @xmath15 ( eqn .
[ eq : old_mbh ] ) .
the calculation of @xmath92 is obtained by weighting the emissivity of the cloud and the time it spends at each radius .
this is roughly equivalent to the observed rm radius .
the mass correction factor is obtained by requiring @xmath93 .
we also show ( in parenthesis ) the values of fwhm and @xmath15 obtained for the case of a thick central disk which represents only a part of a spherical distribution . here
the cloud distribution correspond to a width , relative to the central plane , of @xmath94 radians .
the reduction in fwhm relative to the complete sphere is about a factor of 0.6 and there is a corresponding increase in @xmath15 .
computed line profiles that are typical of this and similar geometries are shown in fig .
[ fig : profiles_sectors ] .
.line widths , mass conversion factor @xmath15 , and emissivity - weighted radii for various models [ cols="<,^,^,^",options="header " , ] [ tab : hb_civ ] in conclusion , the simple single zone models explored here can not reproduce all the observed properties : line intensity ratio , mean emissivity radii and fwhm ratio .
the main reason is that the starting conditions fix the cloud orbit , and hence the line emissivity and fwhm .
simple two - zone models like the ones presented here can account for most observed properties of the and lines . in particular , they can account for the mean line ratio , the mean emissivity weighted radii and the mean relative fwhm of the and lines measured in various rm samples .
obviously , such simple models do not intend to explain all the observed line profile properties that can differ from one object to the next and contain additional components ( see some obvious examples for complex profiles in richards et al .
2002 and sulentic et al . 2007 ) . fitting those is deferred to a forthcoming paper
the above calculations allow us to investigate the intensity , the width and the shape of the broad emission lines and to evaluate various methods used to estimate @xmath1 .
we defer the discussion of specific observed line profiles to a future paper .
assume a system of clouds with a given total amount of gas and a large range of column densities .
such a system will eventually break into three : virialized clouds , non - virialized bound clouds and escaping clouds .
the third group will not contribute significantly to the observed line emission for more than several dynamical times .
the relative contribution of the first and second groups to the line emission depend on the cloud mass distribution . a sudden increase in will increase the importance of radiation pressure and will remove more gas from the system .
a decrease in will drive the system closer to virial equilibrium . a new gas supply , e.g. from a disk - wind , will produce bound as well as unbound clouds .
all aspects of this general scenario must be considered when evaluating the observed line profiles and the various methods developed to use them in estimating @xmath1 .
a major objective of the present paper is to evaluate the accuracy and the normalization of various @xmath1 estimators in type - i agns .
the results presented in tables 1 & 2 suggest the following : 1 .
every agn is likely to contain a large number of clouds with a large range in
. this can be the result of a broad cloud mass distribution and/or due to cloud motion in a radial - pressure dependent environment with a positive value of @xmath72 . a given @xmath40 results in a lower limit on at _ a given location _ for a given orbit eccentricity .
under such conditions , there are always some clouds , e.g. those that are very close to the bh , for which radiation pressure is negligible . for others
, radiation pressure can be very important .
2 . for a small enough
, the effective depends on both @xmath40 and . under these conditions
, the bh mass itself is an important factor in determining . to illustrate this ,
consider two agns with identical sed , , blr geometry , distribution and inclination to the line of sight .
the effective in the two is the same provided they harbor identical bhs .
different will be measured if the two bhs have different masses despite of the fact that is the same in both .
this is the result of the larger @xmath40 in the smaller bh agn .
the effect may not be recognized in a large sample of sources and can , in fact , be attributed to a large intrinsic scatter in the @xmath95 relationship . any derived @xmath95 relationship will depend on the properties of the sources in the chosen rm sample , in particular on the distribution of @xmath40 .
assuming a range in in _ every _ agn , the m08 suggestion to include a luminosity dependent term in the calculation of @xmath1 ( eqn .
[ eq : m08_mbh ] ) is not in accord with our calculation that indicate that and fwhm depend on @xmath2 and not on @xmath3 .
the multi - year rm campaign of ngc5548 is the best example to test some of these ideas in a specific source .
the campaign has been described and analyzed in numerous papers and the ones most relevant to the present study are peterson et al .
( 1999 ) and gilbert and peterson ( 2003 ) .
[ fig : n5548_l_t_jd ] shows the variations in and time lag ( in this case the centroid of the ccf ) in ngc5548 .
each point represents a full observing season which is typically @xmath96 days long .
the data are taken from the recent compilation of bentz et al .
( 2009 ) which provides the best galaxy subtracted flux at 5100 .
the uncertainty on is basically the range of this quantity over the observing season .
this is of the same order as the variation from one season to the next .
as clearly seen from the diagram , ( ) lags the continuum in such a way that more luminous phases are associated with longer lags . this has been noted in earlier publications , e.g. gilbert and peterson ( 2003 ) .
[ fig : n5548_l_t ] shows t(lag ) vs. for the same data set . while the uncertainties are large , some correlation , with a slope of 0.5 - 1 , is evident
. an earlier version of the diagram , with fewer points , is shown in peterson et al .
( 1999 ) . for ngc5548 , @xmath1@xmath97 and ( ) @xmath98 l.d ..
thus , the dynamical time is of order 6 years and the time it takes to change by 50% ( e.g. table 1 ) is approximately 3 years .
this seems to be compatible with the changes in and t(lag ) in fig .
[ fig : n5548_l_t_jd ] , thus some adjustment of ( ) due to the effects discussed in this work are possible .
the measured , with a bolometric correction factor of about 10 , indicates a mean @xmath40 of about 0.02 .
the bottom part of table 1 provides approximate parameters for such a case .
any successful model of ngc5548 should account for the behavior shown in fig .
[ fig : n5548_l_t_jd ] , as well as for the observed fwhms and luminosities of both h@xmath99 and .
while the full investigation is deferred to a future paper , we consider here the predicted lags for @xmath40 = 0.005 , 0.01 , 0.02 assuming a bh mass of 10@xmath100 m@xmath101 and two families of clouds : one with @xmath102 and @xmath103 , and one with @xmath104 and @xmath105 .
the second assumed family of clouds results in pseudo - orbits of higher eccentricity that , as explained earlier , are more strongly affected by radiation pressure . in both cases ,
the predicted lags for @xmath106 are consistent with the observed values ( @xmath107(lag ) @xmath108 1.34 at @xmath109 ) .
however , the calculated slope of @xmath107(lag ) vs. is flatter than observed . as argued earlier , a single family of blr clouds can not provide a full explanation to the observed spectrum of many agns
this must applied to ngc5548 ( to appreciate the complexity of this case see the various components considered by kaspi and netzer ( 1999 ) to explain only the variable line intensities ) .
the simple examples considered here suggests that dynamical scaling of the blr in ngc5548 , due to radiation pressure force , is an additional , physically - motivated mechanism that must be added to any cloud model when attempting to explain the observed variations in @xmath110(lag ) .
current bh mass estimates utilize rm - based measurements of , measured fwhms ( or an equivalent velocity estimator ) of certain broad emission lines , and eqn .
[ eq : old_mbh ] .
the normalization constant @xmath15 is obtained by a comparing @xmath1 obtained in this way with the mass obtained from the @xmath5 method .
having examined a large range of cloud orbits and line profiles under various conditions , and the corresponding values of the effective , we are now in a position to evaluate the merits of this method .
we consider three general possibilities .
the first is the case where all agns contain blr clouds with a wide column density distribution .
a randomly chosen object will have in its blr some clouds that are affected by radiation pressure force and others that are not .
this is the case for _ any _ @xmath40 .
the cloud dynamics and the observed line profiles reflect the ( unknown ) column density distribution .
our calculations suggest that an rm sample drawn randomly from such an agn population can be safely used to determine the best value of @xmath15 by comparing the derived @xmath1 with the @xmath5 method .
this is justified by the fact that @xmath111 even if radiation pressure force is important ( see table 1 ) .
the observed fwhms are , indeed , smaller than the ones that would have been observed if all clouds had extremely large column densities .
this , however , has no practical implication since the column densities are not known and @xmath15 is simply a normalization factor that serves to bring two completely different methods of estimating @xmath1 into agreement .
mass estimates obtained in this ways are reliable provided the properties of the rm sample represent well the population properties .
the second case reflects a situation where the cloud column density distribution is , again , very broad but part of the population is under - represented in the rm sample . for example , the rm sample may contain mostly sources with @xmath112 while the overall distribution of @xmath40 is much wider . in this case , the normalization factor @xmath15 will reflect only the properties of the measured sources and its use will provide poor mass estimates for cases with much larger or much smaller accretion rates .
this may well be the case in the rm sample which is most commonly used ( kaspi et al .
2000 ; bentz et al , 2009 ) that contains only very few agns with @xmath113 .
the numbers in table 1 enable us to evaluate the resulting deviations in the estimated @xmath1 .
for example , if we use the first part of the table and assume a source with a certain and @xmath82 , we find that the mass of a similar source with @xmath114 will be under - estimated by a factor of 1.11/0.75 . regarding the second case , it is important to note that under - estimates and over - estimates of @xmath1 are equally likely .
consider again an rm sample where , for most sources , @xmath82 .
this results in a certain value of @xmath15 which takes into account the effect of radiation pressure force in _ some _ of these sources ( see bottom part of table 1 ) .
assume a second , randomly selected agn sample with a similar bh mass distribution but a typical @xmath40 which is much smaller than 0.1 .
most measured fwhms in this sample are _ broader _ than those in the rm sample because radiation pressure force is not as effective in reducing the cloud velocity . using the value of @xmath15 derived for the rm sample will result in _ over estimating _
@xmath1 in the second sample .
the lower part of table 1 gives some idea about the magnitude of this effect , e.g. an over estimate by a factor of 1.01/0.76 .
the third case is similar to the first one except that large luminosity variations , on time scales that are not too different from the blr dynamical time , are occurring in most sources , including those selected for rm monitoring .
table 1 shows that , like the first case , the deduced @xmath15 represents well the population because @xmath92 follows the variations in .
the mean @xmath1 in such a sample is recovered albeit with a larger uncertainty .
given the above considerations , we now investigate an alternative method to calculate @xmath1 .
the method takes into account the effect of radiation pressure force on the cloud motion and the results will be compared to those obtained by the old method ( eqn .
[ eq : old_mbh ] ) and by the m08 method .
our new calculations indicate that the emissivity weighted depends both on the ( large ) range in @xmath3 across the entire agn population , as well as on short time scale changes in in individual sources .
the first of those depends roughly on @xmath4 and is a manifestation of the observational fact that the ionization parameter , @xmath115 , and the spectral energy distribution ( sed ) , are not changing much with source luminosity .
the second reflects changes in the blr structure due to the reaction of various column density clouds to the ( changing ) radiation pressure force .
this depends on both and @xmath1 .
this is seen for example in eqn .
[ eq : r_min ] for the critical radius where clouds can escape the system and also in the calculations of table 1 .
it is therefore reasonable to assume that is given by an expression of the form , @xmath116 where @xmath117 and @xmath118 are constants and @xmath3 is a measure of the source luminosity , e.g. if = ( ) . obviously , the above approximation is not unique and one can assume other dependences that are consistent with the line profile calculations , e.g. a dependence of fwhm on @xmath2 .
the idea of introducing a second , luminosity dependent term into the calculation of @xmath1 is not new . in particular
, m08 suggested an expression for @xmath1 which depends on both @xmath4 and @xmath119 ( eqn .
[ eq : m08_mbh ] ) . assuming all agns obey the same relationship , and
is the same in all , the m08 expression leads to extremely large values of @xmath1 for the most luminous agns .
the reason is the linear dependence of @xmath1 on @xmath3 at very high luminosities combined with the calibration of the relationship at small @xmath3 , typical of the @xmath5 sample of onken et al .
the additional consequence of this approach is an upper limit of @xmath112 in many high luminosity , large bh mass sources . in their later work
, m09 considered the possibility that can differ from one source to another but is still constant for all clouds in a given blr .
this would result in smaller @xmath1 and larger @xmath40 in some high luminosity sources since in some blrs , can exceed @xmath17 by a large factor thus reducing the importance of radiation pressure force .
the limitation of the m08 mass estimate is the detachment of @xmath3 from @xmath1 . as shown here , this is not the case in more realistic blrs , especially those where the masses of the clouds are conserved . in such cases ,
the location of the outer clouds that still contribute to the line profiles depends on @xmath2 and the 3d - velocities of the marginally bound clouds are such that the product @xmath120 is not very different from what is found in pure gravity dominated systems .
moreover , for pressure confined clouds , the dependence on is likely to be different in different parts of the blr .
thus , we are looking for an expression that will reflect , properly , all these effects and will allow for the possibility of a range of column densities in _ every source_. we also want to avoid biasing in the derivation of @xmath1 in the limits of very large or very small @xmath3 and to retain the experimental results that @xmath121 with @xmath9 .
all the above can be achieved by assuming that is given by eqn .
[ eq : new_rblr ] and requiring that @xmath122 . for the sake of simplicity
, we assume @xmath123 and @xmath124 and substitute eqn .
[ eq : new_rblr ] into the mass expression .
this leads to a simple quadratic equation in @xmath1 with the following solution , @xmath125 \,\ , , \label{eq : new_mbh}\ ] ] where @xmath117 and @xmath118 are the same ones used in eqn .
[ eq : new_rblr ] except for a common multiplicative constant which depends on the units of , 5100 and @xmath1 .
for example , using the measure parameters for the line , @xmath126 , fwhm = fwhm ( ) , then the constant multiplying @xmath117 and @xmath118 in eqn .
[ eq : new_rblr ] is 10@xmath127 when @xmath1 is measured in , 5100 in units of @xmath128 ^-1 ergs s@xmath129 and in cm .
we used eqn .
[ eq : new_mbh ] and the woo et al .
( 2010 ) sample to find @xmath117 and @xmath118 for 29 agns with measured @xmath130 .
the list is an extension of the one used by onken et al .
( 2004 ) that contains only 16 sources .
we have supplemented the data in woo et al . by data from bentz et al .
( 2009 ) on and 5100where this information was missing .
first , we performed a @xmath131 analysis on @xmath1(rm ) vs. @xmath5 using the parameters recommended by gltekin et al .
this gave @xmath132 which is consistent with the values found by onken et al .
( 2004 ) and woo et al .
( 2010 ) by a factor of about @xmath133 .
all these numbers are sensitive to the error estimate in @xmath130 and in the virial product . ]
next we carried out a @xmath131 minimization to solve for @xmath117 and @xmath118 in eqn .
[ eq : new_mbh ] .
since the minimization involves the error estimate on 5100 , and since this error is not very well defined given the combination of observational uncertainly and the intrinsic scatter in 5100 over several long rm campaigns , we decided to adopt a uniform value of @xmath134 .
we also assume a minimum of 0.1 to @xmath135 and a minimum of 0.05 on @xmath136 .
our results depend slightly on these assumptions .
the best values obtained in this procedure are @xmath137 , @xmath138 and @xmath139 .
extensive tests show that the @xmath131 changes very little if @xmath117 or @xmath118 are changing by up to 10% .
this is the result of some degeneracy between @xmath117 and @xmath118 ( see eqn .
[ eq : new_mbh ] ) .
the average deviation between the new mass estimates and those obtained by the @xmath140 method is 0.31 dex .
there is a weak dependence of the deviation on the line width ( larger deviation for larger fwhm ( ) ) which is marginal given the small number of sources in the sample .
the corresponding number for the deviation of masses obtained directly from the rm measurements and the above value of @xmath15 is 0.36 dex . thus the new method is , indeed , superior in this respect .
obviously it is not surprising to find such an improvement when adding a new free parameter to the model . to compare the various mass estimates more thoroughly , we calculated @xmath1 in three different ways : the old method ( eqn .
[ eq : old_mbh ] ) with @xmath132 , the m08 method ( eqn .
[ eq : m08_mbh ] ) with @xmath141 and @xmath142 ( as in m08 ) , and the new method ( eqn .
[ eq : new_mbh ] ) with the above @xmath117 and @xmath118 . for the m08 method
, we followed the m09 recommendation and assumed a log normal distribution of with a mean of @xmath17 and a large standard deviation of 0.5 dex .
we also calculated in the old ( eqn .
[ eq : old_rblr ] ) and new ( eqn .
[ eq : new_rblr ] ) ways .
[ fig : simulation ] compares two mass ratios , @xmath1(new)/@xmath1(old ) ( red points ) and @xmath1(new)/@xmath1(m08 ) ( black points ) , in a large simulated agn sample .
the sample covers , uniformly , the luminosity range = @xmath143 ^-1 ergs s@xmath129and the simulations assume a gaussian , luminosity independent distribution of fwhm ( ) with a mean of 4,500 ^-1 kms@xmath129 and a variance of 1,500 ^-1 kms@xmath129 .
the diagram shows that the new and old estimates are similar at all but @xmath1(m08 ) deviates from both , by a large factor , at both low and high luminosities .
moreover , the slight deviation between the new and old methods at the very high luminosity end , by up to about 0.2 dex in @xmath1 , is most likely due to the fact that the procedure used to obtain @xmath117 and @xmath118 is based on a sample of 29 mostly low - to - intermediate luminosity agns while the simulations reach a much larger value of .
a comparison of the estimated ( eq .
[ eq : old_rblr ] and [ eq : new_rblr ] ) leads to similar conclusions .
we also made a similar test on the netzer and trakhtenbrot ( 2007 ) sample using all three methods .
the luminosity range in this case is smaller but the fwhm ( ) distribution more typical of observed agns .
the results ( not shown here ) are very similar to those of the simulated sample . in conclusion , the new method for estimating @xmath1 gives results that do not deviate much from the old method which is based on a single constant @xmath15 .
this is true at both high and low luminosities and over a large range in fwhm . obviously , the range of parameters tested here ( @xmath72 , orbit eccentricity , several types of cloud distributions , etc . )
is rather limited and more extensive modeling is required to confirm these results .
however , it is our opinion that the main limitation of the @xmath1 determination methods remains observational and is related to the fact that the present agn @xmath5 sample is small ( 29 sources ) and can not possibly represent the entire range of properties , mostly @xmath40 , observed in agns .
we have investigated the motion of blr clouds with time - independent mass under a range of conditions defined by a radial - dependent confining pressure .
these conditions enforce a range of in every blr , even if the intrinsic mass distribution of the cloud is narrow .
we calculated cloud orbits under a central potential that includes a radiation pressure term .
the orbits were then combined to predict emission line profiles in several simple situations .
we only considered uniformly emitted emission lines and the preliminary comparison with with actual observations used realistic emissivity and column density distributions but was limited to the and lines and at most two different cloud distributions .
we found significant changes in cloud locations and velocities for those cases where the column densities are small enough to allow a significant contribution due to radiation pressure .
this can be important in both high and low @xmath40 sources .
however , while cloud orbits are strongly influenced by the radiation pressure force , there is a relatively small change in the mean @xmath120 and hence no large underestimation or overestimation of @xmath1 .
we illustrate this behavior in several cases but note that other cloud distributions , with different mass , location and velocity distributions , may lead to somewhat different conclusions .
we used the new results to suggest a novel method for calculating and @xmath1 by applying two new constants that were calculated by a comparison of the and 5100 observations and the @xmath140 agn sample of woo et al .
we applied the method to several large observed and simulated agn samples and demonstrated good agreement between the new and the old , pure gravity based methods .
the comparison with the m08 methods shows large deviations in the estimates of @xmath1 .
we acknowledge useful comments by an anonymous referee and a detection of a typo in table 1 by j.m .
funding for this work has been provided by the israel science foundation grant 364/07 and by the jack adler chair for extragalactic astronomy .
hn thanks the hospitality of imperial college london and university college london where part of this work has been done .
pm is grateful for the hospitality and support of the school of physics and astronomy at tel aviv university .
baldwin , j. a , et al .
1995 , apj , 455 , l119 bentz , m. , peterson , b.m . ,
netzer , h. , pogge , r.w . , & vestergaard , m , 2009 , 697 , 160 bottorff , m. , korista , k.t . , shlosman , i. , & blandford , r.d .
1997 , , 479 , 200 bottorff , m. , & ferland , g.j 2000 , 316 , 103 chiang , j. , & murray , n. 1996 , , 466 , 704 collin , s. , kawaguchi , t. , peterson , b. m. , vestergaard , m. 2006 , , 456 , 75 everett , j. e. 2003 , ph .
d. thesis , the university of chicago ferland , g.j , hu , chen , wang , jian - min , baldwin , j.a , porter , r.l , vanhoof , p. , peter , a.m , & williams , r.j.r , 2009 , , 707 , l82 korista , k.t . , et al .
1995 , apjs , 97 , 285 gilbert , k. m. , & peterson b. m. 2003 , , 578 , 123 gltekin , k. , et al .
2009 , , 698 , 198 .
jarvis , m. j. , & mclure , r. j. 2006 , mnras 369 , 182 kaspi , s. , & netzer , h. 1999 , , 524 , 71 kaspi , s. , smith , p. s. , netzer , h. , maoz , d. , jannuzi , b. t. , & giveon , u. 2000 , , 533 , 631 kaspi , s. , maoz , d. , netzer , h. , peterson , b. m. , vestergaard , m. , & jannuzi , b. t. 2005 , , 629 , 61 kaspi , s. , et al .
2007 , , 659 , 997 korista , k. t. , & goad , m. r. 2004 , , 606 , 749 maiolino , r. , et al . 2010 , , 517 , a47 marconi , a. , axon , d. j. , maiolino , r. , nagao , t. , pastorini , g. , pietrini , p. , robinson , a. , & torricelli , g. 2008a , , 678 , 693 ( m08 ) marconi , a. , axon , d. j. , maiolino , r. , nagao , t. , pietrini , p. , risaliti , g. , robinson , a. , & torricelli , g. 2009 , , 698 , l103 ( m09 ) marziani , p. , sulentic , j. w. , dultzin - hacyan , d. , calvani , m. , moles , m. 1996 , , 104 , 47 murray , n. , & chiang , j. 1997 , , 474 , 91 netzer , h. 1990 , 20 .
saas - fee advanced course of the swiss society for astrophysics and astronomy : active galactic nuclei , berlin , springer , p. 57 - 160
netzer , h. 2009 , , 695 , 793 netzer , h. , & trakhtenbrot , b. 2007 , , 654 , 754 onken , c. a. , ferrarese , l. , merritt , d. , peterson , b. m. , pogge , r. w. , vestergaard , m. , & wandel , a. 2004 , , 615 , 645 peterson , b. m. , et al .
, 1999 , , 510 , 668 proga , d. , stone , j. m. , kallman , t. r. 2000 , , 543 , 686 rees , m. , netzer , h. , & ferland , g. j. , 1989 , , 347 , 640 richards , g. t. and vanden berk , d. e. and reichard , t. a. and hall , p. b. and schneider , d. p. and subbarao , m. and thakar , a. r. and york , d. g. 2002 , , 124 , 1 risaliti , g. , elvis , m. , bianchi , s. , matt , g. 2010 , mnras , in press , eprint arxiv:1005.3052 smith , j. e. , robinson , a. , alexander , d. m. , young , s. , axon , d. j. , corbett , e. a. 2004 , mnras , 350 , 140 sulentic , j. , et al . 2007 , , 666 , 757 tremaine , s. , et al .
2002 , , 574 , 740 vestergaard , m. , & peterson , b. m. 2006 , , 641 , 689 vestergaard , m. , wilkes , b. j. , barthel p. d. 2000 , , 538 , l103 woo , j .- h . , et al .
2010 , , 716 , 269 young , s. , axon , d. j. , robinson , a. , hough , j. h. , smith , j. e. 2007 , nature , 450 , 74
|
we present a new analysis of the motion of pressure - confined , broad line region ( blr ) clouds in active galactic nuclei ( agns ) taking into account the combined influence of gravity and radiation pressure .
we calculate cloud orbits under a large range of conditions and include the effect of column density variation as a function of location .
the dependence of radiation pressure force on the level of ionization and the column density are accurately computed .
the main results are : a. the mean cloud locations ( ) and line widths ( fwhms ) are combined in such a way that the simple virial mass estimate , @xmath0 , gives a reasonable approximation to @xmath1 even when radiation pressure force is important .
the reason is that @xmath2 rather than @xmath3 is the main parameter affecting the planar cloud motion .
b. reproducing the mean observed , fwhm and line intensity of and requires at least two different populations of clouds . c. the cloud location is a function of both @xmath4 and @xmath2 . given this
, we suggest a new approximation for which , when inserted into the bh mass equation , results in a new approximation for @xmath1 .
the new expression involves @xmath4 , fwhm and two constants that are obtained from a comparison with available @xmath5 mass estimates .
it deviates only slightly from the old mass estimate at all luminosities .
d. the quality of the present black hole mass estimators depends , critically , on the way the present @xmath5 agn sample ( 29 objects ) represents the overall population , in particular the distribution of .
|
define the norm @xmath6 , of a vector @xmath7 by the number of its nonzero components when @xmath8 , the quantity @xmath9 when @xmath10 , and the maximum absolute value @xmath11 of its components when @xmath12 .
we say that a vector @xmath1 is _
@xmath0-sparse _ if @xmath13 , i.e. , the number of its nonzero components is less than or equal to @xmath0 . in this paper
, we consider the problem of compressive sensing in finding @xmath0-sparse solutions @xmath1 to the linear system @xmath14 via solving the @xmath3-minimization problem : @xmath15 where @xmath16 , @xmath17 , @xmath18 is an @xmath19 matrix , and @xmath20 is the observation data ( @xcite ) .
one of the basic questions about finding @xmath0-sparse solutions to the linear system is under what circumstances the linear system has a unique solution in @xmath21 , the set of all @xmath0-sparse vectors .
( @xcite ) let @xmath22 and @xmath18 be an @xmath19 matrix
. then the following statements are equivalent : * the measurement @xmath23 uniquely determines each @xmath0-sparse vector @xmath24 .
* there is a decoder @xmath25 such that @xmath26 for all @xmath27 . * the only @xmath28-sparse vector @xmath29 that satisfies @xmath30 is the zero vector .
* there exist positive constants @xmath31 and @xmath32 such that @xmath33 the first contribution of this paper is to provide another equivalent statement : * _ there exists @xmath16 such that the decoder @xmath25 defined by @xmath34 satisfies @xmath26 for all @xmath27 . _
the implication from ( v ) to ( ii ) is obvious .
hence it suffices to prove the implication from ( iv ) to ( v ) . for this
, we recall the _ restricted isometry property _ of order @xmath0 for an @xmath19 matrix @xmath18 , i.e. , there exists a positive constant @xmath35 such that @xmath36 the smallest positive constant @xmath37 that satisfies , to be denoted by @xmath38 , is known as the _ restricted isometry constant _ @xcite .
notice that given a matrix @xmath18 that satisfies , its rescaled matrix @xmath39 has the restricted isometry property of order @xmath28 and its restricted isometry constant is given by @xmath40 .
therefore the implication from ( iv ) to ( v ) further reduces to establishing the following result : [ maintheorem1 ] let integers @xmath41 and @xmath0 satisfy @xmath22
. if @xmath18 is an @xmath19 matrix with @xmath42 , then there exists @xmath16 such that any @xmath0-sparse vector @xmath24 can be exactly recovered by solving the @xmath3-minimization problem : @xmath43 the above existence theorem about @xmath3-minimization is established in @xcite and @xcite under a stronger assumption that @xmath44 and @xmath45 respectively , as it is obvious that @xmath46 for any @xmath19 matrix @xmath18 . given integers @xmath47 and @xmath48 satisfying @xmath22 and an @xmath19 matrix @xmath18 ,
define @xmath49\big|\ { \rm any\ vector } \ { \bf x}\in \sigma_s\ { \rm can\ be\ exactly\ recovered}\nonumber\\ & & { \rm by \ solving\ the } \
\ell^q-{\rm minimization\ problem}\ \eqref{section1.eq5 } \big\}.\end{aligned}\ ] ] then @xmath50 whenever @xmath51 by theorem [ maintheorem1 ] .
it is also known that any @xmath0-sparse vector @xmath1 can be exactly recovered by solving the @xmath3-minimization problem whenever @xmath52 @xcite .
this establishes the equivalence among different @xmath53 in recovering @xmath0-sparse solutions via solving the @xmath3-minimization problem . hence in order to recover sparsest vector @xmath24 from the measurement @xmath23 , one may solve the @xmath3-minimization problem for some @xmath16 rather than the @xmath54-minimization problem .
empirical evidence ( @xcite ) strongly indicates that solving the @xmath3-minimization problem with @xmath16 takes much less time than with @xmath8 .
the @xmath54-minimization problem is a combinatorial optimization problem and np - hard to solve @xcite , while on the other hand the @xmath55-minimization is convex and polynomial - time solvable @xcite . to guarantee the equivalence between the @xmath54 and @xmath55-minimization problems in finding the sparse vector @xmath24 from its measurement @xmath23
, one needs to meet various requirements on the matrix @xmath18 , for instance , @xmath56 in @xcite , @xmath57 in @xcite , and @xmath58 in @xcite respectively .
many random matrices with i.d.d .
entries satisfy those requirement to guarantee the equivalence @xcite , but lots of deterministic matrices do not . in particular , matrices @xmath59
are constructed in @xcite for any @xmath60 such that @xmath61 and that it fails on the recovery of some @xmath0-sparse vectors @xmath24 by solving the @xmath55-minimization problem with @xmath18 replaced by @xmath59 .
the @xmath3-minimization problem with @xmath62 is more difficult to solve than the @xmath55-minimization problem due to the nonconvexity and nonsmoothness .
in fact , it is np - hard to find a global minimizer in general but polynomial - time doable to find local minimizer @xcite .
various algorithms have been developed to solve the @xmath3-minimization problem , see for instance @xcite .
for any @xmath35 , define @xmath63 then given any positive number @xmath64 and any @xmath19 matrix @xmath18 with @xmath65 , any vector @xmath27 can be exactly recovered by solving the @xmath3-minimization problem .
for any @xmath16 and sufficiently small @xmath66 , matrices @xmath67 of size @xmath68 are constructed in @xcite such that @xmath69 and there is an @xmath0-sparse vector which can not be recovered exactly by solving the @xmath3-minimization problem with @xmath18 replaced by @xmath70 , where @xmath71 is the unique positive solution to @xmath72 .
the above construction of matrices for which the @xmath3-minimization fails to recover @xmath0-sparse vectors , together with the asymptotic estimate @xmath73 as @xmath74 , gives that @xmath75 where @xmath76 is the unique positive solution of the equation @xmath77 .
the second contribution of this paper is a lower bound estimate for @xmath78 as @xmath79 .
[ maintheorem2 ] let @xmath78 be defined as in .
then @xmath80 denote by @xmath81 the vector which equals to @xmath82 on @xmath83 and vanishes on the complement @xmath84 where @xmath85 .
we say that an @xmath19 matrix @xmath18 has the _ null space property _ of order @xmath0 in @xmath3 if there exists a positive constant @xmath86 such that @xmath87 hold for all @xmath88 satisfying @xmath89 and all sets @xmath83 with its cardinality @xmath90 less than or equal to @xmath0 ( @xcite ) .
the minimal constant @xmath86 in is known as the _ null space constant_. for @xmath91 and @xmath35 , define @xmath92 the third contribution of this paper is the following result about the null space property of an @xmath19 matrix .
[ maintheorem ] let @xmath93 be a positive number in @xmath94 $ ] , integers @xmath95 and @xmath0 satisfy @xmath22 , @xmath18 be an @xmath19 matrix with @xmath42 , and set @xmath96 then @xmath18 has the null space property of order @xmath0 in @xmath3 , and its null space constant is less than or equal to @xmath97 .
the fourth contribution of this paper is to show that one can stably reconstruct a compressive signal from noisy observation under the hypothesis that @xmath98 [ maintheorem3 ] let @xmath95 and @xmath0 be integers with @xmath22 , @xmath18 be an @xmath19 matrix with @xmath42 , @xmath99 , @xmath100 $ ] satisfy with @xmath101 given in , and @xmath102 be the solution of the @xmath3-minimization problem : @xmath103 where @xmath104 is the observation corrupted with unknown noise @xmath5 , @xmath105 and @xmath24 is the object we wish to reconstruct . then @xmath106 and @xmath107 where @xmath108 be the best @xmath0-sparse vector in @xmath109 to approximate @xmath110 , i.e. , @xmath111 and @xmath112 , are positive constants independent on @xmath113 and @xmath0 .
the stable reconstruction of a compressive signal from its noisy observation is established under various assumptions on the restricted isometry constant , for instance , @xmath57 and @xmath114 in @xcite , and @xmath115 and @xmath114 in @xcite , @xmath116 for some @xmath117 and @xmath16 in @xcite , and @xmath118 for some @xmath119 and @xmath16 in @xcite .
as an application of theorem [ maintheorem3 ] , any @xmath0-sparse vector @xmath24 can be exactly recovered by solving the @xmath3-minimization problem when @xmath100 $ ] satisfies .
[ maincorollary ] let @xmath95 and @xmath0 be integers with @xmath22 , @xmath18 be an @xmath19 matrix with @xmath42 , @xmath99 , @xmath100 $ ] satisfy with @xmath101 given in
. then any @xmath0-sparse vector @xmath24 can be exactly recovered by solving the @xmath3-minimization problem .
let @xmath120 where @xmath121| \ a(q , \delta_1)<\delta_1\}$ ] , and let @xmath122 be the solution of the equation @xmath123 if it exists and be equal to one otherwise
. then by theorem [ maintheorem3 ] , any @xmath0-sparse vector @xmath24 can be exactly recovered by solving the @xmath3-minimization problem when @xmath124 , while by @xcite there exists a matrix @xmath18 with @xmath65 and an @xmath0-sparse vector @xmath24 such that the vector @xmath24 can not be exactly recovered by solving the @xmath3-minimization problem when @xmath125 .
the functions @xmath126 and @xmath122 are plotted in figure [ figure1 ] . [ cols="^ " , ]
in this section , we give the proofs of theorems [ maintheorem1 ] , [ maintheorem2 ] , [ maintheorem ] and [ maintheorem3 ] .
define @xmath132^m\\ t_1+\cdots+t_m\ge c\end{array } } \frac{n+a+\sum_{k=1}^m t_k}{(n+b+\sum_{k=1}^m t_k^q)^{1/q}}.\ ] ] by the method of lagrange multiplier , the function @xmath133 attains its maximum on the boundary or on those points @xmath134 whose components are the same , i.e. , @xmath135 as the function @xmath136 has at most one critical point and the second derivative at that critical point ( if it exists ) is positive , we then have @xmath137 applying iteratively we obtain @xmath138 then the conclusion follows by letting @xmath139 in the above estimate . [ mainlemma ]
let @xmath16 , @xmath150 be a positive integer , and let @xmath151 be a finite decreasing sequence of nonnegative numbers with @xmath152 for some @xmath35 .
then @xmath153 where @xmath154 is defined as in .
clearly the conclusion holds when @xmath155 for in this case the left hand side of is equal to 0 .
so we may assume that @xmath156 from now on .
let @xmath157 be an arbitrarily number in @xmath158 . to establish , we consider two cases .
_ case i _ : @xmath159 . in this case , @xmath160 where the first inequality holds because @xmath151 is a decreasing sequence of nonnegative numbers , the second inequality follows from lemma [ inequality.lem ] , and the last equality is true as @xmath161 is a decreasing function on @xmath94 $ ] .
let @xmath163 be the smallest integer in @xmath164 $ ] satisfying @xmath165 .
the existence and uniqueness of such an integer @xmath163 follow from the decreasing property of the sequence @xmath166 , the increasing property of the sequence @xmath167 , and @xmath165 when @xmath168 . then from the decreasing property of the sequence @xmath151 and the definition of the integer
@xmath163 it follows that @xmath169 and @xmath170 which implies that @xmath171 applying the decreasing property of the sequence @xmath172 and using the inequality @xmath173 where @xmath174 and @xmath175 $ ] , we obtain @xmath176 and @xmath177 combining and , recalling and the definition of the integer @xmath163 , and applying lemma [ inequality.lem2 ] with @xmath178 and @xmath179 , we get @xmath180 therefore @xmath181 where the third inequality is valid by and the first inequality follows from the following two inequalities : @xmath182 and @xmath183 [ proof of theorem [ maintheorem ] ] let @xmath88 satisfy @xmath184 and let @xmath185 be a subset of @xmath186 with cardinality @xmath187 less than or equal to @xmath0 .
we partition @xmath188 as @xmath189 , where @xmath190 is the set of indices of the @xmath0 largest components , in absolute value , of @xmath88 in @xmath191 , @xmath192 is the set of indices of the next @xmath0 largest components , in absolute value , of @xmath88 in @xmath193 , and so on . applying the parallelogram identity
, we obtain from the restricted isometry property that latexmath:[\[\label{maintheorem.pf.eq3 } for all @xmath0-sparse vectors @xmath195 whose supports have empty intersection @xcite . combining and and using the restricted isometry property yield @xmath196 which implies that @xmath197 applying lemma [ mainlemma ] with @xmath198 gives
@xmath199 then substituting the above estimate for @xmath200 into the right hand side of the inequality and recalling that @xmath201 is an @xmath0-sparse vector lead to @xmath202 the desired null space property .
we follow the argument in @xcite .
set @xmath203 , and denote by @xmath185 the support of the vector @xmath204 , by @xmath191 the complement of the set @xmath185 in @xmath205 .
then @xmath206 and @xmath207 since @xmath208 similar to the argument used in the proof of theorem [ maintheorem ] , we partition @xmath188 as @xmath189 , where @xmath190 is the set of indices of the @xmath0 largest absolute - value component of @xmath88 in @xmath191 , @xmath192 is the set of indices of the next @xmath0 largest absolute - value components of @xmath88 on @xmath191 , and so on .
then it follows from , and that @xmath209 by the continuity of the function @xmath154 about @xmath35 and the assumption , there exists a positive number @xmath210 such that @xmath211 if @xmath216 , then @xmath217 by , where we set @xmath218 . using and applying lemma [ mainlemma ] with @xmath219
give @xmath220 noting the fact that @xmath213 and then applying , and yield @xmath221 this , together with , leads to the following crucial estimate : @xmath222 combining , , and , we obtain @xmath223 and @xmath224 let @xmath226| \ a ( q , \delta_1)<\delta_1\}.\ ] ] take sufficiently small @xmath60 . note that @xmath227 then for any small @xmath228 and sufficiently small @xmath229 , we have that @xmath230 for all @xmath231 . then applying and yields @xmath232 where the last inequality holds since @xmath233 therefore @xmath234 for any sufficiently small @xmath60 .
take @xmath235 and sufficiently small @xmath60 .
then for @xmath236 and sufficiently small @xmath229 , @xmath237 and @xmath238 by , and . therefore @xmath239 by and . combining and and recalling that @xmath60 is a sufficiently small number chosen arbitrarily ,
we have @xmath240 * acknowledgement * part of this work is done when the author is visiting vanderbilt university and ecole polytechnique federale de lausanne on his sabbatical leave .
the author would like to thank professors akram aldroubi , douglas hardin , michael unser and martin vetterli for the hospitality and fruitful discussion , and professor r. chartrand for his comments on the early version of this manuscript .
|
in this paper , it is proved that every @xmath0-sparse vector @xmath1 can be exactly recovered from the measurement vector @xmath2 via some @xmath3-minimization with @xmath4 , as soon as each @xmath0-sparse vector @xmath1 is uniquely determined by the measurement @xmath5 .
|
for almost thirty years after the discovery of the @xmath0 and its excitations , the so - called november revolution @xcite , heavy hadron spectroscopy was at rest .
paraphrasing lord kelvin famous speech @xcite , by 2003 there were only two clouds on the horizon obscuring _ the beauty and clearness of the dynamical theory_. in the hadron spectra these _ two clouds _ were on the one hand the missing resonance problem , i.e. , all quark models predict a proliferation of excited states which have not been measured , and , on the other hand , the observation by babar of an open - charm meson , the @xmath1 , whose properties were quite different from those predicted by quark potential models . since those peaceful days
an increase on both amount and quality of experimental data has shown quite a different picture , far more involved and convoluted , with the observation of the well - known @xmath2 , the @xmath3 , @xmath4 , @xmath5 , @xmath6 , and several other states .
based on gell - mann conjecture @xcite the hadronic experimental data were classified either as @xmath7 or @xmath8 states according to @xmath9 irreducible representations , nowadays this hypothesis may be in question . therefore , the study of the role played by higher order fock space components in the hadron spectra , allowed by the gell - mann classification , is an interesting issue to address . in this talk
we give an overview of a project for getting a coherent understanding of the low - energy hadron phenomenology from the perspective of constituent quark models @xcite .
we will make emphasis on three different aspects .
firstly , we will address the study of meson spectroscopy in an enlarged hilbert space considering many - quark components .
this seems nowadays unavoidable to understand the experimental data and it builds a bridge towards the description in terms of hadronic degrees of freedom .
secondly , the same scheme used to describe the hadron spectroscopy should be valid for describing the low - energy meson - meson scattering .
in other words , contributions to the hadron spectroscopy arising from many - quark components allowed in the gell - mann scheme , should also be recovered by means of hadron - hadron scattering using a full set of states .
finally , we will review our recent efforts to describe the structure of heavy baryons containing charm or bottom quarks .
many - quark systems present a richer , and therefore more complicated , color structure than standard hadrons built with a quark - antiquark pair or three quarks .
while the color wave function for ordinary mesons and baryons leads to a single vector , in the case of four quark states there are different vectors driving to a singlet color state out of either colorless or colored quark - antiquark two - body components .
thus , when dealing with four quark states a basic important question is whether we are in front of a single colorless meson meson molecule or a compact state , defined as a system with two - body colored components . whereas the first structure would be natural in the naive quark model , the second one would open a new area in hadron spectroscopy .
let us start by defining a _ physical channel _ as a four - quark state made of two quark - antiquark color singlets . in ref .
@xcite the formalism to evaluate the probability of the different physical channels for an arbitrary four quark wave function was derived . for this purpose
any hidden
color vector of a four quark color basis , i.e. , vectors with non singlet internal color couplings , was expanded in terms of singlet singlet color vectors .
such a procedure gives rise to a wave function expanded in terms of nonorthogonal vectors belonging to different orthogonal basis , and therefore the determination of the probability of physical channels becomes cumbersome .
such nonorthogonal expansions are common in other fields of science , like chemical physics @xcite , however they have not been properly discussed until now in the context of the multiquark phenomenology , i.e. , _ color chemistry_. we have derived the two hermitian operators that are well defined projectors on the physical singlet
singlet color states @xmath10 and @xmath11 . using these operators the probabilities for finding singlet singlet color components for an arbitrary four - quark state @xmath12
are given by , @xmath13})&=&\frac{1}{2(1-|\,_c\left\langle 11 \mid 1'1 ' \right\rangle_c|^2 ) } \left [ \left\langle\psi\mid p\hat q \mid\psi\right\rangle + \left\langle\psi\mid \hat q p \mid\psi\right\rangle\right ] \nonumber \\ p^{\mid\psi\rangle}({[1'1']})&=&\frac{1}{2(1-|\,_c\left\langle 11 \mid 1'1 ' \right\rangle_c|^2 ) } \left [ \left\langle\psi\mid \hat p q \mid\psi\right\rangle + \left\langle\psi\mid q \hat p \mid\psi\right\rangle\right ] \
, , \end{aligned}\ ] ] where @xmath14 and @xmath15 are the projectors over the different color vectors , @xmath16 and @xmath17 respectively .
the stability of a four quark state relies on @xmath18 , the energy difference between its mass and that of the lowest two - meson threshold .
it is important to emphasize the relevance of comparing four quark energies with respect to thresholds obtained using the same quark - quark interaction as in the four quark case . only in doing
so one may guarantee that no spurious bound states are obtained .
@xmath19 indicates that all fall - apart decay channels are forbidden , and therefore one has a proper bound state .
@xmath20 will indicate that the four quark solution corresponds to an unbound threshold ( two free mesons ) .
as we will discuss below , between these two limits one may find molecular states .
.four quark state properties for selected quantum numbers .
the notation @xmath21 stands for mesons @xmath22 and @xmath23 with relative orbital angular momentum @xmath24 .
@xmath25 $ ] and @xmath26 $ ] stand for the probability of the color components in different basis . [ cols="^,^,^,^,^,^ " , ]
to summarize this talk , we have reviewed some recent results of hadron spectroscopy and hadron - hadron interaction from the point of view of the constituent quark model .
the constituent quark model is a powerful tool at our disposal to face the new issues appearing in an almost daily basis in the heavy quark sector .
unfortunately , the downsides of such robust tools lie in the risk of misusing them .
therefore , we have tried to impress upon the reader the range of validity and proper formalisms of the constituent quark model .
similarly , since there are many different models in the literature , we have tried to present conclusions that are independent of the particularities of the model chosen .
we have seen how the enlargement of the hilbert space when increasing energy , that was seen to be necessary to describe the nucleon - nucleon phenomenology above the pion production threshold , seems to be necessary to understand current problems of hadron spectroscopy .
we have also tried to emphasize that spectroscopy and interaction can , and must , be understood within the same scheme when dealing with quark models .
any other alternative becomes irrelevant from the point of view of learning about properties of qcd .
it is expected that in the coming years better - quality data on all sectors discussed in this talk will become available and our results can be used to analyze these upcoming data in a model independent way .
this work has been partially funded by the spanish ministerio de educacin y ciencia and eu feder under contract no .
fpa2010 - 21750 , and by the spanish consolider - ingenio 2010 program cpan ( csd2007 - 00042 ) .
|
we present recent results of hadron spectroscopy and hadron - hadron interaction from the perspective of constituent quark models .
we pay special attention to the role played by higher order fock space components in the hadron spectra and the connection of this extension with the hadron - hadron interaction .
the main goal of our description is to obtain a coherent understanding of the low - energy hadron phenomenology without enforcing any particular model , to constrain its characteristics and learn about low - energy realization of the theory .
heavy hadrons , constituent quark model , many - quark systems
|
the lattice simulation of quantum chromodynamic equation of state ( eos ) predicts that the properties of nuclear matter at extreme densities and/or temperatures are governed by the partonic degrees of freedom @xcite .
a series of experiments have been performed @xcite and planned @xcite to produce such a partonic state of matter , called quark gluon plasma ( qgp ) by colliding nuclei at ultra - relativistic energies .
rigorous experimental and theoretical efforts are on to create and detect such a novel state of matter @xcite .
various signals have been proposed for the detection of qgp - the pros and cons of these signals are matter of intense debate .
the study of the ratio , @xmath4 is one such currently debated issue .
@xmath5 is measured experimentally @xcite as a function of centre of mass energy ( @xmath2 ) .
it is observed that the @xmath5 increases with @xmath2 and then decreases beyond a certain value of @xmath2 giving rise to a horn like structure , whereas the ratio , @xmath6 increases faster at lower @xmath2 and tend to saturate at higher @xmath2 .
explanation of this structure has ignited intense theoretical activities @xcite .
several authors have attempted to reproduce the @xmath1 ratio using different approaches . while the authors in @xcite use a hadronic kinetic model , in ref .
@xcite high mass unknown hadronic resonances have been introduced through hagedorn formula to describe the data . in ref .
@xcite a transition from a baryon dominated system at low energy to a meson dominated system at higher energy has been assumed to reproduce the ratio @xmath1 . the release of color degrees of freedom is assumed in @xcite beyond a threshold in @xmath0 ( resulting in large pion productions ) or the production of larger number of pions than kaons from higher mass resonance decays has also been employed @xcite to explain the data . in the present work
we employ a microscopic model for the productions and evolution of strange quarks and hadrons depending on the collision energy . here
we examine whether the @xmath1 experimental data can differentiate between the following two initial conditions or two scenarios - after the collisions the system is formed in : ( i ) the hadronic phase for all @xmath0 or ( ii ) the partonic phase beyond a certain threshold in @xmath0 . other possibilities like formation of strangeness in complete thermal equilibrium and evolution in space time ( iii ) without and
( iv ) with secondary productions of quarks and hadrons have been considered .
( v ) results for an ideal case of zero strangeness in the initial state has also been presented . in the context of strangeness enhancement as signal of qgp formation ,
similar approach _
i.e. _ the assumption of an initial state where non - strange sectors are in equilibrium but the strange degrees of freedom are out of equilibrium ( having density much below their equilibrium values ) were considered in the 1980 s .
the strangeness production in a deconfined ( partonic ) phase is enhanced compared to the their production in the confined ( hadronic ) phase primarily because even the lightest strange hadrons , the kaons are much heavier than the strange quarks .
moreover , the strange quark has more degrees of freedom ( six ) in a deconfined matter compared to kaons .
therefore , the strangeness production during the space time evolution of the system for partonic initial state will be enhanced compared to the hadronic initial state , hence the enhanced production of strangeness could be an efficient signal for deconfinement @xcite .
in contrast to these studies gazdzicki and gorenstein @xcite within the ambit of statistical model considered the strangeness production where both the strange and non - strange degrees of freedom are in thermal equilibrium and the production of strangeness during the expansion stage is ignored . in the present work we would like to compare the results on kaon to pion ratio from these two contrasting scenarios .
we assume that the non - strange quarks and hadrons are in complete thermal ( both kinetic and chemical ) equilibrium and the strange quarks and strange hadrons are away from chemical equilibrium .
therefore , the evolution of the strange sector of the system is governed by the interactions between the equilibrium and non - equilibrium degrees of freedom .
the momentum integrated boltzmann equation provides a possible framework for such studies .
similar approach has been used to study the sequential freeze - out of elementary particles in the early universe @xcite . for the strangeness productions in the partonic phase we consider the processes of gluon fusion and light quarks annihilation .
for the production of @xmath7 and @xmath8 an exhaustive set of reactions involving thermal baryons and mesons have been considered .
the time evolution of the densities are governed by the boltzmann equation .
the paper is organized as follows . in the next section the rate of strangeness productions in the partonic and hadronic
phases are discussed .
the space time evolution of the system is presented in section iii .
results are presented in section iv and finally section v is devoted to summary and conclusion .
the productions of @xmath9 and @xmath10 in the qgp and the @xmath7 and @xmath8 in the hadronic system are discussed below .
the two main processes for the strange quark productions are gluon fusion ( @xmath11 and quark(@xmath12)-antiquark ( @xmath13 ) annihilations ( @xmath14 ) .
the cross sections in the lowest order qcd is given by @xcite : _
q|q s|s= ( 1 + ) w(s ) and _ gg s|s= [ g(s)tanh^-1 w(s)-+w(s ) ] where @xmath15 is the mass of strange quark , @xmath16 , is the square of the centre of mass energy of the colliding particles , @xmath17 are the four momenta of incoming particles , @xmath18 , @xmath19=@xmath20 and @xmath21 is the strong coupling constant that depends on temperature @xcite .
quark from @xmath22 and @xmath23 with temperature . ] the rate of @xmath24 ) and @xmath25 ) productions in the hadronic phase can be categorized as due to ( a ) meson - meson ( @xmath26 ) , ( b ) meson - baryon ( @xmath27 ) and ( c ) baryon- baryon ( @xmath28 ) interactions . in the present paper
we quote only the main results for kaon productions in the hadronic matter and refer to @xcite for details .
\(a ) for the first category @xmath29 , we considered the following channels : @xmath30 , @xmath31 , @xmath32 and @xmath33 .
the invariant amplitude for these processes have been calculated from the following lagrangians @xcite . for the @xmath34 vertex
the interaction is given by , @xmath35\ ] ] similarly for the @xmath36 vertex the interaction is , @xmath37\rho^{\mu}\ ] ] the isospin averaged cross section ( @xmath38 ) for @xmath39 ( i.e. , @xmath30 , @xmath31 and @xmath32 , @xmath33 ) is evaluated by using , @xmath40 where @xmath41 and @xmath42 are the three - momenta of the meson and kaons in the centre - of - mass frame , @xmath43 is the cosine of the angle between @xmath41 and @xmath42 .
@xmath44 is the iso - spin averaged squared invariant amplitude .
\(b ) for meson baryon interactions the dominant channels are : @xmath45 , @xmath46 , @xmath47 and @xmath48 .
the isospin averaged cross section is given by @xcite : @xmath49 @xmath50 , @xmath51 and @xmath52 are the spin , width and mass of the resonances , ( @xmath53 ) is the polarization states of the incident particles , @xmath54 is the centre of mass momentum of the initial state .
@xmath55 and @xmath56 are the branching ratios of initial and final state channels respectively .
the index @xmath57 runs over all the resonance states . for interactions @xmath58 , @xmath59 we have considered @xmath60 and @xmath61 as the intermediate states .
values of various hadronic masses and decay widths are taken from particle data book @xcite .
\(c ) for the last category of reactions _
i.e. _ for baryon baryon interactions @xcite the dominant processes are : @xmath62 , @xmath63 , @xmath64 , @xmath65 , @xmath66 and @xmath67 .
the isospin averaged cross section of kaon production from the process like @xmath68 is given by @xcite @xmath69 pion is the intermediate particle for the above interaction , @xmath70 is the mass of @xmath71 , @xmath72 is the total energy in the centre of mass system of pion and @xmath73 , @xmath74 .
@xmath75 where @xmath76 , @xmath77 are the momenta and @xmath78 , @xmath79 are the energies of @xmath80 and @xmath81 respectively .
we take @xmath82 = 1 and to constrain the finite size of the interaction vertices we use the form factor @xmath83 .
@xmath84 is the isospin averaged cross section of @xmath85 .
cross sections for the processes : @xmath86 and @xmath87 have been taken from @xcite .
the cross section of other reactions _ e.g. _
@xmath65 , @xmath66 and @xmath67 have been taken from @xcite . in a baryon rich medium , @xmath8 gets absorbed due to its interaction with the baryons .
the reactions @xmath88 , @xmath89 , @xmath90 , @xmath91 , @xmath92 have been considered for @xmath8 absorption @xcite in the nuclear matter .
and @xmath8 production rates with temperature at center of mass energy=7.6 gev and 200 gev . ]
@xmath93 , the number of @xmath9 quarks produced per unit time per unit volume at temperature @xmath94 and baryonic chemical potential @xmath95 is given by @xmath96 where @xmath17 s are the momenta of the incoming particles and @xmath97 s are the respective phase space distribution functions ( through which the dependence on @xmath94 and @xmath95 are introduced ) , @xmath98 is the relative velocity of the incoming particles and @xmath99 is the production cross sections for the reactions .
the same equation can be used for kaon production by appropriate replacements of phase space factor and cross sections .
the possibility of formation of a fully equilibrated system in high energy nuclear collisions is still a fiercely debated issue because of the finite size and life time of system . in the present work
we assume that the strange quarks or the strange hadrons ( depending on the value of @xmath0 ) produced as a result of the collisions are not in chemical equilibrium .
the time evolution of the strangeness in either qgp or hadronic phase is governed by the momentum integrated boltzmann equation .
we have assumed that the initial density of strange quarks or kaons ( depending on the initial conditions ( i ) or ( ii ) ) is @xmath100 away from the corresponding equilibrium density .
we will comment on the amount of deviations from chemical equilibrium later .
the momentum integrated boltzmann equation has been applied to study the freeze - out of elementary particles during the thermal expansion of the early universe @xcite . in the present work we follow similar procedure to study the evolution of the strange quarks and anti - quarks in the qgp phase or kaons in the hadronic phase .
the coupled equations describing the evolution of @xmath57 ( particle ) and @xmath101 ( anti particle ) with proper time @xmath102 is given by : @xmath103 - \frac{n_{i } } { \tau } \nonumber \\ \frac{dn_{j}}{d\tau } & = & r_{j}(\mu_b , t ) [ 1 - \frac{n_j n_{i}}{n_{j}^{eq}{n_{i}^{eq } } } ] - \frac{n_{j } } { \tau } .
\label{eqn_evol1}\end{aligned}\ ] ] where , @xmath104 ( @xmath105 ) and @xmath106 ( @xmath107 ) are the non - equilibrium and equilibrium densities of @xmath57 ( @xmath101 ) type of particles respectively .
@xmath108 is the rate of production of particle @xmath57 at temperature @xmath94 and chemical potential @xmath95 , @xmath102 is the proper time . first term on the right hand side of eq .
[ eqn_evol1 ] is the production term and the second term represents the dilution of the system due to expansion .
the variation of temperature and the baryonic chemical potential with time is governed by the hydrodynamic equations ( next section ) .
the indices @xmath57 and @xmath101 in eq.[eqn_evol1 ] are replaced by @xmath109 quark in the qgp phase and by @xmath110 in the hadron phase respectively . for higher colliding energies i.e. , @xmath111 @xmath112 8.76 gev an initial partonic phase is assumed .
the hadrons are formed at a transition temperature , @xmath113 mev through a first order phase transition from qgp to hadrons .
the fraction of the qgp phase in the mixed phase at a proper time @xmath102 is given by @xcite : @xmath114 where @xmath115 ( @xmath116 ) is the time at which the qgp ( mixed ) phase ends , @xmath117 is the ratio of statistical degeneracy in qgp to hadronic phase .
the evolution of the kaons are governed by @xcite : @xmath118 - \frac{n_{\bar{s } } } % { \tau } \nonumber \\ % \frac{dn_{s}}{d\tau } & = & r_{s}(\mu , t_c ) % [ 1 - \frac{n_s n_{\bar{s}}}{n_{s}^{eq}{n_{\bar{s}}^{eq } } } ] - \frac{n_{s } } % { \tau } \nonumber \\ \frac{dn_{k^+}}{d\tau } & = & r_{k^+}(\mu_b , t_c ) [ 1 - \frac{n_{k^+}n_{k^-}}{n_{k^+}^{eq}n_{k^-}^{eq } } ] - \frac{n_{k^+ } } { \tau}+ \nonumber \\ & & \frac { 1}{f_h } \frac { d f_h}{d \tau } \left(\delta n_{\bar{s}}-n_{k^+}\right ) \nonumber\\ \frac{dn_{k^-}}{d\tau } & = & r_{k^-}(\mu_b , t_c ) [ 1 - \frac{n_{k^+}n_{k^-}}{n_{k^+}^{eq}n_{k^-}^{eq } } ] - \frac{n_{k^- } } { \tau } + \nonumber \\ & & \frac { 1}{f_h } \frac { df_h}{d \tau } \left(\delta n_{s}-n_{k^-}\right ) \label{eqmix}\end{aligned}\ ] ] similar equation exist for the evolution of @xmath9 and @xmath10 quarks in the mixed phase ( see @xcite for details ) . in the above equations @xmath119 represents the fraction of hadrons in the mixed phase at time @xmath102 .
the last term stands for the hadronization of @xmath120 quarks to @xmath121 @xcite . here
@xmath122 is a parameter which indicates the fraction of @xmath123 quarks hadronizing to @xmath124 .
@xmath125 indicates the formation of @xmath7 and @xmath126 in the mixed phase because half of the @xmath10 form @xmath7 and the rest hadronize to @xmath126 .
the partonic / hadronic system produced in nuclear collisions evolves in space - time .
the space - time evolution of the bulk matter is governed by the relativistic hydrodynamic equation : _
t^=0 [ hydro1 ] with boost invariance along the longitudinal direction @xcite . in the above equation @xmath127 , is the energy momentum tensor for ideal fluid , @xmath128 is the energy density , @xmath41 is the pressure and @xmath129 is the hydrodynamic four velocity .
the net baryon number conservation in the system is governed by : _ ( n_bu^)=0 [ hydro2 ] where @xmath130 is the net baryon density .
[ hydro1 ] and [ hydro2 ] have been solved ( see @xcite for details ) to obtain the variation of temperature and baryon density with proper time .
the initial temperatures corresponding to different @xmath0 are taken from table i. the baryonic chemical potential at freeze - out are taken from the parametrization of @xmath95 with @xmath131 @xcite(see also @xcite ) and the baryonic chemical potential at the initial state is obtained from the net baryon number conservation equation .
the initial temperatures of the systems formed after nuclear collisions have been evaluated from the measured hadronic multiplicity , @xmath132 by using the following relation : @xmath133 where @xmath134 denotes the riemann zeta function , @xmath135 is the transverse radius [ @xmath136 @xmath137 , @xmath138 is the number of participant nucleons ] of the colliding system , @xmath139 is the initial time and @xmath140 is the statistical degeneracy . initial temperatures for different @xmath2 are tabulated in table i. lccccr @xmath0 ( gev)&@xmath141 ( gev ) & @xmath142 ( gev ) + 3.32 & 0.115 & - + 3.83 & 0.128 & - + 4.8 & 0.150 & - + 6.27 & 0.160 & - + 7.6 & 0.187 & - + 8.76 & 0.210 & 0.190 + 12.3 & 0.225 & 0.190 + 17.3 & 0.25 & 0.190 + 62.4 & 0.3 & 0.190 + 130 & 0.35 & 0.190 + 200 & 0.40 & 0.190 + [ tableii ]
the variation of the number of strange anti - quarks produced per unit volume per unit time with temperature has been displayed in fig .
[ fig1 ] for a baryonic chemical potential @xmath143 mev .
it is observed that the process of gluon fusion dominates over the @xmath144 annihilation for the entire temperature range under consideration , primarily because at high @xmath145 the number of anti - quarks is suppressed . in fig .
[ fig2 ] , the production rate of @xmath7 from the @xmath146 type of reactions has been depicted for @xmath147 gev .
the production rate from pion annihilation dominates over the reactions that involves @xmath148 mesons , because the thermal phase space factor of @xmath148 is small due its larger mass compared to pions and smaller production cross section .
results for interactions involving mesons and baryons are displayed in fig .
it is observed that the interactions involving pions and nucleons in the initial channels dominate over that which has a @xmath148 meson in the incident channel .
in fact , contributions from the reactions @xmath46 has negligible effect on the total productions from the meson baryon interactions .
the kaon production from baryon - baryon interaction is displayed in fig .
the contributions from @xmath149 dominates over the contributions from @xmath62 and @xmath150 for the temperature range @xmath151 to 180 mev . in fig.[fig5 ]
the rates of @xmath7 productions from meson - meson interactions has been compared with those involving baryons _
i.e. _ with meson - baryon and baryon - baryon interactions for different @xmath0 ( different @xmath95 )
. the results clearly indicate the dominant role of baryons at lower collision energies which diminishes with increasing @xmath0 . at low temperature
the baryonic contribution is more than the mesonic one for lower beam energy .
rate of productions ( from mb+bb interactions ) at @xmath0=4.8 gev is more compared to the rates at @xmath0=7.6 and 200 gev , since @xmath95 at @xmath0=4.8 gev is more ( see table - ii ) .
production rate from pure mesonic interactions does not depend on @xmath95 hence same for all .
it is quite clear from the results displayed in fig.[fig5 ] that more the baryonic chemical potential ( lower the centre of mass energy ) , more is the rate from bb and mb interactions compared to mm interactions . for a system having lower chemical potential ( higher centre of mass energy )
the rate of production from mesonic interactions is dominant .
cc center of mass energy@xmath152 & chem .
potential ( @xmath153 ) + ( in a gev ) & ( mev ) + 3.32&595 + 3.83 & 568 + 4.8 & 542 + 6.27&478 + 7.6 & 432 + 8.76 & 398 + 12.3 & 321 + 17.3 & 253 + 62.4&86 + 130 & 43 + 200 & 28 + [ tablei ] a comparison is made between rates of kaon productions from meson meson interactions ( mm ) and meson - baryon ( mb ) plus baryon - baryon ( bb ) interactions for @xmath147 gev . at this
energy baryons and mesons are equally important as shown in fig .
[ fig6 ] . in fig .
[ fig7 ] the net rates of productions for @xmath7 and @xmath8 have been depicted for @xmath154 gev ( left panel ) and 200 gev ( right panel ) . at @xmath147 gev the production of @xmath7 dominates over @xmath8 for the entire temperature range . however , for large @xmath0 ( low @xmath95 ) the productions of @xmath7 and @xmath8 are similar .
the strong absorption of the @xmath8 by nucleons in a baryon rich medium resulting in lower production yield of @xmath8 compared to @xmath7 .
this may be contrasted with the experimental findings of brahms experiment @xcite where it is observed that at mid - rapidity ( small @xmath95 due to nuclear transparency at rhic energy ) the @xmath7 and @xmath8 yields are similar but at large rapidity ( large @xmath95 ) @xmath8 yield is smaller than @xmath7 due to large @xmath8 nucleon absorption . in fig .
[ fig8 ] the variations of @xmath5 with @xmath0 are depicted .
the experimental data on @xmath5 is well reproduced if a partonic initial phase ( scenario - ii ) is assumed beyond @xmath0=8.7 gev .
a `` mindless '' extrapolation of hadronic initial state ( scenario - i)for all the @xmath0 up to rhic energy show an increasing trend in disagreement with the experimental data at higher @xmath0 . in both the scenarios ,
i and ii , the curves at higher @xmath0 ( rhic energies ) becomes flatter .
that is because at higher energies the @xmath7 productions in the hadronic phase are dominated by mesonic interactions and the production rates from mesons are same for all @xmath0 for a given temperature range .
but at lower energies the rates of kaon productions are dominated by the effective interactions among the baryonic degrees of freedom . the composition of matter formed in heavy ion collision changes from a matter dominated by baryons to a matter dominated by mesons with the increase in colliding energy . the @xmath95 changes from 86 mev to 28 mev as @xmath0 varies from 62.4 gev to 200 gev ( table ii ) .
the change in the @xmath7 production in the hadronic phase due to the change in @xmath95 mentioned above is marginal - resulting in the flatness in @xmath5 at higher energies .
the decrease of the value of the @xmath5 beyond @xmath0=7.6 gev showing horn like structure happens only when an initial partonic phase is considered .
such a non - monotonic behaviour of @xmath5 can be understood as due to larger entropy productions from the release of large colour degrees of freedom ( resulting in more pions yield ) compared to strangeness beyond energy 7.6 gev . in fig .
[ fig9 ] , the variations of @xmath155 with @xmath0 is displayed .
@xmath155 has a lower value compared to @xmath5 at lower energies since @xmath8 get absorbed in the baryonic medium . at higher energies
@xmath8 is closer to @xmath7 because production of @xmath7 and @xmath8 is similar in baryon free medium , which may be realized at higher collision energies . /@xmath156
ratio for different centre of mass energies .
i represents for pure initial hadronic scenario for all centre of mass energies .
scenario - ii represents for the calculation with hadronic initial conditions for low @xmath2 and partonic initial conditions for higher @xmath2 .
see text for details . ]
/@xmath157 ratio for different centre of mass energies .
scenario - i represents for pure initial hadronic scenario for all centre of mass energies .
scenario - ii represents for the calculation with hadronic initial conditions for low @xmath2 and partonic initial conditions for higher @xmath2 .
see text for details . ]
/@xmath156 ratio for different centre of mass energies .
scenario - iii assumes complete equilibrium of strange quarks and hadrons .
the production through secondary processes have been ignored .
scenario iv is same as iii with secondary productions processes are on and scenario v represents zero strangeness initially but
secondary productions are switched on . ] in fig .
[ fig10 ] the @xmath5 is depicted as a function of @xmath131 for other scenarios ( iii , iv and v ) . on the one hand when the strange quarks and kaons are formed in complete equilibrium but
their secondary productions are neglected during the evolution ( scenario iii ) then the data is well reproduced . on the other hand , in the scenario ( iv ) when the system is formed in equilibrium
( as in iii ) but the productions of strange quarks and kaons are switched on through secondary processes then the data is slightly overestimated at high @xmath131 . however , we have seen that the data is also reproduced well in the scenario ii as discussed above .
this indicate that the deficiency of strangeness below its equilibrium value as considered in ( ii ) is compensated by the secondary productions . in scenario
v we assume that vanishing initial strangeness and observed that the production of strangeness throughout the evolution is not sufficient to reproduce the data . the productions from secondary processes are small but not entirely negligible ( v )
. /@xmath157 . ] in fig .
[ fig11 ] the @xmath155 has been displayed as a function of @xmath131 .
a trend similar to the results shown in fig .
[ fig10 ] is observed .
the data is overestimated for the intermediate @xmath131 in the scenario iv , reproduced well in scenario iii and underestimated for the scenario v.
the evolution of the strangeness in the system formed in nuclear collisions at relativistic energies have been studied within the framework of momentum integrated boltzmann equation .
the boltzmann equation has been used to study the evolution of @xmath9 and @xmath10 in the partonic phase and @xmath8 and @xmath7 in the hadronic phase .
the calculation has been done for different centre of mass energies ranging from ags to rhic .
we get a non - monotonic variation of @xmath1 with @xmath2 when an initial partonic phase is assumed for @xmath158 gev and beyond .
a monotonic rise of @xmath1 is observed when a pure hadronic scenario is assumed for all centre of mass energies .
the @xmath3 data is unable to differentiate between the two initial conditions mentioned before .
some comments on the values of the initial parameter are in order at this point .
we have seen that a @xmath159 variation in the initial temperature does not change the results drastically .
we have assumed that the initial density of strange quarks or kaons depending on the scenario ( i ) or ( ii ) is about @xmath100 away from the corresponding equilibrium density .
results from a scenario where strange quarks or kaons are formed in complete equilibrium and the production is ignored during the evolution then the data is well reproduced ( scenario iii ) . if the the strangeness is produced in equilibrium and the production is included during the expansion stage then the data is overestimated .
however , if the system is formed with zero strangeness then the theoretical results underestimate the data substantially .
this indicate that the production of strangeness during the expansion of the system is small but not entirely negligible .
the deficiency assumed in scenario ii is compensated by the production during evolution .
i. arsene _
( brahms collaboration ) , nucl .
a * 757 * , 1 ( 2005 ) ; b. b. back _ et al . _
( phobos collaboration ) , nucl .
phys . a * 757 * , 28 ( 2005 ) ; j. adams _ et al . _ ( star collaboration ) , nucl .
phys . a * 757 * , 102 ( 2005 ) ; k. adcox _
( phenix collaboration ) , nucl .
phys . a * 757 * , 184,(2005 ) . c. w. fabjan ( for the alice collaboration ) , j. phys .
g * 35 * , 104038 ( 2008 ) ; d. denterria ( for the cms collaboration ) , j. phys.g * 35 * , 104039 ( 2008 ) ; n. grau ( for the atlas collaboration ) , j. phys.g * 35 * , 104040 ( 2008 ) .
j. alam , s. chattopadhyay , t. nayak , b. sinha and y .
p. viyogi ( eds ) , j. phys .
* 35 * ( 2008 ) ( proc .
quark matter 2008 ) ; p. stankus , d. silvermyr , s. sorensen and v. greene ( eds . ) , nucl .
a * 830 * ( proc .
quark matter ) ( 2009 ) .
|
a microscopic approach has been employed to study the kaon productions in heavy ion collisions .
the momentum integrated boltzmann equation has been used to study the evolution of strangeness in the system formed in heavy ion collision at relativistic energies .
the kaon productions have been calculated for different centre of mass energies ( @xmath0 ) ranging from ags to rhic .
the results have been compared with available experimental data .
we obtain a non - monotonic horn like structure for @xmath1 when plotted with @xmath0 with the assumption of an initial partonic phase beyond a certain threshold in @xmath2 .
however , a monotonic rise of @xmath1 is observed when a hadronic initial state is assumed for all @xmath0 .
experimental values of @xmath3 are also reproduced within the ambit of the same formalism .
results from scenarios where the strange quarks and hadrons are formed in equilibrium and evolves with and without secondary productions have also been presented . .
|
a knot can be embedded in many different ways in @xmath8-space , smooth or piecewise linear .
polygonal knots are those which consist of finite line segments , called _ sticks _ , attached end - to - end .
a _ lattice knot _ is a polygonal knot in the cubic lattice @xmath9 . for further studies on lattice knots the readers
are referred to @xcite .
a quantity that we may naturally be interested on lattice knots is the minimum length necessary to realize a knot as a lattice knot .
edge _ is a line segment of unit length joining two nearby lattice points in @xmath10 .
obviously a stick with length @xmath11 consists of @xmath11 edges .
the minimum number of edges necessary to realize a knot @xmath1 as a lattice knot is called the _ minimum lattice length _ , denoted by @xmath0 .
diao @xcite introduced this term ( he used minimal edge number " instead ) , and proved that the minimum lattice length of the trefoil knot @xmath12 is 24 and all the other nontrivial knots need more than 24 edges . ishihara and shimokawa
( see @xcite ) proved that the minimum lattice length of @xmath13 and @xmath14 are @xmath15 and @xmath16 , respectively .
lattice knot presentations with minimum lattice length of the knots @xmath12 , @xmath13 , and @xmath14 are depicted in figure [ fig1 ] .
also many numerical estimations of the minimum lattice length for various knots are addressed in @xcite . ,
@xmath13 and @xmath14 with minimum lattice length ] in the paper we are interested in finding upper bounds for the minimum lattice length of a given nontrivial knot @xmath1 in terms of the minimum crossing number @xmath2 .
et al _ @xcite established an o@xmath17 upper bound for the minimum lattice length as follows : @xmath18 recently the authors @xcite obtained another inequality for nontrivial knots except the trefoil knot @xmath12 : @xmath19 in this paper we improve this upper bound .
[ thm : len ] let @xmath1 be any nontrivial knot . @xmath20 moreover if @xmath1 is a non - alternating prime knot , @xmath21 an essential question in physical knot theory concerns the minimum amount of rope ( of unit thickness ) needed to tie a given knot .
we measure the _ ropelength _ of a knot @xmath1 as the quotient of its length by its thickness where thickness is the radius of the largest embedded normal tube around the knot .
we define the minimum ropelength @xmath4 to be the minimum ropelength of all embeddings of @xmath1 .
many results about finding lower bounds for the ropelength of @xmath1 as a function of crossing number can be found in @xcite . in this paper
we are interested in a converse problem : finding upper bounds for the minimum ropelength of @xmath1 . indeed , the minimum ropelength of a knot is less than or equal to twice the minimum lattice length of the knot . in the same paper ,
diao _ et al _ also established an o@xmath17 upper bound for the minimum ropelength which is twice the upper bound for the minimum lattice length they found : @xmath22 also cantarella _ et al _ @xcite obtained another inequality : @xmath23 we obtained an improved inequality from theorem [ thm : len ] .
[ thm : rop ] let @xmath1 be any nontrivial knot .
@xmath24c(k)^2 + ( + 6 ) c(k ) + 2 + @xmath6c(k)^2 + ( + 13 ) c(k ) + + @xmath7 moreover if @xmath1 is a non - alternating prime knot , @xmath25c(k)^2 + c(k ) - @xmath6c(k)^2 + ( + 8) c(k ) - @xmath7 note that the first inequality in theorem [ thm : rop ] guarantees the following inequality which slightly improves the cantarella s bound : @xmath51.5 c(k)^2 + 9.15 c(k ) + [email protected] c(k)^2 + 14.58 c(k ) + 16.90@xmath7
a _ grid diagram _ of a knot @xmath1 is a knot diagram of vertical strands and the same number of horizontal strands with the properties that at every crossing the vertical strand crosses over the horizontal strand and no two horizontal strands are co - linear and no two vertical strands are co - linear .
it is known that every knot admits a grid diagram @xcite .
the minimum number of vertical strands in all grid diagrams of @xmath1 is called the _ grid index _ of k , denoted by @xmath26 . an _ arc presentation _ of a knot @xmath1
is an embedding of @xmath1 in finitely many pages of an open - book decomposition so that each of the pages meets @xmath1 in a single simple arc .
note that this open - book decomposition has open half - planes as pages and the standard @xmath27-axis as the binding axis . and the _ arc index _
@xmath28 is the minimum number of pages among all possible arc presentations of @xmath1 .
since a grid diagrams is a way of depicting an arc presentation @xcite , the arc index equals the grid index , i.e. @xmath29 . figure [ fig2 ] presents an arc presentation , a grid diagram , and how they are related in an example of the trefoil knot .
[ h ] in this paper we first find upper bounds for the minimum lattice length and the minimum ropelength of knots in terms of the grid index .
the following proposition has a key role to convert these bounds to upper bounds in terms of the crossing number .
[ prop : grid ] let @xmath1 be any nontrivial knot . then @xmath30 .
moreover if @xmath1 is a non - alternating prime knot , then @xmath31 .
bae and park @xcite established an upper bound on arc index in terms of the crossing number . in their paper , corollary @xmath32 and theorem @xmath33 provide that @xmath34 , and moreover @xmath35 if @xmath1 is a non - alternating prime knot . later jin and park @xcite improved the second part of bae and park s theorem .
they proved @xmath36 for a non - alternating prime knot @xmath1 .
the equality @xmath29 guarantees the proposition .
a stick in @xmath10 parallel to the @xmath37-axis is called an _
@xmath37-stick _ and an edge parallel to the @xmath37-axis is called an _ @xmath37-edge_. the plane with the equation @xmath38 for some integer @xmath39 is called an _ @xmath37-level @xmath39_.
the related notations concerning the @xmath40 and @xmath27-coordinates will be defined in the same manner as the @xmath37-coordinate .
note that each @xmath40-stick or @xmath27-stick lies on some @xmath37-level .
in this section we prove theorem [ thm : len ] by actually constructing a lattice knot and counting the number of its edges .
let @xmath1 be a nontrivial knot with the grid index @xmath41 .
the proof follows four steps : settling a grid diagram of @xmath1 into @xmath10 , folding it twice horizontally and vertically to reduce the number of edges , and apply proposition [ prop : grid ] .
* step 1 . * _ natural settlement of a grid diagram of @xmath1 into @xmath10 .
_ we begin by considering a grid diagram of @xmath1 which consists of @xmath41 horizontal strands and @xmath41 vertical strands .
we realize this grid diagram into the cubic lattice @xmath10 as follows . we regard these @xmath41 horizontal strands as @xmath37-sticks lying on @xmath27-level 1 and @xmath40-levels @xmath42 in the same order as they appeared in the original grid diagram .
similarly we regard the other @xmath41 vertical strands as @xmath40-sticks lying on @xmath27-level 2 and @xmath37-levels @xmath42 .
now connect each pair of an @xmath37-stick and a @xmath40-stick , which came from adjacent strands in the grid diagram , by a @xmath27-edge whose boundary lies on @xmath27-levels 1 and 2 .
we finally get a lattice presentation of @xmath1 as shown in the leftmost of figure [ fig3 ] .
now we count the number of edges . by the definition of the grid diagram , there are exactly two @xmath37-edges each between @xmath37-levels 1 and 2 , and between @xmath37-levels @xmath43 and @xmath41 , four @xmath37-edges each between @xmath37-levels 2 and 3 , and between @xmath37-levels @xmath44 and @xmath43 , and so on .
every time we pass over each @xmath37-level , we add at most two @xmath37-edges until we reach the middle level because the number of @xmath37-edges increases by two usually , but sometimes is unchanged or decreases by two .
thus the maximum number of @xmath37-edges is @xmath45 if @xmath41 is odd , or @xmath46 if @xmath41 is even .
we count the number of @xmath40-edges in the same manner .
obviously the number of @xmath27-edges is equal to @xmath47 .
therefore the maximum number of edges of this lattice presentation is @xmath48 if @xmath41 is odd , or @xmath49 if @xmath41 is even .
so we have the following inequality : @xmath50 * step 2 . *
_ horizontal folding to reduce one fourth of the square growth .
_ we would like to fold this lattice presentation structure along the line @xmath51 , @xmath52 if @xmath41 is odd , and @xmath51 , @xmath53 if @xmath41 is even as shown in figure [ fig3 ] . during this folding argument ,
any stick does not pass another but only some @xmath37-sticks can be overlapped . to get a proper lattice presentation , remove such all overlapped @xmath37-edges which are expressed by dotted line segments in the last figure .
note that almost half of @xmath40-sticks which are on the right side in the leftmost figure moved into @xmath27-level 0 and all @xmath37-sticks stayed on @xmath27-level 1 .
this movement does not change the knot type of @xmath1 .
[ h ] again we count the number of edges .
note that the numbers of @xmath40-edges and @xmath27-edges are unchanged .
if @xmath41 is odd , then each @xmath37-level @xmath54 contains exactly two @xmath40-sticks and @xmath37-level @xmath55 contains one @xmath40-stick .
this means that there are four @xmath37-edges between @xmath37-levels 1 and 2 , at most eight @xmath37-edges between @xmath37-levels 2 and 3 , and so on .
similarly there are two @xmath37-edges between @xmath37-levels @xmath56 and @xmath55 , at most six @xmath37-edges between @xmath37-levels @xmath57 and @xmath56 , and so on .
keep counting until we reach the middle level .
then the maximum number of @xmath37-edges is @xmath58 .
moreover we can reduce two more @xmath27-edges in @xmath37-level @xmath55 .
if @xmath41 is even , then each @xmath37-level @xmath59 contains two @xmath40-sticks and each @xmath37-level 1 and @xmath60 contains one @xmath40-stick .
thus there are two @xmath37-edges each between @xmath37-levels 1 and 2 , and between @xmath37-levels @xmath61 and @xmath60 , at most six @xmath37-edges each between @xmath37-levels 2 and 3 , and between @xmath37-levels @xmath62 and @xmath61 , and so on .
then the maximum number of @xmath37-edges is @xmath63 if @xmath60 is odd , or @xmath64 if @xmath60 is even .
moreover we can reduce two @xmath27-edges each in @xmath37-levels 1 and @xmath60 .
therefore the maximum number of edges of this new lattice presentation is @xmath65 if @xmath41 is odd , @xmath66 if @xmath67 for some positive integer @xmath68 , or @xmath69 if @xmath70 .
so we have the following inequality : @xmath71 * step 3 . *
_ vertical folding to reduce one eighth of the square growth . _
we fold again the new lattice presentation structure along the line @xmath72 , @xmath73 if @xmath41 is odd , and @xmath72 , @xmath74 if @xmath41 is even as shown in figure [ fig4 ] .
when we fold , all @xmath40-sticks on @xmath27-level 2 stayed on the same @xmath27-level , upper half of @xmath37-sticks on @xmath27-level 1 moved into @xmath27-level 3 and the rest half stayed on @xmath27-level 1 , and upper half of @xmath40-sticks on @xmath27-level 0 moved into @xmath27-level 4 and the rest half stayed on @xmath27-level 0 . during this folding argument , a stick does not pass another , only some of @xmath40-sticks on @xmath27-level 2 are overlapped , and some of @xmath40-sticks on @xmath27-level 0 are broken into two pieces . to get a proper lattice presentation , remove such all overlapped @xmath40-edges , and connect each pair of the broken @xmath40-sticks by two @xmath40-edges and four @xmath27-edges as shown in the last figure in figure [ fig4 ] .
still this movement does not change the knot type of @xmath1 .
[ h ] in this step , counting the number of edges is more complicated .
first consider the @xmath40-sticks on @xmath27-level 2 which are at least half among @xmath41 @xmath40-sticks . indeed
, the exact number of such @xmath40-sticks is @xmath55 for odd @xmath41 case , and @xmath60 for even @xmath41 case .
folding these @xmath40-sticks is exactly in the same way as folding @xmath37-sticks in step 2 .
when we fold @xmath37-sticks in step 2 , the maximum number of @xmath37-edges decreases by almost half as follows : @xmath75 replaced by @xmath76 for odd @xmath41 , @xmath77 replaced by @xmath78 for @xmath67 , and @xmath77 replaced by @xmath79 for @xmath70 .
so we can apply this calculation to the above half class of @xmath40-sticks on @xmath27-level 2 .
more precisely the decrease by almost half is guaranteed when we fold all @xmath40-sticks as proceeded in step 2 .
but we can surely apply this argument to either the set of almost half @xmath40-sticks on the left side or the set of almost half @xmath40-sticks on the right side in the first figure in figure [ fig3 ] .
we may assume that the above half class of @xmath40-sticks on @xmath27-level 2 fits this condition .
note that for the other half class of @xmath40-sticks on @xmath27-levels 0 and 4 , we add two more @xmath40-edges and four more @xmath27-edges to connect each pair after broken
. then the maximum number of @xmath40-edges is @xmath80 for odd @xmath41 case , @xmath81 for @xmath67 case , and @xmath82 for @xmath70 case . after counting the @xmath27-sticks used to connect each pair of the broken @xmath40-sticks ,
the maximum number of @xmath27-edges in total is @xmath83 for odd @xmath41 , and @xmath84 for even @xmath41 , while the number of @xmath37-edges are unchanged from step 2 .
therefore the maximum number of edges of the final lattice presentation is @xmath85 if @xmath41 is odd , @xmath86 if @xmath67 , or @xmath87 if @xmath70 .
we have the following inequality : @xmath88 * step 4 .
* _ apply proposition [ prop : grid ] to complete the proof of theorem [ thm : len ] .
_ since we can take any upper bounds of step 1 , 2 and 3 , we finally get : @xmath89 now apply proposition [ prop : grid ] to complete the proof of theorem [ thm : len ] .
in this section we prove theorem [ thm : rop ] . since a lattice knot can be changed to a smooth knot by taking a twice enlargement and replacing every corner with a suitable quartile circle as illustrated in figure [ fig5 ] .
apply this argument to each steps in section [ sec : len ] .
this means that we take twice the length and subtract @xmath90 the number of corners times .
abgwqq c. adams , m. chu , t. crawford , s. jensen , k. siegel , and l. zhang , _ stick index of knots and links in the cubic lattice _ , preprint arxiv:1205.5256v1 .
y. bae and c. y. park , _ an upper bound of arc index of links _ , math .
* 129 * ( 2000 ) 491500 .
g. buck and j. simon , _ thickness and crossing number of knots _ , topology appl .
* 91(3 ) * ( 1999 ) 245257
. j. cantarella , x.w.c .
faber , and c. mullikin , _ upper bounds for ropelength as a function of crossing number _
, topology appl .
* 135 * ( 2004 ) 253264 .
j. cantarella , r. kusner , and j. sullivan , _ on the minimum ropelength of knots and links _
, invent . math .
* 150(2 ) * ( 2002 ) 257286 .
p. cromwell , _ embedding knots and links in an open book i : basic properties _ , topology appl . * 64 * ( 1995 ) 3758 .
y. diao , _ minimal knotted polygons on the cubic lattice _ , j. knot theory ramif . * 2 * ( 1993 ) 413425 .
y. diao , _ the lower bounds of the lengths of thick knots _ , j. knot theory ramif .
* 12 * ( 2003 ) 116 .
y. diao and c. ernst , _ the complexity of lattice knots _ , topology appl .
* 90 * ( 1998 ) 19 .
y. diao , c. ernst , and x. yu , _ hamiltonian knot projections and lengths of thick knots _ , topology appl . * 136 * ( 2004 ) 736 .
k. hong , s. no , and s. oh , _ upper bound on lattice stick number of knots _ , math . proc .
* 155 * ( 2013 ) 173179 .
k. hong , s. no , and s. oh , _ upper bounds on the minimum length of cubic lattice knots _ , j. phys . a : math
* 46 * ( 2013 ) 125001 .
x. hua , d. nguyen , b. raghavan , j. arsuaga , and m. vazquez , _ random state transitions of knots : a first step towards modeling unknotting by type ii topoisomerases _ , topology appl . * 154 * ( 2007 ) 13811397 .
y. huh and s. oh , _
lattice stick numbers of small knots _ , j. knot theory ramif . * 14 * ( 2005 ) 859867 .
y. huh and s. oh , _ knots with small lattice stick numbers _ , j. phys . a : math .
* 43 * ( 2010 ) 265002(8pp ) .
e. j. janse van rensburg and s. d. promislow , _ minimal knots in the cubic lattice _ , j. knot theory ramif .
* 4 * ( 1995 ) 115130 .
g. t. jin and w. k. park , _ prime knots with arc index up to 11 and an upper bound of arc index for non - alternating knots _ , j. knot theory ramif .
* 19 * ( 2010 ) 16551672 .
r. scharein , k. ishihara , j. arsuaga , y. diao , k. shimokawa , and m. vazquez , _
bounds for the minimum step number of knots in the simple cubic lattice _ , j. phys . a : math . theor .
* 42 * ( 2009 ) 475006 .
|
let @xmath0 be the minimum length of a knot on the cubic lattice ( namely the minimum length necessary to construct the knot in the cubic lattice ) .
this paper provides upper bounds for @xmath0 of a nontrivial knot @xmath1 in terms of its crossing number @xmath2 as follows : @xmath3 the ropelength of a knot is the quotient of its length by its thickness , the radius of the largest embedded normal tube around the knot .
we also provide upper bounds for the minimum ropelength @xmath4 which is close to twice @xmath0 : @xmath51.5 c(k)^2 + 9.15 c(k ) + [email protected] c(k)^2 + 14.58 c(k ) + 16.90@xmath7
|
agb stars with low initial mass are the major responsible for the nucleosynthesis of solar @xmath0 isotopes with @xmath2 @xmath3 90 @xcite .
+ the main neutron source of low - mass agb models is the @xmath1c(@xmath4 , n)@xmath5o reaction , which burns radiatively during the interpulse in a thin layer of the he intershell , the so - called _
@xmath7 @xcite .
the formation of the @xmath1c pocket requires an unknown mixing mechanism that allows partial mixing of a few protons from the convective envelope into the top layers of the radiative he- and c - rich intershell .
this is assumed to occur at the quenching of a third dredge up ( tdu ) episode .
when the star contracts , the h shell reignites and protons in the intershell are captured by the abundant @xmath8c nuclei to yield primary @xmath1c . if more protons than @xmath8c nuclei are diffused in the outer layers , a region of primary @xmath9n may form by further proton captures on @xmath1c .
subsequently , the temperature in the @xmath8c pocket increases to @xmath101@xmath1110@xmath12 k , and neutrons are released radiatively within the pocket via @xmath1c(@xmath4 , n ) reactions at quite low neutron densities .
various physical mechanisms have been explored for the formation of @xmath1c pocket ( e.g. , overshooting , rotation , magnetic fields , gravity waves ; @xcite ) .
the details of how the @xmath1c pocket forms are still debated , making its mass extent and the h profile largely uncertain .
+ spectroscopic observations provide key information to constrain theoretical models : chemically peculiar @xmath0-rich stars have evidenced a dispersion of the @xmath0 abundances for a given metallicity ( e.g. , ms , s , c(n ) , ba , cemp - s and post - agb stars , planetary nebulae ; see the recent review by @xcite and @xcite ) .
this dispersion has been recognized since first studies by @xcite and @xcite , but the reason(s ) are not definitely identified .
a variation in the stellar rotational velocity may be regarded as a possible explanation ( see @xcite , and references therein ) .
owing to the present uncertainties , the @xmath1c pocket is artificially introduced in our post - process agb models , following the observational constraints .
the shape of the @xmath1c and @xmath9n profiles and the mass involved in the pocket are regulated by a free parametrization .
the internal structure of the @xmath1c pocket adopted so far has been calibrated to represent the solar main component @xcite : it is a three - zone pocket ( each zone has defined @xmath13(@xmath1c ) and @xmath13(@xmath9n ) abundances ) with a total mass of about 0.001 @xmath14 ( see table [ tab1 ] ; first group of data ) .
a range of @xmath1c - pocket strengths is assumed to reproduce the spectroscopic @xmath0-process dispersion : we parametrically vary the concentration of @xmath1c ( and @xmath9n ) of each zone given in table [ tab1 ] by different factors , leaving the mass of the pocket constant .
we refer to @xcite for a detailed discussion .
+ this systematic approach appears justified by the present uncertainties : the formation of diverse @xmath1c pockets may result from the interplay between different physical processes in stellar interiors .
post - process models should be considered useful tests to address full stellar evolutionary models ( and in general multidimensional / hydrodynamical simulations ) against observational constraints .
the solar @xmath0-process abundances must account for the complex chemical evolution of the galaxy , which includes agb yields of various masses and metallicities .
the chemical evolution model adopted to reproduce the solar @xmath0 distribution has been exhaustively described by @xcite . + in the framework of gce , we showed that the solar @xmath0 distribution of isotopes with 130 @xmath15 @xmath2 @xmath16 208 can be accurately reproduced once we consider a proper weighted average among the @xmath1c - pocket strengths - only isotope @xmath17sm is taken as reference nuclide for the whole @xmath0-process distribution .
the high production of @xmath18pb in low - mass low - metallicity agb stars plays another indicative gce constraint @xcite .
the solar abundance of @xmath18pb is matched once the @xmath0-process occurring in low - metallicity agb stars is properly considered in the context of the chemical evolution of the galaxy . ] .
this is consistent with the observed spectroscopic @xmath0-process dispersion . a deficit ( of about 25% ) between gce predictions of @xmath0-process elements and the solar abundances was found for isotopes with 90 @xmath15 @xmath2 @xmath15 130 ( solar lepp - s ) . @xcite
have investigated a possible connection between this deficit and the @xmath1c - pocket structure . on the basis of their tests , solar gce predictions of @xmath0-process elements are marginally affected . the aim of this work is to implement the analysis carried out on the @xmath1c - pocket structure by @xcite , ( section [ tests ] ) . in section [ yields ] , we consider the sensitivity of agb yields to metallicity , focusing on the contribution by metal - rich agb stars to the light elements ( see discussion by @xcite for open clusters ) .
although the solar composition is fundamental to constrain agb yields , it only provides a single piece of information about the galactic history .
the reliability of the @xmath1c - pocket tests needs to be verified by considering the complex framework of galactic chemical enrichment .
the gce predictions of selected neutron - capture elements versus metallicity are compared with updated spectroscopic observations in section [ z ] .
recently , rotation - induced mixing in low - metallicity massive stars has been proposed as an explanation of the observed [ sr / ba ] dispersion in extremely metal - poor stars , being efficient primary producers for @xmath0 isotopes heavier than sr , up to the ba neutron - magic peak @xcite . in this context
, we investigate the impact of recent weak @xmath0-process yields by @xcite , available from a large grid of rotating massive stars ( @xmath19 from 10@xmath20 to solar ) , on the galactic chemical enrichment ( section [ rot ] ) .
+ in section [ conclusions ] our results are briefly summarized and future outlook are discussed .
we focus on specific additional tests carried out on the basis of recent advice available in literature .
the internal @xmath1c - pocket structure adopted in each test is given in table [ tab1]c pocket in agb stars with intermediate mass ( 4 @xmath16 @xmath21/@xmath14 @xmath15 8) is negligible for gce solar predictions ( see section 2 by @xcite ; @xcite ) . ] .
* starting from the three - zone @xmath1c - profile adopted so far , we investigate the impact of a substantially extended @xmath1c - pocket mass than assumed in our previous computations ( up to four times larger , corresponding to a total mass of @xmath22(pocket)@xmath104@xmath1110@xmath23 @xmath14 ; see tests described as * case 1 * and * case 2 * in table [ tab1 ] ) .
+ in this regard , @xcite have proposed such a @xmath1c - pocket mass to represent the abundances of neutron - capture elements in young open clusters .
magnetic buoyancy ( or other forced mechanisms ) are suggested as good candidates to form such a @xmath1c reservoir @xcite .
rotation models by @xcite indicate that low - metallicity agb stars and fast - rotating metal - rich stars might produce such an extended @xmath1c pocket .
comparison between theoretical models and the strontium and barium isotopic signatures measured in mainstream sic grains require @xmath1c pockets with @xmath22(pocket)@xmath241@xmath1110@xmath23 @xmath14 @xcite . * in * case 3 * ( see table [ tab1 ] ) , we test the effect of an additional parametrized @xmath9n - pocket . +
as anticipated in section [ intro ] , an outer @xmath9n - rich layer may form in the pocket once enough protons are mixed in the external zone of the he intershell @xcite .
the presence of rotation may widen the @xmath9n pocket by inducing partial mixing of the @xmath9n - rich region with the inner @xmath1c - rich zone . * in * case 4 * , we have performed a set of 1.5 , 2 and 3 @xmath14 models computed with a more efficient reimers mass loss than in previous models .
this implies that agb models experience about half of the tdu episodes compared to case 3 .
the @xmath1c pocket structure is the same as adopted in case 3 ( see table [ tab1 ] ) . +
this test simulates the recent prescriptions of updated fruity models by @xcite .
fruity models include an efficient agb mass - loss rate , which has been calibrated using the infrared luminosity function of galactic carbon stars , and improved radiative c - enhanced opacity tables .
accordingly , these new models with a reduced number of tdu episodes are in better agreement with observations ( @xcite and references therein ) . * finally , in * case 5 * ( see table [ tab1 ] ) , we test the impact of an additional extended inner zone of the pocket with a mass of 2@xmath1110@xmath23 @xmath14 and a correspondingly lower @xmath1c abundance ( @xmath13(@xmath1c)=2.75e@xmath25 ) .
+ recently , @xcite found that a different convective / radiative boundary condition allows a deeper penetration of protons with a very low mixing efficiency during tdu episodes .
the resulting @xmath1c pocket displays an extended tail with a smooth decrease of the @xmath1c profile ( see @xmath26 model in their fig .
case 5 roughly approximates the @xmath26 model by @xcite .
ccccccccc + & mass(@xmath14 ) & @xmath27 & 5.50e@xmath274 & 5.30e@xmath274 & 7.50e@xmath276 & @xmath27 & @xmath27 & + & @xmath13(@xmath1c ) & @xmath27 & 3.20e@xmath273 & 6.80e@xmath273 & 1.60e@xmath272 & @xmath27 & @xmath27 & + & @xmath13(@xmath9n ) & @xmath27 & 1.07e@xmath274 & 2.08e@xmath274 & 2.08e@xmath273 & @xmath27 & @xmath27 & + + & mass(@xmath14 ) & @xmath27 & 1.65e@xmath273 & 1.59e@xmath273 & 2.25e@xmath275 & @xmath27 & @xmath27 & + & @xmath13(@xmath1c ) & @xmath27 & 3.20e@xmath273 & 6.80e@xmath273 & 1.60e@xmath272 & @xmath27 & @xmath27 & + & @xmath13(@xmath9n ) & @xmath27 & 1.07e@xmath274 & 2.08e@xmath274 & 2.08e@xmath273 & @xmath27 & @xmath27 & + + & mass(@xmath14 ) & @xmath27 & 2.20e@xmath273 & 2.12e@xmath273 & 3.00e@xmath275 & @xmath27 & @xmath27 & + & @xmath13(@xmath1c ) & @xmath27 & 3.20e@xmath273 & 6.80e@xmath273 & 1.60e@xmath272 & @xmath27 & @xmath27 & + & @xmath13(@xmath9n ) & @xmath27 & 1.07e@xmath274 & 2.08e@xmath274 & 2.08e@xmath273 & @xmath27 & @xmath27 & + + & mass(@xmath14 ) & @xmath27 & 1.20e@xmath273 & 1.05e@xmath273 & 3.30e@xmath274 & 3.30e@xmath274 & 3.30e@xmath274 & + & @xmath13(@xmath1c ) & @xmath27 & 3.20e@xmath273 & 6.80e@xmath273 & 1.60e@xmath272 & 4.00e@xmath272 & 4.00e@xmath272 & + & @xmath13(@xmath9n ) & @xmath27 & 1.07e@xmath274 & 2.08e@xmath274 & 2.08e@xmath273 & 8.00e@xmath272 & 1.49e@xmath271 & + + & mass(@xmath14 ) & 2.00e@xmath273 & 5.50e@xmath274 & 5.30e@xmath274 & 7.50e@xmath276 & @xmath27 & @xmath27 & + & @xmath13(@xmath1c ) & 2.75e@xmath273 & 3.20e@xmath273 & 6.80e@xmath273 & 1.60e@xmath272 & @xmath27 & @xmath27 & + & @xmath13(@xmath9n ) & 5.73e@xmath275 & 1.07e@xmath274 & 2.08e@xmath274 & 2.08e@xmath273 & @xmath27 & @xmath27 & + + + + + + the resulting solar gce predictions of @xmath0-process elements are displayed in fig .
a proper weighted average among the various @xmath1c - pocket strengths must be adopted for each test to reconcile gce predictions with 100% of solar @xmath17sm ( see @xcite ) .
this approach allows us to reproduce the solar @xmath17sm within 5% uncertainty @xcite . +
the @xmath0-only isotopes with atomic mass @xmath2 @xmath28 90 show variations smaller than @xmath1010% , thus confirming the need of a solar lepp - s mechanism in order to increase the solar @xmath0 abundances in the range 90 @xmath15 @xmath2 @xmath15 130 , as predicted by @xcite .
note that agb yields computed with a single @xmath1c - pocket choice do not provide accurate interpretations neither of solar @xmath0 abundances nor of observations of peculiar @xmath0-rich stars . by working in a larger range of uncertainties ,
potential missing contributions are not necessarily highlighted .
accordingly , the result of this paper does not disagree with a recent study by @xcite that provides a meticulous discussion about the uncertainties affecting stellar models and the solar gce distribution . within the estimated uncertainties
a lepp mechanism is not necessarily required .
however , the authors suggest how their representation of the solar distribution could be improved once models with different initial rotational velocities ( or different prescriptions for convective overshoot during the tdu ) will be included in gce computations for an extended metallicity grid .
these upcoming progresses in stellar models will assess whether additional contributions are needed .
the complex dependence of @xmath0-process yields on the initial stellar metallicity is well known . in fig .
[ fig2 ] ( top panel ) , we display the agb production factors of selected @xmath0-only isotopes . starting from [ fe / h ] = @xmath270.5 , which corresponds to @xmath29 = 3.5 gyr ,
the s contribution to light isotopes with a = 96 130 is favored . as suggested by @xcite and @xcite ,
it is crucial to verify that the delayed contribution of metal - rich agb stars to the solar epoch is properly considered in the framework of our gce model .
otherwise , the solar lepp - s may derive from a relevant underestimation of the metal - rich agb component .
+ the theoretical minimum agb initial mass that can contribute to the solar system is rather uncertain and model dependent . in our gce computations , @xmath21 @xmath30 1.3 @xmath14 models with [ fe / h ] @xmath28 @xmath270.4
do not contribute to the chemical evolution of the galaxy because the conditions for the activation of the tdu episodes are never reached in our models ( see e.g. , @xcite , and references therein ) .
+ we have analyzed here a test case , in which we have accounted for the longest life of the agb stars that may contribute to the solar distribution ( e.g. , our 1.3 @xmath14 models with disc metallicity have lifetime of @xmath104 gyr ) . in fig .
[ fig2 ] ( bottom panel ) , we compare the @xmath0 distribution obtained at the formation time of the solar system ( @xmath31 = 9.2 gyr ) and that computed at @xmath29 = 13.1 gyr , in which the overall metal - rich agb component that may contribute to the solar s distribution is included .
the two distributions show marginal variations ( @xmath323% ) for @xmath2 @xmath3 90 .
indeed , our gce model adopts a very efficient star formation rate in the halo , which produces a fast metallicity increase at the early evolutionary epochs ( [ fe / h ] = @xmath272 at 0.1 gyr ) .
an extended thin disc phase ( @xmath108 gyr ) assures that the contribution of metal - rich stars is not overlooked .
moreover , the small star formation rate after 9.2 gyr results in a slow metallicity increase ( by + 0.11 dex from solar to 13.1 gyr ) . for this reason ,
the solar distribution ( normalized to @xmath17sm ) shows small variations at the two selected epochs .
this indicates that the solar distribution properly accounts for the contribution of agb yields with [ fe / h ] @xmath32 @xmath270.2 .
+ non - negligible variations are found for @xmath33sr and @xmath34y : at @xmath29 @xmath1013 gyr , @xmath35(@xmath33sr)@xmath36 + 8% and @xmath35(@xmath34y)@xmath36 + 4% .
the @xmath0-only isotopes with @xmath2 @xmath15 90 are mainly produced by agb stars with [ fe / h]@[email protected] , which corresponds to @xmath29 @xmath10 6.3 gyr , only about 2.9 gyr before the solar system formation ( see fig .
[ fig2 ] , top panel ) .
this is comparable to the lifetime of agb stars with @xmath21 = 1.4 @xmath27 1.5 @xmath14 with [ fe / h]@[email protected] ( @xmath37 @xmath10 3.0 to 2.4 gyr , respectively ) .
+ in table [ tab2 ] , we list the @xmath0 contributions to elements from sr to cd at @xmath31 = 9.2 gyr and @xmath29 @xmath1013 gyr : although non - negligible variations are found for sr ( + 5% ) and y ( + 4% ) , these differences are compatible with the solar uncertainties . as discussed in the previous section
, the lepp - s is still required .
top panel : agb yields ( obtained with a weighted average of standard @xmath1c - pockets ) of a 1.5 @xmath14 model at different metallicities .
four selected @xmath0-only isotopes are shown : @xmath17sm is representative of heavy @xmath0-only isotopes with @xmath2 = 142 @xmath27 204 ( peaked at [ fe / h ] @xmath10 @xmath270.8 , which in our gce model corresponds to a galactic time of @xmath29 @xmath10 2 gyr ) ; @xmath38cd illustrates the behavior of light @xmath0-only isotopes with @xmath2 = 96 @xmath27 130 ( [ fe / h ] @xmath10 @xmath270.5 ; @xmath29 @xmath10 3.5 gyr ) ; last are the @xmath0-only isotopes with @xmath2 = 70 @xmath27 87 , represented by @xmath39sr ( [ fe / h ] @xmath10 @xmath270.15 ; @xmath29 @xmath10 6.3 gyr ) .
@xmath18pb is efficiently produced by low - metallicity agb stars ( [ fe / h ] @xmath10 @xmath271.3 ; @xmath29 @xmath10 1 gyr ) .
bottom panel : the @xmath0-process distribution obtained at the epoch of the solar system formation ( @xmath31 = 9.2 gyr at [ fe / h ] = 0 ; black symbols ) compared with the @xmath0-process distribution obtained at a galactic time of @xmath29 = 13.1 gyr ( at [ fe / h ] = + 0.11 ; gray symbols ) .
both results have been normalized to @xmath17sm .
, width=366 ]
in this section we present our results for the chemical evolution of selected neutron - capture elements as a function of metallicity . we analyze the chemical evolution of y , zr , ba , la and pb ( representative of the three @xmath0-process peaks ) , and eu ( typical of the @xmath40-process elements ) .
gce predictions are compared with updated high - resolution spectroscopic observations . the galactic enrichment of the three @xmath0-process peaks is followed by accounting for the @xmath0-process agb yields discussed in sections [ tests ] and [ yields ] .
an additional contribution to the first @xmath0 peak is ascribed to the ` classical ' weak @xmath0 component from ( non - rotating ) massive stars @xcite .
given the secondary - like nature of the @xmath41ne(@xmath4 , n)@xmath42 mg neutron source in standard non - rotating massive stars ( with yields scaling quite linearly with the initial metallicity ) , the weak @xmath0 process is mainly relevant at solar metallicity , accounting for up to @xmath1010% of solar sr - y - zr @xcite .
we will discuss the supplementary impact of fast - rotating low - metallicity massive stars on gce predictions in section [ rot ] .
+ concerning the treatment of eu and @xmath40-process elements , we have adopted the same prescription discussed by @xcite . actually , the origin of the @xmath40 process is still unclear and largely uncertain .
spectroscopic observations have evidenced the existence of multiple @xmath40-process components operating in the early galaxy ( see , e.g. , @xcite , and references therein ) .
at least two @xmath40-processes have been bf proposed : the main @xmath40 process ( @xmath2 @xmath3 130 ) , which is currently ascribed to mergers of compact objects and/or magneto - rotationally driven supernovae , and the weak @xmath40 process ( or @xmath43p - process , rp - process ; @xmath2 @xmath10 80 to 120 ) , which has been proposed as an alternative nucleosynthesis source of lighter heavy elements in neutrino - driven supernova outflows .
noteworthy progresses have been reached in modeling these extremely energetic and complex events , but the unclear physical conditions of the astrophysical sites and the inaccessibility of nuclear physics inputs lead to very uncertain @xmath40-process predictions ( see e.g. , @xcite , and references therein ) .
+ the lack of a comprehensive @xmath40-process view forced us to systematically include an @xmath40-process contribution in gce computations . for elements heavier than ba , the solar @xmath40-process contribution is derived by subtracting the @xmath0 fractions from the solar abundances ( the so - called @xmath40-process residuals method ; @xcite ) .
following the observed decreasing trend of heavy neutron - capture elements in the early galaxy , we ascribe the @xmath40 contribution to a primary process occurring in snii with a limited range of progenitor masses , @xmath21 @xmath10 8@xmath2710 @xmath14 @xcite . in light of the present uncertainties ,
we do not exclude different hypotheses ( e.g. , jet - sne , neutron star mergers ) .
chemo - dynamical evolution models are needed to adequately explore the chemical origin of an inhomogeneous galactic halo ( see e.g. , discussion by @xcite ) .
nevertheless , by adopting different ranges of stellar mass progenitors for the @xmath40 process , the gce predictions show marginal variations for disk metallicities , and for the solar @xmath0 distribution .
+ even more puzzling is the origin of light neutron - capture elements in the early galaxy , for which simple @xmath40-process residuals provide inadequate descriptions @xcite .
accordingly , a different treatment is required for sr , y and zr .
we derive an @xmath40 fraction of @xmath1010% from observations of very metal - poor @xmath40-rich stars @xcite , under the hypothesis that these peculiar objects show the signature of a pure main @xmath40-process .
an additional lepp contribution was evoked by @xcite to explain the missing abundances of solar sr , y , and zr .
the flat [ sr , y , zr / fe ] trend observed at low metallicities suggests that lepp is a primary process , likely occurring in snii with an extended range of mass progenitors compared to the main @xmath40 process .
as discussed by @xcite , we confirm that the solar lepp - s described in section [ tests ] and the metal - poor lepp may originated in different stellar environments . figures [ fig3 ] and [ fig4 ] illustrate the galactic evolution of [ ba / fe ] , [ la / fe ] , [ y / fe ] , [ zr / fe ] , [ eu / fe ] , [ pb / fe ] and their ratios [ ba / eu ] , [ la / eu ] , [ eu / y ] , [ eu / zr ] , [ ba / y ] , [ pb / eu ] versus [ fe / h ] .
+ spectroscopic observations are taken from aoki & honda ( 2008 ; @xmath44-@xmath45 @xmath46-@xmath47 ) , nissen & schuster ( 2011 ; @xmath48-@xmath49 @xmath50 ) , andrievsky et al .
( 2011 ; @xmath51 @xmath52 ) , aoki et al .
( 2013 ; @xmath53-@xmath54 @xmath52 ) , hansen et al .
( 2012 ; @xmath55 and @xmath56 @xmath49 @xmath52 for dwarfs and giants , respectively ) , hansen et al .
( 2014b ; @xmath57 @xmath58 ) , roederer et al .
( 2010,2012,2014,2014a ; @xmath59 @xmath60 , @xmath44-@xmath45 and @xmath45 @xmath52 , @xmath54 @xmath46-@xmath50 ) , ishigaki et al .
( 2013 ; @xmath45 , @xmath61 , and @xmath62 @xmath63-@xmath64 @xmath65 for thick - disc , inner - halo and outer - halo stars , respectively ) , mishenina et al .
( 2013 ; @xmath53-@xmath59 and @xmath56 @xmath52 for thin and thick disc stars , respectively ) , cohen et al .
( 2013 ; @xmath66 @xmath58 ) , yong et al .
( 2013 ; @xmath59 @xmath67 ) , bensby et al .
( 2014 ) and battistini & bensby ( 2016 ; @xmath51 , @xmath59 , @xmath66 , @xmath68 @xmath65 for thin disc stars , unclassified stars , thick - disc and halo stars , respectively ) , mashonkina et al .
( 2014 ; @xmath57 @xmath65 ) , siqueira mello et al .
( 2014 ; @xmath57 @xmath69-@xmath64 @xmath65 ) . stars with [ ba / fe ] @xmath24 + 0.6 and [ ba / eu ] @xmath3 0 are excluded ( e.g. , possible binaries as ba stars , c - stars , cemp - s stars ) . the well studied sample of @xmath40-rich stars at [ fe / h ] @xmath15 @xmath272.2 ( r - i and r - ii stars , with [ eu / fe ] @xmath24 1 and [ ba / eu]@xmath70 @xmath10 @xmath270.8 ; @xcite ) are peculiar objects likely born from a cloud polluted by a pure @xmath40-process production event .
they do not represent the average chemical evolution of our galaxy .
+ gce predictions account for the @xmath0-process , the @xmath40-process , and the lepp contributions , in the halo , thick disk , and thin disk ( @xmath71 , @xmath72 and @xmath73 @xmath74 , respectively ) .
agb yields with a standard @xmath1c - pocket choice ( first group of data in table [ tab1 ] ) are included . the shaded area at [ fe / h ] @xmath16 @xmath272.5 indicates that homogeneous gce models can not properly represent the [ el / fe ] scatter observed in the early galaxy for neutron - capture elements .
it is noteworthy that low [ sr , ba , eu / fe ] abundances in metal - poor field stars may be affected by higher uncertainties owing to observational detection thresholds @xcite .
gce contributions of [ ba / fe ] , [ y / fe ] and [ pb / fe ] versus [ fe / h ] .
different lines correspond to gce results for thick ( @xmath45 @xmath74 ) and thin disc ( @xmath57 @xmath74 ) obtained with different @xmath1c - pocket tests in table [ tab1 ] .
we also display the result of a flat @xmath1c - pocket profile obtained by excluding the two external @xmath1c - rich zones of the pocket ( see @xcite ) .
spectroscopic observations are the same shown in fig .
[ fig3 ] . , width=1058 ] in figure [ fig5 ] , we illustrate the impact of agb yields computed with different @xmath1c - pocket tests selected from section 2 on gce predictions of @xmath0-process elements .
we consider [ ba / fe ] , [ y / fe ] and [ pb / fe ] versus [ fe / h ] as representative of the three @xmath0 peaks .
similar results are obtained for [ la / fe ] and [ zr / fe ] . + as discussed in section [ tests ] , [ el / fe ] and their ratios are plausibly reproduced at [ fe / h ] = 0 .
variations up to @xmath35[el / fe]@[email protected] dex are shown by gce predictions of the three @xmath0 peaks in the metallicity range of @xmath271.6 @xmath15 [ fe / h ] @xmath15 0.1 . at [ fe / h ] @xmath15 @xmath271.6 ,
the @xmath0 contribution is negligible , and marginal variations are expected .
eu , which receives a dominant @xmath40-process contribution ( 94% ) , is unaffected by different tests .
the growing number of spectroscopic data currently suggests a possible underestimation of the observed [ ba / fe ] for [ fe / h ] @xmath10 @xmath272 , independently of the @xmath1c profile adopted in the pocket .
for instance , the observed [ ba / eu ] ratio smoothly increases starting from [ fe / h ] @xmath10 @xmath272.5 @xcite , about 1 dex earlier than predicted by our model ( [ fe / h ] @xmath10 @xmath271.6 ) . despite a smaller dispersion
is observed for [ la / fe ] , similar indications are given by [ la / eu ] , which on average show an even flatter and higher trend for [ fe / h ] @xmath15 @xmath272 .
this trend appears a characteristic of the second @xmath0 peak rather than other @xmath0 elements ( see , e.g. , [ y , zr / fe ] and [ ba / y ] versus [ fe / h ] ) .
otherwise , galactic thin and thick disk dwarfs by mishenina et al . ( 2013 ; little filled squares in the metallicity range of @xmath271 @xmath16 [ fe / h ] @xmath16 + 0.3 ) show a delayed and steeper [ ba , la / eu ] increase , in agreement with our gce @xmath0 process enrichment .
+ current agb yields seem not give adequate hints to account for this behavior . within a gce framework , several studies agree that the rise of the agb @xmath0 process occurs beyond [ fe / h ] @xmath10 @xmath271.6 @xcite , given the fast iron enrichment occurring in the early galaxy .
however , an anticipated rise for the @xmath0-process was questioned by previous spectroscopic studies @xcite .
recent observational and theoretical studies have evidenced that an additional intermediate neutron - capture process ( the @xmath76 process ) may play an important role in the evolution of the galaxy ( see , e.g. , peculiar chemical signatures discussed by @xcite ) . despite its existence
was hypothesized 30 years ago by @xcite , observational hints have only recently been revealed .
the @xmath76 process occurs when protons are ingested in he - burning convective stellar regions to form @xmath1c , which is the main source of neutrons via the ( @xmath4 , n ) reaction , driving neutron densities of @xmath1010@xmath77 neutrons cm@xmath23 .
the physical conditions leading to the @xmath76 process may be found in different stellar environments , as super - agb and post - agb stars , he - core and he - shell flashes in low - metallicity low - mass stars , and massive stars @xcite . improved stellar models with the guidance from hydrodynamics simulations
will help to establish the impact of the @xmath76 process in the milky way .
the important contribution of fast - rotating massive stars in the early galaxy has been evidenced by several studies ( see e.g. , the review by @xcite ) .
first investigations highlighted that rotational induced - mixing boost the production of primary @xmath9n in the h - burning shell , subsequently converted to @xmath41ne via 2@xmath4 captures in the he - burning core @xcite .
the large amount of @xmath41ne primarily produced at low metallicities leads to an efficient nucleosynthesis of light @xmath0 elements , and may extend up to the barium peak @xcite .
the recent studies by @xcite and @xcite show a large sensitivity of the first two @xmath0 peaks to both metallicity and rotation , with a [ sr / ba ] predicted ratio that covers more 2 dex in most metal - free models .
rotating low - metallicity massive stars are a promising site to explain the [ sr / ba ] dispersion observed in the halo .
@xcite first advanced the hypothesis that metal - poor fast - rotating massive stars may offer a plausible explanation for the large [ ba / y ] dispersion observed in the early universe .
later on , the inhomogeneous chemical evolution model by @xcite provided a quantitative estimation in support of this hypothesis .
the aim of this section is to test the effect of weak @xmath0 process yields from fast - rotating massive stars on gce predictions of @xmath0-process elements .
specifically , new yields by @xcite are analyzed in the framework of our gce model in order to explore the impact on the solar and metal - poor lepp mechanisms .
+ @xcite provide a grid of four stellar masses ( @xmath21 = 15 , 20 , 25 and 40 @xmath14 ) and three metallicities ( @xmath19 = @xmath79 , 10@xmath23 , 10@xmath20 ) , computed with a full s - process network .
we have linearly interpolated the weak @xmath0 yields between the mass range of 15 @xmath15 @xmath21/@xmath14 @xmath16 40 and the metallicity range .
the authors compare non - rotating and standard - rotating models ( @xmath80/@xmath81 = 0.4 , where the critical velocity @xmath81 assumed by the authors corresponds to an average equatorial velocity of @xmath10200 km s@xmath82 on the main sequence at @xmath79 , and increases to @xmath10400 km s@xmath82 at low metallicity , see @xcite ) .
they also provide additional yields on the 25 @xmath14 model down to @xmath19 = 10@xmath83 , with fast - rotation rates ( @xmath80/@xmath81 = 0.5 and 0.6 , with @xmath81 up to @xmath10590 km s@xmath82 ) . for fast - rotating low - metallicity models of 25 @xmath14 ,
the authors discussed a further test in which the @xmath84o(@xmath4 , @xmath85 ) rate is divided by a factor of ten , in better agreement ( within the uncertainties ) with the recent measurement by @xcite .
+ although only pre - explosive sn yields are given by @xcite , they are suited for gce studies because strong variations on the total yields of s - process nuclei are not expected after sn explosion ( see e.g. , @xcite ) .
we have considered three tests : + a first test with non - rotating weak @xmath0-process yields ( label @xmath86 ; corresponding to models ` @xmath87 ' in table 1 by @xcite ) , which provides a more exhaustive analysis of the ` classical ' weak-@xmath0 process contribution included in the gce predictions presented in section [ z ] ; + a second test with standard - rotating weak @xmath0-process yields ( label @xmath88 ; corresponding to models ` @xmath89 ' in table 1 by @xcite , with @xmath80/@xmath81 = 0.4 ) ; + the same as the second test , but with fast - rotating yields ( @xmath80/@xmath81 = 0.6 ; @xmath90 ) coupled with the @xmath84o(@xmath4 , @xmath85 ) rate divided by a factor of ten for the 25 @xmath14 model at @xmath19 = 10@xmath20 and 10@xmath83 ( corresponding to models ` @xmath91 ' in table 1 by @xcite ) .
we applied the same scaling factors to 40 @xmath14 models . in fig .
[ fig6 ] , the impact of new weak @xmath0 yields is shown for the solar abundance distribution of @xmath0 isotopes .
non - rotating weak @xmath0 yields mainly increase the solar abundances of isotopes with @xmath2 @xmath32 80 , with a small contribution to sr , y and zr .
this is in agreement with previous ` classical ' analyses , which have been obtained starting from the solar weak @xmath0 contribution and following a secondary - like behavior by decreasing the metallicity .
fast - rotating models produce an evident increase of isotopes lighter than @xmath2 @xmath32 90 , with an additional contribution of @xmath1030% to @xmath33sr , and @xmath1010% to @xmath34y , @xmath105% to @xmath92zr among non @xmath0-only isotopes .
this corresponds to a weak @xmath0 contribution from fast - rotating massive stars of about 17% , 10% and 5% to solar sr , y , zr , respectively . in the atomic mass region between 90 @xmath15 @xmath2 @xmath15 140 , @xmath0-only isotopes
are marginally affected , with variations within the solar uncertainties .
this suggests that fast - rotating massive stars may only partially account for the solar lepp , being inefficient for @xmath2 @xmath3 90 . by considering the influence of the fast - rotating weak - s process over the whole galactic enrichment
, we display in fig .
[ fig7 ] the impact on [ ba / fe ] , [ y / fe ] , [ ba / y ] and [ eu / y ] versus [ fe / h ] , selected as most representative .
black lines account for both the ` classical ' weak-@xmath0 process and for the estimated lepp .
other lines are computed by setting the lepp contribution to zero , and by including the three tests listed above for the weak-@xmath0 process .
we adopt our standard treatment described in section [ z ] for the @xmath40 contribution .
the @xmath40 process dominates the [ ba / fe ] predictions for @xmath273 @xmath16 [ fe / h ] @xmath32 @xmath271.5 .
the impact of fast - rotating models only affects very metal - poor gce predictions ( see @xmath51 @xmath93 at [ fe / h ] @xmath15 @xmath273 ) .
as expected , eu ( which is not shown in figure ) does not present variations . for [ y / fe ] , different initial rotation rates ( represented by @xmath45 , @xmath59 , and @xmath51 lines in fig .
[ fig7 ] ) seem to cover a more extended range of stellar observations at [ fe / h ] @xmath15 @xmath273 .
this also partly explains the observed [ eu / y ] and [ ba / y ] dispersions .
it is noteworthy that fast - rotating massive stars offer a promising nucleosynthesis site for the metal - poor lepp .
for instance , gce predictions that account for fast - rotating weak - s process yields ( @xmath51 line ) move towards the original model with the estimated lepp ( @xmath57 line ) .
although our homogeneous gce model can not provide a realistic representation of the observed s - element dispersion , the averaged gce predictions shown in fig .
[ fig7 ] are in agreement with previous results presented by @xcite , who indicated that low - metallicity fast - rotating massive stars may reproduce the [ sr / ba ] inhomogeneities observed in extremely metal - poor stars . + note that large uncertainties affect massive - star models , as discussed by @xcite : on the nuclear point of view , owing to the uncertain @xmath84o+@xmath4 and @xmath41ne+@xmath4 rates , and on the modeling point of view , mainly because of the treatment of the physics associated to magnetic fields and rotation - induced mixing . in the framework of gce models , a more detailed grid of stellar yields may help to better assess an upper limit for the fast - rotating weak - s process contribution .
+ as evidenced by different gce approaches , the chemical enrichment of the stellar halo is likely the result of a merger of different ( primary ) processes , including fast - rotating weak - s process and @xmath40 processes ( e.g. , @xcite ; see also section [ z ] ) .
we have implemented the study on gce predictions of @xmath0-process elements presented by @xcite by including recent pivotal suggestions provided in literature .
the main result is that the predicted chemical compositions in the interstellar medium at any epoch are grossly independent of the particular @xmath1c profile adopted inside the pocket , once a different weighted average of @xmath1c - pocket strengths is considered to reproduce solar @xmath17sm ( within 5% accuracy ) .
+ the impact of agb uncertainties on gce computations may be partially reduced by assuming a range of @xmath1c - pocket strengths , as indicated by observational constraints .
actually , this approach allows us to obtain more accurate information about the complementary contribution from other nucleosynthesis processes that may compete over the galactic history . in the framework of these @xmath1c - pocket prescriptions , we confirm that the additional lepp - s mechanism invoked by @xcite is required to represent the observed solar distribution within the estimated solar uncertainties . in this view , only 3d or hydrodynamical agb modeling may shed light on the @xmath1c - pocket scenario .
+ updated gce calculations accounting for recent fast - rotating metal - poor weak @xmath0-process yields by @xcite may partially solve the missing contribution to the solar sr - y - zr abundances .
the [ sr / ba ] dispersion observed in the early galaxy may be reconciled with the contribution of metal - poor massive stars with different rotational velocity , in agreement with previous results by @xcite and @xcite .
a more detailed grid of stellar yields would help to assess the total @xmath0 contribution . from the present analysis ,
a combination of different @xmath40 and @xmath0 process components is the most promising scenario to explain the solar and the metal - poor lepp .
+ in this view , additional processes have to be invoked to explain the observed trends . the intermediate neutron - capture process ( @xmath76 process ) may play a decisive role in the nucleosynthesis of the galaxy ( see @xcite , and references therein ) .
this issue deserves a more comprehensive analysis in order to establish whether the present discrepancies between gce predictions and the analyzed spectroscopic sample are confirmed by other elements .
future investigations with a more extended and homogeneous data - set of neutron - capture elements ( e.g. , cu , ga , ge , among weak @xmath0 process elements ; mo , ru , ag , pd , for weak @xmath40 process elements ; ce , which is expected to behave as ba ) are planned to distinguish between different nucleosynthesis processes . the improved quality and statistics of the ongoing surveys ( e.g.
, gaia - eso , segue , rave , apogee , hermes - galah ) will also assure a more solid picture of the galactic sub - structures .
first studies by @xcite , @xcite and @xcite evidenced that stars belonging to the thin and thick discs have different ages , kinematics , locations on the galactic plane , and chemical properties .
although a complex image of the galactic structure emerges from the most recent investigations ( e.g. , perturbation , accretion , merging or heating processes , radial migrations , gas flows ; see the review by @xcite ) , the main features observed in the solar neighborhood are confirmed by several surveys .
for instance , it has been well established in literature that various @xmath4-elements show clear separate behaviors in the thick and the thin disc .
+ in this view , a detailed analysis of our gce prescriptions is required in order to provide an improved representation of the chemical abundances observed in thick / thin disc stars ( see , e.g. , [ o / fe ] vs [ fe / h ] in fig . 1 by @xcite ) .
+ a preliminary study of the main gce ingredients ( e.g. , star formation rate , initial mass function , delay - time distribution function for snia , and the present uncertainties affecting snia stellar yields ) highlights that the theoretical interpretation of the thick / thin disc observed trends is largely affected by the poorly known snia scenario ( see , e.g. , @xcite , and references therein ) . specifically ,
our gce thick / thin disc predictions may be improved by assuming a factor of 2 uncertainty in the snia stellar yields adopted so far @xcite , coupled with an updated treatment of the delayed - time distribution ( dtd ) function ( as suggested by @xcite ) , in which we assume a dominant snia contribution starting from [ fe / h ] @xmath3 @xmath271 .
+ additionally , distinct enrichment histories of the thick / thin disc components may be obtained by adopting a different star formation rate , with a delayed thin disc phase ( as hypothesized by the two - infall model by @xcite ) . to this purpose
, we will provide a comprehensive investigation of several gce ingredients against a large number of elements in a forthcoming paper .
the underway study of snia yields for an extended metallicity grid will further constraint the disc chemical enrichment .
new snia inputs , combined with high - resolution studies of several chemical elements and the reliable age determination expected from astroseismology ( corot and kepler , @xcite ) , will provide a real breakthrough in the understanding of the galactic history .
we acknowledge the anonymous referee for offering helpful suggestions .
we thank d. yong for precious information about spectroscopic observations .
the present work has been supported by jina ( nd fund # 202476 ) .
numerical calculations have been sustained by b2fh association ( http://www.b2fh.org/ ) .
lodders , k. , palme , h. , & gail , h .- p .
2009 , landolt - brnstein - group vi astronomy and astrophysics numerical data and functional relationships in science and technology , edited by j.e .
trmper , 4b : solar system , 4.4
|
the solar @xmath0-process abundances have been analyzed in the framework of a galactic chemical evolution ( gce ) model .
the aim of this work is to implement the study by @xcite , who investigated the effect of one of the major uncertainties of asymptotic giant branch ( agb ) yields , the internal structure of the @xmath1c pocket .
we present gce predictions of @xmath0-process elements computed with additional tests in the light of the suggestions provided in recent publications .
+ the analysis is extended to different metallicities , by comparing gce results and updated spectroscopic observations of unevolved field stars .
we verify that the gce predictions obtained with different tests may represent , on average , the evolution of selected neutron - capture elements in the galaxy .
the impact of an additional weak @xmath0-process contribution from fast - rotating massive stars is also explored .
|
-type and @xmath17-type four - body interactions on alternate plaquettes of a square lattice , and the ( exact ) ribbon operators are the string - like products of @xmath18 and @xmath17 shown above .
these ribbon operators commute with the hamiltonian terms but anticommute with each other , thus identifying a phase with @xmath19 topological order.,title="fig : " ] in this section , we review the physical properties of anyon models that characterize a topological phase . of course
, not all topological phases have an associated anyon model , for example the cubic code,@xcite however we will tailor our method to those phases with anyonic excitations .
this discussion of anyon models will motivate our definition in the subsequent section of a _ ribbon operator _ for a spin model defined by some local hamiltonian @xmath20 on a lattice in two dimensions .
we will argue that this definition captures the essential features outlined below of a topologically ordered model by thinking about anyons as quasiparticle excitations of the spin model .
the simplest realization of this is kitaev s toric code,@xcite and we will refer to that model to clarify key features .
creating a pair of anyons , and moving them to the boundary of the lattice leaves no excitations , and so maps the vacuum to the vacuum . in the toric code , anyons are created and transported using string operators , as shown in fig . [ fig : toricstrings ] .
these operators commute with the hamiltonian in the bulk , but have excitations localized to the ends . in this way , if the anyons are moved off the edge of the lattice or are fused together , the system remains in its ground state . more generally , we expect anyonic excitations to be particle like , but to have nontrivial braiding relations . on sufficiently large length scales , it is expected that these braid statistics are governed solely by an anyon model.@xcite for this reason , the creation operators in this limit should be one dimensional with the particles localized around the end points .
away from a renormalization - group fixed point , we expect these operators to be dressed , and supported on slightly fattened regions , though still effectively one - dimensional .
more precisely , the width of these ribbons should be set by the microscopic details of the model , and should not scale with system size .
the family of local commuting projector codes ( lcpc ) generalizes the toric code and provides a concrete instantiation of these ideas .
the lcpcs are families of hamiltonians where the terms are pairwise commuting projectors onto an unfrustrated ground state . in these models
the string operators described above always exist.@xcite this class of models includes exactly solvable points of most known topological phases@xcite including the levin - wen string nets.@xcite away from these exactly solvable points , we can dress the string operators using quasiadiabatic continuation.@xcite given a point in a phase for which an initial string operator exists , such as an lcpc , we can dress this operator to any other point in the phase .
generically this operator will have extensive support , however the error made in truncating the width is exponentially small in @xmath21 , where @xmath22 is the width and @xmath23 is the correlation length .
thus string - like operators with width proportional to the correlation length can be found anywhere within the phase .
since the string operators act as creation operators for pairs of anyons , we argue that they should be only slightly entangling operators.@xcite these are operators whose schmidt rank is independent of the system size on any bipartition that cuts the string into two contiguous pieces . in the lcpc case ,
the string operators are known to be slightly entangling.@xcite since the quasiparticles are simply dressed by moving away from these fixed points , we expect this property to remain throughout the phase ; although the required schmidt rank can increase substantially , it will still be independent of the system size .
when we remove the dressed quasiparticles from the lattice ( by annihilating them or bringing them out to infinity ) , the system returns to its ground space , which is expected to obey the area law of eq .
( [ eqn : arealaw ] ) .
this also supports the slightly entangling nature of such operators .
some of the defining features of an anyon model are the relations that are encoded in the topological @xmath15 matrix and the @xmath24 matrices .
let @xmath25 label the anyonic particles in the anyon model describing the low energy excitations of the hamiltonian @xmath26 .
the @xmath24 matrices are defined by @xmath27 and the @xmath15 matrix is @xmath28 where the lines trace out the worldlines of the particles .
another approach is to consider the related @xmath29 matrix that can be obtained via the twist product.@xcite if the @xmath15 matrix is an object that captures dynamical information from braiding anyons in 2 + 1 dimensions , then the @xmath29 matrix is a static object that captures correlations of ground states in 2 + 0 dimensions .
if @xmath30 is an operator supported on an annulus that projects onto a state with total charge @xmath31 inside , then @xmath32 where @xmath33 is any ground state of the spin model . for bipartite operators @xmath34 and @xmath35 , the twist product with respect to the partition @xmath36
is defined as @xmath37 where the order is reversed on the @xmath38 region relative to the @xmath39 partition .
this is clearly closely related to eq .
( [ eqn : smatrix ] ) , where the worldline loops are replaced by the corresponding closed loop operators on the lattice . for abelian models where @xmath40 for all particle types @xmath31 , the @xmath15 and @xmath29 matrices coincide up to a multiple of @xmath41 .
finally , anyonic quasiparticles should be able to move around the lattice , and the state should not depend upon any smooth deformations of the path they take .
in particular , when they are fused back together or moved off the lattice , the system should return to a vacuum state .
this is realized in the toric code since the string operators can be deformed by dressing with local operators that are symmetries of the hamiltonian .
this preserves commutativity with @xmath26 and the @xmath29 matrix and clearly creates the same particle at the ends .
let us summarize the physical intuition that we ve gained in the previous section into a few simple properties that the string - like operators seem to possess in general .
these properties hold for known exactly solvable models with to .
given a two - dimensional quantum system with to , we expect that there are operators @xmath2 supported on one - dimensional strips through the bulk where the following four properties hold , at least approximately : 1 .
@xmath2 commutes with the bulk hamiltonian in the low energy sector.[item : commuting ] 2 .
@xmath2 is supported on a strip of spins with bounded width @xmath22.[item : bounded ] 3 .
distinct ribbons @xmath2 and @xmath42 should respect the data ( e.g. the @xmath24 or @xmath15 matrix ) for an underlying anyon model in the low energy sector.[item : logical ] 4 .
@xmath2 is smoothly deformable , that is , the ability to satisfy [ item : commuting]-[item : logical ] should not depend on the chosen strip of spins defining the support of @xmath2 , given sufficient width.[item : deformable ] since finding a low energy projector appears to be a challenging task , we make the following assumption . rather than asking
for [ item : commuting]-[item : deformable ] to hold only in the low energy sector , we ask for them to hold on the whole spectrum . at first glance
, this seems to be too strong for characterizing to away from an exactly solvable rg fixed point .
nonetheless , by making this assumption , we can arrive at an effective numerical procedure for detecting to in 2d models .
moreover , in sec .
[ s : heuristic ] , we provide a heuristic physical justification of this assumption , and some natural avenues for relaxing it in sec . [ s : future ] .
here we introduce what we call the ribbon operator method for detecting to .
the strategy is simple : we will write a cost function that tries to satisfy the above requirements simultaneously . since the support of a string - like operator is on a 1d strip , and is expected to be only slightly entangling , we will use the ansatz class of mpos@xcite to describe candidate operators .
we will then use the highly successful methods of 1d systems such as dmrg@xcite for optimizing over this ansatz class to find the lowest cost mpo .
we can then _ define _ a ribbon operator as any mpo supported on any one - dimensional strip of some fixed width . by drawing on the lattice hamiltonian and an underlying anyon model
, we can define a cost function quantifying the fitness of candidate ribbons given the above criteria .
a _ good _ ribbon operator then corresponds to a local minimum of the cost function .
there is clearly considerable scope within this method for how to use it .
the art will be to choose a cost function that is tractable to optimize and gives clear signals of to when appropriate , such that the distinction between which local minima are `` good '' and `` bad '' is obvious .
we do not claim to have a unique or best choice of cost function or optimization routine .
the remainder of this section details one particular choice of cost function , and as we show in sec .
[ s : results ] this particular choice performs quite well for a variety of simple models .
we now define our proposed cost function that quantifies how well a candidate ribbon satisfies the above criteria .
let @xmath20 be some local hamiltonian and @xmath2 some candidate ribbon with @xmath43 ( the schatten 2-norm , or frobenius norm @xmath44 ) and width @xmath22 , which ideally should be chosen to be comparable to the correlation length .
we will also use the symbol @xmath2 to denote the region of support of the ribbon @xmath2 since no confusion should result .
the first condition is that the ribbon should commute with the hamiltonian of the model .
this gives us two terms in the cost function .
we can quantify the violation of this condition using @xmath45}\right\rvert}^2 $ ] .
this can be decomposed into two distinct contributions by writing the hamiltonian as @xmath46 where the first term contains hamiltonian terms whose support is completely within the support of @xmath2 , the second contains those whose support has no overlap with @xmath2 and are hence supported entirely in the complementary region @xmath47 .
the final sum contains all those terms in the hamiltonian with support on both @xmath2 and @xmath47 .
clearly the second term trivially commutes with @xmath2 as a consequence of our second condition .
commutation with the first term can not be simplified further .
however , the term on the boundary @xmath48 can be simplified as follows .
let @xmath49 be some term on the boundary of the ribbon . then using the operator schmidt decomposition , we can split the operator @xmath49 into a sum over interior and exterior components @xmath50 where @xmath51 . using the orthonormality of the @xmath52 terms and the definition of the frobenius norm , we find that @xmath53}\right\rvert}^2&={\left\lvert\sum_k{\left[r , h_{j , k}^{\text{in}}\right]}\otimes h_{j , k}^{\text{out}}\right\rvert}^2\nonumber\\ & = \sum_k{\left\lvert{\left[r , h_{j , k}^{\text{in}}\right]}\otimes h_{j , k}^{\text{out}}\right\rvert}^2\nonumber\\ & = \sum_k { \left\lvert{\left[r , h_{j , k}^{\text{in}}\right]}\right\rvert}^2,\end{aligned}\ ] ] by appropriate grouping of boundary terms , we can ensure that @xmath54 .
( this is equivalent to applying the schmidt decomposition to @xmath55 instead of just one term . )
thus , we can express the boundary contribution as @xmath56}\right\rvert}^2\ , .
\label{eqn : boundarycost}\end{aligned}\ ] ] we find that the commutation condition together with the bounded width condition gives us a term in the cost function proportional to @xmath57}\right\rvert}^2 = { \left\lvert{\left[r , h_r\right]}\right\rvert}^2 + \sum_{j , k\in\partial r}{\left\lvert{\left[r , h_{j , k}^\text{in}\right]}\right\rvert}^2\,.\end{aligned}\ ] ] because commutation with the hamiltonian can always be achieved by choosing the ribbon to be the identity operator , we require a competing term to enforce the topological properties such as having a nontrivial @xmath15 or @xmath24 matrix . for the moment , suppose that we already have a given nontrivial ribbon operator @xmath42 that crosses @xmath2 as in fig .
[ fig : toricstrings ] .
( we will discuss how to relax this prior - knowledge assumption below . )
we can incorporate anyon data with the following term in the cost function , @xmath58}_\eta\bigr\|^2&={\left\lvertrl-\eta lr\right\rvert}^2\,,\label{eqn : etacomm}\end{aligned}\ ] ] where @xmath59 is a complex number different from 1 .
this choice of topological term in the cost function intuitively reminds us of the @xmath24 matrix relations of eq .
( [ eqn : rmatrix ] ) .
we can see that such a term is in fact sufficient to give a nontrivial @xmath24 matrix as follows . here
we specialize to an abelian model , but the discussion could be generalized . using the @xmath24 matrix relation for an abelian model with a fixed total anyon charge @xmath60
, we have the relation @xmath61 thus if @xmath62 then @xmath63 and @xmath64 can not both be 1 .
note that in the abelian case , @xmath65 can be written as@xcite @xmath66 , so @xmath59 is related to the @xmath15 matrix and is therefore gauge invariant ( invariant under a change of basis of the fusion space ) .
since the anyon charge @xmath10 is fixed already by the labels @xmath31 and @xmath67 in an abelian model , this switching relation should hold even when considering this as just a different operator product order , analogous to the case of a twist product for the @xmath29 matrix : @xmath68 since there are many equivalent strips on which @xmath42 and @xmath2 can be supported , the cost should be computed for all possible intersection regions .
this is achieved by summing @xmath69}_\eta\bigr\|^2 $ ] over all translates of @xmath42 , labelled @xmath70 , that lead to an inequivalent intersection with @xmath2 .
this translation invariance gives us an additional motivation for our choice of topological term : we can use translation - invariant mpos in our optimization routines . by comparison , an _ a priori
_ equally attractive term would be to encode the @xmath15 matrix relations , but then finite - size effects from the periodic boundary conditions might add additional complications to the numerics .
now , we define the @xmath59-_cost _ of a ribbon @xmath2 given some fixed secondary ribbon @xmath42 as @xmath71}\right\rvert}^2+&\sum_{j , k\in\partial r}{\left\lvert{\left[r , h^{\text{in}}_{j , k}\right]}\right\rvert}^2 \nonumber\\ & + \sum_{\mathcal{t}(l)}\bigl\|{\left[r , l\right]}_\eta\bigr\|^2\biggr),\label{eqn : costfunction}\end{aligned}\ ] ] where @xmath5 is the number of spins on which @xmath2 is supported .
note that due to the nonsymmetric nature of @xmath72}_\eta$ ] , it is convenient to define the @xmath59-cost of @xmath42 as @xmath73}\right\rvert}^2+&\sum_{j , k\in\partial l}{\left\lvert{\left[l , h^{\text{in}}_{j , k}\right]}\right\rvert}^2 \nonumber\\ & + \sum_{\mathcal{t}(r)}\bigl\|{\left[r , l\right]}_\eta\bigr\|^2\biggr),\end{aligned}\ ] ] where the appropriate hamiltonian terms are taken . here
the normalization term is chosen so that the cost of a ribbon is approximately independent of the volume .
one could consider summing these two terms to build a total cost function for the pair of ribbons , however we linearize the problem for a fixed @xmath42 ( or fixed @xmath2 ) and instead do an alternating minimization algorithm . in the case of the exact toric code , the form of the cost function is relatively simple and gives a concrete instantiation of the translation operation @xmath74 .
it is easy to check that the standard logical operators of this code have zero @xmath59-cost when @xmath75 , as expected .
we present the explicit cost function derivation in appendix [ s : costfunc ] for illustration .
although this cost function only involves a pair of ribbons , it is possible to incorporate braid relations with other particle types by adding terms of the form of eq .
( [ eqn : etacomm ] ) with additional ribbon operators . in this way
, it is possible to completely reconstruct the @xmath15 matrix of the anyon model up to a column permutation .
this ambiguity is due to the difficulty of identifying equivalence of operators on incomparable supports ( i.e. those with vertical and horizontal supports in fig .
[ fig : toricstrings ] ) .
we discuss possible avenues to removing this freedom in sec .
[ s : future ] .
as discussed in sec . [ s : ribbons ] , a key assumption which enables our method to be numerically efficient is that the commutation relations , which are expected to hold only on a topologically ordered ground space , hold on the entire spectrum . while this appears to be a very strong assumption , we believe that our method should continue to work even when this assumption breaks . in this section , we will provide a physically motivated , heuristic justification for this assumption . at high energies ,
we expect a gas of short - range interacting anyons .
when ribbon operators are used to create additional anyons , braid them , and fuse them in a specific way , the resulting process will in general be affected by the presence of the background anyon gas .
specifically , the expectation value of the braid can be affected by 1 ) dynamical phases acquired from short range interactions with the background anyons , and 2 ) topological terms caused by the background anyons enclosed in the braid . to eliminate these effects , we need to make sure that 1 ) there are no background anyons near the support of the ribbon operator , and 2 ) the topological charge enclosed in the braid is trivial .
it happens that both of these conditions can be enforced with small width operators .
indeed , 1 ) requires projecting on the local ground state ( lgs ) , i.e. the subspace with no charges .
similarly , it is possible to fix the topological charge of a region using an operator which acts only on the boundary of this region and is described by a finite bond dimension mpo.@xcite let @xmath76 denote the corresponding projector .
suppose that @xmath2 is an mpo ribbon operator which obeys the right commutation relations on the ground space , but not on the entire spectrum .
then , @xmath77 should be a slightly wider and larger - bond - dimension mpo ribbon operator which obeys the commutation on the entire spectrum . thus , while our approach can not find @xmath2 , it should be able to find @xmath78 without any assumptions .
we stress that the above argument is not key to our method , and no ground state projectors are incorporated directly in our algorithm or in the numerical results presented in sec .
[ s : results ] .
one could alter the method to explicitly make use of information about the low energy space .
we discuss this further in sec .
[ s : future ] .
in many exactly solvable models , an analytic form for the ribbon operators can be found . in more general models , for example ones with a noncommuting hamiltonian
, the ribbons must be found numerically . in this section ,
we describe how to find ribbon operators using dmrg.@xcite we parameterize a width @xmath22 ribbon by a block - translationally invariant infinite mpo .
this mpo is `` snaked '' along the support of the ribbon to cover the two - dimensional region .
that is , the coordinates of the ribbon are treated in a linearized lexicographic order moving along the ribbon lengthwise .
this mpo can be vectorized , i.e. interpreted as a vector instead of a matrix .
an mpo is also built for the cost function , with the same snaking pattern , and standard dmrg can then be applied to find the ribbon that minimizes the constraints .
this vectorization procedure corresponds to the following transformation , written in tensor network diagram notation : @xmath79 given any operator @xmath80 in mpo form , we can explicitly write the form of the constraint term @xmath81}\right\rvert}^2 $ ] for one site of the mpo as @xmath82 by concatenating this expression and closing the boundaries appropriately , we see that this is equivalent to the expectation of some mpo for some matrix product `` state '' given by the vectorization of @xmath2 .
the total cost function can be obtained by summing up the mpos defining each cost term , corresponding to direct sum on the mpo matrices .
the cost function mpo will not generally correspond to a local hamiltonian , which is the regime where dmrg is usually applied . indeed , due to the squaring required on the internal commutator
, we will have terms like @xmath83 for every pair @xmath84 . despite this potential challenge
, we will demonstrate that dmrg is an effective algorithm for this minimization .
we attempt to optimize @xmath85 for various choices of @xmath59 .
in particular , we restrict ourselves to the unit disk as this is sufficient for the models below , although extending to other @xmath59 does not change any of the results obtained . when attempting to optimize a pair , we used alternating minimization .
first @xmath86 was minimized in the presence of a random @xmath87 .
then @xmath88 was optimized given the @xmath89 obtained from the previous optimization .
the @xmath87 obtained was then used to seed the next iteration .
this process was repeated until convergence . for each choice of @xmath59 , several ( usually 5 or 10 ) ribbons or pairs of ribbons were optimized from random initial pairs and the lowest cost pair was retained .
this proved necessary to avoid premature convergence to a local minimum , particularly in the alternating minimization , leading to a pair that was orthogonal to the optimal solution . in most cases
a total of 5 - 10 restarts appeared sufficient to ensure that an optimal pair was obtained .
for each value of @xmath59 , an independent random start was generated , rather than using a warm start from the previous nearby values of @xmath59 .
in this section we use ribbon operators to study several standard models that exhibit both topologically ordered and trivial phases .
we demonstrate the efficacy of ribbon operators in identifying to both at exactly solvable points and away from such points . model with @xmath90 .
a pair of ribbons were optimized on two @xmath91 strips with bond dimension @xmath92 for the paramagnet and @xmath93 in the other cases .
we observe two distinct failures to observe topological order . in the case of the paramagnet
, a zero - cost ribbon can be found for all @xmath94 .
in fact , this can be extended to all @xmath59 . in the other cases ( closeup in inset plot ) , there is no @xmath59 for which a nontrivial low - cost ribbon can be obtained .
it is possible that this could be due to insufficient width , although we have not observed this to be the case .
note that averaging over many trials ( 20 ) to give a typical picture of the algorithms performance . in the case
where the variational minimum is used instead , still no signal of topological order is observed.,title="fig : " ] we define the @xmath95 quantum double - ising ( qdi@xmath96 ) model as follows.@xcite let @xmath97 be the bicolorable lattice shown in fig .
[ fig : toricstrings ] .
place a qudit ( @xmath98 level quantum system ) at each vertex and define the generalized unitary pauli operators @xmath18 and @xmath17 such that @xmath99 , @xmath100 , and @xmath101 .
we define plaquette operators @xmath31 and @xmath67 as @xmath102 on the dark and light plaquettes respectively . the hamiltonian is given by [ eqn : qdi ] @xmath103 where at each plaquette
@xmath104 we have the operators @xmath105 the hamiltonian contains the topological terms in eq .
( [ subeq : to ] ) with strength @xmath106 , a transverse onsite @xmath18 field in eq .
( [ subeq : pm ] ) with strength @xmath107 , and a ferromagnetic ising - type term in eq .
( [ subeq : fm ] ) with strength @xmath108 .
model of eq .
( [ eqn : qdi ] ) with @xmath109 using bond dimension 5 and plotted with @xmath90 .
the other ribbon was kept as a string of @xmath17s as shown in fig .
[ fig : toricstrings ] .
+ b ) cost for ribbons found using alternating minimization on the qdi@xmath110 model with @xmath111 . in the @xmath112 case
a pair of @xmath91 , bond dimension 1 ribbons were sought , in the other cases , the pair consisted of a @xmath91 and a @xmath113 ribbon with bond dimensions @xmath92 and @xmath114 respectively . at the exactly solvable points ( @xmath112 ) , we see a dramatic cost decrease at @xmath98th roots of unity , signalling both the presence of and type of topological order present in these models .
this feature remains even with the addition of an @xmath18 field or ising type term which destroys the solvability of the model .
+ c ) cost of a width 4 ribbon of bond dimension 2 in the qdi@xmath110 model . the cost obtained from our numerical optimization is smaller than that of the fixed point string for all ( nonzero ) values of @xmath107 .
this demonstrates the nontrivial nature of the ribbon operators we obtain , even in this perturbative regime.,title="fig : " ] + model of eq .
( [ eqn : qdi ] ) with @xmath109 using bond dimension 5 and plotted with @xmath90 .
the other ribbon was kept as a string of @xmath17s as shown in fig .
[ fig : toricstrings ] .
+ b ) cost for ribbons found using alternating minimization on the qdi@xmath110 model with @xmath111 . in the @xmath112 case a pair of @xmath91 ,
bond dimension 1 ribbons were sought , in the other cases , the pair consisted of a @xmath91 and a @xmath113 ribbon with bond dimensions @xmath92 and @xmath114 respectively . at the exactly solvable points ( @xmath112 ) , we see a dramatic cost decrease at @xmath98th roots of unity , signalling both the presence of and type of topological order present in these models .
this feature remains even with the addition of an @xmath18 field or ising type term which destroys the solvability of the model .
+ c ) cost of a width 4 ribbon of bond dimension 2 in the qdi@xmath110 model . the cost obtained from our numerical optimization is smaller than that of the fixed point string for all ( nonzero ) values of @xmath107 .
this demonstrates the nontrivial nature of the ribbon operators we obtain , even in this perturbative regime.,title="fig : " ] + model of eq .
( [ eqn : qdi ] ) with @xmath109 using bond dimension 5 and plotted with @xmath90 .
the other ribbon was kept as a string of @xmath17s as shown in fig .
[ fig : toricstrings ] . + b ) cost for ribbons found using alternating minimization on the qdi@xmath110 model with @xmath111 . in the @xmath112 case
a pair of @xmath91 , bond dimension 1 ribbons were sought , in the other cases , the pair consisted of a @xmath91 and a @xmath113 ribbon with bond dimensions @xmath92 and @xmath114 respectively . at the exactly solvable points ( @xmath112 ) , we see a dramatic cost decrease at @xmath98th roots of unity , signalling both the presence of and type of topological order present in these models .
this feature remains even with the addition of an @xmath18 field or ising type term which destroys the solvability of the model .
+ c ) cost of a width 4 ribbon of bond dimension 2 in the qdi@xmath110 model .
the cost obtained from our numerical optimization is smaller than that of the fixed point string for all ( nonzero ) values of @xmath107 .
this demonstrates the nontrivial nature of the ribbon operators we obtain , even in this perturbative regime.,title="fig : " ] when @xmath115 , this is a topologically ordered commuting model which reduces to the toric code when @xmath116 .
this model also captures a number of other models , including the paramagnet ( @xmath117 ) and a ferromagnet ( @xmath118 ) .
these latter correspond to distinct topologically trivial phases . at the fixed point of each of these phases , the model is exactly solvable . in the @xmath116 case , the generic model with @xmath119
is given by the hamiltonian @xmath120 at the rg fixed points of these phases ( corresponding to either @xmath112 or @xmath121 ) , we can set @xmath122 and @xmath123 , where @xmath124/@xmath125 correspond to the @xmath126/@xmath127 eigenstates of the relevant operator ( i.e. the eigenstates of @xmath18 when @xmath121 and of @xmath17 when @xmath112 ) .
these ribbons commute with the hamiltonian .
since @xmath128 , their @xmath59-commutation relation is not unique , so they do not correspond to good ribbon operators .
we therefore expect to be able to find zero - cost ribbons for all @xmath59 at these points .
we note that constraining the ribbons to be unitary would be a natural way to eliminate these spurious solutions in the trivial phases , but this is computationally expensive ; we discuss this point further in sec .
[ s : future ] .
costs for ribbons at various non - topological points in the qdi@xmath96 model are shown in fig .
[ fig : topologicallytrivialcostdata ] .
we observe two distinct ways in which trivial order can be signalled .
as we have already discussed , one of these corresponds to mutually annihilating operators as in the para- and ferromagnetic phases . in the other case
we simply fail to observe any low cost ribbons which form a nontrivial algebra .
there are two explanations for this .
either there are no ribbon operators to be found or the chosen width @xmath22 is too narrow . model with @xmath109 .
a pair of @xmath91 ribbons with bond dimension 1 were optimized .
we observe a signal of topological order only at @xmath129.,title="fig : " ] model with @xmath130 where @xmath59 runs over points in the complex unit disk .
a pair of ribbons with @xmath91 and @xmath113 and bond dimensions 1 and 5 respectively were optimized and the total cost is shown .
we observe a signal of topological order at @xmath131 .
note that fig .
[ fig : zqdicostdata]b corresponds to the edge of this disk .
the raw data is displayed , but some contours were subjected to gaussian smoothing for clarity in the nearly flat region where the noise in the data and the gradient of the function are comparable in magnitude.,title="fig : " ] the phase which includes the point @xmath109 is topologically ordered , and we expect it to show a signal of @xmath95-topological order .
we therefore expect to observe low cost ribbons only when @xmath90 is a @xmath98th root of 1 .
this property is expected to persist away from the exactly solvable point . in fig .
[ fig : zqdicostdata ] we show the data obtained at the fixed points of the @xmath132 and @xmath19 models and when an @xmath18 field is turned on in the latter . as expected , at the fixed point a very strong signal
is observed . using a large density of points
we see that the cost drops dramatically over a very narrow region around the @xmath98th root of unity .
when we turn on the field term , the signal decreases in magnitude but is still unmistakable , with more than two orders of magnitude in total contrast . despite the clear signal , we observe that the value of the cost function at the minimum is comparable to the ( square of ) the perturbation strength .
this may lead one to suspect that the method is simply returning the analytically obtainable fixed point ribbon . to refute this hypothesis , in fig .
, we compare the cost of numerically obtained ribbons to those of the fixed point strings .
encouragingly , we see that the cost is smaller for the optimized ribbon , indicating that the method is finding a nontrivial result . in fig .
[ fig : z3diskcostdata ] , we present cost data for the fixed point of the qdi@xmath133 model over the entire unit disk .
as expected , low cost ribbons can only be found around the third root of unity , with no minima occurring inside the disk .
this property persists even when an @xmath18 field is turned on as shown in fig .
[ fig : z2diskcostdata ] .
this figure shows the cost of ribbons in the qdi@xmath110 model with @xmath130 . even at this nonintegrable point
, no low cost ribbons can be found anywhere away from the @xmath98th roots of unity .
it is interesting to consider the 2d quantum compass ( qc ) or bacon - shor model@xcite in the context of ribbon operators .
this model is defined on a square lattice @xmath97 with hamiltonian @xmath134 this model fails to exhibit many of the features expected of a topologically ordered model . for example , the ground state degeneracy does not depend on the underlying lattice topology .
moreover , this appears to be a gapless model with exponential ground state degeneracy.@xcite thus , this hamiltonian is usually described as not having topological order .
however , this model can be used to encode a qubit which is protected against arbitrary local errors , having a code distance extensive in the ( linear ) lattice size,@xcite a property which is clearly required of a topological code .
this qubit can be defined by the pair of logical operators @xmath135 and @xmath136 .
these operators commute with each hamiltonian term , and intersect at a single point , hence they anticommute , @xmath137_{-1}=0 $ ] , and therefore almost fulfill the conditions we give to be good ribbon operators . we will see below that they are rigid objects that exist due to special lattice symmetries , and they are not deformable . the hamiltonian eq .
( [ eqn : qcmodel ] ) also commutes with all elements of the _ stabilizer group _ @xmath138 it is therefore clear that @xmath139(@xmath140 ) is related to @xmath141(@xmath142 ) by application of a stabilizer element , and so defines the same encoded qubit in the ground space .
hence , the logical operators can be moved along to parallel strips . to investigate topological order in the qc model ,
we seek to find logical operators supported on rotated regions as shown in fig .
[ fig : rotatedribbons ] or widened versions thereof .
we compare this to the @xmath19 quantum double model where topological order is well understood . for two models : a ) the quantum compass model and b ) the qdi@xmath110 model .
the cost is parameterized with @xmath143 , and all ribbons were found using bond dimension 5 .
+ a ) cost of ribbons in the quantum compass model of eq .
( [ eqn : qcmodel ] ) . because ribbons in this model are sensitive to their support
, we conclude that they do not represent genuine topological order .
+ b ) cost of ribbons in the qdi@xmath110 model of eq .
( [ eqn : qdi ] ) using @xmath109 .
note that the unit cell in the rotated case is @xmath144 rather than @xmath145 for convenience .
low cost ribbons can be found on both unrotated and rotated supports indicating genuine topological order.,title="fig : " ] + for two models : a ) the quantum compass model and b ) the qdi@xmath110 model .
the cost is parameterized with @xmath143 , and all ribbons were found using bond dimension 5 .
+ a ) cost of ribbons in the quantum compass model of eq .
( [ eqn : qcmodel ] ) . because ribbons in this model are sensitive to their support
, we conclude that they do not represent genuine topological order .
+ b ) cost of ribbons in the qdi@xmath110 model of eq .
( [ eqn : qdi ] ) using @xmath109 .
note that the unit cell in the rotated case is @xmath144 rather than @xmath145 for convenience .
low cost ribbons can be found on both unrotated and rotated supports indicating genuine topological order.,title="fig : " ] in fig .
[ fig : rotatedribbonscostdata ] we show the cost for ribbons obtained in these two models where the vertical ribbons are fixed to be @xmath146 and a bond dimension @xmath147 ribbon is optimized with both horizontal and rotated support . as discussed above , zero cost ribbons can be obtained for both models in the unrotated case when @xmath131 .
thus we see that each model supports a topologically encoded qubit .
once we rotate the ribbons , we find that in both cases the ribbon disappears when we restrict to @xmath91 . in the qdi@xmath110
, the ribbon is recovered once we allow the support to grow to @xmath113 , however in the qc model there is no indication of a good ribbon existing even when we allow ribbons of width @xmath148 .
this indicates that the low cost ribbons in the qc model are anomalies associated to lattice symmetries instead of genuine signatures of topological order .
the honeycomb model@xcite is a frustrated spin model on the honeycomb lattice with hamiltonian @xmath149 where the @xmath18 , @xmath150 , and @xmath17 links refer to the three orientations of edges on the lattice . for @xmath151
, this model supports a phase with the same order as the @xmath19 quantum double model.@xcite writing a pair of anticommuting logical operators for this phase is a nontrivial task however , since they can not both be symmetries of the hamiltonian and will hence not have completely localized support .
we can instead attempt to describe them with ribbons having some width @xmath22 .
topological order known to be present in this phase are visible .
as expected , the cost decreases as the allowed width or bond dimension is increased , however the cost rapidly saturates with increasing bond dimension . + a ) cost for a bond dimension 5 ribbon on the honeycomb model of eq .
( [ eqn : honeycombh ] ) with @xmath152 using @xmath143 .
a second ribbon can be obtained analytically , and was fixed as a product of @xmath17s along the @xmath18 and @xmath150 links of the lattice .
the dip at @xmath153 is characteristic of @xmath19 order , and indicates the topological nature of this phase .
+ b ) cost for a width 2 ribbon at @xmath131 on the honeycomb model with @xmath152 as a function of bond dimension .
the cost rapidly saturates to a minimum , justifying our assumption that only small bond dimension is required .
+ c ) as we increase the strength of @xmath154 , we eventually cross a phase transition to a gapless phase at @xmath155 ( indicated by @xmath156 ) . approaching this point
, the signal of @xmath157 topological order essentially vanishes . here
the strength of @xmath106 is increased in steps of @xmath158 from @xmath119 ( @xmath159 ) to @xmath111 ( @xmath160 ) .
note that the line marked by @xmath161 corresponds to plots @xmath162 and a bond dimension of @xmath163 is used for a width @xmath148 ribbon.,title="fig : " ] + topological order known to be present in this phase are visible . as expected
, the cost decreases as the allowed width or bond dimension is increased , however the cost rapidly saturates with increasing bond dimension . + a ) cost for a bond dimension 5 ribbon on the honeycomb model of eq .
( [ eqn : honeycombh ] ) with @xmath152 using @xmath143 .
a second ribbon can be obtained analytically , and was fixed as a product of @xmath17s along the @xmath18 and @xmath150 links of the lattice .
the dip at @xmath153 is characteristic of @xmath19 order , and indicates the topological nature of this phase .
+ b ) cost for a width 2 ribbon at @xmath131 on the honeycomb model with @xmath152 as a function of bond dimension .
the cost rapidly saturates to a minimum , justifying our assumption that only small bond dimension is required .
+ c ) as we increase the strength of @xmath154 , we eventually cross a phase transition to a gapless phase at @xmath155 ( indicated by @xmath156 ) .
approaching this point , the signal of @xmath157 topological order essentially vanishes . here
the strength of @xmath106 is increased in steps of @xmath158 from @xmath119 ( @xmath159 ) to @xmath111 ( @xmath160 ) .
note that the line marked by @xmath161 corresponds to plots @xmath162 and a bond dimension of @xmath163 is used for a width @xmath148 ribbon.,title="fig : " ] + topological order known to be present in this phase are visible . as expected
, the cost decreases as the allowed width or bond dimension is increased , however the cost rapidly saturates with increasing bond dimension . + a ) cost for a bond dimension 5 ribbon on the honeycomb model of eq .
( [ eqn : honeycombh ] ) with @xmath152 using @xmath143 .
a second ribbon can be obtained analytically , and was fixed as a product of @xmath17s along the @xmath18 and @xmath150 links of the lattice .
the dip at @xmath153 is characteristic of @xmath19 order , and indicates the topological nature of this phase .
+ b ) cost for a width 2 ribbon at @xmath131 on the honeycomb model with @xmath152 as a function of bond dimension .
the cost rapidly saturates to a minimum , justifying our assumption that only small bond dimension is required .
+ c ) as we increase the strength of @xmath154 , we eventually cross a phase transition to a gapless phase at @xmath155 ( indicated by @xmath156 ) . approaching this point
, the signal of @xmath157 topological order essentially vanishes . here
the strength of @xmath106 is increased in steps of @xmath158 from @xmath119 ( @xmath159 ) to @xmath111 ( @xmath160 ) .
note that the line marked by @xmath161 corresponds to plots @xmath162 and a bond dimension of @xmath163 is used for a width @xmath148 ribbon.,title="fig : " ] in fig .
[ fig : honeycombcostdata ] , we investigate ribbon operators for the honeycomb model . as before , we observe the characteristic dip at @xmath131 signalling a topologically ordered phase ( fig . ) .
we also observe the effect of increasing the allowed ribbon width in fig . .
as expected the cost goes down as the width increases , although this is not a smooth decrease .
we suggest that this is due to the nature of the widening .
each time the width is increased by three , an additional plaquette lies completely within the support of @xmath2 .
these plaquettes commute with the hamiltonian , and so correspond to conserved quantities . in the perturbative regime , these plaquettes become those in the toric code model , which are associated with the location of anyonic charge .
we note that the correlation length for @xmath152 is approximately @xmath164,@xcite so all of the widths correspond to several correlation lengths .
we also investigate how restricting the bond dimensions affects the cost . in fig . , we use a fixed width ( 3 ) and a fixed strength of @xmath152 . at this point
, we see that the cost rapidly saturates to a minimum for a small bond dimension of 4 - 5 .
this justifies the relatively small bond dimensions used in this work .
finally , we examine the effect of increasing the strength of @xmath154 in fig . .
when @xmath165 , the model undergoes a transition from a phase supporting @xmath19 topological order to a gapless phase .
the latter supports ising anyons if a time reversal symmetry breaking term is added.@xcite although we do not see a strong signal , such as a discontinuity , at this point , we see that the dip associated with @xmath131 becomes very shallow .
we expect the width of the required ribbons to be comparable to the correlation length , which we expect to increase as we move out of the perturbative regime .
thus the width @xmath148 ribbon used here is expected to be inadequate for large @xmath166 even within the phase .
it remains to be seen whether ribbon operators can be used to identify phase transitions , possibly using a scaling analysis of ribbon width and bond dimension .
one of the motivations for the ribbon operator methods is an application to quantum error correcting codes . in this section
we show how nontrivial ribbon operators with sufficiently low cost can be used to certify the existence of an approximate topological quantum error correcting code in the ground space of the model .
the ribbon operators function as approximate logical operators for these approximate codes . for gapped phases
, this will also provide a natural justification for the method , although to apply this justification to the numerical results of sec .
[ s : results ] would require similar proofs with weaker assumptions than we are currently able to make here .
let @xmath26 be a hamiltonian with gap @xmath167 and suppose that for some @xmath168 , @xmath169 and @xmath59 , there exist ribbon operators @xmath42 and @xmath2 such that @xmath170}\right\rvert_\text{op}}\leq\epsilon$ ] , @xmath171}\right\rvert_\text{op}}\leq\epsilon$ ] , and @xmath172}_\eta}\bigr\|_{\text{op}}\leq \delta$ ] . here
@xmath173 , we restrict to hermitian ribbons , and we let @xmath174 denote the operator norm .
let @xmath175 be some ground state of @xmath26 with energy @xmath176 .
then the expected energy of @xmath177 is @xmath178}{|{g}\rangle}\nonumber\\ & \leq e_0+\epsilon.\end{aligned}\ ] ] thus , @xmath179 is a low energy state , and we can interpret @xmath2 as a generator for a unitary approximate logical operator . it is important to check that @xmath2 is not mapping out of the ground space , so we need to ensure that there is a large overlap .
let @xmath180 denote the ground space projector for @xmath26 .
by assumption the model has a spectral gap @xmath167 , so we find that @xmath181 rearranging this , and noting that @xmath176 can always be set to @xmath182 , we obtain @xmath183 thus , the action of the ribbon operators approximately preserves the ground space as long as they approximately preserve the norm of the individual ground states ( @xmath184 ) .
if this condition does not hold , @xmath2 is annihilating the ground space .
given an almost commuting pair of hermitian operators @xmath2 and @xmath26 , we can perturb to an exactly commuting hermitian pair @xmath185 and @xmath186 such that the following holds:@xcite @xmath187 and @xmath188 , where @xmath189 can be taken to be at most @xmath190 .
it is convenient to work with @xmath186 and @xmath185 since they can be simultaneously exactly diagonalized .
we need to check that the assumed twisted commutation relation between @xmath2 and @xmath42 is approximately maintained when we consider @xmath185 and @xmath42 . @xmath191 rearranging and using standard inequalities , we find @xmath192 and so the twisted commutator is approximately preserved .
the same argument shows that @xmath42 approximately commutes with @xmath186 and thus approximately preserves its ground space .
we now check that the action of @xmath42 on eigenstates of @xmath185 maps to a nearly orthogonal state and can therefore be used to manipulate information in the code space .
since @xmath186 and @xmath185 exactly commute , we can find a joint eigenstate @xmath193 such that @xmath194 and @xmath195 , then @xmath196 and therefore latexmath:[\ ] ] notice that @xmath218 .
care should be taken with this term since @xmath72}_\eta$ ] is not symmetric in its arguments .
|
we introduce a numerical method for identifying topological order in two - dimensional models based on one - dimensional bulk operators . the idea is to identify approximate symmetries supported on thin strips through the bulk that behave as string operators associated to an anyon model .
we can express these ribbon operators in matrix product form and define a cost function that allows us to efficiently optimize over this ansatz class .
we test this method on spin models with abelian topological order by finding ribbon operators for @xmath0 quantum double models with local fields and ising - like terms . in addition , we identify ribbons in the abelian phase of kitaev s honeycomb model which serve as the logical operators of the encoded qubit for the quantum error - correcting code .
we further identify the topologically encoded qubit in the quantum compass model , and show that despite this qubit , the model does not support topological order .
finally , we discuss how the method supports generalizations for detecting nonabelian topological order . despite the apparent simplicity of quantum spin models
, they can exhibit a wide variety of interesting and potentially useful phenomena .
these range from conventional magnetic order to the more novel topological@xcite and symmetry - protected@xcite and symmetry - enriched@xcite topological orders which are of interest in both condensed matter physics@xcite and quantum information theory.@xcite these states are disordered in the sense of landau - ginzburg - wilson , however they do exhibit properties distinct from the usual disordered phases .
for example , topological phases possess quasiparticle excitations , known as anyons , whose braid relations can be far more exotic than those of fermions or bosons.@xcite these phases also have ground state degeneracy which depends on the topology of the lattice.@xcite this protected degeneracy has prompted the investigation of topologically ordered models as quantum memories.@xcite quantum information stored in the degenerate subspace can be protected from arbitrary local noise when error correction techniques are employed.@xcite distinguishing topological phases can be an especially challenging task precisely because of their topological nature : there is no broken symmetry and no local order parameter signalling the phase transition.@xcite there has been a large amount of previous work which attempts to identify topological order ( to ) .
the existing key techniques , such as the topological entanglement entropy,@xcite the entanglement hamiltonian@xcite , topological degeneracy@xcite and associated properties,@xcite symmetries of particular representations of the ground states,@xcite and specific properties of particular to states,@xcite have been highly successful in various domains of applicability .
we review these methods below .
a common feature of these methods is that they utilize the ground state of the model , which is unfortunately a challenging computational task in general .
here we propose a numerical method that we call the _ ribbon operators _ method for identifying to in the ground state of a given 2d hamiltonian using only the hamiltonian , without reference to the ground state .
we reduce the search for to to a 1d problem through the bulk of the material , and we present a variational approach based on standard dmrg@xcite to identify certain operators the ribbon operators supported in the bulk that satisfy the commutation relations relevant for a candidate anyon model .
we demonstrate the power of this approach by identifying to in both integrable and non - integrable models , and contrast this with topologically trivial hamiltonians .
we also demonstrate the ability to identify topologically encoded qubits and logical operators of quantum error - correcting codes , even in a non - integrable model .
all of our calculations are focused on the case of abelian to in spin models , however the ribbon operators method suggests several natural extensions beyond this case , which we leave open for future work .
the following subsection reviews prior approaches , while sec .
[ s : anyons ] reviews anyon models .
the expert reader can skip to sec .
[ s : ribbons ] .
one important tool is the _ topological entanglement entropy _
( tee)@xcite of the ground state wave function .
given the reduced density matrix @xmath1 of some many - body ground state on a region @xmath2 , the von neumann entropy @xmath3 typically obeys the area law @xmath4 where @xmath5 and @xmath6 are the number of spins in the region @xmath2 and on the boundary of the region @xmath2 respectively , and @xmath7 is the tee . in a topologically
ordered model , a physical argument suggests that @xmath8 , where @xmath9 is the quantum dimension of the anyon with charge @xmath10 in the associated anyon model . the tee is clearly nonzero if the ground state is topologically ordered,@xcite so @xmath11 can be used as a signal of such ordering.@xcite one can compute @xmath11 by obtaining an explicit ground state wavefunction , for example as a peps .
additionally , a rnyi entropy variant of @xmath11@xcite can be computed by sampling the ground state wavefunctions using quantum monte carlo.@xcite there are two major challenges associated with this approach .
firstly , there exist examples of topologically trivial states for which computing @xmath11 leads to nonzero values.@xcite secondly , any two topological phases whose associated anyon models have the same total quantum dimension @xmath12 will have the same tee , and so it can not be used to distinguish them .
this second problem has already led to difficulty in fully identifying the phase of physically interesting models including the heisenberg antiferromagnet on the kagome lattice , a model thought to describe the low - energy physics of several naturally occurring and synthetic minerals.@xcite a less coarse approach is to investigate the full entanglement spectrum , that is , the spectrum of the effective hamiltonian defined by @xmath13.@xcite it has been suggested that the universal properties of @xmath14 are intimately linked to the structure of the physical edge state of the model , however @xmath14 can undergo phase transitions without the physical model doing so,@xcite and at least for models in the same phase as a string - net model@xcite the spectrum is expected to contain the same universal information as the tee.@xcite one of the characteristic features of topologically ordered models is the topology - dependent ground state degeneracy.@xcite after obtaining a full set of ground states , one can observe transitions out of topological phases via loss of topological degeneracy,@xcite and , in the topological phase , compute the @xmath15 and @xmath16 matrices defining the braiding relations in the associated anyon theory.@xcite unfortunately , demonstrating topology - dependent ground state degeneracy is not sufficient to ensure robust topological order.@xcite given a projected entangled pair state ( peps ) description of the ground state , one can identify the topological order by understanding the symmetry properties of the parent hamiltonian@xcite or the environment tensor.@xcite one can also use the peps formalism to identify matrix product operators ( mpos ) that ` pull through ' the peps tensors on the virtual level.@xcite these mpos are in close analogy to the physical operators that act to create and transport anyons . as we will describe , these physical operators are central to our approach .
finally , given access to a ground state , specific structure such as certain correlation functions or distribution of bond energies can provide evidence for topological ordering.@xcite as a prerequisite for each of these methods , one must obtain an efficient description of the ground space ( e.g. via tensor networks ) or obtain expectations with respect to ground states ( e.g. via monte carlo sampling ) .
tensor network ( tn ) methods have proven very useful for this purpose,@xcite however in many cases the result does not conclusively determine the topological order.@xcite the tn states obtained suffer from several drawbacks .
in particular , they are usually computed on infinite cylinders of small circumference .
they can also be biased towards low - entanglement states.@xcite additionally , properties of a ground state alone are not sufficient to identify a gapped topological phase .
for example , a one can obtain a gapless hamiltonian sharing the toric code ground space.@xcite of these methods , the ones closest in spirit to our current approach are the tensor network - based methods,@xcite however our approach differs substantially in that we do not require a peps description of the ground state wavefunction .
we instead variationally create tensor network representations of certain ribbon operators that are supported on 1d strips through the truly two - dimensional bulk ; this dimensional reduction is what makes our method numerically tractable .
in contrast to state - based variational approaches , where a similar reduction is often included for numerical convenience , we will argue that in our operator - based approach this reduction is an expected feature of the operators .
additionally , if one wishes to use the topologically protected ground space as a quantum memory , knowledge of these ribbon operators is required for information manipulation and extraction .
thus , this method could be used to augment a state - based approach to obtain this additional data .
this paper is organized as follows . in sec .
[ s : anyons ] , we review some of the features of anyon models which describe the low energy excitations of topologically ordered spin models . in sec . [
s : ribbons ] , we define a ribbon operator , and use the properties of the anyons to define a cost function which quantifies how well a candidate ribbon realizes their behavior .
we then describe how the cost function can be minimized numerically in sec .
[ s : algorithm ] . using this algorithm , in sec .
[ s : results ] we obtain ribbon operators in a number of topologically ordered and topologically trivial models .
these numerics demonstrate that the cost function we define allows identification of abelian topological order in nonintegrable spin models . in sec .
[ s : logicaloperators ] , we prove that , under certain assumptions , ribbon operators can be used as approximate logical operators in quantum error correcting codes .
we conclude in sec .
[ s : future ] with a discussion of extensions of the method to nonabelian topological order and more complex spin models .
|
x - ray techniques have long been applied to image the electronic charge density of atoms , molecules , and materials @xcite .
recently - developed x - ray free electron laser sources , which generate short ( attosecond ) , intense pulses , open up numerous potential applications for high temporal and spatial resolution studies @xcite .
one exciting application is the determination of molecular structure by x - ray diffraction of nanocrystals @xcite avoiding the crystal growth process which is often the bottleneck in structure determination @xcite ; it may take decades to crystallize a complex protein .
it is much easier to grow nanocrystals than the many - microns - sized samples required by conventional crystallography .
this has been demonstrated experimentally in nanocrystals for the water splitting photosynthetic complex ii @xcite , a mimivirus @xcite , and a membrane protein @xcite .
extending this idea all the way to the single - molecule level , totally removing crystal " from crystallogrpaphy is an intruiging possibility @xcite .
obtaining a protein structure by scattering from a single molecule is revolutionary .
many obstacles need to be overcome to accomplish this ambitious goal .
for instance , the molecule will typically break down when subjected to such high fluxes .
however , it has been argued @xcite that , for sufficiently short pulses , the scattering occurs prior to photon damage , so that this should not affect the measured charge density .
this point is still under debate . in this paper
, we consider the scattering of a broad - band x - ray pulse from a system composed of a single , few , or many molecules prepared in a superposition of electronic states and show under what conditions the signal may be described solely by the time - dependent charge density .
we assume that the x - ray pulse is off - resonance from any material transitions and that the pulse is sufficiently short so that the electrons in the sample do not appreciably re - arrange during the pulse time .
the fluence is assumed sufficiently low so that the scattering is linear in pulse intensity .
scattering off non - stationary evolving states is an area of growing interest .
when the time - dependent state of matter merely follows some classical parameter ( as in the case of tracking the time - dependent melting of crystals @xcite ) no coherence of electronic states is prepared and the analysis of scattering is simplified considerably .
time - dependent diffraction can then be described by simply replacing the charge density in stationary diffraction by the time - dependent charge - density .
the situation is more complex in pump - probe experiments in which a superposition of electronic states is prepared by the pump and is then probed by x - ray scattering .
such superpositions , which involve electronic coherence , can be prepared e.g. by inelastic stimulated raman processes @xcite , a photoionization process @xcite , off - resonant femtosecond pulses @xcite or high - intensity optical pumping @xcite .
diffraction is a macroscopic , classical effect that involves the interference of wavefronts emanating from different sources treated at the level of maxwell s equations @xcite .
a fundamental difficulty in extending it to single molecules is that x - ray scattering ( as any light scattering ) from a single molecule may not be thought of simply as a diffraction since it has both elastic ( thompson / rayleigh ) and an inelastic ( compton / raman ) components @xcite . using a quantum electrodynamic ( qed ) approach , we discuss and analyze the single - particle vs. the cooperative contributions to the signal .
we find that the two terms carry different particle form factors that permit raman scattering in different frequency ranges .
simulations are presented for the scattering of a broad - band x - ray pulse from a single molecule of the amino acid cysteine either in the ground state or when prepared in a nonstationary state by a stimuated resonant x - ray raman process with various delay times between preparation and the scattering event .
as the expressions developed in this paper bear a resemblence to the standard classical theory of diffraction , we review it briefly .
the diffraction signal from a system initially in the ground state @xmath0 is @xmath1 where @xmath2 is the ground state charge density in @xmath3-space and @xmath4 is the momentum transfer ( @xmath5 is the outgoing mode and @xmath6 is the incoming mode ) .
more generally , @xmath7 is the fourier transform of the charge - density operator @xmath8 equation ( [ eq : skclassic ] ) assumes that the scattering is elastic and treats the entire sample as a single system ( i.e. @xmath9 is the electron density of the entire sample ) .
a common approach to understanding diffraction patterns involves making an approximation on the structure of the electron density .
for example , we may presume that the complete wavefunction is built up from single - electron wavefunctions and the total electron density is the sum of the densities associated with each electon wavefunction .
if the system is composed of @xmath10 identical particles and each electron is bound to a particle , then the amplitude of the signal from each particle carries a relative phase related to the particle position . the term
particle " is used here as a generic partitioning of the system and may stand for atoms as well as large ( molecules ) or small ( unit cell ) groups of atoms . practically speaking , they should be large enough that the electron density between particles may be safely neglected and small enough that electronic structure calculations can be performed . separating single- and multi - particle contributions yields for the intensity @xcite : @xmath11\ ] ] where @xmath12 is the position of particle @xmath13 relative to @xmath14 . here
, @xmath15 stands for the electron density in a single particle and @xmath16 is known as the particle form factor .
thus , the signal is factored into a product of the signals from the particle electron density and from the distribution of particles .
the two terms in equation ( [ eq : skclassicsep ] ) both contribute linearly in @xmath10 to the integrated signal @xcite .
however , the former yields a signal that is generally distributed throughout reciprocal space while the various terms in the @xmath17 summation carry different phases that cause a redistribution of the signal into points of constructive and destructive interference ( a bragg pattern for crystaline samples ) .
we calculate the x - ray scattering by a sample initially prepared in a non - stationary state using a quantum description of the field , in which radiation back - reaction ( which leads to inelastic scattering ) is naturally built in .
incorporation of the broad frequency bandwidth of incoming x - rays is an important point as ultra - short pulses need to outrun destruction in single - molecule scattering .
we assume that the molecule is initially prepared in an arbitrary density matrix , representing a pure or mixed state such as photo - ions .
we extend the quantum field formalism developed in @xcite for spontaneous emission of visible light to off - resonant x - ray scattering .
our calculation starts with the minimal - coupling hamiltonian , which contains a term proportional to @xmath18 as well as @xmath19 where @xmath20 is the electronic current operator .
the former describes instantaneous scattering where the electrons do not have the time to respond during the scattering process .
the second term dominates resonant processes and allows for a delay between absorption and emission during which electronic rearrangement , ionization , and breakdown can occur . since hard x - rays
are always resonant with the electronic continua representing various ionized states , the relative role of the @xmath19 term should be investigated further in order to clarify how important are these processes . in our study
, we treat only off - resonant scattering and therefore focus on the @xmath21 term , which dominates such processes @xcite . in the interaction picture ,
the hamiltonian is given by @xmath22 where @xmath23 is the bare field and matter hamiltonian while @xmath24 is the field - matter coupling .
assuming the diffracting x - ray pulse is not resonant with any material transitions and its intensity is not too high , its interaction with the matter is given by @xcite @xmath25 we consider a sample of @xmath10 non - interacting , identical particles ( molecules or atoms ) indexed by @xmath13 with non - overlapping charge distributions so that the total charge - density operator can be partitioned as @xmath26 .
the signal is given by the electric field intensity arriving at the detector and may be generally expressed as the overlap integral of a detector spectrogram ( given in terms of the detection parameters ) and a bare spectrogram ( equation [ eq : sspectrograms ] ) .
the scattering signal mode is initially in the vacuum state @xmath27 .
therefore an interaction on both bra and ket of the field density matrix is required to generate the state @xmath28 which gives the signal .
we thus expand the signal as @xmath29 it is given in terms of the gated electric fields ( defined in appendix [ ap : gating ] ) , to second order in @xmath30 ( equation [ eq : hprime ] ) .
this naturally leads to a double sum @xmath31 over the scatterers .
terms with @xmath32 arise when the probe pulse is scattered off a single particle ( fig .
[ fig:1d].*a * ) and terms with @xmath33 describe two - particle scattering events ( fig .
[ fig:1d].*b * ) .
the former contains @xmath10 terms which add incoherently ( at the intensity level ) , giving a virtually isotropic signal .
the latter , in contrast , is governed by @xmath34 terms which carry different phase - factors and can interfere destructively or constructively ( yielding the bragg peaks in a crystalline sample or , more generally , a speckle pattern ) @xcite . in general , both contributions must be considered as it is frequently not sufficient to sample a signal only at the points of constructive interference ( bragg diffraction ) where the two - particle terms dominate @xcite . in this paper
, we refer to the single - particle contribution as `` incoherent '' and the two - particle contribution as coherent " reflecting the way in which these contributions add ( at the intensity versus amplitude level ) . in the x - ray community ,
incoherent is commonly taken to refer to inelastic contributions while coherent refers to elastic .
the two nomenclatures coincide when considering scattering from the ground state because , as will be shown below , two - particle scattering from the ground state ( or any population ) is necessarily elastic while single - particle scattering from the ground state has only a single elastic term ( or one term for each initially populated state ) .
two - particle scattering from a nonstationary superposition state , in contrast , has both elastic and inelastic terms which add coherently ( since they are both proportional to a spatial phase - factor @xmath35 dependent on the distance between the two particles ) .
on the other hand , the elastic terms from the single - particle scattering add incoherently since their spatial phase factors are canceled by the opposite - hermiticity interactions on the ket / bra . in the coherent terms the product of charge densities on different particles can be factorized and the signal is proportional to the modulus square of the single - particle electron density @xmath36 .
such factorization is not possible for the incoherent terms where the signal is given by a correlation function of the charge density rather than the charge density itself . as discussed below
, restricting attention to elastic scattering reduces the correlation function to the modulus - square result and eliminates the need to consider the correlation function .
thus , the signal is expressible in terms of the charge - density alone either when the coherent contribution dominates or when attention is restricted to elastic scattering .
figure [ fig:1d ] illustrates the incoherent ( fig .
[ fig:1d].*a * ) and coherent ( fig .
[ fig:1d].*b * ) scattering processes from a sample following preparation in a non - stationary electronic superposition state described by the density matrix @xmath37 .
the preparation process is represented by the gray box . in this paper
, we will consider a rama preparation process ( in another work , we examine the case where the preparation process is itself an off - resonant scattering process @xcite ) .
after the preparation process ( which terminates at @xmath38 ) the system evolves freely until an x - ray probe pulse with an experimentally - controlled envelope centered at time @xmath39 impinges on the sample and is scattered into a signal mode that is initially in a vacuum state .
this signal photon is then finally absorbed by the detector .
time translation invariance of the matter correlation function implies the basic energy - conservation condition @xmath40 for an incoherent scattering event ( fig .
[ fig:1d].*a * ) .
coherent scattering events ( fig .
[ fig:1d].*b * ) give two such conditions corresponding to the two diagrams @xmath41 and @xmath42 . for
a broad - band pulse in which @xmath43 and @xmath44 can differ appreciably , the signal will contain contributions from paths in which @xmath45 and @xmath46 take all possible values within this bandwidth
. these can be controlled by pulse - shaping techniques as well as by the choice of detection parameters @xcite .
it follows from the diagrams that , since each particle must end the process in a population state and has only interactions on one side of the loop , coherent scattering from populations will always be rayleigh type while raman ( inelastic ) processes result from scattering off coherences . in contrast
, the single - particle energy conservation condition for an initial population ( i.e. @xmath47 ) allows for inelastic processes .
complete expressions for the signal which include arbitrary gating and pulse envelopes ( i.e. , the bare spectrograms ) are given in the appendix [ ap : barederivation ] . in this section , we discuss the simpler signal that results when no time - gating is applied .
the formulas are simplest when we employ delta functions for the detector spectrograms . as shown appendix [ ap : gating ] we have separate detector spectrograms for the time - frequency gating and the space - propagation gating .
as seen in appendix [ ap : barederivation ] ( equations ( [ eq : wbcohcf ] ) and ( [ eq : wbinccf ] ) ) both coherent and incoherent bare spectrograms carry the delta function factor @xmath48 .
this connects @xmath49 to @xmath50 in the usual way ( though this is not automatic since the two are not a priori related in this way but rather both begin as seperate gating variables ) as well as fixing the direction of @xmath50 . for this reason ,
the logical choice for the spatial - propagation detector spectrogram is @xmath51 this represents a spatially resolved signal ; that is , the location of the detection event ( the pixel location ) is resolved .
all signals considered in this paper use this choice . if the detector spectrogram does not depend on @xmath52 ( no time - gating is applied ) , we may separate the time - dependent phase factors from the auxilliary functions and carry out the time integration to give a factor of @xmath53 .
the signals are therefore given by @xmath54 @xmath55 where @xmath56 is the momentum transfer , @xmath57 is the frequency gating function of appendix [ ap : gating ] and @xmath58 stands for the set of parameters that define the external pulse envelopes ( including @xmath59 ) .
we approximate @xmath60 as a constant on the assumption that all pixels are roughly equidistant from the sample . in approaches to x - ray scattering
that do not incorporate the detection event , the differential scattering cross section is calculated and found to be proportional to @xmath61 with @xmath62 the classical electron radius .
our incorporation of the detection event included a summation over polarizations of the signal field and an averaging over initial polarizations and emission directions .
this was shown to lead to the replacement @xmath63 while the use of atomic units equates @xmath64 .
finally , since we calculate the signal ( defined as the expectation value of the gated electric field ) by explicitly incorporating the detection event ( which is linear in @xmath45 ) our result is proportional to @xmath65 . recalling that @xmath66 , we see that our result carries the appropriate proportionality factors compared to the usual differential scattering cross section ( @xcite ) . that the arguments of eqs .
( [ eq : scohwr])-([eq : sincwr ] ) are @xmath67 and @xmath68 reflects the fact that they correspond to taking a spectrum at every pixel . since the final observed signal frequency is @xmath67 , we may as well relabel it @xmath45 to make the interpretation more intuitive . aside from the expected inverse - square dependence on @xmath69 ,
the signal only depends on @xmath68 through @xmath70 , i.e. , the direction vector pointing from the sample to the pixel .
since @xmath70 is the same as the direction of propagation of scattered light , this suggests representing the directional dependence by defining @xmath71 .
finally , it is important to note that , althought these signals do not depend on time directly since we have assumed no time resolution ( i.e. , the pixels are simply left open to collect incoming light ) , the signal does depend parametrically on the central time of the incoming pulse through the field envelope @xmath72 which carries a phase factor @xmath73 . here
, @xmath39 is the central time of the pulse envelope and the zero of time is set at the end of the state preparation process ( where the prepared state is presumed to be known ) . since @xmath39 therefore represents the time separation between state preparation and arrival of the center of the probe pulse and this is a key experimental control , we explicitly write this dependence in future expressions . in the following ,
we focus on the frequency - resolved signal because of its relative ease of interpretation .
while this has not yet been demonstrated in the x - ray regime , it has been shown possible to discriminate a single wavelength component from multiwavelength scattering data in the euv range @xcite . from supplementary eqns .
( [ eq : scohwr ] ) and ( [ eq : sincwr ] ) ) with @xmath74 , this signal is given by the sum of a coherent and an incoherent contribution which are related to the transition charge density @xmath75 @xmath76 where @xmath77 is the spectral envelope of the scattered pulse , @xmath78 are fourier transformed matrix elements of the charge - density operator and @xmath79,@xmath80,@xmath81 , and @xmath82 represent electronic states .
we have also defined the momentum - transfer vector @xmath83 .
the signal is not generally related to the time - dependent , single - particle charge density but rather to its correlation function @xcite .
a compact expression for the total signal is @xmath84 where the correlation function of the total charge density operators may be expanded in terms of the single - particle densities as @xmath85 for a macroscopic sample initially in the ground electronic state , the standard classical theory of diffraction gives the signal as the product of the modulus square of the single - particle momentum - space electron density ( the form factor ) and a structure factor which describes the interparticle distribution ( equation ( [ eq : skclassicsep ] ) ) @xcite .
this formula is used to invert the x - ray diffraction signal to obtain the ground state electron density once the phase problem " is resolved @xcite .
substituting equation ( [ eq : cofunc ] ) into ( [ eq : stot ] ) yields a close resemblence to the classical expression for x - ray diffraction except that the coherent and incoherent terms now carry different form - factors .
note that , if we restrict attention to elastic scattering , the correlation function in the incoherent term separates into a modulus square form and the two form factors are equal .
cysteine is a sulfur - containing amino acid which affects the secondary structure of many protiens because of the disulfide bonds it forms .
it has been implicated in biological charge transfer in respiratory complexes @xcite .
we have previously explored various resonant x - ray spectroscopic signals from this molecule , including stimulated x - ray raman scattering ( sxrs ) and x - ray photon echo @xcite .
below , we present calculations of off - resonant scattering from a single cysteine molecule ( chemical structure and oriention shown in fig .
[ fig : gssignal]c ) .
details of the computational methodology can be found in section [ sec : methods ] .
the scattering signal from the ground state ( the second term in equation [ eq : skmixed ] with @xmath86 , the ground state ) is depicted in figs .
[ fig : gssignal]*a * and [ fig : gssignal]*b*. we also show the ground - state electron density , @xmath87 , the pulse power spectrum , and the pulse wave vector for reference .
as these calculations are for a single isolated molecule , we can restrict our scattering calculations to the second ( incoherent ) term in equation ( [ eq : skmixed ] ) .
we take the scattering pulse to be a transform - limited gaussian @xmath88 the center frequency @xmath89 is set to 10 kev , and we take the direction of propagation to be in the positive @xmath90 direction ( for molecule orientation , see fig . [ fig : gssignal].*c * ) .
the pulse duration is @xmath91 @xmath92 which orresponds to a fwhm bandwidth of 3.65 ev .
future progress in pulse - generation may make such experiments realizable .
for this frequency range , the difference between @xmath93 and @xmath94 for any two states @xmath79 and @xmath80 is negligibly small , and is ignored in the calculations presented herein .
we take the signal detectors to be on square grid , 2 cm in length on each side , located 1 cm from the molecule in the positive @xmath90 direction ( i.e. we detect forward - scattered light ) .
this corresponds to a maximum detected scattering angle of @xmath95 .
we consider two different values for the detection frequency @xmath45 , one inside and one outside the pulse bandwidth .
when we set the detection frequency equal to the pulse center frequency , we get the signal shown in fig . [
fig : gssignal].*a*. this signal is dominated by the elastic scattering terms , where the scattering process does not change the state of the molecule . at this detection frequency , the elastic contribution is @xmath96 larger than the inelastic .
the elastic scattering term can be eliminated by moving the detection frequency outside the pulse bandwidth . in fig . [
fig : gssignal].*b * , we show the scattering signal with a detection frequency @xmath97 . with this detection frequency , we see inelastic terms from valence states @xmath98 whose excitation energy satisfies the condition that @xmath99 is within the pulse bandwidth .
therefore all states with an energy between 4 and 12 ev will contribute .
the scattering pattern resulting from the elastic and inelastic process are vastly different .
the former is more strongly centered around the origin , corresponding to @xmath100 , and elongated in the @xmath101 direction .
the inelastic term , in addition to the feature at the origin , has two equal - intensity peaks at @xmath102 and @xmath103 , which , when converted to reciprocal space corresponds to @xmath104 and @xmath105 , respectively .
we next turn to time - resolved scattering , in which an x - ray raman preparation pulse @xmath106 , resonant with the sulfur k edge , arrives at @xmath38 followed by off - resonance scattering at time @xmath107 . in this process ,
the raman pulse acts twice on the same side of the loop , first promoting a sulfur 1s electron to the valence band before the transient core hole is filled by another valence electron . because the raman pulse is broadband
, these two dipole interactions leave the molecule in a superposition of valence - excited states .
this wavepacket is initially localized in the region surrounding the atom whose core is in resonance ( sulfur in this case ) , but becomes delocalized across the molecule in a @xmath108 time scale @xcite . the molecular density matrix immediately following the interaction with the first pulse is @xmath109 where @xmath110 is the effective polarizability operator and @xmath111 is the initial ( equilibrium ) density matrix . in equation ( [ eq : alpha ] ) ,
@xmath112 is the polarization vector for the raman pulse and @xmath113 is the transition dipole between the valence - excited state @xmath98 and the core - excited state @xmath81 . for a single - molecule system prepared in this manner ( and with the simplification @xmath114 ) , equation ( [ eq : skmixed ] )
assumes the form @xmath115 note that any amplitude in the ground state after the raman pulse has passed ( terms in equation ( [ eq : tdxrd ] ) where @xmath116 ) will contribute to a background , delay - time - independent signal , which can be filtered out .
the remaining time - dependent scattering signal is a difference signal and will have positive and negative features , unlike the ground - state scattering signals from fig .
[ fig : gssignal ] which were only positive .
the largest contributions will come from terms where @xmath117 in equation ( [ eq : tdxrd ] ) is equal to either @xmath98 or @xmath118 .
we take the raman pulse center frequency at the sulfur k - edge frequency @xmath119 , and polarized along the @xmath90 direction .
the x - ray raman signal is highly dependent upon the choice of polarization vector , and the nature of the underlying wavepacket is quite different for a @xmath120 or @xmath101 polarized pulses @xcite .
we take both the raman and scattering pulses to be gaussian with duration @xmath121 ( fwhm of 10.96 ev ) .
the broad bandwidth connects the ground state with the set of valence excited states , with energies between 5.7 ev and 9.0 ev .
figure [ fig : tdxrd1 ] shows the time - dependent x - ray scattering signal for interpulse delays ranging from 0 to 20 fs . for each signal
, we also show the transition density for the raman wavepacket prior to interaction with the scattering pulse .
this is defined by @xmath122= \sum_{e } i \alpha_{eg } e^{-i \omega_{eg } t } \sigma_{eg}(\boldsymbol{r } ) + \mathrm{c.c.}\ ] ] the left panel of fig .
[ fig : tdxrd1 ] shows that the transition density is localized near the sulfur atom at @xmath123 in supplementary material we show a movie of the time - dependent scattering signal and transition density for interpulse delays up to 20 fs . from the movie ( see supplemental material @xcite ) and from fig .
[ fig : tdxrd1 ] we see that there is a good deal more structure in the scattering signal along the @xmath120 direction than along the @xmath101 direction .
this is consistent with the fact that the electronic motion induced by the raman pulse is mostly in the @xmath120 direction .
while the correspondence between electronic motion in real space and the resulting scattering pattern is highly suggestive , it is not immediately apparent whether the transition density can be recovered from the scattering pattern alone .
this is because the scattering pattern ( equation [ eq : tdxrd ] ) is not simply the fourier transform of the density ( equation [ eq : rsdm ] ) .
the scattering signal shows a complex dependence on time , reflecting interference between the many different electronic states which make up the superposition .
the signal may not simply be thought of as a snapshot of the instantaneous time - dependent charge - density .
the time variation strongly depends on the scattering direction , as can be seen in fig .
[ fig : tdxrd2 ] .
here we depict the time evolution of six points from the @xmath124 signal , corresponding to the highest and lowest peaks therein .
each trace has a beating pattern , representing a spatially resolved interferogram .
decay due to finite lifetime and dephasing is not included in the time - domain signals presented here .
the contribution to the signal at a given detector due to a particular electronic coherence can be determined by fourier transforming with respect to the delay time .
this would give information on the transition density for the contributing excited states .
however , we do not pursue this analysis here . in fig .
[ fig:8features ] we show the variation of the time traces in fig .
[ fig : tdxrd2 ] with the detection frequency @xmath45 . in the previous figures ,
@xmath45 was set equal to the scattering pulse center frequency , @xmath89 .
however , since purely elastic processes do not contribute to the time - dependent signal , the signal is larger for @xmath125 .
the signal is maximized when , for a given @xmath98 and @xmath117 from equation ( [ eq : tdxrd ] ) , both @xmath126 and @xmath127 lie within the pulse bandwidth .
coherent two - particle scattering from populations is an elastic process while coherent scattering from matter coherences is inelastic .
incoherent scattering , in contrast , generally produces both elastic and inelastic contributions regardless of the initial material state , as evident from equation ( [ eq : skmixed ] ) .
thus , the coherent terms can only induce transitions between states populated by the material superposition state while the incoherent terms can induce transitions to any electronic state .
notably , the incoherent and coherent signals come with different particle form factors .
thus , the total signal may not be simply factored into the product of a particle form factor and a structure factor as in the classical theory .
this issue had been addressed for x - ray scattering from a single hydrogen atom when it is prepared in a superposition state @xcite .
our qed approach generalizes previous treatments @xcite to properly account for arbitrary pulse bandshape , non - impulsive pulses , and detection details .
the role of electronic coherence requires full account of frequency , time , and wavevector gated detection as is done here .
the present treatment fully incorporates inelastic scattering effects , which must be taken into account for single - molecule scattering .
if the sample is initially in the ground state , the coherent scattering is entirely elastic and if the sample is prepared perturbatively the coherent is dominated by elastic scattering while the incoherent terms are affected equally by the transition charge densities ( @xmath128 ) .
since light scattered from a single particle is not necessarily eleastic and can change the state of the particle , obtaining the charge density from single - particle scattering will require distinguishing between the rayleigh and raman components . furthermore , the total scattered intensity may not be fully described by the ground state charge density alone ; it requires more information about electronic excited states of the particle , i.e. the transition charge densities .
our approach and simulations can provide valuable insight for future structural studies of proteins and nano - devices .
the details of the electronic structure calculations can be found in ref . , and are recounted briefly here .
the optimized geometry of cysteine was obtained with the gaussian09 package@xcite at the b3lyp@xcite/6 - 311 g * * level of theory .
all tddft calculations were done at the cam - b3lyp@xcite/6 - 311++g(2d,2p ) level of theory , and with the tamm - dancoff approximation ( tda ) @xcite .
it was found that tddft with this type of long - range - corrected density functionals and diffused basis functions can describe rydberg states well@xcite . in these calculations ,
we include 50 valence excited states , with energies ranging from 5.4 ev to 9.0 ev .
core - excited states , in which a sulfur 1s electron is excited to the valence band , were calculated using restricted excitation windown ( rew ) tddft with a locally modified version of nwchem code @xcite .
we also include 50 core - excited states for core excitations , with energies ranging from 2473.5 ev to 2495.9 ev ( shifted to match experimental xanes results ) .
transition density matrices between different valence excited states , which contribute to the summation in equation ( 9 ) , are calculated using the ci coefficients from the tddft / tda results , and are therefore in an unrelaxed sense .
more accurate relaxed state - to - state transition density matrices could be calculated using the z - vector method @xcite , and this research is ongoing .
the support of the chemical sciences , geosciences , and biosciences division , office of basic energy sciences , office of science , u.s .
department of energy as well as from the national science foundation ( grants che-1058791 and che-0840513 ) , and the national institutes of health ( grant gm-59230 ) is gratefully acknowledged .
kochise bennett and yu zhang were supported by doe .
the signal is defined as the intensity of the detected electric field @xmath129 where the detected electric field is represented as @xmath130 following the procedure outlined in ref .
@xcite , we add a series of gating functions to the detected electric field : @xmath131 this gives : @xmath132 the signal is then given by equation ( [ eq : sgatedintensity ] ) .
we define the bare and detector spectrograms via : @xmath133 @xmath134 where we have defined the auxilliary functions @xmath135 and @xmath136 the signal is then given by the overlap of the two spectrograms : @xmath137 for brevity , the following definitions are used in the derivations : @xmath138 @xmath139
beginning with equation ( [ eq : wb01 ] ) , we expand it to leading order in @xmath140 ( equation ( 1 ) ) .
this requires two interactions ( one each for the ket and bra ) since the signal mode is initially in a vacuum state .
@xmath141 here , the total density matrix is the direct product of field and matter density matrices immediately following state preparation ( i.e @xmath142 ) .
the vector potential of the vacuum modes , @xmath143 , is expanded as : @xmath144 here , @xmath145 is the annihilation operator for mode @xmath146 and polarization @xmath147 , @xmath148 is a unit vector in the direction of polarization , and @xmath149 is the field quantization volume .
the field operator is given by @xmath150 while the vector potential for the classical probe beam @xmath151 is represented as : @xmath152 where @xmath153 is the fraction of the probe pulse in polarization state @xmath147 and @xmath154 is a unit vector in the direction of direction of polarization @xmath147 .
henceforth , we will use the shorthand @xmath155 and assume a narrow beam so that @xmath156 . note that , by starting the @xmath157 and @xmath158 integrations at @xmath159 , we assume that the scattered pulse is well separated from the state preparation process . inserting these definitions into equation ( [ eq : wb02 ] ) ,
separating matter and field correlation functions ( and evaluating the latter with the conditions described above ) , we obtain : @xmath160 where we have taken a dipolar - interaction model for the detection event with @xmath161 the dipole moment of the detector .
we have also defined @xmath162 so that @xmath163 .
we first examine the @xmath33 terms in the above . assuming that the particles are uncorrelated
, we have @xmath164 . the correlation function
therefore splits and we can separately collect factors associated with each particle .
that is , we define : where @xmath166 $ ] is the trace over the product of the argument and the density matrix immediately after the state preparation process ( @xmath167 ) .
the coherent spectrogram is then given by : @xmath168 in order to carry out the integration over @xmath169 we use the fourier transform @xmath170 we thus have @xmath171 where we have also expanded the dot products of the polarizations and used the identity @xmath172 we are now free to carry out the time integration : @xmath173 where @xmath174 is a positive infinitesimal .
we change the summation over @xmath5 to an integration via @xmath175 and make use of the relation @xcite @xmath176 this gives : @xmath177 where we have defined @xmath178 .
the @xmath179 integral has poles at @xmath180 and @xmath181 .
the term arising from the residue at the first pole will have a factor @xmath182 . because the interaction between the sample and the field is off - resonant
, @xmath43 will not be close to any material frequency .
the term arising from the residue of the pole at @xmath180 is negligible in such a process .
thus , we may perform the @xmath179 integration : @xmath183 using the identity @xmath184 and rotationally averaging so that @xmath185 results in @xmath186 placing the origin within the sample and taking the detector to be far away ( in comparison to the size of the sample ) allows the approximation @xmath187 .
where we have substituted @xmath188 since that is the point at which we will eventually evaluate @xmath189 .
although we will later formally integrate over all @xmath190 , this represents different detection locations and thus should only be carried out over the area of a detector pixel . the assumption that @xmath190 is small compared to @xmath191 ( the distance to the detector ) is thus justified . dropping the retardation due to @xmath191 ( since this uniformly delays the signal by some constant due to travel time ) and replacing the @xmath192 in the denominator by @xmath191 simplifies the expression yielding : @xmath193 where we have also carried out the @xmath194 integration via @xmath195 with @xmath196 .
we are now in a position to perform the integrations over @xmath197 and @xmath198 in equation ( [ eq : wbcoh1 ] ) : @xmath199 @xmath200 with @xmath201 defined for convenience .
the bare coherent spectrogram is then : @xmath202 the incoherent ( @xmath32 ) terms contain the correlation function : @xmath203 since there is only one operator on each side of the density matrix , there is no time ordering ambiguity and we may drop @xmath204 .
the hilbert space expression is then @xmath205\ ] ] although we may not factor this into a product of correlation functions , we may still go through the same series of simplifications as in the coherent case resulting in : @xmath206 the total ( incoherent plus coherent ) bare spectrogram may also be written in a form similar to this @xmath207 when given in terms of the total ( many - particle ) charge density .
henry n chapman , petra fromme , anton barty , thomas a white , richard a kirian , andrew aquila , mark s hunter , joachim schulz , daniel p deponte , and uwe weierstall .
femtosecond x - ray protein nanocrystallography . , 470(7332):7377 , february 2011 .
c. w. siders , a. cavalleri , k. sokolowski - tinten , cs tth , t. guo , m. kammler , m. horn von hoegen , k. r. wilson , d. von der linde , and c. p. j. barty .
detection of nonthermal melting by ultrafast x - ray diffraction .
, 286(5443):13401342 , november 1999 .
rudolf koopmann , karolina cupelli , lars redecke , karol nass , daniel p deponte , thomas a white , francesco stellato , dirk rehders , mengning liang , jakob andreasson , et al .
_ in vivo _ protein crystallization opens new routes in structural biology .
, 9(3):259262 , march 2012 . jan kern , roberto alonso - mori , julia hellmich , rosalie tran , johan hattne , hartawan laksmono , carina glckner , nathaniel echols , raymond g sierra , jonas sellberg , et al .
room temperature femtosecond x - ray diffraction of photosystem ii microcrystals . , 109(25):97219726 , june 2012 .
d starodub , a aquila , s bajt , m barthelmess , a barty , c bostedt , jd bozek , n coppola , rb doak , sw epp , et al .
single - particle structure determination by correlations of snapshot x - ray diffraction patterns .
, 3:1276 , december 2012 .
henry n chapman , anton barty , michael j bogan , sbastien boutet , matthias frank , stefan p hau - riege , stefano marchesini , bruce w woods , saa bajt , and w henry benner .
femtosecond diffractive imaging with a soft - x - ray free - electron laser .
, 2(12 ) , 2006 .
m. bargheer , n. zhavoronkov , y. gritsai , j. c. woo , d. s. kim , m. woerner , and t. elsaesser .
coherent atomic motions in a nanostructure studied by femtosecond x - ray diffraction .
, 306(5702):17711773 , december 2004 .
j. stingl , f. zamponi , b. freyer , m. woerner , t. elsaesser , and a. borgschulte .
electron transfer in a virtual quantum state of libh@xmath208 induced by strong optical fields and mapped by femtosecond x - ray diffraction .
, 109(14):147402 , october 2012 .
spots of constructive interference ( known as the speckle pattern ) are determined by the interparticle structure factor @xmath209 .
the peak intensity at the points of constructive interference ( such as the bragg peaks for crystals or small angle scattering ( saxs ) ) scales as @xmath210 .
this scaling facilitates the measurement of the scattered intensity at bragg peaks or for small @xmath211.@xcite .
jason d. biggs , judith a. voll , and shaul mukamel .
coherent nonlinear optical studies of elementary processes in biological complexes ; diagrammatic techniques based on the wavefunction vs. the density matrix . , 370:37093727 , 2012 .
jianwei miao , henry n. chapman , janos kirz , david sayre , and keith o. hodgson . taking x - ray diffraction to the limit : macromolecular structures from femtosecond x - ray pulses and diffraction microscopy of cells with synchrotron radiation .
, 33(1):157176 , 2004 .
ruben a. dilanian , bo chen , garth j. williams , harry m. quiney , keith a. nugent , sven teichmann , peter hannaford , lap v. dao , and andrew g. peele .
diffractive imaging using a polychromatic high - harmonic generation soft - x - ray source .
, 106(2):023110 , 2009 .
yu zhang , jason d. biggs , daniel healion , niranjan govind , and shaul mukamel .
core and valence excitations in resonant x - ray spectroscopy using restricted excitation window time - dependent density functional theory . , 137:194306 , 2012 .
daniel healion , yu zhang , jason d. biggs , niranjan govind , and shaul mukamel .
entangled valence electron
hole dynamics revealed by stimulated attosecond x - ray raman scattering . , 3(17):23262331 , 2012 .
m. j. frisch , g. w. trucks , h. b. schlegel , g. e. scuseria , m. a. robb , j. r. cheeseman , g. scalmani , v. barone , b. mennucci , g. a. petersson , h. nakatsuji , m. caricato , x. li , h. p. hratchian , a. f. izmaylov , j. bloino , g. zheng , j. l. sonnenberg , m. hada , m. ehara , k. toyota , r. fukuda , j. hasegawa , m. ishida , t. nakajima , y. honda , o. kitao , h. nakai , t. vreven , j. a. montgomery , jr .
, j. e. peralta , f. ogliaro , m. bearpark , j. j. heyd , e. brothers , k. n. kudin , v. n. staroverov , r. kobayashi , j. normand , k. raghavachari , a. rendell , j. c. burant , s. s. iyengar , j. tomasi , m. cossi , n. rega , j. m. millam , m. klene , j. e. knox , j. b. cross , v. bakken , c. adamo , j. jaramillo , r. gomperts , r. e. stratmann , o. yazyev , a. j. austin , r. cammi , c. pomelli , j. w. ochterski , r. l. martin , k. morokuma , v. g. zakrzewski , g. a. voth , p. salvador , j. j. dannenberg , s. dapprich , a. d. daniels , ? .
farkas , j. b. foresman , j. v. ortiz , j. cioslowski , and d. j. fox .
gaussian 09 .
gaussian inc .
wallingford ct 2009 .
p. j. stephens , f. j. devlin , c. f. chabalowski , and m. j. frisch .
ab initio calculation of vibrational absorption and circular dichroism spectra using density functional force fields .
, 98:1162311627 , 1994 .
m. valiev , e.j .
bylaska , n. govind , k. kowalski , t.p .
straatsma , h.j.j .
van dam , d. wang , j. nieplocha , e. apra , t.l .
windus , and w.a .
de jong . : a comprehensive and scalable open - source solution for large scale molecular simulations .
, 181:14771489 , 2010 .
|
using a quantum electrodynamic framework , we calculate the off - resonant scattering of a broad - band x - ray pulse from a sample initially prepared in an arbitrary superposition of electronic states .
the signal consists of single - particle ( incoherent ) and two - particle ( coherent ) contributions that carry different particle form factors that involve different material transitions .
single - molecule experiments involving incoherent scattering are more influenced by inelastic processes compared to bulk measurements .
the conditions under which the technique directly measures charge densities ( and can be considered as diffraction ) as opposed to correlation functions of the charge - density are specified .
the results are illustrated with time- and wavevector - resolved signals from a single amino acid molecule ( cysteine ) following an impulsive excitation by a stimulated x - ray raman process resonant with the sulfur k - edge .
our theory and simulations can guide future experimental studies on the structures of nano - particles and proteins .
|
decades of exponential growth in commercial data services has ushered in the so - called bigdata " era , to which the expansive mobile wireless network is a critical data contributor . as of @xmath0 , the global penetration of mobile subscribers has reached @xmath1 , producing staggeringly @xmath2 exabytes ( @xmath3 ) of mobile data worldwide .
the surge of mobile data traffic in recent years is mainly attributed to the popularity of smartphones , phone cameras , mobile tablets and other smart mobile devices that support mobile broadband applications , e.g. , online music , video and gaming as shown in fig . [ 71 ] . with a compound annual growth rate of over @xmath4
, it is expected that the mobile data traffic will increase by @xmath5 times from @xmath6 to @xmath7 .
in addition to the vast amount of wireless source data , modern wireless signal processing often amplifies the system s pressure from bigdata in pursuit of higher performance gain .
for instance , mimo antenna technologies are now extensively used to boost throughput and reliability at both mobile terminals ( mts ) and base stations ( bss ) of high speed wireless services .
this , however , also increases the system data traffic to be processed in proportion to the number of antennas in use .
moreover , the @xmath5 g ( the fifth generation ) wireless network presently under development is likely to migrate the currently hierarchical , bs - centric cellular architecture to a cloud - based layered network structure , consisting of a large number of cooperating wireless access points ( aps ) connected by either wireline or wireless fronthaul links to a bigdata capable processing central unit ( cu ) .
new wireless access structures , such as coordinated multipoint ( comp or networked mimo ) @xcite , heterogeneous network ( hetnet ) @xcite and cloud - based radio access network ( c - ran ) @xcite , are under development to achieve multi - standard , interference - aware and energy - friendly ( green ) wireless communications . in practice , the use of cooperating wireless aps could easily generate multiple gbps data from a single user s fronthaul links due to the need for baseband joint processing , such that the high traffic load may overwhelm the fronthaul link or the system computing unit for signal processing and coordination . such intensely high system traffic volume , together with the rapidly growing mobile data source volume , surpasses both the processing power improvement speed of our current computing capabilities and the fronthaul / backhaul link rate increase pace of our networking systems .
it necessitates a new wireless architecture along with efficient signal processing methods to make wireless systems _ scalable _ to continued growth of data traffic . on the other hand ,
timely and cost - efficient information processing is made possible by the fact that the vast - volume mobile data traffics are not completely chaotic and hopelessly beyond management .
rather , they often exhibit strong _ insightful features _ , such as user mobility pattern , spatial , temporal and social correlations of data contents . these special characteristics of mobile traffic present us with opportunities to harness and exploit bigdata for potential performance gains in various wireless services . to effectively utilize and exploit these characteristics ,
they should be identified , extracted , and efficiently stored .
for instance , caching popular contents at wireless hot spots could effectively reduce the real - time traffic in the fronthaul links .
additionally , network control decisions , such as routing , resource allocation , and status reporting , instead of being rigidly programmed , could be made data - driven to fully capture the interplay between bigdata and network structure .
presently , however , these advanced data - aware features could not be efficiently implemented in current wireless systems , which are mainly designed for content delivery , instead of analyzing and making use of the data traffic .
bearing in mind of the aforementioned challenges and opportunities brought by bigdata traffic , we address in this article two important problems of wireless communication system design in the bigdata era : * what may constitute a _ scalable _ wireless network architecture for efficient handling of bigdata traffic ? * how to effectively incorporate and utilize the _ bigdata awareness _ to improve the wireless system performance ?
specifically , to answer * q1 * , we introduce in section [ sec : scalable_structure ] a hybrid signal processing paradigm to enable flexible data processing at both the bs / ap and the cu levels , and correspondingly a number of scalable data traffic management techniques to serve the conflicting needs between the overall system performance and the data processing complexity . for * q2
* , we first discuss in section [ sec : capital_big_data ] typical bigdata features and efficient data analytics to extract these features .
next , we introduce a number of bigdata - aware signal processing methods and wireless networking structures to capitalize from bigdata interplay , such as mobile cloud processing , crowd computing , and software - defined networking , etc .
we also suggest in section [ sec : future ] several future research directions for wireless communications in the bigdata era .
finally , we conclude this article in section [ sec : conclusion ] . before proceeding to detailed discussions , it is worth mentioning that the considered scalable network structure and bigdata awareness are both important mechanisms for accommodating mobile bigdata in future wireless networks , though they focus differently on the physical and network / application layers , respectively
. nonetheless , the two solutions can also be complementary to each other .
for instance , as we will discuss later , we can optimize the overall caching strategy by combining long - term cache provisioning ( network / application layer ) and real - time cache - assisted signal processing ( physical layer ) techniques .
in addition , although this article focuses on the design aspects of cellular networks in the bigdata era , most of the key enabling mechanisms for mobile bigdata processing are also applicable to other wireless networking structures , such as wireless local area networks ( wlans ) and heterogeneous networks .
some representative system designs are also discussed in this article .
neither the current cellular systems nor the next - generation cloud - based c - ran @xcite under development was designed to provide a scalable solution for the arrival of the bigdata era .
the current @xmath8 g and @xmath9 g cellular systems exemplify a bs - centric design , in which a bs bears much the responsibilities of radio access , baseband processing and radio resource control execution to serve the mobile users in the vicinity . to meet the fast growing mobile data service demand , a smaller cell size is commonly used to improve frequency reuse , which may generate complex and severe inter - cell interference .
furthermore , small cells can also be costly because of cost from densely deployed bss . the cloud - centric network proposed for @xmath5 g
mitigates the inter - cell interference by centralized signal processing and reduces the unit cell deployment cost by moving computations to the cloud " . at the same time , only inexpensive relay - like remote radio heads ( rrhs ) are used for radio frequency ( rf ) level wireless access .
however , such fully centralized scheme may be overwhelmed by the huge wave of data traffic beyond its fronthaul link capacities and its computational power .
alternatively , a _ hybrid _ structure could take advantage of the benefits from the two design paradigms ; that is , a wireless system that could adaptively choose only local processing at the bs - level , or only central processing at the cu - level , or parallel processing at both levels , based on , for instance , physical channel conditions and correlations in the data contents , etc .
we thus consider such a generic network structure shown in fig .
[ 72 ] , which mainly inherits the skeleton of c - ran , but has integrated several programmable modules to carry out intelligent signal processing at the bs level . in the radio access network , mobile users could be served simultaneously by multiple bss , where each bs is equipped with multiple antennas and is linked to the cu via high - speed fiber / wireless fronthauls for exchanging user data and control signals .
the cu is further connected to the backhaul core network for external content access . in the proposed hybrid network structure ,
baseband processing units ( bpus ) are available at both the bss and cu , which enable user message encoding / decoding at both levels .
in addition , learning units ( lus ) are installed for data traffic analytics , whose functions will be detailed in section [ sec : capital_big_data ]
. caches are also installed at the bss and cu to save the fronthaul bandwidth consumed for frequent retransmissions of popular contents . before entering the discussions of hybrid signal processing models ,
it is worth mentioning that applicable fronthaul data management methods are directly constrained by fronthaul technologies in use .
specifically , the system could choose between optical analog and optical / wireless digital fronthaul technologies .
optical analog modulation using radio frequency ( rf ) signal as input is commonly referred to as the radio - over - fiber ( rof ) .
alternatively , analog rf input signal could also be quantized and encoded into binary codewords for digital wireless or fiber - optic communication ( dfc ) . in practice
, rof is simpler and less expensive than dfc .
furthermore , it also exhibits lower processing delays and better interoperability with multiple wireless standards , e.g. , @xmath8 g , lte and wifi , as it is oblivious to the user codebooks and wireless modulation schemes .
however , its limitations are also evident , e.g. , susceptibility to noise and signal distortion , and difficulty of synchronization . more importantly , available signal processing techniques for fronthaul traffic management using rof are less sophisticated , generally limited to simply transforming , or removing certain parts of the received rf signals , e.g. , sub - channel and antenna selection methods .
in contrast , dfc could be combined with data compression , opportunistic decoding and many other advanced digital signal processing techniques . in the following ,
we mainly focus on data traffic management methods using digital fronthaul .
the wireless / fiber - optic link has its own throughput limit .
for instance , a commercial fiber - optic link normally operates at a link rate in the order of @xmath10 gbps for digital communication over a single optical carrier .
transmission rates beyond the link rate capacity may lead to severe signal distortions , and consequently poor decoding performance .
therefore , the system performance must be optimized under the fronthaul link capacity constraints . with respect to the hybrid network structure in fig .
@xmath11 , we now introduce some scalable fronthaul data management techniques in three major categories : uplink direction would require unlimited fronthaul capacity to transmit an analog rf signal perfectly without any distortion from a bs to the cu .
an analog signal could be more efficiently transmitted through the fronthaul if it is quantized and compressed into binary codewords . from an information theoretic perspective , the effect of data compression could be modeled as a test channel ( often gaussian for simplicity of analysis ) for which uncompressed signals as the input and compressed signals are the output .
the compression design is equivalent to setting the variance of the additive compression noise @xcite . to achieve successful compression
, the encoder needs to transmit to the decoder at a rate at least equal to the mutual information between the input and the output over the gaussian test channel .
intuitively , a tighter fronthaul capacity constraint would therefore require a more coarse compression " with a larger compression noise .
existing compression designs in general take the following approaches .
* _ joint compression across different bss : _ when multiple bss compress and forward their received signals to the cu in uplink , the compression design requires setting the covariance of the compression noises across different bss .
a common objective is to maximize the information rate under the fronthaul capacity constraints . in this
setting , _ distributed wyner - ziv lossy compression _ can be used at the bss , exploiting signal correlation across the multiple bss @xcite .
the distributed wyner - ziv compression scheme is shown to yield significant capacity gains over independent quantization methods especially in the low backhaul capacity region @xcite .
similar data compression methods could also be applied in downlink .
interestingly , it has been shown in @xcite that downlink compression and multi - user precoding design ( for interference mitigation ) could be designed separately without compromising maximum system throughput , which is achieved by an optimal but much more complicated joint compression - precoding design . * _ independent bs - level compression : _ the practical implementation of distributed wyner - ziv compression is difficult mainly because of the high complexity in determining the optimal joint compression codebook and the joint decompressing / decoding at the cu . accordingly ,
independent compression methods , where the quantization codebook at a bs is only determined by its local signal - to - noise ratio ( snr ) , can be used to reduce the computational complexity and the signaling exchange overhead in the fronthaul . * _ uniform scalar quantization : _ even when using independent bs - level compression , real - time computation and exchange of quantization codebooks using the information - theoretical source coding approaches are often difficult to realize in practice .
instead , simple uniform scalar quantization methods compatible with a / d modules are proposed to reduce the implementation cost @xcite .
interestingly , it is shown in @xcite that the achievable rate using simple uniform scalar quantization in fact performs closely to that of the gaussian test channel model .
this indicates that efficient fronthaul capacity usage is achievable in practical systems with simple quantization methods .
besides acting as relays to compress / decompress and forward the user signals , bss with advanced baseband processing capabilities could also encode / decode the received messages to further improve the system performance under stringent fronthaul capacity constraints .
* _ partial cooperation : _ in uplink , one direct method to reduce fronthaul traffic is to limit the number of cooperating elements when serving mobile users .
many sparsity inducing optimization methods could be applied to satisfy a certain quality of service level using minimum numbers of sub - channels , antennas or cooperating bss . in downlink ,
similar _ sparse precoding _ methods could be studied to optimize precoders by jointly maximizing the user utilities ( e.g. , data rate ) and minimizing the total number of data streams in the fronthaul @xcite . *
_ distributed encoding / decoding : _ distributed decoding allows the bss to decode user messages locally without forwarding quantized signals to the cu .
for instance , @xcite considers a rate - splitting approach to divide an mt s message into two parts , where one part is decoded locally by the serving bs and the remainder is compressed and jointly decoded by the cu . in another case
, @xcite proposes an opportunistic hybrid decoding method , where a user s message is either decoded locally at a bs when its snr is sufficiently high , or jointly decoded by the cu based on signals forwarded from a subset of cooperating bss when the snr at each individual bs is too low .
note that the locally decoded user messages can be used to cancel their interferences to the received rf signals at the bss , which can effectively reduce the amount of data transmitted to the cu over the fronthaul links . in the downlink case ,
bss could _ encode and modulate _ the baseband symbols to rf signals before transmitting them to the mts .
therefore , instead of transmitting complete signal waveforms ( or waveform samples ) to the bss , cu could save fronthaul bandwidth by transmitting separately the information symbols and the beamforming vectors , while leaving rf modulation to the bss . , @xmath12 , and @xmath13 mbps , respectively.,scaledwidth=90.0% ] to show the performance advantage of the hybrid signal processing model
, we present a numerical example in fig . [
73 ] to compare the throughput performance among the bs - centric , the cloud - centric , and the hybrid processing networks .
let us consider a cellular uplink , where @xmath8 mts transmit over orthogonal sub - channels , each with @xmath14 mhz bandwidth .
besides , each fronthaul link has @xmath15 gbps capacity .
the decoding methods of the three networks are described as follows .
* bs - centric network : bs@xmath16 decodes the messages from mt@xmath16 and mt@xmath17 , and bs@xmath17 decodes the message from mt@xmath18 . then both the bss send the decoded messages to the cu ; * cloud - centric network : both bss compress the received signals using the scalar quantization method considered in @xcite .
they then forward the compressed signals to the cu for joint decoding . in particular , each user
is equally allocated @xmath19 mbps fronthaul bandwidth at a bs to transmit its compressed signal ; * hybrid processing network : bs@xmath16 and bs@xmath17 first decode the messages from mt@xmath16 and mt@xmath18 , respectively , before transmitting the decoded messages to the cu .
meanwhile , each bs uses the remaining fronthaul bandwidth to compress and transmit the signal from mt@xmath17 to the cu for the joint decoding of mt@xmath17 s message . from the aforementioned network setups , we calculate in fig .
[ 73 ] the achievable user data rates under a random channel realization , and compare the common - throughput performance ( the minimum data rate among the three users ) in different cases .
we can see that the bs - centric network achieves the lowest common - throughput , owing to the low data rate of the cell - edge user mt@xmath17 , which is only @xmath20 mbps .
the cloud - centric network slightly improves the data rate of mt@xmath17 and hence the common - throughput to @xmath12 mbps , thanks to its joint processing gain .
however , the data rates of mt@xmath16 and mt@xmath18 are severely degraded , since the limited fronthaul capacity introduces high compression noises to the useful signals .
the hybrid processing network achieves the highest common - throughput ( @xmath13 mbps ) among the three schemes that we considered , which is @xmath21 and @xmath22 higher than those of the bs - centric and cloud - centric networks , respectively .
compared to the cloud - centric networks , by decoding the messages from mt@xmath16 and mt@xmath18 at the bs - level , the hybrid processing network has a larger fronthaul bandwidth to spare for transmitting mt@xmath17 s signals to the cu with more refined compression , thus achieving a higher joint processing gain . in downlink transmission ,
caching at the bss is cost - effective to reduce real - time traffic on fronthaul , thereby enabling significant improvement on the overall c - ran performance .
cache - assisted wireless resource allocation is a cross - layer approach that incorporates the status of application - layer data flow in wireless physical - layer design . as an illustrative example in fig .
[ 74 ] , bs@xmath17 serves two requests from the two mts , whereas caches of the other two bss are empty .
although mt@xmath16 is closer to bs@xmath16 with a better wireless channel condition , the maximum downlink data rate is only @xmath23 unit per second if bs@xmath16 is selected to transmit directly , due to the constraint of link congestion between bs@xmath16 and the cu . instead , the cu could select bs@xmath17 to send the cached contents to mt@xmath16 at a rate of @xmath24 units per second , whose end - to - end data rate is not constrained by the congestion level of the cu - to - bs@xmath17 link . on the other hand , mt@xmath17 could be served by two cooperating bss ( bs@xmath17 and bs@xmath18 ) with an improved wireless channel gain from coordinated beamforming . in particular , the cu only needs to transmit the content requested by mt@xmath17 to bs@xmath18 before the cooperative transmissions of the two bss .
thanks to such wireless cooperation , mt@xmath17 could achieve a higher data rate at @xmath8 units per second . in a more general setting
, caches could be located at not only the bss , but also the routers and the cu .
furthermore , distributed caching could also be adopted at mts to allow mobile users to serve popular contents requested by nearby peer users in a device - to - device ( d2d ) manner .
we could foresee that cache - assisted resource allocation method becomes a key enabling factor of significant bandwidth saving , since frequent overlapping of requested objects will occur as the volume of mobile traffic increases .
however , it also becomes a more challenging problem to optimize system - wide resource allocation due to the interleaving among cache placement , wireless interference , routing , and the combinatorial nature of node selections in the wireless network .
a more comprehensive understanding on the design tradeoff remains open for future study .
another interesting topic on cache - assisted resource allocation is on _ cache provisioning _ for popular contents to reduce the real - time backhaul traffic . in particular
, cache provisioning addresses the questions of what , where and when to cache in the wireless infrastructure . in this case ,
accurate knowledge of the mobile user demand profiles is a key to efficient cache provisioning .
the extraction of user demand profiles from mobile data traffic is performed by wireless bigdata analytics , which will be discussed in the next section .
instead of viewing mobile bigdata as a pure burden , we investigate in this section the potential performance gain from developing a bigdata - aware intelligent wireless network . however , its efficient operation relies on the in - depth knowledge of the wireless bigdata traffic characteristics .
as most of such characteristics are implicit , we first introduce data - analytical methods necessary to extract these bigdata features .
we then discuss how to leverage these bigdata characteristics in designing wireless networks to capitalize from the mobile bigdata traffic .
there is clearly a strong connection between wireless service usage and human behavioral patterns in the physical world .
for this reason , wireless data traffic contains strong correlative and statistical features in various dimensions , such as time , location and the underlying social relationship , etc . on one hand ,
mobile traffic has strong _ aggregate features_. for instance , there exist severe load imbalances spatially and temporally , such that , presently , @xmath25 of popular " bss carry about @xmath26 traffic load .
the peak traffic volume at a given location is much higher than the regular average .
these aggregate features could be exploited to reduce real - time fronthaul / backhaul traffic and to improve wireless network efficiency .
example applications include : cell planning according to geographical data usage distribution , peak load shifting via load - dependent pricing , and cache provisioning based on aggregate demand profile , among others .
on the other hand , each mobile user s data usage profile also exhibits a unique set of _ individual features _
, such as mobility pattern , preference of various data applications , and service quality requirements .
for instance , a mobile user s trajectories often consist of a very limited number of frequent positions and quasi - repetitive patterns .
besides , the recent popularity of mobile social networking interconnects seemingly uncorrelated individual data usages into a unified social profile , thereby presenting a novel perspective to analyze the mobile traffic pattern .
these individual and social features are useful for system operators to personalize and improve wireless service quality .
many intelligent data - aware services could be provided according to user profiles .
examples include resource reservation in handoff using location prediction , context - aware personal wireless service adaptation , and mobility - based routing and paging control .
the ability to acquire , analyze , and exploit mobile traffic characteristics can be accomplished by specially designed learning units ( lus ) installed at both the bss and cus ( see fig . [ 72 ] ) .
their core enabling factors are the embedded data - analytical algorithms .
some commonly used algorithms for wireless traffic analysis and their main applications to wireless communications are classified as follows and summarized in table i. stochastic modeling methods use probabilistic models to capture the _ explicit _ features and dynamics of the data traffic .
commonly used stochastic models include : order-@xmath27 markov model , hidden markov model , geometric model , time series , linear / nonlinear random dynamic systems , etc .
for example , markov models and kalman filters are widely used to predict user mobility and service requirements @xcite .
the collected user data are often used for parameter estimation of stochastic models , such as estimating the transition probability matrix of a markov chain .
data mining focuses on exploiting the _ implicit _ structures in the mobile data sets . also taking the mobility prediction problem as an example
, individual user s mobility pattern could be extracted and discovered by finding the most frequent trajectory segments in the mobility log .
prediction could be made accordingly by matching the current trajectory to the mobility profile .
clustering is another useful technique to identify the different patterns in the data sets .
it is widely used in context - aware mobile computing , where a mobile user s context and behavioral information , such as sleeping and working , are identified from wireless sensing data for providing context - related services @xcite .
the main objective of machine learning is to establish functional relationship between input data and output actions , thus achieving _ auto - processing _
capability for unseen patterns of data inputs . among the many useful techniques in
machine learning applied to wireless communications , classification ( determining the type of input data ) and regression analysis ( data fitting ) are two common methods , whose applications include context identification of mobile usage and prediction of traffic levels ( classification ) , or fitting the distributions of trajectory length , mobile user location , and channel holding times ( regression ) .
besides , reinforcement learning , such as q - learning @xcite , is useful for taking proper real - time actions to maximize certain long - term rewards .
a typical example is making the handoff and admission control decision ( action ) , given the current traffic load ( state ) and incoming new requests ( event ) , in which the reward could be evaluated against the reduction of dropped calls or failed connections .
.summary of common wireless bigdata analytic tools and example applications [ cols="^,^,^",options="header " , ] wireless bigdata poses many challenges to the aforementioned conventional data - analytical methods due to its high volume , large dimensionality , uneven data qualities , and the complex features therein . to improve signal processing efficiency
, one can combine the following complexity reduction techniques with the conventional data analytical tools for large - scale data processing .
* _ distributed optimization algorithms _ , such as primal / dual decomposition and alternating direction method of multipliers ( admm ) , are very useful to decouple large - scale statistical learning problems into small subproblems for parallel computations so as to relieve both the computational burden at the cu and the bandwidth pressures to the fronthaul / backhaul links .
* _ dimension reduction _ methods are useful to reduce the data volume to be processed while capturing the key features of bigdata . among various methods , principle component analysis ( pca ) , along its many variants , is the mostly used method today .
in addition , tensor decomposition methods are also popular in mobile data processing , which seek to approximately represent a high - order multi - way array ( tensor ) as a linear combination of outer products of low - order tensors . by doing so
, the hardware requirement and cost for storing the high - order arrays of mobile data could be reduced . * _ other advanced learning methods _ could be used to handle incomplete or complex data sets .
interesting examples include active learning , which deals with partially labeled data set ; online learning for responding in real - time to sequentially received data ; stochastic learning that makes a decision periodically in each time interval ; and deep learning for modeling complex behaviors contained in a data set .
once identified and extracted , data characteristics could be used to improve wireless service quality and generate new mobile applications . for simplicity of illustration ,
we have postulated in fig .
@xmath28 a structure of bigdata aware wireless network , consisting of several mutually complementary components that enable data - driven mobile services , whose functionalities are described below . * _ data - aware cache management : _ for quick access under high traffic volumes , cached contents need to be carefully categorized , compactly organized and timely updated .
many types of content objects , such as music and video files , are embedded with _
metadata labels _ that describe the properties of the contents , from which the data contents could be well classified . by classifying data into a number of sub - classes based on contents , such as sports videos and news pictures , the lus could achieve more accurate evaluation of the content popularity by jointly considering its own access count and the total access count of its type , which reflects the average frequency of potential future accesses . accordingly , popular contents are continuously cached while unpopular contents are removed regularly to maximize the effective system bandwidth given limited cache size .
* _ crowd computing : _ mobile users of similar interests could share their resources with peers in their vicinity , either with or without taking advantage of the wireless infrastructure .
for instance , a complete 3d street view could be generated by a bs from relevant photos contributed by users from different angles .
meanwhile , when mt - to - bs connection is unavailable , an mt could ask for assistance from its neighboring mts to share available contents and applications , or to even act as relays to the cellular network , etc .
such an idea is explored in @xcite , where a crowd - enabled data transmission mechanism is proposed to let mobile users assist the data dissemination of other users .
in particular , it makes use of personal social information and market incentives to enhance the willingness " of mobile users for acting as a data broker of others such that higher chance of successful data delivery could be achieved .
essentially , this peer - to - peer nature of crowd computing exploits user mobility and spatial correlation of data traffics , which also helps us reduce the conventional cellular traffic to and from the wireless infrastructure . *
_ mobile cloud processing : _
multiple interconnected c - rans constitute a mobile cloud , which could optimize the wireless services based on knowledge with respect to mobile traffic patterns , especially when user mobility spans across different c - ran clusters .
for instance , based on the mobility pattern of an mt , a cu could reserve channel resource in advance and pre - feed the contents to the bss along the anticipated mt s route .
as such , chunks of contents could be sent from different bss to achieve seamless handoffs .
similarly , aggregate characteristic behavior of data traffic could also be used to allocate resources such as bandwidth and cache space to some popular locations ahead of some real - time events .
this approach could evidently reduce connection time , delay jitter , and burden of real - time traffic bursts on both cellular fronthaul and backhaul . *
_ wireless cloudlet : _ the concept of cloudlets introduced in @xcite defines a self - organized light cloud with limited storage and computing power installed at the bss to enhance their local data processing capability .
the deployment of cloudlets could effectively reduce the packet round - time delay by an order of magnitude .
a cloudlet may be owned by the network operator but leased to commercial clients for improving performance of delay - sensitive applications , such as online gaming .
besides , a cloudlet could also allow commercial clients to access local cache to provide better location - based services . for instance , an advertising company could send to its subscribers in the vicinity the latest deals based on the information posted by local stores and queries made by prospective customers .
with cloudlet , real - time traffic in the backhaul network could also be largely reduced , since many services could be provided locally instead of burdening the core network . * _ context / social - aware processing : _ context / social - aware computing is an emerging paradigm for exploiting complex data characteristics besides conventional user profiles such as mobility pattern and demand distribution @xcite .
the idea of context - aware computing is to provide personalized services adaptive to the mt s real - time context " , such as traveling , working , and recreation , either directly reported by the mt or inferred from various available data .
social computing , on the other hand , calls for wireless resource allocation to follow closely the interaction within and among social groups @xcite .
conceptually , a social group is a subset of users that share some similar interests , professions , hobbies , and life experiences , etc . in general
, a social group has unique eigenbehaviors " , such that the group members require and generate similar data contents .
the knowledge of a social community s composition , activities and interests could be used to improve the wireless services for the targeted social group members . * _ software - defined - network ( sdn ) : _ sdn replaces the conventional hardware - configured routing and forwarding devices by software programmable units .
in particular , it decouples the user s data plane ( u - plane ) from the control and management plane ( c - plane ) , such that the network is managed by a central controller while the underlying devices are only responsible for simple functions such as packet forwarding .
such decoupling provides unprecedented flexibility to network traffic management , where packet forwarding decisions may now be programmed based on many new considerations such as qos ( quality of service ) requirement , application types , and payload length , in addition to the conventional destination oriented and distance - based metrics . for sdn - enabled wireless networks , @xcite proposes a flow - based resource management framework in c - ran , where the packet routing in the backhaul network and beamforming design in the wireless access network are jointly optimized based on individual data flow s source - destination pair , wireless channel condition , backhaul link capacity , and user qos requirements , etc . in the case of wlan networks , @xcite introduces a sdn - based enterprise wlans framework named odin , which is built with programable functions , global knowledge of network status , and direct control of network devices .
the sdn - based system makes many difficult or costly tasks in conventional wlans easier and less inexpensive , including seamless user handoffs , global load balancing , and hidden terminal problem mitigation .
in the mobile bigdata era , wireless system designs contain rich research problems of important applications and impact that are yet to be studied . beyond the many research issues that arise among the number of topics we have discussed so far , here in this section , we highlight several interesting research topics that we particularly find exciting . in many data compression proposals , real - time calculation of the optimal compression noise covariance matrix
is often impeded by the large number of fronthaul capacity constraints and the non - convex nature of many fronthaul - constrained problems .
the problem is further exacerbated by the difficulty in generating practical joint compression codebooks based on the obtained covariance matrix .
therefore , sub - optimal but practical compression schemes , such as scalar quantization , should be given more consideration in future study of fronthaul - constrained compression design .
similarly , cu - level encoding and decoding also suffers from high computational complexity on large - scale multi - user detection and the combinatorial nature of many limited cooperation schemes , such as optimal antenna , relay , modulation and coding combinations , as well as bs selections .
it therefore calls for practical complexity - reduction algorithms that are truly scalable to the number of mobile users and network entities .
bs - level caching is expected to play an important role in future wireless bigdata processing , due to its simplicity , low cost , and natural integration with bigdata analytical tools .
however , research on cache - assisted wireless resource allocation is still in its infancy . for cache - assisted cellular networks with bs - level caching , currently there is a shortage of both concrete theoretical analysis on the capacity gain of cache - assisted processing and practical optimization frameworks for cache - assisted resource allocation .
furthermore , effective and optimized integration of various identified bigdata characteristics in cache - assisted network design is an interesting problem that awaits future investigations . in large - scale wireless networks ,
distributed control / computing algorithms could be integrated to alleviate computational complexity of the cu , to reduce backhaul traffic volume and to mitigate the risk of single point failures without compromising overall system performance .
owing to the programmability of sdn - enabled system infrastructure , distributed control mechanisms could be implemented with much better flexibility and lower cost .
however , the feasibility and complexity reduction of distributed algorithms are often constrained by the underlying problem structure , such as the coupling constraints in the backhaul and the partial knowledge of data traffic , etc . distributed control , or a mixed centralized and decentralized control framework ,
is a promising working direction towards a future wireless networking design supporting mobile bigdata .
additionally , the sdn - based design may also incorporate distributed caching ( at bss and routers ) to enhance the efficiency of the routing decision .
harvesting over large mobile data sets and data analytics naturally give rise to concerns with respect to data security and privacy . in a cloud - based wireless network ,
large amount of data is stored in the fronthaul / backhaul network either for customers personal use or as commercial database for future analytical purposes .
the system operators or commercial entities that collect the user data should be responsible for data security and privacy .
for example , personal data should be only available for legitimate and authenticated users .
similarly , data integrity should be guaranteed such that no data is lost or modified by unauthorized entities .
furthermore , it is also important to maintain confidentiality of user data when they are either in storage or during processing .
it is therefore important to develop secure yet efficient data processing and storage methods .
promising security measures may include privacy aware distributed data storage and decentralized processing , which aim to maintain local data confidentiality .
this article addresses challenges and opportunities that we face in the era of wireless big data .
we first reviewed state - of - the - art signal processing methods and networking structures that may allow us to effectively manage and in fact take advantage of wireless bigdata traffic .
we outlined the major obstacles of bigdata signal processing and network design with respect to the scale of problem size and the complex problem structures .
nevertheless , research on big data for wireless communications and networking is not only promising but also inevitable in light of the continuing data volume explosion .
we also suggested several interesting research problems aimed at stimulating future wireless research innovations in the bigdata era .
s. h. park , o. simeone , o. sahin , and s. shamai , joint precoding and multivariate backhaul compression for the downlink of cloud radio access networks , " _ ieee trans .
signal process .
_ , vol .
61 , no .
22 , pp . 5646 - 5658 , nov . 2013 .
l. liu , s. bi , and r. zhang , joint power control and fronthaul rate allocation for throughput maximization in ofdma - based cloud radio access network , " available online at http://arxiv.org/abs/1407.3855 m. hong , r. sun , h. baligh , and z. q. luo , joint base station clustering and beamformer design for partial coordinated transmission in heterogeneous networks " , _ ieee j. sel .
areas commun .
_ , vol . 31 , no . 2 , pp . 226 - 240 , feb
. 2013 .
a. krause , a. smailagic , and d. p. siewiorek , context - aware mobile computing : learning context - dependent personal preferences from a wearable sensor array , " _ ieee trans .
mobile comput .
_ , vol . 5 , no.2 , pp . 113 - 127 , feb
k. l. a. yau , p. komisarczuk , and p. d. teal , reinforcement learning for context awareness and intelligence in wireless networks : review , new features and open issues , " _ journal of network and computer applications _ ,
35 , no . 1 ,
pp . 253 - 267 , jan . 2012 .
w. c. liao , m. hong , h. farmanbar , x. li , z. q. luo , and h. zhang , min flow rate maximization for software defined radio access networks , " _ ieee j. sel .
areas commun .
_ , vol.32 , no.6 , pp . 1282 - 1294 , june 2014 .
|
the rapidly growing wave of wireless data service is pushing against the boundary of our communication network s processing power .
the pervasive and exponentially increasing data traffic present imminent challenges to all the aspects of the wireless system design , such as spectrum efficiency , computing capabilities and fronthaul / backhaul link capacity . in this article , we discuss the challenges and opportunities in the design of scalable wireless systems to embrace such a bigdata " era . on one hand , we review the state - of - the - art networking architectures and signal processing techniques adaptable for managing the bigdata traffic in wireless networks . on the other hand , instead of viewing mobile bigdata as a unwanted burden ,
we introduce methods to capitalize from the vast data traffic , for building a bigdata - aware wireless network with better wireless service quality and new mobile applications .
we highlight several promising future research directions for wireless communications in the mobile bigdata era .
|
present day colliders test the standard model at an energy scale of order 100 gev .
precision experiments at lep , slc and tevatron ( with some additional measurements at lower energies ) have measured more than twenty separate experimental observables , and have confirmed the standard model predictions with an accuracy of one part in a thousand @xcite .
a few anomalies in the data could suggest hints of new physics beyond the standard model @xcite , although no deviations have been rigorously confirmed . nevertheless , the verification of the standard model is not yet complete .
absent to date is any experimental signal that sheds light on the dynamics responsible for electroweak symmetry breaking .
any consistent theory of electroweak symmetry breaking must generate goldstone bosons which are absorbed by the @xmath0 and @xmath1 gauge bosons , thereby generating the gauge boson masses .
the standard model posits that electroweak symmetry breaking is due to the dynamics of a weakly - coupled complex doublet ( with hypercharge one ) of elementary scalar fields .
the physical consequence of this model is the existence of a cp - even neutral higgs boson with mass roughly of order @xmath2 .
extensions of this model can easily be constructed , in which the scalar sector is enlarged .
the resulting model then contains a _ non - minimal _ higgs sector consisting of neutral higgs bosons ( of definite or indefinite cp depending on the model ) and charged higgs bosons @xcite .
the best motivated non - minimal higgs sector is the two higgs doublet model . starting with two complex scalar doublets of hypercharge @xmath3 respectively
, one finds a higgs sector ( after three goldstone bosons are absorbed to give mass to the @xmath0 and @xmath1 ) consisting of five states : a light cp - even higgs scalar , @xmath4 , a heavy cp - even scalar , @xmath5 , a cp - odd scalar @xmath6 , and a charged higgs pair , @xmath7 .
this is the higgs sector of the minimal supersymmetric extension of the standard model ( mssm ) @xcite . in the global fits of lep , slc , and tevatron data based on the standard model ,
there is weak ( but non - trivial ) sensitivity to the higgs boson mass by virtue of higgs mediated radiative corrections .
the most recent global fits find that @xmath8 gev at @xmath9 cl @xcite , although some care needs to be taken in interpreting this limit @xcite .
the potential for improving this bound at future colliders is discussed in section ii . in the context of the standard model ,
the higgs boson in this mass range is necessarily weakly - coupled .
moreover , such fits also apply to non - minimal higgs sectors in which the lightest higgs scalar ( @xmath4 ) is separated in mass from heavier non - minimal higgs states .
therefore , there is a strong motivation to conduct a vigorous experimental search for weakly - coupled higgs bosons at lep and future colliders .
if a higgs boson with standard model properties were discovered , then one might naively conclude that the search for the model of the elementary particles has been completed .
however , theorists strongly believe that the standard model can not be the fundamental model of particles . apart from the many parameters of the standard model which must be inserted by hand ( with no explanation )
, there is a theoretical problem in the standard model associated with the very large hierarchy of energy scales .
we know that the planck scale , @xmath10 gev , exists in nature ; it characterizes the energy scale above which gravitational interactions can not be neglected relative to the strong and electroweak interactions of the elementary particles .
given the existence of such a large energy scale , one must explain how the scale of electroweak symmetry breaking , which is so small when expressed in units of the planck scale ( @xmath11 ) , could be generated by a fundamental theory of particles that includes gravity .
related to this question is the theoretical problem of generating a `` naturally '' light higgs boson ( with a mass of order @xmath12 ) , since in the standard model , there is no symmetry that can protect the mass of an elementary scalar from being driven up to @xmath13 via radiative corrections .
these problems are intimately connected with the dynamics that generates electroweak symmetry breaking .
attempts to solve the problem of hierarchy and the related problem of the unnaturally light higgs boson inevitably lead to the existence of new physics at the 1 tev energy scale or below .
possible mechanisms invoke either supersymmetry @xcite ( a symmetry that can protect the masses of elementary scalars ) or dynamical electroweak symmetry breaking @xcite ( which typically eliminates elementary scalar fields completely ) .
which path nature chooses can only be determined through experimentation .
thus the central goals of the future colliders program are : to explore the dynamics of electroweak symmetry breaking , and to determine its implications for the structure of the standard model and the nature of physics that lies beyond the standard model . in this report
, we assume that nature chooses a weakly - coupled higgs sector as the source of electroweak symmetry breaking dynamics .
we do not address the alternative approach which invokes strong interaction dynamics as the source of electroweak symmetry breaking .
the phenomenology of the electroweak symmetry breaking sector in this latter case is explored by the strongly interacting electroweak symmetry breaking working group @xcite .
the focus of this working group is the weakly - coupled higgs sector of the standard model , and possible non - minimal higgs sector extensions ( including the higgs sector of the mssm ) .
although there is considerable freedom for the structure of the scalar sector ( even after imposing all known theoretical and phenomenological constraints ) , models of the scalar sector often exhibit the following structure : ( i ) the lightest scalar ( @xmath4 ) is a cp - even neutral higgs boson with couplings closely approximating those of the standard model higgs boson ( @xmath14 ) , and ( ii ) additional higgs scalars ( neutral higgs bosons with definite or indefinite cp quantum numbers and charged higgs bosons ) are expected to be heavier ( perhaps significantly heavier ) than @xmath4 , although still weakly - coupled .
this is the so - called _ decoupling limit _ which will be discussed in section iiia . in this case
, the discovery of @xmath15 is not sufficient to probe the underlying structure of the electroweak symmetry breaking sector .
the essence of the decoupling limit is that the existence of a light cp - even higgs boson with properties closely approximating those of @xmath14 is consistent with many possible non - minimal higgs sectors .
thus , the discovery of the heavy non - minimal higgs scalars is essential in order to probe the details of the electroweak symmetry breaking dynamics .
the mssm provides a natural framework for light elementary higgs scalars .
the higgs sector of the mssm is a constrained two - higgs - doublet model , whose tree - level properties are determined by two free parameter ( typically chosen to be the mass of the cp - odd state , @xmath6 , and the ratio of vacuum expectation values , @xmath16 ) .
the decoupling limit of the model corresponds to @xmath17 ; in this limit , the properties of @xmath4 become identical to those of @xmath14 .
extensions of the mssm higgs sector are also possible . for example , the simplest non - minimal supersymmetric extension of the standard model ( nmssm ) consists of a higgs sector with two doublets and one singlet of complex higgs fields @xcite .
thus , a detailed exploration of the scalar sector has the potential for probing both the electroweak symmetry breaking dynamics and the underlying supersymmetric structure of the theory .
present experimental data tells us that the higgs sector must be compatible with ( ) .5 in .7 in @xmath18 ; the absence of significant higgs mediated flavor changing neutral currents ; the absence of significant virtual charged higgs mediated effects ( which can contribute , _
e.g. _ , to @xmath19@xmath20 mixing , @xmath21 and @xmath22 ) .
even after imposing such model constraints , there is still significant freedom in the structure of the higgs sector .
exotic higgs sectors ( beyond those mentioned above ) are easily constructed that satisfy all present day phenomenological constraints .
such higgs sectors could arise in models with extended gauge groups , models with exotic scalar multiplets , or models with a lepton number violating sector ( _ e.g. _ , in r - parity violating models of low - energy supersymmetry @xcite , in which there is no distinction between scalar lepton superpartners and higgs bosons ) . sorting out the details of the scalar sector
will be one of the fundamental challenges for future collider experimentation .
the weakly - coupled higgs boson and precision electroweak physics working group program consisted of the following tasks : 1 .
extend present day precision tests of the standard model + this will serve to tighten constraints on the higgs sector and perhaps uncover deviations from the standard model and provide evidence for new physics beyond the standard model .
2 . evaluate the higgs boson discovery reach of future colliders + high energy colliders are needed to directly produce the massive higgs bosons .
however , the cleanest decay channels of the higgs boson usually have rather small branching ratios .
thus , high luminosity is critical to insure that higgs signals can be extracted from the standard model backgrounds . in table
[ colliders ] , we list the approved and possible future collider facilities considered in our study . furthermore , special features of the collider detectors ( such as the high resolution for the electromagnetic calorimeter for @xmath23 , and high @xmath24 tagging efficiency ) are also required in order to maximize the significance of the higgs signal .
thus , establishing the discovery reach for future colliders is an important and non - trivial first step in the pursuit of the higgs boson .
3 . consider precision measurements of @xmath4 properties + the discovery of the higgs boson will complete the experimental verification of the standard model .
once the higgs boson is discovered , one must check that it does indeed possess couplings to particles proportional to their masses .
one should quickly be able to verify that the properties of the scalar state roughly match those expected for the standard model higgs boson .
more precise measurements may be required to detect deviations of the observed higgs properties from that of the @xmath14 .
the difficulty of this latter task will depend on how close one is to the decoupling limit ( see section iiia ) .
4 . evaluate the potential for direct detection of the non - minimal higgs states and the measurement their properties + this is essential for probing the nature of the electroweak symmetry breaking dynamics .
in addition , the non - minimal higgs states may be sensitive to physics associated with the hierarchy problem ( for example , the properties of the non - minimal higgs states in supersymmetric models can provide important checks of the supersymmetric dynamics ) .
the non - minimal higgs sector imposes the most stringent requirements on the collider facility . to accomplish
this task may require the highest energies and luminosities now being considered .
lccc & type & @xmath25 & annual @xmath26 + & & & + lep-2 & @xmath27 & 192 gev & 170 pb@xmath28 + tevatron ( m.i . ) & @xmath29 & 2 tev & 2 fb@xmath28 + lhc & @xmath30 & 14 tev & 100 fb@xmath28 + & & & + tev-33 & @xmath29 & 2 tev & 30 fb@xmath28 + nlc@xmath31 & @xmath27 & 0.51.5 tev & 50200 fb@xmath28 + + fmc & @xmath32 & 0.54 tev & 501000 fb@xmath28 + this report consists of four parts . following this introduction
, section ii briefly summarizes the results of the precision electroweak physics subgroup .
section iii discusses some theoretical issues that are important for the considerations of the higgs discovery and properties subgroups .
section iv summarizes the essentials of higgs phenomenology at future colliders .
the conclusions and some final thoughts are given in section v. the details underlying sections ii and iv can be found in the subgroup reports that follow this summary report @xcite .
in the electroweak standard model , there are two coupling parameters , @xmath33 and @xmath34 , of @xmath35 gauge interactions .
the vacuum expectation value of the scalar field , @xmath36 , sets the mass scale . at tree level ,
the @xmath0 and @xmath1 boson masses @xmath37 , as well as the weak mixing angle @xmath38 , are determined by these three parameters . alternatively , one may use the precisely measured quantities the electromagnetic coupling constant @xmath39 , the muon decay constant @xmath40 and @xmath2as inputs to evaluate the other electroweak parameters .
when the radiative corrections are taken into account , the relations among these parameters become dependent on @xmath41 , @xmath42 as well as other possible contributions from new physics .
therefore , precision electroweak measurements not only check the consistency of the standard model , but also constrain @xmath42 and other new physics @xcite .
the precision electroweak physics subgroup @xcite paid special attention to the measurements of @xmath43 , @xmath41 and @xmath38 at future collider experiments .
the implications for the constraints on @xmath42 are also discussed .
currently , the world average values for @xmath44 and @xmath41 are @xmath45 the precision which can be achieved for @xmath44 and @xmath41 measurements at different colliders is summarized in table [ tab : wmass ] and table [ tab : top ] , respectively .
table entries are taken from ref .
@xcite unless otherwise indicated .
.expected @xmath46 mass precision at future colliders .
[ cols="<,^ " , ] -3pt @xmath47near the @xmath1 peak , the expected fmc uncertainty in @xmath48 is about 30% .
[ widthbrs ] in models of non - minimal higgs sectors , precision measurements of the branching ratios and partial ( and total ) decay rates of the lightest cp - even higgs boson could prove that @xmath49 , thereby providing indirect evidence of the non - minimal higgs states .
once the non - minimal higgs bosons are directly discovered , detailed measurements of their properties would yield significant clues to the underlying structure of electroweak symmetry breaking .
for example , if the higgs sector arises from a two - doublet model , then precision studies of the heavy higgs states can provide a direct measurement of the important parameter @xmath16 ( the ratio of higgs vacuum expectation values ) . can not be distinguished from @xmath14 ) , measurements of processes involving @xmath4 alone can not yield any information on the value of @xmath16 . ]
the measurement of @xmath16 can also provide a critical self - consistency test of the mssm , since the parameter @xmath16 also governs the properties of the charginos and neutralinos ( and can in principle be determined in precision measurements of supersymmetric processes ) .
moreover , the couplings of higgs bosons to supersymmetric particles will provide invaluable insights into both the physics of electroweak symmetry breaking and the structure of low - energy supersymmetry .
the possibility that the heavy non - minimal higgs states have non - negligible branching ratios to supersymmetric partners can furnish an additional experimental tool for probing the higgs boson supersymmetry connection . as in the case of the @xmath14 discussed above , the lepton colliders ( assuming @xmath50 for the nlc and @xmath51 for the fmc ) provide the most powerful set of tools for extracting the magnitudes of the higgs couplings to fermion and vector boson pairs .
the higgs couplings to vector boson pairs directly probe the mechanism of electroweak symmetry breaking [ via the sum rule of eq .
( [ sumrule ] ) ] .
the higgs coupling to two photons , depends ( through their one - loop contributions ) on all charged states whose masses are generated by their couplings to the higgs sector .
precision measurements of the higgs couplings to fermions are sensitive to other higgs sector parameters ( _ e.g. _ , @xmath52 and the neutral higgs mixing parameter @xmath39 in a two - higgs - doublet model ) .
additional information can be ascertained if higgs self - interactions could be directly measured .
this would in principle provide direct experimental access to the higgs potential .
unfortunately , there are very few cases where the measurement of higgs self - couplings has been shown to be viable @xcite .
finally , one should also consider the possible effects of virtual higgs interactions @xcite . in some models ,
flavor changing neutral currents mediated by neutral higgs bosons may be observable .
the cp - properties of the heavy higgs states could be mixed , leading to higgs mediated cp - violating effects that could be observed in processes with heavy flavor . in some cases , precision measurements of low - energy observables can be quite sensitive to the heavy higgs states .
the canonical example is the process @xmath21 , which can be significantly enhanced due to charged higgs boson exchange . if there are no other competing non - standard model contributions ( and this is a big _ if _ ) , then present data excludes charged higgs masses less than about 250 gev @xcite .
eventually , when non - minimal higgs states are directly probed , it is essential to check for the consistency between their properties as determined from direct observation and from their virtual effects .
this working group has examined the potential of a program of future precision electroweak measurements @xcite and the search for weakly - coupled higgs boson at future colliders @xcite .
the goal of such a program is to address the outstanding problem of elementary particle physics : what is the origin of electroweak symmetry breaking and the nature of the dynamics responsible for it ? improvements of precision electroweak measurements can provide an important consistency check of the higgs interpretation of a higgs signal , in the same way that the lep electroweak data provided support for the tevatron interpretation of the top - quark events .
discovery of a higgs - like signal alone may not be sufficient to earn a place in the particle data group ( pdg ) tables .
some basic measurements of the properties of the higgs candidate will be essential to confirm a higgs interpretation of the discovery .
higgs searches at lep-2 , run - ii of the tevatron and/or the lhc will almost certainly discover a higgs signal if electroweak symmetry breaking dynamics is weakly coupled .
moreover , measurements at these machines will yield evidence for the higgs interpretation that is sufficient to pass pdg muster in some fraction of the higgs parameter space .
the nlc ( and fmc ) can discover or definitively rule out the existence of a higgs boson in the intermediate higgs mass regime ( the mass region most problematical for the higgs search at hadron colliders ) .
once a higgs boson is discovered , the lepton colliders would play a decisive role in the precision measurement of higgs sector properties .
it is not unlikely that the first higgs state to be discovered will be experimentally indistinguishable from the standard model higgs boson .
this occurs in many theoretical models that exhibit the decoupling of heavy scalar states . in this decoupling limit , the lightest higgs state
, @xmath4 is a neutral cp - even scalar with properties nearly identical to the @xmath14 , while the other higgs bosons of the non - minimal higgs sector are heavy ( compared to the @xmath1 ) and are approximately mass - degenerate . thus , discovery of @xmath15 may shed little light on the dynamics underlying electroweak symmetry breaking .
it is then crucial to directly detect and explore the properties of the non - minimal higgs states .
in particular , precision measurements are critical in order to distinguish between @xmath4 and @xmath14 and/or to map out the properties of the non - minimal higgs states . to accomplish these goals , future colliders of the highest energies and luminosities , considered in this report , are essential .
we have entered a new era in higgs phenomenology .
the methods by which the first higgs signal will be identified are well known and have been studied in great detail .
however , the most outstanding challenge facing the future higgs searches lies in identifying and exploring in detail the properties of the non - minimal higgs states .
a successful exploration will have a profound effect on our understanding of tev - scale physics .
99 , fermilab - conf-96/354 ( 1996 ) [ hep - ex/9611019 ] , invited talk given at the _ meeting of the division of particles and fields _ , minneapolis , mn , 1015 august 1996 .
[ lep electroweak working group ] and b. schumm and d. su [ sld heavy flavor group ] , cern - ppe/96 - 183 ( 1996 ) , prepared from contributions of the lep and sld experiments to the 28th international conference on high energy physics , warsaw , poland , 2531 july 1996 . see , _
e.g. _ , s. godfrey , _ et al .
_ , summary of the new phenomena working group , in these proceedings . for a review of higgs boson physics ,
see , _ e.g. _ , j.f .
gunion , h.e .
haber , g.l .
kane and s. dawson , _ the higgs hunter s guide _ ( addison - wesley , redwood city , ca , 1990 ) . for reviews on low - energy supersymmetry ,
e.g. _ , h.e .
haber and g. kane , _ phys .
_ * 117 * , 75 ( 1985 ) ; h.e .
haber , in _ recent directions in particle theory _ , proceedings of the 1992 theoretical advanced study institute in elementary particle physics , edited by j. harvey and j. polchinski ( world scientific , singapore , 1993 ) pp .
589686 ; r. arnowitt and p. nath , in _ particles and fields _ , proceedings of the vii jorge andre swieca summer school , sao paulo , brazil , 1023 january , 1993 , edited by j.p .
eboli and v.o .
rivelles ( world scientific , singapore , 1994 ) pp . 363 . , summary of the supersymmetry working group , in these proceedings ; j. amundson _
et al . _ ,
report of the supersymmetry theory subgroup , in these proceedings .
e.g. _ , j. rosner , cern - th/96 - 245 ( 1996 ) [ hep - ph/9610222 ] , to appear in _ masses of fundamental particles _ , proceedings of the cargse summer institute on particle physics , august 1996 , to be published by plenum press . for a review of theories of dynamical electroweak symmetry
breaking , see _
e.g. _ , k. lane , in _ the building blocks of creation _ , proceedings of the 1993 theoretical advanced study institute in elementary particle physics , edited by s. raby and t. walker ( world scientific , singapore , 1994 ) pp .
381408 . , summary of the strong coupling electroweak symmetry breaking working group , in these proceedings .
j. ellis , j.f .
gunion , h.e .
haber , l. roszkowski and f. zwirner , _ phys .
rev . _ * d39 * , 844 ( 1989 ) ; m. drees , _ int . j. mod
. phys . _ * a4 * , 3635 ( 1989 ) ; u. ellwanger , m. rausch de traubenberg and c.a .
savoy , _ phys .
lett . _ * b315 * , 331 ( 1993 ) ; _ z. phys . _ * c67 * , 665 ( 1995 ) ; lpthe - orsay-96 - 85 [ hep - ph/9611251 ] ; t. elliot , s.f .
king and p.l .
white , _ phys .
* b305 * , 71 ( 1993 ) ; _ phys .
rev . _ * d49 * , 2435 ( 1994 ) ; s.f .
king and p.l .
white , _ phys .
rev . _ * d52 * , 4183 ( 1995 ) .
e.g. _ , f. de campos , m.a .
garcia - jareno , a.s .
joshipura , j. rosiek , and j.w.f .
valle , _ nucl .
phys , _ * b451 * , 3 ( 1995 ) ; t. banks , y. grossman , e. nardi , and y. nir , _ phys .
rev . _ * d52 * , 5319 ( 1995 ) .
s. meyers and c. wyss , in _ physics at lep2 _ , volume 1 , edited by g. altarelli , t. sjstrand and f. zwirner , cern yellow report 96 - 01 ( 1996 ) pp .
2343 . , precision electroweak physics subgroup report , in these proceedings .
j.f . gunion _
et al . _ , higgs boson discovery and properties subgroup report , in these proceedings . , _ phys
_ * 65 * , 964 ( 1990 ) ; _ phys .
* d46 * , 381 ( 1992 ) . , invited talk given at the _ meeting of the division of particles and fields _ , minneapolis , mn , 1015 august 1996 ; g. altarelli , cern - th-96 - 265 ( 1996 ) [ hep - ph/9611239 ] , talk given at the nato advanced study institute on techniques and concepts of high - energy physics , st .
croix , u.s .
virgin islands , july 10 - 23 , 1996 .
( ccfr / nutev collaboration ) , in _ electroweak interactions and unified theories _ , proceedings of the xxxi rencontres de moriond , les arcs , france , 1623 march , 1996 , edited by j. trn thanh vn ( editions frontires , gif - sur - yvette , france , 1996 ) pp .
grivaz , t. sjstrand and p.m. zerwas , conveners of the physics working group , to appear in the proceedings of the joint ecfa / desy study : physics and detectors for a linear collider , february november , 1996 .
v. barger , m. berger , j.f .
gunion and t. han , madph-96 - 963 and ucd-97 - 01 ( 1997 ) [ hep - ph/9702334 ] .
m. breidenbach _
et al . _ , slac - cn-409 ( 1996 ) .
kniehl , _ z. phys . _ * c72 * , 437 ( 1996 ) .
smith and s.s .
willenbrock , illinois preprint ( 1996 ) [ hep - ph/9612329 ] .
phys . _ * b335 * , 363 ( 1990 ) . , in _ beyond the standard model iv _ , proceedings of the fourth international conference on physics beyond the standard model , granlibakken , lake tahoe , ca , 1318 december , 1994 , edited by j.f .
gunion , t. han and j. ohnemus ( world scientific , singapore , 1995 ) pp .
151163 ; and in _ perspectives for electroweak interactions in @xmath27 collisions _ , proceedings of the ringberg workshop , ringberg castle , tegernsee , germany , 58 february , 1995 , edited by b.a .
kniehl ( world scientific , singapore , 1995 ) pp .
for a recent review , see , _ e.g. _ , u.m .
heller , _ nucl .
phys b. ( proc . suppl . )
_ * 34 * , 101 ( 1994 ) .
n. cabibbo , l. maiani , g. parisi and r. petronzio , _ nucl .
phys . _ * b158 * , 295 ( 1979 ) ; m. lindner , _ z. phys . _ * c31 * , 295 ( 1986 ) . for a recent review ,
see b. schrempp and m. wimmer , _ prog .
_ * 37 * , 1 ( 1996 ) .
lett . _ * b337 * , 141 ( 1994 ) .
_ * b342 * , 171 ( 1995 ) ; j.r .
espinosa and m. quiros , _ phys .
lett . _ * b353 * , 257 ( 1995 ) .
e.g. _ , chapter 4 of ref .
lett . _ * 66 * , 1815 ( 1991 ) ; y. okada , m. yamaguchi and t. yanagida , _ prog .
phys . _ * 85 * , 1 ( 1991 ) ; j. ellis , g. ridolfi and f. zwirner , _ phys . lett . _ * b257 * , 83 ( 1991 ) . , _ phys .
* b274 * , 191 ( 1992 ) ; _ nucl .
phys . _ * b423 * , 437 ( 1994 ) ; a. dabelstein , _ z. phys . _ * c67 * , 495 ( 1995 ) ; d.m .
pierce , j.a .
bagger , k. matchev and r .- j .
zhang , slac - pub-7180 ( 1996 ) [ hep - ph/9606211 ] , _ nucl .
b _ ( 1997 ) , in press . , _ phys . rev . _ * d48 * , 4280 ( 1993 ) .
lett . _ * b331 * , 99 ( 1994 ) . , cern - th/95 - 216 ( 1996 ) [ hep - ph/9609331 ]
, _ z. phys .
c _ ( 1997 ) , in press .
m. carena , j.r .
espinosa , m. quiros and c.e.m .
wagner , _ phys .
lett . _ * b355 * , 209 ( 1995 ) ; m. carena , m. quiros and c.e.m .
wagner , _ nucl .
_ * b461 * , 407 ( 1996 ) .
_ * 70 * , 2686 ( 1993 )
. y. pan [ for the aleph collaboration ] , invited talk given at the international symposium on recent developments in phenomenology , madison , wisconsin , 1719 march 1997 .
martin , lycen-9644 ( 1996 ) , invited talk given at the 28th international conference on high energy physics , warsaw , poland , 2531 july 1996 .
, cdf collaboration , fermilab - conf-96 - 232-e ( 1996 ) , talk presented at the dpf-96 conference , minneapolis , mn , 1015 august 1996 .
m. carena , p.m. zerwas _ et al .
_ , in _ physics at lep2 _ , volume 1 , edited by g. altarelli , t. sjstrand and f. zwirner , cern yellow report 96 - 01 ( 1996 ) pp . 351462 .
d. amidei and r. brock , editors , _ future electroweak physics at the fermilab tevatron _
, report of the tev-2000 study group , fermilab - pub-96/082 ( 1996 ) .
s. kim , s. kuhlmann and w .- m .
yao , `` improvement of signal significance in @xmath53 search at tev-33 '' , cdf - anal - exotic - public-3904 ( 1996 ) , contribution to these proceedings .
yao , `` prospects for observing the higgs boson in the @xmath54 channel at tev-33 '' , fermilab - conf-96 - 383-e ( 1996 ) , contribution to these proceedings .
w.w . armstrong _
[ atlas collaboration ] , _ atlas technical proposal _ , cern - lhcc-94 - 43 ( 1994 ) .
[ cms collaboration ] , _ cms technical proposal _ , cern - lhcc-94 - 38 ( 1994 ) .
glover , j. ohnemus and s.s.d willenbrock , _ phys .
rev . _ * d37 * , 3193 ( 1988 ) ; v. barger , g. bhattacharya , t. han and b.a .
kniehl , _ phys .
* d43 * , 779 ( 1991 ) ; m. dittmar and h. dreiner , _ phys .
rev . _ * d55 * , 167 ( 1997 ) .
haber , in _ physics and experiments with linear colliders _ , volume i , proceedings of the linear collider workshop , saariselk , finland , 114
september 1991 , edited by r. orava , p. eerola , and m. nordberg ( world scientific , singapore , 1992 ) pp .
s. kuhlman _
_ , _ physics and technology of the next linear collider _ , a report submitted to snowmass 96 , slac - report-485 ( 1996 ) .
a. djouadi et al . , `` higgs physics '' [ hep - ph/9605437 ] , prepared by the european higgs working group on physics with @xmath27 linear colliders . , _ phys .
* 75 * , 1462 ( 1995 ) ; ucd-96 - 6 ( 1996 ) [ hep - ph/9602415 ] , to appear in _ phys .
_ r. palmer _
et al . _ [ the @xmath32 collider collaboration ] , _ @xmath32
collider : a feasibility study _ , bnl-52503 ( 1996 ) .
d.j . miller , in proceedings of the second workshop on the physics potential and development of @xmath32 colliders , sausalito , ca , 1994 , edited by d. cline ( american institute of physics conference proceedings 352 ) p. 191 .
, _ phys . rev . _ * d43 * , 904 ( 1991 ) . , _ phys .
* b328 * , 67 ( 1994 ) . , ucd-96 - 26 ( 1996 ) [ hep - ph/9610337 ] , contribution to these proceedings .
rev . _ * d48 * , 5109 ( 1993 ) ; d.l .
borden , d.a .
bauer and d.o .
caldwell , _ phys .
rev . _ * d48 * , 4018 ( 1993 ) .
r. van kooten and d. jackson , work reported in ref .
@xcite ( see fig . 3 ) .
, private communication and atlas technical note in preparation .
p. janot , private communication ( analysis based on the work of ref .
a. djouadi , h.e .
haber and p.m. zerwas , _ phys .
_ * b375 * , 203 ( 1996 ) ; f. boudjema and e. chopin , _ z. phys .
_ * c73 * , 85 ( 1996 ) .
hewett , t. takeuchi and s. thomas , in _ electroweak symmetry breaking and new physics at the tev scale _ , edited by t.l .
barklow , s. dawson , h.e .
haber , and j.l .
siegrist ( world scientific , singapore , 1996 ) , pp . 548 - 649 .
hewett , _ phys .
lett . _ * 70 * , 1045 ( 1993 ) ; v. barger , m. berger and r.j.n .
phillips , _ phys .
lett . _ * 70 * , 1368 ( 1993 ) ; j.l .
hewett , in _
b physics : physics beyond the standard model at the b factory _ , proceedings of the international workshop on @xmath55 physics , nagoya , japan , 2628 october 1994 , edited by a.i . sanda and s. suzuki ( world scientific , singapore , 1995 ) pp .
|
we examine the prospects for discovering and elucidating the weakly - coupled higgs sector at future collider experiments .
the higgs search consists of three phases : ( i ) discovery of a higgs candidate , ( ii ) verification of the higgs interpretation of the signal , and ( iii ) precision measurements of higgs sector properties .
the discovery of one higgs boson with standard model properties is not sufficient to expose the underlying structure of the electroweak symmetry breaking dynamics .
it is critical to search for evidence for a non - minimal higgs sector and/or new physics associated with electroweak symmetry breaking dynamics .
an improvement in precision electroweak data at future colliders can play a useful role in confirming the theoretical interpretation of the higgs search results .
|
the heavy - ion collisions around fermi energies have been a subject of intensive theoretical and experimental investigations in recent times @xcite .
this is primarily due to the onset of multifragmentation @xcite , collective flow and its disappearance @xcite as well as a mixture of fusion , decay , and fission @xcite dominating the physics at these incident energies . among all these phenomena ,
collective flow has been found to be sensitive towards the nuclear matter equation of state ( eos ) as well as towards nucleon - nucleon ( nn ) cross section @xcite . at low incident energies , dominance of attractive mean field
prompts the emission of particles into backward hemi - sphere whereas particle emission at higher incident energies is dominated by the forward scattering . in other words ,
the net transverse flow disappears at a certain incident energy termed as the energy of vanishing flow ( evf ) @xcite .
this energy of vanishing flow has been of rich importance because of the fact that it is independent of the particles under study @xcite and is also insensitive towards the apparatus corrections @xcite .
this energy of vanishing flow has also been robusted for the search of equation of state @xcite . up to now
, one has measured the energy of vanishing flow in the reactions of @xmath0c+@xmath0c @xcite , @xmath2ne+@xmath3al @xcite , @xmath4ar+@xmath3al @xcite , @xmath5ar+@xmath3al @xcite , @xmath5ar+@xmath6sc @xcite , @xmath5ar+@xmath7v @xcite , @xmath8zn+@xmath3al @xcite , @xmath5ar+@xmath9ni @xcite , @xmath8zn+@xmath10ti @xcite , @xmath9ni+@xmath9ni @xcite , @xmath9fe+@xmath9fe @xcite , @xmath8zn+@xmath9ni @xcite , @xmath11kr+@xmath12nb @xcite , @xmath12nb+@xmath12nb @xcite , @xmath13xe+@xmath14sn @xcite , @xmath15la+@xmath15la @xcite , and @xmath16au+@xmath16au @xcite .
a power law mass dependence of the energy of vanishing flow has also been reported in theoretical @xcite as well as in experimental studies @xcite .
looking at these mass dependence analyses , one observes that the power law behaviour reported for the mass dependence is not absolute , instead , it explains an average general trend of the mass dependence @xcite .
for example , the maximum deviation of the experimental energy of vanishing flow from the power law is @xmath17 @xcite whereas it varies between 8 - 19% in theoretical analysis @xcite .
these cases are not isolated ones , rather , they occur very frequently .
these huge deviations from the mass power law can be due to the following reasons .
\(i ) the choice of impact parameter is not uniform in the above listed reactions .
it varies from @xmath18 to 0.4 ( @xmath19 ; @xmath20 is the radius of either target or projectile ) .
it has been reported by many authors that the energy of vanishing flow depends significantly upon the impact parameter @xcite .
for example , the measured energy of vanishing flow varies from @xmath21 mev / nucleon to @xmath22 mev / nucleon for the reaction of @xmath5ar+@xmath6sc between @xmath23 and 0.48 @xcite . whereas theoretically , it varies from 65 mev / nucleon to 90 mev / nucleon .
therefore , it is utmost important to have a concrete and accurate information about the impact parameter .
\(ii ) the asymmetric reacting partners do not follow the power law behaviour of symmetric nuclei at the first place .
the asymmetry of the reaction plays a dramatic role in heavy - ion observables .
therefore , one can not expect to have an absolute power law behaviour for a mixture of symmetric and asymmetric nuclei as reported in refs .
@xcite . in the above mass dependent studies , the asymmetry of colliding nuclei was as high as 0.4 ( = @xmath24 ) . for a better understanding , one needs to have a complete systematic study of the mass dependence of energy of vanishing flow where impact parameter as well as asymmetry of the reacting nuclei is kept under control so that the above mentioned ambiguities can be removed .
our present aim , therefore , is at least twofold .
\(a ) to present a systematic study of the mass dependence of energy of vanishing flow by analyzing symmetric reacting partners at fixed reduced impact parameters .
this will remove all the above mentioned discrepancies .
\(b ) as reported in the literature , the studies of energy of vanishing flow give confused picture about the equation of state , momentum dependent interactions ( mdi ) as well as about the in - medium nn cross section .
some studies suggest hard equation of state producing more flow than soft @xcite whereas others have contradictory experiences @xcite .
some studies deny the role of momentum dependent interactions @xcite whereas others point toward their dominant role @xcite .
the above systematic study shall also be extended to shed light on the equation of state , momentum dependent interactions as well as on the in - medium nn cross section throughout the periodic table , for the first time .
this study will also help to clear many unresolved questions about the nuclear matter equation of state as well as in - medium nn cross section .
the above study is carried out within the framework of quantum molecular dynamics ( qmd ) model @xcite .
in the qmd model @xcite , each nucleon propagates under the influence of mutual two and three - body interactions .
the propagation is governed by the classical equations of motion : @xmath25 where h stands for the hamiltonian which is given by : @xmath26 here @xmath27 , @xmath28 , @xmath29 , and @xmath30 are , respectively , the skyrme , yukawa , coulomb , and momentum dependent potentials .
the momentum dependent interactions are obtained by parameterizing the momentum dependence of the real part of the optical potential .
the final form of the potential reads as @xcite @xmath31\delta({\bf r_{1}}-{\bf r_{2}}).\ ] ] here @xmath32 = 1.57 mev and @xmath33 = @xmath34 . a parameterized form of the local plus momentum dependent interaction potential is given by @xmath35\rho/\rho_{0}.\ ] ]
the parameters @xmath36 , @xmath37 , @xmath38 , @xmath39 , and @xmath40 are listed in ref .
the present systematic study is carried out by simulating the symmetric reactions of @xmath0c+@xmath0c , @xmath2ne+@xmath2ne , @xmath5ca+@xmath5ca , @xmath9ni+@xmath9ni , @xmath12nb+@xmath12nb , @xmath41xe+@xmath41xe , @xmath16au+@xmath16au , and @xmath1u+@xmath1u at their corresponding reduced impact parameters @xmath42 and 0.2 using the hard , soft , and hard equation of state with momentum dependent interactions .
we shall also use constant nn cross sections of 40 and 55 mb strength for the present analysis .
the static hard and soft equations of state are labeled as hard and soft whereas momentum dependent inclusion is written as hmd and smd , respectively .
the nn cross sections are written as superscripts to these abbreviations .
the above mentioned reactions were simulated at incident energies between 30 mev / nucleon and 250 mev / nucleon in small steps and energy of vanishing flow was calculated using straight line interpolations . in table 1
, we list the incident energies and corresponding transverse flow over which straight line interpolation was made to find out the energy of vanishing flow .
[ fig1 ] , we display the energy of vanishing flow as a function of combined mass of the system ( between 24 and 476 ) for different equations of state and nn cross sections along with available experimental data . -1 cm whereas part _ f _ is for @xmath43 .
open symbols represent our theoretical calculations whereas solid stars represent the experimental data at @xmath42 .
lines are power law fit @xmath44 .
the solid line in part _ a _ is a power law fit excluding @xmath0c+@xmath0c reaction.,title="fig:",width=491 ] first of all , baring few cases , all the calculated energies of vanishing flow fall on the same line that is a fit of the power law nature @xmath44 .
this is very encouraging since as has been reported in refs .
@xcite , these power law dependencies were not absolute and actual energy of vanishing flow was as far as by 23 mev / nucleon from the mass power law fit . here
we notice that the power law curve passes through all the calculated points justifying our above discussion that the deviation from the power law behaviour was due to different impact parameters and asymmetric colliding nuclei . while comparing figs .
[ fig1]c and [ fig1]f , one also sees the role of impact parameter in energy of vanishing flow .
we see a significant variation in the value of energy of vanishing flow while going from @xmath42 to 0.2 .
this is particularly true for lighter colliding nuclei .
for example , one sees that for the reaction of @xmath0c+@xmath0c , the energy of vanishing flow at @xmath42 is 143 mev / nucleon which reduces to 76 mev / nucleon for @xmath43 .
this huge variation in the energy of vanishing flow due to impact parameter variation was not glorified in earlier publications of mass dependence @xcite .
this huge variation also explains why mass dependence of energy of vanishing flow is not explained by power laws reported in earlier publications .
.the incident energy ( mev / nucleon ) as well as collective transverse flow ( mev/_c _ ) for all the reactions reported above using @xmath45 at @xmath42 and 0.2 and @xmath46 at @xmath42 .
[ cols="^,^,^,^,^,^,^ " , ] this also points toward the need of accurate impact parameter analysis in experimental measurements where one often either gives the range or the upper value of the estimated impact parameters .
the insufficient and inaccurate estimates of impact parameters in experiments can play a havoc for pinning down the nuclear matter equation of state from the study of energy of vanishing flow .
for instance , hard@xmath47 and soft@xmath47 equations of state at @xmath42 , for @xmath0c+@xmath0c reaction give energy of vanishing flow equal to 143 and 158 mev / nucleon whereas hard@xmath47 at @xmath43 has an energy of vanishing flow equal to 76 mev / nucleon .
a small error in the estimation of impact parameter can lead to completely different interpretation of the nature of equation of state and other physical conclusions .
+ from the figure , it is also evident that the average difference between hard and hmd equations of state ( for @xmath48 and 55 mb ) is 15% whereas for hard and soft equations of state , it is 26% . on the other hand , it is 42% for hard equation of state at @xmath42 and 0.2 .
the role of momentum dependent interactions is stronger for lighter colliding nuclei compared to heavier nuclei due to larger energy of vanishing flow in lighter cases in agreement with ref . @xcite .
further , larger cross section leads to smaller energy of vanishing flow which is in agreement with refs .
the average role of momentum dependent interactions ( in deciding the energy of vanishing flow ) is smaller due to low energy of vanishing flow .
this is also in agreement with ref .
@xcite . this may , however be , significant for peripheral reactions where energy of vanishing flow with static equation of state is much higher than the experimental values @xcite .
one also sees that both the hard@xmath47 and hmd@xmath47 equations of state are able to explain the energy of vanishing flow ( including @xmath0c+@xmath0c ) throughout the periodic table at @xmath42 whereas this is not possible with smaller cross section .
it is worth mentioning that the @xmath0c+@xmath0c reaction is very rarely discussed in the literature .
one also sees that the power factors @xmath49 in all the cases are close to each other expect for hard@xmath5 . in all the other cases , these are close to -1/3 defending the mutual dominance of attractive nuclear forces @xmath50 and repulsive nuclear forces @xmath51 unity , therefore , leading to net -1/3 dependence . we have also tested the role of asymmetry of colliding nuclei on energy of vanishing flow which is also found to be significant
. -1 cm , title="fig:",width=377 ] in fig .
[ fig2 ] , we display the division of collective in - plane flow at energy of vanishing flow into parts resulting from the repulsive nn collisions and attractive mean field .
we see that the mean field is always attractive whereas it is counterbalanced by the repulsive collision part leading to net zero collective transverse in - plane flow at the energy of vanishing flow .
this figure also depicts the mass independent nature of the contribution of mean field towards collective transverse in - plane flow at the energy of vanishing flow .
this is in agreement with earlier calculation @xcite .
summarizing , we presented the systematic study of the energy of vanishing flow by considering symmetric colliding nuclei at normalized impact parameters using variety of equations of state ( with and without momentum dependent interactions ) as well as different cross sections .
we obtained a perfect power law mass dependence in all the cases which passes through all the calculated points nicely throughout the periodic table .
further , the choice of impact parameter affects the energy of vanishing flow drastically , demanding a very accurate measurement of impact parameter .
however , the energy of vanishing flow is found to be less sensitive towards the equation of state as well as its momentum dependence .
999 d. sisan _ et al .
c * 63 , * 027602 ( 2001 ) ; h. y. zhang , w. q. shen , y. g. ma , x. z. cai , l. p. yu , c. zhong , y. b. wei , and j. g. chen , j. phys .
* 28 , * 2397 ( 2002 ) ; c. b. das , l. shi , and s. das gupta , phys . rev .
c * 70 , * 064610 ( 2004 ) .
|
we present a systematic study of the energy of vanishing flow by considering symmetric colliding nuclei ( between @xmath0c and @xmath1u ) at normalized impact parameters using variety of equations of state ( with and without momentum dependent interactions ) as well as different nucleon - nucleon cross sections .
a perfect power law mass dependence is obtained in all the cases which passes through calculated points nicely .
further , the choice of impact parameter affects the energy of vanishing flow drastically , demanding a very accurate measurement of the impact parameter .
however , the energy of vanishing flow is less sensitive towards the equation of state as well as its momentum dependence .
pacs number : 25.70.-z , 25.70.jj + electronic address : [email protected]
|
in this paper we continue the lineage of work on quantum brownian motion via the influence functional path - integral method of feynman and vernon@xcite used by caldeira and leggett@xcite to derive a master equation for a high - temperature ohmic environment , which corresponds to the markovian regime . following this , caldeira , cerdeira and ramaswamy ( ccr)@xcite derived the markovian master equation for the system with weak coupling to an ohmic bath , that was claimed to be valid at arbitrary temperature ( sec .
[ sec - caldeira ] ) . at the same time unruh and
zurek@xcite derived a more complete and general master equation that incorporated a colored noise at finite temperature .
finally hu , paz and zhang ( hpz ) @xcite derived an exact master equation for a general environment ( arbitrary temperature and spectral density ) which preserves the positive definiteness of the reduced density matrix , an important property missed out in many earlier derivations . for many calculations ,
physicists often invoke markovian master equations as they are easier to solve owing to their having compact , functional representations .
but one runs into trouble if one is interested in low temperatures , short times or non - ohmic baths , which likely fall under the non - markovian regime ( see ref .
@xcite for a discussion ) .
the hpz equation is capable of dealing with the full range of parameters for a general environment but its form is somewhat involved .
for example the coefficients are represented by solutions to integro - differential equations and multiple integrals over green s functions . in this paper
we present the correct ( _ cf .
_ for instance ccr ) late - time non - markovian master equation in compact functional representation for arbitrary temperature and coupling strength ( sec .
[ sec : latediff ] ) .
additionally we obtain the coefficients for a large class of spectral densities and their time dependent form at times before the oscillator and reservoir have equilibrated .
solutions to the hpz equation have been attempted before @xcite and the role of the master equation coefficients ( renormalized frequency , dissipation , diffusion , and anomalous diffusion coefficients ) has been discussed in the past .
here we present complete solutions of the master equation .
we consider a quantum oscillator under the influence of an external classical force that is linearly coupled to a thermal reservoir of quantum oscillators .
the particle begins decoupled from the reservoir and uncorrelated with it , with a short switch - on time @xmath0 for the coupling .
the master equation coefficients are derived for spectral functions which are laurent series in the frequency with ultraviolet ( uv ) cut - off @xmath1 much larger than the system frequencies , and infrared ( ir ) cut - off @xmath2 much smaller than the system frequencies .
these coefficients are calculated in detail for arbitrary temperature and arbitrary time after the switch - on time .
the master equation is solved explicitly for this class of spectra , and the general solution for arbitrary coefficients is reduced to solving the classical equation of motion of a parametric oscillator .
it is clearly shown how each coefficient enters into the solution and how the state evolves in time .
the anomalous diffusion coefficient is actually an `` anti - diffusion '' term that keeps the position uncertainty finite even when the uv cut - off tends to infinity ( in contrast with the momentum uncertainty ) .
having the solution gives us a plethora of information such as the late time thermal covariance [ eq . ] and uncertainty function ( see sec .
[ sec : uncertainty ] ) .
these results generalize the work of anastopoulos and halliwell @xcite , who already found the late time state to be a gaussian and placed lower bounds on the uncertainty function which we discuss in sec .
[ sec : uncertainty ] .
this also crystallizes the work of hu and zhang @xcite on the generalized uncertainty function for gaussian states .
finally , we derive the master equation that includes the influence of a classical external force and solve the master equation in that case as well .
the paper is organized as follows .
we begin our derivation of the master equation in sec .
[ sec : master - ohmic ] with a quantum oscillator linearly coupled to a thermal reservoir of quantum oscillators . the spectral density here is assumed to be ohmic with high - frequency cut - off @xmath1 .
the particle begins both _ decoupled _ from and _ uncorrelated _ with the reservoir .
the coupling is turned off at the initial time @xmath3 and is switched on within a short time scale @xmath0 . at moderate times , @xmath4 ( but not necessarily @xmath5 ) , we have obtained the master equation coefficients as an expansion from zero temperature ( sec . [ sec : master - low ] ) , as an expansion from high temperature ( sec .
[ sec : master - high ] ) , and in closed form as an approximation that is exact at both zero and extreme temperatures ( sec .
[ sec : master - approx ] ) .
exact closed form solutions are possible in terms of various special functions , but they do not straightforwardly reveal the behavior seen in our expansions and approximations .
most importantly , we have closed form solutions for the master equation coefficients at late times , @xmath6 , and arbitrary temperature ( sec . [
sec : master - late ] ) . in sec .
[ sec : master - laurent ] we extend these results first to analytic spectral functions ( sec .
[ sec : master - power ] ) and then to laurent - series spectra ( sec .
[ sec : master - laurent2 ] ) . with the inclusion of subohmic terms
we must introduce a low frequency cut - off @xmath2 .
we take both the uv and ir cut - offs to be very large and very small respectively and we only consider spectra that give contributions to the master equation coefficients no more divergent than in the ohmic case ( which has logarithmic dependence on the uv cut - off ) .
this limits our study to equations of motion for the system trajectories with the same form as in the ohmic case and whose contributions to the master equation coefficients are not markedly different either .
spectral densities of arbitrary powers have been studied numerically in the past , but we do not consider them as they correspond to classical paths that involve fractional calculus ( _ i.e. _ , integro - differential equations ) and would be difficult to solve analytically . in sec .
[ sec : wigner ] we solve the master equation for this class of spectra .
it is seen that the initial solution undergoes damped oscillations while evolving into a gaussian state of thermal equilibrium .
all the cumulants of the wigner distribution are easily determined as functions of time ( sec .
[ sec : wigner - sol ] ) .
in particular we can provide the exact late - time uncertainty function ( sec .
[ sec : uncertainty ] ) . in sec .
[ sec : wigner - gen ] we extend this solution to more general spectra , provided that one knows the form of the master equation coefficients .
the solution to this master equation is reduced to solving for the motion of a classical parametric oscillator .
it is seen that simple damped oscillations may now be parametric damped oscillations , otherwise the form of the solution is the same . finally in sec .
[ sec : force ] we extend the method of calzetta , roura , and verdaguer @xcite to include the influence of a classical force acting upon the oscillator .
the master equation turns out to be what one would naively guess .
we solve this master equation in sec .
[ sec : force - sol ] and it is seen that the external force drives the mean around while the evolution of the variance remains unchanged and determined by the reservoir . in the last section we conclude with a list of our findings and suggestions for their applications .
the lagrangian of a system consisting of a quantum brownian oscillator with mass @xmath7 , natural frequency @xmath8 and coordinate @xmath9 coupled with coupling constants @xmath10 to an environment at temperature @xmath11 made up of @xmath12 oscillators with mass @xmath13 , natural frequency @xmath14 and coordinates @xmath15 is given by @xmath16 where @xmath17 is a switch - on function with a very short characteristic time - scale @xmath0 . in this section
we work with an ohmic bath so that the spectral density function is given by @xmath18 where @xmath1 is a high - frequency cut - off of the bath and @xmath19 is the dissipation frequency .
we will be using this hard cut - off for all our calculations .
gaussian and exponential decay cut - off functions , of the form @xmath20 and @xmath21 respectively , are also common choices .
the system and environment begin decoupled at @xmath3 , but are fully engaged in a time @xmath22 which is assumed to be very short but non - zero . this time - dependent coupling
is easily dealt with following the work of hu and matacz@xcite on the time - dependent problems where all parameters of the system and bath oscillators and their couplings are allowed to be time - dependent .
when only the system - environment couplings are time - dependent , as in our case , the dissipation and noise kernels are given respectively by @xmath23 } \coth{\left ( \frac { \omega}{2 t } \right ) } i(\omega ) \mbox{~ } \theta_\tau(t ) \theta_\tau(s ) , \end{aligned}\ ] ] where @xmath24 and it tends to a dirac delta distribution in the limit @xmath25 . had we chosen a gaussian high - frequency cut - off function instead of a hard cut - off , we would have found the dissipation kernel to be the derivative of a gaussian distribution that corresponds to an alternative representation of the delta distribution in the infinite cut - off limit . on the other hand ,
the integral for the noise kernel is more complicated , even without the cut - off .
although it can also be expressed in terms of distributions , for the purposes of this paper it is best left to the last . in hpz@xcite
the system - environment coupling was taken to be constant in time , yet the system and environment were initially uncorrelated .
this gives rise to initial `` jolts '' in the normal diffusion coefficient of the master equation with a characteristic time - scale of order @xmath26 and an amplitude proportional to @xmath1 , which diverges in the limit @xmath27 .
it also gives rise to a delta - like term at the initial time in the equation of motion : @xmath28 in the limit of infinite cut - off this sudden frequency change gives an initial kick to the homogeneous solutions of the equation of motion , which can be seen as follows : @xmath29 thus , the classical paths experience a finite velocity change within an infinitesimal time . following the approach in ref .
@xcite one can easily see that this kick translates into a distortion of the wigner distribution from the bare initial state to a shifted one @xmath30 one expects that this result , which corresponds to replacing @xmath31 with a dirac delta in eq .
( [ eq : eom1 ] ) , is a good approximation whenever all the relevant time - scales ( @xmath32 , @xmath19 and @xmath33 ) are much larger than @xmath34 . the physical origin of the jolts in the coefficients of the master equation as well as other initial time divergences ( such as the divergent contributions to correlation functions of system observables that are due to divergent boundary terms at the initial time ( see appendix d in ref .
@xcite ) can be understood as follows . in general
when a system couples to an environment with an infinite number of modes , well - behaved states exhibit correlations with arbitrarily high - frequency modes .
in contrast , states that are uncorrelated for sufficiently high frequencies ( and hence completely factorizable states in particular ) are pathological .
for instance , they have infinite energy and their hilbert space is even unitarily inequivalent to the space of physical states , spanned by the basis of energy eigenvectors of the whole system hamiltonian including the system - environment interaction .
( of course for a finite uv cut - off there are no divergences , but the potentially divergent terms are very sensitive to changes in the value of the cut - off . )
physically acceptable initial states that correspond to the thermal equilibrium state for the whole system can be obtained using euclidean path integrals @xcite .
however , the instantaneous preparation functions employed in ref .
@xcite to produce other states in addition to the thermal equilibrium state still give rise to initial divergences , as explained in ref . @xcite . in order to obtain finite results
, one needs to prepare the new initial state within a non - vanishing time @xcite , which corresponds to a physically more realistic situation .
the alternative approach that we follow here is to switch on the interaction smoothly within a time @xmath0 much longer than @xmath26 but shorter than any other relevant time - scale of the problem . in this way
the factorized initial state , which is perfectly acceptable in the uncoupled case , becomes adequately correlated with the arbitrarily high - frequency modes in a regular fashion . when adding the short time switch - on function to the spectral density to turn on the interaction gradually , the initial jolt is no longer present in the results for the master equation coefficients .
moreover , the equation of motion exhibits a smooth transition between the decoupled and the coupled system without the term proportional to @xmath31 on the right - hand side : @xmath35 however , since the constant bare frequency @xmath36 is of order @xmath1 so as to cancel out the divergent contribution that arises when integrating out the environment and give a finite value of the renormalized frequency at late times ( larger than the switch - on time ) , the frequency will change significantly in a short period of time of order @xmath0 .
as long as @xmath0 is much shorter than all the other relevant time - scales in the problem ( except for @xmath26 ) , the dynamics for @xmath4 can be obtained by approximating the time - dependent frequency by a delta function . more specifically , since the renormalized frequency changes within a characteristic time - scale of order @xmath0 and with an amplitude of order @xmath1 , it can be written as @xmath37 where @xmath38 is the asymptotic constant value for times much larger than @xmath0 , @xmath39 is a constant of order one , and @xmath40 is a function peaked at @xmath3 with amplitude of order @xmath41 and width of order @xmath0 that becomes a dirac delta in the limit @xmath42 .
therefore , for @xmath4 the dynamics is governed by @xmath43 note that @xmath44 if the switch - on time is sufficiently large to cure the initial divergences discussed above . the term on the right - hand side of eq .
( [ eq : eom3 ] ) has the same form as in eq .
( [ eq : eom1 ] ) .
it will , therefore , give the same kind of initial kick to the solutions of the equation and generate the same kind of transformation of the reduced wigner function : @xmath45 this phase - space transformation has a jacobian with determinant equal to unity : @xmath46 therefore , it is simple to calculate renormalized expectation values in terms of bare expectation values and vice versa : @xmath47 we can immediately see that the normalization , linear entropy ( see sec .
[ sec : entropy ] ) and state overlap are all unchanged by the kick .
we can also check that the heisenberg uncertainty relation is also preserved as follows .
first , we start with the covariance matrix for @xmath9 and @xmath48 corresponding to the wigner distribution @xmath49 with @xmath50 , @xmath51 and @xmath52 , and which transforms in the following way under linear phase space transformations : @xmath53 hence , from eq .
( [ eq : jacobian ] ) we have @xmath54 finally , one takes into account that @xmath55 corresponds to the formulation in terms of the wigner function of the generalized heisenberg uncertainty relation due to schrdinger @xcite : @xmath56 where @xmath57 .
in addition to being real and normalized , a phase space distribution corresponds to the wigner function of a physical state only when the corresponding density matrix ( undoing the wigner transformation ) satisfies the condition @xmath58 . for gaussian distributions this is guaranteed if and only if eq .
( [ eq : uncertainty1 ] ) is satisfied @xcite .
thus , the kick does not change whether a given gaussian distribution corresponds to a physical state . on the other hand , for non - gaussian distributions
( [ eq : uncertainty1 ] ) is no longer a sufficient condition for that , but one can use the following argument .
for a finite switch - on time , the frequency change simply corresponds to unitary evolution associated with a time - dependent hamiltonian , which leaves the eigenvalues of the density matrix non - negative . by continuity ,
the limit of a very short switch - on time can not make these eigenvalues negative .
if one is interested in studying the evolution of a certain state of the system properly correlated with the environment ( such as the global equilibrium states considered in ref . @xcite or states prepared from those in a finite time ) one can always consider the reduced wigner function associated with that state and invert eq .
( [ eq : uncertainty1 ] ) to obtain the corresponding initial wigner function before the interaction was switched on .
( note that for times larger than the switch - on time the result is then essentially equivalent to having introduced a time - dependent counterterm for the frequency so that the renormalized frequency was constant in time . )
the hpz master equations for the reduced density matrix @xmath59 and the reduced wigner function are given respectively by @xmath60 + \gamma \left[\hat{x},\{\hat{p},\hat{\rho}_r\}\right ] + \imath d_{xp } \left ( \left[\hat{x},\left[\hat{p},\hat{\rho}_r\right]\right ] + \left[\hat{p},\left[\hat{x},\hat{\rho}_r\right]\right ] \right ) - \imath d_{pp } \left[\hat{x},\left[\hat{x},\hat{\rho}_r\right]\right ] , \\
\frac{\partial}{\partial t }
w_r & = & \ { h_r , w_r \ } + 2\gamma \frac{\partial } { \partial p } p w_r -2 d_{xp } \frac{\partial^2}{\partial x \partial p } w_r + d_{pp } \frac { \partial^2}{\partial p^2 } w_r \label{eq : wigner},\end{aligned}\ ] ] where the detailed form of the dissipation function @xmath61 and the diffusion functions @xmath62 and @xmath63 can be found in @xcite . following the derivation of calzetta , roura , and verdaguer @xcite
the coefficients in the hpz master equation for an ohmic bath with a high cut - off @xmath64 and evaluated at times larger than the switch - on time are given by : @xmath65 where @xmath66 and @xmath67 is the retarded green function associated with the differential equation that results from multiplying eq .
( [ eq : eom2 ] ) by @xmath7 . according to the discussion in the previous subsection , when considering times much larger than the switch - on time , its effect on the equation of motion can be approximated by a delta function , as seen in eq .
( [ eq : eom3 ] ) . since @xmath67 for @xmath68 , the contribution from the delta term on the right - hand side of eq .
( [ eq : eom3 ] ) vanishes when solving for the retarded propagator .
therefore , it is indeed independent of the initial kick and given by eq .
( [ eq : eom3 ] ) .
there is still some dependence on the switch - on function @xmath69 in the noise kernel for @xmath70 , but for @xmath4 the contribution from @xmath71 to the time integrals in eqs .
( [ eq : dxp1])-([eq : dpp1 ] ) is negligible provided that the behavior during the switch - on time is sufficiently regular , which is indeed the case .
similarly , the behavior of the master equation coefficients during switch - on time , which we do not calculate here , will be sufficiently regular so that its contribution to the evolution of the reduced wigner function ( or density matrix ) is also negligible when considering times larger than @xmath0 and one only needs to include the effect of the kick on the initial wigner function , given by eq .
( [ eq : phase_trans2 ] ) .
some clarification is in order here about the use of a finite frequency cut - off @xmath1 for the environment spectrum . having a finite cut - off in the frequency integral for the noise kernel , given by eq .
( [ eq : noise ] ) , is necessary to obtain a finite result for the coefficients of the master equation because , as it will be seen below , it gives rise to contributions to the diffusion coefficients that become logarithmically divergent in the limit of large @xmath1 .
in contrast , the use of a finite cut - off in the frequency integral of the dissipation kernel is not required to obtain a finite result .
in fact , for computational convenience we are going to take @xmath72 in this case because then the dissipation kernel becomes local and we just have to deal with an ordinary differential equation rather than an integro - differential one .
one may object that treating the noise and dissipation kernels on a different footing could lead to inconsistencies ( _ e.g. _ the fluctuation - dissipation relation will not hold exactly any more , as pointed out by hpz @xcite with regard to the treatment of this problem in ref .
however , our results are consistent when properly understood .
if one considered an expansion of the exact master equation coefficients in terms @xmath1 , one would have leading terms of order @xmath73 , terms of order one and subdominant terms involving inverse powers of @xmath1 .
our approach gives the right result for the @xmath73 and order - one terms : using the local approximation for the dissipation kernel only alters the result for the subdominant terms .
this can be seen as follows . when integrating @xmath74 , given by eq .
( [ eq : delta1 ] ) , with a sufficiently well - behaved function @xmath75 , one has @xmath76 this condition will indeed be satisfied in our case since all the other relevant time - scales in the problem are much larger than @xmath26 and all the functions that @xmath77 will be integrated with are regular and extremely uniform over that scale .
therefore , the discrepancy between using the dirac delta function and using @xmath78 corresponds to terms of order @xmath26 or higher inverse powers , which give contributions to the master equation coefficients of order @xmath79 or higher . in order to compute the coefficients of the master equation , we first perform the relatively simple time integrals in eqs .
( [ eq : dxp1])-([eq : dpp1 ] ) . on the other hand ,
the frequency integral in the noise kernel , as given by eq .
( [ eq : noise ] ) , is particularly difficult due to the @xmath80 in @xmath81 .
it is this integral that we save for last . others@xcite
have also seen the utility or at least the simplicity of reducing the master equation to a collection of one - dimensional frequency integrals .
we are able to obtain four sets of new results : ( a ) exact results at late time for arbitrary temperature , ( b ) asymptotic expansions for high and low temperatures , ( c ) approximate general results for arbitrary temperature and all times ( larger than the switch - on time ) , ( d ) time - dependent diffusion functions in closed form at high temperatures for all times ( larger than the switch - on time ) .
after computing the time integrals in eqs .
( [ eq : dxp1 ] ) and ( [ eq : dpp1 ] ) , the diffusion coefficients become @xmath82 in terms of a single family of frequency integrals @xmath83 , @xmath84 , @xmath85 : @xmath86 fortunately the only integral that needs to be computed is @xmath87 .
all other integrals can be generated from this one integral .
in general the integrals can not be solved directly , therefore it is necessary to expand the hyperbolic cotangent into a series of simpler functions .
this will be done with low and high - temperature expansions . when computing the time integrals that led to eqs .
( [ eq : dxp2])-([eq : dpp2 ] ) , we just took the switch - on functions equal to one , so that sufficiently simple analytical results could be obtained .
this means that , within our approach , the results in eqs .
( [ eq : dxp2])-([eq : dpp2 ] ) are only valid for @xmath88 .
in fact , for @xmath89 , they essentially coincide with those of ref .
@xcite for an ohmic environment .
for instance , the term @xmath90 in eq .
( [ eq : dpp2 ] ) gives rise to the same initial jolt , with a width of order @xmath26 and an amplitude proportional to @xmath1 , found in ref .
however , as we discussed above , the results should be valid for times larger than the switch - on time because the contribution from the switch - on period to both the master equation coefficients and the evolution of the density matrix at @xmath4 is negligible .
it is interesting to note that the coefficients @xmath62 and @xmath63 both exhibit logarithmic divergences in the limit @xmath25 ( for times larger than the switch - on time and thus larger than @xmath26 ) due to the term proportional to @xmath91 . ) , the term logarithmic in @xmath1 is not present in @xmath91 when considering a fixed finite cut - off @xmath1 and temperatures much higher than @xmath1 . ]
this has been pointed out for @xmath62 in ref .
@xcite , where the coefficients of the master equation were calculated perturbatively to second order in the system - environment coupling constants ( linear order in @xmath19 ) .
the fact that there is also a logarithmic divergence in @xmath63 was not seen in that reference because it is quartic in the system - environment coupling constants ( quadratic in @xmath19 ) , as it can be seen in eq .
( [ eq : dpp2 ] ) .
moreover , such kinds of perturbative calculations can not be employed to study the long time behavior since they are only valid for @xmath92 and they miss for instance the exponential decay of the second and third terms on the right - hand side of eqs .
( [ eq : dxp2])-([eq : dpp2 ] ) . for the late - time diffusion coefficients ,
only two integrals need to be performed and they can be expressed in closed form with reasonably intuitive functions : many of the expressions derived throughout this paper assume underdamping with an extreme cut - off , _
i.e. _ , @xmath93 with @xmath66
. they can be used for the overdamping regime by making the following analytical continuation : @xmath94 with @xmath95 real .
+ therefore , eqs . - can be applied to the overdamping case if the @xmath96 and @xmath97 terms are first expanded assuming @xmath98 is real , and then the analytical continuation @xmath94 is made . ]
\label{eq : fi_1 } , \\
\mbox{fi}_3 & = & \frac{\pi t}{2 \gamma_0 } + \frac{\tilde{\omega}^2-\gamma_0 ^ 2 } { 2 \gamma_0 \tilde{\omega } } \mbox{im}\left [ \mbox{h}\left ( \frac{\gamma_0 + \imath \tilde{\omega } } { 2 \pi t } \right ) \right ] \notag \\ & & + \mbox{re}\left [ \mbox{h}\left(\frac{\lambda}{2 \pi t}\right ) - \mbox{h}\left ( \frac{\gamma_0 + \imath \tilde{\omega}}{2 \pi t } \right)\right ] \label{eq : fi_3 } , \end{aligned}\ ] ] where terms involving negative powers of the cut - off have been neglected and @xmath100 is the harmonic number function defined in appendix [ sec : harmonic ] . it behaves like @xmath101 except near the origin , where it does not diverge and actually vanishes .
the @xmath102 $ ] term in @xmath103 is effectively the @xmath104 divergence .
it disappears at extreme temperature ( higher than the cut - off ) .
a low - temperature expansion of the master equation coefficients can be obtained using the following expansion for the hyperbolic cotangent in the corresponding frequency integrals : @xmath105 moreover , since the results for all the relevant integrals can be obtained from @xmath106 simply by differentiating with respect to @xmath33 , we will concentrate on computing eq . with @xmath107 . in order to do that
, one first decomposes the integrand in eq . into partial fractions and
splits the integral @xmath108 into @xmath109 . finally , using eq .
they become a linear combination of exponential integrals .
the result can be written as @xmath110 \nonumber \\ & & + \frac{\pi}{4 \gamma_0 \tilde{\omega } } \cos(\tilde{\omega } t ) e^{-\gamma_0 t } + \delta \mbox{fc}_1^\mathrm{pt}(t ) + \delta \mbox{fc}_1^\mathrm{l\lambda}(t ) \label{eq : low - approx } , \end{aligned}\ ] ] where @xmath111 is the exponential integral detailed in appendix [ sec : expint ] .
the first three terms come from the @xmath112 integral of the constant term in eq . .
on the other hand , @xmath113 , which vanishes at zero temperature , is the contribution from the @xmath112 integral of the remaining terms in eq .
: @xmath114 finally , the term @xmath115 contains all the dependence on the cut - off @xmath1 , which arises from all the @xmath116 integrals .
in particular , its value for zero temperature , which corresponds to taking only the first term in eq . , can be calculated explicitly : @xmath117 .\end{aligned}\ ] ] note that , since @xmath106 is finite in the limit @xmath118 , the contribution from @xmath119 will be very small for sufficiently large @xmath1 . however , when differentiating several times with respect to @xmath33 , this term gives rise to the divergences that @xmath120 and @xmath121 exhibit during a short initial period of order @xmath26 , which are responsible for the initial jolt of the master equation coefficients found in ref .
moreover , even in those cases it would be enough to consider just @xmath122 rather than @xmath115 because the integral in eq .
is convergent for arbitrary positive powers of the frequency due to the exponential factor .
in fact , given that the divergent behavior for short times is irrelevant when considering a switch - on time sufficiently longer than @xmath26 , it turns out that in our case we can completely neglect the contribution from the cut - off dependent terms for sufficiently large @xmath1 .
it would only be necessary when obtaining @xmath91 by evaluating @xmath120 at @xmath3 , but it is much simpler and more accurate in general to use the exact result for @xmath123 and @xmath91 computed earlier and given by eqs . - .
to sum up , taking into account the remarks in the previous paragraph together with the fact that the contribution from the third term in eq . cancels out when adding all the time - dependent terms in eqs .
- , all that one needs [ in addition to eqs . - ] to calculate the master equation coefficients is @xmath124 \notag \\ & & + 2 \sum_{k=1}^\infty \int_0^\infty \frac{\omega \cos(\omega t ) e^{-k \frac{\omega}{t } } } { \left ( \omega^2-\omega_r^2 \right)^2 + 4 \gamma_0 ^ 2 \omega^2 } d\omega \label{eq : low - approx2 } .\end{aligned}\ ] ] given the asymptotic behavior of @xmath111 for large @xmath125 , one can see that the @xmath126 $ ] terms decay only weakly in time like @xmath127 ; they both have complicated @xmath98 frequency oscillations .
the integral in the last term , which vanishes when the temperature tends to zero , can be performed explicitly , but the sum becomes particularly complicated . a high - temperature expansion can be obtained by making use of the following expression for the hyperbolic cotangent : @xmath128 which will yield a series solution that is best in the high temperature , late time regime @xmath129 .
@xmath87 can then be calculated in several steps .
first , one calculates the @xmath112 integral corresponding to eq . considering only the first term in eq .
, decomposes the resulting integrand into partial fractions and ends up with a linear combination of exponential integrals .
after simplification , the result can be written as @xmath130 next , we perform the @xmath112 integral corresponding to eq . with the general term in the sum of eq . replacing the hyperbolic cotangent .
after decomposing into partial fractions we end up again with a linear combination of exponential integrals , and the final result after simplification is @xmath131 .\end{aligned}\ ] ] the sum over @xmath132 can be performed explicitly for the terms multiplying the sine and the cosine , so that we get @xmath133 with @xmath134 which satisfies the following property : @xmath135 putting everything together , we have @xmath136 where @xmath137 contains all the dependence on the cut - off @xmath1 , and comes from all the @xmath116 integrals . nevertheless , for the same reasons given in the previous subsection , in our case one can neglect this term for sufficiently large values of @xmath1 .
furthermore , since the contribution from the first term in eq . cancels out when adding all the time - dependent terms in eqs . - , and
provided that we use eqs . - to compute @xmath123 and @xmath91 , it is sufficient to consider @xmath138 in order to calculate the master equation coefficients .
all terms on the right - hand side of eq .
decay with temperature - dependent time - scales . since it can not be expressed in closed form with any intuitive functions , the sum is not explicitly performed .
although useful for numerical evaluation , this form does not immediately reveal the @xmath139 and @xmath140 behavior .
therefore , it is convenient to approximate the sum by an integral that can be computed explicitly , as follows : @xmath141 \label{eq : high - approx2}.\end{aligned}\ ] ] from this expression one can see that the most lingering terms decay exponentially as @xmath142 .
there are also oscillations with frequency @xmath98 . by inspecting the corresponding terms in the high and low - temperature expansions
, one can get a good idea of the general solution : @xmath143 \notag\\ & & + \delta \mbox{fc}_1^{\mathrm{h}\lambda } ( t ) . \ ] ] this expression is exact at both zero temperature and high temperature , and it has only minor discrepancies within a small range of intermediate temperatures .
moreover , as explained above , all that one really needs to calculate the master equation coefficients is @xmath144 .\ ] ] from our analysis of the integrals it can be seen that there is only one fluctuation frequency @xmath98 and it is always damped .
there are only two damping rates , @xmath19 and @xmath145 .
all temperature - damped terms also have coupling damping , so one can say that @xmath146 is always late time , although the temperature can hasten this . since we have solved the integrals in the different temperature regimes , we will now plot the lowest order terms of each expansion in the different temperature regimes .
the coupling will be set at a moderate level to enhance their differences .
calculated , to first order in the , to first order in the , and with the .
all of the approximate solutions are fairly close with some minor discrepancy at short times
. the low - temperature expansion will continue to be slightly off into late times.,width=384 ] note that the high - temperature expansion ( including the first order correction ) gives very good agreement even at much lower temperatures @xmath147 ( much lower than the regime @xmath148 where the high - temperature expansion is clearly expected to be very accurate ) .
when we say @xmath149 order in the high - temperature expansion , we more acurately mean up to the @xmath150 term in the high - temperature expansion . strictly speaking this is not a temperature expansion but an expansion in @xmath151 , the only terms which decay at a temperature - dependent rate ,
so it is really a high - temperature , late - time expansion .
similarly , when we say @xmath149 order in the low - temperature expansion , we mean keeping the @xmath149 order term in the sum appearing in the last term of eq . . considering just the first term on the right - hand side of eq .
for the hyperbolic cotangent , one gets the following result for temperatures much higher than the cut - off : @xmath152 this is the common , but questionable high - temperature limit .
there is no first order term for the anomalous diffusion coefficient , but there are lower order terms that vanish only for infinite temperature .
they include the term that gives rise to the @xmath153 divergence when the temperature becomes much smaller than @xmath1 . neglecting the terms multiplied by the factor @xmath154 in eqs . - , we obtain the following result , which is valid for times much larger than the relaxation time ( _ i.e. _ , @xmath146 ) : @xmath155 \notag \\ & & + \frac{\gamma_0}{\pi } \mbox{re}\left [ \mbox{h}\left(\frac{\lambda}{2 \pi t}\right ) - \mbox{h}\left ( \frac{\gamma_0 + \imath \tilde{\omega}}{2 \pi t } \right)\right ] \label{eq : latedxp } , \\ d_{pp } & = & + 2 m \gamma_0 t + \frac{2 m \gamma_0}{\pi \tilde{\omega } } ( \tilde { \omega}^2 - \gamma_0 ^ 2 ) \mbox{im}\left [ \mbox{h } \left ( \frac{\gamma_0 + \imath \tilde { \omega}}{2 \pi t } \right)\right ] \notag \\ & & + \frac{4 m \gamma_0 ^ 2}{\pi } \mbox{re}\left [ \mbox{h}\left(\frac{\lambda}{2 \pi t } \right ) - \mbox{h}\left ( \frac{\gamma_0 + \imath \tilde{\omega}}{2 \pi t } \right)\right ] \label{eq : latedpp } .\end{aligned}\ ] ] the @xmath156 $ ] terms become @xmath157 at zero temperature and vanish in the infinite temperature limit . on the other hand ,
the value of the @xmath158 $ ] terms is @xmath159 at zero temperature , whereas it also vanishes in the infinite - temperature limit .
we now compare our results to ccr s weak coupling master equation @xcite , which differs from the hpz equation and has the following diffusion coefficients : @xmath160 the ccr master equation is frequently used for its simplicity and is believed to be accurate at late times for weak coupling and arbitrary temperature .
for , , , and .,width=384 ] one can see from fig .
[ fig : ccr1 ] that , ignoring the contribution from the cut - off , the ccr approximation matches extremely well with our exact results for the _ normal diffusion coefficient _ at weak coupling . the ccr error from neglecting the cut - off dependence
is determined by the order of magnitude of the cut - off scale since it is logarithmic in @xmath1 . in any case
, the ccr approximation underestimates the magnitude of the diffusion coefficient . for , , and .,width=384 ] unlike the normal diffusion coefficient , the difference is much more severe for the _ anomalous diffusion coefficient _ , which is _ completely absent _ in ccr . the largest contribution ( in the weak coupling regime ) to
the anomalous diffusion coefficient comes from the cut - off and it does not vanish at finite temperature ( see fig . [
fig : ccr2 ] ) .
moreover , it can not be regarded as vanishing at weak coupling because it is only proportional to one power of the coupling constant , which is the order to which ccr s master equation should be valid .
the final point that should be made is that ccr s arbitrary temperature master equation is only valid in the weak coupling regime for late times and , unfortunately , late time always means @xmath161 ( according to our exact results for the hpz master equation ) .
therefore , the weaker the coupling , the longer one must wait for ccr s master equation to be valid .
we start by considering the independent effect of power - law terms in the spectral function with ultraviolet cut - off @xmath1 and infrared cut - off @xmath2 whenever necessary .
we take these cut - offs to be extremely large and extremely small respectively , not because that should necessarily be the case for any relevant physical situation , but to restrict the model to something that we can solve explicitly .
hence , we take a spectral function of the form @xmath162 where @xmath163 is an integer number and @xmath164 is a product of characteristic frequencies chosen so that well - behaved results are obtained for the master equation coefficients in the limits @xmath118 and @xmath165 . traditionally @xmath166 has been chosen to be @xmath1 whenever a uv cut - off was necessary ; we will end up making a similar choice here .
given a general spectral function the equation of motion for the system trajectories corresponds to an integro - differential equation : @xmath167 it is thus convenient to perform a laplace transform @xmath168 under which the equation becomes purely algebraic , since the integral involving the dissipation kernel is merely a convolution .
we then have @xmath169 or , equivalently , @xmath170 where @xmath171 are the initial conditions for @xmath172 .
given the expression of the dissipation kernel @xmath173 one can easily compute its laplace transform : @xmath174 the results for different integer values of @xmath163 are listed in table [ tab : intdiss ] .
the column labeled with @xmath175 corresponds to the outcome of the integral in eq . after taking the limits @xmath176 and @xmath165 wherever that gave a finite result
the values of @xmath166 were chosen so that the integrals that appear when calculating the diffusion coefficients were finite ( more details can be found in the next two subsections ) .
finally , the column labeled with @xmath175 follows from substituting @xmath166 in @xmath175 with the values of the previous column and taking the limits @xmath72 and @xmath165 whenever they give a finite result .
had we chosen fractional powers of the frequency in the spectrum or even a different functional dependence such as a logarithm , this would have resulted in an integro - differential equation for the classical system trajectories .
in addition to being less classical , they would also be much more difficult to solve .
.[tab : intdiss]laplace transform of the dissipation kernels for various power - law spectral functions after discarding @xmath177 and @xmath178 terms .
@xmath163 is the power of the spectrum : @xmath107 is ohmic , @xmath179 is supraohmic and @xmath180 is subohmic .
@xmath175 is the laplace transform of the dissipation kernel before a reasonable choice of characteristic frequencies @xmath164 is made , whereas @xmath181 is the choosen physical dissipation kernel .
constant terms are renormalizations to the bare frequency . [ cols="^,^,^,^",options="header " , ] integer _ subohmic _ spectra , ( corresponding to @xmath182 in table [ tab : intdiss ] ) all exhibit a similar behavior , except for the @xmath183 case . for @xmath184
it is never possible to change the classical ordinary differential equation ( ode ) while keeping the diffusion coefficients finite in the limit @xmath165 .
indeed , the only way to guarantee that all the frequency integrals involving the noise kernel that appear when calculating the diffusion coefficients are finite is by taking @xmath185 ( higher powers of @xmath2 are allowed , but would give vanishing results for the diffusion coefficients ) .
in that case , there are no additional contributions to the equation of motion left . in contrast , for @xmath183 the diffusion coefficients only exhibit a logarithmic ir divergence ( proportional to @xmath186 ) when considering a finite non - vanishing value of @xmath166 independent of @xmath2 . since we tolerated logarithmic dependences on the uv cut - off @xmath1 in the ohmic case , it would be natural to allow a similar situation with the ir cut - off @xmath2
. however , in that case the laplace transform of the dissipation kernel exhibits a logarithmic term that makes the equation of motion for the system trajectories ( which corresponds to an integro - differential equation ) much more difficult to solve .
therefore , we will not consider the @xmath183 case , for the same reason why we did not consider fractional powers of the frequency in the spectral function or even a non - logarithmic dependence on the frequency . on the other hand , integer _
supraohmic _ spectra ( corresponding to @xmath187 in table [ tab : intdiss ] ) all exhibit a similar behavior .
none of them are capable of changing the nature of the classical ode while keeping the diffusion coefficients finite in the limit @xmath72 .
indeed , the only way to guarantee that all the frequency integrals involving the noise kernel that appear when calculating the diffusion coefficients are finite is by taking @xmath188 ( higher powers of @xmath1 are allowed , but would give vanishing results for the diffusion coefficients ) . in that case ,
the only contributions to the equation of motion that are left are frequency renormalization terms . in this subsection
we consider spectral functions that are analytic in the frequency ( and vanish in the limit of zero frequency ) .
they have the following form : @xmath189 where @xmath190 is an analytic function .
eq . can be written as @xmath191 which corresponds to a linear combination of terms with @xmath192 among those considered in the previous subsection .
the classical equation of motion associated with these spectra is an ode that differs trivially from the ohmic case : @xmath193 which can be written in the more compact form @xmath194 by introducing the renormalized frequency @xmath195 we can see that only the @xmath196 power ( from the ohmic - like term ) gives rise to the dissipation term , is missing , the inverse laplace transform of @xmath197 gives not only the local dissipation term @xmath198 , but also a term proportional to @xmath199 , which corresponds to the term on the right - hand side of eq .
and is responsible for the initial kick when no smooth switching - on function is present . ]
whereas all the remaining supraohmic terms merely contribute to an additional renormalization of the bare frequency . in fact , one could rig supraohmic anticoupling terms to eradicate the divergent difference between the bare and renormalized frequency , although there is no physical motivation to do so . for times longer than the switch - on time , when our explicit expressions for the diffusion coefficients can be used , the only difference between the master equation for an analytic spectrum and the ohmic spectrum lies in a modification of the contribution from the @xmath91 integrals to the diffusion coefficients .
this can be seen as follows .
when considering a term proportional to @xmath200 in the spectral function , instead of the @xmath91 integral of the ohmic case , one gets @xmath201 and the only way to get a finite non - vanishing contribution is by taking @xmath166 proportional to @xmath1 ( which was the choice already made in eqs . - ) .
therefore , one just needs to introduce the following simple substitution in the expressions for the diffusion coefficients of the ohmic case : @xmath202 where we took @xmath203 and discarded terms involving negative powers of @xmath1 .
note that all other integrals contributing to the diffusion coefficients will exhibit lower powers of the frequency in the numerator so that their result will be proportional to inverse powers of @xmath1 and can be neglected .
the integrals that correspond to @xmath120 and @xmath121 in the ohmic case also exhibit divergences that would lead to a non - vanishing result ( and even a divergent one for @xmath121 ) when dividing by @xmath204 , but that is only for short times of order @xmath26 after the initial time .
this is the regime where our explicit expressions are not accurate if the interaction is gradually switched on with a much longer characteristic time - scale , in which case the contribution to the diffusion coefficients from that period is very small .
we close this subsection by pointing out that , since @xmath91 is where the @xmath153 divergence arises for the ohmic case , one could rig an infinite number of finite supraohmic anticoupling terms to cancel the divergence , though there is no physical motivation to do so .
one could even ask whether we can both renormalize the diffusion coefficients and keep the bare frequency finite and unmodified , which amounts to requiring the following two conditions to hold simultaneously : @xmath205 the answer is obviously in the negative , at least not with finite couplings . in this subsection
we extend the form of the spectral function considered in the previous subsection to that of a laurent series .
this is done by adding an analytic function of @xmath206 to the analytic function of @xmath207 already considered there , so that we have @xmath208 where @xmath209 is an analytic function .
eq . can be written as @xmath210 which corresponds to a linear combination of terms with @xmath211 among those considered in sec .
[ sec : master_general ] . as we saw in sec .
[ sec : master_general ] , in the limit @xmath212 there is no modification of the equations of motion due to the terms in eq . with negative powers of @xmath207 .
hence , the dissipation and renormalized frequency terms , @xmath213 , in the master equation are exactly the same as with ohmic and , more generally , analytic spectra : @xmath214 , @xmath215 .
however , there will be some non - trivial contribution to the diffusion coefficients that we analyze below . for times
longer than the switch - on time the only difference between the master equation for the laurent - series spectrum and the analytic spectrum lies in a modification of the contribution from the terms most sensitive to ir divergences , namely @xmath123 and @xmath106 .
this can be seen as follows .
when considering a term proportional to @xmath216 in the spectral function , instead of the @xmath106 integral of the ohmic case , one gets @xmath217 \cos ( \omega t ) \nonumber \\ & = & \frac{2t}{\omega_r^4 } \ , \omega_c^n \left [ \frac{1}{n \lambda^n } + \mathcal{o}(\lambda^{-(n-2 ) } ) \right ] \cos ( \lambda t ) , \end{aligned}\ ] ] where we used the expansion @xmath218 in the first equality .
the only way to get a finite non - vanishing contribution is by taking @xmath166 proportional to @xmath2 ( which was the choice already made in eqs . - ) .
while @xmath123 can be obtained by evaluating @xmath106 at @xmath3 , all the remaining integrals appearing in the expressions for the diffusion coefficients will involve a less negative power of the frequency and , when dividing them by @xmath219 , will give a result corresponding to positive powers of @xmath2 , which can be neglected in the limit @xmath165 .
thus , in order to obtain the diffusion coefficients for a laurent - series spectral function of the type considered in this subsection , one just needs to introduce these simple substitutions in the expressions for the diffusion coefficients of the ohmic case : @xmath220 in summary , the supraohmic terms shift the bare frequency and @xmath91 integrals , and can be used to shift their values in the ohmic theory , whereas the subohmic terms shift the @xmath221 and @xmath123 integrals .
first , we express the master equation @xmath222 in a more compact form : @xmath223 with @xmath224 & { \mbox{\boldmath$\nabla$ } } _ \mathbf{q } & = \left [ \begin{array}{c } \frac{\partial}{\partial x } \\
\frac{\partial}{\partial p } \end { array } \right]\\ \mathbf{h } & = \left [ \begin{array}{cc } 0 & -\frac{1}{m } \\ m \omega_r^2 & 2\gamma_0 \end{array } \right ] & \mathbf{d } & = \left [ \begin{array}{cc } 0 & -d_{xp } \\ -d_{xp } & d_{pp } \end{array } \right]\end{aligned}\ ] ] this is a hyperbolic second order partial differential equation ( pde ) . as @xmath225 is a function of time , in general
the equation is not seperable in time .
it is not seperable in phase space either .
the nature of the pde suggests a fourier transform of the phase - space variables since derivatives are more complicated than algebraic parameters .
furthermore , not only does a fourier transform reduce the pde to first order , but the computation of expectation values also becomes trivial since we are then working with the characteristic function of the distribution .
the fourier transform is defined as @xmath226 and it exhibits the usual properties : @xmath227 the master equation then becomes @xmath228 where @xmath229 and the normalization of @xmath230 implies @xmath231 .
finally , it is convenient to group all the derivatives on the left - hand side , so that we have @xmath232 the method of characteristic curves involves looking for parameterized curves in the domain @xmath233 along which the first order pde becomes a set of first order odes . for each one of those curves we have @xmath234 next
, we attempt to match the right - hand side of eq . to the left - hand side of eq . .
this results in a system of odes in the parameter @xmath235 .
we will look for curves that synchronize with the initial time so that @xmath236 , @xmath237 .
the time solution is simple : @xmath238 on the other hand , the solution for the fourier transform of the phase - space variables is a bit more involved : @xmath239 in order to calculate the exponential of the @xmath240 matrix we diagonalize it , so that @xmath241 the eigenvalues , eigenvectors , and exponential matrix are given by @xmath242\label{eq : heigen } , \\
e^{-t \mathbf{h } } & = & \left ( \begin{array}{cc } 2 m \gamma_0 + m ( \partial / \partial t ) & 1 \\ -m^2 \omega_r^2 & m ( \partial / \partial t ) \end{array } \right ) g_\mathrm{ret}(t,0 ) \label{eq : exponential},\end{aligned}\ ] ] where @xmath243 is the retarded green function in eq . .
we now have the rules for transforming back and forth between the domain coordinates @xmath244 and the characteristic curve coordinates @xmath245 .
@xmath246 does not change along the characteristic curve , but for a given @xmath235 it uniquely specifies a particular curve ( except at the origin , where all curves intersect ) . using these results , we can immediately apply the method of characteristic curves to solving eq .
as follows : @xmath247 the last equation is a linear ode whose solution can easily be found to be @xmath248 where @xmath249 is the initial reduced wigner function at @xmath3 .
we can now express the solution back in terms of @xmath250 and get the final result @xmath251 with @xmath252 one can check that this result agrees with those in ref .
@xcite if one considers equal times in the two - point correlation functions there . from eqs .
and one can see that all the integrals in eq .
involve a factor @xmath253 times some oscillatory factor .
taking into account that the diffusion coefficients tend to constant asymptotic values @xmath254 and @xmath255 for sufficiently late times , one can see that for @xmath256 the integral is dominated by large values of @xmath235 .
therefore , the asymptotic value of @xmath257 can be calculated using the constant asymptotic values of the diffusion coefficients in eq . and taking the limit @xmath258 , with the following result : @xmath259 \label{eq : ltsig } .\ ] ] the solution has two factors .
the first one tends to one in the long time limit and encodes the disappearance of the initial state ( we will call it the _ death factor _ ) .
the second factor describes the appearance of a gaussian state that evolves in time and tends asymptotically to a state that corresponds to thermal equilibrium ( we will refer to this as the _ birth factor _ ) .
all initial distributions evolve towards this final gaussian state , whose covariance matrix is given by eq . .
this state does not look like the thermal state of a free harmonic oscillator because of the coupling to the environment .
it results from considering the thermal equilibrium state for the whole system ( system plus environment ) including the system - environment interaction , which gives rise to a non - trivial correlation between them , and tracing out the environment .
the _ death factor _ contains the information on the initial conditions , describes the gradual disappearance of the initial distribution and it is always temperature independent .
the initial distribution undergoes damped oscillations with characteristic time - scales @xmath260 .
the higher cumulants of the distribution , discussed in the next subsection , oscillate and decay more rapidly .
this is responsible for the inspirals in phase space of the evolution of gaussians plotted by unruh and zurek @xcite , which are calculated in the next subsection and plotted in fig .
[ fig : spiral ] . of course
this is all assuming a non - vanishing ohmic term in the spectral function .
if the spectral function is purely non - ohmic ( _ i.e. _ , with @xmath261 ) then this factor will describe the initial state oscillating with a renormalized frequency , but which never decays away .
the _ birth factor _ describes the complicated birth of and settlement into a state of thermal equilibrium .
this factor is always gaussian with a covariance matrix given by eq . , which involves a convolution of the diffusion matrix with propagators that reflect the natural oscillatory decay of the system .
this covariance matrix vanishes at the initial time and tends at late times to the equilibrium covariance matrix , with the diffusion coefficients being their asymptotic constant values given by eqs . - .
this covariance matrix is positive definite for all reasonable diffusion constants .
moreover , the anomalous diffusion coefficient actually acts as an `` anti - diffusion '' term that makes @xmath262 ( and the uncertainty in @xmath9 ) free of the @xmath263 divergence , as it will be discussed in the next subsection .
as we have already mentioned , the fourier transfom of the reduced wigner function corresponds to its characteristic function , from which the correlation functions for the phase - space variables can be easily derived using eq . .
the general expressions for the cumulants can be obtained straightforwardly from the logarithm of the reduced wigner function in fourier space as follows : @xmath264 where @xmath265 denotes the components of the vector @xmath250 and we used the einstein summation convention for pairs of repeated indices ( _ i.e. _ , it is implicitly understood that a sum @xmath266 should be preformed over each pair of repeated indices @xmath267 ) .
@xmath268 is the @xmath269 cumulant and acts as a tensor of order @xmath12 contracted with @xmath12 copies of @xmath250 . using the result for @xmath270 from eq .
we have @xmath271 where @xmath272 are the cumulants associated with the initial distribution .
. implies @xmath273 we can see that the only cumulant with a non - vanishing asymptotic value , which is a consequence of the thermal fluctuations , is the covariance matrix ( with @xmath274 ) .
the closely related second momenta of the distribution are given by @xmath275 where @xmath276 denotes expectation value with respect to the reduced wigner function at the initial time .
all other cumulants experience oscillatory decay with time scales of @xmath277 with @xmath12 being the order of the cumulant . in particular
, the expectation value @xmath278 follows a trajectory plotted in fig .
[ fig : spiral ] , where one can see that the trajectory of the expectation values @xmath279 for any initial distribution inspiral into the origin .
the trajectory of the expectation values @xmath280.,width=384 ] as we have seen above , any specific features of the initial distribution decay away ( assuming a non - vanishing ohmic term in the spectral function ) and at late times the state tends generically to a gaussian with a covariance matrix given by eq . .
therefore , from eq .
it follows that at late times @xmath281 and @xmath282 . from eq . and using eqs .
- we can express the position and momentum uncertainties at late times as @xmath283 the product of the two uncertainties @xmath284 can be expanded for high temperatures as @xmath285 inspecting the terms in powers of @xmath11 immediately reveals the high - temperature result of classical statistical mechanics for the case of an ohmic spectrum .
integer subohmic terms would strictly decrease this amount .
one can also see that at weak coupling the uncertainty function agrees with the weak coupling approximation for moderate values of the cut - off scale , as shown in fig .
[ fig : dxdp ] .
late time @xmath286 for , , , and .,width=384 ] had one naively tried to have finite diffusion coefficients in the limit @xmath27 subtracting by hand the @xmath157 term , one would find a violation of the heisenberg uncertainty principle at low temperature and strong coupling ( see fig.[fig : dxdp3d - ren ] ) , which renders the theory unphysical .
of course this does not happen with the unsubtracted theory , as seen in fig .
[ fig : dxdp3d - unren ] .
it is thus clear that the logarithmic dependence on the ultraviolet cut - off that appears in the diffusion coefficients is a physically important parameter and not something that can be subtracted away .
late time @xmath286 for the subtracted theory.,width=384 ] late time @xmath286 for the unsubtracted theory.,width=384 ] from eqs .
- [ or alternatively from eq .
] one can see that mentioned earlier , only the momentum uncertainty contains a logarithmically divergent cut - off dependence .
in contrast , the position uncertainty is much smaller and finite in the limit @xmath27 ( this had already been noticed for gaussian wave - packets in ref .
in fact , if the brownian particle is coupled to a reservoir at arbitrarily low temperature and arbitrarily strong coupling , .
moreover , one needs to keep @xmath287 , which means that @xmath1 also tends to infinity as @xmath288 .
alternatively , one should use eqs .
- plus the terms involving negative powers of @xmath1 which were neglected there , so that the restriction @xmath289 does not apply . in both cases
one gets the result quoted in eqs . - .
] the uncertainty in position will be arbitrarily small : @xmath290 hence , strong coupling to a low temperature reservoir tends to localize the wigner distribution in position . from this result , one would qualitatively expect strong coupling to suppress quantum tunneling while high temperature would aid both tunneling and classical escape ( thermal activation ) . in this subsection
we investigate the linear entropy @xcite , which can be easily obtained from the wigner distribution as follows : @xmath291 in fourier space it becomes @xmath292 and using the result in eq .
we finally get @xmath293 at the initial time the linear entropy is that of the initial state , and at late times it tends to @xmath294 . alternatively , one can express the linear entropy in terms of an integral of the fourier - transformed reduced wigner function at the initial time by introducing the change of variables @xmath295 .
eq . can then be written as @xmath296 } \int d^2\mathbf{k_0 } { \mbox{\boldmath$\sigma$}}_t(t ) e^{+t \mathbf{h}^{{\!\scriptscriptstyle\mathrm t } } } ; \mathbf{k_0 } \right ) , \end{aligned}\ ] ] where @xmath297 is a normalized gaussian distribution for the variable @xmath298 with mean @xmath299 and covariance @xmath300 . for small times
this integral is similar to that for the initial state , whereas for long times the normalized gaussian distribution becomes increasingly close to a delta function . for a gaussian initial state
@xmath301 $ ] the integral in eq .
can be explicitly computed : @xmath302 } .\end{aligned}\ ] ] for these gaussian states , reasonable linear entropy is synonymous with reasonable uncertainty functions ( _ i.e. _ , the linear entropy will be positive if and only if the heisenberg uncertainty principle is satisfied ) .
we have already found that the late time uncertainty is well behaved .
the uncertainty at the initial and intermediate times should not violate the heisenberg uncertainty principle either . as a particular example , let us consider an initial state that corresponds to the ground state associated with the bare frequency . at the initial time
the determinant is completely specified by the determinant of @xmath303 , which equals @xmath304 as a consequence of the cancellation between the large @xmath305 factor ( of the order of @xmath306 with @xmath307 ) and the small @xmath308 factor ( of order @xmath309 ) . however , after a very short time , when @xmath257 starts acquiring non - vanishing values , it will dominate the total contribution to @xmath308 and the linear entropy will increase suddenly ( this kind of behavior for the entropy was found in ref .
@xcite ) . at later times
it will relax to its thermal equilibrium values . in situations more general than the laurent series spectrum and where the classical equation of motion involves fractional calculus ( _ i.e. _
, it becomes an integro - differential equation rather than an ode ) , the master equation can be of a slightly more general nature , with additional time dependence in the frequency and the dissipation coefficients : @xmath310 the method for finding its solutions will be almost the same , except that solving for the characteristic curves of the fourier - transformed phase - space variables is now less straightforward due to the time dependence of @xmath240 , and the procedure followed to solve eq . needs to be generalized .
the system of odes that corresponds to @xmath311 can be decoupled through a process of differentiation and substitution that leads to the equations for a pair of independent parametric oscillators where solving one effectively solves the other up to an integral .
the second order ode satisfied by @xmath312 is @xmath313 after solving it , @xmath314 can be simply obtained as @xmath315 alternatively , one can proceed in the reverse order as follows : @xmath316 solving for @xmath312 first appears to be simpler since everything is then linear in the coefficients .
it is convenient to factor out the exponential growth , which reduces the ode to its undamped form : @xmath317 if one is able to solve this differential equation , which for simple enough functions could be performed with floquet analysis and variation of parameters , then the characteristic curves can be expressed in terms of the following matrix equation : @xmath318 where @xmath319 is the so - called transition matrix , which has many of the properties of an exponential and contains the parametric oscillatory behavior of the characteristic curves .
if @xmath61 and @xmath320 tend to some asymptotic values at large times , then the behavior of @xmath319 tends to that of a damped harmonic oscillator .
once the transition matrix is available one can apply exactly the same approach as above in order to solve the master equation .
one starts by writing the master equation as @xmath321 next , one solves this linear ode in @xmath235 and , reexpressing the result in terms @xmath33 and @xmath250 , one can finally write the solution of the master equation as @xmath322 with @xmath323^{{\!\scriptscriptstyle\mathrm t } } { \mbox{\boldmath$\phi$}}^{{\!\scriptscriptstyle\mathrm t } } ( s ) \mathbf{d}(s ) { \mbox{\boldmath$\phi$}}(s ) { \mbox{\boldmath$\phi$}}^{-1}(t ) ds \right ) .\ ] ] for ohmic - like coefficients , the qualitative behavior of the solution is very similar to that of the ohmic solution .
the moments of the initial distribution will experience oscillatory decay and the solution will tend to a thermal equilibrium state at late times . in purely supraohmic regimes
where @xmath324 vanishes , the initial state would remain with renormalized frequency . in purely subohmic regimes where @xmath324 grows quickly ,
the initial state would vanish quickly .
in this section we consider the case of a classical force @xmath325 acting on the quantum oscillator .
this is done by introducing a time - dependent potential @xmath326 : @xmath327 to derive the form of the master equation we follow the method of calzetta , roura , and verdaguer @xcite as it is of a very general nature and can be adapted quite straightforwardly . in the presence of an external force the system action takes the following form in terms of the `` center of mass '' and `` relative '' coordinates , @xmath328 and @xmath329 : @xmath330 + f(s ) \delta(s ) \right ) .\end{aligned}\ ] ] the reduced wigner function can still be expressed as an average over a stochastic process and a distribution for the initial conditions as follows : @xmath331 where we used the notation @xmath332 in the last equation , and @xmath333 is now a solution to the langevin equation with the external force : @xmath334 differentiating with respect to time reveals the usual dissipation , renormalized harmonic potential and the classical force potential @xmath326 : @xmath335 after a functional integration by parts , the last term gives rise to the diffusion terms , with the following coefficients : @xmath336 eq .
( [ eq : langevin1 ] ) can be solved in the same way as in the case without external force to obtain @xmath333 .
it is simply given by a homogeneous solution that contains all the information on the initial conditions , plus a convolution of the external force with the retarded propagator associated with the homogeneous part of the equation , plus a convolution of the stochastic source with the same retarded propagator . by the linearity of the functional derivative ,
@xmath337 only depends on the retarded propagator , which is the same as in the case of no external force .
therefore , the master equation has exactly the same form , except for the addition of the classical force in the potential derivative term of the poisson bracket : @xmath338 fourier transforming the phase - space variables , the master equation becomes @xmath339 this equation can also be solved via the method of characteristic curves , following the same method of the previous section : @xmath340 with the solution @xmath341
which can be finally written as @xmath342 where @xmath343 one can see that just as all temperature dependence only appears in the second cumulant , or covariance , the external force only affects the first cumulant , or mean .
the change in the evolution of the mean , given by eq . , simply corresponds to adding the response to the external force . as with an ordinary driven oscillator ,
the response is a convolution of the retarded propagator , which exhibits the natural oscillatory decay of the system , and the driving force .
quantum brownian motion of an oscillator coupled to a thermal reservoir of quantum oscillators has been the canonical model for studying the environmental effects on a quantum system , even of macroscopc scale , such as quantum dissipation , diffusion , decoherence and entanglement .
it also provides important information on quantum measurement , such as noise , fluctuations , correlations , uncertainty relation and standard quantum limit in mesoscopic systems .
many experiments have been carried out for testing these processes .
fifteen years ago an exact master equation @xcite for the reduced density matrix of the system were derived for a general environment of arbitrary spectral density and temperature .
there are claims that exact solutions have been found @xcite . in this paper
we report on solutions to this equation for a fairly general set of physical conditions and analyze its salient features .
we expect these solutions to be useful in realistic settings for the analysis of many problems which can be described by this model . the first question we addressed concerns the initial time divergences .
these divergences result from the sudden coupling of an initially uncorrelated system and environment .
this situation is unphysical and can be resolved by switching on the system - environment interaction smoothly over a time - scale longer than the inverse of the characteristic uv cut - off scale of the environment .
having switched on the coupling , the initial frequency of the oscillator , as it quickly evolves from the bare to the renormalized frequency , is the only remaining initial singular behavior .
but this merely implies a transition from the bare initial state to a `` renormalized '' version of that state , which can be taken as the effective initial state for the subsequent regular evolution .
we then solved the master equation coefficients for a general class of bilinear system - environment couplings , which includes any combination of ohmic and integer supraohmic and subohmic spectral densities with extreme cut - offs .
although it is by no means a complete class of spectral densities , it is a privileged class in that the classical trajectories are determined by ordinary differential equations and not integro - differential or fractional differential equations . in that sense ,
their dynamics is more classical . for these system - environment couplings ,
we have solved the master equation coefficients for all temperature ranges , beyond weak coupling , and at all times after the short initial time - scale during which the coupling is switched on .
this has enabled us to clarify the validity and shortcomings of previously obtained approximations and provide more general results .
perhaps the most useful are our simple expressions for the master equation coefficients at late time [ eq . - ] .
they clearly reveal the existence of logarithmic divergences in the limit of infinite uv cut - off not only in the anomalous diffusion coefficient , but also in the normal one .
these divergences can not be consistently subtracted and condition the possibility of obtaining meaningful results upon the existence of a physically well - motivated cut - off . for the system - environment couplings under discussion , which all have local dissipation , we have also solved the master equation in its entirety .
all initial states evolve into a given gaussian state corresponding to thermal equilibrium .
we have obtained the covariance matrix for this state [ eq . ] , and the corresponding uncertainty function ( sec . [ sec : uncertainty ] ) .
interestingly , arbitrarily strong coupling to a zero temperature reservoir will tend to localize the position with arbitrary precision . for more general systems with non - local dissipation
, we have drastically reduced the task of solving the master equation to that of solving a one - dimensional classical parametric oscillator problem .
the master equation solutions are parametrically similar to the case of local dissipation .
the same kind of terms arise , including a thermal covariance , but it is not guaranteed that the system will relax to a thermal state .
that of course depends upon the behavior of the classical trajectories . finally , we extended the model of the quantum oscillator linearly coupled to a thermal reservoir of oscillators by including a classical driving force .
this modifies the dynamics by driving the mean position and momentum around just as with a classical driven system . in this model we found that the force has no effect upon the width of the wave - packet or any cumulant other than the mean .
these results may be useful for the study of low - temperature measurements of forced oscillators , which are relevant for experiments with nanomechanical resonators @xcite .
they also play a crucial role in future schemes for the detection of gravitational waves with high - intensity laser interferometers , where the radiation pressure effects on the cavity mirrors are important @xcite . c. f. and b. l. h.
are supported in part by grants from the nsf - itr program ( phy-0426696 ) , nist and nsa - lps .
a. r. is supported by ldrd funds from los alamos national laboratory .
the harmonic number @xmath344 is a function similar to a logarithm .
its analytical continuation to the complex plane is similar too .
@xmath345 where eq .
( [ psi2 ] ) is the asymptotic expansion for @xmath346 .
the following definition and properties of the exponential integral are used in this paper : @xmath347 where @xmath348 denotes the cauchy principal value , corresponding to the singularity at @xmath3 of the integrand in eq .
( [ ei1 ] ) , and eq . ( [ ei3 ] ) is an asymptotic expansion for @xmath346 .
99 r. p. feynman and f. l. vernon , ann .
( n.y . ) * 24 * , 118 ( 1963 ) .
a. o. caldeira and a. j. leggett , physica a * 121 * , 587 ( 1983 )
. h. grabert , p. schramm and g. l. ingold , phys .
rep . * 168 * , 115 ( 1988 ) .
a. o. caldeira , h. a. cerdeira and r. ramaswamy , phys .
a * 40 * , 3438 ( 1989 ) . w. g. unruh and w. h. zurek , phys . rev .
d * 40 * , 1071 ( 1989 ) .
l. d. romero and j. p. paz , phys .
a * 55 * , 4070 ( 1997 ) .
j. r. anglin , j. p. paz and w. h. zurek , phys .
a * 55 * , 4041 ( 1997 ) .
h. p. robertson , phys . rev . * 46 * , 794 ( 1934 ) .
d. a. trifonov , eur .
j. b * 29 * , 349 ( 2002 ) .
j. krger , phys .
a * 46 * , 5385 ( 1992 ) .
f. c. lombardo and p. i. villar , phys .
a * 336 * , 16 ( 2005 ) .
b. l. hu , j. p. paz and y. zhang , phys .
d * 45 * , 2843 ( 1992 ) . b. l. hu and y. zhang , mod .
lett . a * 8 * 3573 ( 1993 ) ; int .
* 10 * , 4537 ( 1995 ) .
b. l. hu and a. matacz , phys .
d * 49 * , 6612 ( 1994 ) . c. anastopoulos and j. j. halliwell , phys .
d * 51 * , 6870 ( 1995 ) .
j. j. halliwell and t. yu , phys .
d * 53 * , 2012 ( 1996 ) . b. l. hu , a. roura and e. verdaguer , phys .
d * 73 * , 044002 ( 2004 ) .
e. calzetta , a. roura and e. verdaguer , physica a * 319 * , 188 ( 2003 )
. g. manfedi and m. r. feix , phys . rev .
e * 62 * , 4665 ( 2000 ) .
s. y. lin and b. l. hu , `` where is the unruh effect ? - new insights from exact solutions of uniformly accelerated detectors '' , .
k. shiokawa and b. l. hu , phys . rev .
a * 70 * , 062106 ( 2004 ) .
g. w. ford and r. f. oconnell , phys .
d * 64 * , 105020 ( 2001 ) .
et al . _ ,
nature * 443 * , 193 ( 2006 ) .
lahaye , o. buu , b. camarota and k.c .
schwab , science 304 , 74 ( 2004 ) .
a. buonanno and y. chen , phys . rev .
d * 64 * , 042006 ( 2001 ) . h. j. kimble _ et al .
d * 65 * , 022002 ( 2002 ) .
|
we revisit the model of a system made up of a brownian quantum oscillator under the influence of an external classical force and linearly coupled to an environment made up of many quantum oscillators at zero or finite temperature .
we show that the hpz master equation for the reduced density matrix derived earlier [ b.l .
hu , j.p .
paz , y. zhang , phys .
rev .
d * 45 * , 2843 ( 1992 ) ] with coefficients obtained from solutions of integro - differential equations can assume closed functional forms for a fairly general class of spectral densities of the environment at arbitrary temperature and coupling strength .
as an illustration of these new results we solve the corresponding master equation and calculate , among other physical quantities , the uncertainty function whose late time behavior can be obtained fully .
this produces a formula for investigating the standard quantum limit which is central to addressing many theoretical issues in macroscopic quantum phenomena and experimental concerns related to low temperature precision measurements .
we find that any initial state always settles down to a gaussian density matrix whose covariance is determined by the thermal reservoir and whose mean is determined by the external force . for more general spectra
we show that the solution of the master equation can be reduced to solving for the motion of a classical parametric oscillator with parametric frequency determined by the _
unsolved for _ master equation coefficients .
states in these systems experience evolution that is parametrically similar to the simpler evolution explicitly determined for in the case of laurent - series spectra .
|
studies of solitons in spatially periodic ( lattice ) potentials have grown into a vast area of research , with profoundly important applications to nonlinear optics , plasmonics , and matter waves in quantum gases , as outlined in recent reviews @xcite . in ultracold bosonic and fermionic gases ,
periodic potentials can be created , in the form of optical lattices , by coherent laser beams illuminating the gas in opposite directions @xcite .
effective lattice potentials for optical waves are induced by photonic crystals , which are built as permanent structures by means of various techniques @xcite , or as laser - induced virtual structures in photorefractive crystals @xcite .
parallel to the progress in the experiments , the study of the interplay between the nonlinearity and periodic potentials has been an incentive for the rapid developments of theoretical methods @xcite .
both the experimental and theoretical results reveal that solitons can be created in lattice potentials , if they do not exist in the uniform space [ this is the case of gap solitons ( gss ) supported by the self - defocusing nonlinearity , see original works @xcite and reviews @xcite ] , and solitons may be stabilized , if they are unstable without the lattice ( multidimensional solitons in the case of self - focusing , as shown in refs .
bbb , ym , ym2,bbb2,barcelona , bbb3,bbb4,thawatchai , see also reviews rev0,rev2,rev3 ) .
the stability of gss has been studied in detail too chiefly , close to edges of the corresponding bandgaps in one stability1,stability2,jc and two @xcite dimensions alike .
an essential extension of the theme is the study of two - component solitons in lattice potentials . in particular ,
if both the self - phase - modulation and cross - phase - modulation ( spm and xpm ) nonlinearities , i.e. , intra- and inter - species interactions , are repulsive , one can construct two - component gss of _ intra - gap _ and _ inter - gap _ types , with chemical potentials of the components ( or propagation constants , in terms of optical media ) falling , respectively , into the same or different bandgaps of the underlying linear spectrum @xcite . in the case of the attractive spm , a family of stable _ semi - gap solitons _
was found too , with one component residing in the infinite gap , while the other stays in a finite bandgap @xcite .
the gss supported by the xpm repulsion dominating over the intrinsic ( spm - mediated ) attraction may be regarded as an example of _ symbiotic solitons_. in the free space ( without the lattice potential ) , symbiotic solitons are supported by the xpm attraction between their two components , despite the action of the repulsive spm in each one @xcite .
this mechanism may be additionally enhanced by the linear coupling ( interconversion ) between the components sadhan2 .
another case of the symbiosis " was reported in ref .
@xcite , where the action of the lattice potential on a single component was sufficient for the stabilization of two - dimensional ( 2d ) two - component solitons against the collapse , the stabilizing effect of the lattice on the second component being mediated by the xpm interaction .
in addition , the attraction between the components , competing with the intrinsic repulsion , may cause spatial splitting between two components of the gs , as for these components , whose effective masses are negative @xcite , the attractive interaction potential gives rise to a repulsion force @xcite
. the ultimate form of the model which gives rise to two - component gss of the symbiotic type is the one with no intra - species nonlinearity , the formation of the gss being accounted for by the interplay of the repulsion between the components and the lattice potential acting on both of them . in optics ,
the setting with the xpm - only interactions is known in the form of the holographic nonlinearity " , which can be induced in photorefractive crystals for a pair of coherent beams with a small angle between their wave vectors , giving rise to single- @xcite and double - peak holokiv solitons .
both beams are made by splitting a single laser signal , hence the power ratio between them ( which is essential for the analysis reported below ) can be varied by changing the splitting conditions .
the creation of 2d spatial holographic solitons " in a photorefractive - photovoltaic crystal with the self - focusing nonlinearity was demonstrated in ref .
@xcite ( such solitons are stable , as the collapse is arrested by the saturation of the self - focusing ) .
to implement the situation considered here , the sign of the nonlinearity may be switched to self - defocusing by the reversal of the bias voltage , and the effective lattice potential may be induced by implanting appropriate dopants , with the concentration periodically modulated in one direction , which will render the setting quasi - one - dimensional . in binary bosonic gases , a similar setting may be realized by switching off the spm nonlinearity with the help of the feshbach resonance , although one may need to apply two different spatially uniform control fields ( one magnetic and one optical ) to do it simultaneously in both components . on the other hand ,
the same setting is natural for a mixture of two fermionic components with the repulsive interaction between them , which may represent two states of the same atomic species , with different values of the total atomic spin ( @xmath0 ) .
if spins of both components are polarized by an external magnetic field , the spm nonlinearity will be completely suppressed by the pauli blockade while the inter - component interaction remains active stoof , hence the setting may be described by a pair of schrdinger equations for the two wave functions , coupled by xpm terms .
the objective of this work is to present basic families of one - dimensional symbiotic gss , supported solely by the repulsive xpm nonlinearity in the combination with the lattice potential , and analyze their stability , via the computation of eigenvalues for small perturbations and direct simulations of the perturbed evolution .
the difference from the previously studied models of symbiotic solitons @xcite is that the solitons where created there by the spm nonlinearity separately in each component , while the xpm interaction determined the interaction between them and a possibility of creating two - component bound states . here
, the two - component gss may exist solely due to the repulsive xpm interactions between the components .
we conclude that the symmetric solitons , built of equal components , are destabilized by symmetry - breaking perturbations above a certain critical value of the soliton s power .
the analysis is chiefly focused on asymmetric symbiotic gss , and on breathers into which unstable solitons are transformed .
the model is introduced in section ii , which is followed by the analytical approximation presented in section iii .
it is an extended ( tailed " ) version of the thomas - fermi approximation , tfa , which may be applied to other models too . in section iv , we report systematic numerical results obtained for fundamental solitons of both the intra- and inter - gap types , hosted by the first two finite bandgaps of the system s spectrum .
the most essential findings are summarized in the form of plots showing the change of the gs stability region with the variation of the degree of asymmetry of the two - component symbiotic solitons , which is a new feature exhibited by the present system .
in particular , the stability area of intra - gap solitons shrinks with the increase of the asymmetry , while inter - gap solitons may be stable only if the asymmetry is large enough , in favor of the first - bandgap component , and intra - gap solitons in the second bandgap are completely unstable .
the paper is concluded by section v.
the model outlined above is represented by the system of xpm - coupled schrdinger equations for local amplitudes of co - propagating electromagnetic waves in the planar optical waveguide , @xmath1 and @xmath2 , where @xmath3 and @xmath4 are the transverse coordinate and propagation distance , without the spm terms , and with the lattice potential of depth @xmath5 acting on both components:@xmath6the variables are scaled so as to make the lattice period equal to @xmath7 , and the coefficients in front of the diffraction and xpm terms equal to @xmath8 . in the case of matter waves , @xmath9 and @xmath10 are wave functions of the two components , and @xmath11 is replaced by time @xmath12 .
direct simulations of eqs .
( 1dpsi ) and ( [ 1dphi ] ) were performed with the help of the split - step fourier - transform technique . ) and ( [ ode2 ] ) for @xmath13 .
shaded areas are occupied by the bloch bands , where gap solitons do not exist.,width=288 ] stationary solutions to eqs .
( [ 1dpsi ] ) , ( [ 1dphi ] ) are looked as@xmath14where the real propagation constants , @xmath15 and @xmath16 , are different , in the general case , and real functions @xmath17 and @xmath18 obey equations@xmath19with the prime standing for @xmath20 .
numerical solutions to eqs .
( [ ode1 ] ) and ( [ ode2 ] ) were obtained by means of the newton s method .
solitons are characterized by the total power,@xmath21with both @xmath22 and @xmath23 being dynamical invariants of eqs .
( [ 1dpsi ] ) , ( [ 1dphi ] ) , and by the _ asymmetry ratio_,@xmath24the total power and asymmetry may be naturally considered as functions of the propagation constants , @xmath15 and @xmath16 . the well - known bandgap spectrum of the linearized version of eqs . ( [ ode1 ] ) , ( [ ode2 ] ) ( see , e.g. , book @xcite )
is displayed in fig .
[ fig1 ] , the right edge of the first finite bandgap being @xmath25the location of gss is identified with respect to bandgaps of the spectrum . in this work , results are reported for composite gss whose two components belong to the first and second finite bandgaps .
stability of the stationary solutions can be investigated by means of the linearization against small perturbations stability , stability1,stability2 . to this end ,
perturbed solutions of eqs .
( [ 1dpsi ] ) and ( [ 1dphi ] ) are looked for as@xmath26 , \nonumber \\
v\left ( x , z\right ) & = & e^{iqz}\left [ v(x)+v_{1}(x)e^{-i\lambda z}+v_{2}^{\ast } ( x)e^{i\lambda ^{\ast } z}\right ] , \label{pert}\end{aligned}\]]where @xmath27 and @xmath28 are wave functions of infinitesimal perturbations , and @xmath29 is the respective instability growth rate , which may be complex ( the asterisk stands for the complex conjugate ) .
the instability takes place if there is at least one eigenvalue with @xmath30 .
the substitution of ansatz eqs .
( [ pert ] ) into eqs .
( [ 1dpsi ] ) , ( [ 1dphi ] ) and the linearization with respect to the small perturbations leads to the eigenvalue problem based on the following equations:@xmath31these equations were solved by means of the fourth - order center - difference numerical scheme .
results for the shape and stability of gss of different types are presented below for lattice strength @xmath32 , which adequately represents the generic case .
note , in particular , that fig .
[ fig1 ] was plotted for this value of the lattice - potential s strength .
it is well known that , close to edges of the bandgap , gss feature an undulating shape , which may be approximated by a bloch wave function modulated by a slowly varying envelope @xcite .
on the other hand , deeper inside the bandgap , the gss are strongly localized ( see , e.g. , figs .
[ fig4 ] and [ fig7 ] below ) , which suggests to approximate them by means of the variational method based on the gaussian ansatz sadhan , physicad .
this approximation was quite efficient for the description of gss in single - component models @xcite , while for two - component systems it becomes cumbersome @xcite .
explicit analytical results for well - localized patterns can be obtained by means of the tfa @xcite , which , in the simplest case , neglects the kinetic - energy terms , @xmath33 and @xmath34 , in eqs .
( [ ode1 ] ) , ( [ ode2 ] ) .
assuming , for the sake of the definiteness , @xmath35 and also @xmath36 ( the tfa is irrelevant for @xmath37 ) , the approximation yields the fields inside the _ inner layer _ of the solution:@xmath38thus , the tfa predicts the core part of the solution in the form of peaks in the two components with the same width , @xmath39 , but different heights , @xmath40 .
this structure complies with numerically generated examples of asymmetric solitons displayed below in figs .
[ fig7](a , b ) .
further , the expansion of expressions ( [ in ] ) around the soliton s center ( @xmath41 ) , yields@xmath42the substitution of the second derivatives of the fields at @xmath41 , calculated as per eq .
( [ expand ] ) , into eqs .
( [ ode1 ] ) , ( [ ode2 ] ) yields a corrected expression for the soliton s amplitudes:@xmath43 ^{-1 } ,
\\ \sqrt{\varepsilon -k}-\varepsilon \left [ 2\left ( \varepsilon -q\right ) \sqrt{% \varepsilon -k}\right ] ^{-1},% \end{array}% \right\ } ~ \label{corr}\]]along with the condition for the applicability of the tfa:@xmath44for the symmetric - gs families in the two first finite bandgaps , the amplitude predicted by the improved tfa in the form of eq .
( [ corr ] ) is displayed , as a function of @xmath45 , in fig .
[ fig2](a ) and compared to its numerically found counterpart .
it is worthy to note that the correction terms in eq .
( [ corr ] ) essentially improve the agreement of the tfa prediction with the numerical findings . at @xmath46 ,
the tfa gives @xmath47 , which is a continuous extension of the respective expression eq .
( [ in ] ) in the @xmath48 component , while the continuity of fields @xmath17 and @xmath49 makes it necessary to match the respective expression eq .
( [ in ] ) to tails " , which , in the lowest approximation , satisfy equation @xmath50 .
the continuity is provided by the following tail solution:@xmath51 ^{2},~ \\ \mathrm{at}~~0<|x|-x_{0}<\left ( k - q\right ) /\sqrt{\varepsilon ^{2}-k^{2 } } ; \\ 0,~\mathrm{at}~~|x|-x_{0}>\left ( k - q\right ) /\sqrt{\varepsilon
^{2}-k^{2}}.% \end{array}% \right\ } \label{outer}\ ] ] the integration of expressions eq .
( [ in ] ) and eq .
( [ outer ] ) yields the following approximation for the powers of the two components:@xmath52the substitution of approximation eqs .
( [ pp ] ) into definitions eq .
( p ) and eq .
( [ r ] ) of the total power and asymmetry demonstrates an agreement with numerical results .
for instance , the slope of the curve @xmath53 for the intra - gap gss at fixed @xmath15 ( see fig . [ fig8 ] below ) at the symmetry point ( @xmath45 ) , as predicted by eqs .
( [ pp ] ) for @xmath32 and @xmath54 , is @xmath55 while its numerically found counterpart is @xmath56 .
further , the analysis of eqs .
( [ pp ] ) readily demonstrates that the strongly asymmetric solitons may exist up to the limit of @xmath57 , which is corroborated by the existence area for the intra - gap solitons shown below in fig .
[ fig11](b ) .
.,width=288 ] another corollary of eqs .
( [ pp ] ) is the prediction for the total power for the symmetric solitons,@xmath58 , \label{ptf}\]]which is plotted in fig .
[ fig2](b ) , along with its numerically found counterpart .
although the tfa does not predict edges of the bandgaps , the overall analytical prediction for @xmath59 runs quite close to the numerical curve , except for near the right edge , where , indeed condition eq .
( [ < < ] ) does not hold for @xmath32 and @xmath60 , see eq .
( [ k_max ] ) .
note that this very simple analytical approximation was not derived before in numerous works dealing with single - component gss . and
@xmath61 ( left and right panels , respectively ) . here and in similar figures below , the background pattern ( green sinusoid ) represents the underlying periodic potential .
both solitons are stable as solutions of the single - component model , but only the one corresponding to @xmath62 remains stable in the two - component system , while its counterpart pertaining to @xmath63 is destabilized by symmetry - breaking perturbations.,width=288 ]
obviously , the shape of symmetric solitons , built of two equal components [ with @xmath45 and @xmath64 , see eq .
( [ gs ] ) ] , is identical to that of gss in the single - component model .
however , there is a drastic difference in the stability of the symmetric gss between the single- and two - component systems .
almost the entire symmetric family is stable against symmetric perturbations , i.e. , it is stable in the framework of the single - component equation ( in agreement with previously known results @xcite ) , except for a weak oscillatory instability , accounted for by quartets of complex - conjugate eigenvalues , in the form of @xmath65 ( with two mutually independent signs @xmath66 ) , which appears near the left edge of the second bandgap namely , at @xmath67 .
an example the development of the latter instability is displayed below in fig .
[ fig3 ] . on the other hand , fig .
[ fig2 ] demonstrates that a considerable part of the family in the first finite bandgap , and the entire family in the second bandgap are unstable against symmetry - breaking perturbations in the two - component system .
the boundary separating the stable and unstable subfamilies of the fundamental symmetric gss in the first finite bandgap corresponds to the power and propagation constants is found at @xmath68the symmetric solitons being stable in the intervals of @xmath69 , @xmath70 [ see eq .
( [ k_max ] ) ] .
these results were produced by a numerical solution of eqs .
( [ u1])-([v2 ] ) ( the instability is oscillatory , characterized by complex eigenvalues ) . typical examples of stable and unstable fundamental symmetric gss , found in the first and second bandgaps ( not too close to their edges ) , are displayed in fig .
further , direct simulations demonstrate that the evolution transforms the unstable symmetric solitons into persistent localized breathers , as shown in figs .
[ fig5 ] and [ fig6 ] , in accordance with the fact that the corresponding instability eigenvalues are complex .
although the emerging breather keeps the value of @xmath71 , see eq .
( [ r ] ) , the @xmath9- and @xmath10- components of the breather generated by the symmetry - breaking instability are no longer mutually identical .
this manifestation of the symmetry - breaking instability is illustrated by fig .
[ fig5](c ) , which displays the evolution of the difference between the peak powers of the two components , and the separation between their centers .
the latter is defined as@xmath72 it is relevant to mention that the second finite bandgap also contains a branch of the so - called subfundamental solitons , whose power is smaller than that of the fundamental gss @xcite .
these are odd modes , squeezed , essentially , into a single cell of the underlying lattice potential .
the subfundamental solitons are unstable , tending to rearrange themselves into fundamental ones belonging to the first finite bandgap , therefore they are not considered below .
[ defined as per eq . ( [ r ] ) ] , versus propagation constant @xmath16 , at fixed values of @xmath73 , @xmath74 , and @xmath75 ( the top , middle , and bottom curves , respectively ) , for asymmetric fundamental solitons of the intra - gap type .
stable and unstable branches are shown by solid and dashed lines , respectively.,width=288 ] and @xmath76 .
fields @xmath17 and @xmath18 , which pertain to propagation constants @xmath15 and @xmath16 , are shown , respectively , by the magenta ( lower ) and blue ( taller ) profiles.,width=192 ] as said above , two - component asymmetric fundamental gss , with different propagation constants , @xmath77 , may be naturally classified as solitons of the intra- and inter - gap types if @xmath15 and @xmath16 belong to the same or different finite bandgaps @xcite . in this subsection
, we report results for asymmetric intra - gap solitons with both @xmath15 and @xmath16 falling into the first finite bandgap , as well as for asymmetric breathers developing from such solitons when they are unstable . examples of stable asymmetric gss of the intra - gap type are displayed in fig .
[ fig7 ] , for a fixed asymmetry ratio , @xmath78 , defined as per eq .
( [ r ] ) . the gs family , along with the family of persistent breathers into which unstable solitons are spontaneously transformed ,
is represented in fig .
[ fig8 ] by dependences @xmath53 at different fixed values of the other propagation constant , @xmath15 .
it is possible to explain the fact that all the @xmath53 curves converge to @xmath79 , as @xmath16 approaches the right edge of the bandgap in fig .
[ fig8 ] . in this case
, the @xmath48 component turns into the delocalized bloch wave function with a diverging power , @xmath23 , that corresponds to @xmath80 [ it is tantamount to @xmath81 as per eq . ( [ r ] ) ] .
an example of a stable gs , close to this limit , with @xmath82 , @xmath83 and @xmath84 , is shown in fig .
the central core of the @xmath48-component is described by the tfa , based on eq .
( [ in ] ) , as the corresponding necessary condition ( [ < < ] ) holds in this case , while the tfa does not apply to the @xmath85-component . the presence of undulating tails , which are close to the bloch functions , rather than the simple approximation ( [ outer ] ) , which is valid far from the edge of the bandgap , is also visible in fig .
[ fig8 ] .
those asymmetric intra - gap gss which form unstable subfamilies in fig .
fig8 are destabilized by oscillatory perturbations .
the instability transform the solitons into breathers , see a typical example in fig .
fig10 [ cf . the examples of the destabilization of the symmetric gss shown in fig .
[ fig5](b , c ) ] .
the emerging breathers keep values of the asymmetry ratio ( [ r ] ) almost identical to those of their parent gss ; for instance , in the case displayed in this figure , the unstable soliton with @xmath86 _
_ _ evolves into the breather with @xmath87 .
it is relevant to stress that the transformation of unstable stationary gss into the breathers gives rise to little radiation loss of the total power , @xmath88 .
on the other hand , in the general case a given unstable gap soliton does not have a stable counterpart with a close value of @xmath89 , hence this unstable soliton can not transform itself into a slightly excited state of another stable gs .
thus , the breathers represent a distinct species of localized modes .
the most essential results of the stability analysis for the asymmetric solitons of the intra - gap type , and for breathers replacing unstable solitons , are summarized by diagrams in the planes of @xmath90 and @xmath91 , which are displayed in fig .
[ fig11 ] .
the predictions of the analysis based on the computation of the stability eigenvalues for the stationary solitons , as per eqs .
( [ u1])-([v2 ] ) , always comply with stability tests provided by direct simulations of eqs .
( [ 1dpsi ] ) and ( [ 1dphi ] ) . as mentioned above , the instability of a part of the branch of the symmetric solitons along the line of @xmath71 in fig .
[ fig11](b ) implies that the symmetry - breaking perturbations destabilize the symmetric solitons in the first finite bandgap at @xmath92 , see eq .
( [ cr ] ) , while their counterparts are stable in the single - component system .
another clear conclusion is that the stability region gradually shrinks with the increase of the asymmetry .
all the gss of the inter - gap type , with two propagation constants belonging to the two different finite bandgaps , are naturally asymmetric , even if their components have equal powers .
examples of stable and unstable inter - gap solitons are displayed in figs .
[ fig12 ] and [ fig13 ] , respectively .
a noteworthy feature exhibited by these examples is a split - peak structure of the component belonging to the second finite bandgap . and
the single - peak and split - peak profiles , @xmath17 and @xmath18 , represent , respectively , the components in the first and second finite bandgaps.,width=192 ] the asymmetry measure , @xmath53 , for families of the inter - gap solitons is plotted in fig .
[ fig14](a ) versus the propagation constant @xmath16 in the second bandgap , at fixed values of @xmath15 ( the propagation constant in the first bandgap ) .
the stability of the respective gs families is also shown in fig . [ fig14 ] . for the inter - gap solitons versus propagation constant @xmath16 in the second finite bandgap , at fixed values @xmath73
, @xmath94 , @xmath95 , and @xmath96 ( from the top to the bottom ) of wavenumber @xmath15 in the first bandgap .
solid and dashed lines designate stable stationary solitons and breathers , respectively.,width=288 ] the ( in)stability of the inter - gap solitons is summarized by the diagrams in the planes of @xmath90 and @xmath91 presented in fig .
[ fig15 ] , cf .
similar diagrams for intra - gap solitons shown above in fig .
[ fig11 ] . as well as in that case ,
unstable inter - gap solitons are spontaneously replaced by robust localized breathers .
the spontaneous transformation increases the initial degree of the asymmetry : for instance , an unstable inter - gap soliton with @xmath97 is converted into a breather with @xmath98 . in the present case too , the existence region of stable modes shrinks with the increase of the asymmetry ; note also that the stationary inter - gap solitons may be stable solely at sufficiently large values of the asymmetry , @xmath99 .
the asymmetric shape of the stability diagram in fig .
[ fig15](b ) with respect to @xmath100 and @xmath101 [ unlike the symmetry of the diagram for the intra - gap solitons implied in fig .
[ fig11](b ) ] is explained by the fact that , in definition ( [ r ] ) , @xmath100 implies that the dominant component resides in the first finite bandgap , where it is more robust than in the second bandgap .
finally , out additional analysis has demonstrated that all the stationary gss not only the symmetric ones [ see fig .
[ fig2 ] ] , but also all the asymmetric solitons of the intra - gap type are completely unstable in the second finite bandgap .
they too tend to spontaneously rearrange themselves into breathers , which is not shown here in detail .
we have introduced the model of symbiotic two - component gss ( gap solitons ) , based on two nonlinear schrdinger equations coupled by the repulsive xpm terms and including the lattice potential acting on both components , in the absence of the spm nonlinearity .
the model has a realization in optics , in terms of holographic solitons " in photonic crystals , and as a model of binary quantum gases ( in particular , a fully polarized fermionic one ) loaded into the optical - lattice potential .
families of fundamental asymmetric gss have been constructed in the two lowest finite bandgaps , including the modes of both the intra - gap and inter - gap types , i.e. , those with the propagation constants of the two components belonging to the same or different bandgaps , respectively .
the existence and stability regions of the symbiotic gss and breathers , into which unstable solitons are transformed , have been identified .
a noteworthy finding is that symmetry - breaking perturbations destabilize the symmetric gss in the first finite bandgap , if their total power exceeds the critical value given by eq .
( [ cr ] ) , along with all the symmetric solitons in the second bandgaps .
it was demonstrated too that the stability area for the intra - gap gss shrinks with the increase of the asymmetry ratio , @xmath102 . on the other hand ,
inter - gap gss may be stable only for sufficiently large ratio , @xmath103 .
the intra - gap solitons are completely unstable in the second bandgap
. some features of the gs families were explained by means of the extended tfa ( thomas - fermi approximation ) , augmented by the tails attached to the taller component , in the case of asymmetric solitons .
a natural extension of the analysis may deal with 2d symbiotic gap solitons , supported by the square- or radial - lattice potentials . in that case
, it may be interesting to consider two - component solitary vortices too .
the work of t.m . was supported by the thailand research fund through grant rmu5380005 .
b.a.m . appreciates hospitality of the mahanakorn university of technology ( bangkok , thailand ) .
|
we consider a two - component one - dimensional model of gap solitons ( gss ) , which is based on two nonlinear schrdinger equations , coupled by repulsive xpm ( cross - phase - modulation ) terms , in the absence of the spm ( self - phase - modulation ) nonlinearity .
the equations include a periodic potential acting on both components , thus giving rise to gss of the symbiotic " type , which exist solely due to the repulsive interaction between the two components .
the model may be implemented for holographic solitons " in optics , and in binary bosonic or fermionic gases trapped in the optical lattice .
fundamental symbiotic gss are constructed , and their stability is investigated , in the first two finite bandgaps of the underlying spectrum .
symmetric solitons are destabilized , including their entire family in the second bandgap , by symmetry - breaking perturbations above a critical value of the total power .
asymmetric solitons of intra - gap and inter - gap types are studied too , with the propagation constants of the two components falling into the same or different bandgaps , respectively .
the increase of the asymmetry between the components leads to shrinkage of the stability areas of the gss .
inter - gap gss are stable only in a strongly asymmetric form , in which the first - bandgap component is a dominating one .
intra - gap solitons are unstable in the second bandgap .
unstable two - component gss are transformed into persistent breathers .
in addition to systematic numerical considerations , analytical results are obtained by means of an extended ( tailed " ) thomas - fermi approximation ( tfa ) .
99 b. a. malomed , d. mihalache , f. wise , and l. torner , spatiotemporal optical solitons , " j. opt .
b : quant . semicl . opt .
* 7 * , r53r72 ( 2005 ) .
f. lederer , g. i. stegeman , d. n. christodoulides , g. assanto , m. segev , and y. silberberg , discrete solitons in optics , " phys .
rep . * 463 * , 1 ( 2008 ) .
y. v. kartashov , v. a. vysloukh , and l. torner , soliton shape and mobility control in optical lattices , " progr . opt . * 52 * , 63148 ( 2009 ) .
y. v. kartashov , b. a. malomed , and l. torner , solitons in nonlinear lattice , " rev . mod . phys . * 83 * , 247306 ( 2011 ) . l. pitaevskii and s. stringari ,
_ bose - einstein condensate _
( clarendon press : oxford , 2003 ) .
h. t. c. stoof , k. b. gubbels , and d. b. m. dickerscheid , _ ultracold quantum fields _ ( springer : dordrecht , 2009 ) .
o. morsch and m. oberthaler , dynamics of bose - einstein condensates in optical lattices , " rev .
mod .
phys . * 78 * , 179215 ( 2006 ) .
j. d. joannopoulos , s. g. johnson , j. n. winn , and r. d. meade , _ photonic crystals : molding the flow of light _
( princeton university press : princeton , 2008 ) .
a. szameit , d. blmer , j. burghoff , t. schreiber , t. pertsch , s. nolte , and a. tnnermann , 2005 , discrete nonlinear localization in femtosecond laser written waveguides in fused silica , " opt .
express * 13 * , 1055210557 ( 2005 ) .
n. k. efremidis , s. sears , d. n. christodoulides , j. w. fleischer , and m. segev , discrete solitons in photorefractive optically induced photonic lattices , " phys .
rev .
e * 66 * , 046602 ( 2002 ) .
j. yang , _ nonlinear waves in integrable and nonintegrable systems _
( siam : philadelphia , 2010 ) .
d. e. pelinovsky , _ localization in periodic potentials _
( cambridge university press : cambridge , uk , 2011 ) .
b. b. baizakov , v. v. konotop , and m. salerno , regular spatial structures in arrays of bose
einstein condensates induced by modulational instability , " j. phys .
b : at . mol .
. phys . * 35 * , 51055119 ( 2002 ) .
p. j. y. louis , e. a. ostrovskaya , c. m. savage , and y. s. kivshar , bose - einstein condensates in optical lattices : band - gap structure and solitons , " phys .
rev .
a * 67 * , 013602 ( 2003 ) .
h. sakaguchi and b. a. malomed , dynamics of positive- and negative - mass solitons in optical lattices and inverted traps , " j. phys .
b * 37 * , 14431459 ( 2004 ) .
b. baizakov , b. a. malomed , and m. salerno , `` matter - wave solitons in radially periodic potentials , '' phys . rev .
e * 74 * , 066615 ( 2006 ) .
v. a. brazhnyi and v. v. konotop , theory of nonlinear matter waves in optical lattices , " mod .
phys .
lett .
b * 18 * , 627651 ( 2004 ) .
b. b. baizakov , b. a. malomed and m. salerno , multidimensional solitons in periodic potentials , " europhys .
lett . * 63 * , 642648 ( 2003 ) .
j. yang , and z. h. musslimani , fundamental and vortex solitons in a two - dimensional optical lattice , " opt
. lett . * 28 * , 20942096 ( 2003 ) .
z. h. musslimani and j. yang , self - trapping of light in a two - dimensional photonic lattice , " j. opt .
soc . am .
b * 21 * , 973981 ( 2004 ) .
b. b. baizakov , b. a. malomed and m. salerno , multidimensional solitons in a low - dimensional periodic potential , " phys .
rev . a * 70 * , 053613 ( 2004 ) .
d. mihalache , d. mazilu , f. lederer , y. v. kartashov , l .- c .
crasovan , and l. torner ,
stable three - dimensional spatiotemporal solitons in a two - dimensional photonic lattice , " phys .
rev .
e * 70 * , 055603(r ) ( 2004 ) .
b. b. baizakov , b. a. malomed , and m. salerno , multidimensional semi - gap solitons in a periodic potential , " eur .
phys .
j. * 38 * , 367374 ( 2006 ) .
b. baizakov , b. a. malomed , and m. salerno ,
matter - wave solitons in radially periodic potentials , " phys .
rev .
e * 74 * , 066615 ( 2006 ) .
t. mayteevarunyoo , b. a. malomed , b. b. baizakov , and m. salerno ,
matter - wave vortices and solitons in anisotropic optical lattices , " physica d * 238 * , 14391448 ( 2009 ) .
k. m. hilligsoe , m. k. oberthaler , and k. p. marzlin , stability of gap solitons in a bose - einstein condensate , " phys .
rev .
a * 66 * , 063605 ( 2002 ) .
d. e. pelinovsky , a. a. sukhorukov , and y. s. kivshar , bifurcations and stability of gap solitons in periodic potentials , " phys .
rev .
e * 70 * , 036618 ( 2004 ) .
g. hwanga , t. r. akylas , and j. yang , gap solitons and their linear stability in one - dimensional periodic media , " physica d * 240 * , 10551068 ( 2011 ) .
j. cuevas , b. a. malomed , p. g. kevrekidis , and d. j. frantzeskakis , solitons in quasi - one - dimensional bose - einstein condensates with competing dipolar and local interactions , " phys .
rev .
a * 79 * , 053608 ( 2009 ) .
z. shi , j. wang , z. chen , and j. yang , linear instability of two - dimensional low - amplitude gap solitons near band edges in periodic media , " phys .
rev .
a * 78 * , 063812 ( 2008 ) .
a. gubeskys , b. a. malomed , and i. m. merhasin , two - component gap solitons in two- and one - dimensional bose - einstein condensate , " phys .
rev .
a * 73 * , 023607 ( 2006 ) .
s. k. adhikari and b. a. malomed , symbiotic gap and semigap solitons in bose - einstein condensates , " phys .
rev .
a * 77 * , 023607 ( 2008 ) . v. m. prez - garca and j. b. beitia ,
symbiotic solitons in heteronuclear multicomponent bose - einstein condensates , " phys .
rev .
a * 72 * , 033620 ( 2005 ) .
s. k. adhikari , bright solitons in coupled defocusing nls equation supported by coupling : application to bose - einstein condensation , " phys .
lett .
a * 346 * , 179185 ( 2005 ) .
s. k. adhikari , fermionic bright soliton in a boson - fermion mixture , " phys .
rev .
a * 72 * , 053608 ( 2005 ) .
s. k. adhikari and b. a. malomed , two - component gap solitons with linear interconversion , " phys . rev . a * 79 * , 015602 ( 2009 )
. o. v. borovkova , b. a. malomed and y. v. kartashov , two - dimensional vector solitons stabilized by a linear or nonlinear lattice acting in one component , " epl * 92 * , 64001 ( 2010 ) .
m. matuszewski , b. a. malomed , and m. trippenbach , competition between attractive and repulsive interactions in two - component bose - einstein condensates trapped in an optical lattice , " phys .
rev .
a * 76 * , 043826 ( 2007 ) .
o. cohen , t. carmon , m. segev , and s. odoulov , holographic solitons , " opt . lett . * 27 * , 20312033 ( 2002 ) .
o. cohen , m. m. murnane , h. c. kapteyn , and m. segev , cross - phase - modulation nonlinearities and holographic solitons in periodically poled photovoltaic photorefractives , " opt .
lett . *
31 * , 954956 ( 2006 ) .
j. r. salgueiro , a. a. sukhorukov , and y. s. kivshar , spatial optical solitons supported by mutual focusing , " opt
. lett .
28 , 14571459 ( 2003 ) .
j. liu , s. liu , g. zhang , and c. wang , observation of two - dimensional holographic photovoltaic bright solitons in a photorefractive - photovoltaic crystal , " appl .
. lett . * 91 * , 111113 ( 2007 ) .
s. adhikari and b. a. malomed ,
gap solitons in a model of a superfluid fermion gas in optical lattices , " physica d * 238 * , 14021412 ( 2009 ) .
n. k. efremidis and d. n. christodoulides , lattice solitons in bose - einstein condensates , " phys .
rev .
a * 67 * , 063608 ( 2003 ) .
t. mayteevarunyoo and b. a. malomed ,
stability limits for gap solitons in a bose - einstein condensate trapped in a time - modulated optical lattice , " phys .
rev .
a * 74 * , 033616 ( 2006 ) .
j. cuevas , b. a. malomed , p. g. kevrekidis , and d. j. frantzeskakis , solitons in quasi - one - dimensional bose - einstein condensates with competing dipolar and local interactions , " phys .
rev .
a * 79 * , 053608 ( 2009 ) .
|
a second order phase transition has been observed in the heavy fermion metal uru@xmath0si@xmath0 at @xmath3=17.5 k via a sharp discontinuous jump in the specific heat with about 24% entropy loss.@xcite considering an associated sharp anomaly in the @xmath4axis magnetic susceptibility , it was initially assumed to be an antiferromagnetic ( af ) phase.@xcite however , in subsequent works,@xcite including a very recent one,@xcite the presence of any static magnetic moment has been eliminated .
this apparent ` magnetic dichotomy ' has remained a smoking gun feature of the phase , known as the ` so - called ' hidden order ( ho ) phase of uru@xmath0si@xmath0 .
additional key fingerprints of the ho state include a relatively weak intensity peak at the commensurate wavevector @xmath5=(001),@xcite and a dispersive collective excitation at an incommensurate vector @xmath6=([email protected],0,0),@xcite as probed by inelastic neutron scattering ( ins ) measurement .
furthermore , the ins spectral weight loss at the incommensurate wavevector , as opposed to the commensurate one , can fully account for the entropy loss at the ho transition .
direct spectroscopies including angle - resolved photoemission spectroscopy ( arpes)@xcite scanning tunneling microscopy ( stm)@xcite and point contact spectroscopy@xcite have found the presence of a fermi surface ( fs ) gapping at the ho state , which is consistent with the bulk measurement of about 40% loss of quasiparticle weight at this phase.@xcite finally , a shubunikov de - hass ( sdh ) measurement finds that the quantum oscillation frequency does not change by any significant amount in going from ho phase at ambient pressure to the large moment af ( lmaf ) phase at high pressure.@xcite considering also the high - magnetic field phase diagram of uru@xmath0si@xmath0 where the ho phase is intertwined with the lmaf and possibly with other unknown phases,@xcite this sdh data can be interpreted along the same line . therefore , a feasible solution to the ho problem lies in a theory which can explain the ` magnetic dichotomy ' , while accounting for the large fs gapping and entropy change , among others .
according to the fermi - liquid theory , the ho phase transition is associated with a symmetry breaking static order parameter .
numerous evidence of anomalies probed by ins@xcite and spectroscopies@xcite at either @xmath5 and/ or at @xmath6 suggest that ho breaks translational symmetry .
recent evidence from ultrasonic measurement@xcite ( and also from arpes@xcite ) points to the presence of a lattice anomaly ( @xmath8-type lattice symmetry breaking ) at the ho transition , suggesting the involvement of orbital degrees of freedom , in addition to spin . as mentioned before ,
no evidence of time - reversal symmetry breaking is reported to date.@xcite interestingly , a recent magneto - torque measurement has reported the presence of an in - plane rotational symmetry breaking in a single crystal sample , which , however , has been placed on hold until reproduced.@xcite this interesting problem has triggered the development of many novel proposals .
while these proposals are mainly targeted to solve the ho state in uru@xmath0si@xmath0 , they are also interesting in a general ground . among them , many multipolar orders , including quadrupoles,@xcite octupoles,@xcite hexadecapoles@xcite to dotriacontapoles or rank-5 orders@xcite can be highlighted .
however , an interesting group analysis study have shown that such higher - rank order parameters are not allowed in uru@xmath0si@xmath0.@xcite among dipole orders conventional and unconventional spin,@xcite orbital,@xcite and other unconventional orders@xcite have been proposed .
variety of theoretical models aiming to explain the ` nematic ' behavior of the ho state ( if indeed exists ) are given in refs .
@xcite . in a third paradigm ,
the typical concept of hybridization between light and heavy bands has been extended to be modulated with a specific nesting wavevector without spin,@xcite with 1/2 spins in both bands,@xcite and with integer and 1/2 spins in heavy and light bands , respectively@xcite .
however , multiple spectroscopic data including stm,@xcite point - contact tunneling,@xcite and arpes@xcite as well as resistivity , specific heat and magnetic susceptibility@xcite have demonstrated that the hybridization phenomena is insensitive to the ho temperature , and it starts at a much higher temperature ( about 60 - 100 k ) .
this is supported by good consistency of the arpes data in the paramagnetic phase with first - principles band structure calculation , with some discrepancies lying mainly along the @xmath9 direction.@xcite more interestingly , only the hybridization theory@xcite and the present sodw theory assuming incommensurate nesting can only explain the ins dispersion and the spectral weight loss at this wavevector , while no other theory has succeeded in explaining this crucial experimental signature of the ho phase . taken together , a consistent and widespread research activity from both theoretical and experimental perspectives have been channeled in the ho research for three decades , and the recent boost to it has been triggered by the high - quality sample preparation and spectroscopic data .
these high - quality data have been successful in narrowing down the possible origin of the ho state , however no consensus about the nature of ho phase has yet been reached . in present author s opinion ,
the solution is hidden in the ` magnetic dichotomy ' associated with the ho state .
these clues have led to the proposal of the sodw which explains this apparent paradox .
the paper includes a coherent combination of new calculations , a review on the prior works , and also hints of future works on sodw .
the first - principles bandstructure calculation and the low - energy modelling presented in secs . 3 , 4 , and a major part of the discussions in sec .
5 are taken from ref .
@xcite . a substantial overlap between secs . 6 , 8 and ref . @xcite can also be expected .
some parts of the application of sodw in other materials which deal with the arpes results for two - dimensional electron gas with rashba - type soc is taken from ref .
the rest of the calculations and figs . 3 , 4 , 5 , 6(e)-(h ) , 7 , 10 are produced here .
recently , it has been shown that soc can drive new states of matter , including quantum spin - hall effect,@xcite and topological insulator,@xcite without electron - electron interaction via symmetry invariance .
another relevant physics of our interest is the theory of weyl semimetal , originally refereed to ` accidental degeneracy in bands'@xcite , in which multiple relativistic dirac points are scattered in the momentum space ( and thus termed ` semimetal ' ) owing to a broken crystal symmetry , but they remain topologically protected via other subjective symmetry invariance . considering the ` magnetic dichotomy ' of the ho phase which apparently breaks time - reversal symmetry ( magnetic anomaly ) and preserves time - reversal symmetry ( no magnetic moment ) , we introduce the concept of sodw .
it breaks translational symmetry by fs nesting , but respects time - reversal invariance owing to the soc.@xcite while the model is originally applied to uru@xmath0si@xmath0,@xcite it has subsequently been extended to two - dimensional electron gas ( 2deg ) systems in which interface induced rashba - type soc is strong , and finite interaction strength can give rise to the sodw and fs gapping,@xcite consistent with corresponding arpes data.@xcite the essential ingredient for the sodw is the presence of ( at least ) two spin - orbit split bands in the low - energy spectrum , and the fs nesting between them is to be significantly large
. if the soc strength is larger than the coulomb interaction , an interband spin density wave ( sdw ) which breaks time - reversal symmetry is disfavored .
however , above a critical value of the interorbital coulomb strength , a sodw can emerge .
the schematic view of the sodw is shown in figs .
[ fig1](a ) and [ fig1](b ) , in momentum and real - space , respectively . in the @xmath10-space ,
thin solid blue and red lines are two bands , split by soc @xmath11 , while the corresponding dashed lines are the folded bands shifted by a nesting vector @xmath12 .
the sodw order gaps out the ` accidental degenerate ' points on the fs , and the weight of each orbital is mixed between them .
it is evident that sodw can be thought of as two concomitant sdws with opposite spin state at each spatial point , owing to large soc .
for rashba - type soc , the spin - chirality of each band is different to fig .
[ fig1 ] , and thus the low - energy hamiltonian changes slightly ( see sec . [
sec : other ] ) , however , the general concept is the same .
the consistency of the sodw properties with many experimental signatures of the ho phase in uru@xmath0si@xmath0 is appealing , particularly as it reconciles the apparently contradictory magnetic properties of this phase . among them , we highlight several robust features as : ( 1 ) the present order breaks translational symmetry,@xcite and spin - rotational symmetry , but thanks to soc , the order parameter respects time - reversal symmetry and charge conservation at each lattice point .
( 2 ) as a result , no magnetic moment is induced , in agreement with measurements.@xcite ( 3 ) time - reversal symmetry invariance fosters magnetic field to be a perturbation to the order parameter , and the corresponding critical field primarily depends on the gap value and the @xmath13-factor .
experiments validate the vanishing of the ho around @xmath14 t.@xcite .
( 4 ) as a consequence , as field increases the ho gap decreases ( given other parameters such as temperature is constant ) , and thus the resistivity decreases .
in other words , a large magnetoresistance effect is expected within the sodw framework .
( 5 ) finally , we offer a detectable prediction of a second spin-1 collective mode localized inside the gap at @xmath15 in the sodw state which can be probed via electron spin resonance ( esr ) or nuclear quadrupole resonance ( nqr ) at zero magnetic field or polarized neutron scattering measurements .
finally , we envisage to take this model further and apply to laalo@xmath2/srtio@xmath2 ( lao / sto)@xcite and iridium - oxide ( iridate)@xcite where soc and electronic interaction are large and tunable . in these materials , we present evidence of quasiparticle gapping@xcite , and strong reduction of preexisting magnetic moment below critical temperature , magnetic field or carrier density , as seen in uru@xmath0si@xmath0 along the pressure axis .
in order to determine the symmetry properties of the low - lying states , we begin with investigating the _ ab - initio _ ` parent ' band dispersion and the fs of uru@xmath0si@xmath0@xcite in fig . [ fig2 ] . the electronic structure in the vicinity of the fermi level ( @xmath16 ) ( @xmath70.2ev ) is dominated by the 5@xmath17 states of u atom in the entire brillouin zone.@xcite due to the lack of the ` so - called ' itinerant states of @xmath18- or @xmath19-characters within the @xmath7500 mev vicinity of @xmath16 , the typical hybridization process is weak in this system , and is pushed to a higher temperature @xmath2060 - 80 k @xcite .
furthermore , pertaining to larger atomic size of 5@xmath17-electrons in u - atoms , than in the largely localized smaller 4@xmath17-atoms , the inter - atomic electron hopping is larger here and mobile 5@xmath17 electrons stem within the single - electron picture . in other words , the electronic structure of this actinide compound can be described adequately by first - principles calculation , without invoking kondo - like physics .
in fact , the good consistency between the dft calculation@xcite and arpes data@xcite has excluded the presence of @xmath18-electrons within the @xmath7500 mev vicinity of @xmath16 to cause any significant hybridization . owing to the soc and the tetragonal symmetry , the 5@xmath17 states split into the octet @xmath21=@xmath22 ( @xmath23 ) states and the sextet @xmath21=@xmath24 ( @xmath25 ) states .
uru@xmath0si@xmath0 follows a typical band progression in which the @xmath23 bands are pushed upward to the empty states while the @xmath25 states drop to the vicinity of @xmath16 .
the dft band structure , calculated by wien2k software,@xcite is shown by blue solid line in fig .
[ fig2](a ) along several representative high - symmetry momentum directions in the body - centered tetragonal ( bct ) lattice of the crystal .
the corresponding fs topology is depicted in fig .
[ fig2](b ) , overlayed by a shifted fs with the ` hot - spot ' wavevector @xmath26 ( red lines ) in this unit cell notation .
paramount fs nesting is evident here between the two spin - orbit split bands , and hence identified it to be responsible for the sodw state .
the fs nesting property is confirmed by static susceptibility calculation , plotted in the basal plane momentum space , in fig .
[ fig2](c ) , exhibiting a dominant peak at this @xmath12 vector .
the band structure shifted by @xmath12 is then plotted in fig .
[ fig2](a ) in red color which reveals that the low - energy bands are linear in momentum , and the contact points between main- and folded - bands at @xmath16 are in the particle - hole channel along the ( 110 ) direction .
therefore , the nesting condition in this system resembles the so - called ` accidental degenerate points ' or weyl - type dirac cones@xcite .
since the nesting is between different spin - orbit split bands , a typical sdw is prohibited by time - reversal symmetry and the strong soc in this case . on the other hand , a spin - orbit entangled order parameter in the two - particle channel can collectively propagate with alternating sign in the total angular momentum at the wavelength determined by the modulation vector .
this is the guiding instability that drives spontaneous rotational symmetry breaking , while the time - reversal symmetry remains intact .
this is because , both @xmath27 groups for spin and orbital separately are odd under time - reversal , but their product @xmath28 is even .
as the parent state is not a non - trivial topological phase , a gap is opened to lift the fs instability .
motivated by the above - mentioned experimental results and band structure symmetry properties , we write down the effective theory for the sodw using two soc split bands .
we restrict our discussion to the low - lying @xmath25 bands and neglect the unfilled @xmath23 bands . due to the @xmath29-@xmath29-type soc and time - reversal symmetry , the @xmath25 atomic states consist of three doublets , characterized by up and down ` pseudospins ' : @xmath30=@xmath7@xmath24 , @xmath7@xmath31 , @xmath7@xmath32 , where @xmath30 is the @xmath33 component of @xmath21 .
although we consider all relevant bands that cross @xmath16 directly from the dft results for the numerical calculations , however , to explicate the physical mechanism of sodw , and the role of soc in it , we start from a two - orbital basis with atomistic soc , @xmath11 .
if no other symmetry is broken , the other band ( mostly of @xmath21=@xmath24 character ) remains ungapped and thus not considered in the discussion .
we , however , note that a nesting between @xmath21=@xmath24 and @xmath21=@xmath31 states as reported in ref .
@xcite , can also lead to sodw at a different interaction strength .
the purpose here is to derive a general hamiltonian of such kind , in which no particular assumption of the orbital symmetry and the relevant details of the soc are imposed .
our starting noninteracting hamiltonian is thus @xmath34 , \label{h0}\ ] ] where @xmath35 is the electron creation operator in the @xmath36 orbital with spin @xmath37 , and @xmath38 is its corresponding dispersion spectrum in a translational invariant crystal of bloch momentum @xmath10 . with a translational symmetry breaking at the reduced reciprocal wavevector or the `` hot - spot '' vector @xmath12 , the new spin - orbit basis in the nambu - space yields @xmath39 .
the spin - polarized nambu operator is an important consideration which distinguishes the present sodw order parameter from an inter - orbital sdw , and promotes a _ zero _ magnetic moment as will be evaluated rigorously below . in this basis , we express our interacting hamiltonian in a matrix form@xcite as @xmath40 the sodw field parameter @xmath41 is taken to be complex for generality and will be defined below .
this sodw hamiltonian ( [ hint ] ) is very similar to the one proposed for the rashba - type soc system,(see sec .
[ sec : other ] ) and also for 5@xmath18 system iridates.@xcite also , it is interesting to compare the hamiltonian with that for the topological insulator ( eq . 1 of ref
. @xcite ) in which we replace the two - particle gap term @xmath41 by their hybridization term .
we diagonalize the above hamiltonian in two steps by the bogolyubov method .
first , we define the bogolyubov spinor for the soc term as @xmath42 after this basis transformation , the hamiltonian in eq .
[ h0 ] becomes diagonal with eigenstates @xmath43 , and the corresponding spin - orbit helical weight is @xmath44 , where @xmath45 , and @xmath46 .
the subsequent diagonalization onto the nambu - basis on the reduced brillouin zone follows similarly , and the spin and orbital notations are retained explicitly . here the bogolyubov basis transformation is @xmath47 where @xmath48 are the band indices , and @xmath49 is the spin - orbit helicity index which changes sign when orbitals are interchanged . here
the canonical density operators @xmath50 take equivalent forms as before , and the corresponding quasiparticle states are replaced by @xmath51 , and @xmath52 . a schematic view of the sodw split bands and the shared orbital character , determined by coherence factors @xmath53 between them is given in fig .
[ fig1](a ) .
it is interesting to notice that soc allows the many - particle wavefunction to have a dynamical spin flip on the same orbital at the same momentum as indicated by vertical arrow , which governs a collective @xmath54 mode as derived below .
it is worthwhile to combine the above two steps to visualize how the four - vector sodw nambu operator transforms under translational symmetry breaking while retaining time - reversal symmetry : @xmath55 an immediate ansatz emerges from the above transformation @xmath56 guided by the same symmetry of the hamiltonian in eq .
[ hint ] @xmath56 that the system is time - reversal invariant under the representation of this symmetry @xmath57 , where @xmath58 is the complex conjugation , @xmath59 is the 2@xmath602 identity matrix , and @xmath61 is the second pauli matrix . under time - reversal ,
the hamiltonian transforms as @xmath62 .
this important symmetry consideration renders a _ zero _ magnetic moment as depicted schematically in fig .
[ fig2]b and is evaluated in eq .
( [ magmoment ] ) below . in a multiorbital setup ,
the full interacting hamiltonian includes intra- and interorbital coulomb interactions , @xmath63 and @xmath64 , hund s coupling @xmath21 , and pair - exchange term @xmath65 : @xmath66 . \label{rpaham}\end{aligned}\ ] ] where @xmath67 .
the linearized equations for the two order parameters sdw and sodw are presented graphically in fig .
[ fig4 ] which leads to the self - consistent equations as @xmath68 where @xmath69=`sdw ' , and ` sodw ' . for a given order
, we consider all possible contractions of the interaction term written in eq . 2 , which lead to the following vertices @xmath70 the corresponding susceptibilities in the particle - hole channel
are then defined as @xmath71 it is interesting to see that hund s coupling does not contribute to the sodw , while the pair scattering term does .
however , in many materials , this term has very small value and can be neglected for some practical purposes .
the vertices in eqs .
( [ sceqn ] ) are different here from a typical multiband system ( see , for example , ref .
@xcite ) due to the presence of soc . based on eqs . [ sceqn ] and [ chi ]
, we can deduce a general phase diagram in fig .
[ fig4 ] for the competition and coexistence of sodw and sodw . for a single orbital system ,
only an intraband nesting @xmath72 is present and thus only sdw is allowed .
when soc is included , the inter - orbital susceptibility @xmath73 is turned on .
the relative competition between these two nestings determines the dominant instability to win in a given band structure . at the limit of @xmath74 ,
when soc @xmath75 , no sodw is possible , similarly , as @xmath76 , no sdw is possible from interorbital susceptibility ( since @xmath77 is also imposed ) . for large @xmath64 , i.e. small @xmath78 , sdw is more stable for finite @xmath11 . when @xmath79 ( assuming both intra- and interorbital nesting are comparable ) , two phases can coexist . from eq .
[ sceqn ] , it is also easy to deduce that while inter - orbital susceptibility helps sdw via hund s coupling @xmath21 , no intra - orbital term is involved in the sodw phase . if there is no orbital fluctuation present in the system , for the two fully spin - polarized bands , the magnetic moment along any direction vanishes
this is achievable in materials with strong soc as in uru@xmath0si@xmath0 in which @xmath801 ev .
this can be verified by a trivial analytical computation of the spin - operator : @xmath81 = 0 . \label{magmoment}\end{aligned}\ ] ] magnetic moment vanishes in each band for any value of the interaction strength , owing to the presence of two opposite sdw coupled by soc .
we should carefully distinguish the induced magnetic moment of sodw in eq .
( [ magmoment ] ) , from a sdw terms ( a finite gap term to be placed in the ` 13 ' component of hamiltonian in eq .
( [ hint ] ) which has of course different interaction term as discussed in eq .
[ sceqn ] ) .
our conclusion of zero moment is only valid when no sdw term is explicitly present .
finally , the sodw order parameter can then be written in an explicit form as @xmath82\left[f(\tilde{e}^+_{\bf k})-f(\tilde{e}^-_{\bf k})\right ] , \label{sodw}\end{aligned}\ ] ] where the fermion occupation number is defined as @xmath83 and so on , and @xmath84 are the degenerate eigenstates of the interaction hamiltonian in eq .
the notation ` prime ' over the momentum summation indicates that the summation is restricted within the reduced brillouin zone .
the complex gap parameter that enters in the hamiltonian is @xmath85 . for the experimental value of gap amplitude @xmath86 mev
, we estimate the critical inter - orbital interaction strength to be @xmath87 ev which is a reasonable number for 5@xmath17 electrons estimated earlier.@xcite a fascinating property of the ho state in uru@xmath0si@xmath0 is that it transforms into the lmaf phase via a first - order phase transition as pressure ( and also magnetic field ) is increased .
sdh oscillation measurement@xcite has demonstrated that the angle - dependence of the oscillation frequency , which is taken to be proportional to the closed fs areas at finite magnetic field , does not change as we go from the ho phase to the lmaf phase via increasing pressure .
this observation suggests that both ho and lmaf phases coexist in uru@xmath0si@xmath0 , at least at high magnetic field .
based on this result , we study the stability of the two phases , attributing the lmaf phase as a sdw state , within a ginzburg - landau ( gl ) treatment by integrating out the diagonal and the soc terms ( single - particle terms ) . here
we denote @xmath88 , and @xmath89 for the sodw and sdw order parameters .
the gl free energy can be expressed as @xmath90 = -t\sum_{{\bf k},n}{\rm ln}~{\rm det}g^{-1}_{{\bf k},n } + \frac{|\delta_{\rm sodw}|^2}{v } + \frac{|\delta_{\rm sdw}|^2}{u},\nonumber\\ \label{glf0}\end{aligned}\ ] ] where the inverse green s function in the nambu - grassmann space is @xmath91 , with @xmath92 is the matsubara frequency for the fermions , and @xmath93 is the identity matrix .
expanding eq .
[ glf0 ] up to the fourth order in the gap fields , we obtain the essential gl functional for a homogeneous system of real fields as @xmath90 = \sum_{i={\rm sdw , sodw}}\left[\alpha_i|\delta_i|^2+\beta_i|\delta_i|^4\right]+ \gamma|\delta_{\rm sodw}|^2|\delta_{\rm sdw}|^2 .
\label{glf}\end{aligned}\ ] ] all expansion coefficients @xmath94 , @xmath95 , and @xmath96 are shown in fig . [ fig5 ] which lead to @xmath97 for simplicity the sdw order parameter for both states
are assumed to be the same .
the two , four and six order susceptibilities @xmath98 are defined in the band basis as @xmath99 here the green s functions @xmath13 denote the diagonal components of the full green s function matrix as @xmath100 .
the phase transition for each order parameter is determined as usual by the corresponding quadratic terms as @xmath101 , where @xmath102 are the transition temperatures for them without phase competition . for multiple competing order parameters ,
a general criterion was derived before in various contexts@xcite that both phases can share a uniform coexistence region if @xmath103 .
whereas for @xmath104 , the phase fluctuation is sufficiently strong that both phases are separated by a first - order phase transition . given the experimental fact in uru@xmath0si@xmath0 that ho and lmaf phases share a first - order phase boundary as a function of pressure , the second criterion is expected in this material
. however , the numerical calculations of the susceptibilities and the effects of pressure are beyond the scope of the present paper .
we give two representative results of the electronic and magnetic structure of the sodw in the ho state and compare them with arpes and ins data , respectively .
we first evaluate the electronic signature of ho gap opening .
the sodw order induced ho state is evaluated by using dft - based band structure as the input frame of reference , and downfolding the full dft results in the reduced space near @xmath16 with the ` hot - spot ' wavevector @xmath26 in the bct crystal structure .
figure [ fig6 ] gives single - particle maps along several representative momentum cuts , and the fs topology before and after the ho transition .
gapping of the fs is clearly visible for the two bands aligned along the ( 110 ) direction at the degenerate dirac points at @xmath16 , introduced in fig .
[ fig1](a ) .
this gapping process truncates the paramagnetic fs into small pockets aligned along the bond - direction .
we require to pay particular attention to the location of the fs pockets in this system , since the original crystal structure is bct ( not simple tetragonal which is often used for the simplification of computation ) . in this unit cell notation , the location of the gapping and fs pockets governed in the sodw state are in direct agreement with the arpes spectral function map on the fs.@xcite we find that the loci of the gapped states move away from @xmath16 as we increase the @xmath9 value .
arpes has also reported an interesting result that a conduction band collapses below @xmath16 at the @xmath105-point upon entering into the ho state.@xcite this unusual behavior is also reproduced within the present framework , without invoking the previously - thought hybridization phenomena . as mentioned before , the optimum fs nesting occurs at the @xmath106-plane , and beyond it , the same nesting condition moves to empty states as the @xmath105-centered electron pocket gradually shrinks .
this is the manifestation of the itinerant nature of the ho gap which generates an anisotropic gap on the electronic structure , and can also induce momentum dependence in the @xmath13-factor measured in quantum oscillation measurements.@xcite we locate the @xmath107 plane where the bottom of the paramagnetic electron pocket resides just above @xmath16 , and upon the ho gap opening in the empty state , this pocket drops into the fermi sea .
this result , shown in the bottom row of fig .
[ fig6 ] , agrees well with arpes measurements.@xcite since arpes can probe different @xmath9-planes with the variation of incident photon energy , our theoretical result is consistent with the experimental fact that such phenomena is observed at a finite photon energy of @xmath108 21.2 ev .
our explanation can further be confirmed by observing the possible flickering presence of the @xmath105-centered electron pocket as a function of the photon energy .
the corresponding density of states ( dos ) before and after the ho transition are shown in fig .
[ fig7 ] . without invoking any hybridization ,
we see two flat band - like peaks ( or the van - hove singularities ) below and above @xmath16 ( which are pushed further towards lower energy by renormalization ) .
the ho gap opens around @xmath16 as seen from the _ inset _ box in fig .
[ fig7 ] and the corresponding spectral weight shifts to the states at a higher energy below @xmath16 .
the results are consistent with stm measurements.@xcite in this context , we note an interesting consequence of the two flat - bands in the particle - hole channel .
as shown in pu-115 and u-115 families,@xcite such particle - hole channel , triggered by soc , causes large dynamical spin and pseudospin fluctuations which dress the electronic states in the intermediate coupling regime .
a trademark feature of the intermediate coupling scenario is the duality between renormalized itinerant quasiparticle states and localized high energy states in the same or different orbitals , separated by the ( pseudo- ) spin fluctuation energy scale , which is seen in arpes.@xcite the pseudospin fluctuation is also reported in a recent nmr study in this system.@xcite this is a future problem to study in uru@xmath0si@xmath0 .
the entropy release of the ho transition can be easily evaluated from the free - energy estimation .
the free - energy of the ho state is deduced from @xmath109 , where @xmath110 is boltzmann constant , @xmath111 is number of filled states , and @xmath112 is the chemical potential . in the diagonal basis with a mean - field order , the free energy translates into @xmath113+n(\mu+|\delta|^2/v)$ ] . and the corresponding entropy release at the ho transition is evaluated to be @xmath114 , which is close to its experimental estimate from the specific heat jump.@xcite now we present the results of the spin - excitation spectrum as a function of both momentum and energy .
ins measures the imaginary part of the spin - susceptibility @xmath56 modulo materials specific form factor @xmath56 which is theoretically computed from eq . ( [ chi ] ) . in the sodw state ,
the susceptibility is a 2@xmath602 matrix ( to be precise it is a 4@xmath604 matrix including both intra and interband contributions separately as is done in the main computation ) , made of direct transitions ( in the diagonal terms ) and umklapp scattering terms placed in off - diagonals@xcite : @xmath115 here the direct term @xmath116 means that the scattering between the two main bands and between the two shadow bands while the umklapp scattering term @xmath117 arises from the scattering betwee then main and shadow bands .
all scattering processes included here are in the particle - hole channel .
although , the mechanism and structure of the ins results presented in fig . [ fig7 ] are tied mainly to the details of the electronic structure and the ho ` hot - spot ' wavevector embedded within this bare susceptibility , however , for completeness we invoke the many - body correction within the random - phase approximation ( rpa ) .
rpa calculation is justified in this heavy - fermion systems at low-@xmath118 where itinerant electronic structure is well - established .
the rpa interaction hamiltonian is same as given in eq .
[ rpaham ] , with @xmath74 .
different bandwidths of different orbitals amount to different critical values of @xmath63 , and @xmath64 , however , to justify that the final results are parameter free , we take the lowest critical value of @xmath63=1 ev and @xmath64=0.6 ev which shifts the energy scale of all the excitation mode slightly to a lower energy .
figure [ fig8 ] presents the computed direct and umklapp spin excitation spectra in ( a ) and ( c ) , respectively , along the zone diagonal direction , and compares with the corresponding experimental data@xcite obtained along the zone boundary direction given in fig .
[ fig8](b ) . for direct transition in the particle - hole continuum ( trace of the susceptibility matrix ) , we clearly mark a prominent @xmath54 collective mode with a dispersion that resembles gapped goldstone mode or a roton - like spectrum . for the case of the discrete symmetry breaking due to the complex order parameter @xmath119 , such gapping of the goldstone mode
can bee justified .
the mode is localized around the @xmath120 value , slightly shifted from the original ho wavevector @xmath12 due to band structure effect . however , the shift of the mode below @xmath121 is a many - body correction as discussed above . to match the mode energy of the experimental data about @xmath122 5 mev , a gap amplitude of @xmath123 mev
is invoked in our calculation which is close to the spectroscopic value of the ho gap.@xcite although a static neutron signal is reported in ref .
@xcite at @xmath124 , but in the inelastic spectrum , both experiment@xcite and our theory find a weaker and featureless intensity at this wavevector . by construction ,
the sodw is associated with an interaction induced spin - orbit entangled electronic structure , rendering conceptual similarity with the spin - orbit order in the particle - particle channel proposed by leggett@xcite for the liquid @xmath125he superconducting phase , dynamical generation of soc,@xcite , or the single - particle quantum spin - hall state@xcite which do not break time - reversal symmetry . in
what follows , the umklapp scattering process between different spin - orbital states brings out two zeeman - like spin - split states at the same momentum * k * , on both sides of the ho gap with coherence factors @xmath53 , without any external magnetic field , as discussed in fig .
[ fig1](a ) .
therefore , a second spin - flip collective mode is expected in the off - diagonal susceptibility in eq .
( [ chisodw ] ) as @xmath126 to localize inside the sodw gap @xmath127 around @xmath15 .
indeed , our computation confirms the existence of this collective mode at energy @xmath128 , as shown in fig .
[ fig8](c ) . since the intensity of this mode is about an order of magnitude lower than that of the @xmath12-mode presented in fig .
[ fig8](a ) , it will be difficult to simultaneously detect them together in the ins measurement
. however , esr or nqr , having the capability of detecting lineshifts of resonance without the application of a static magnetic field , will be able to measure our proposed @xmath129 mode .
the present spin - orbit order is not only topologically protected by time - reversal symmetry , but most importantly it is thermodynamically shielded with a condensation energy equal to the quasiparticle gap . the ho gap is insensitive to any time - reversal invariant perturbation such as pressure ( with sufficient pressure the ho transforms into the lmaf phase ) , while time - reversal breaking perturbation such as magnetic field will destroy the order . remarkably , these are the hallmark features of the ho states,@xcite which find a natural explanation within our sodw order scenario . in
what follows , the magnetic field will destroy the ho state even at @xmath130 k , that means at a quantum critical point ( qcp ) as the ho is a spontaneously broken symmetry phase .
however , due to the finite gap opening at the ho state , it requires finite field value to destroy the order .
the thermodynamical critical field can be approximately determined by the critical zeeman term @xmath131 ( @xmath13 is ` g'-factor , @xmath132 is bohr magneton ) which can overcome the condensation gap energy ( i.e. @xmath133 ) . assuming a mean - field like magnetic field dependence of the sodw gap @xmath41
, we obtain an approximate phase diagram of the ho state as a function of @xmath134 .
the critical field is obtained to be @xmath135 = 43 t for an isotropic @xmath13-factor of @xmath201 for 5@xmath17 electrons and zero field gap of @xmath136 5 mev .
the result is shown in fig .
[ fig9 ] . experimental value of @xmath135 is 35 t.@xcite ho interaction can increase the value of the @xmath13-factor which will reduce @xmath135 .
in fact , a sdh oscillation measurement reports the presence of an anisotropic @xmath13-factor in the ho state,@xcite which whether can be reproduced by the sodw order or not is to be explored in a future study . the present model can not deduce the phase diagram for the superconducting ( sc ) state , possibly intervening the ho state , or any other phases that may arise around or above the qcp.@xcite however , from the study of sec .
[ sec : holmaf ] , we can expect the appearance of a sdw phase near this phase boundary .
experimentally , it is known that the second order phase transition at low field transforms into a first order one at high field , which may play a role in eliminating the qcp of the sodw state in this system .
however , the qcp as a function of magnetic field is a general result for sodw and can thus be searched in other materials . in this context
, we note an interesting result in the laalo@xmath2/srtio@xmath2 ( lao / sto ) interface.@xcite this interface is known to host local ferromagnetism and 2deg state .
it is shown in a recent magnetoresistance and anomalous hall effect experiment that below a critical value of the magnetic field , the magnetic polarization is dramatically reduced while resistivity sharply rises , see fig .
[ fig10](c ) .
this unique phenomenon indicates the emergence of a new phase which onsets below a critical field and quenches magnetic moment while reduces the electric conductivity via fs gapping .
these are the key properties of the sodw .
similar effect of large magnetoresistance is also expected in uru@xmath0si@xmath0 .
finally we discuss the possible evidence of the sodw in other systems .
we recall that sodw quenches any preexisting magnetic moment .
this is evident in the uru@xmath0si@xmath0 phase diagram as a function of pressure , see fig .
[ fig10](a ) .
pressure increases the noninteracting bandwidth ( @xmath137 ) , and thus the ratio of @xmath138 decreases , and the possibility of the existence of sodw is reduced while that of the sdw increases , as illustrated in fig .
[ fig4 ] .
a very similar behavior of the magnetic moment is observed as a function of @xmath118 for a typical bilayer iridate sr@xmath2ir@xmath0o@xmath139@xcite and in the lao / sto interface@xcite as shown in figs .
[ fig10](b ) and ( c ) , respectively .
iridates exhibit metal - insulator transition ( mit ) promoted by strong soc of 5@xmath18 orbitals,@xcite but no consensus on the origin of mit has emerged yet . for this given system ,
an antiferromagnetic ( af ) order sets in below 280 k for both field cooled ( fc ) and zero field cooled ( zfc ) samples , however , the resistivity shows a power - law temperature dependence below @xmath140 ( a behavior expected for a semi - metallic phase).@xcite the resistivity exhibits an abrupt exponential rise which marks the onset of the mit below @xmath141 70 k , and simultaneously the magnetic moment begins to drop .
the new experimental data in the lao / sto interface which has strong rashba - type soc , shows that below a critical magnetic field and carrier density , the preexisting ferromagnetic order is quenched , while the resistivity increases , signaling the onset of a new non - magnetic order parameter , as shown in fig .
[ fig10](c).@xcite based on the similarity of the phase diagram of magnetic moment of uru@xmath0si@xmath0 , sr@xmath2ir@xmath0o@xmath139 , and lao / sto , we envisage that a sodw phase is also present in these materials.@xcite another motivation for the search of sodw in 2deg is the arpes evidence of a large quasiparticle gapping on the thin film of biag@xmath0 even when the spin - degenerate point of the rashba - split bands remain ungapped.@xcite to study the emergence of sodw in this case , we consider a system of two - component fermi gas in the presence of rashba - type soc , which can be represented in the two component fermion field @xmath142^{t}$ ] as @xmath143\psi_{\bm k}$ ] . here
@xmath144 is the free - fermion dispersion term , modeled by nearest - neighbor electronic hopping @xmath145 as @xmath146-e_f$ ] , where @xmath16 is the chemical potential .
the second term is the 2d lattice generalization of the standard rashba soc term @xmath147 , with @xmath148 being the pauli matrices and @xmath149 being the soc strength .
the helical dispersion spectrum of hamiltonian @xmath150 is @xmath151^{1/2}$ ] .
the corresponding fs exhibits paramount nesting instability at @xmath152 as shown in ref .
@xcite . in this translational symmetry broking case
, we define the the nambu - gorkov spinor as @xmath153^t$ ] , where @xmath154 corresponds to four - different @xmath155-vectors . in this basis , the auxiliary sodw field @xmath156_{\nu\nu^{\prime}}\psi_{{\bm k},\nu^{\prime}}$ ] , where @xmath64 is the contact potential .
correspondingly , the inverse single - particle green s function in the fermionic matsubara frequency is obtained as @xmath157 the other terms belong to the three values of @xmath158 . at the mean - field level
, @xmath159 represents the gap parameter which is same for all values of @xmath155 . as in the case of the hamiltonian for uru@xmath0si@xmath0 in eq . [ hint ]
, the above hamiltonian also has resemblance with the hamiltonian for the topological insulator in heterostructure of rashba - type soc 2deg proposed in ref .
@xcite in which the sodw term is replaced by the interlayer tunneling amplitude . taking the trace of the above green s function ,
we obtain the excitation spectrum as @xmath160^{1/2},\nonumber\\ & { \rm with}\hspace{0.15cm}s^{\nu}_{\pm}({\bm k})=(e_{{\bm k},0}^{\nu}\pm e_{{\bm k},0}^{\bar{\nu}})/2 , \label{eq : eig}\end{aligned}\ ] ] where @xmath161 is the helical index and @xmath162 is the split band index due to translational symmetry breaking . for a self - consistent value of @xmath163310 mev , we obtain a quasiparticle gap of @xmath136120 mev , which agrees well with the gap amplitude measure for 2 monolayers of biag@xmath0.@xcite . experimentally it was also demonstrated that the quasiparticle gap is tunable by changing the number of layers of biag@xmath0 layer which we modeled by changing interaction parameter @xmath64 , while all other parameters including rashba soc remain constant .
we plot the corresponding calculated band dispersion on top of the experimental dispersion in fig .
[ fig11 ] , and find good correspondence between the theory and experiment . in both experiment and theory ,
two key features should be noted here .
we notice the presence of shadow bands ( or reconstructed bands due to translational symmetry breaking ) which convey the fact that the quasiparticle gap is not a single - particle gap . furthermore , since the kramer s degeneracy at the @xmath105-momentum point remains ungapped , this particular phase is not a sdw .
such a reconstruction of fs and fs nesting has also been documented in a quantum wire of atomic pb grown on si(557 ) having strong soc spitting.@xcite our sodw model for the rashba - type soc which is a generic feature arising in many interfaces and surface states ( sometimes in bulk states as well ) establishes that the sodw phase may be present in many systems other than uru@xmath0si@xmath0 .
the similarity of the hamiltonian for the sodw in uru@xmath0si@xmath0 in eq .
[ hint ] and rashba - type sodw in eq .
[ greenfn ] with that of the non - interacting ti in refs .
@xcite motivates us to evaluate the topological invariant index of the interacting hamiltonian in eq .
[ hint ] . to characterize the topological phenomena
, we revisit the fu - kane classification scheme@xcite which implies that if a time - reversal invariant system possess an odd value of @xmath164 invariant index , then the system is guaranteed to be topologically non - trivial .
@xmath164 index is evaluated by the time - reversal invariant index @xmath165 , if defined , for all filled bands as @xmath166 , where @xmath69 is the total number of orbitals in the fermi sea
. a more efficient method of determining the topological phase is called the adiabatic transformation scheme used earlier in realizing a large class of topological systems , especially when @xmath164 calculation is difficult.@xcite in this method , the non - trivial topological phase of a system can be realized by comparing its band - progression with respect to an equivalent trivial topological system .
uru@xmath0si@xmath0 is topologically trivial above the ho state , i.e. @xmath167 . the gap opening makes the top of the valence band ( odd parity ) to drop below @xmath16 as shown in fig .
[ fig6](b ) .
thereby , an odd parity gained in the occupied level endows the system to a non - trivial topological metal . to see that we evaluate the topological index for the ho term as @xmath168 , where the corresponding berry curvature can be written in terms of @xmath169-vector as @xmath170 with @xmath171 .
the components of the @xmath169 vector is obtained from the sodw order parameter given in eq .
( [ sodw ] ) . when the @xmath169 vector has a odd parity symmetry , it is easy to show that @xmath172 which makes the total @xmath164 value of the ho phase to be @xmath173 , and hence we prove that the ho gap is a topologically non - trivial phase .
the consequence of a topological bulk gap is the presence of surface states.@xcite in our present model , we expect two surface states of opposite spin connecting different orbitals inside the ho gap . as the system is a weak topological semimetal , the surface states are unlikely
to be topologically protected .
the spin - orbit locking of these states can be probed by arpes using circular polarized incident photon which will be a crucial test of this postulate .
furthermore , the topological nature of the ho state can lead to non - trivial ground state in the lmaf and sc state , if they uniformly coexist .
for example , topological ho with lmaf phase may host non - abelian axion particle , or in the sc state may give rise to a topological sc.@xcite
to conclude , in this article we presented detailed analytical calculation and numerical results of sodw as a candidate proposal of the long - sought ho phase of the heavy - fermion metal uru@xmath0si@xmath0 .
the proposal of the sodw is motivated by the ` magnetic dichotomy ' phenomena at this phase transition as well as the fs instability between two spin - orbit split bands as deduced consistently by dft calculation and arpes results .
dft calculation demonstrated the presence of a logarithmic - like divergent susceptibility at @xmath174 in the bct crystal structure which connects inter - orbital fermi momenta . as a result ,
a density wave originates which involves both spin and orbital flips at the same time between two in - plane sublattices of u atoms which is thus termed as sodw .
the corresponding sodw gap opens in the electronic and magnetic spectrum , but no static magnetic moment is induced due to time - reversal invariance imposed by soc .
we formulated a general two orbital model hamiltonian including soc .
the similarity of the sodw hamiltonian ( eq . [ hint ] ) with that of the three - dimensional topological insulator@xcite is noted .
the numerical results were deduced by including dft bandstructure for the noninteracting bands .
we found that fs gapping occurs along the zone diagonal direction in the bct crystal , in agreement with arpes data.@xcite we also computed the spin - excitation dispersion which showed a gapped upward dispersion at the incommensurate wavevector , in agreement with ins data.@xcite finally , we showed that the sodw order yields a second resonance mode at @xmath1750 , which can be probed by polarized ins or esr experiment at zero magnetic field .
most importantly , despite dynamical magnetic signal , the static magnetic moment was shown analytically to be zero .
the feedback effect of the time - reversal invariance can be tested by applying magnetic field to the ho state . due to symmetry reason
, the ho can be destroyed by magnetic field , but to overcome the finite gap opening , a finite magnetic field is required to compensate for the ho gap at zero temperature .
for this reason , sodw predicts the presence of a qcp along the field axis . with a rough estimation of the critical field ( assuming the zeeman energy to be equal to the gap energy )
, we found a good agreement with experiment.@xcite we also gave evidence of the presence of sodw in other soc systems such as iridates , biag@xmath0 thin film , and lao / sto interface .
for biag@xmath0 thin film , arpes exhibits the presence of unusual quasiparticle gapping and shadow bands while the spin - degenerate point at @xmath105-momentum is ungapped.@xcite these results are not reproduced by corresponding dft band structure .
we extended the sodw concept to the rashba - type soc bandstructure and explained the nature of this unusual gapping phenomena.@xcite similarly , in iridates@xcite and lao / sto interface,@xcite we found that a preexisting magnetic moment is quenched below a characteristic temperature and magnetic field , respectively , indicating the presence of an unusual phase .
more interestingly , for iridates , it was also shown that a metal - insulator transition occurs at the same temperature where the magnetic moment decreases , suggesting that magnetic ordering is unlikely to be responsible for the mit behavior in this compound .
due to the presence of soc and the possible presence of fs nesting , we envisioned to explain these unusual properties by sodw .
however , no theoretical calculation for iridates and lao / sto were presented in this article . taken together ,
sodw was shown to be a new state of matter which breaks translational symmetry , but not time - reversal symmetry .
this is why sodw is fundamentally distinct from an inter - orbital sdw .
also the contributing interaction terms for sodw are inter - orbital interaction @xmath64 and pair interaction @xmath65 , as given in eq .
( [ sceqn ] ) , while for the inter - orbital sdw the pair scattering is replaced by the hund s coupling @xmath21 .
in the present sodw model both contributing orbitals are itinerant and they are split by soc . finally , we comment on two future research directions to further explore the nature of sodw in uru@xmath0si@xmath0 .
( 1 ) the competition between ho and lmaf phase as a function of pressure has remained unexplained . as shown in the phase diagram in fig .
[ fig9 ] , the ho phase at ambient pressure goes through a first order phase transition around 0.5 gpa , and at higher pressure it coexists and/ or competes with the lmaf phase . with pressure , however , the ho transition temperature increases , rather than decreases . to explain this phenomenon , our strategy is to study the competition between sdw and sodw .
we expect that as pressure increases , the effective non - interacting bandwidth increases
. therefore , @xmath138 decreases which , according to fig .
[ fig4 ] weakens sodw , and favors sdw .
the corresponding ginzburg - landau criterion for this problem is deduced in sec
( 2 ) as magnetic field is expected to reduce the sodw order for all systems due to time - reversal invariance , one would expect to gain more quasiparticle state on the fs with increasing field at a fixed temperature . at @xmath130 the resistivity of a system
is determined by the quasiparticle scattering on the fs without the inclusion of scattering lifetime or the imaginary part of the self - energy , therefore , with decreasing field , one would expect a larger fs gapping and thus larger resistivity .
this is seen in lao / sto interface@xcite and in iridates,@xcite ; and we expect to see similar effect in uru@xmath0si@xmath0 .
the author acknowledges valuable discussion with a. leggett , j. mydosh , p. coleman , m. j. graf , p. wolfle , j .- x .
zhu , j. haraldsen , p. riseborough , a. f. santander - syro .
the author also expresses gratitude to a. v. chubukov and a. rahmanian for discussion of sodw formalism in iridates .
the work is supported by the u.s .
doe through the office of science ( bes ) and the ldrd program and faciliated by nersc computing allocation .
ramirez , p. coleman , p. chandra , e. brck , a.a .
menovsky , z. fisk , and e. bucher , phys .
68 ( 1992 ) p. 2680 . c. broholm , h. lin , p.t .
matthews , t.e .
mason , w.j.l .
buyers , m.f .
collins , a.a .
menovsky , j.a .
mydosh , and j.k .
kjems , phys .
b 43 ( 1991 ) p. 12809 .
y. j. jo , l. balicas , c. capan , k. behnia , p. lejay , j. flouquet , j. a. mydosh , and p. schlottmann , phys .
98 ( 2007 ) p. 166404 .
p. das , r.e .
baumbach , k. huang , m.b .
maple , y. zhao , j.s .
helton , j.w .
lynn , e.d .
bauer , and m. janoschek , new j. phys 15 ( 2013 ) p. 053031 .
a.villaume , f. bourdarot , e. hassinger , s. raymond , v. taufour , d. aoki , and j. flouquet , phys .
b 78 ( 2008 ) p. 012504 .
c. r. wiebe , j. a. janik , g. j. macdougall , g. m. luke , j. d. garrett , h. d. zhou , y .- j .
jo , l. balicas , y. qiu , j. r. d. copley , z. yamani , and w. j. l. buyers , nat .
phys . 3 ( 2007 )
a. f. santander - syro , m. klein , f. l. boariu , a. nuber , p. lejay , and f. reinert , nat .
phys . 5 ( 2009 )
r. yoshida , y. nakamura , m. fukui , y. haga , e. yamamoto , y. nuki , m. okawa , s. shin , m. hirai , y. muraoka , and t. yokoya , phys .
b 82 ( 2010 ) p. 205108 .
s. chatterjee , j. trinckauf , t. hanke , d. e. shai , j. w. harter , t. j. williams , g. m. luke , k. m. shen , and j. geck , phys .
( 2013 ) p. 186401 .
s. chatterjee , j. trinckauf , t. hnke , d. e. shai , j. w. harter , t. j. williams , g. m. luke , k. m. shen , and j. geck phys .
( 2013 ) p. 186401 .
meng , p. m. oppeneer , j. a. mydosh , p. s. riseborough , k. gofryk , j. j. joyce , e. d. bauer , y. li , and t. durakiewicz , phys .
( 2013 ) p. 127002 .
a. r. schmidt , m. h. hamidian , p. wahl , f. meier , a. v. balatsky , j. d. garrett , t. j. williams , g. m. luke , and j. c. davis , nature 465 ( 2010 ) p. 570 - 576 .
p. aynajian , e. h. da silva neto , c. v. parker , y. huang , a. pasupathy , j.mydosh , and a. yazdani , proc .
usa 107 ( 2010 ) p. 10383 .
w. k. park , p. h. tobash , f. ronning , e. d. bauer , j. l. sarrao , j. d. thompson , and l. h. greene , phys .
( 2012 ) p. 246403 .
m. b. maple , j. w. chen , y. dalichaouch , t. kohara , c. rossel , m. s. torikachvili , m. w. mcelfresh , and j. d. thompson , phys .
56 ( 1986 ) p. 185
e. hassinger , g. knebel , t.d .
matsuda , d. aoki , v. taufour , and j. flouquet , phys .
105 , ( 2010 ) p. 216409 .
k. h. kim , n. harrison , m. jaime , g. s. boebingera , and j. mydosh , phys .
( 2003 ) p. 256401 .
j. levallois , k. behnia , j. flouquet , p. lejay , and c. proust , europhys .
85 ( 2009 ) p. 27003
. b. wolf , w. sixl , r. graf , d. finsterbusch , g. bruls , b. liithi , e. a. knetsch , a. a. menovsky , and j. a. mydosh , j. low temp .
94 ( 1994 ) p. 307 .
t. yanagisawa , s. mombetsu , h. hidaka , h. amitsuka , m. akatsu , s. yasin , s. zherlitsyn , j. wosnitza , k. huang , and m. brian maple , j. phys .
82 ( 2012 ) p. 013601 .
a. f. santander - syro , private communication .
r. okazaki , t. shibauchi , h. j. shi , y. haga , t. d. matsuda , e. yamamoto , y. onuki , h. ikeda , and y. matsuda , science 331 ( 2011 ) p. 439 .
p. santini , phys .
b 57 ( 1998 ) p. 5191 - 5199 .
a. tsuruta , a. kobayashi , t. matsuura , and y. kuroda , j. phys .
69 ( 2000 ) p. 663
h. harima , k. miyake , and j. flouquet , j. phys .
79 ( 2010 ) p. 033705 .
k. hanzawa , j. phys . : condens .
matter 19 ( 2007 ) p. 072202 .
k. haule , and g. kotliar , nat .
phys . 5 ( 2009 )
p. 796 - 799 .
h. kusonose , and h. harima , j. phys .
80 ( 2011 ) p. 084702 .
j. g. rau , and h .- y .
kee , phys .
b 85 ( 2012 ) p. 245112 .
h. ikeda , m .-
suzuki , r. arita , t. takimoto , t. shibauchi , and y. matsuda , nat .
( 2012 ) p. 528 - 533 .
fazekas , a. kiss , and k. radnczi , prog .
( 2005 ) p. 114 - 133 .
v. p.mineev , and m. e. zhitomirsky , phys . rev .
b 72 ( 2005 ) p. 014432 .
y. okuno , and k. miyake , j. phys .
67 ( 1998 ) p. 2469 .
s. elgazzar , j. rusz , m. amft , p. m. oppeneer , and j. a. mydosh , nat . mat . 8
( 2009 ) p. 337 - 341 .
t. kasuya , j. phys .
66 ( 1997 ) p. 3348 . c. ppin , m.r .
norman , s. burdin , and a. ferraz , phys .
106 ( 2011 ) p. 106601 .
p. s. riseborough , b. coqblin , and s. g. magalhes , phys .
b 85 ( 2012 ) p. 165116 .
s. fujimoto , phys .
106 ( 2011 ) p. 196407 .
p. chandra , p. coleman , and r. flint , nat .
493 ( 2013 ) p. 621 - 626 .
y. dubi , and a. v. balatsky , phys .
106 ( 2011 ) p. 086401 .
t. t. m. palstra , a. a. menovsky , and j. a. mydosh , phys .
b 33 ( 1986 ) p. 6527 .
t. das , phys .
b 89 ( 2014 ) p. 045135 .
t. das , sci .
rep . 2 ( 2012 )
t. das , phys .
( 2012 ) p. 246406 .
b. a. bernevig , t. l hughes , and s .- c .
zhang , science 314 ( 2006 ) p. 1757 .
m. z. hasan , and c. l. kane , rev .
82 ( 2010 ) p. 3045 - 3067 .
qi , and s .- c .
zhang , rev .
83 ( 2011 ) p. 1057 - 1110 . c. herring , phys .
52 ( 1937 ) p. 365
h. bentmann , s. abdelouahed , m. mulazzi , j. henk , and f. reinert , phys .
( 2012 ) p. 196801 .
a. joshua , j. ruhman , s. pecker , e. altman and s. ilani , proc .
( usa ) , 110 ( 2013 ) p. 9633 - 9638 .
c. dhital , s. khadka , z. yamani , c. de la cruz , t. c. hogan , s. m. disseler , m. pokharel , k. c. lukas , w. tian , c. p. opeil , z. wang , and s. d. wilson , phys .
b 86 ( 2012 ) p. 100401(r ) .
p. blaha , k. schwarz , g. k. h. madsen , d. kvasnicka , and j. luitz , _
j. p. perdew , k. burke , and m. ernzerhof , phys .
77 ( 1996 ) p. 3865 .
p. m. oppeneer , s. elgazzar , j. rusz , q. feng , t. durakiewicz , and j. a. mydosh , phys .
b 84 ( 2011 ) p. 241102(r ) .
t. das , and a. v. balatsky , phys . rev .
b 84 , ( 2011 ) p. 014521 .
t. das , a. rahmanian , and a. v. chubukov , manuscript under preparation .
h. zhang , c .- x .
qi , xi dai , z. fang , and s .- c .
zhang , nat .
phys . 5 ( 2009 )
t. das , j .- x .
zhu , and m. j. graf , phys .
( 2012 ) p. 017001 .
t. das , t. durakiewicz , j .- x .
zhu , j. j. joyce , j. l. sarrao , and m. j. graf , phys . rev .
x 2 ( 2012 ) p. 041012 .
r. m. fernandes , and jrg schmalian , phys .
b 82 ( 2010 ) p. 014521 .
a. b. vorontsov , m. g. vavilov , and a. v. chubukov , 81 ( 2010 ) p. 174538 .
t. das , phys .
b 87 ( 2013 ) p. 144505 .
m. m. altarawneh , n. harrison , s. e. sebastian , l. balicas , p. h. tobash , j. d. thompson , f. ronning , and e. d. bauer , phys .
106 ( 2011 ) p. 146403 .
i. kawasaki , s. fujimori , y. takeda , t. okane , a. yasui , y. saitoh , h. yamagami , y. haga , e. yamamoto , and y. nuki , phys .
b 83 ( 2011 ) p. 235121 .
k. r. shirer , j. t. haraldsen , a.p .
dioguardi , j. crocker , n. aproberts - warren , a.c .
shockley , c .- h .
lin , d.m .
nisson , j.c .
cooley , m. janoschek , k. huang , n. kanchanavatee , m.b .
maple , m.j .
graf , a.v .
balatsky , and n.j .
curro , phys .
b 88 ( 2013 ) p. 094436 .
j. r. schrieffer , x. 6 .
wen , and s. c. zhang , phys .
b 39 ( 1989 ) p. 11663 .
a. j. leggett , j. phys .
c : solid state phys . 6 ( 1973 )
3187 - 3204 .
c. wu , and s .- c .
zhang , phys .
( 2004 ) p. 36403 .
e. hassinger , g. knebel , k. izawa , p. lejay , b. salce , j. and flouquet , phys .
b 77 ( 2008 ) p. 115117 .
h. amitsuka , k. matsuda , i. kawasaki , k. tenya , m. yokoyama , c. sekine , n. tateiwa , t. c. kobayashi , s. kawarazaki , and h. yoshizawa , j. magn .
310 ( 2007 ) p. 214
b. j. kim , h. jin , s. j. moon , j .- y .
park , c. s. leem , j. yu , t. w. noh , c. kim , s .- j .
oh , j .- h . park , v. durairaj , g. cao , and e. rotenberg , phys .
( 2008 ) p. 076402 .
h. lin , t. das , y. j. wang , l. a. wray , s .- y .
xu , m. z. hasan , and a. bansil , phys . rev .
b 87 ( 2013 ) p. 121202(r ) . c. tegenkamp , d. lkermann , h. pfnur , b. slomski , g. landolt , and j. h. dil , phys .
( 2012 ) p. 266401 .
l. fu , c. l. kane , and e. j. mele , phys .
98 ( 2007 ) p. 106803 .
p. goswami , and l. balicas , arxiv:1312.3632 .
|
we provide a brief review and detailed analysis of the spin - orbit density wave ( sodw ) , proposed as a possible explanation to the ` hidden order ' phase of uru@xmath0si@xmath0 . due to the interplay between inter - orbital coulomb interaction and spin - orbit coupling ( soc ) in this compound ,
the sodw is shown to arise from fermi surface nesting instability between two spin - orbit split bands .
an effective low - energy hamiltonian including single - particle soc and two - particle sodw is derived , while numerical results are calculated by using density - functional theory ( dft ) based band structure input .
computed gapped quasiparticle spectrum , entropy loss and spin - excitation spectrum are in detailed agreement with experiments .
interestingly , despite the fact that sodw governs dynamical spin - excitations , the static magnetic moment is calculated to be _
zero _ , owing to the time - reversal invariance imposed by soc . as a consequence
, sodw can be destroyed by finite magnetic field even at zero temperature .
our estimation of the location of the quantum critical point is close to the experimental value of @xmath1 35 t. finally , we extend the idea of sodw to other soc systems including iridium oxides ( iridates ) and two - dimensional electronic systems such as biag@xmath0 surface and laalo@xmath2/srtio@xmath2 interface . we show hints of quasiparticle gapping , reduction of preexisting magnetic moment , large magneto - resistance etc . in these systems which can be explained consistently within the sodw theory .
spin - orbit density wave ; hidden order ; heavy fermion ; quantum phase transition , uru@xmath0si@xmath0 .
|
propositional linear time temporal logic , ltl , is important for hardware and software verification@xcite .
ltl satisfiability checking ( ltlsat ) is receiving renewed interest with advances computing power , several industry ready tools , some new theoretical techniques , studies of the relative merits of different approaches , implementation competitions , and benchmarking : @xcite .
common techniques include automata - based approaches @xcite and resolution @xcite as well as tableaux @xcite .
each type of approach has its own advantages and disadvantages and each can be competitive at the industrial scale ( albeit within the limits of what may be achieved with pspace complexity ) .
the state of the art in tableau reasoners for ltl satisfiability testing is the technique from @xcite which is used in portfolio reasoners such as @xcite .
many ltl tableau approaches produce a very untraditional - looking graph , as opposed to a tree , and need the whole graph to be present before a second phase of discarding takes place . within the stable of tableau - based approaches to ltlsat
, the system of @xcite stands out in in being tree - shaped ( not a more general graph ) , and in being one - pass , not relying on a two - phase building and pruning process .
it also stands out in speed @xcite . however , there are still elements of communication between separate branches and a slightly complicated annotation of nodes with depth measures that needs to be managed as it feeds in to the tableau rules .
so it is not in the traditional style of classical tableaux @xcite .
this paper presents a new simpler tableau for ltl .
it builds on ideas from @xcite and is influenced by ltl tableaux by @xcite and @xcite .
it also uses some ideas from a ctl * tableau approach in @xcite where uselessly long branches are curtailed .
the general shape of the tableau and its construction rules are mostly unsurprising but the two novel prune rules are perhaps a surprisingly simple way to curtail repetitive branch extension and may be applicable in other contexts .
the tableau search allows completely independent searching down separate branches and so lends itself to parallel computing .
in fact this approach is `` embarrassingly parallel '' @xcite .
thus there is also potential for quantum implementations .
furthermore , only formula set labels need to be recorded down a branch , and checked back up the one branch , and so there is great potential for very fast implementations . the soundness , completeness and termination of the tableau search is proved .
the proofs are mostly straightforward . however , the completeness proof with the prune rules has some interesting reasoning .
we provide a simple demonstration prototype reasoning tool to allow readers to explore the tableau search process . the tool is not optimised for speed in any way but we report on a small range of experiments on standard benchmarks which demonstrate that the new tableau will be competitive with the current state of the art @xcite .
in this paper we give some context in section [ sec : other ] , before we briefly confirm our standard version of the well - known syntax and semantics for ltl in section [ sec : synsem ] , describe our tableau approach is general terms section [ sec : tab ] , present the rules section [ sec : rules ] , make some comments and provide some motivation for our approach section [ sec : motiv ] , prove soundness in section [ sec : sound ] , prove completeness in section [ sec : complete ] , and briefly discuss complexity and implementation issues section [ sec : complex ] , and detailed comparisons with the schwendimann approach in section [ sec : compare ] , before a conclusion section [ sec : concl ] .
the latest version of this long report can be found at http://www.csse.uwa.edu.au/~mark/research/online/ltlsattab.html with links to a java implementation of the tool and full details of experiments .
ltl is an important logic and there has been sustained development of techniques and tools for working with it over more than half a century .
many similar ideas appear as parts of different theoretical tools and it is hard for a researcher to be across all the threads .
thus it is worth putting the ideas here in some sort of context .
we will define structures and formulas more carefully in the next section but a structure is essentially a way of definitively and unambiguously describing an infinite sequence of states and an ltl formula may or may not apply to the sequence : the sequence may or may not be a model of the formula .
see section 3 or @xcite for details .
we are addressing a computational task called _ satisfiability checking_. that is , given an ltl formula , decide whether or not there exists any structure at all which is a model of that formula .
input is the ltl formula , output is yes or no .
the procedure or algorithm must terminate with the correct answer .
this problem is pspace complete in the size of formula @xcite .
there is a related but separate task called _ model checking_. model checking is the task of working out whether a given structure is a model or not of a given formula .
input is the ltl formula and a description of a system , output is yes or no , whether or not there is a behaviour generated by the system which is a model of the formula ( or some variant on that ) .
model checking may seem to be more computationally demanding in that there is a formula and system to process . on the other hand ,
as described in @xcite and outlined below , it is possible to use model checking algorithms to do satisfiability checking efficiently ( in pspace ) .
model - checking itself , with a given formula on a given structure , is also pspace - complete @xcite , but in @xcite we can find an algorithm that is exponential in the size of the formula and linear in the size of the model .
the thorough experimental comparison in @xcite finds that across a wide range of carefully chosen benchmarks , none of the usual three approaches to ltl satisfiability testing dominates .
here we briefly introduce tableaux and distinguish the automata - based model - checking approach .
resolution is a very different technique and we refer the reader elsewhere , for example to @xcite .
tableau approaches trace their origins to the semantic tableaux , or truth trees , for classical logics as developed by beth @xcite and smullyan @xcite .
the traditional tree style remains for many modal logics @xcite but , as we will outline below , this is not so for the ltl logic .
the tableau itself is ( typically ) a set of nodes , labelled by single formulas or sets of formulas , with a successor relation between nodes .
the tree tableau has a root and the successor relation gives each node @xmath0 , @xmath1 or @xmath2 children .
the tableau is typically depicted with the root at the top and the children below their parent : thus giving an upside tree - shape .
building a tableau constitutes a decision procedure for the logic when it is governed by rules for labelling children nodes , for counting a node as the leaf of a successful branch and for counting a node as the leaf of a failed branch .
finding one successful branch typically means that the original formula is satisfiable while finding all branches failing means it is not satisfiable .
there are efficiently implemented tableau - based ltl satisfiability reasoning tools , which are easily available , such as pltl @xcite and lwb @xcite .
we describe the approach and its variations in more detail in the next subsection .
the other main approach to ltl satisfiability checking is based on a reduction of the task to a model - checking question , which itself is often implemented via automata - based techniques . the traditional automata approach to model checking
seen , for example in @xcite , is to compose the model with an automaton that recognizes the negation of the property of interest and check for emptiness of the composed structure .
recently there has been some serious development of satisfiability checking on top of automata - based model - checking that ends up having some tableau - like aspects .
the automata approach via model - checking as described in @xcite shows that if you can do model checking you can do satisfiability checking .
one can construct an automaton which will represent a universal system allowing all possible traces from the propositions .
a given formula is satisfiable iff the universal system contains a model of the formula . by not actually building the universal system
entirely the overall task can be accomplished in pspace in the size of formula . some fast and effective approaches to satisfiability checking
can then be developed by reasoning about the model checking using symbolic ( bounded ) sat or bdd - based reasoners such as cadencesmv and nusmv @xcite .
alternatively , _
explicit _ automata - based model - checkers such as spin @xcite could be used for the model checking but @xcite shows these to not be competitive .
an important part of the symbolic approach is the construction of a symbolic automaton from the formula and this involves a tableau - like construction with sets of subformulas determining states .
the symbolic automaton or tableau presented as part of the model - checking procedure in @xcite has been commonly used but this has been extended in @xcite to a portfolio of translators . also an ltl
to tableau tool used in @xcite is available via the first author s website .
we discuss these sorts of tableaux again briefly below .
the model - checking approach to ltl satisfiability checking is further developed in @xcite where a novel , on - the - fly process combines the automata - construction and emptiness check .
the tool out - performs previously existing tools for ltlsat which use model - checking .
comprehensive experimental comparisons with approaches to ltlsat beyond automata - theoretic model - checking based ones is left as future work . in some modal logics , and in temporal logics in particular , variations on the traditional tableau idea have been prevalent and the pure tree - shape is left behind as a more complicated graph of nodes is constructed . as we will see , this may happen in these variant tableaux , when the successor relation is allowed to have _ up - links _ from descendants to ancestors , _ cross - links _ from nodes to nodes on other branches , or where there is just an arbitrary directed graph of nodes .
such a _ graph - shaped _ approach results if , for example , we give a _ declarative _ definition of a node and the successor relation determined by the labels on the node at each end .
typically , the node is identified with its label .
certain sets of formulas are allowed to exist as nodes and we have at most one node with a given label . examples of this sort include the tableaux in @xcite and ( the symbolic automaton of ) @xcite .
alternatively , the tableau may have a traditional form that is essentially _ tree - shped _ with a root and branches of nodes descending and branching out below that
. usually a limited form of up - link is allowed back from a node ( leaf or otherwise ) to one of its ancestors
. there may be tableaux which contain several different nodes with the same label .
examples , include @xcite and the new one .
the new style tableau even allows multiple nodes down the same branch with the same labels while this is not permitted in @xcite . the tableau construction may be _ incremental _ , where only reachable states are constructed , versus _ declarative _ , when we just define what labels are present in the tableau and which pairs of labels are joined by a directed edge .
a tableau is said to be _ one - pass _ if the construction process only build legitimate nodes as it proceeds . on the other hand ,
it is _ multi - pass _ if there is an initial construction phase followed by a culling phase in which some of the nodes ( or labels ) which were constructed are removed as not being legitimate .
a construction of a tableau , or a search through a tableau , will often proceed in a _ depth - first _ manner starting at a chosen node and then moving successively to successors .
an alternative is via some sort of _ parallel _ implementation in which branches are explored concurrently .
search algorithms may make use of heuristics in guessing a good branch to proceed on .
undertaking a depth - first search in a graph - shaped label - determined tableau may seem to be similar to building a tree - shaped tableau but it is likely that the algorithm will behave differently when it visits a label that has been seen before down an earlier branch . in a tree - shaped tableau
this may not need to be recorded . which is faster ?
trees @xcite may have 2exptime worst case complexity .
graphs re - use labels and make many possible branches at one time : in general exptime .
however , @xcite demonstrated that the tree - shaped approach of @xcite ( consistently and sometimes drastically ) outperformed the graph - shaped approach of @xcite .
wolper s @xcite was the first ltl tableau .
it is a multi - pass , graph - shaped tableau .
the nodes are labelled with sets of formulas ( from the closure set ) with a minimal amount of extra notation attached to record which formulas already have been decomposed .
one builds the graph starting at @xmath3 and using the standard sorts of decomposition rules and the transition rule .
the tableau may start off looking like a tree but there must not be repeated labels so edges generally end up heading upwards and/or crossing branches .
after the construction phase there is iterated elimination of nodes according to rules about successors and eventualities .
( this approach was later extended to cover the inclusion of past - time operators in @xcite ) .
there was a similar but slightly quicker proposal for a multi - pass , graph - shaped tableau for ltl in @xcite .
the similar tableau in @xcite is incremental but multi - pass as it builds a graph from initial states , then looks for strongly connected components ( to satisfy eventualities ) . other graph - like tableaux include those in @xcite .
a graph - shaped tableau for ltl also forms part of the model - checking approach suggested in @xcite . here
the need to check fulfilment of eventualities is handed over to some ctl fairness constraint checking on a structure formed from the product of the tableau and the model to be checked .
the symbolic model checker smv is used to check the property subject to those fairness constraints . in @xcite ,
the ` symbolic automaton ' approach based on the tableau from @xcite was adapted to tackle ltl satisfiability checking .
figure [ fig : cgh97tab ] shows a typical graph - style tableau from @xcite .
schwendimann s tableau @xcite is close to being purely tree - shaped ( not a more general graph ) .
it is also one - pass , not relying on a two - phase building and pruning process .
however , there are still elements of communication between separate branches and a slightly complicated annotation of nodes with depth measures that needs to be managed as it feeds in to the tableau rules . in general , to decide on success or not for a tableau we need to work back up the tree towards the root , combining information from both branches if there are children .
thus , after some construction downwards , there is an iterative process moving from children to parents which may need to wait for both children to return their data .
this can be done by tackling one branch at a time or in theory , in parallel .
the data passed up consists in an index number and a set of formulas ( unfulfilled eventualities ) .
we describe this approach in more detail in section [ sec : compare ] .
figure [ fig : schwetab ] shows a typical tree - style tableau from @xcite .
the main available tools for ltl satisfiability checking are listed and described in @xcite .
these included pltl which ( along with another tableau option ) implements schwendiammn s approach .
we should also mention @xcite here .
the tool uses a novel , on - the - fly approach to ltlsat via model - checking and out performs previously existing tools for ltlsat which use the model - checking approach .
there is an open source ltl to tableau translator available from rozier s website used for the @xcite approach .
also , a more limited implementation was released open - source following the publication of @xcite : it is called ltl2smv and distributed with the nusmv model checker .
there is a brief comparison of the tableau approaches of schwendimann and wolper in @xcite and in @xcite there is a comparison symbolic model - checking approaches .
however , the most thorough benchmarking exercise is as follows .
most known implemented tools for deciding satisfiability in ltl are compared in @xcite .
the best tools from three classes are chosen and compared : automata - based reduction to model - checking , tableau and resolution . a large range of benchmark patterns are collected or newly proposed .
they find that no solver , no one of the three approaches , dominates the others .
the tableau tool `` pltl '' based on schwendimann s approach is the best of the tableaux and the best overall on various classes of pattern .
a portfolio solver is suggested and also evaluated .
the benchmark formulas , rendered in a selection of different formats , is available from schuppan s webpage .
the two prune rules are novel .
they force construction of branches to be terminated in certain circumstances .
they depend only on the labels at the node and the labels of its ancestors . in general they may allow a label to be repeated some number of times before the termination condition is met .
the overall tableau shape is novel .
although tableaux are traditionally tree - shaped , no other tableau system for ltl builds graphs that are tree shaped .
most tableaux for ltl are more complicated graphs .
the schwendimann approach is close to being tree - shaped but there are still up - links from non - leaves .
the labels on the tableau are just sets of formulas from the closure set of the original formula ( that is , subformulas and a few others ) . other approaches ( such as schwendimann s ) require other annotations on nodes . overall tableau : one that is in completely traditional style ( labels are sets of formulas ) , tree shaped tableau construction , no extraneous recording of calculated values , just looking at the labels .
completely parallel development of branches .
no communication between different branches .
this promises interesting and useful parallel implementations .
the reasoning speed seems to be ( capable of being ) uniquely fast on some important benchmarks .
we assume a countable set @xmath4 of propositional atoms , or atomic propositions . a ( transition ) structure is a triple @xmath5 with @xmath6 a finite set of states , @xmath7 a binary relation called the transition relation and labelling @xmath8 tells us which atoms are true at each state : for each @xmath9 , @xmath10 .
furthermore , @xmath11 is assumed to be total : every state has at least one successor @xmath12 given a structure @xmath5 an @xmath13-sequence of states @xmath14 from @xmath6 is a _ fullpath _ ( through @xmath5 ) iff for each @xmath15 , @xmath16 . if @xmath17 is a fullpath then we write @xmath18 , @xmath19 ( also a fullpath ) .
the ( well formed ) formulas of ltl include the atoms and if @xmath20 and @xmath21 are formulas then so are @xmath22 , @xmath23 , @xmath24 , and @xmath25 .
we will also include some formulas built using other connectives that are often presented as abbreviations instead .
however , before detailing them we present the semantic clauses .
semantics defines truth of formulas on a fullpath through a structure .
write @xmath26 iff the formula @xmath20 is true of the fullpath @xmath27 in the structure @xmath28 defined recursively by : + [ cols= " < , < , < " , ] we use benchmark formulas from the various series described in @xcite .
the only new series we add is @xmath29 as follows : for all @xmath30 , @xmath31 originally , the @xmath29 series was invented by the author to obtain an idea experimentally how much longer it might take the new tableau compared to schwendimann s on testing examples but the experiments show the opposite outcome .
in this section we give a detailed comparison of the new tableau s operation compared to that of the schwendimann s tableau . from @xcite .
the tableau is a labelled tree .
suppose we are to decide the satisfiability or not of @xmath32 .
labels are of the form @xmath33 where @xmath34 is a set of formulas ( subformulas of @xmath32 ) , @xmath35 is a pair ( described shortly ) and @xmath11 is a pair of the form @xmath36 where @xmath37 and @xmath38 . the @xmath34 part of a node label @xmath33 is a set of subformulas of @xmath32 serving a similar purpose to our node labels .
the second component @xmath35 is a pair where @xmath39 is a set of formulas , and @xmath40 is a list of pairs each of the form @xmath41 .
the set @xmath39 records the eventualities cured at the current state .
the list @xmath42 is a record of useful information from the state ancestors of the current node along the current branch .
each @xmath43 records the poised label @xmath44 of the @xmath15th state down the branch and the set @xmath45 of eventualities fulfilled there .
the second component @xmath6 essentially contains information about the labels on the current branch of the tree and so is just a different way of managing the same sort of `` historical '' recording that we manage in our new tableau .
we manage the historical records by having the branch ancestor labels directly available for checking .
the third component @xmath36 , a pair consisting of a number and a set of formulas , is specific to the schwendimann tableau process and has no analogous component in our new tableau .
the number @xmath46 records the highest index of ancestor state in the current branch that can be reached directly by an up - link from a descendent node of the current node .
the set @xmath47 contains the eventualities which are in the current node label but which are not cured by the time of the end of the branch below .
the elements of the third component are not known as the tableau is constructed until all descendants on all branches below the current node are expanded .
then the values can be filled in using some straightforward but slightly lengthy rules about how to compute them from children to parent nodes .
the rules for working out these components are slightly complicated in the case of disjunctive rules , including expansion of @xmath48 formulas .
the third component helps assess when a branch and the whole tableau can be finished .
apart from having to deal with the second and third components of the labels , the rules are largely similar to the rules for the new tableau . however , there are no prune rules
. the schwendimann tableau does not continue a branch if there is a node with the same label as one of its proper ancestors .
the branch stops there .
the other main difference to note is that the schwendimann tableau construction rules for disjunctive formulas do in general need to combine information from both children s branches to compute the label on the parent .
consider the example @xmath49 technically schwendimann tableau assumes all formulas ( the input formula and its subformulas ) are in _ negation normal form _ which involves some rewriting so that negations only appear before atomic propositions .
however , we can make a minor modification and assume the static rules are the same as our new ones .
we start a tableau with the label @xmath50 with @xmath51 , @xmath52 and @xmath53 with both @xmath46 and @xmath47 unknown as yet .
tableau rules decompose @xmath54 and subsequent subformulas in a similar way to in our new tableau .
neither @xmath6 nor @xmath11 change while we do not apply a step rule ( known as @xmath55 rule ) .
the conjunctions and @xmath56 rules are just as in our new tableau .
thus the tableau construction soon reaches a node with the label @xmath50 as follows : @xmath52 , @xmath53 with both @xmath46 and @xmath47 unknown as yet and @xmath57 there are six choices causing branches within this state caused by the disjuncts and the @xmath58 formulas .
this could lead to 64 different branches before we complete the state but many of these choices lead immediately to contradictions .
the details depend on the order of choice of decomposing formulas .
a typical expansion leads to a branch with a node labelled by the state @xmath50 as follows : @xmath52 , @xmath53 with both @xmath46 and @xmath47 unknown as yet and @xmath59 notice that the first component of @xmath6 is still empty as , in this case , none of the eventualities , @xmath60 nor @xmath61 , is fulfilled here .
the next rule to use in such a situation is the transition rule , which is very similar to that in our new tableau , except that the second component part of the label is updated as well .
we find @xmath62 , @xmath53 with both @xmath46 and @xmath47 unknown as yet and @xmath63 a further series of expansions and choices leads us to the following new state , @xmath64 , @xmath53 with both @xmath46 and @xmath47 unknown as yet and @xmath65 applying the step rule here gives us the following : @xmath66 , @xmath53 with both @xmath46 and @xmath47 unknown as yet and @xmath67 when this is branch is expanded further then we find ourselves at @xmath68 with @xmath66 and @xmath53 with both @xmath46 and @xmath47 unknown as yet .
notice the main first part of the label has ended up being @xmath69 again .
thus we have a situation for the loop rule .
the loop rule uses the facts that the state repeated has the index @xmath1 in the branch above and that the eventuality @xmath61 has not been fulfilled in this branch .
thus we can fill in @xmath70 and @xmath71 at all the intervening pre - states which have been left unknown so far .
these values do not as yet transfer up to the top state as yet as there are still other undeveloped branches from within that state .
the other two states that we find are a minor variation on @xmath72 , @xmath73 and a mirror image of @xmath72 when @xmath74 is true instead of @xmath75 : @xmath76 then all eventualities are fulfilled and the tableau succeeds after 3933 steps .
the overall picture of states ( poised labels ) is as follows ( with @xmath1 standing for @xmath72 etc ) .
@xmath77 by comparison , the new tableau visits the same states but takes only 3087 steps to proceed as follows : @xmath78 the new tableau can always decide on the basis of a single branch , working downwards .
schwendimann s needs communication up and between branches , and will generally require full development of several branches .
the new tableau just needs to store formula labels down current branch .
however , if doing depth - first search , also needs information about choices made to enable backtracking .
schwendimann s tableau needs to pass extra sets of unfulfilled eventualities back up branches also keeping track of indices while backtracking .
it also needs to store information about choices made to enable backtracking .
we have introduced novel tableau construction rules which support a new tree - shaped , one - pass tableau system for ltlsat .
it is traditional in style , simple in all aspects with no extra notations on nodes , neat to introduce to students , amenable to manual use and promises efficient and fast automation . in searching or constructing the tableau
one can explore down branches completely independently and further break up the search down individual branches into separate independent processes .
thus it is particularly suited to parallel implementations .
experiments show that , even in a standard depth - first search implementation , it is also competitive with the current state of the art in tableau - based approaches to ltl satisfiability checking .
this is good reason to believe that very efficient implementations can be achieved as the step by step construction task is particularly simple and requires on minimal storage and testing .
because of the simplicity , it also seems to be a good base for more intelligent and sophisticated algorithms : including heuristics for choosing amongst branches and ways of managing sequences of label sets .
valentin goranko , angelo kyrilov , and dmitry shkatov .
tableau tool for testing satisfiability in ltl : implementation and experimental analysis .
, 262(0):113 125 , 2010 .
proceedings of the 6th workshop on methods for modalities ( m4m-6 2009 ) .
yonit kesten , zohar manna , hugh mcguire , and amir pnueli .
a decision algorithm for full propositional temporal logic . in costas courcoubetis ,
editor , _ cav _ , volume 697 of _ lecture notes in computer science _ , pages 97109 .
springer , 1993 .
kristin y. rozier and moshe y. vardi .
ltl satisfiability checking . in dragan bosnacki and stefan
edelkamp , editors , _ spin _ , volume 4595 of _ lecture notes in computer science _ , pages 149167 .
springer , 2007 .
k. y. rozier and m. y. vardi .
a multi - encoding approach for ltl symbolic satisfiability checking . in _
17th international symposium on formal methods ( fm2011 ) _ , volume 6664 of _ lncs _ , pages 417431 .
springer verlag , 2011 .
s. schwendimann . a new one - pass tableau calculus for pltl . in harrie c. m. de swart , editor , _ proceedings of international conference , tableaux 1998 , oisterwijk _ , lnai 1397 , pages 277291 .
springer , 1998 .
viktor schuppan and luthfi darmawan .
evaluating ltl satisfiability solvers . in tevfik bultan and pao - ann
hsiung , editors , _
atva11 _ ,
volume 6996 of _ lecture notes in computer science _ , pages 397413 .
springer , 2011 .
|
propositional linear time temporal logic ( ltl ) is the standard temporal logic for computing applications and many reasoning techniques and tools have been developed for it .
tableaux for deciding satisfiability have existed since the 1980s . however , the tableaux for this logic do not look like traditional tree - shaped tableau systems and their processing is often quite complicated .
we present a new simple traditional - style tree - shaped tableau for ltl and prove that it is sound and complete . as well as being simple to understand , to introduce to students and to use manually , it also seems simple to implement and promises to be competitive in its automation .
it is particularly suitable for parallel implementations .
note : the latest version of this report can be found via http://www.csse.uwa.edu.au/~mark/research/online/ltlsattab.html .
|
the recently demonstrated @xcite high - fidelity preparation of a single exciton state in a self - assembled quantum dot ( qd ) by means of adiabatic evolution induced by a chirped laser pulse ( referred to as _ adiabatic rapid passage _ , arp ) opens new possibilities of charge control in qds .
in contrast to the more traditional rabi flops @xcite as well as to methods based on voltage control @xcite , arp is much less sensitive to the details of the driving field . in particular , in the ideal case , the arp technique ensures the full inversion of occupation between the empty dot and exciton states as soon as the pulse intensity reaches the threshold for the adiabatic passage .
however , as qds are embedded in a semiconductor crystal matrix , carrier - phonon interactions are usually found to considerably limit the fidelity of various optical control schemes .
the phonon - induced dephasing process is inevitable in such systems and leads to loss of information .
these mechanisms have been investigated theoretically @xcite and have been confirmed experimentally @xcite . as we have recently shown @xcite , in the case of the arp
the role of phonon coupling in the state preparation is essential .
as the adiabatic evolution follows one of the spectral branches corresponding to dressed exciton states , carrier phonon coupling leads to transitions to the other branch . depending on whether the upper or lower branch is followed ( which , in turn , depends on the direction of the chirp )
, the transition is assisted by phonon emission or absorption , respectively . as a consequence , at low temperatures
the phonon - induced decoherence leads to a strong asymmetry in the final occupation of the exciton state depending on the sign of the chirp @xcite . while the analysis performed in ref .
was based on a two - level model , corresponding to the excitation with a circularly polarized laser pulse , in actual experiments linearly polarized beams have been used @xcite . in this case , the selection rules allow the coupling to the biexciton state and phonon - assisted biexciton generation @xcite may occur even at low temperatures , as the biexciton state is typically shifted to lower energies . in this paper
, we study the evolution of the biexciton system driven by a linearly polarized chirped pulse in the regime of the arp in the presence of the coupling to acoustic phonons .
we show that including the biexciton state , which becomes relevant for linearly polarized laser pulses @xcite , indeed opens a new decoherence path related to phonon - assisted transitions to the biexciton state . as a result ,
the fidelity of exciton state preparation drops down considerably in particular in most of the parameter areas where almost ideal evolution was predicted by the two - level model .
thus , high - fidelity control becomes restricted to the range of moderate pulse areas for positive chirps and at low temperatures only , unless circular polarized pulses are used to suppress biexcitonic excitations .
the paper is organized as follows . in sec .
[ sec : model ] , we describe the system and define the model used in our study . sec .
[ sec : simul ] introduces our method of simulation of the open system dynamics .
the results are presented in sec .
[ sec : results ] . finally , sec .
[ sec : concl ] concludes the paper .
the biexciton system in the qd , restricted to the lowest bright states , is modeled as a four - level system with @xmath0 representing the empty dot , @xmath1,@xmath2 being the two exciton states with different linear polarizations and @xmath3 denoting the biexciton state .
the corresponding hamiltonian is @xmath4 where @xmath5 is the exciton transition energy , @xmath6 is the biexciton shift , and we neglect the fine structure splitting which is irrelevant on the picosecond time scales considered here .
this charge system is driven by a chirped gaussian pulse with the original envelope @xmath7 where @xmath8 is the original pulse area ( before chirping ) and @xmath9 is the initial duration ( pulse length ) .
the central frequency @xmath10 is chosen at resonance with the single exciton transition .
we assume that the linear chirp is effected by passing this pulse through a gaussian chirp filter @xcite with the chirp coefficient @xmath11 .
this results in the chirped pulse envelope @xmath12 with the chirped pulse length @xmath13 , and the frequency chirp rate @xmath14 .
the hamiltonian describing the coupling to the chirped pulse in the rotating wave approximation is @xmath15 thus , for one linear polarization , the coupling is @xmath16 and the other single - exciton state @xmath17 is decoupled ( it is not coupled by phonon transitions , either@xcite ) .
therefore , only a three - level model is needed@xcite .
( color online ) the adiabatic spectral branches of the optically driven qd system for @xmath18 , @xmath19 ps , @xmath20 ps@xmath21 and @xmath22 mev : ( a ) positive chirp ; ( b ) negative chirp .
the evolution starting from the ground state follows the red line .
the gray solid line represents the pulse envelope @xmath23 which has been shifted upwards for clarity.,width=336 ] the idea of the adiabatic passage @xcite is to drive the system state adiabatically along one of the spectral branches corresponding to the instantaneous eigenstates of the hamiltonian @xmath24 .
these instantaneous eigenstates are shown in fig .
[ fig : branches ] for the biexciton shift @xmath25 mev .
the branch followed by the system state is plotted with a thicker line . in fig .
[ fig : branches](a ) , we assume a positive chirp ( @xmath26 ) , while in fig . [
fig : branches](b ) a negative chirp is assumed . for pulse intensities above the arp threshold , the central anticrossing at @xmath27 ( between the vacuum and exciton states ) is passed adiabatically along a single spectral branch . on the contrary , the two other anticrossings away from @xmath27 ( exciton - biexciton and vacuum biexciton ) are met when the pulse is already very weak , hence they are very narrow ( in the case shown here , the widths of both of these anticrossings are below 10 @xmath28ev , unresolved in fig . [
fig : branches ] ) . at such narrow anticrossings ,
the sweep rate is very high compared to the distance between the two levels , which leads to a nearly purely non - adiabatic evolution : these anticrossings are almost completely crossed by the evolving system state , that is , the system nearly completely jumps to the other branch ( as shown by color coding of the lines in fig .
[ fig : branches ] ) . as a result ,
the charge configuration does not change here , as opposed to the central anticrossing . for the parameter range corresponding to the experiments @xcite excitation of optical phonon modes
is excluded by the spectral properties of the system . in consequence
, the coupling to acoustic phonons can be expected to dominate as a source of decoherence .
the carrier - phonon interaction is described in terms of the usual independent boson hamiltonian , @xmath29 , with @xmath30 and @xmath31 where @xmath32 are phonon annihilation operators .
we assume that the biexcitonic and excitonic wave functions can be factorized into products of single particle wave functions which are the same for both exciton and biexciton .
this is well justified in the strong confinement limit where the wave functions are mostly determined by the confinement potentials .
the coupling constants , accounting for the deformation potential interaction , can then be written as @xmath33,\ ] ] where @xmath34 is the normalization volume , @xmath35 is the crystal density , @xmath36 is the speed of sound , @xmath37 is the deformation potential constant for an electron ( hole ) and @xmath38 are the form factors for electron ( hole ) wave function @xmath39 .
finally , the phonon reservoir is described by the hamiltonian @xmath40 where @xmath41 are the phonon frequencies .
the system is conveniently described in the non - uniformly rotating frame defined by the unitary transformation @xmath42 , where @xmath43 the resulting hamiltonian is @xmath44 , with @xmath45 where @xmath46 , and @xmath47
to simplify the numerical simulation , we perform the unitary transformation defined by the operator @xmath48 where @xmath49.\ ] ] upon this transformation and expanding to the leading order in the coupling constants , the hamiltonian may be written as @xmath50 where @xmath51 and @xmath52 here @xmath53 accounts for the phonon - induced renormalization of the pulse amplitude in the slow driving limit @xcite .
we assume that the central frequency of the driving pulse is corrected for the phonon - induced energy shifts .
the evolution of the reduced density matrix of the charge subsystem is then found by numerically solving the lowest - order time - convolutionless ( tcl ) evolution equation @xcite @xmath54 \right],\ ] ] where @xmath55 is the interaction hamiltonian @xmath56 in the interaction picture with respect to @xmath57 , @xmath58 is the phonon density matrix at the thermal equilibrium , and @xmath59 denotes the partial trace with respect to phonon states .
( color online ) the final occupation of the single - exciton state ( color coded ) as a function of the original pulse area @xmath8 and the chirp @xmath11 in the absence of coupling to phonons : ( a ) for @xmath60 mev ( b ) for @xmath25 mev.,width=321 ] ( color online ) the final occupation of the single - exciton state ( color coded ) as a function of the original pulse area @xmath8 and the chirp @xmath11 with phonon effects included : ( a , c , e ) for @xmath25 mev ( b , d , f ) for @xmath61 mev.,width=321 ] ( color online ) the final occupation of the single - exciton state as a function of the original pulse area @xmath8 at two different temperatures : ( a , b ) for @xmath25 mev ; ( c , d ) for @xmath61 mev .
blue dotted line shows the results with chirp @xmath62 ps@xmath63 , the red solid line represents @xmath64 , while the green dashed line corresponds to @xmath65 ps@xmath63 .
the final occupation of the biexciton state for @xmath25 mev is shown in ( e , f ) .
, width=321 ] ( color online ) the total number of excitons ( color coded ) as a function of the original pulse area @xmath8 and the chirp @xmath11 for @xmath25 mev ( a ) at @xmath66 k and ( b ) at @xmath67 k.,width=321 ]
with the simulation method outlined in sec . [
sec : simul ] , we study the exciton dynamics focusing on the possibility of phonon - assisted transitions to the biexciton state , when the system is excited with a linearly polarized laser pulse . in our simulations , we assume @xmath68 kg / m@xmath69 , @xmath70 m / s , @xmath71 ev and @xmath72 ev .
the electron wave functions are assumed to be gaussians with the extension @xmath73 nm in the @xmath74 plane and @xmath75 nm along the growth direction . in the case of holes ,
we take @xmath76 and @xmath77 respectively . the pulse length ( before chirping ) is taken to be @xmath19 ps .
first , we need to assess the degree of perturbation to the adiabatic evolution which stems from the additional level crossings with the biexciton ( see fig . [
fig : branches ] ) and is not related to phonons . to this end , in fig . [
fig : maps - bf ] we show the single - exciton occupation without carrier - phonon coupling for @xmath60 mev and @xmath25 mev .
the results show a reduction of the final occupation of the single exciton state at higher pulse areas for a very small biexciton shift ( @xmath60 mev ) and positive chirps . we have found that under these conditions , a biexciton occupation on the order of @xmath78 occurs ( not shown ) . on the other hand , for the still small value of @xmath25 mev ,
the evolution is nearly unperturbed by the presence of the biexciton state .
thus we conclude that for @xmath79 mev no direct population of the biexciton occurs with the pulses studied here , as the driving of the exciton to biexciton transition is too off - resonant . in the next step ,
we include phonons and consider two cases : a ) the biexciton state is shifted by a typical energy ( @xmath25 mev ) below the single - exciton state , b ) for comparison , the biexciton is shifted by a very large ( and rarely realized ) energy ( @xmath61 mev ) . in the first case , one expects that phonon - induced transitions to the biexciton state are efficient . in the second case , due to the large energy separation between the states ( beyond the cut - off frequency of the carrier - phonon coupling )
, the biexciton generation is expected to be negligible and the system can effectively be treated as a two - level system . the simulated evolution of the single exciton occupation for both cases is shown in fig . [ fig : maps ] .
the left panels ( fig .
[ fig : maps](a , c , e ) ) contain results for @xmath25 mev , and the right panels ( fig .
[ fig : maps](b , d , f ) ) for @xmath61 mev .
the final occupation is calculated as a function of the pulse area @xmath8 and the chirp rate @xmath11 at three temperatures ( @xmath66 , @xmath80 , and @xmath81 k ) .
let us start the discussion for the case of the large biexciton shift @xmath61 mev and low temperature @xmath66 k shown in fig .
[ fig : maps](b ) . as expected ,
the results from the two - level model @xcite are retrieved and a strong asymmetry in the final occupation with respect to the chirp is seen . in the two - level model , for negative chirp
the evolution follows the upper branch and phonon emission leads to damping , while for positive chirps the lower branch is followed and since phonon absorption is almost absent at 1 k a full exciton occupation is reached .
in contrast , in fig . [
fig : maps](a ) , for the small biexciton shift @xmath25 mev , the influence of the biexciton state can be clearly seen . in particular , for positive chirps a decoherence at high pulse areas is visible due to the opening of the new decoherence channel consisting in transitions to the biexciton state ( phonon - assisted biexciton generation ) . as can be seen in fig .
[ fig : branches ] , for a negative biexciton shift ( positive biexciton binding energy ) , this process corresponds to phonon emission no matter which adiabatic branch is followed , hence it is effective irrespective of the sign of the chirp even at low temperatures . also for negative chirps
the damping has become more effective .
thus , only for moderate pulse areas around @xmath18 a high - fidelity preparation of the single exciton state is possible . when considering higher temperatures , as seen in fig .
[ fig : maps](c - f ) , phonon absorption becomes more likely and now also for positive chirps decoherence due to transitions between the main branches is stronger .
hence , considerable phonon - induced perturbation is seen even in the case of strongly detuned biexciton ( fig . [
fig : maps](d , f ) ) . in consequence ,
the additional effect of the biexciton shift is relatively weak . for a more quantitative analysis
we show the final occupation as a function of pulse area for different chirps @xmath82 and @xmath80 ps@xmath63 in fig .
[ fig : cs](a - d ) for @xmath25 mev ( fig .
[ fig : cs](a , b ) ) and for @xmath61 mev ( fig . [ fig : cs](c , d ) ) .
the most remarkable difference for the two biexciton shifts is again seen for positive chirp @xmath83
ps@xmath63 at @xmath66 k. while in fig .
[ fig : cs](a ) , for the small biexciton shift , a strong damping of the exciton occupation is found , in fig .
[ fig : cs](c ) , for the large biexciton shift , an exciton occupation of one is reached for all pulse areas above the adiabatic threshold .
also the decoherence for the negative chirp @xmath84
ps@xmath63 and damping of the rabi oscillations at zero chirp are stronger due to the phonon - assisted biexciton generation .
for higher temperature at @xmath85 k in fig .
[ fig : cs](b , d ) the difference between small and large biexciton shift is less pronounced .
the strong decoherence for the small biexciton shift @xmath25 mev is accompanied by an occupation of the biexciton state as shown in fig .
[ fig : cs](e , f ) confirming our previous statements .
since in some experiments @xcite a photocurrent detection scheme is used , in which the detection signal is proportional to the total number of photocreated excitons , we have calculated also the total number of excitons in the system , @xmath86 , where @xmath87 and @xmath88 are the occupation of the single- and biexciton state respectively .
the results are shown in fig .
[ fig : maps - total ] for @xmath89 mev at @xmath66 k ( fig .
[ fig : maps - total](a ) ) and @xmath67 k ( fig .
[ fig : maps - total](b ) ) .
one can see that at a certain level of phonon - assisted biexciton generation the result exceeds 1 , which has no sense if interpreted in terms of the two - level model .
moreover , it is clear that even below this limit , the biexciton contribution leads to an overestimated occupation of the single exciton state .
in summary , we have studied the open system evolution of a qd adiabatically controlled by a chirped linearly polarized laser pulse .
we have shown that the presence of a biexciton state , which becomes optically coupled for linearly polarized laser pulses @xcite , opens new decoherence paths related to phonon - assisted biexciton generation .
if the biexciton state is shifted down by @xmath90 or @xmath91 mev it becomes accessible via phonon emission processes and the resulting decoherence is strong even at low temperatures .
this reduces the achieved occupation of the single exciton state in the parameters range where the chirped pulse control was very efficient in a model without the biexciton state .
since the phonon - assisted biexciton generation increases with the pulse area moderate pulse areas are optimal for the state preparation . on the other hand ,
if the biexciton shift is larger , the two - level evolution is essentially recovered .
this work was supported in part by a research group linkage project of the alexander von humboldt foundation and by the team programme of the foundation for polish science , cofinanced from the european regional development fund .
|
we study the evolution of an exciton confined in a quantum dot adiabatically controlled by a frequency - swept ( chirped ) laser pulse in the presence of carrier - phonon coupling .
we focus on the dynamics induced by a linearly polarized beam and analyze the decoherence due to phonon - assisted biexciton generation .
we show that if the biexciton state is shifted down by a few mev , as is typically the case , the resulting decoherence is strong even at low temperatures . as a result ,
efficient state preparation is restricted to a small parameter area corresponding to low temperatures , positive chirps and moderate pulse areas .
|
the fundamental quantity to be considered in the driven motion of helical particles is the propulsion velocity @xmath14 induced by the rotational motion from the application of a constant external torque .
the angular velocity of the helix for a constant torque depends on viscous drag coefficients which are presented in fig .
we first present results for optimizing the geometric shape parameters for the helical particles shown in fig .
[ directions ] such that @xmath14 is maximized by using method 1 as described in the methods section . during optimization
the hydrodynamic interactions are fully taken into account , but thermal noise is neglected .
the velocity @xmath14 depends on several parameters . the drag force experienced by a helical particle as it propels in the fluid depends on the shape of its cross - section .
we consider here three different shapes as shown in the methods , where the helix has a circular cross - section and helicoids a and b have elliptic cross - sections .
the helicoid a ( b ) has its long axis parallel ( perpendicular ) to the radial direction of the long axis ( see diagram in fig . [ directions ] ) .
the volumes of the helical particles are kept constant here and the elliptic cross - sections of helicoids a and b have a linear eccentricity of @xmath15 .
other relevant parameters are the number of helical turns @xmath16 and the pitch length ( defined as the length @xmath9 for one full turn of a helix ) .
we keep the helical radius fixed at @xmath17 nm and thus optimize the ratio @xmath18 . in the shape
optimization a fluid density and viscosity of water at room temperature are used as @xmath19 kg / m@xmath20 and @xmath21 mpa s , respectively . comparing the different helical particles in fig .
[ directions ] we find that for all the pitch lengths considered and for the same applied external torque , the angular velocity of the helix is the highest of them all ( cf .
s3 ) because it displaces the least fluid during rotation .
also , the helix yields the greatest propulsion velocity when the pitch length is varied , as shown in fig .
[ fig2].a because it has the smallest circular area of cross - section resulting in the lowest drag along the long axis .
the maximum propulsion velocity for the helix is obtained around @xmath22 ( fig . [ fig2].a ) for @xmath8 .
keaveny and co - workers optimized the shape of propellers with an attached spherical cargo for maximum propulsion .
using the boundary integral formulation @xmath23 ( @xmath24 ) they observe maximum propulsion for a helical filament radius and helical radius ratio of 0.2 ( 0.4 ) @xcite .
as shown in fig .
s4 , due to the strong spatial asymmetry of the helices there is significant wobbling for @xmath25 . in order to achieve maximum propulsion , both wobbling and the drag force
must be minimized , and thus @xmath16 must be larger than unity . in fig .
[ fig2].b we show the propulsion velocity for two values of @xmath18 in the region of the velocity maximum as a function of @xmath16 , showing that the optimal number of turns is indeed @xmath26 .
this is consistent with the experimental result of walker _
@xcite , where maximum propulsion is achieved for about one full helical turn for micron sized particles .
in addition to maximizing the propulsion velocity , it is also useful to characterize the efficiency of the helix .
the dynamics at the low reynolds number of a particle is described by the generalized resistance tensor @xmath27 where @xmath28 and @xmath29 are the net force , torque , translational velocity , angular velocity and the friction matrix of the particle , respectively @xcite .
particles with spherical symmetry have zero non - diagonal terms in the friction matrix and diagonal terms correspond to the stokes translational and rotational drag coefficients . for a helix , when a torque is applied along its long axis , the rotational motion drives it forward along this same axis through the non - zero coupling term @xmath30 or @xmath31 . to characterize the tr coupling or the amount of propulsion for a given angular velocity
, we define the efficiency @xmath32 to be the constant of proportionality of the propulsion velocity and the angular velocity , @xmath33 , where @xmath34 is the helical radius . within the resistive force theory ( rft ) , there exists an approximate analytic result for the ( relative ) efficiency given by @xmath35 where @xmath36 is the pitch angle that is related to the pitch length and the helical radius by @xmath37 @xcite . in fig .
[ fig2].c we compare the numerically obtained results for the efficiency of a helix ( @xmath8 ) to that in eq .
( 2 ) and the efficiency calculated using the slender - body theory ( for the efficiencies of different shapes see fig .
s5 ) . in our data ,
the maximum efficiency is achieved at @xmath38 , which is about @xmath39 and @xmath40 times larger than that predicted by the rft and sbt , respectively .
these results clearly justify the use of the quantitatively accurate lbmd method to calculate the efficiency . in fig .
[ fig2].d we show numerical results for the efficiency as a function of the number of turns @xmath16 for two values of @xmath18 close to the optimum in fig .
[ fig2].a .
in contrast to eq .
( 2 ) , where there is no @xmath16 dependence , with full hydrodynamics we find nontrivial dependence of @xmath32 on @xmath16 .
the stability of the rotational motion and propulsion , and consequently the efficiency , increases with the number of turns from @xmath41 to @xmath42 , where there is a local maximum . beyond @xmath8 ,
the efficiency exhibits damped oscillations before saturating to a constant value , where both @xmath43 and @xmath14 asymptotically decrease as @xmath44 ( see fig .
s5 ) .
thermal fluctuations affect the motion of the helices by changing the orientation of the helix and interfere with the translational and rotational motion .
in this section we characterize the strength of the thermal fluctuations with respect to the rotational and translational motion . for the rotational motion ,
the rotational pclet number characterizes the strength of the driven rotation and the thermal effects , @xmath45 , where @xmath46 is the rotational diffusion coefficient along the long axis which is the ratio of the thermal energy and the viscous rotational drag coefficient @xmath47 . at steady state ,
the external torque is proportional to the angular coefficient with a constant of proportionality of @xmath48 , and therefore @xmath49 . in all of the simulations of the driven helix ,
@xmath50 hence the strength of the thermal fluctuations is weaker compared to the induced rotation by the external torque .
the effects of the thermal fluctuations on the rotational motion were observed to be negligible . in esi fig .
s6 , we show the orientations of the perpendicular axis in the presence of thermal fluctuations for the applied external torques that are @xmath51 kg m@xmath52/s@xmath52 . for the directed motion of the helical particles , the translational pclet number ( @xmath53 , where @xmath54 is the brownian tracer diffusion coefficient of the particle ) quantifies the strength of the propulsive versus random diffusive motion and is thus a relevant measure of thermal fluctuations .
it can be written as @xmath55 where @xmath56 is the effective size ( hydrodynamic radius ) of the particle .
as can been seen from eq .
( [ pecletnum ] ) , the effect of temperature is typically negligible for micron size particles due to the large pe for relevant velocities , and has not been considered to date .
however , at the nanoscale pe is not necessarily large and thermal fluctuations need to be fully accounted for .
helical turns and @xmath57 in simulations without thermal fluctuations , and with thermal fluctuations .
inset shows details of the data for the small pe number regime.,height=264 ] in the esi ( fig .
s9 ) we show how @xmath54 depends on @xmath16 .
as expected , pe grows linearly with the applied torque ( fig .
s10.a ) . to obtain directed motion , @xmath16 must be greater than unity to minimize wobbling . in fig .
[ method1-method2-propulsion - velocities - thermalized ] we present data for the propulsion velocity of an optimal nanohelix ( @xmath8 , @xmath17 nm and @xmath58 ) as a function of the external torque with and without thermal fluctuations .
the velocities are averaged over the ratio of the displacements and a time interval of @xmath59 s for 80 independent trajectories . from the data
we can conclude that directed motion becomes very well established for pclet numbers greater than approximately @xmath60 .
separation of the magnetic helical colloids with a radius @xmath61 @xmath3 m and number of turns @xmath62 has been experimentally demonstrated at high ratios ( @xmath63 ) of torque strengths and thermal energies @xcite .
we present results for the nanoscale helices driven at constant torque ( @xmath17 nm and @xmath57 ) with positive @xmath64 and negative @xmath65 helicities with thermal fluctuations at @xmath66 k. in the absence of the external torque , purely diffusive motion is observed . at constant external torque in the presence of thermal fluctuations , chiral particles with @xmath64 and @xmath65 helicities move in the two opposite directions parallel to their long axis . in fig .
[ displacement - thermalized - pn - helix - rev1p25].a , we show typical displacement trajectories of the @xmath64 and @xmath65 helices with @xmath67 and @xmath68 helical turns for various torque strengths . at @xmath69 kg m@xmath70s@xmath52 ,
the angular velocity obtained is not sufficient to separate the chiral particles with @xmath41 and @xmath8 . at a higher torque @xmath71 kg m@xmath70s@xmath52 ,
the longer helical particles ( @xmath8 ) get separated while chiral separation for the short particles is still not possible . in general
, we conclude that pclet numbers below about @xmath72 are definitely not sufficient to separate chiral particles .
this result is consistent with the experimentally demonstrated chiral separation of microscale bacteria in sheared flows @xcite .
we have shown here that helical particles with a circular cross - section characterized by a minimum surface area result in optimal propulsion in terms of maximising the propulsion velocity .
a circular cross - section of the flagella , often naturally encountered in many organisms , is thus optimized for propulsion . controlled propulsion in the presence of thermal fluctuations requires stability in the rotational motion ( _ i.e. _ , minimum wobbling ) which increases with the number of turns @xmath16 .
however , increase in the length adds to the hydrodynamic drag .
therefore , maximum propulsion is achieved by balancing the competing requirements for maintaining the stability of the rotational motion and minimizing viscous drag .
thus , the optimal number of turns , @xmath8 , required for obtaining maximum propulsion velocities in the high pclet number limit reported here , is consistent with the previously published experimental results for microscale helices @xcite . at the nanoscale where thermal effects are significant
, we have shown that for well - defined directed motion and chiral separation of magnetic particles , the pclet number should definitely be larger than @xmath72 , in good agreement with the experiments on the chiral separation of microscale objects reported earlier @xcite .
the multiscale hybrid lattice - boltzmann - molecular dynamics ( lbmd ) method implemented in lammps is used in simulating the propulsion of the helical nanoparticles in the fluid @xcite .
the method incorporates full navier - stokes hydrodynamics with consistent coupling of the md particles to thermal fluctuations in the fluid .
the method has been extensively benchmarked and tested for colloids and polymers @xcite .
we use physically extended helical nanoparticles .
they are represented as composite particles with surface nodes that are treated as md particles .
the nodes are equally distributed on the surface with the center line following the parametric equation @xmath73 .
the cross - sectional shapes are chosen to cover the extreme ranges of the flat shaped helices with elliptical cross - sections , helicoid a ( b ) , that have the long - axis that are parallel ( perpendicular ) to the radial direction and the circular - cross section case , the helix fig .
[ cross - section ] . the cross - sectional shape is given by @xmath74 where @xmath75 and @xmath76 for the helix , @xmath77 and @xmath78 for helicoids a and b. the node masses are set such that the total mass of the helix is equivalent to the mass of the displaced fluid and the area per node is given by 18.5 nm@xmath52 . in order to couple the md particles to the fluid
, the method introduces forces on both the md particles and the fluid that result from an energy and momentum conserving interaction @xcite .
these forces are calculated by assuming an elastic 3d collision between an md node and a representative fluid particle .
the representative fluid mass and momentum are interpolated at the location of the node that it interacts with using a peskin interpolating stencil .
@xmath79 where @xmath80 and @xmath81 ( @xmath82 and @xmath83 ) are the node and fluid mass ( velocity ) , respectively @xcite .
an explicit no - slip boundary condition at the fluid - particle interface is implemented in all the simulations .
this boundary condition sets the sum of the initial and the final relative fluid - node velocities to zero .
all simulations are carried out with a lattice discretization of @xmath84 nm and @xmath85 ps .
the rotation of a helical particle in the simulations can be induced using two different methods .
a constant torque is applied along the long axis of the helix .
this corresponds to the synchronous regime wherein the frequency of rotation of the helix matches the frequency of rotation of the external magnetic field that exerts a torque on it @xcite . in the absence of thermal fluctuations , this method is employed as a means to optimize the geometrical shape for the helical particle and the fluid viscosity that result in optimal propulsion and efficiency .
mariana medina - snchez , lukas schwarz , anne k meyer , franziska hebenstreit , and oliver g schmidt .
cellular cargo delivery : toward assisted fertilization by sperm - carrying micromotors .
, 16(1):555561 , 2016 .
debora schamel , andrew g mark , john g gibbs , cornelia miksch , konstantin i morozov , alexander m leshansky , and peer fischer .
nanopropellers and their actuation in complex viscoelastic media . , 8(9):87948801 , 2014 .
this work was supported in part by the academy of finland through its centres of excellence programme ( 2012 - 2017 ) under project no .
251748 and aalto energy efficiency research programme .
we acknowledge the computational resources provided by aalto science - it project and csc - it .
the graphical representations in the fig . [ directions ] were rendered using vmd @xcite .
|
controlling the motion of nano and microscale objects in a fluid environment is a key factor in designing optimized tiny machines that perform mechanical tasks such as transport of drugs or genetic material in cells , fluid mixing to accelerate chemical reactions , and cargo transport in microfluidic chips .
directed motion is made possible by the coupled translational and rotational motion of asymmetric particles . a current challenge in achieving directed and controlled motion at the nanoscale lies in overcoming random brownian motion due to thermal fluctuations in the fluid .
we use a hybrid lattice - boltzmann molecular dynamics method with full hydrodynamic interactions and thermal fluctuations to demonstrate that controlled propulsion of individual nanohelices in an aqueous environment is possible .
we optimize the propulsion velocity and the efficiency of externally driven nanohelices .
we quantify the importance of the thermal effects on the directed motion by calculating the pclet number for various shapes , number of turns and pitch lengths of the helices .
consistent with the experimental microscale separation of chiral objects , our results indicate that in the presence of thermal fluctuations at pclet numbers @xmath0 , chiral particles follow the direction of propagation according to its handedness and the direction of the applied torque making separation of chiral particles possible at the nanoscale .
our results provide criteria for the design and control of helical machines at the nanoscale .
most natural organisms and molecules exhibit chiral structures that significantly influence their chemical and physical behavior .
for example , most proteins are left - handed while sugars are right - handed .
drug molecules which target proteins , therefore , have different effects depending on their chirality .
thalidomide is perhaps the most infamous example of a drug whose devastating health effects have been ascribed to the presence of both left and right handed enantiomers @xcite .
motion and kinetics of chiral objects are largely determined by their spatial anisotropy since it leads to a coupling between the translational and rotational ( tr ) degrees of freedom . in biological systems ,
tr coupling is utilized by microorganisms that propel themselves in fluids by rotating their flagella and generating a chirality .
artificially manufactured chiral particles with coupled tr motion provide many opportunities for technological applications @xcite .
recently , clemens _
et al . _ have experimentally demonstrated separation of racemic mixtures by exposing left and right - handed chiral molecules to a rotating electric field .
the dipole moments of the molecules in the racemic mixture align with the rotating electric field leading to their rotational motion .
this propels the chiral molecules via the tr coupling and gives rise to enantiomer separation depending on their handedness @xcite .
controlled and directed motion of chiral particles has important applications in targeted delivery of genetic material and drugs , too , because such particles can be tailored to facilitate their efficient uptake by various cells or tissues @xcite .
tottori _
et al . _
demonstrated controlled motion and cargo transport of magnetic helical particles @xcite .
application of time - dependent magnetic fields on magnetic screw - like particles induces a torque that rotates the particle and propels it in the fluid .
manipulation of the direction of the rotating magnetic fields makes it possible to load microspheres as cargo into the holder of a helical particle , transport the cargo and then release it at a preferred site @xcite .
medina - snchez _ et al .
_ have even demonstrated the possibility of assisted sperm delivery on immotile live sperm using externally driven microhelices @xcite .
walker _
et al .
_ , on the other hand , focused on optimizing the propulsion velocity of helices with a payload by varying the helical length @xcite . in the experiments above
, the propellers are typically @xmath1 nm wide and @xmath2 @xmath3 m long @xcite . on the theoretical side ,
exact analytic results for the drag coefficients of certain high - symmetry shapes such as spheres and ellipsoids can be obtained in the stokes limit . for more complex shapes such as chiral particles ,
the boundary integral formulation of the stokes equation can be used @xcite .
the typical rotational reynolds number in experiments of driven magnetic helices at the micron scale and at the nanoscale are about @xmath4 and @xmath5 @xcite .
hence zero reynolds number approximation is justified .
the resistive force theory ( rft ) and the slender body theory ( sbt ) are the usual approaches used to study the propulsive motion of rigid helices and flagellar propulsion .
the rft works under the assumption that the segments of the helical body do not generate fluid disturbances and consequently these segments do not influence the motion of the other segments of the helical body .
the hydrodynamic forces on an infinitesimal segment are proportional to the local velocity of the body @xcite by drag coefficients .
the calculations for the propulsive velocity of rigid helices are carried out by summing the drag and thrust contributions of the rotating helical particles @xcite .
the main weakness of the rft lies in the neglect of the long - range hydrodynamic interactions .
the sbt incorporates these hydrodynamic interactions by considering the fluid response due to the moving segments of the helical body by assuming a local point force represented by stokeslets that are uniformly distributed along the center line of the helical body . to account for the no - slip boundary condition ,
corresponding doublets are also uniformly distributed along the center line @xcite .
the regularized sbt assumes that the force is not a dirac delta at a single point but uses a shape function that modulates the force to be spatially distributed in a small blob around a point @xcite . in terms of the propulsion of microorganisms by a helical flagellum ,
the numerical results of the sbt and the regularized sbt are similar @xcite .
recently , artificial helical particles at the nanoscale with radii less than @xmath6 nm have become available @xcite .
controlled propulsion has been experimentally demonstrated in a biological gel and in a high viscosity glycerol - water mixture ( 25 cp ) but at low viscosities such as pure water , brownian motion was greater than the propulsion of the nanohelix such that no directed motion was observed @xcite .
the challenge in controlling nanohelices is overcoming thermal effects which can significantly change the direction of motion and hinder the propulsion . their coupled tr motion in aqueous solutions
have not been characterized yet .
open questions include how the chiral geometry and the fluid properties affect the efficiency and the propulsion speed in cargo delivery . here , we address these issues by employing the recently developed fluctuating lattice - boltzmann - molecular dynamics ( lbmd ) method @xcite .
the method incorporates full navier - stokes hydrodynamics with consistent thermal fluctuations and a coupling of the fluid to extended md particles of arbitrary shapes .
since its development , the fluctuating lbmd has been extensively benchmarked for colloids and polymers @xcite .
we consider externally driven helices with radii of @xmath7 nm , optimize the shape parameters and identify the conditions for controlled and directed motion , and chiral separation of helices at the nanoscale in general ( typical values of the reynolds numbers in the simulations are shown in the fig .
s1 ) .
our results provide clear criteria for the physical parameters such as the pclet number and fluid viscosity essential to studying applications of driven helical particles in nanoscale systems . , helical turns @xmath8 and pitch length @xmath9 .
the unit vectors @xmath10 , @xmath11 and @xmath12 represent coordinates in the laboratory frame , while @xmath13 and @xmath13 represent the coordinates in the body frame of the helical particles .
, scaledwidth=35.0% ]
|
in this paper i present an overview of the predictions for pb - pb collisions at @xmath0 tev at the large hadron collider , mainly based on the material presented during the workshop heavy ion collisions at the lhc - last call for predictions , held in the frame of the cern theory institute from may 14th to june 10th 2007 @xcite . such compilation ( see also @xcite , and @xcite for the analogous one for rhic ) should be useful for : ( a ) distinguishing pre- from post - dictions ; ( b ) assuming that a model tested at rhic ( and eventually sps ) energies can be extrapolated to the lhc , the huge lever arm in energy provides very strong constraints ; and ( c ) providing a frozen image of our present understanding of ultra - relativistic heavy ion collisions .
i have classified the predictions in : bulk properties ( section 2 ) : multiplicities ( see pre - rhic predictions in @xcite ) , azimuthal asymmetries , hadronic flavor observables and correlations at low transverse momentum ; and hard and electromagnetic probes ( section 3 ) : high transverse momentum observables and jets , quarkonium and heavy quarks , and leptonic probes and photons , extensively documented in @xcite .
unless otherwise stated , the predictions presented here can be found in @xcite and will be referenced by the name of their first author .
finally i draw some conclusions .
multiplicity is a first - day observable .
it plays a central role , as practically all other predictions require directly or indirectly such input . in fig .
[ armestofig1 ] sixteen predictions for charged hadron multiplicities at mid - rapidity in central collisions can be found . to compare the different predictions more accurately ,
we re - scale them to a common observable ( @xmath1 ) and centrality class ( @xmath2 ) using the model @xcite .
the re - scaling factors can be read from table [ armestotable1 ] and the corrected results found in fig .
[ armestofig2 ] .
.results in the monte carlo code in @xcite of the mean impact parameter , number of participants and binary nucleon - nucleon collisions , and charged multiplicity at mid-(pseudo-)rapidity , for different centrality classes defined by the number of participants . [ cols="^,^,^,^,^,^",options="header " , ] predictions can be roughly classified into those based on saturation physics ( abreu et al . ,
albacete , armesto et al .
, eskola et al . , fujii et al . , kharzeev et al . ) , data - driven predictions ( busza , jeon et al . ) , those based on percolation ideas ( dias de deus et al . ) , those containing strong shadowing ( capella et al . ) , monte carlo models ( topor pop et al .
, bopp et al . , chen et al . , porteboeuf et al . )
in which many physical mechanisms are combined , parton cascades ( el et al . ) , and those based on diffusion equations ( wolschin et al . ) .
they tend to lie in the range @xmath3 , the lowest values corresponding to extreme saturation models and data - driven predictions .
these values are generically much lower than pre - rhic predictions @xcite .
most models contain now a large degree of collectivity in the form of saturation , strong gluon shadowing , percolation , etc .
azimuthal asymmetries are another first - day observable .
generically , @xmath4-integrated @xmath5 is expected to increase in all models .
on the other hand , data - driven extrapolations @xcite tend to disagree with hydrodynamical models , and there are differences in the expectations for @xmath6 .
gathering the predictions in rough categories and focusing on the difference between @xmath5 at small @xmath7 gev / c at rhic and at the lhc , we find : * hydrodynamical predictions : drescher et al .
, on the basis of a simple model , predict deviations from ideal hydro seen at rhic , to diminish at the lhc .
bluhm et al .
kestin et al . and eskola et al .
predict @xmath5 at small @xmath4 to be similar or slightly smaller than at rhic on the basis of ideal hydrodynamics . increasing viscous corrections
would further diminish @xmath6 , although the initial conditions for evolution have to be settled .
* non - hydro predictions : porteboeuf et al .
find a similar @xmath6 at the lhc than at rhic , in the epos model . on the other hand , chen et al . in the ampt model , and the absorption model of capella et al .
, show a moderate to strong increase .
the parton cascade of molnar indicates that at fixed viscosity - to - entropy ratio , @xmath6 should decrease with increasing energy .
this offers a possible interpretation of the results in ampt or absorption models in terms of an effective diminishing of the viscosity - to - entropy ratio due to the increasing parton / particle density .
three different groups of observables can be discussed : * particle ratios are usually studied in the framework of statistical hadronization models . both andronic et al . and kraus et al .
predict all antiparticle - to - particle ratios to be very close to 1 for the small @xmath8 gev expected at the lhc at mid - rapidity , with only @xmath9/p being slightly smaller .
the latter group shows the sensitivity of multi - strange - to - non - strange hadron ratios on the temperature , and to deviations from the grand - canonical ensemble .
finally , rafelski et al .
show how non - equilibrium scenarios reflect on an increase of multi - strange hadron yields and a decrease of non - strange resonances , for several total multiplicity predictions .
so different statistical hadronization scenarios could be distinguished .
this becomes of capital importance in regeneration models for open and hidden heavy flavor production , as shown by the differences between the equilibrium ( andronic et al . ) and the non - equilibrium ( kuznetsova et al . ) predictions .
* net baryon number at mid - rapidity : p@xmath9 is generically predicted @xmath10 in a wide variety of models : those including the baryon junction mechanism like hijing b / bbar ( topor pop et al . ) or dpmjet ( bopp et al . ) , ideal hydro ( eskola et al . ) , epos ( porteboeuf et al . ) and rdm ( wolschin et al . ) . * baryon - to - meson ratios : ideal hydro ( kestin et al . ) and recombination ( chen et al . )
models predicts larger values than models which consider the creation of large color fields ( topor pop et al . , cunqueiro et al . ) .
the latter predict the cronin effect for protons to survive at the lhc .
both observables will clarify the hadronization and baryon - transport mechanisms . concerning the hbt radii , all models ( chen et al .
in ampt , kestin et al . with ideal hydro , the non - equilibrium scenarios of sinyukov et al . and karpenko et al . , and the recent work @xcite ) predict an increase , although the density ( i.e. total multiplicity ) dependence of such increase varies from model to model .
nevertheless , the predictive power is limited by the problems that the models face to reproduce rhic data ( i.e. the @xmath11-dependence of @xmath12 and @xmath13 , and their relative magnitude ) , and the poor understanding of the role of dissipative effects .
before discussing the predictions , let me stress that the control of the benchmark for hard and electromagnetic probes will be at the lhc as key an aspect as it was at rhic . for example , a basic ingredient as the extrapolation of nuclear parton densities to the lhc kinematical region , is under poor control .
this can be seen e.g. in the factor @xmath14 uncertainty in the nuclear effects for the gluon distribution in pb at @xmath15 that can be extracted from the comparison of the results shown in @xcite and the recent work @xcite . in fig .
[ armestofig3 ] , i gather thirteen predictions for the nuclear suppression factor @xmath16 in central collisions , for two values of @xmath17 and 50 gev / c .
predictions can be roughly classified into those based on radiative and/or collisional energy loss in pqcd ( dainese et al .
, renk et al .
, qin et al . , wicks et al . , jeon et al . , liu et al . ,
lokhtin et al . , vitev , and wang et al . ) , those based on absorption scenarios ( capella et al . and
kopeliovich et al . ) , those based on percolation ( cunqueiro et al . ) and the opaque core - transparent corona model of pantuev .
the generic expectation in pqcd - based models is that @xmath18 and increasing with increasing @xmath4 . at @xmath17 (
filled symbols ) and 50 ( open symbols ) gev / c from different models @xcite . on the right
the authors of the predictions , the hadron species for which @xmath19 has been calculated , the centrality class , the density / multiplicity input , an indication in case the highest @xmath4 is 40 gev / c instead of 50 , and a brief description of the model ingredients , are provided .
the error bars reflect the uncertainties coming from the multiplicity / density input but in capella et al .
, qin et al . and wang et al . , where they correspond to the variation of kinematics for particle production , @xmath20 and nucleon / nuclear parton densities respectively.,height=427 ] another aspect is the hadro - chemistry at large @xmath4 , which in pqcd - based models is expected to be modified both by the different energy loss of quarks and gluons ( barnafoldi et al . ) , by the modification of the parton cascade inside the medium ( sapeta et al . ) , and by additional mechanisms as conversions ( liu et al . ) .
finally , jet reconstruction and other differential observables like the yield in the away - side peak ( see predictions in lokhtin et al . and wang et al .
respectively ) will offer new possibilities to constraint models .
all these aspects will help to verify the mechanism underlying the jet quenching phenomenon observed at rhic .
concerning heavy quarks , radiative energy loss models predict that the modification of the suppression pattern due to the mass of the quark will be visible in @xmath21 for heavy mesons and double ratios @xmath22 up to quite large @xmath23 gev / c ( armesto et al .
, wicks et al . ) .
these doubles ratios have also discriminatory power on the mechanism for energy loss ( horowitz et al . ) .
a common feature to all pqcd - based mechanisms is that hadronization is supposed to happen outside the medium , which could not be the case for low @xmath4 , where different mechanisms might be at work ( van hees et al . ,
vitev ) .
concerning quarkonium suppression and production , the main uncertainty in recombination models for the magnitude of the suppression or enhancement lies in the heavy quark cross section ( andronic et al . ) . on the other hand ,
@xmath4-broadening as a tool to verify the recombination mechanism ( thews et al . ) suffers from large uncertainties due to cold nuclear matter effects which should be clarified ( kang et al . ) before any firm conclusion can be drawn .
the pattern of the suppression of the different quarkonium states versus @xmath4 should be measurable at the lhc .
it offers a possibility to distinguish ( vogt ) the different scenarios of sequential melting derived from either potential models or lattice computations , as well as new possibilities suggested in the framework of super - symmetric versions of qcd at strong coupling ( h. liu et al . ) .
finally , predictions which could be directly compared with experimental data , based on co - mover dissociation plus recombination , can be found in @xcite . in any case , the considerable uncertainties in the mechanisms of quarkonium production and suppression limit our predictive power . concerning photon production at large @xmath4 , predictions in the frame of pqcd - based models containing radiative energy loss are available ( arleo , vitev ) . at small @xmath4 , other mechanisms as thermal production ( arleo et al . ) , conversions @xcite or different types of factorization ( rezaeian et al . )
could be at work .
the measurement of @xmath5 for photons also offers some discriminatory power .
ideal hydro models ( chatterjee et al . ) indicate that @xmath5 below @xmath24 gev / c should be dominated by the partonic phase of the expansion of the medium . at the lhc , both high- and low-@xmath4 photons
will be measurable , but an accurate understanding of the benchmark will be key to disentangle a thermal contribution over the other possible , background effects . concerning dileptons , nayak et al .
propose the ratio real - to - virtual photons as sensitive to the temperature of the initial phase and weakly dependent on the details of expansion , eos , etc .
dremin proposed the cherenkov effect as a mechanism which could mimic an apparent broadening of the @xmath25 .
finally , predictions for the dilepton mass spectrum for both large- ( fries et al . ) and small-@xmath4 ( van hees et al . ) regions , are available .
the intermediate mass region @xmath26 gev / c@xmath27 suffers from a large background from heavy meson semi - leptonic decays .
therefore the understanding of the benchmark looks crucial again to disentangle a thermal component .
we are getting closer and closer to the moment when rhic - tested models will be confronted to lhc experimental results .
for every observation at rhic which is considered one of its major discoveries , an interpretation has been proposed and the corresponding predictions for the lhc have been computed . in this way : * the charged multiplicity at mid - rapidity , sizably lower than expectations @xcite , is interpreted in terms of a large degree of coherence or collectivity in particle production .
it leads to predictions for central collisions at the lhc smaller than 2000 charged particles per unit pseudo - rapidity at @xmath28 . *
@xmath5 in agreement with ideal hydrodynamical models is interpreted as the creation of an almost perfect fluid , leading to the prediction that @xmath5 at small @xmath4 will be at the lhc similar or smaller than at rhic . *
the observed strong jet quenching leads to the conclusions that the created medium is dense , partonic and opaque , resulting in an @xmath18 for central collisions at the lhc .
this picture has already motivated many theoretical developments : applications of ads / cft , mechanisms for early thermalization , viscous hydro , the cgc , etc .
major deviations from the generic expectations outlined here will enlarge our understanding of the physics of ultra - relativistic heavy ion collisions - note that naive , data - driven extrapolations @xcite tend to disagree with predictions from successful models at rhic .
besides , the lhc experiments , with their unprecedented capabilities @xcite , will provide new observables like jets or higher quarkonium states - not or only marginally measurable at rhic - for which a large theoretical effort is still required .
this work has been supported by mec of spain under a ramn y cajal contract and grant fpa2005 - 01963 , by the spanish consolider - ingenio 2010 programme cpan ( csd2007 - 00042 ) , and by xunta de galicia ( consellera de educacin and grant pgidit07pxib206126pr ) .
i thank the organizers for such a nice conference , j. albacete , f. bopp , w. busza , l. cunqueiro , a. dainese , a. el , k. eskola , u. heinz , c .- m .
ko , i. lokhtin , g. milhano , c. pajares , v. pantuev , t. renk , v. topor pop , r. venugopalan , i. vitev , x. n. wang , k. werner and g. wolschin for feedback on their predictions , and j. albacete , j. casalderrey - solana , k. eskola , e. ferreiro , u. heinz , p. jacobs , x. n. wang and u. wiedemann for discussions .
special thanks go to n. borghini , s. jeon and u. a. wiedemann who co - organized together with me the cern workshop , and to c. salgado for discussions and a critical reading of this manuscript .
10 armesto n _ et al . _ 2008 054001 borghini n and wiedemann u a 2008 023001 bass s a _ et al .
_ 1999 a * 661 * 205 armesto n and pajares c 2000 _ int .
a * 15 * 2019 accardi a _ et al .
_ 2003 _ preprint _ arxiv : hep - ph/0308248 accardi a _ et al .
_ 2003 _ preprint _ arxiv : hep - ph/0310274 bedjidian _
_ m 2003 _ preprint _ arxiv : hep - ph/0311048 arleo f _ et al .
_ 2003 _ preprint _ arxiv : hep - ph/0311131 amelin n s , armesto n , pajares c and sousa d 2001 _ eur .
c * 22 * 149 ahmad n _ et al . _
[ alice collaboration ] 1995 _ preprint _ cern / lhcc 1995 - 071 chojnacki m , florkowski w , broniowski w and kisiel a 2007 _ preprint _ arxiv:0712.0947 [ nucl - th ] eskola k j , paukkunen h and salgado c a 2008 _ preprint _
arxiv:0802.0139 [ hep - ph ] capella a , bravina l , ferreiro e g , kaidalov a b , tywoniuk k and zabrodin e 2007 _ preprint _
arxiv:0712.4331 [ hep - ph ] liu w and fries
r j 2008 _ preprint _ arxiv:0801.0453 [ nucl - th ] carminati f _ et al . _
[ alice collaboration ] 2004 1517 alessandro b _
[ alice collaboration ] 2006 1295 denterria d g _ et al .
_ 2007 2307 steinberg p [ atlas collaboration ] 2007 s527
|
i present an overview of predictions for the heavy ion program at the large hadron collider .
it is mainly based on the material presented during the workshop heavy ion collisions at the lhc - last call for predictions , held in the frame of the cern theory institute from may 14th to june 10th 2007 .
predictions on both bulk properties and hard probes are reviewed .
|
spatial disorder is a generic feature of pattern formation . whereas the simplest patterns that appear in numbers of physical or chemical systems driven out of thermodynamic equilibrium are periodic structures ( such as stripes , squares or hexagons ) , in practice , many steady patterns observed in spatially extended systems are more complex and disordered at large scale @xcite .
why disordered configurations spontaneously form ( without the help of impurities or any quenched disorder ) and do not asymptotically evolve with time toward more ordered states is still not fully understood .
for instance , systems where various ( say , two ) patterns of different symmetries coexist can easily lead to disordered states . a basic example is rayleigh - bnard convection of non - boussinesq fluids , where hexagonal patterns bifurcate subcritically and are stable along with the uniform conductive state within a whole interval of rayleigh numbers . in bistability regions , asymptotic configurations may be either homogeneous ( containing a single phase ) or heterogeneous , with both phases present and distributed in a non - regular way .
heterogeneous states have been identified experimentally in rayleigh - bnard convection @xcite , vibrated granular materials @xcite , or turing patterns and related gas - discharge systems @xcite .
spatial disorder has been also proposed as a generic feature of pattern formation in ecosystems , as illustrated by vegetation patches in arid regions @xcite . on the theoretical side ,
the usual nonlinear model equations for periodic patterns often predict metastable disordered states as well .
regarding bistable / subcritical patterns , a lot of attention has been devoted in the past years to localized structures , that are axisymmetric isolated structures of well - defined shape and size immersed in a uniform quiescent background ( of vanishing order parameter ) @xcite .
localized structures have been identified as a possible origin of disorder in bistable systems , as asymptotic steady states are in some case composed of many localized structures randomly distributed in space @xcite .
the stability of these spatially chaotic " states , as opposed to periodic , can be explained by the particular nature of the effective pair interaction between two localized structures @xcite . in the present paper , we consider the swift - hohenberg model for hexagonal patterns in the bistability zone mentioned above and study spatially more extended objects , namely , islands with hexagonal order surrounded by the quiescent background .
a localized structure can be seen as a limiting case where such an domain is of very small size ( of order @xmath0 , the wavelength of the base pattern ) .
an interesting question to ask is up to what point localized structures and large islands differ from each other , although they probably share closely related features .
we therefore analyze in the following the dynamics of a planar interface separating an extended domain of hexagons and a uniform phase in the bistability region .
although interfaces have been relatively less studied than localized structures in this context @xcite , we show that they play a central role in the large time evolution of two - dimensional extended systems , and in the formation of heterogeneous disordered states . after recalling a few known results on the weakly nonlinear regime ( section [ recall ] ) , we show that in the nonlinear regime interfaces can either move ( with two modes : expanding or shrinking domains ) , or be pinned , depending on the value of the main control parameter ( section [ ampeq ] ) .
the theoretical treatment is based on an extension of the usual amplitude equation formalism to include non - adiabatic `` effects , according to some ideas presented in @xcite .
some analogies of this system with interfaces in solid crystals are pointed out .
the results are further confronted with numerical resolution of the swift - hohenberg equation in section [ numerical ] , where a dynamical diagram for interfaces is drawn in parameter space . in section [ disorder ] , we discuss the stability of heterogeneous frozen states .
we show that the stability region of localized structures only partially overlap the pinning regime of interfaces .
we deduce that the existence of steady disordered configurations ( and therefore of localized structures of infinite life - time ) is possible in practice only in that pinning regime .
disordered frozen configurations obtained from random initial conditions can be of various types : they can contain localized structures , or localized structures and extended domains of hexagons ( mixed '' disorder ) , or extended domains and no localized structures .
changing the control parameters can drive the system outside of the pinning zone , keeping localized structures stable : these are absorbed by growing domains and therefore become metastable .
conclusions are presented in section [ conclusions ] .
we consider a system described by a dimensionless local order parameter @xmath1 , representing , for example , the mid - plane vertical velocity in convection problems .
the spatio - temporal evolution of @xmath2 in non - boussinesq convection can be described by a generalized swift - hohenberg equation @xcite , @xmath3 the parameter @xmath4 is the main control parameter and represents the reduced rayleigh number @xmath5 , where @xmath6 is the critical rayleigh number .
@xmath7 is the characteristic wavenumber of the patterns , and @xmath8 the bare coherence length . in the case of convection
, @xmath8 is of order @xmath9 , its precise value depending on the boundary conditions at the fluid layer .
the coefficient @xmath10 of the quadratic term in eq.([sh ] ) models the strength of non - boussinesq effects ; @xmath10 will be taken as positive in the following without restricting generality ( @xmath11 ) .
equation ( [ sh ] ) , or close variants of it , have been studied in many contexts , for instance , to describe the microphase separation of asymmetric block - copolymer melts in the weak segregation limit @xcite or for degenerate optical parametric oscillators @xcite .
recently , a similar equation was introduced to describe the evolution of vegetation biomass ( the local order parameter in that case ) in arid climates @xcite .
eq.([sh ] ) can be written under a frequently used form : @xmath12 with @xmath13 we will take in the following @xmath14 and @xmath8 ( rather than @xmath14 and @xmath10 ) as the two main tunable parameters of eq.([sh ] ) , @xmath10 being fixed to an arbitrary small value .
the coherence length @xmath8 is thus considered as independent of @xmath0 _ a priori_. from ( [ rescal ] ) , decreasing @xmath8 is equivalent to increasing the bifurcation parameters @xmath15 and @xmath16 . keeping @xmath17
, the weakly nonlinear regime corresponds to @xmath18 , while the nonlinear regime is reached with @xmath19 . for @xmath20 ,
the uniform solution @xmath21 ( conductive state ) is linearly stable , while it becomes unstable for @xmath22 .
standard weakly nonlinear analysis shows that a subcritical bifurcation to stable hexagonal patterns takes place at @xmath23 .
hexagonal patterns remain stable up to @xmath24 , while stripes pattern become stable for @xmath25 @xcite . in the following ,
we restrict our study to the coexistence zone of the hexagonal and conductive states ( @xmath26 ) .
we come back in more details to a problem presented in @xcite .
we study a planar interface separating two semi - infinite domains , one composed of hexagonal patterns and the other in the uniform state @xmath21 .
the local order parameter can be approximated at leading order ( @xmath27 ) as @xmath28 where the amplitudes @xmath29 are slowly varying envelopes depending on @xmath30 , the coordinate along the axis @xmath31 normal to the interface .
the solutions of eq.([sh ] ) satisfy the condition @xmath32 the wave vectors @xmath33 are such that @xmath34 , and make an angle of @xmath35 with each other .
we choose @xmath36 with an orientation perpendicular to the interface : this situation corresponds a faceted " hexagonal pattern having a raw parallel to the interface , as schematically displayed in fig.[interface ] .
faceted configurations are the more relevant ones , as further discussed in section [ numerical ] .
when @xmath17 , @xmath37 and @xmath38 , the equations satisfied by the amplitudes @xmath29 can be derived from eq.([sh ] ) by standard multiple scale analysis @xcite : @xmath39 ( @xmath40 is the complex conjugate of @xmath41 ) .
note that @xmath42 from the planar geometry of the interface .
we next recall some known results on the velocity @xmath43 of the interface .
use the ansatz @xmath44 , multiply eq.([amp10 ] ) ( resp .
eqs.([amp20 ] ) and ( [ amp30 ] ) ) by @xmath45 ( resp . @xmath46 and @xmath47 ) , add the equations and integrate their real part over @xmath30 .
the velocity is obtained as @xcite : @xmath48}{2d},\ ] ] where @xmath49 is the amplitude of the regular hexagonal pattern @xmath50 and @xmath51 the numerator of eq.([v0 ] ) can be interpreted as a constant driving force acting on the interface , while @xmath52 ( which depends on the whole interface profile ) plays the role of a friction term .
the system of equations ( [ amp10])-([amp30 ] ) has a potential form : there exists a functional free - energy " @xmath53 such that @xmath54 . the driving force in eq.([v0 ] )
( plotted below in fig.[fig : pth ] as a function of @xmath14 ) is proportional to the difference of free - energy density @xmath55 between the regular hexagonal state and the uniform state .
therefore , the sign of @xmath43 is such that the domain with the lower free - energy expands at the expense of the other . from eq.([v0 ] ) , @xmath10 being fixed , the velocity changes sign at @xmath56 , a value located in the coexistence region .
the uniform phase spreads if @xmath57 , whereas the hexagonal phase does in the much wider interval @xmath58 .
the interface is stationary only at @xmath59 .
therefore , except at the marginal value @xmath60 , the amplitude equation formalism at leading order predicts that spatially heterogeneous distributions of hexagonal and conductive regions should eventually evolve toward a single state , either hexagonal or conductive , through interface motion .
from eqs.([amp10])-([amp30 ] ) , the length scale of variation of the amplitudes , defining the interface thickness @xmath61 , is roughly of order @xmath62 suppose now that @xmath63 ( or @xmath64 ) is not that small compared to 1 , and/or @xmath8 is significantly smaller than @xmath0 : the assumption @xmath65 may not be fulfilled in that case .
the spatial variations of @xmath66 are not necessarily much slower than the local oscillations of the patterns , and so - called non - adiabatic `` effects can appear .
the effects of non - adiabaticity on moving interfaces have been studied theoretically for one - dimensional patterns @xcite , stripe - stripe interfaces @xcite , square - hexagon interfaces @xcite , and other cases @xcite .
non - adiabatic effects can be sufficient to pin interfaces and therefore to stabilize spatially heterogeneous configurations of the order parameter .
a weakly nonlinear analysis taking into account these effects in the present case is reported in appendix [ ap : a ] .
we derive the modified amplitude equations , with ( small ) non - adiabatic corrections that illustrate the coupling between the fast '' and slow " length scales . given the geometry depicted in fig.[interface ] , one obtains @xmath67 \frac{e^{-ik_0x}}{k_0}\nonumber\\ \frac{\partial a_2}{\partial t}&=&-\frac{\delta f}{\delta \bar{a}_2 } + i\partial_x\left(2g_2a_2\bar{a}_1 - 3a_2 ^ 2a_3 -3\bar{a}_1 ^ 2\bar{a}_3\right)\frac{e^{-ik_0x}}{k_0}\label{amp22}\\ a_3&=&a_2\label{amp32}.\end{aligned}\ ] ] in eqs.([amp12])-([amp32 ] ) , the @xmath68 s are the right - hand sides of eqs.([amp10])-([amp20 ] ) .
the above equations involve directly the space variable @xmath30 and therefore break the invariance by translation of standard amplitude equations .
this is a consequence of the coupling between the amplitudes and the local phases of the waves .
the additional terms in ( [ amp12])-([amp32 ] ) can not be derived from a usual multiple scale analysis .
equations ( [ amp12])-([amp32 ] ) have been derived in appendix [ ap : a ] in a lowest mode approximation ( valid if @xmath61 is still @xmath69 ) : the non - adiabatic terms oscillate in that case with the same wavenumber than the base pattern , @xmath70 .
other terms oscillating as @xmath71 should be kept if @xmath72 , a situation corresponding to the strongly nonlinear regime .
fixing the non - boussinesq parameter @xmath10 to an arbitrary small value , we determine in this section the finite interval of values of the control parameter @xmath14 such that interfaces do not move , or are pinned .
this interval depends on the coherence length @xmath8 , and can represent a significant part of the coexistence region when @xmath8 is small enough . to derive the modified law of motion of the interface , we use the method described in section [ recall ] . denoting @xmath74 the position of the interface and using the ansatz @xmath75 , one obtains from eqs.([amp12])-([amp32 ] ) , @xmath76 with @xmath77,\ ] ] and @xmath78,\end{aligned}\ ] ] @xmath52 being given by ( [ d ] ) . in eq.([v ] ) ,
the quantity @xmath79 is a dimensionless positive number ( @xmath80 ) that can be identified with the magnitude of a periodic pinning potential : the motion of the interface is analogous to that of a viscous particle sliding down over an inclined plane with undulations . from eq.([lawmotion ] ) , if @xmath81 , the interface moves at speed @xmath82 , although non - constant in time . when @xmath83 , the effective potential @xmath84 seen by the interface has many local minima : there exist a set of discrete stable positions @xmath85 for which @xmath86 .
the interface is pinned and hexagonal domains do not grow nor shrink .
two successive steady positions are separated by a distance @xmath0 .
this value for the periodicity ( result of a single mode approximation , see appendix [ ap : a ] ) has a simple geometrical interpretation : the closest steady configuration of the faceted hexagonal domain represented in figure [ interface ] is the one obtained by addition ( or symmetrically , removal ) of an interfacial layer " of dots parallel to the interface ( the gray dots of fig.[interface ] ) .
this situation is analogous to that of a slow solidification ( or melting ) process controlled by activation barriers @xcite . in expressions ( [ d ] ) and ( [ p ] ) we have assumed that the @xmath29 s are real ( their imaginary part is supposed to be small ) . to evaluate @xmath79
, we solve numerically the first order equations ( [ amp10])-([amp20 ] ) for the moving interface , and substitute in ( [ p ] ) .
although this solution is only approximate ( in the pinning regime , we expect some changes in the amplitude profiles due to the fact that interfaces are not moving ) , it captures the correct physical picture . a closer comparison with numerical solutions of eq.([sh ] ) is performed in section [ numerical ] below . in figure
[ fig : pth ] , we plot the driving force @xmath87 and the pinning potential @xmath79 given by eqs.([force])-([p ] ) as a function of @xmath14 , for a fixed value of @xmath10 . we show some curves of @xmath79 obtained for different values of the coherence length @xmath8 ( @xmath88 is independent of @xmath8 ) .
decreasing @xmath8 increases non - adiabatic effects and the pinning potential : from eq.([p ] ) , @xmath79 roughly behaves as @xmath89 with @xmath90 a numerical constant of order unity and @xmath61 given by ( [ l ] ) . at fixed @xmath8
, interfaces are pinned in a range @xmath91 , while hexagons shrink for @xmath92 and expand for @xmath93 , see fig.[fig : pth ] . for @xmath94 ( the weakly nonlinear regime ) ,
the size of the pinning interval is extremely small compared with @xmath95 , and limited to the vicinity of @xmath60 .
however , @xmath96 blows up and covers a significant part of the coexistence region at a cross - over value @xmath97 .
the present calculation predicts that nearly the whole coexistence region ( at least the interval @xmath98 $ ] ) has pinned interfaces for @xmath8 smaller than a critical value ( see fig.[fig : pth ] ) .
we define the crossover value @xmath73 as given by this critical value . in figure
[ fig : xi0p ] , we plot @xmath73 as a function of @xmath10 .
a linear relation is observed : @xmath99 since @xmath37 , @xmath100 in eq.([xic ] ) .
we conclude that pinning can be observed in practice only if the scale of variation of the amplitude is very short , corresponding to the _ nonlinear _ regime of eq.([sh ] ) , or equivalently if @xmath101 in eq.([sh2 ] ) . unfortunately , in that regime , the equations ( [ amp10])-([amp30 ] ) used to derive these results are no longer valid : higher order terms should be included to improve accuracy .
we therefore do not expect this calculation to give a quantitative picture of pinning effects , but rather a qualitative one .
an example of discrepancy between theoretical and numerical results ( presented in the following section ) is that the predicted pinning in the nearly whole coexistence region at low @xmath8 is not observed numerically . instead , pinning only occurs in the lower fraction of the coexistence interval , as further shown in sec .
[ numerical ] below .
nevertheless , the present weakly nonlinear analysis correctly captures the onset of a significant pinning regime for @xmath8 below a value @xmath73 , as well as the right order of magnitude for @xmath73 . in figure
[ fig : deps ] , we plot the relative size of the pinning regime @xmath102 , determined from theory and numerics , as a function of @xmath103 . due to the exponential form of eq.([estimp ] ) , the theoretical curve is nearly a step function at @xmath104 .
the numerical curve is softer , but with significant variations around @xmath104 as well .
we numerically solve eq.([sh ] ) by using a pseudo - spectral method and a time integration procedure described in @xcite .
the space is discretized on a square lattice , with @xmath105 nodes along @xmath31 , @xmath106 nodes along @xmath107 , and periodic boundary conditions .
the initial condition is a rectangular hexagonal domain of length @xmath108 and width @xmath106 , comprised in - between two quiescent regions of @xmath21 .
the two interfaces are faceted and parallel to the @xmath107 direction .
the lattice size @xmath109 set to unity .
the base period @xmath110 of the pattern is @xmath111 . to determine whether interfaces are pinned or
not , we compute the time evolution of the quantity @xmath112 with @xmath113 and @xmath114 the box dimensions .
@xmath115 is proportional to the length of the hexagonal domain , and therefore can be used to record the relative positions of the interfaces .
figure [ position ] shows transitions between moving and pinned boundaries as @xmath14 is decreased , for a given @xmath10 and @xmath8 .
as expected , hexagonal domains are either expanding , steady or shrinking . in the moving regimes near a pinning transition , the motion is noticeably non - uniform in time , in agreement with the form of the law of motion ( [ v ] ) .
figure [ fig : diag.int ] displays a @xmath116-diagram representing the pinned and moving regimes of hexagon / uniform interfaces .
the parameter @xmath10 is fixed to 0.2 , but the shape of the diagram is independent of this value , which just sets the scales of both axis ( @xmath64 for the @xmath30-axis and @xmath10 for the @xmath117-axis , from sections [ recall ] and [ ampeq ] ) . for comparison , we also report the characteristic values of @xmath14 given by weakly nonlinear analysis ( @xmath18 ) , introduced in section [ recall ] , and re - noted @xmath118 and @xmath119 here . when @xmath120 , the coexistence region delimited by the lower value @xmath121 becomes wider as @xmath8 decreases :
@xmath121 deviates significantly from the value predicted by weakly nonlinear analysis . yet
, the picture drawn in the previous section is qualitatively reproduced . at any @xmath8
, one can always identify two pinning - depinning transitions : for @xmath92 , hexagonal domains shrink ; for @xmath122 , interfaces are pinned ; for @xmath123 , hexagons expand . for @xmath8 small enough , the pinning interval @xmath124 quickly grows from almost zero and becomes comparable to @xmath95 .
the numerical cross - over value found numerically compares well with that given by formula ( [ xic ] ) , as shown in figure [ fig : deps ] .
the curve @xmath125 converges toward the value @xmath118 as @xmath8 increases , while @xmath126 and @xmath127 converge toward @xmath119 . on the other hand , @xmath121 and @xmath128
tend to merge as @xmath129 , a trend also reproduced by the theoretical curves of fig.[fig : pth ] . in the region located below the dotted line of figure [ fig : diag.int ] ( low @xmath8 and @xmath63 ) , fronts are moving , with possible instabilities toward labyrinthine patterns or non - trivial uniform phases .
this region is not of interest here and was not investigated .
one can show from section [ ampeq ] that interfaces have to be faceted to be pinned .
as detailed in the appendix , if none of the wavenumbers @xmath130 of the base pattern is directed along the normal to the interface , the nonadiabatic terms ( [ i ] ) will always integrate to zero along the transverse coordinate @xmath117 . in @xcite
, it was similarly argued that oblique interfaces could not be pinned .
we show here that oblique interfaces actually modify their shapes to become pinned .
more generally , in the pinning regime , an hexagonal domain surrounded by a conductive state @xmath21 evolves toward a polygonal shape compatible with its three wavenumber directions .
this situation can be seen as a nonequilibrium analogue of faceting in crystals due to strong anisotropy in surface tension @xcite . as an example
, we have solved numerically eq.([sh ] ) with an initial condition composed of a planar interface with a wavenumber @xmath36 making an angle @xmath131 with the interface normal ( figure [ facet]a ) .
the interface evolves toward a wavy , faceted shape ( figure [ facet]b ) .
from the preceding analysis , we conclude that spatially disordered configurations of the order parameter can be stabilized via pinning effects as soon as the coherence length is of order @xmath73 ( or @xmath16 in eq.([sh2 ] ) of order @xmath132 , given by ( [ g2p ] ) ) .
these configurations can contain immobile , extended hexagonal domains with faceted boundaries . in this subsection
, we wish to compare the previously determined dynamical regimes for planar interfaces with the stability diagram of localized structures .
localized structures are axisymmetric ( dot - like ) structures of well defined shape and size frequently encountered in bistability regions .
they have been widely studied and have been recognized to play an important role in the production of disordered states @xcite .
the stability of assemblies of localized structures irregularly distributed in space is relatively well understood .
the effective interaction force between two localized structures changes sign and oscillates with their separation distance ( a property reminiscent of eq.([v ] ) for interfaces ) , therefore allowing many stable discrete positions that are unlikely to lead to periodic arrangements of structures @xcite .
we show below that the likelihood of observing localized structures in disordered solutions of eq.([sh ] ) in fact largely depends on whether planar boundaries are pinned or not .
we find that localized structures can actually coexist with pinned extended domains .
in other cases , they are metastable , _
i.e. _ susceptible to be swept by any expanding domain of hexagonal order .
we have determined the stability diagram for a single localized structure .
calculations were performed on a square of @xmath133 grid points over an extended period of time ( 20000 time units ) , and with a small additive noise term in eq.([sh ] ) .
the initial localized structure ( a region of size @xmath134 with @xmath135 ) can either shrink , converge to a stable form , or expand .
when localized structures expand , they generate hexagonal order , as described in details in refs .
@xcite . similarly to extended domains , we find that localized structures shrink if @xmath136 , are stable if @xmath137 , and expand if @xmath138 .
the results are reported in the @xmath139-plane in figure [ fig : diag.int.ls ] , together with the data of figure [ fig : diag.int ] for planar fronts .
note that the stability zone of localized structures is located toward the right of the pinning regime of interfaces .
the stability interval @xmath140 and the pinning interval @xmath96 for boundaries become significant in the same range of values of the coherence length @xmath8 .
the hatched region of fig.[fig : diag.int.ls ] indicates where both intervals overlap . in this region of parameters ,
localized structures actually have an infinite life - time . anywhere outside this zone , localized structure are either unstable , or metastable , _
i.e. _ susceptible to be swept by a moving front of expanding hexagons . also notice that when @xmath141 is of order 1 , @xmath142 tends to @xmath143 .
therefore , localized structures always shrink in the weakly nonlinear regime . to further illustrate these results ,
we have solved the swift - hohenberg equation ( [ sh ] ) with @xmath144 and random initial conditions , in various regions of figure [ fig : diag.int.ls ] . for convenience ,
we fix @xmath145 in the vicinity of @xmath73 .
the initial condition for @xmath2 is given by a random gaussian field of zero mean and variance @xmath146 , with @xmath147 a dimensionless tunable parameter of order 10 .
inside the hatched region of figure [ fig : diag.int.ls ] @xmath148 and with @xmath14 _ relatively closer to @xmath149 _ , the asymptotic patterns obtained are mainly composed of aperiodic assemblies of localized structures of liquid - like character , with no visible hexagonal order .
the example shown in figure [ frozen]a at @xmath150 ( for a system of size @xmath151 ) is similar to the patterns presented in various studies on spatial chaos " @xcite . given a random initial condition , at short times , most of the pattern quickly relax to the conductive state @xmath21 , with only a few remaining patches that further lead to immobile localized structures .
one can control the final density of localized structures by means of the parameter @xmath147 in the initial condition .
the configurations become more dilute as @xmath147 decreases .
if @xmath14 is increased and set _ closer to @xmath152 _ in @xmath153 $ ] , random initial conditions gradually tends to form other kinds of frozen disordered states .
these states are mixed " , characterized by the presence of both localized structures and faceted extended hexagonal domains ( figure [ frozen]b ) .
the interfaces in fig .
[ frozen]b are pinned , as expected .
the enhancement of hexagonal order for quenches closer to @xmath152 can be qualitatively understood from section [ pinning ] : the driving force @xmath88 increases with @xmath14 faster than the pinning potential @xmath79 , even in the pinning zone , see fig.[fig : pth ] .
the time evolution of a typical configuration is the following : a few stable localized structures form at intermediate times .
small clusters , composed of 2,3,4 or more dots , are also present .
clusters with 2 dots do not grow ( nor move ) .
some clusters with 3 or more dots are unstable : some of them , that are initially irregular , expand slowly by the creation of successive layers of dots ( they may absorb in the process stable localized structures standing in their neighborhood ) .
growing clusters eventually reach steady states , corresponding to faceted polygons with an underlying hexagonal structure .
these extended hexagonal domains ( generalizations of the 7-dots structures observed in ref.@xcite ) coexist with the remaining single localized structures .
it takes a long time ( typically @xmath154 times units ) for the whole system to reach a steady state .
the normal to an interface can be directed along non - crystalline directions , due to , for instance , merging processes taking place between neighboring clusters .
domains can be fairly extended by the time they reach a faceted shape , that further prevent their growth .
the facets ensure stability , although some defects ( penta - hepta dislocations ) can be present in the bulk of a domain .
a slight cusp inward is visible on an interface in the upper left corner of fig.[frozen]b .
this feature is reminiscent of a grooving effect with sharp angles , a phenomenon well known in polycrystals with a strong anisotropy in surface tension @xcite .
further increase in @xmath14 beyond @xmath152 leads , as expected , to homogeneous states of hexagons , with no conductive regions at large times .
figures [ grow ] shows successive configurations obtained at intermediate times in the case @xmath155 .
small hexagonal clusters grow , in a way analogous to a stable phase in a metastable one .
although the single localized structures present are individually stable , they are absorbed during growth , as hexagon / uniform interfaces sweep them in a finite time .
hence , localized structures can not be observed at arbitrary large times and are metastable in this range .
the absolute " stability region of localized structures is in practice restricted to the hatched zone of fig.[fig : diag.int.ls ] , included in the pinning zone of interfaces .
when the control parameter exceeds @xmath156 , transient short - lived localized structures may still emerge from random initial conditions .
they further destabilize by forming structures of six - fold symmetry and then growing bubbles , as illustrated in the sequence of figure [ unstable ] .
localized structures nucleate hexagonal order in this case , as already observed in the lengyel - epstein model of turing patterns @xcite and experimentally in semiconductor - gas discharge systems @xcite .
an other possible case , illustrated by figure [ frozen]c , corresponds to pinned interfaces and unstable ( shrinking ) localized structures .
it happens in the roughly triangular region delimited by the curves @xmath128 , @xmath152 and @xmath149 in fig.[fig : diag.int.ls ] . in fig.[frozen]c , the smallest structures observed are clusters with two dots
. finally , for @xmath157 , any initial random condition converges toward the conductive state @xmath21 .
we have studied the dynamics of interfaces in the bistability region of conductive and hexagonal states of the swift - hohenberg equation ( [ sh ] ) .
we have examined the consequences of the various dynamical regimes of interfaces on the occurrence of spatial complexity and on the metastability of smaller structures like localized structures .
steady disordered states are possible in the nonlinear regime only , and for a choice of parameters such that planar interfaces remain pinned . in the expansion regime ( @xmath158 )
localized structure may be stable but are susceptible to be swept by any traveling interface in a finite time .
the size of the pinning interval @xmath159 $ ] is significant as soon as the bare coherence length @xmath8 is of the order of ( or smaller than ) a characteristic value @xmath73 , approximately given by formula ( [ xic ] ) . an extended domain with pinned interfaces can be seen as a nonequilibrium analogue of a solid crystal .
domains can contain dislocations in the bulk , and even vacancies .
defects can induce grooving - like effects at the interface ( non - planar interfaces with a cusp inward )
. on general grounds , decreasing the coherence length @xmath8 ( keeping the other parameters constant ) increases nonlinear effects and enhances hexagonal order ( @xmath160 ) , as illustrated by the monotonic curves of @xmath125 and @xmath126 in the diagram of fig.[fig : diag.int ] .
the quantity @xmath161 can be seen as a measure of the pattern strength `` , a weak '' pattern yielding the phase @xmath21 .
though , one of the nontrivial features of fig.[fig : diag.int ] is the _ non - monotonic _ behavior of @xmath152 with @xmath8 .
there exist a small interval of @xmath14 s such that pinned states undergo a depinning transition ( the system becoming completely filled with hexagons ) if one slightly _ increases _ @xmath8 . in this case ,
a weaker " pattern fills space better .
heterogeneous frozen configurations can be of various kinds : they can include localized structures ( liquid``-like disorder ) , or both localized structures and extended faceted domains of the condensed phase ( mixed '' disorder ) ; or even , faceted domains with small clusters and no isolated localized structures .
as the swift - hohenberg equation ( [ sh ] ) is the minimal model allowing subcritical patterns and bistability , we expect the results presented here , specially the conclusions drawn from the diagrams of figs.[fig : diag.int ] and [ fig : diag.int.ls ] , to be quite general and applicable to a variety of systems .
let us mention turing chemical patterns , nonlinear optical systems and related devices @xcite .
( rayleigh - bnard convection and block - copolymer melts are unfortunately less promising candidates as their @xmath8 is relatively large @xcite . )
our results show in particular that crystal - like patterns can be contained inside polygonal domains compatible with symmetries , and of arbitrary shape .
such patterns could be used as lithographic templates for technological applications ( like micro - electronics ) that require patterned surfaces over well - defined regions of space .
we thank m. clerc for fruitful discussions .
this work was supported by the consejo nacional de ciencia y tecnologa ( conacyt , mexico ) grant number 40867 .
the equations satisfied by the amplitudes @xmath29 can be derived from eq.([sh ] ) by standard multiple scale analysis @xcite .
the spatial variations of @xmath2 , see eq.([decomp ] ) , involve two kinds of length scales : a short scale associated with the periodicity of modulations , @xmath0 , and a large one associated with the scale of variation of the amplitudes , of order @xmath61 ( the interface thickness ) . when @xmath162 and @xmath163 , @xmath65 , and it is possible to decompose spatial variations into fast `` variables , @xmath164 , and a slow '' variable @xmath165 ( as well a a slow temporal variable @xmath166 ) , such that @xmath167 .
the equations for the @xmath168 are derived from a solubility condition for @xmath2 at the higher order following @xmath169 .
the solubility condition for the amplitude @xmath29 reads , @xmath170e^{-i\vec{k}_n\cdot \vec{r}^{\,\prime}}=0,\ ] ] where the linear operator @xmath171 ^ 2 $ ] follows from eq.([sh ] ) , @xmath169 being given by ( [ decomp ] ) and @xmath172 . in ( [ solvc ] ) , we have taken the limit @xmath173 for commodity , since the interface profile is invariant along @xmath117 . close to onset , @xmath174 and @xmath30 can be considered as two independent variables and the only non - vanishing contributions of the integral ( [ solvc ] ) come from the terms oscillating as @xmath175 in the brackets .
coming back to physical variables , this leads to the well - known coupled amplitude equations ( [ amp10])-([amp30 ] ) according to relation ( [ l ] ) , when @xmath8 is smaller than @xmath0 ( or when @xmath14 and @xmath10 are not small compared to unity ) , the fast `` and slow '' variables ( @xmath30 and @xmath174 , respectively ) can not be considered as independent in the solubility condition ( [ solvc ] ) for @xmath29 . @xmath169
being given by the decomposition ( [ decomp ] ) , if one expands , for instance , the nonlinear term @xmath176 in eq.([solvc ] ) , some following integrals appear : with the notation @xmath178 and @xmath179 . in eq.([i ] ) , @xmath180 , with the notation @xmath181 .
this time , let us consider the non - resonant terms only , _
i.e. _ those with @xmath182 . the @xmath29 profiles of the planar interface depending on @xmath183 only , the contribution @xmath184 is exactly zero unless @xmath185 . if @xmath185 and @xmath186 , then @xmath184 reduces to @xmath187 where @xmath188 is actually a multiple ( positive or negative ) of @xmath70 .
the usual amplitude formalism assumes that @xmath189 varies slowly , so that @xmath190cst in the interval @xmath191 $ ] . in that case , the non - resonant @xmath184 given by eq.([na1 ] ) always vanishes .
however , as soon as @xmath192 is large but finite , the complex exponential does not exactly integrate to zero if one sits in the interface region ; @xmath184 , although small , does not vanish .
anticipating on section [ pinning ] ( see eq.[estimp ] ) , one can show that a term like ( [ na1 ] ) in the solubility condition ( [ solvc ] ) gives an additional contribution of order @xmath193 to the velocity @xmath43 of the boundary given by eq.([v0 ] ) .
( see further eq.([lawmotion])-([p ] ) . ) in eq.([estim ] ) , @xmath90 is a constant of order unity , @xmath74 the position of the interface with respect to an arbitrary origin , and @xmath52 given by ( [ d ] ) .
@xmath194 can take the values @xmath70 , @xmath195 , @xmath196 ... in ( [ estim ] ) , @xmath197 is finite but still assumed to be @xmath198 : one can keep among the nonadiabatic terms @xmath184 those with the smallest @xmath194 ( @xmath199 ) , and neglect those involving higher wavenumbers .
one now derives the amplitude equations with their non - adiabatic corrections .
one develops @xmath200 ( with @xmath169 given eq.([decomp ] ) ) , and , for each solubility condition ( [ solvc ] ) ( @xmath201 ) , one has to find out , with the notation introduced above , which are the products @xmath202 ( and @xmath203 ) such that @xmath204 ( or @xmath205 ) has @xmath185 and @xmath206 .
note that one must eliminate among the eligible products those that are resonant in any of the leading order amplitude equations ( [ amp10])-([amp30 ] ) or their complex conjugates .
these resonant terms , added to similar ones coming from the linear operator @xmath207 in ( [ solvc ] ) , constitute one of the 6 standard amplitude equations : the sum of these terms cancels out at order @xmath208 ( the order of the terms in the brackets of eq.([solvc ] ) ) , and thus would contribute in non - adiabatic effects at the following order , not considered here . for example , in deriving the nonadiabatic contributions of the equation for @xmath209 , one finds an eligible term ( [ i ] ) involving a product @xmath210 , characterized by @xmath211 . but this term is resonant in the equation for @xmath212 , and hence needs to be discarded
. the products of amplitudes appearing in the leading nonadiabatic terms can not be already known products . in order to calculate the integrals ( [ na1 ] )
, we next assume that the amplitudes vary slowly in the interval @xmath191 $ ] and @xmath213 .
after some tedious but straightforward algebra , equations ( [ amp12])-([amp32 ] ) are obtained .
|
we revisit the swift - hohenberg model for two - dimensional hexagonal patterns in the bistability region where hexagons coexist with the uniform quiescent state .
we both analyze the law of motion of planar interfaces ( separating hexagons and uniform regions ) , and the stability of localized structures .
interfaces exhibit properties analogous to that of interfaces in crystals , such as faceting , grooving and activated growth or melting " . in the nonlinear regime , some spatially disordered heterogeneous configurations do not evolve in time .
frozen states are essentially composed of extended polygonal domains of hexagons with pinned interfaces , that may coexist with isolated localized structures randomly distributed in the quiescent background .
localized structures become metastable at the pinning / depinning transition of interfaces . in some region of the parameter space ,
localized structures shrink meanwhile interfaces are still pinned .
the region where localized structures have an infinite life - time is relatively limited .
|
many fundamental and practical scientific problems involve a wide range of length scales .
typical examples may include subsurface flows and geophysical domains with microscopic structures . because there exist both natural randomness and lack of knowledge about the physical properties , it is often necessary to incorporate uncertainties into the model inputs .
one way to address the uncertainties is to model the random inputs as a random field / process , and in turn , such problems are often modeled as stochastic partial differential equations ( spdes ) . then the model s output can be accurately predicted by efficiently solving the associated spdes .
it is challenging to solve the spdes when the random inputs vary over multiple scales in space and contain inherent uncertainties .
the interest in developing stochastic multiscale methods for the spdes has steadily grown in recent years ( see e.g. , @xcite ) .
let @xmath0 be a set of outcomes and @xmath1 be a bounded domain in @xmath2 with a lipschitz boundary .
we consider the stochastic elliptic boundary value problem : seek a random field @xmath3 such that @xmath4 almost surely ( a.s ) satisfies the following equation @xmath5 where @xmath6 is a scalar random field . in particular , we assume that @xmath7 exhibits heterogeneity in multiple scales over space . since @xmath6 varies over different spatial scales , resolving the finest scale is not computationally feasible .
thus , we use multiscale methods . in practice
, a high dimensional random field can be used to approximate the stochastic field @xmath7 , yet computing the statistical output quantities of interest remains a difficult task . during the last decade several multiscale methods have been developed , see e.g. @xcite .
the idea of multiscale methods is to divide the fine scale field into many local sub - problems and solve these in order to form a global coarse scale equation .
this leads to a coarse scale equation in which the fine scale effects are taken into account .
one such multiscale method is the multiscale finite element method ( msfem ) @xcite .
the main idea of msfem is to incorporate the small - scale information into the finite element basis functions and capture their effects on the large scale through the discrete variational formulation . in many cases ,
the multiscale basis functions can be pre - computed and used repeatedly in subsequent computations with different source terms , boundary conditions and even modified coefficients .
the goal in this paper is to quantify the uncertainty through computing the statistical moments ( e.g. , expectation and variance ) of the stochastic solution .
we note that the variance of the solution gives a measure for confidence of the solution expectation .
numerical solution strategies for stochastic pdes generally follow three main steps .
first , the random inputs are approximately parameterized by a finite number of random variables .
this can be achieved by a truncated karhunen - love expansion and/or truncated polynomial chaos expansion @xcite .
second , a numerical approximation for the resulting high - dimensional deterministic pde is used to approximate the solution with the respective input parameters . finally , the solution is reconstructed as a random field and the statistical quantities of interest are computed .
the second step is most difficult because the pdes involve high - dimensional random parameter inputs .
there exist many methods for the second step .
a broad survey of these methods can be found in @xcite . among these methods ,
monte carlo and stochastic collocation methods have been extensively studied and widely used .
monte carlo methods and stochastic collocation methods generate completely decoupled systems , each of which is the same size as the deterministic system .
this is suitable for parallel computing and amendable for relatively high - dimensional random inputs . in a monte carlo context
, a large number of samples are randomly chosen and separate solves for each of the samples are used to determine the statistical behavior of solutions . however , a limitation is that convergence of monte carlo methods is usually slow .
unlike monte carlo methods , stochastic collocation requires independent solves at fixed collocation points which are specifically chosen . in turn
, this type of method has the capability to provide better accuracy than monte carlo with a fewer number of realizations . moreover , to overcome the _ curse of dimensionality _ imposed by high - dimensional input parameters , we can use smolyak sparse grids ( see e.g. , @xcite ) to reduce the number of collocation points . in this paper
, we consider the monte carlo method and the smolyak sparse grid collocation method for stochastic approximation . in the paper ,
we consider both multiscale features and uncertainties simultaneously .
a main focus is the use of a resourceful splitting technique to compute msfem basis functions . for the problem ( [ model - spde ] ) , we assume that the coefficient @xmath8 can be split into two parts , @xmath9 .
we then construct green s functions using the differential operator associated with @xmath10 .
the green s functions are used to construct a sequence of multiscale `` bubble functions , '' which are employed to build the multiscale basis functions for msfem in an iterative manner .
the green s function technique provides an modified framework to compute the bubble functions , and is suitable for parallel computing due to the independent construction .
the splitting of @xmath8 is flexible and can be easily controlled to lead to fast convergence of the bubble function sequence .
compared to standard msfem @xcite , the proposed msfem approach can accurately approximate multiscale solutions .
the new multiscale approach is applied to spdes and may result in new stochastic multiscale methods . since the green s functions essentially generate the msfem basis functions , this will reduce the dimension of random parameter space if the dimension of the random field for @xmath10 is smaller than that of the random field for @xmath8 .
we note that using karhunen - love expansion or polynomial chaos expansion usually yields an inherent splitting .
the new msfem can efficiently solve spdes with high - dimensional input parameters , and combining the approach with sparse grid collocation methods alleviates the need for a prohibitive number of deterministic solves .
we present convergence analysis of the proposed msfem approach for deterministic elliptic equations and stochastic elliptic equations .
complexity analysis is also presented for deterministic msfem basis functions and stochastic msfem basis functions .
the rest of the paper is organized as follows . in section @xmath11 we present the splitting technique which is used to compute the new msfem basis functions for deterministic elliptic pdes , and provide the associated computational algorithm . in section
@xmath12 , convergence analysis is rigorously derived for deterministic elliptic pdes .
section @xmath13 is devoted to the applications to stochastic elliptic pdes .
we present convergence analysis and complexity analysis using stochastic collocation methods in the section . in section @xmath14 , a number of numerical examples are presented to confirm the theoretical results . some conclusions and closing remarks
are made in section @xmath15 .
we consider the deterministic elliptic equation @xmath16 where @xmath8 is a heterogenous scalar function .
we note that our method can immediately be extended to the case of tensor coefficient function .
we assume that @xmath8 admits the splitting , @xmath17 where @xmath18 and @xmath19 are bounded below and above , specifically , @xmath20 here , @xmath10 often represents the coarse scale information of @xmath8 , and @xmath21 the fine scale information of @xmath8 . the multiscale finite element method ( msfem ) for eq .
( [ model - eq ] ) was introduced in @xcite and further analyzed in @xcite .
the key ingredient of msfem is the construction of an appropriate multiscale finite dimensional space in which the solution is sought .
in particular , the fine scale heterogeneity in @xmath8 will be imbedded in this finite dimensional space .
this information is incorporated into the coarse scale formulation through the coarse scale stiffness matrix . in this section ,
we develop a msfem basis function , which is constructed in a different way from the previous works ( e.g. , @xcite ) .
we introduce some notation for presentation .
@xmath22 ( @xmath23 ) denotes the lebesgue space .
the norm of @xmath24 is denoted by @xmath25 .
@xmath26 is the usual sobolev space equipped with norm @xmath27 and seminorm @xmath28 . in the paper
, @xmath29 is the usual @xmath30 inner product .
we define an energy norm on a sub - domain @xmath31 by @xmath32 .
if @xmath33 , then @xmath34 simply represents @xmath35 .
we let @xmath36 be a quasi - uniform partition of @xmath0 and @xmath37 be a representative coarse mesh with @xmath38 .
let @xmath39 .
following @xcite , we define the standard multiscale basis functions by @xmath40 for vertices @xmath41 of coarse cell @xmath37 , which satisfy @xmath42 where @xmath43 is the boundary condition associated with node @xmath44 .
there exist some options for the boundary condition @xmath43 ( see @xcite ) .
@xmath45 defines basis functions for local msfem if @xmath43 is a linear / bilinear function .
incorporating global information into @xmath46 produces global msfem @xcite .
we define the finite element space for the standard msfem by @xmath47 we note that the idea of using basis functions satisfying certain differential equations has been used before , see e.g. @xcite and the references therein .
since we discuss a generic multiscale basis function , hereafter , we will remove the subindex @xmath37 and @xmath44 from ( [ msfe - basis1 ] ) for simplicity of presentation .
next we use the splitting ( [ k - split ] ) to derive a new msfem basis function . on each coarse cell @xmath48
, we define a projection operator @xmath49 by @xmath50 the definition of @xmath51 implies @xmath52 .
we extend the function for the boundary condition in ( [ msfe - basis1 ] ) onto @xmath37 and denote the extended function by @xmath53 .
let @xmath54 , where @xmath55 is the identity operator .
then by ( [ msfe - basis1 ] ) we can derive an equation for @xmath56 @xmath57 we are going to construct a series to approximate @xmath56 . to this end , we set @xmath58 to satisfy @xmath59 then we recursively define a sequence of function @xmath60 , @xmath61 which satisfies @xmath62 the function @xmath63 and the sequence @xmath64 are bubble functions " containing microstructure information , which are localized to a coarse cell by imposing zero dirichlet boundary conditions .
let @xmath65 .
we define the new multiscale basis function @xmath66 and the finite element space for the new msfem by @xmath67 since the proposed multiscale basis function is defined as @xmath68 , the computation of @xmath69 depends on the construction of @xmath63 and the bubble sequence @xmath70 .
we find that @xmath63 and @xmath60 ( @xmath71 ) are all associated with the differential operator @xmath72 .
moreover , @xmath63 or @xmath60 ( @xmath71 ) can be formally written as @xmath73 , where @xmath74 is the source term in the equation on @xmath63 or @xmath60 .
it is well - known that green s function can be viewed as generalized inverses of differential operators .
we use green s functions to obtain @xmath75 in the present work .
as far as making an efficient implementation , we use the green s function @xmath76 for the operator @xmath77 .
the green s function @xmath78 solves the equation @xmath79 since the green s function @xmath76 offers the fundamental solution for the differential operator @xmath77 , the green s function @xmath76 can efficiently generate @xmath63 and @xmath60 ( @xmath71 ) . by eq .
( [ def - pi ] ) , we have @xmath80 then we similarly compute @xmath58 by performing @xmath81 and compute @xmath82 , @xmath83 , by performing @xmath84 to discuss the complexity of computation for the proposed msfem basis function , we investigate the computation in terms of matrix operations .
we use the vector function @xmath85 , where @xmath86 ( @xmath87 ) is a standard finite element basis function at the underlying vertex @xmath88 of a underlying fine grid in @xmath37 . here
@xmath89 is the number of the internal fine vertices in @xmath37 .
we define vectors @xmath90 and @xmath91 by @xmath92 where @xmath93 represents the tensor product .
we use @xmath94 to denote the differential operator associated with @xmath21 .
let @xmath95 and @xmath96 be the stiffness matrices associated with the operators @xmath77 and @xmath97 , respectively .
then @xmath98 we have the following theorem to represent @xmath63 and @xmath60 ( @xmath71 ) in the finite element space of fine grid .
the notations @xmath63 and @xmath60 are slightly abused in the following theorem .
[ thm - v - m ] let @xmath99 and @xmath100 ( @xmath71 ) be the finite element approximations on the underlying fine grid in @xmath37 .
then @xmath101 and for @xmath102 , @xmath103 let us still use @xmath78 to represent the numerical green s function on the underlying fine grid .
then direct calculation implies that @xmath104 thanks to eq .
( [ pi - l ] ) and eq .
( [ comp - g ] ) , it follows that @xmath105 for @xmath106 , we have @xmath107m_0^{-1 } v_0-\big(m_0^{-1 } \vec{b}(x)\big)^t v_1\\ & = \big(m_0^{-1 } \vec{b}(x)\big)^t ( m_1m_0^{-1 } v_0- v_1 ) .
\end{split}\end{aligned}\ ] ] using eq .
( [ comp - tx0 ] ) , we obtain @xmath108 m_0^{-1}(m_1 m_0^{-1 } v_0-v_1)\\ & = - \big(m_0^{-1 } \vec{b}(x)\big)^t(m_1 m_0^{-1})(m_1 m_0^{-1 } v_0-v_1 ) .
\end{split}\end{aligned}\ ] ] by repeating the procedure of ( [ comp - tx1 ] ) , it follows immediately that for @xmath109 , @xmath110
the proof is complete .
we pre - compute vectors @xmath90 , @xmath91 and matrices @xmath95 , @xmath96 . since @xmath90 , @xmath91 , @xmath95 and @xmath96 only depend on the local information in @xmath37 , their construction is suitable for parallel computation . by theorem [ thm - v - m ] , we obtain @xmath99 and @xmath100 ( @xmath102 ) by performing a direct matrix - vector multiplication . moreover , we find that the computations of @xmath99 and @xmath100 ( @xmath102 ) are independent of each other and suitable for parallel computation as well . by theorem [ thm - v - m ] , the numerical representation of @xmath69 can be written as @xmath111 if we still use @xmath112 to denote the numerical approximation of a standard msfem basis function in the underlying fine grid of @xmath37 , then it is easy to show that @xmath113 where @xmath114 and @xmath115 . compared eq .
( [ comp - new - msfe ] ) and eq .
( [ comp - msfe ] ) , we find that the computation of @xmath116 is comparable to the computation of @xmath112 in a parallel setting since each of the terms in @xmath117 may be obtained independently . when @xmath118 , eq .
( [ comp - new - msfe ] ) and eq .
( [ comp - msfe ] ) imply that @xmath119 , which means that the proposed msfem coincides with standard msfem .
application of this procedure is summarized in algorithm [ alg : det - msfem ] . 1 .
assemble the vectors @xmath90 and @xmath91 by ( [ def - v01 ] ) , and assemble the matrices @xmath95 and @xmath96 by ( [ def - m01 ] ) ; 2 .
compute each term of ( [ comp - new - msfe ] ) independently ; 3 .
construct the proposed msfem basis function @xmath117 by summing the terms obtained in step ( 2 ) .
let @xmath56 be the finite element solution ( i.e. , evaluated at fine vertices ) of eq .
( [ eq - xi ] ) . here
the notation @xmath56 is slightly abused . then by eq .
( [ eq - xi ] ) we have @xmath120 .
this gives the neumann series of @xmath56 , @xmath121 consequently , the last term in eq .
( [ comp - new - msfe ] ) is actually the truncated neumann series . by the neumann series and eq .
( [ comp - new - msfe ] ) , if the spectral radius of @xmath122 is less than @xmath123 , then @xmath69 is convergent as @xmath124 .
this will be consistent with assumption 3.1 .
we make the following assumption for convergence analysis . _
assume that the splitting @xmath125 on @xmath37 satisfies @xmath126 _
if the property of the splitting @xmath127 fails , then we can introduce a slightly modified splitting of @xmath8 such that the assumption ( [ assum1 ] ) is satisfied . as a matter of fact
, we have the following proposition . [ sprop ]
suppose that @xmath128 .
let constant @xmath129 satisfy the following inequality @xmath130 then the modified splitting @xmath131 satisfies the assumption ( [ assum1 ] ) if we switch @xmath132 to @xmath133 in ( [ assum1 ] ) .
if @xmath128 and ( [ s - eq ] ) holds , then direct calculation implies that @xmath134 .
consequently , the assumption ( [ assum1 ] ) is satisfied .
we note that the modified splitting , @xmath135 , is essentially the same ( up to a constant ) as the original splitting , @xmath125 .
now we provide the main convergence result for deterministic msfems .
[ det - thm ] define @xmath136 and @xmath137 .
let @xmath138 and @xmath139 be the msfem solution for eq .
( [ model - eq ] ) . then + ( 1 ) @xmath140 . +
( 2 ) @xmath141 .
thanks to the boundedness of @xmath10 and @xmath8 , it follows that @xmath142 . by theorem [ det - thm ] , it follows that @xmath143 this shows convergence of the proposed msfem .
the arguments of the proof theorem [ det - thm ] include the following lemma [ lem - pi-2 ] , lemma [ lem - pi-3 ] and lemma [ lem - pi-4 ] . in the rest of the section
, we will formulate these lemmas . to describe and prove lemma [ lem - pi-2 ] , we need a couple of lemmas .
the following lemma gives an upper bound of the sequence @xmath64 .
[ lem - xi - j ] let @xmath144 and @xmath60 ( @xmath145 ) solve eqs .
( [ eq - xi0 ] ) and ( [ eq - xi - j ] ) , respectively .
then @xmath146 where @xmath147 is defined in ( [ assum1 ] ) .
the proof of lemma [ lem - xi - j ] is presented in appendix [ app1 ] .
the following lemma shows that the sequence of the new msfe basis functions converge to a standard msfe basis function .
[ singlethm ] the sequence of basis functions @xmath148 is convergent .
moreover , @xmath149 where @xmath150 solves the standard msfe basis equation ( [ msfe - basis1 ] ) .
the proof of lemma [ singlethm ] is presented in appendix [ app2 ] . from the proof of lemma [ singlethm ] , we can obtain an explicit convergence rate for @xmath69 , which is stated as following : [ prop3.2 ] if the coarse function @xmath53 in eq .
( [ eq - xi0 ] ) is in @xmath151 , then @xmath152 where the constant @xmath153 .
the proof of proposition [ prop3.2 ] is presented in appendix [ app3 ] .
to describe lemma [ lem - pi-2 ] , we need to introduce elliptic projection operators .
specifically , we define the elliptic projection @xmath154 by @xmath155 similarly , the elliptic projection @xmath156 is defined by @xmath157 since @xmath158 by lemma [ singlethm ] , the standard density argument @xcite implies that @xmath159 .
we can show that the projection operators @xmath160 and @xmath161 are self - adjoint and idempotent operators .
let @xmath162 and @xmath163 be the msfem solution in @xmath164 and @xmath165 , respectively , for eq .
( [ model - eq ] ) .
then @xmath166 using lemma [ lem - xi - j ] and the technique presented in @xcite , we have the following result .
[ lem - pi-2 ] let @xmath167 .
then @xmath168 and @xmath169 are defined in theorem [ det - thm ] . for any @xmath167
, we write @xmath170 we define the operator @xmath171 by @xmath172 then we have @xmath173 by ( [ base - dif ] ) , we have @xmath174 and @xmath175 we first estimate @xmath176 .
let @xmath177 .
thanks to ( [ eq1-lem3.3 ] ) and the proof of lemma [ lem - xi - j ] , it follows that @xmath178 this implies that @xmath179 and @xmath180 , we have @xmath181 from which we get @xmath182 by ( [ eq3-lem3.3 ] ) and ( [ zk - bound ] ) , it follows immediately that @xmath183 next we estimate @xmath184 . by ( [ eq2-lem3.3 ] )
, we have @xmath185 .
\end{split}\end{aligned}\ ] ] let @xmath186 .
then @xmath187 and @xmath188 combining ( [ eq10-lem3.3 ] ) and ( [ eq11-lem3.3 ] ) implies that @xmath189 because @xmath190 , it follows @xmath191 since @xmath192 , we get @xmath193
. a direct calculation gives @xmath194 by ( [ eq12-lem3.3 ] ) and ( [ eq13-lem3.3 ] ) , we have @xmath195 since @xmath196 , it follows immediately that @xmath197 ) and ( [ eq14-lem3.3 ] ) completes the proof .
a straightforward application of theorem 3.6 in @xcite gives rise to the following lemma .
[ lem - pi-3 ] let the inequality ( [ est - pi2 ] ) in lemma [ lem - pi-2 ] hold .
then @xmath198 where @xmath199 is defined in lemma [ lem - pi-2 ] .
lemma [ lem - pi-3 ] gives the first convergence result in theorem [ det - thm ] .
following the proof of theorem 3.8 in @xcite and lemma [ lem - pi-3 ] , we have the following lemma .
let @xmath200 be the solution to eq .
( [ model - eq ] ) and @xmath138 the standard msfem solution of eq .
( [ model - eq ] ) . then [ lem - pi-4 ] @xmath201 because @xmath202 , the second convergence result in theorem [ det - thm ] follows lemma [ lem - pi-4 ] immediately .
we consider the stochastic elliptic equation ( [ model - spde ] ) . to make @xmath6 positive , we consider a logarithmic stochastic field , @xmath203 , where @xmath204 is a stochastic field with second moment .
we assume that @xmath204 admits the following truncated karhunen - love expansion ( see @xcite for details ) , i.e. , @xmath205+\sum_{i=1}^n \sqrt{\lambda_i } b_i(x ) \theta_i(\omega).\ ] ] here @xmath206 let @xmath207 , where @xmath208 and @xmath209 .
then stochastic field @xmath7 can be parameterized to a finite - dimensional random field @xmath210 .
we define the splitting of @xmath211 by @xmath212 where @xmath213+\sum_{i=1}^m \sqrt{\lambda_i } b_i(x ) \theta_i(\omega))$ ] ( @xmath214 ) and @xmath215 . in the subsection , we will present convergence analysis when the random field @xmath210 admits the splitting described in ( [ split - random - k ] ) .
we define @xmath216 to be the square integrable space with the probability measure @xmath217 , where @xmath218 is the joint probability density function of @xmath219 .
we can use theorem [ det - thm ] to derive an error estimate of the stochastic elliptic equation ( [ model - spde ] ) .
[ stochbound ] let @xmath162 and @xmath220 be the msfem solution of stochastic elliptic equation in space @xmath221 and @xmath222 , respectively , for the stochastic equation ( [ model - spde ] ) . if @xmath223 , then @xmath224 where @xmath225 .
the eigenvalues @xmath226 play an important role to control @xmath227 . to this end
, we define the energy ratio @xmath228 by @xmath229
. then we can show that @xmath230 is proportional to @xmath231 under certain conditions .
[ k1/k0-random - prop ] let @xmath232 uniformly for all @xmath233 .
if @xmath234(x_1,x_2)$ ] are piecewise analytic in @xmath235 , then there exists constant a @xmath236 such that for @xmath237 large enough @xmath238 where @xmath239 . since @xmath240
, then it follows that @xmath241 because @xmath234(x_1,x_2)$ ] is piecewise analytic in @xmath235 , then there exist positive constants @xmath242 independent of @xmath237 and @xmath243 ( see @xcite ) such that for given @xmath244 @xmath245 where @xmath246 . if @xmath237 is large enough , inequality ( [ ineq01-prop4.1 ] ) implies that @xmath247 due to the inequality @xmath248 for @xmath249 , then it follows @xmath250 the proof is completed .
theorem [ stochbound ] and proposition [ k1/k0-random - prop ] show that the convergence rate of the proposed msfem for the stochastic equation ( [ model - spde ] ) depends on the energy ratio @xmath228
. then we immediately have the following theorem .
suppose that the assumptions in proposition [ k1/k0-random - prop ] hold .
if @xmath237 is large enough such that @xmath251 is less than 1 , then @xmath252 if the stochastic field @xmath204 is gaussian , then its covariance function can analytically be extended to the whole complex plane @xmath253 , which is stronger than piecewise analytic .
the eigenvalues @xmath226 associated with gaussian fields decay very fast .
consequently , @xmath231 will decay very fast as @xmath237 increases .
due to central limit theorem , the gaussian stochastic process is very interesting for applications . for stochastic simulation
, we can use monte carlo methods .
the main disadvantage of monte carlo methods is slow convergence . to overcome the disadvantage , here we use stochastic collocation methods to discretize random parameter space .
the proposed msfem is used to discretize the spatial variable .
combined with the new msfem , we develop modified stochastic collocation methods to reduce the dimension of the random parameter space .
computational complexity is addressed for the modified stochastic collocation method . in this subsection
, we use stochastic collocation methods to discretize the random parameter space and show that combining the proposed multiscale method with stochastic collocation methods can reduce the dimension of the random parameter space , which is very important for simulations in high - dimensional random space .
let @xmath254 be @xmath129 collocation points scattered in random parameter space associated with an interpolation operator @xmath255 .
let @xmath256 be a deterministic solution depending on random parameters @xmath257 . then given a realization @xmath258 , the collocation solution @xmath259 is defined by @xmath260 .
we usually use the roots of an orthogonal polynomial ( e.g. , hermite polynomial or chebyshev polynomial ) to find the interpolation points .
one can select different collocation points and use a different interpolation operator @xmath255 to obtain different stochastic collocation methods , for example , full tensor product collocation @xcite and smolyak sparse grid collocation @xcite .
suppose that the green s function @xmath261 in ( [ eq - green ] ) depends on the random parameter @xmath258 and @xmath262 is the collocation solution for an arbitrary @xmath263 .
for any arbitrary realization @xmath264 , we define a modified interpolation operator @xmath265 for @xmath63 and @xmath60 ( @xmath102 ) . we define them as follows : @xmath266:=\int_k k_0(y , \theta_0 ) \nabla_y i_m g(x , y , \theta_0)\cdot \nabla_y l dy.\ ] ] we note that we can get the value of @xmath267 and do not employ interpolation @xmath255 for @xmath267
. then we similarly compute @xmath268 $ ] by @xmath269 : = \int_k \ , k_1(y , \theta ) \nabla_y i_m g(x , y , \theta_0 ) \cdot \bigg(\nabla_y \tilde{i}_m [ ( \pi l)(y , \theta_0 ) ] -\nabla_y l(y ) \bigg ) dy,\ ] ] and compute @xmath270 $ ] , @xmath83 by performing @xmath271 : = - \int_k \ , k_1(y ,
\theta ) \nabla_y i_m g(x , y , \theta_0 ) \cdot \nabla_y \tilde{i}_m [ { \tilde{\xi}}_{j-1}(y,\theta ) ] dy.\ ] ] by the definitions of @xmath272 $ ] and @xmath273 $ ] ( @xmath71 ) , we have only used @xmath274 to compute @xmath272 $ ] and @xmath273 $ ] ( @xmath71 ) .
this interpolation is performed on the relatively low dimensional parameter space @xmath275 .
since the new multiscale basis function is defined as @xmath276 , the computation of the interpolation for @xmath277 only involves the @xmath237-dimensional interpolation @xmath274 . to address the complexity
, we use matrix and inner products to discuss the computation of the interpolations in ( [ interp - pi - l ] ) and ( [ greeninterp ] ) . due to ( [ pi - l - fe ] ) and ( [ interp - pi - l ] )
, it follows that @xmath278=\bigg ( [ i_m m_0^{-1}(\theta_0 ) ] \vec{b}(x)\bigg)^t v_0(\theta_0).\ ] ] by ( [ tx - fe ] ) and ( [ greeninterp ] ) , we have for @xmath71 , @xmath279&= ( -1)^j \bigg([i_m m_0^{-1}(\theta_0 ) ] \vec{b}(x)\bigg)^t\bigg(m_1(\theta ) [ i_m m_0^{-1}(\theta_0)]\bigg)^j\\ & \times
\bigg(m_1(\theta ) [ i_m m_0^{-1}(\theta_0 ) ] v_0(\theta_0)-v_1(\theta)\bigg ) . \end{split}\end{aligned}\ ] ] by ( [ d - green1 ] ) and ( [ d - green2 ] ) , we compute @xmath280 , @xmath281 , @xmath282 and @xmath283 to obtain @xmath284 and @xmath285 $ ] ( @xmath102 ) .
the computations of @xmath281 , @xmath286 and @xmath283 are independent each other and very efficient in parallel .
the dominant computation lies in @xmath280 and solely depends on the dimension @xmath237 of the random parameter space .
we can very efficiently compute @xmath280 in parallel as well .
if we use the standard multiscale basis function defined in equation ( [ msfe - basis1 ] ) , then the basis function is @xmath287 for an arbitrary realization @xmath288 .
we interpolate the basis function @xmath289 in the full random space @xmath290 ( @xmath291 ) .
if @xmath292 is large , the interpolation on @xmath290 is computationally expensive and prohibitive .
if we use full tensor product collocation and polynomials with degree @xmath293 for each component of @xmath288 , then we have @xmath294 collocation points for full - space interpolation .
consequently , we need to compute @xmath294 multiscale basis equations @xmath150 for each vertex for the collocation . however , if we use the technique of the green s function for the new multiscale basis functions , then we have @xmath295 collocation points and compute only @xmath295 green s functions ( or green s matrix @xmath95 ) to generate the proposed multiscale basis functions .
consequently , the ratio @xmath296 between the number of collocation points in the proposed multiscale basis function and in the standard multiscale basis function is @xmath297 for example , if @xmath298 , @xmath299 and @xmath300 , then the @xmath301 .
this means that the computational effort for using the proposed multiscale method may be considerably decreased compared to the standard multiscale method when the full tensor product collocation is employed .
let @xmath302 denote the interpolation nodes for smolyak sparse grid collocation @xcite at dimension @xmath292 and interpolation level @xmath303 .
although smolyak sparse grid collocation requires much fewer nodes than the full tensor product collocation to achieve the similar accuracy , the number of nodes @xmath302 increases very quickly as @xmath292 increases .
the ratio @xmath296 between the number of collocation points in the proposed multiscale basis function and in the standard multiscale basis function is given by @xmath304 where we have used the fact @xmath305 for @xmath306 @xcite .
for example , if we take interpolation level @xmath307 , @xmath299 and @xmath298 , then @xmath308 .
this means that the computation time of the proposed multiscale method is almost @xmath309 of the standard multiscale method in parallel setting .
smolyak sparse grid collocation is known to have the same asymptotic accuracy as full tensor product collocation , while requiring many fewer interpolation points as the parameter dimension increases
. we will use smolyak sparse grid collocation for the numerical tests .
the stochastic approximation of the smolyak sparse grid collocation method , @xmath310 , depends on the total number of sparse grid collocation nodes and the dimension @xmath237 of the random parameter space .
the convergence analysis in @xcite implies that the convergence of smolyak sparse grid collocation with respect to the number of smolyak nodes is exponential , but depends on the parameter dimension @xmath237 .
if @xmath311 , then the exponential convergence rate behaves algebraically . using the modified stochastic collocation method described in ( [ interp - pi - l ] ) and ( [ greeninterp ] ) , we define the corresponding collocation basis function @xmath312 + \sum_{j=0}^j \tilde{i}_m [ { \tilde{\xi}}_{j } ( x , \theta)]$ ] .
let @xmath313 be the collocation solution using the basis @xmath314 .
then the total error @xmath315 includes two parts : splitting error @xmath316 and collocation error @xmath317 .
it can be formally expressed by @xmath318 where @xmath319 and @xmath320 .
there exists a trade - off between the splitting error @xmath321 and the collocation error @xmath322 .
the numerical results in section @xmath14 illustrate the finding . using the proposed stochastic collocation methods , computation of the msfem basis function @xmath323
is summarized in algorithm [ alg : col - msfem ] .
choose collocation samples @xmath324 ; 2 .
assemble green s matrices @xmath325 ( @xmath326 ) for all collocation samples independently ; 3 .
given an arbitrary realization @xmath327 , assemble vectors @xmath281 and @xmath286 and matrix @xmath283 independently ; 4 .
compute @xmath280 indpendently ; 5 .
compute @xmath284 by ( [ d - green1 ] ) and @xmath285 $ ] ( @xmath102 ) by ( [ d - green2 ] ) independently ; 6 .
construct the interpolated basis function @xmath328 $ ] .
in this section we offer a number of representative numerical results to verify the analysis and evaluate the performance of the proposed method . to begin , we consider the analysis offered in sect .
[ converge ] .
in particular , we are initially interested in verifying the convergence properties of a single , deterministic basis function as described in lemma .
[ singlethm ] ( or prop .
[ prop3.2 ] ) . in this subsection
we consider two distinct cases of coefficient examples .
we first consider a coefficient generated by a karhunen - love expansion and a coefficient constructed from a log - normal distribution .
see fig .
[ logperm ] for illustrations of each coefficient plotted on the log scale .
mesh ( log scale ) , scaledwidth=60.0% ] to generate the first test coefficient on an arbitrary coarse element @xmath37 we employ the kle expansion from eq . . in our case
we use the correlation function @xmath329(\mathbf{x}_1 , \mathbf{x}_2 ) : = r(x_1 , y_1 ; x_2 , y_2 ) = \sigma^2 \text{exp } \left ( - \frac{|x_1 - x_2|^2}{2 l_x } - \frac{|y_1 - y_2|^2}{2 l_y } \right),\ ] ] where @xmath330 is the variance , and @xmath331 denote the correlation lengths in the @xmath332 and @xmath333directions , respectively .
we consider an elliptic coefficient which is generated on a @xmath334 mesh . for the variance we use @xmath335 and for the correlation lengths we use @xmath336 and @xmath337 .
for all examples we truncate the original kl expansion at @xmath298 terms to obtain the full coefficient @xmath338 .
then , in order to split the coefficient accordingly , we may choose a variety of @xmath237 values and employ eq .
for the splitting .
for example , we may use @xmath339 terms to obtain @xmath10 , and @xmath340 . see fig .
[ permsplit ] for a representative example of a kle coefficient splitting .
mesh ; @xmath298 , @xmath341,scaledwidth=85.0% ] as the initial analysis is built in a deterministic setting , we use the same , fixed @xmath342 @xmath343 in for all related examples . to begin ,
we recall the error estimate @xmath344 and set @xmath345 from the last lines in the proof of lemma .
[ singlethm ] .
in particular , we recall that when @xmath346 , convergence of the basis function sequence @xmath347 is expected from the analysis . to test the error bound in eq .
we consider a variety of field splitting configurations . in fig .
[ boundplot ] we illustrate two cases of splitting where @xmath346 .
these examples result from the cases where @xmath348 and @xmath349 terms are used in the kle splitting .
first , it is important to note that the bounds presented in the analysis are clearly represented in the figure . in particular , for either case we see that the energy norm of the error is always bounded above by the theoretical estimate provided in eq . .
we also note that as @xmath350 ( the number of terms in the approximate basis function sequence ) increases the error and associated bounds rapidly decrease .
this behavior is expected as the term @xmath351 quickly decreases as @xmath350 increases .
we also point out that a smaller value of @xmath147 yields a tighter bound .
terms ( left ) , @xmath349 terms ( right ) , scaledwidth=90.0% ] to conclude this coefficient example , we offer an illustration of the actual basis functions which are obtained through the proposed computational method .
[ phixiplot ] includes the benchmark basis function @xmath150 as well as @xmath69 @xmath352 from the sequence @xmath353 .
we also plot the benchmark perturbation @xmath354 and @xmath355 for @xmath356 .
all plots were obtained from the case where @xmath339 terms were used in the kle splitting .
we note that for a relatively pronounced splitting , we see a noticeable convergence trend . and the corresponding perturbation @xmath354 for the kle coefficient ; @xmath339 , scaledwidth=100.0% ] to generate a second test coefficient on an arbitrary coarse element @xmath37 we assume that the full coefficient @xmath8 follows a log - normal distribution .
that is , we assume that @xmath357 , where @xmath204 is a normal random variable with zero mean and variance one .
we also assume that @xmath358 , where @xmath359 is the strength factor which will determine the final splitting @xmath338 .
in particular , we set @xmath360 to create the coefficient decomposition . for the following examples , we choose various values of @xmath359 within the interval @xmath361 .
see fig .
[ permsplit_norm ] for an example of the coefficient splitting in eq . for the case
when @xmath362 .
we note that the @xmath10 portion of the decomposition is much less heterogeneous than @xmath21 .
mesh ; @xmath359 = 0.4,scaledwidth=85.0% ] to validate the convergence properties outlined in thm .
[ singlethm ] , we consider two approaches and recall the estimate from eq . . in fig .
[ boundplot_normal ] we illustrate two representative plots of splitting where @xmath346 .
these examples result from the cases where strength factors of @xmath363 and @xmath364 are used in the log - normal coefficient splitting .
as before , we note that the bounds presented in the analysis are clearly represented in the figure . for either case
we see that the energy norm of the error is always bounded above by the theoretical estimate provided , and that as @xmath350 increases , the error and associated bounds decrease rapidly . here
, we are also interested in the rate of convergence offered in eq . .
to address this , we fix a value of @xmath350 and plot @xmath147 vs. @xmath365 on the log scale .
fig [ orderplot_normal ] illustrates the log plots as well as the slopes obtained from a linear trend line . in this case
we obtain slopes which are close to the value of @xmath366 .
in particular , for @xmath367 we obtain a slope of @xmath368 , for @xmath369 we obtain a slope of @xmath370 and for @xmath371 we obtain a slope of @xmath372 . these results are consistent with the convergence rate results from prop .
[ prop3.2 ] .
in particular , we see from the plots in fig .
[ orderplot_normal ] that exponents of @xmath373 are recovered for all values of @xmath350 .
because the estimate in prop .
[ prop3.2 ] depends on a constant ( maybe not sharp ) , there exist a slight difference @xmath374 of the convergence rate in the numerical results compared to the estimate in in prop .
[ prop3.2 ] .
( left ) , @xmath364 ( right ) , scaledwidth=90.0% ] vs. @xmath365 for the log - normal coefficient ; @xmath375 ( left ) , @xmath376 ( center ) , @xmath377 ( right ) , scaledwidth=100.0% ] to conclude these coefficient examples , we offer a representative illustration of the actual basis functions which are obtained through the proposed computational method .
[ phixiplot_normal ] includes the benchmark basis function @xmath150 as well as @xmath69 @xmath378 from the sequence @xmath353 .
we also plot the benchmark perturbation values @xmath354 and @xmath355 for @xmath379 .
all plots were obtained from the case where a strength factor of @xmath362 is used in the coefficient splitting .
we note that this is a rather extreme splitting where @xmath10 is much less heterogenous compared to @xmath21 ( refer back to fig . [ permsplit_norm ] ) for the original log - normal coefficient configuration .
the figure illustrates that successive approximations offer an accurate counterpart to the benchmark basis and perturbation values . and the corresponding perturbation @xmath354 for the log - normal coefficient ; @xmath380 , scaledwidth=100.0% ] in this subsection we assess the convergence of the respective solutions of eq . . in particular , we are interested in comparisons between the standard msfem solution @xmath381 , and the proposed msfem solution @xmath382 .
we first recall that thm .
[ det - thm ] suggests convergence of @xmath163 to @xmath383 as @xmath124 . in order to verify this theoretical result
we test two separate permeability configurations .
in particular , we use a kl expansion ( with a set of fixed random parameters ) , and a single realization of a log - normal field analogous to those in subsection [ detbasis ] .
however , the fine scale fields are now posed on at least a @xmath384 fine mesh .
see fig .
[ logperm_fine ] for a log - plot of each respective permeabililty field . until otherwise noted , all msfem solutions in this subsection
are obtained by using a @xmath385 coarse mesh .
mesh ( log scale ) , scaledwidth=60.0% ] first , we consider a kle coefficient which is posed on @xmath384 , @xmath386 , and @xmath387 fine meshes . for the variance we use @xmath335 and for the correlation lengths we use @xmath388 and @xmath389 , respectively ( see eq . ) .
for all examples we truncate the original kl expansion at @xmath298 terms to obtain the full coefficient @xmath338 . in order to test the convergence properties in thm .
[ det - thm ] we are interested in computing a variety of proposed msfem solutions @xmath163 ( for @xmath390 ) . then
, we compute the energy norm @xmath391 and the associated bound @xmath392 .
we note that for @xmath393 , we compute the energy norm of the standard fem elliptic solution .
[ fineboundplot_kl ] illustrates the energy norm values and bounds for increasing @xmath350 values , and two different kle splitting configurations ( @xmath394 and @xmath395 ) .
we note that , as before , a smaller @xmath169 value yields a tighter bound .
in addition , we see the pronounced convergence of the proposed msfem solution . leaving the coarse mesh intact , fig .
[ fineboundplot_kl ] also serves to illustrate the effect that different local meshes have on the construction of the operators @xmath95 and @xmath96 from thm .
[ thm - v - m ] .
in essence , the fine meshes give three successive local mesh refinements ( from @xmath396 to @xmath397 to @xmath334 ) on which the local operators will be constructed .
as shown in the figure , there is a negligible difference in the respective errors , thus illustrating that the fine scale on which the operators are computed does not have a significant effect on the error resulting from the proposed method .
furthermore , the results show that the original @xmath384 mesh is sufficiently fine to resolve the necessary scales in the kl expansion . to further illustrate the convergence of the proposed msfem approach , we offer various elliptic solution plots in fig .
[ ellipcompare_kl ] .
in addition to the solution plots we offer the relative errors @xmath398 . as @xmath350 increases , we see that the relative error steadily decreases from @xmath399 to @xmath400 . terms ( left ) , @xmath395 terms ( right ) ; @xmath384 ,
@xmath386 , and @xmath387 fine meshes , scaledwidth=90.0% ] ( labeled ` reference ' ) , and the corresponding new msfem elliptic solutions @xmath163 for @xmath356 ; kle coefficient , @xmath401 , scaledwidth=100.0% ] we also consider the log - normal coefficient to test the convergence properties of the proposed approach .
the energy norm and bound plots may be seen in fig .
[ fineboundplot_normal ] , where strength values of @xmath402 and @xmath403 are used .
as expected , we see that the solutions converge and that a smaller @xmath169 value yields a tighter bound . for this case
, we also offer an illustration of various solution plots along with the relative errors values .
see fig .
[ ellipcompare_normal ] for the convergence illustration .
we note that as @xmath350 , the relative error decreases from @xmath404 to @xmath405 . in this case
we see that the initial error is less than its counterpart from the kle coefficient .
this is not unexpected since the kle expansion represents a stronger form of heterogeneity ( refer back to fig .
[ logperm_fine ] ) .
( left ) , @xmath406 ( right),scaledwidth=90.0% ] ( labeled ` reference ' ) , and the corresponding new msfem elliptic solutions @xmath163 for @xmath356 ; log - normal coefficient , @xmath407 , scaledwidth=100.0% ] in addition to the results presented above , we also consider a variety of coarse mesh configurations for comparison . to recall , all previous examples in this subsection were obtained from a fixed @xmath385 coarse mesh .
however , it is also fitting to illustrate the effects of different mesh configurations on the solution error .
in particular , fig .
[ errorplot_varycoarse ] contains the errors quantities @xmath408 ( left ) and @xmath391 ( right ) for a variety of mesh configurations and @xmath350 values .
the errors are obtained from the same kle coefficient data with @xmath409 .
we note that the left set of errors ( the errors between the standard fem solution and the proposed msfem solution ) are essentially constant regardless of the value of @xmath350 . in other words ,
the dominant source of error is clearly from the multiscale solution method , and the error between standard msfem and the proposed method is negligible . in addition , we see that a refinement of the coarse mesh yields a smaller error , which is to be expected from a multiscale solution technique . of particular interest
are the right set of errors @xmath391 , which are computed using the same coarse mesh configurations .
most importantly , we see that a successive refinement of the coarse mesh does not significantly affect the error between standard msfem and the proposed method .
we note that a refined coarse mesh does lead to a slight decrease in the error , however , it is clear from fig .
[ errorplot_varycoarse ] that the proposed method is not sensitive with respect to the coarse mesh configuration .
in particular , it is evident that @xmath350 is the parameter which dictates the respective errors . ;
@xmath410 , and @xmath411 coarse meshes , scaledwidth=90.0% ] in this subsection we address the stochastic problem described in sect .
[ stochanalysis ] . to begin ,
we are first interested in testing the stochastic bounds which are proved in thm .
[ stochbound ] .
we remark that this stochastic result is analogous to the deterministic bounds offered in thm .
[ det - thm ] . in order to compute the left hand side of the inequality in thm .
[ stochbound ] we use the monte carlo approximation @xmath412 where the index @xmath129 denotes a fixed sample value .
in particular , we generate a stochastic field , compute the corresponding energy norms for @xmath413 , and obtain an average over the samples .
for the right hand side of the inequality @xmath414 , we use the same type of monte carlo approximation for the stochastic integrals and we use the fully resolved fem solution for @xmath200 . in order to verify the bounds , we focus on a stochastic field which is generated from a truncated kl expansion .
for the results in this subsection we employ correlation lengths of @xmath415 and a variance of @xmath416 for the field construction .
the series expansion is truncated at @xmath298 terms , and the coefficient is posed on a @xmath417 fine mesh .
we assume that the random coefficients @xmath342 ( @xmath418 ) represent a @xmath419 dimensional vector in the hypercube @xmath420^{20}$ ] .
in other words , for sampling we draw @xmath419 i.i.d .
uniform random variables from the interval @xmath420 $ ] .
see fig .
[ permsample ] for a typical coefficient sample in this context . for all stochastic computations @xmath421 samples are used , and for all msfem computations we use a @xmath422 coarse mesh .
mesh ( log scale ) , scaledwidth=40.0% ] to verify the bounds in thm .
[ stochbound ] we offer two representative plots in fig . [ stochbounds ] .
the left hand side was obtained by keeping @xmath423 terms in the original @xmath419 term expansion , and the right hand side was obtained by keeping @xmath424 terms in the original expansion .
we note that more terms in the kl expansion ( resulting in a smaller average @xmath169 value ) yields a tighter bound , and that the expected energy norms deplete rapidly .
these results are consistent with those obtained from the deterministic fields .
terms ( left ) , @xmath424 terms ( right ) , scaledwidth=90.0% ] to further illustrate the behavior of the stochastic problem we recall the energy ratio @xmath229 , where @xmath425 are the eigenvalues from the kl expansion .
we note that as @xmath237 increases this value is expected to quickly increase ( in other words , @xmath231 will quickly decrease ) .
as more terms in the kl expansion typically yield smaller errors from the series approximation , we expect consistent behavior between @xmath231 vs. the mean and variance quantities of the energy norm of the error , @xmath391 .
[ statplot ] illustrates the relationship between the statistics of the energy norm quantities and the energy ratio . for both plots we use @xmath369 for the basis function approxmations .
for both the energy norm mean and variance comparisons , we note that as @xmath426 increases , the respective statistical values also increase . in particular , we see that less terms in the kl expansion yield errors that grow algebraically with respect to a decreasing energy ratio . vs. the mean of @xmath427 ( left ) , and @xmath231 vs. the variance of @xmath427 ( right ) ; @xmath376 , monte carlo , scaledwidth=90.0% ] to finish this subsection we offer statistical comparisons obtained from the reference ( standard ) msfem solution @xmath162 and the proposed msfem solution @xmath163 .
see fig .
[ statcompare ] for a comparison between the mean and variance of the respective elliptic solutions .
we note that any differences are nearly undetectable , further verifying the accuracy of the proposed method . in this subsection
we address the alternative parameter reduction collocation approach as described in subsection [ paramreduc ] .
in particular , we are interested in testing the accuracy of monte carlo sampling with standard msfem , versus random parameter reduction collocation with the proposed msfem approach using the green s kernel . throughout this section
we recall that the total error can be decomposed into two main components .
namely , the error can be decomposed into the splitting error @xmath321 , and the collocation error @xmath428 . in order to assess the significance of the error contributions , we thoroughly test a variety of scenarios resulting from values of @xmath350 ( number of terms in the series expansion ) , @xmath237 ( splitting configuration ) , and @xmath303 ( collocation level ) . in a parallel setting the computational cost
is decreased when the parameter reduction approach is implemented , and demonstrated accuracy of the technique would solidify it as a suitable sampling alternative . throughout this subsection
we use the same kle configurations as in subsection [ stochresults ] . to motivate further disussion , we first offer table [ error ] which compares the monte carlo sampling approach with the new parameter reduction sampling approach . the values in the table result from keeping @xmath409 terms for construction of @xmath10 combined with interpolation levels @xmath429 , @xmath307 , and @xmath430 .
in particular , we tabulate the values @xmath431 for @xmath14 fixed @xmath219 ( random parameter ) values , where @xmath162 denotes the standard msfem solution with monte carlo sampling , and @xmath432 denotes the proposed msfem solution with the parameter reduction approach . as we can see from table [ error ] , the parameter reduction approach closely recovers each individual sample of the elliptic solution , and a higher interpolation level leads to smaller errors . in all cases we note that the relative errors do not exceed @xmath433 .
[ htbp ] |c|c|c|c| & + & level 1 & level 2 & level 3 + @xmath123 & 0.64 & 0.12 & 0.05 + @xmath11 & 0.30 & 0.11 & 0.10 + @xmath12 & 0.33 & 0.18 & 0.17 + @xmath13 & 0.67 & 0.19 & 0.08 + @xmath14 & 0.25 & 0.13 & 0.12 + although the individual sample results are promising , we are most interested in the statistical behavior of the respective solutions . in fig .
[ collstatplot ] we offer analogous results to those found in the previous subsection .
namely , we plot @xmath231 vs. the mean and variance quantities of the relative errors , @xmath434 and @xmath435 , where we recall that @xmath163 denotes the standard msfem solution technique combined with monte carlo sampling .
in essence , we use these expected errors as benchmarks for comparison with the proposed method . more specifically ,
we are interested in comparing standard msfem with monte carlo sampling against the proposed msfem approach with parameter reduction collocation .
we first note that an increase in interpolation level clearly yields a decrease in the relative error values .
the level 2 interpolation errors nearly match the monte carlo results , and this slight discrepancy may be viewed as a trade of the increased efficiency of the proposed method . in other words , for a moderate collocation level
it is natural to expect some minimal error contributions from collocation .
however , the level 3 interpolation yields errors that are nearly identical to the monte carlo results .
this is due to the fact that the collocation error is essentially dimished , and the only remaining error is due to the splitting .
this will be discussed in further detail below . in fig .
[ meanvarl3 ] we single out the @xmath430 results and plot them with the monte carlo results .
we again note that the discrepancies are neglible , and point out the similarities in the statistical behavior which is illustrated in fig .
[ statplot ] .
in particular , we again encounter an increase of the mean and variance quantities with respect @xmath231 which is solely due to the splitting configuration . vs. the expected value ( left ) , and variance ( right ) of the relative error quantities ; @xmath436 , @xmath437 and monte carlo , scaledwidth=100.0% ] vs. the expected value ( left ) , and variance ( right ) of the relative error quantities ; @xmath436 , @xmath430 and monte carlo , scaledwidth=90.0% ] in addition to the high level of accuracy we obtain for a larger interpolation level
, we emphasize that _ all _ plots from fig .
[ collstatplot ] illustrate relative errors that do not exceed @xmath433 .
even though the level 1 results do not closely match the monte carlo results , a neglible error of @xmath438 may be completely acceptable for many applications . in particular , the small vertical scale should be duely noted . from the @xmath429 results we conclude that the proposed sampling method is not sensitive with respect to @xmath228 , and fewer terms
may be kept in the kl expansion without a significant loss in accuracy .
this is due to the fact that @xmath322 is dominant for the basic collocation level .
these results can be viewed as a potential limitation and advantage of the approach if a level 1 interpolant is used .
it may be a potential limitation because the addition of terms does not decrease the expected error significantly .
however , these results may also be viewed as an advantage since 14 terms in a 20 term kl expansion ( for example ) yields similar errors as 18 terms in a 20 terms kl expansion .
thus , fewer terms may be used in the decomposition with a neglible loss in accuracy . however , as the results show , no sacrifice in accuracy is necessary if a higher interpolation level is used .
this would simply amount to more precomputation steps in which additional sparse grid points are considered for the smolyak interpolant .
we also test the sensitivity of the approach with respect to the number of terms @xmath350 which are kept in eq . .
all previous results in this subsection were obtained from a value of @xmath436 in the expansions ; however , it is fitting to offer a comparison between solutions obtained by keeping more terms in the expansion .
for the following comparisons we use the same level 1 interpolation results as seen in fig .
[ collstatplot ] and test the results corresponding to @xmath436 and @xmath371 . in fig .
[ jcompare ] we note that the method does not exhibit sensitivity with respect to @xmath350 .
this is again due to the fact that the collocation error is dominant for a level @xmath429 interpolation .
we see that the mean errors may be slightly decreased by keeping more terms in the series expansion , yet this increase in accuracy is subtle .
this may be attributed to the fact that the @xmath429 green s function interpolant @xmath439 in eq .
does not guarantee faster convergence depending on the number of terms kept in the series expansion .
more specifically , the dominant collocation error is inhereted through the iterative procedure . as more terms
do not yield a significant gain in accuracy , keeping less terms is preferable in this low interpolation level setting .
however , as the splitting error becomes dominant ( i.e. , a higher interpolation level is considered ) using more @xmath350 terms would , in fact , result in more pronounced accuracy .
we next elaborate on some additional effects of a higher interpolation level .
, @xmath440 , scaledwidth=90.0% ] to finish this section we offer detailed relative error comparisons for @xmath441 ( total error ) , @xmath442 ( splitting error ) , and @xmath443 ( collocation error ) . in doing so ,
the aim is to solidify the contention that a higher interpolation level indeed diminishes the the collocation error . in turn , if the neglible total error ( @xmath438 ) which results from a lower level interpolation ( e.g. , @xmath429 ) is not suitable , a higher level interpolation ( e.g. , @xmath430 ) may be used to negate the collocation error . to reiterate , @xmath220 denotes a solution obtained from the proposed msfem approach with monte carlo sampling , and @xmath313 denotes a solution obtained from the proposed msfem approach with parameter reduction .
see fig .
[ methodcompare ] for an illustration of the various errors .
we introduce a slight abuse of the original notation from subsection [ paramreduc ] , and use @xmath444 to denote the total _ relative _ error , @xmath321 to denote the _ relative _ splitting error , and @xmath322 to denote the _ relative _ collocation error in fig .
[ methodcompare ] . in the left hand side of fig .
[ methodcompare ] we note that the smallest errors result from the proposed msfem combined with monte carlo sampling ( i.e. , the splitting error is small ) .
in addition , the total error and collocation error are comparable .
thus , we conclude that the collocation approach yields the dominant soure of error for a low level interpolant . however , we note the significant difference in the right hand side of fig .
[ methodcompare ] . in this case
we note two important factors .
first , the total error from a @xmath430 interpolant is smaller than its @xmath429 counterpart ( as expected ) .
furthermore , we see that the splitting error is now the dominant source of error , and the collocation error is much smaller . in other words ,
an increase in the interpolation level accomplishes the task of significantly reducing the collocation error .
thus , we conclude that for a higher interpolation level the parameter reduction approach behaves much like the standard monte carlo counterpart due to the minimal effect of collocation error . , @xmath436 , @xmath429 ( left ) , @xmath430 ( right ) , scaledwidth=100.0% ]
in this paper we present a new msfem approach for solving elliptic equations with coefficients that vary on many length scales and contain uncertainties . through considering a coefficient decomposition combined with a green s function approach ,
we are able to construct new msfem solutions which closely recover traditional msfem solutions . in a deterministic setting ,
rigorous error estimates and bounds are first presented for a representative basis function within the msfem approximation space . using the initial basis function results ,
we offer a rigorous error analysis describing the behavior of the elliptic solutions that are sought in the space that is spanned by the new multiscale basis functions . under appropriate assumptions on the coefficient splitting configuration , we are ultimately able to construct approximate solutions which nicely converge to a benchmark solution . the basis function and elliptic solution analysis are thoroughly verified through a number of representative numerical examples . in a stochastic
setting , the proposed solution method is shown to reduce the number of sample solutions that must be constructed for assessing the statistical behavior of the system .
in particular , the splitting gives rise to a situation where the random parameter dimension can be reduced , and where stochastic collocation becomes an efficient alternative to direct monte carlo sampling . as the parameter reduction sampling approach involves a number of pre - computation steps that are completely independent , this approach is especially desirable in a parallel setting .
analogous error bounds are derived for the stochastic problem , and the successful performance of the proposed method is verified through a variety of numerical examples .
we would like to formally thank professor v. ginting for helpful discussions on the theory and implementation of the proposed method .
his expertise in multiscale methods made the current work more manageable .
l. jiang also thanks professor y. efendiev and dr .
j. david moulton for useful discussions and comments .
we thank the reviewers for their comments to improve the paper .
let @xmath144 solve equation ( [ eq - xi0 ] ) .
then integration by parts gives @xmath445 this implies that @xmath446 let @xmath60 ( @xmath447 ) solve eq .
( [ eq - xi - j ] ) .
then a similar argument shows @xmath448 using ( [ ineq - xi - j ] ) recursively and ( [ ineq - xi0 ] ) , we have @xmath449 the proof is completed . since @xmath450 and @xmath54 , it suffices to show that @xmath451 by adding eq .
( [ eq - xi0 ] ) and the sequence of equations in @xmath452 , it follows that @xmath453 because @xmath454 , eq .
( [ thm1-eq1 ] ) is simplified to @xmath455 applying integration by parts to ( [ thm1-eq2 ] ) , we have @xmath456 as a consequence , it follows immediately that @xmath457 where we have used lemma [ lem - xi - j ] in the last step . hence , @xmath458 is convergent , and @xmath69 is convergent as well . subtracting eq .
( [ thm1-eq2 ] ) from eq .
( [ eq - xi ] ) , we have @xmath459 performing integration by parts and using the cauchy - schwarz inequality for eq .
( [ thm1-eq3 ] ) , then we obtain @xmath460 this completes the proof .
because @xmath450 and @xmath54 , it suffices to show @xmath461 in fact , the proof of lemma [ singlethm ] implies that @xmath462 this completes the proof .
|
in this paper we use a splitting technique to develop new multiscale basis functions for the multiscale finite element method ( msfem ) .
the multiscale basis functions are iteratively generated using a green s kernel .
the green s kernel is based on the first differential operator of the splitting .
the proposed msfem is applied to deterministic elliptic equations and stochastic elliptic equations , and we show that the proposed msfem can considerably reduce the dimension of the random parameter space for stochastic problems . by combining the method with sparse grid collocation methods , the need for a prohibitive number of deterministic solves is alleviated .
we rigorously analyze the convergence of the proposed method for both deterministic and stochastic elliptic equations .
computational complexity discussions are also offered to supplement the convergence analysis .
a number of numerical results are presented to confirm the theoretical findings .
multiscale finite element methods , green s function , stochastic elliptic equations , reduction of parameter space dimension 65n15 , 65n30 , 65n99
|
there have been some exciting developments recently since the four groups brahms , phenix , phobos and star have analyzed rhic data .
these four groups reporting on rhic , with gold on gold , show that quark gluon plasma with asymptotic freedom and chiral symmetry restoration may never be realized in rhic although a non - hadronic phase is reached .
unfortunately , their conclusions are negative in so far as finding of asymptotically free chirally symmetric qcd state is not possible in these reactions - although a new phase is formed which is quite distinct from the hadronic phase .
this phase is strongly interacting and is not fully understood theoretically but it is certainly not qgp with asymptotic freedom and chiral symmetry restoration .
two decades back witten@xcite has proposed the existence of strange matter and strange stars , and even today it is still difficult to prove or disprove the existence of a state of strange quark matter in its purest form . in literature , there are several eoss for strange quark matter , starting from the bag model @xcite to the recent models like the mean field model with interacting quarks@xcite , the perturbative qcd approach@xcite , chiral chromodielectric model@xcite , dyson - schwinger model@xcite , etc .
subsequently , rajagopal and wilczek combined asymptotic freedom and bcs theory to arrive at the color - flavor locked state of quark matter@xcite , and ever since , a lot of studies on this state and their application to quark matter eoss are made@xcite . here
, we developed a set of new eoss using a two parameter interaction potential .
the original qq potential of richardson@xcite was designed to obtain the mass spectrum of heavy mesons ( charmonium and upsilon ) .
it takes care of two features of qq force , af and confinement , however , with the same scale , @xmath3 : @xmath4 with @xmath3=400 mev .
it was further applied to light meson spectroscopy and baryon properties with the same value of @xmath3@xcite . in strange stars ,
the af part is important and there @xmath3 would be much smaller , around 100 mev which is the scale for asymptotic freedom as obtained from perturbative qcd . indeed , for the self bound high density
strange quark matter dey et al.@xcite found that @xmath3 needed to be @xmath5 mev .
quarks are deconfined at high density , as the bare potential is screened .
confinement is softened and the af part takes over .
this bare potential in a medium will be screened due to gluon propagation .
the temperature dependence of the screening in quark matter is taken from alexanian & nair @xcite .
the inverse debye screening length ( the gluon mass ) becomes : @xmath6 where @xmath7 , the fermi momentum of the _ i - th _ quark is obtained from the corresponding number density @xmath8 and @xmath9 is the perturbative quark gluon coupling . in our model ,
chiral symmetry restoration at high density is incorporated by assuming that the quark masses are density dependent : @xmath10 where @xmath11 is the baryon number density ; @xmath12 is the normal nuclear matter density ; @xmath13 , @xmath14 , @xmath15 are number densities of u , d and s quarks respectively and @xmath16 is a parameter .
the current quark masses ( @xmath17 ) are taken as : @xmath18 .
it is ensured that in strange matter , the chemical potentials of the quarks satisfy @xmath19 equilibrium and charge neutrality conditions .
the parameters @xmath20 and @xmath16 are adjusted in such a way that the minimum value of @xmath21 for u , d , s quark matter is less than that of fe@xmath22 , so that u , d , s quark matter can constitute stable stars .
the minimum value of @xmath21 is obtained at the star surface where the pressure is zero .
the surface is sharp since strong interaction dictates the deconfinement point .
however , the minimum value of @xmath21 for u - d quark matter is greater than that of fe@xmath22 so that fe@xmath22 remains the most stable element in the non - strange world . recently the magnetic moments of @xmath23 and @xmath24 with three u and three s valence quarks respectively have been found in accurate experiments .
these sensitive properties are fitted with a two parameter modified richardson potential with different scales for confinement ( @xmath25 mev ) and af ( @xmath26 mev)@xcite .
the baryonic properties depend more on the confinement part and less on the af part as baryons are confined quark systems .
it is natural to apply this new potential to star calculation since they are constrained by baryonic data . in our present calculation
, we have modified the richardson potential as : @xmath27 \times
\frac{1}{({\bf k}_i - { \bf k}_j)^2 } \label{eq : vtl}\end{aligned}\ ] ] with @xmath28 taking care of the confinement property and @xmath3 that of asymptotic freedom .
the term @xmath29 is asymptotically zero for large momentum transfer @xmath30 and the term @xmath31 explains the confinement reducing to a linear confinement for small @xmath32 .
the appropriate values of @xmath3 and @xmath33 as obtained from a fit to baryonic properties calculations@xcite are @xmath34 and @xmath35 , which we used in the present calculation .
finite temperature t is incorporated in the system through the fermi function@xcite : @xmath36 with the flavour dependent single particle energy @xmath37
@xmath38 i = number density for @xmath39 and i = energy density for @xmath40 .
@xmath41 is the spin - colour degeneracy , equal to 6 .
the system is highly degenerate even at very high @xmath42 which is around 80 @xmath43 since the chemical potential is very high , of order several hundred @xmath43 .
the entropy is calculated as follows : @xmath44 dk \label{eq : s}\end{aligned}\ ] ] the pressure is calculated from the free energy @xmath45 as follows : @xmath46 with the obtained eos ( pressure vs density ) , we solve the tolman oppenheimer volkov equation , to get the stellar structure .
eos for the set of parameters @xmath47 and @xmath48 , @xmath49 and @xmath50 and for temperatures 0 , 50 , 80 and 90 mev ( bottom to top ) .
we see that for 90 mev , there is no stable configuration.,width=283 ] eos for the set of parameters @xmath47 and @xmath48 , @xmath49 and @xmath50 and for temperatures 0 , 50 , 80 and 90 mev ( bottom to top ) .
we see that for 90 mev , there is no stable configuration.,width=283 ] the entropy per baryon decrease with density . in the nuclear matter limit ( say 1.5@xmath51 normal nuclear matter density ) , it is found to match exactly with the experimental results , width=283 ] the entropy per baryon decrease with density . in the nuclear matter
limit ( say 1.5@xmath51 normal nuclear matter density ) , it is found to match exactly with the experimental results , width=283 ]
a set of new eos for strange matter is presented , using the richardson potential with the value of @xmath52 and the value of @xmath53 , the two scales for confinement and af respectively .
this is then a good inter - quark potential as it also explains both the properties of the deconfined quark matter ( our present work ) and the properties of confined 3u and 3s baryons@xcite .
although the confinement part @xmath54 is stronger , leading to a sharp surface , - it is softened by medium effect , developing a screening length . inside the star ,
the af part is more important .
we have found that for a wide range of parametric variation , the strange matter eos gives minimum energy of e / a , much less than compared to @xmath55 which ensures that the system is absolutely stable , and unlike neutron star like structure , it is not gravitational force alone that binds the system .
however , for this present article , we have presented with only one such eos .
considering temperature dependent screening of the potential , we have found that strange stars can sustain stable configurations up to a temperature of 80 @xmath43 ; this value of the temperature is very close to witten s scenario of cosmic phase separation .
witten e. 1984 _ phys . rev .
d _ * 30 * 272 .
|
we calculate strange star properties , using large @xmath0 approximation with built - in chiral symmetry restoration ( csm ) .
we used a relativistic hartree fock mean field approximation method , using a modified richardson potential with two scale parameters @xmath1 and @xmath2 , to find a new set of equation of state ( eos ) for strange quark matter .
we take the effect of temperature ( t ) on gluon mass , in addition to the usual density dependence , and find that the transition t from hadronic matter to strange matter is 80 mev .
therefore formation of strange stars may be the only signal for formation of qgp with asymptotic freedom ( af ) and csm .
|
extremely high resolution observations of the powerful extragalactic double radio sources @xcite reveal that they are formed by well - collimated outflows or jets that continuously emerge from the nuclear region of the host active galaxies ( agns ) or quasars , believed to harbor supermassive black holes . similarly micro - quasars @xcite discovered in recent times reveal that outflows are generated from stellar mass black holes ( or black holes candidates ) . the well - collimated outflow in ss433 observed for several decades , which is our galactic , persistent , super - critical accretor , is an well known evidence of cosmic jet @xcite .
further , highly collimated line jets are seen in young stellar objects @xcite .
relativistic jets are also observed in neutron star low mass x - ray binaries ( lmxbs ) @xcite .
the outflows / jets extract matter , energy and momentum from the accretion disc that forms around the compact object , and thus the dynamics of outflow leading to jet is intrinsically coupled with the accretion dynamics through the conservation laws .
also several observations ( precisely the simultaneous observations of disc and jet ; see e.g. ghosh & mukhopadhyay 2009 , and references therein ) reveal that accretion processes and outflows are strongly correlated and they eventually control the accretion process , precisely , the accretion dynamics in the vicinity of the central star . the relativistic outflowing matter , in the case of quasars or micro - quasars , should indeed come only from the inner region of the accretion disc .
this is particularly suggestive as the quasars or the micro - quasars do not have an atmosphere of their own .
however , most of the models of accretion disc and outflow / jet have been evolved separately , considering these two to be apparently dissimilar objects .
the outflow / jet models have been evolved over the years , from speculative ideas such as de laval nozzles @xcite to electrodynamic acceleration model @xcite , centrifugally driven outflows @xcite , etc . moreover
, mhd simulations of outflow / jet have been performed both in non - relativistic as well as relativistic limits , mostly in the keplerian paradigm ( e.g. hawley & balbus 2002 , mizuno et al .
2006 , hawley & krolik 2006 , and references therein ) to see how the matter gets deflected from the equatorial plane . nevertheless , the definitive understanding of the origin of outflows / jets is sill unknown .
the radiatively driven outflow or jet whose origin is better understood , can be envisaged when the accretion disc is highly radiation pressure dominated or precisely `` radiation trapped '' .
this is likely to occur when the accretion rate is super - eddington or super - critical @xcite as in ultra luminous x - ray ( ulx ) sources such as ss433 ( with luminosity @xmath4 erg / s or so ) @xcite ; a prototype of ulxs in external galaxies having relativistic jets .
the super - critical flows are optically thick and have a strong advective component @xcite .
the unusual signatures of ulxs , compared to their counterparts , lie in their extreme high luminosity ( @xmath5 erg / s ) , strong spectral variability and their association with the actively star forming regions , that make their nature a subject of controversy @xcite .
recently , an ultra - luminous accretion disc with a high kinetic luminosity radio jet has been discovered in quasar pks 0743 - 67 @xcite , which indicates that the ultra - luminous sources may be extended to quasars or even agns . on the other extreme end
, the under - luminous agns and quasars ( e.g. sagittarius @xmath6 ) had been described by the advection dominated accretion flow model ( in short adaf @xcite ) , where the flow is substantially sub - critical / sub - eddington .
the model hypothesized a plausible emergence of strong outflows / jets , which lead to reveal a strong inter - connection between the outflow and advection .
very few models exist which simultaneously deal with the accretion and outflow / jet dynamics on the same platform , both in analytical regime as well as through numerical simulations ( see ghosh & mukhopadhyay 2009 , and references therein ) .
the difficulty of simultaneous simulation of the disc and outflow arises due to the difference in time scales between accretion and outflow .
also the results strongly depend on the initial conditions @xcite .
the 2.5-dimensional disc - outflow coupled model given by ghosh & mukhopadhyay ( 2009 ) has been formulated by incorporating explicit information of the outflow in a fully analytical regime through a self - similar approach in a general advective paradigm , by upholding the conservation laws .
it has been further shown that with mass , energy and momentum conservation , and a few scaling arguments , the correlation of the jet with the disc could be successfully modeled .
the authors solved a complete set of partial differential fluid equations ( with the variation of flow parameters in both radial and vertical direction ) without assuming the vertical hydrostatic equilibrium , with the explicit inclusion of the vertical velocity ( representing outflow ) and all the relevant components of the stress tensor apart from the usual @xmath7 , from the first principle .
they used their solutions to study two extreme cases of the geometrically thick advective accretion flows : super - critical and high sub - critical , which are more probable regimes of the strong outflows and jets .
they found that the flow parameters of the accretion - induced outflow / jet strongly depend on shakura & sunyaev viscosity parameter @xmath0 @xcite and cooling factor @xmath1 . in the present paper ,
we use and extend the above mentioned work @xcite to study the detailed energetics of the accretion - induced outflow and then plausible jet . in order to do that
, we stick to two extreme regimes of the black hole accretion : super - critical and high sub - critical , to describe ultra / highly luminous and under - luminous sources respectively . for obvious reasons ,
we do not repeat the calculation and the formulation of the disc - outflow model given earlier @xcite , but will recall them appropriately .
we arrange the paper in the following manner . in the next section ,
we formulate the equations for the energetics of the accretion - induced outflow . in 3 , we study the properties of the disc - outflow energetics
. finally we end in 4 with a discussion and implications .
in order to study the energetics of the outflow , we need to compute the mass outflow rate .
blandford & begelman ( 1999 ) generalized the adaf solution @xcite by including wind and outflow , assuming the mass inflow rate to be proportional to @xmath8 , where @xmath9 ( @xmath10 be the radial coordinate ) .
chakrabarti and his collaborators @xcite also attempted to calculate the mass outflow rate from a disc involving a shock without including the vertical flow explicitly . in the present paper , we derive this self - consistently . integrating the continuity equation @xcite vertically about the equatorial plane
we obtain @xmath11 where @xmath12 and @xmath13 are the radial and the vertical velocity respectively , @xmath14 is an arbitrary disc scale height and @xmath15 the minimum ( finite ) height of the disc - outflow system , which is different for super - eddington and sub - eddington accretion flows , @xmath16 be the vertical coordinate . here
we deliberately focus on the disc - coronal region ( sub - keplerian halo ) , from where the mass loss takes place in the form of wind or outflow , and hence the solution at equatorial plane looses significance .
the integral in the first term of the left - hand side of eqn .
( [ 1a ] ) represents the disc mass accretion rate given by @xmath17 the second term then can be attributed to the rate of change of outflow ( and then plausible jet ) mass , given by @xmath18 here , @xmath19 is analogous to the mass outflow rate represented as @xmath20 where the constant @xmath21 is to be determined by an appropriate boundary condition .
thus from eqn .
( [ 1c ] ) the total mass accretion rate can be written as @xmath22 eqn .
( [ 1f ] ) entails that under the stationary condition the radial mass flux in the disc decreases as the inflowing matter approaches the central object , at the same rate at which the vertical mass flux increases to maintain a constant @xmath23 which is exactly the net mass accretion rate at infinity . to compute the power of the outflow and then plausible jet , following previous work @xcite
, we integrate the disc - energy conservation equation for an accretion - induced outflow over the disc scale height which yields @xmath24 here , @xmath25 is the vertical component of the total energy flux given by @xmath26 where @xmath27 , @xmath28 is the gravitational potential , @xmath29 the radiative flux from the disc surface , @xmath30 the @xmath31 component of the stress tensor , @xmath32 the ratio of the specific heats of the gas - radiation mixture .
we thus obtain the power of the outflow from eqn .
( [ 2a ] ) , which is the total power removed from the disc by the outflow , as @xmath33 we then calculate the disc luminosity in presence of outflow ( and jet ) using disc - energy conservation equation described earlier @xcite @xmath34 where @xmath35 is the total viscous heat generated in the disc .
we now introduce three dimensionless parameters which correlate the disc and the outflow leading to jet : @xmath36 , @xmath37 and @xmath38 .
we do not describe here the generalized bernoulli s number @xmath39 , explicitly shown earlier @xcite , which represents the total energy of the system .
accretion by a black hole ( or any other central star ) is the primary source of the mass outflow and jet formed in the inner hot region of the disc . at the first instant we neglect the contribution of magnetic field which is likely to be the origin of collimation and acceleration of jet .
however , doubts can be raised about the significance of the magnetic field for production , collimation and acceleration of jets in ulxs and highly luminous agns and quasars , which are super - critical and highly radiation trapped systems ( see fabrika 2004 , for details ) .
we argue that the radiation pressure is likely to be the plausible reason for strong outflows and then plausible jets in high mass accretion flows @xcite .
the inflowing matter is expected to be ejected and accelerated through a funnel like region @xcite by the strong radiation pressure leading to strong outflow and then plausible jet from the disc . in
early , based upon the geometrically thick accretion disc model , it was shown that very narrow and deep funnels are formed around the rotation axis of the accretion disc and most of the disc energy flux is radiated from the surface of the funnels .
the radiation pressure may accelerate the outflowing mass in funnels up to relativistic velocities in the form of two jets pointing in opposite directions as observed in ss433 @xcite . in the inner accretion disc , a few schwarzschild radii away from the black hole ,
as the infall time scale is much smaller than the viscous time scale to transport angular momentum outside , matter has no time to loose angular momentum and attains a near constant value . around this region
the centrifugal force becomes comparable to the gravitational force , and thus the incoming matter slows down at this centrifugally dominated region and gets puffed up @xcite .
therefore , a funnel type geometry is likely to form around the region and the radiation pressure blows up the matter through them .
however , the magnetic field is likely to play the role in generating turbulence in the hot disc systems @xcite .
in fact , for critical or sub - critical accretion flows , the jet is likely to form due to the magnetic activity in the disc @xcite .
we do not intend to discuss here the possible reason of the emanation of the jet .
whatever the explanation / mechanism be , the potential dynamics of the accretion - induced outflow and then plausible jet always upholds the conservation laws which shows several interesting features .
the dimensionless parameters @xmath40 , @xmath41 , @xmath42 and @xmath43 carry the information of the disc - outflow energetics . in the next two subsections we analyse their features for super - critical and sub - critical accretion flows respectively . to study the high mass accretion flows , we consider @xmath44 corresponding to the central mass @xmath45 , which is a probable estimate of the accretion rate of our own galactic super - critical accretor ss433 @xcite .
we express @xmath23 in terms of critical eddington unit so that , for our case , @xmath46 .
this makes the energetic profiles independent of mass of the central star and one can extend our model to the case of high or ultra - luminous agns and quasars .
figures 1a and 1b depict the variation of mass outflow rate and power of the outflow and then plausible jet respectively as functions of @xmath10 .
the nature of profiles is similar to that of vertical component of velocity @xmath13 , explained in detail earlier @xcite . at @xmath47 , a substantial amount of matter and radiation
is extracted in the form of strong outflow leading to jet and thus the power of the outflow is significantly high , which falls off rapidly for @xmath48 . the luminosity profile ( fig .
1c ) exhibits an interesting feature . at small @xmath10
, it increases slowly ( although not very clear from the figure that is given in the logarithmic scale ) with @xmath10 and eventually attains a near constant value . at high @xmath0
, the luminosity corresponds to the eddington or the super - eddington accretion .
however the system becomes sub - eddington at low @xmath49 , even if @xmath46 and @xmath1 low .
this extreme sensitivity is due to the fact that @xmath0 directly controls the viscous heat generation . at @xmath50 , for @xmath51
, @xmath52 reaches @xmath53 , resembling highly luminous and ultra - luminous xrbs ( e.g. cxou j095550.2 + 694047 @xcite , grs 1915 + 105 , gro j1655 - 40 @xcite ) . at @xmath48
, @xmath52 decreases in orders of magnitude but still lies within the range of the high or ultra - luminous regime ( as in the case of ss433 ) . for a black hole of mass @xmath54
, it can be shown that @xmath52 attains a value @xmath55 erg / s for @xmath56 and @xmath50 .
such luminosity is commonly observed in highly luminous agns and ultra - luminous quasars ( e.g. pks 074367 ) , possibly in ulirs @xcite and narrow - line seyfert 1 galaxies ( e.g. @xcite ) .
the nature of the variation of @xmath43 has been shown in fig .
1d , e , already explained by ghosh & mukhopadhyay ( 2009 ) with three - dimensional plots .
the truncation of the curves attributes to the truncation of the disc due to evaporation to corona . with the decrease of @xmath1 , which corresponds to more increase of radiation pressure
, the disc gets more radiation pressure dominated and the matter gets blown up due to the strong radiation pressure further out . at this situation , the disc tends to become more keplerian and thus centrifugally dominated .
therefore the centrifugal dominated boundary layer is likely to form at a larger radius . as a result the outflow rate and then the outflow / jet power increase significantly and the disc gets truncated at a larger radius . in fig .
2 , we show a comparison of @xmath57 , @xmath58 and @xmath42 between two values of @xmath32 at the same set of @xmath0 and @xmath1 . with a small increase in the value of @xmath32 from @xmath59 to @xmath60 ,
the above parameters decrease by an enormous margin .
the abrupt change of the corresponding parameters reflects that the system is very sensitive to @xmath32 and then @xmath61 which is the ratio of the gas pressure to total pressure . and @xmath61 is given in ghosh & mukhopadhyay 2009 .
] an increase of @xmath32 from @xmath59 to @xmath60 corresponds to an increase of @xmath61 from @xmath62 to @xmath63 , which further corresponds to the conversion from radiation dominated to marginally gas dominated system .
this signifies that the gas pressure in the system increases , which means that the radiation pressure of the system suddenly decreases .
the decrease in the radiation pressure inhibits strong outflow . hence the power of the outflow and mass outflow rate fall in order of magnitudes .
the same reasons are applicable in order to explain the variation of luminosity profiles . here
we consider that the flow has a mass accretion rate @xmath64 corresponding to @xmath65 , which renders the flow to be thick and gas pressure dominated .
figure 3 shows that the mass outflux and the power extracted from the disc by the outflow are much less compared to that of the high mass accretion flows . even for a near extreme @xmath66
, @xmath43 is @xmath67 times larger for super - critical flows compared to the ideal case of an advection dominated @xmath68 sub - critical accretion flow .
this signifies that outflows are more probable for super - eddington accretion flows compared to that of sub - eddington flows . for low @xmath0 and @xmath69
, we obtain a luminosity @xmath70 corresponding to @xmath71 ( fig .
3c ) . for high
@xmath51 , @xmath52 increases by three orders of magnitude ( not shown in the figure ) at the same @xmath1 . such low luminosities are observed in many under - luminous agns ( e.g. sagittarius @xmath6 ; @xcite ) .
based on a 2.5 dimensional hydrodynamical formulation of a disc - outflow coupling system with a set of self - similar solutions @xcite , we have studied the energetics of the system .
the self - similar approach is not just a mere tool to solve the non - trivial coupled partial differential equations .
the accretion flow indeed may exhibit self - similar behaviour .
it was shown earlier that the black hole system grs 1915 + 105 is chaotic in nature @xcite that supports the idea of inner disc instability and then turbulence .
it was also found that the corresponding correlation / fractal dimension to be similar to that in the lorenz system which is a model example of an ideal chaos .
the low correlation / fractal dimension implies possible self - similarity into the system . in reality
the accretion flows with strong outflows are likely to be geometrically thick and advective , which more possibly occur in ultra - luminous and under - luminous accreting sources .
the standard keplerian model @xcite fails to describe these two opposite traits of observational signatures , which are possible sources of powerful outflows and jets .
the energetic profiles strongly depend on @xmath0 and @xmath1 .
they also reveal that outflows and jets are more probable and powerful for super - eddington accretion flows compared to the sub - critical ones , which correspond to ultra - luminous and under - luminous sources respectively . for super - eddington flows ( fig.1 ) , the disc gets truncated at larger radii compared to the sub - eddington ones .
this means that the outflow / jet may occur at the inner region of the disc and the disc - corona transition region may form in the vicinity to the central star in case of sub - eddington flows .
however , a slight departure of @xmath32 from @xmath72 to a higher value leads to an abrupt decrease in the energetics of the flow .
we have shown that at super - eddington accretion flows @xmath73 , with an appropriate choice of @xmath0 and @xmath1 , the luminosity calculated from our model is @xmath74 erg / s corresponding to a black hole of mass @xmath45 . such a high luminosity is commonly observed in exotic ulxs in external galaxies or in our own super - critical accretor ss433 . with the appropriate choice of free parameters @xmath75 and @xmath23 , we have shown that our model can reproduce the luminosities observed in ultra luminous sources and other family of high luminous agns ( fig .
1c ) , which probably accrete super - critically . similarly , the luminosity observed in under - luminous agns ( e.g. sagittarius @xmath6 ) can also be reproduced by our model at @xmath76 ( fig .
this particular situation is similar to adaf @xcite .
the limitation of adaf can be addressed by incorporating outflow from the disc ( as of ghosh & mukhopadhyay 2009 ) , which may turn the disc into a non - radiative accretion flow and thus can explain luminosity profile of under - luminous sources . as our solutions
can explain the observed luminosities of both ultra - luminous sources and under - luminous agns with the appropriate choice of free parameters , we should put our model to more observational tests .
although we have neglected the contribution of the magnetic field , whose importance for super - critical accretion flows is probably not significant @xcite , the inclusion of the magnetic field will render the system to be more realistic .
such a work will be pursued in the course of time . in the follow up
work , we plan to analyse a full scale numerical solution to conform / verify the results discussed here .
this work is partly supported by a project , grant no .
sr / s2hep12/2007 , funded by dst , india .
+ m. abramowicz , & t. piran , apj , 241 , 7 ( 1980 ) s. a. balbus , & j. f. hawley , apj , 376 , 223 ( 1991 ) m.c .
begelman , r. d blandford , & m.j .
rees , rvmp , 56 , 255 ( 1984 ) m. c. begelman , a. r. king , & j. e. pringle , mnras , 370 , 399 ( 2006 ) r. d. blandford , & m. c. begelman , mnras , 303 , 1 ( 1999 ) r. d. blandford , & d.g .
payne , mnras , 199 , 883 ( 1982 ) r. d. blandford , & m. j rees , mnras , 169 , 395 ( 1974 ) r. d. blandford , & l. znajek , mnras , 179 , 433 ( 1977 ) m. camenzind , a&a , 156 , 137 ( 1986 ) s. k. chakrabarti , a&a , 351 , 185 ( 1999 ) t. k. das , & s. k. chakrabarti , cqg , 16 , 3879 ( 1999 ) g. fabbiano , rmxac , 20 , 46 ( 2004 ) s. fabrika , asprv , 12 , 1 ( 2004 ) a. ferrari , ara&a , 36 , 539 ( 1998 ) j. fukue , pasj , 39 , 679 ( 1987 ) r. genzel et al . , apj , 498 , 579 ( 1998 ) s. ghosh , & b. mukhopadhyay , raa , 9 , 157 ( 2009 ) j. f. hawley , & s. a. balbus , apj , 573 , 738 ( 2002 ) j. f. hawley , & j. h. krolik , apj , 641 , 103 ( 2006 ) m. jaroszyski , m. a. abramowicz , & b. paczyski , aca , 30 , 1 ( 1980 ) p. kaaret et al . , mnras , 321 , 29 ( 2001 ) g. v. lipunova , astl , 25 , 508 ( 1999 ) r. v. e. lovelace , m. m. romanova , & w. i. newman , apj , 437 , 136 ( 1994 ) r. mahadevan , nature , 394 , 651 ( 1998 ) b. margon , ara&a , 22 , 507 ( 1984 ) s. migliari , & r. p. fender , mnras , 366 , 79 ( 2006 ) s. mineshige , t. kawaguchi , m. takeuchi , & k. hayashida , pasj , 52 , 499 ( 2000 ) i. f. mirabel , new ast .
, 47 , 471 ( 2003 ) i. f. mirabel , & l. f rodriguez , nature , 371 , 46 ( 1994 ) i. f. mirabel , & l. f rodriguez , nature , 392 , 673 ( 1998 ) r. misra , k. p. harikrishnan , b. mukhopadhyay , g. ambika , & a. k. kembhavi , apj , 609 , 313 ( 2004 ) y. mizuno , k - i .
nishikawa , s. koide , p. hardee , & g. j. fishman , arxiv : astro.ph.9344 ( 2006 ) r. mundt , obs , 105 , 224 ( 1985 ) r. mushotzky , prog .
suppl . , 155 , 27 ( 2004 ) r. narayan , & i. yi , apj , 428 , 13 ( 1994 ) k. ohsuga , m. mori , t. nakamoto , s. mineshige , apj , 628 , 368 ( 2005 ) r. e. pudritz , & c. a. norman , apj , 301 , 571 ( 1986 ) b. punsly , & s. j. tingay , apj , 633 , 89 ( 2005 ) n. shakura , & r. sunyaev , a&a , 24 , 337 ( 1973 ) g. v. ustyugova , a. v. koldoba , m. m. romanova , v. m. chechetkin , & r. v. e. lovelace , apj , 516 , 221 ( 1999 ) , ( d ) bernoulli s constant , as functions of radial coordinate for super - eddington accretion flows .
( e ) variation of bernoulli s constant as a function of @xmath16 .
solid , dotted , dashed curves are for @xmath56 and @xmath77 respectively .
long - dashed , dot - dashed curves represent flow with @xmath78 and @xmath79 respectively .
other parameters are @xmath80 , corresponding @xmath81 , and @xmath82 . ]
|
we study the energetics of the accretion - induced outflow and then plausible jet around black holes / compact objects using a newly developed disc - outflow coupled model .
inter - connecting dynamics of outflow and accretion essentially upholds the conservation laws .
the energetics depend strongly on the viscosity parameter @xmath0 and the cooling factor @xmath1 which exhibit several interesting features .
the bolometric luminosities of ultra - luminous x - ray binaries ( e.g. ss433 ) and family of highly luminous agns and quasars can be reproduced by the model under the super - eddington accretion flows . under appropriate conditions , low - luminous agns ( e.g. sagittarius @xmath2 )
also fit reasonably well with the luminosity corresponding to a sub - eddington accretion flow with @xmath3 .
keywords : accretion , accretion disc black hole physics x - rays : binaries galaxies : jets galaxies : nuclei [ firstpage ]
|
single crystals of ( c@xmath0h@xmath1n)@xmath2cu@xmath3zn@xmath4br@xmath5 were grown from solution by the temperature gradient method @xcite , according to the same procedure as for the original material ( c@xmath0h@xmath1n)@xmath2cubr@xmath5 @xcite but with replacement of the relevant amount @xmath18 of cubr@xmath2 by znbr@xmath2 . for the magnetic measurements
single crystals with typical masses of 15 mg were used .
the measurements were carried out with the help of a standard quantum design magnetic properties measurement system ( mpms - xl squid magnetometer ) . for measurements below 1.7 k ,
the @xmath68he cryostat insert for mpms ( iquantum ihelium3 ) was employed .
for all the samples , the magnetic field was applied along the @xmath69 axis of the structure . for the synthesis of single crystals for neutron scattering ,
all the hydrogen - containing chemicals were replaced by the deuterated analogues .
similar to our studies on the clean material @xcite , the measured sample was composed of three co - aligned crystals of ( c@xmath0d@xmath1n)@xmath2cu@xmath42zn@xmath43br@xmath5 with typical masses of 800 mg each .
they were fixed on a similar aluminum sample holder as previously used ( see e. g. fig . 2 of ref .
@xcite ) , with the @xmath70 direction being vertical .
hereby , we would like to provide an additional figure , showing the part of reciprocal space accessed in the in5 neutron experiment as well as the 1d spin island scattering .
data in fig .
[ fig : neutron ] ( measured with the in5 instrument @xcite , with @xmath71 mev and @xmath46 mk ) was integrated along the out - of - plane direction @xmath72 and in the energy range @xmath54 . the spin island scattering is readily observed as a stripe of intensity , located at @xmath73 r.l.u .. d@xmath1n)@xmath2cu@xmath3zn@xmath4br@xmath5 with @xmath56 , as measured with @xmath71 mev and @xmath46 mk .
uncorrected data is shown as a function of @xmath74 ( along the ladder leg ) and the perpendicular in - plane direction .
data was integrated in the out - of - plane direction @xmath75 ( @xmath72 ) and in the energy range @xmath54 . ]
in this section , further details about the numerical simulations are provided .
we follow closely the previous work @xcite , studying the general @xmath76 coupling case .
the shape of the emergent magnetization profile around an impurity was studied by placing a single @xmath77 site in the center of a ladder with @xmath78 spins and by calculating the local magnetization @xmath79 of the ground state with @xmath80 .
[ fig : shape ] illustrates the obtained local magnetization @xmath79 using the density matrix renormalization group ( dmrg ) algorithm from the alps libraries @xcite ( @xmath81 mev and @xmath82 mev , @xmath83 dmrg states kept ) .
the local magnetization follows an exponential law without further corrections as expected @xcite . by fitting @xmath84 with a linear function ,
the exponential decay length was determined to be @xmath85 .
if two impurities are introduced , two spin islands emerge , represented by effective spins @xmath27 at the impurity sites , with @xmath91 @xcite .
their interaction depends on the distance along the leg ( @xmath92 ) and rung @xmath93 and due to their effective interaction , they pair up to a dimer with a @xmath77 singlet and a @xmath7 triplet state at energies well below the gap energy of the ladder system @xcite .
depending on whether @xmath94 is odd or even , the effective interaction is ferro- or antiferromagnetic ( fm or afm ) .
the mutual effective interactions were determined based on matrix product state ( mps ) @xcite calculations of ladder systems with @xmath95 sites and two non - magnetic impurities placed at a predefined distance @xmath92 along the leg and either in cis or trans configuration ( impurities on the same leg or different legs @xcite ) .
the two impurities were centered around the middle of the ladder . in order to minimize boundary effects , the distance to the ladder ends is larger than 4 - 5 correlation lengths .
following the approach described in refs .
@xcite , the two lowest energies in the symmetry sectors with total spin projection of @xmath96 and @xmath97 were calculated .
if spin - island interactions are afm , the lowest state corresponds to a singlet ( @xmath77 , @xmath96 ) while the first excited state is a triplet ( @xmath7 , @xmath98 ) and vice versa for fm interactions .
the energy difference between the lowest two states in the @xmath96 sector thus corresponds to @xmath29 while the sign is determined from the lowest energy state of the @xmath97 sector , as discussed in refs .
@xcite .
calculations were performed with the matrix mps algorithm from the alps package @xcite ( 250 states kept ) . as an independent check - up , the results published in table ii of ref .
@xcite were successfully reproduced first . in the latter ,
a similar numerical study was performed for the @xmath99 coupling case , close to the dimpy coupling ratio of @xmath100 .
the determined effective interactions are shown in fig .
[ fig : int ] . as previously noted in refs .
@xcite , they follow @xmath101 where @xmath102 and @xmath103 . the parameters @xmath104 and @xmath61 were determined to be @xmath105 and @xmath106 unit cells for the dimpy coupling case , in agreement with the trend shown in fig . 2 and fig
5 of ref .
@xcite .
since the original heisenberg spin ladder hamiltonian is non - frustrated , the thermodynamic properties are accessible with numerical quantum monte carlo ( qmc ) calculations . the temperature - dependent susceptibility @xmath107 and the magnetization @xmath108 were calculated using the qmc algorithms provided by the alps package @xcite .
calculations were performed for @xmath109 mev and @xmath82 mev , on depleted ladders with @xmath110 sites and a fixed number @xmath111 of randomly placed non - magnetic sites .
similar to ref .
@xcite , all thermodynamic quantities were averaged over 300 random configurations ( total computation time : 2 months , parallelized to 32 - 64 cores on the brutus cluster at eth zrich ) .
the susceptibility was calculated using the loop algorithm , for @xmath112 ( fig .
[ fig : comparison ] , coloured symbols in upper panel ) .
the field - dependent magnetization was calculated as a function of magnetic field , for @xmath113 , for @xmath114 k ( fig .
[ fig : comparison ] , coloured symbols in lower panel ) . for the latter , the stochastic series expansion ( sse ) algorithm with directed loops
@xcite was applied , similar to ref .
@xcite .
a small number of effective spins @xmath27 was randomly placed on @xmath117 non - equal positions of a ladder with @xmath39 rungs . from this random arrangement , the mutual distances @xmath118 and the corresponding interactions @xmath29 were determined . before setting up the hamiltonian , two special cases were considered ( fig . [ fig : specialcases ] ) .
first , impurities on adjacent rungs in trans position ( @xmath119 and @xmath120 ) split the system into two non - interacting segments , releasing a spin island on both sides .
second , impurities on the same rung ( @xmath121 and @xmath120 ) cut the ladder into to non - interacting fragments as well .
however , in contrast , the latter release no additional spin island . based on these requirements , the hamiltonian was set up according to eq .
( [ eq : hameff ] ) and by exactly diagonalizing ( ed ) the latter , both eigenvalues @xmath122 with spin projection quantum number @xmath123 and eigenvectors @xmath124 were obtained . from the former , thermodynamic quantities such as the magnetization and the susceptibility
were calculated , @xmath125 with @xmath126 , @xmath127 and @xmath128 \notag\\ \langle m^\alpha \rangle & = \frac{1}{\mathcal{z}}\ , \sum\limits_{\mu } m_{s,\mu}^\alpha\,\exp\biggl[-\frac{e_\mu - h\,m_{s,\mu}}{k_\mathrm{b}t}\biggr].\end{aligned}\ ] ] the results were averaged over many repetitions @xmath129 of randomly chosen configurations of @xmath117 impurities on a ladder with @xmath39 rungs . the susceptibility and field - dependent magnetization ( @xmath130 ) are shown in fig .
[ fig : comparison ] as black lines , for different @xmath18 .
calculations were performed for a fixed number of @xmath37 impurities randomly placed on ladders with @xmath131 , @xmath132 , @xmath133 , @xmath26 , @xmath134 and @xmath135 sites for @xmath14 , @xmath136 , @xmath137 , @xmath138 , @xmath139 and @xmath140 respectively .
the result was averaged over @xmath141 random configurations . by performing calculations for different number of impurities @xmath142
, we estimate that despite the small number of impurities , the mean deviation of the thermodynamic quantities from the large-@xmath117 limit is less than 2% for @xmath15 and even less for @xmath143 .
calculations according to the effective model are similar to the ones published in ref .
@xcite for the general @xmath144 case ( with @xmath145 effective spins and @xmath146 repetitions ) , except for the additional special cases considered in this study ( fig . [ fig : specialcases ] ) . notably and within the range of applicability ,
the agreement between the ed calculations according to the effective model and the qmc calculation of the full depleted ladder hamiltonian is excellent , for all concentrations considered and both for the susceptibility and magnetization .
we note that the ed treatment of the effective model provides access to the eigenvectors @xmath124 as well , thus enabling to calculate dynamical properties such as local correlations @xmath147 , measurable with inelastic neutron scattering @xcite . finally , it is instructive to compare the qmc and ed numerical results in fig .
[ fig : comparison ] to the simplified model based on the naive assumption that each non - magnetic ion is accompanied by a spin island and that the latter are non interacting , thus leading to a brillouin response ( gray lines in fig .
[ fig : comparison ] ) @xmath148 .
\label{eq : fs}\end{aligned}\ ] ] at very low concentration , this model becomes a faithful description since spin islands are well separated on average ( mean distance @xmath149 ) and thus non - interacting .
however , at @xmath150 , the free spin model becomes inappropriate in two ways .
first , interactions between mutual spin islands become progressively more important and deviations from the brillouin response apparent .
second , the probability of impurities occupying the same rung increases as well . in the latter case ( fig .
[ fig : specialcases]b ) , no spin island is created at all .
the probability that , in a system with impurity concentration @xmath18 , an impurity shares a rung with another impurity is @xmath151 @xcite . since two impurities are needed to form such a non - magnetic rung pair and since both do not release a spin island , the saturated magnetic moment of the effective model is @xmath152 instead of @xmath18 as in eq .
( [ eq : fs ] ) .
14ifxundefined [ 1 ] ifx#1 ifnum [ 1 ] # 1firstoftwo secondoftwo ifx [ 1 ] # 1firstoftwo secondoftwo `` `` # 1''''@noop [ 0]secondoftwosanitize@url [ 0 ]
+ 12$12 & 12#1212_12%12@startlink[1]@endlink[0]@bib@innerbibempty link:\doibase 10.1080/14786435.2012.669072 [ * * , ( ) ] link:\doibase 10.1021/ja066330 m [ * * , ( ) ] link:\doibase 10.1103/physrevb.84.144421 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.108.167201 [ * * , ( ) ] link:\doibase 10.1103/physrevb.88.094411 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.111.107202 [ * * , ( ) ] link:\doibase 10.1103/physrevb.91.020406 [ * * , ( ) ] link:\doibase 10.1143/jpsjs.80sb.sb003 [ * * , ( ) ] link:\doibase 10.1103/physrevb.55.2955 [ * * , ( ) ] link:\doibase 10.1103/physrevb.88.134420 [ * * , ( ) ] link:\doibase 10.1088/1742 - 5468/2011/05/p05001 [ * * , ( ) ] link:\doibase 10.1016/j.cpc.2014.08.019 [ * * , ( ) ] link:\doibase 10.1103/physreve.66.046701 [ * * , ( ) ] @noop
|
properties of the depleted heisenberg spin ladder material series ( c@xmath0h@xmath1n)@xmath2cu@xmath3zn@xmath4br@xmath5 have been studied by the combination of magnetic measurements and neutron spectroscopy .
disorder - induced degrees of freedom lead to a specific magnetic response , described in terms of emergent strongly interacting `` spin island '' objects .
the structure and dynamics of the spin islands is studied by high - resolution inelastic neutron scattering .
this allows to determine their spatial shape and to observe their mutual interactions , manifested by strong spectral in - gap contributions . in the solid state
, even weak perturbations may lead to qualitatively new physics described in terms of entirely new emergent degrees of freedom and quasiparticles @xcite .
one such perturbation , known to open the door to a variety of novel and competing ground states and rich phase diagrams , is structural or chemical disorder @xcite .
an exciting example of disorder - induced emergent degrees of freedom are the _
magnetic _ objects that appear upon the introduction of _ non - magnetic _ impurities in gapped quantum - disordered antiferromagnets ( afs ) @xcite . these entities may be understood as spins released from non - magnetic af singlets by removing their partner spins .
the short - range correlations in the underlying quantum af spread these spin degrees of freedom over extended regions ( `` spin - islands '' ) around each impurity site @xcite .
the size of the spin islands is controlled by the correlation length in the parent system , and may be as large as dozens of nanometers . due to their partial overlap ,
these spin islands interact .
the original quantum af thus acts as a `` medium '' that hosts a new magnetic system of _ mesoscopic _ objects and carries interactions between them . due to these interactions
, the emergent system may have its own unique correlations and dynamics .
the impurity - induced formation of localized @xmath6 spin objects has been thoroughly studied in gapped @xmath7 haldane spin chains @xcite .
unfortunately , in these highly one - dimensional chain materials , each impurity completely severs the host system .
the emergent spin islands located at the chain segment ends merely pair up into isolated dimers @xcite , and have no collective dynamics .
this is in contrast to expectations for heisenberg spin ladders , which are composed of _ two _ neighboring chains linked by rung interactions @xmath8 . within this topology ,
the non - magnetic impurities have a low chance to disrupt the continuity of the system and a finite - size segment typically contains a large number of mutually interacting islands . to date , the study of non - magnetic impurities in spin ladders was limited to rather complex materials with prohibitively large energy scales , such as sr(cu@xmath3zn@xmath9)@xmath2o@xmath10 @xcite or bi(cu@xmath3zn@xmath9)@xmath2po@xmath11 @xcite .
these cases are difficult to describe theoretically , as they involve additional interactions such as cyclic exchange , frustration , or 3d coupling .
what is missing is an experimental study of emergent and interacting spin islands in a _
clean _ realization of the spin ladder model , such that could be compared to theoretical calculations at the quantitative level .
fortunately , in recent years , a number of new exceptionally good organic spin ladder compounds were discovered @xcite . among them , ( c@xmath0h@xmath1n)@xmath2cubr@xmath5(dimpy ) @xcite realizes the rare case of a strong - leg spin ladder , which is of special interest in the context of the present study .
its dominant leg interactions imply a significant correlation length and hence significant interactions between emerging spin objects , even at low concentration . in the present letter
we show that dimpy , diluted with nonmagnetic zinc , is indeed a perfect prototype for testing this physics . by combining thermodynamic and high - resolution inelastic neutron scattering experiments with numerical simulations ,
we are able to directly study the shape and interactions of the spin islands , and to quantitatively describe the impurity - induced _ collective _ response in the language of these emergent objects . h@xmath1n)@xmath2cu@xmath3zn@xmath4br@xmath5 .
( b ) spin ladder model microscopic description .
the random replacement of magnetic cu@xmath12 by non - magnetic zn@xmath12 renders some @xmath6 sites missing .
( c ) emerging local spin degrees of freedoms ( spin islands ) , pinned to the impurity position but with a magnetization profile extending over many unit cells .
( d ) effective spin island hamiltonian ( [ eq : hameff ] ) with interactions controlled by the mutual distances as given by eq .
( [ eq : jeff ] ) . ] h@xmath1n)@xmath2cu@xmath3zn@xmath4br@xmath5 .
( a ) susceptibility as a function of temperature .
( b ) isothermal magnetization at @xmath13 mk for the same samples .
symbols are the experimental data .
numerical qmc simulations and results from the diagonalization of the effective model [ eq .
( [ eq : hameff ] ) ] are shown as dashed and solid lines correspondingly .
dotted lines in ( b ) illustrate the free - spin response for @xmath14 and @xmath15 cases.,scaledwidth=50.0% ] the parent material dimpy was shown to be described by a simple heisenberg spin - ladder hamiltonian with @xmath16 mev along the leg and @xmath17 mev ( see fig .
[ figspinisland ] ) @xcite and with exceptionally weak additional interactions @xcite .
we were able to chemically introduce non - magnetic impurities by replacing a fraction @xmath18 of magnetic @xmath6 cu@xmath12 by non - magnetic zn@xmath12 ions @xcite .
the magnetic susceptibility and isothermal magnetization data are presented in fig .
[ figtd ] , as measured on a series of ( c@xmath0h@xmath1n)@xmath2cu@xmath3zn@xmath4br@xmath5 single crystals with varying impurity concentration @xmath18 @xcite .
while the susceptibility of the @xmath19 parent compound is being dominated by a spectral gap of @xmath20 mev and exponentially decays to zero at low temperatures , the susceptibility of the zn - diluted material acquires an additional paramagnetic - like contribution , progressively increasing with the impurity concentration .
this behavior becomes even more apparent in the low - temperature magnetization ( fig .
[ figtd]b ) .
the latter remains suppressed by the gap below the critical field of @xmath21 t @xcite in the clean material but acquires an impurity - induced contribution in the zn - substituted derivatives .
nonetheless , the observed response is not simply described by the @xmath6 brillouin function , as it would in case of free magnetic moments @xcite . while deviations from the brillouin function are small at low impurity concentrations , they become up to 45% at @xmath15 .
qualitatively , this is explained by the mean impurity distance @xmath22 @xcite which is as large as 50 unit cells for @xmath14 but becomes comparable to the correlation length @xmath23 @xcite at @xmath15 .
the probability of finding close and strongly interacting islands thus rapidly increases with impurity concentration . as a first step , we compare our measurements to numerical calculations @xcite based on the full parent spin ladder hamiltonian @xcite , depleted with randomly placed non - magnetic sites ( fig .
[ figspinisland]b ) . following ref .
@xcite , we calculate the susceptibility and magnetization with quantum monte carlo ( qmc ) simulations of ladder systems with @xmath24 rungs , @xmath25 randomly placed non - magnetic sites and averaged over @xmath26 random impurity configurations @xcite . the numeric results ( dashed lines in fig .
[ figtd ] ) quantitatively reproduce the measured data in the entire temperature- and field - range .
this proves not only that we are able to chemically control the zinc concentration but also that its introduction does not lead to significant local distortion effects altering the superexchange interactions .
extensive qmc calculations of large depleted systems , averaged over hundreds of configurations , are expensive , time - consuming and yield little physical insight .
an alternative , much simpler approach is provided by the effective low - energy description in terms of interacting spin islands ( fig .
[ figspinisland]c , d ) @xcite .
the latter are represented by @xmath6 spins @xmath27 at random impurity positions @xmath28 , interacting with distance - dependent effective interactions @xmath29 and described by the hamiltonian @xcite @xmath30 with @xmath31 and @xmath32 being the bohr magneton and the electron @xmath32-factor .
mutual spin - island interactions are mediated by the spin correlations in the hosting ladder medium .
since the latter are short - ranged , effective interactions exponentially decay with distance .
in addition , due to their antiferromagnetic nature , @xmath29 are ferro- or antiferromagnetic , depending on whether the mutual interaction path contains an odd or even number of sites @xcite , reminiscent of the oscillatory ruderman - kittel - kasuya - yosida ( rkky ) interaction @xcite .
being fully controlled by the spin ladder medium , the effective interactions can be numerically obtained from a system with two non - magnetic sites at distance @xmath33 in a long ladder system . following refs .
@xcite we have performed matrix product state calculations @xcite and confirmed that @xmath29 is described by the simple law @xmath34 with the energy scale @xmath35 mev and the decay length @xmath36 for the exchange constants of dimpy @xcite . by including the numerically calculated effective interactions , eq .
( [ eq : hameff ] ) becomes a _
parameter - free _ description of the emergent spin island system .
the magnetization and susceptibility of the effective model were determined @xcite by exactly diagonalizing ( ed ) the hamiltonian in eq .
( [ eq : hameff ] ) , for a system of @xmath37 sites @xmath38 , randomly placed on a ladder with @xmath39 rungs and with the effective interactions given by eq .
( [ eq : jeff ] ) .
the magnetization , averaged over 5000 random configurations , is in excellent agreement with measured data ( fig .
[ figtd]b ) .
notably , a model described by not more than 12 mutually interacting spin operators is sufficient to model the entire thermodynamic response of this complex disordered many - body quantum system . nevertheless , the description in terms of eq .
( [ eq : hameff ] ) remains valid only as long as the mutual interactions controlled by the spin island correlations do not change .
while the latter remain unaffected by temperature and applied field as long as the spin ladder remains in its quantum disordered regime , they fundamentally change once the system crosses the critical field @xmath40 t @xcite to the quantum critical tomonaga - luttinger spin liquid ( tlsl ) state . above @xmath41
the effective description naturally fails .
so far , we have successfully described the measured thermodynamic properties in terms of an effective model of emergent spin island .
however , can such objects and their interactions be observed directly ? to answer this question , we present a series of high - resolution inelastic neutron scattering experiments enabling to access the spectral properties of both the spin ladder medium and the low - energy spin island system . in fig .
[ figneutron1]a , c , we show the background - subtracted magnetic neutron spectrum of ( c@xmath0d@xmath1n)@xmath2cu@xmath42zn@xmath43br@xmath5 , as measured with the neutron time - of - flight ( tof ) technique ( let spectrometer @xcite , isis facility , u.k . ) .
data was gathered in distinct low- and high - resolution setups with incident energies @xmath44 mev and @xmath45 mev ( @xmath46 mk ) .
neutron intensity is shown as a function of energy transfer @xmath47 and momentum transfer @xmath48 along the leg direction .
corresponding data from the parent compound is shown in figs .
[ figneutron1]b , d , measured under identical experimental conditions @xcite and treated in the same way .
mev ) .
( c , d ) low - energy sector in the higher resolution mode ( @xmath49 mev ) .
background corrected magnetic neutron scattering intensity is shown as a function of energy transfer @xmath47 and momentum transfer along the leg @xmath50 .
hatched regions are contaminated by parasitic nuclear incoherent scattering . ] first , we observe that the properties of the spin ladder medium hosting the spin island system remains nearly unaffected by the presence of impurities ( fig .
[ figneutron1]a , b ) . in both the parent and disordered compound , identical
gapped and dispersive magnon and two - magnon bound state branches @xcite are observed .
in contrast , clear changes are found in the low - energy sector ( fig .
[ figneutron1]c , d ) .
the magnon gap is slightly shifted from @xmath20 to @xmath51 mev , reminiscent of the magnon blue - shift observed in 1d quantum - disordered afs at finite temperature @xcite . however , as a main feature , we observe a stripe of enhanced in - gap intensity developing around @xmath52 r.l.u . in the disordered compound .
this contribution originates in strongly interacting spin islands , leading to a disorder - induced finite density of states inside the gap .
mev .
neutron intensity is shown as a function of energy transfer @xmath47 and momentum transfer @xmath50 .
( b ) neutron intensity as a function of @xmath47 for @xmath53 between 0.45 and 0.55 r.l.u .
data in the hatched area is contaminated by parasitic nuclear incoherent scattering .
( c ) scattering intensity as a function of @xmath50 , after integrating in @xmath54 .
instrumental resolution ( shaded area ) is shown along with the resolution - convoluted fit ( [ eq : shape ] ) [ solid line ] . ]
it is even more pronounced in the spectrum presented in fig .
[ figneutron2]a , measured with a lower incident energy of @xmath55 mev and thus better resolution ( in5 instrument @xcite , ill , france ) .
the spin island contribution progressively increases with decreasing energy @xmath47 , as shown in fig .
[ figneutron2]b .
this is readily explained since at @xmath56 , the probability of finding spin islands only weakly interacting with others remains large and there is a vast number of quantum states at low energies . as a main characteristics , these in - gap contributions exhibit an intrinsic @xmath50 linewidth , as shown in fig .
[ figneutron2]c .
this intrinsic linewidth is a measure of the spin island real - space magnetization profile .
similar to conventional magnetic neutron scattering @xcite , where intensity is modulated by the magnetic form factor ( the fourier transform of the magnetic ions unpaired electron density ) , scattering by spin islands is modulated by the square of the spin island form factor @xmath57 the fourier transform of its real space shape .
a similar reasoning was applied to determine the spatial shape of emergent local objects in the substantially different cases of the charge - doped or depleted haldane spin chains y@xmath58ca@xmath4banio@xmath59 @xcite , y@xmath2bani@xmath3mg@xmath9o@xmath59 @xcite . for a spin island with exponentially decaying staggered magnetization on a ladder ,
the form factor reads as @xmath60 here , @xmath61 denotes the real space decay length and @xmath62 is the momentum transfer along the rung @xcite .
we compared @xmath63 , averaged over @xmath62 , to our experimental data and by convolution with experimental resolution , we were able to quantify the real space decay length to be @xmath64 .
the magnetization profile thus decays to 10% of its initial value only after 15 unit cells and the size of an emergent spin island is of the order of 20 nm .
notably , the determined decay length is slightly shorter than the numerically calculated value of @xmath65 @xcite .
however , this is not surprising since in - gap states observed in experiments are located at a comparatively large energy transfer on the order of 0.1 mev and thus mainly originate from dimers or clusters of spin islands close to each other .
influenced by the nearby spin islands , the magnetization profile thus deviates from the one of an isolated spin island . in conclusion , by combining thermodynamic and high - resolution inelastic neutron scattering techniques with numerical simulations
, we were able to quantitatively study the properties of a system of strongly interacting spin islands formed in a strong - leg heisenberg spin ladder depleted with non - magnetic impurities .
an effective description in the language of these objects explains and quantitatively reproduces the measured data .
as long as the spin - ladder medium remains quantum disordered at @xmath66 , this effective description is faithful and , in principle , would allow not only to describe the thermodynamic response but even to access dynamical quantities such as time - dependent correlation functions @xcite .
beyond @xmath67 , however , the effective description breaks down due to the quantum phase transition to the tlsl state @xcite , rendering the spin - ladder correlation functions fundamentally different . to the authors knowledge ,
the problem of non - magnetic impurities and the fate of spin islands in the tlsl phase of a spin ladder or their influence on the magnetically ordered state @xcite has not been addressed experimentally or theoretically .
we hope that our study stimulates further efforts and experiments towards understanding the impurity - induced physics in such systems .
this work is partially supported by the swiss national science foundation , division ii .
we would like to thank dr .
s. mhlbauer ( frm ii , technische universitt mnchen ) for his involvement at the early stages of this project and dr . s. gvasaliya ( eth zrich ) for assistance with the magnetic measurements .
simulations were performed on the brutus cluster ( eth zrich ) .
42ifxundefined [ 1 ] ifx#1 ifnum [ 1 ] # 1firstoftwo secondoftwo ifx [ 1 ] # 1firstoftwo secondoftwo `` `` # 1''''@noop [ 0]secondoftwosanitize@url [ 0 ]
+ 12$12 & 12#1212_12%12@startlink[1]@endlink[0]@bib@innerbibempty link:\doibase 10.1126/science.177.4047.393 [ * * , ( ) ] link:\doibase 10.1126/science.1107559 [ * * , ( ) ] link:\doibase 10.1103/physrevb.55.2955 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.90.087202 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.103.047201 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.71.4059 [ * * , ( ) ] link:\doibase 10.1143/jpsj.65.1392 [ * * , ( ) ] http://stacks.iop.org/0953-8984/14/i=10/a=201 [ * * , ( ) ] link:\doibase 10.1038/nphys3186 [ * * , ( ) ] link:\doibase 10.1038/ncomms8061 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.67.1614 [ * * , ( ) ] link:\doibase 10.1016/0921 - 4526(94)90416 - 2 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.83.412 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.83.632 [ * * , ( ) ] link:\doibase 10.1103/physrevb.66.174416 [ * * , ( ) ] link:\doibase 10.1103/physrevb.55.r8658 [ * * , ( ) ] link:\doibase 10.1002/ejic.201300133 [ * * , ( ) ] link:\doibase
10.1021/ja066330 m [ * * , ( ) ] link:\doibase 10.1103/physrevlett.108.167201 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.111.107202 [ * * , ( ) ] link:\doibase 10.1103/physrevb.91.020406 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.111.106404 [ * * , ( ) ] link:\doibase 10.1103/physrevb.92.184403 [ * * , ( ) ] link:\doibase 10.1103/physrevb.92.241113 [ * * , ( ) ] link:\doibase 10.1080/14786435.2012.669072 [ * * , ( ) ] @noop `` , '' @noop _ _ , ed .
( , , ) link:\doibase 10.1103/physrevb.88.134420 [ * * , ( ) ] link:\doibase 10.1088/1742 - 5468/2011/05/p05001 [ * * , ( ) ] link:\doibase 10.1103/physrev.96.99 [ * * , ( ) ] link:\doibase 10.1143/ptp.16.45 [ * * , ( ) ] link:\doibase 10.1103/physrev.106.893 [ * * , ( ) ] link:\doibase 10.1016/j.cpc.2014.08.019 [ * * , ( ) ] link:\doibase 10.1016/j.nima.2011.01.173 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.100.157204 [ * * , ( ) ] link:\doibase 10.1143/jpsjs.80sb.sb003 [ * * , ( ) ] @noop _ _ ( , , ) link:\doibase 10.1126/science.289.5478.419 [ * * , ( ) ] link:\doibase 10.1103/physrevb.84.144421 [ * * , ( ) ] @noop @noop _ _ ( , , ) link:\doibase 10.1038/nphys893 [ * * , ( ) ] * supplementary material : + emergent interacting spin islands in a depleted strong - leg heisenberg ladder *
|
in the currently favored cosmological constant plus cold dark matter ( @xmath3cdm ) model of cosmology , structure forms hierarchically : small dark matter haloes collapse at high redshifts , while larger and more massive objects form through a series of minor mergers that accrete smaller mass haloes and major mergers with comparable mass haloes . since dark matter haloes provide the gravitational potential wells necessary for galaxy formation @xcite , a detailed understanding of the effects of mergers is necessary in order to test models of galaxy formation and evolution within the @xmath3cdm paradigm .
recent numerical studies indicate that the hierarchical build - up process results in haloes with significant substructure populations : hundreds to thousands of subhaloes to the smallest resolvable mass scales appear to reside within galaxy sized haloes , comprising approximately 10% of a halo s mass @xcite .
a number of recent papers have investigated the evolution and detailed properties of the substructure using both semi - analytic models and numerical simulations .
depending on factors such as the relative mass , pericentric distance , and halo concentration , the orbits and masses of the subhaloes can be strongly affected by dynamical friction and tidal stripping ( e.g. @xcite ) .
dense subhaloes that survive the tidal process can sink to the centre of the parent halo , impacting its inner structure ( e.g. @xcite ) .
in contrast to minor mergers , where large haloes swallow up smaller ones in a relatively gentle fashion , major mergers are more violent events involving systems of approximately equal mass that can be studied in detail with numerical simulations only .
the interest in major mergers has a long history .
@xcite noted that many observed features of galaxy pairs can be explained by tidal interactions that could promote merging .
they further speculated that if such interactions were common enough , elliptical galaxies might naturally be explained as the merger remnants of spiral galaxies .
this suggestion has led to numerous studies of galaxy merging and has been quite successful in helping explain the origin of ellipticals . in some of the earliest simulations
, @xcite studied several types of mergers using 250 simulation particles per halo and concluded that products of mergers of cored galaxies were significantly more centrally concentrated than their progenitors , with more extended haloes having an approximate power - law density profiles @xmath4 that resembled the de vaucouleurs empirical profile for the surface - brightness of ellipticals @xcite .
the higher resolution simulations by @xcite with @xmath5 particles per halo also exhibited the @xmath6 behavior with evidence for homology .
simulations with multiple components representing dark matter haloes , stellar disks , and stellar bulges have been a source of much theoretical progress ( e.g. @xcite and references therein ) and are essential for elucidating the dynamical coupling between the dark matter and stellar components in galaxies and for comparing simulation results with observations .
pure dark halo simulations , however , are required for a controlled study of the non - baryonic dark matter that dominates the gravitational potential of seemingly all galactic systems . in this paper
we perform high - resolution @xmath0-body simulations of mergers of two equal - mass dark matter haloes and study how the phase - space distribution of dark matter particles is affected by such violent events and how the merger product relaxes into equilibrium .
we pay particular attention to the central regions of the haloes and examine whether the initial inner structure of a dark matter halo , either with a flat core or a sharp cusp , is disrupted or preserved by major collisions .
our study complements the numerous recent cosmological @xmath0-body studies of the properties and evolution of dark matter haloes , most of which have reached the somewhat surprising conclusion that over a wide range of mass scales , these haloes are universally cuspy and have radial density profiles with @xmath7 ( with @xmath8 ) in the central region and @xmath4 near the virial radius @xcite .
the seemingly generic prediction of an inner cusp agrees with earlier simulations ( e.g. @xcite ) but is contrary to the inferred dark matter distribution in some dwarf and low surface brightness galaxies ( e.g. @xcite , though also see @xcite ) as well as in spiral galaxies @xcite and has lead to speculation that the @xmath3cdm model is in crisis .
the simulations of major mergers performed in this paper provide a set of controlled experiments for us to address questions such as : how does the phase space density of dark matter haloes evolve and re - equilibrate during major mergers ?
is the cuspy profile an attractor region of the phase space so that merger remnants of two non - cuspy ( i.e. cored ) haloes will end up cuspy
? do cuspy haloes , once formed ( by any process ) , retain their inner structure after inevitable major mergers with other haloes ?
unlike cosmological simulations that are designed to incorporate many complicated processes at once , the simulations performed here allow us to separate out the effects of minor mergers and to focus on how major collisions change the spatial and velocity structures of any initial haloes of our choice .
additionally , cosmological parameters are unimportant for the simulations themselves since major mergers are local processes of two fully collapsed objects . in this case , all we need is the initial phase space distribution of each galaxy halo and the initial relative positions and velocities of the haloes themselves . by simulating collisions of two galaxy haloes with high spatial resolution and up to 5 million particles ( for each halo ) , we are able to follow their phase - space evolution and the density profiles of the remnants down to @xmath9 @xmath10 in our highest resolution simulation .
section 2 of this paper describes the initial setup and the parameters used in the numerical simulations .
we also discuss in some detail the tests performed to quantify the numerical effects of two - body relaxation on the central regions of the haloes and the technical issue of how to determine the centre of a merged halo .
section 3 contains our results from the ten production runs for the energy , density profiles , velocity distributions , and phase space structure of merger remnants .
section 4 discusses the energy distribution and the relaxation of a merged halo to an equilibrium state , with an emphasis on violent relaxation and phase mixing .
section 5 contains a summary of our results .
we study two general types of inner density profiles for galaxy haloes : cored and cuspy . for the cored profile , we use an isothermal sphere with a core radius and an exponential cut - off @xcite : @xmath11 where @xmath12 is the halo s mass ,
@xmath13 is a normalization constant given in terms of @xmath14 as @xmath15 , @xmath16 is the outer exponential fall - off radius and @xmath17 is the core radius . for the cuspy haloes , we use the navarro , frenk , and white ( nfw ) profile @xcite : @xmath18 with scale radius @xmath19 and characteristic density @xmath20 given by @xmath21}\end{aligned}\ ] ] where @xmath22 is the halo concentration parameter and
@xmath10 is the radius at which the average interior density is 200 times the critical density of the universe @xmath23 .
( we also adopt @xmath10 as a halo s virial radius and therefore @xmath24 as its virial mass . )
we generate our initial conditions using the methods described by @xcite .
the initial positions of the particles in a halo are drawn randomly from its density profile .
all of our initial haloes are chosen to be @xmath25 . in the case of the nfw profile
the total mass is formally infinite , so it is necessary to introduce a truncation radius for the halo .
we choose to truncate our haloes at @xmath26 . for our choice of @xmath27 ,
the virial and truncation radius is 162.6 @xmath28 kpc and the scale radius is @xmath29 @xmath28 kpc .
for the cored profile , we choose @xmath30 @xmath28 kpc , @xmath31 @xmath28 kpc , and truncate the halo at @xmath32 .
this choice of parameters allows us to start with two types of galaxies , equal in mass and radius , but with very different inner density profiles ( core vs. cusp ) . for the initial velocities of particles in the frame of a given halo ,
we first use the jeans equation ( the first moment of the collisionless boltzmann equation ) to relate the density profile @xmath33 and potential @xmath34 to the velocity dispersions @xcite : @xmath35 where the velocity anisotropy is parameterized by @xmath36 and @xmath37 is the dispersion of velocity component @xmath38 . if we further assume initial velocity isotropy , @xmath39 , we can integrate eqn .
( [ jeans ] ) to yield the dispersion of the radial velocity as a function of radius , @xmath40 from there , we randomly draw individual velocity components from a gaussian distribution with variance @xmath41 .
following @xcite , we impose the additional constraint @xmath42 , which insures that all particles are initially bound to the halo .
we set up the two colliding haloes to be just touching at the truncation radii initially .
the initial relative speed of the two haloes is taken to be the speed one of the haloes would have if the two were in circular orbit about their common centre of mass , i.e. , @xmath43 , where @xmath44 is the reduced mass and @xmath45 is the distance between the halo centres . if the two haloes were point particles , this configuration would give @xmath46 ; all our simulations are therefore relatively high - speed encounters for bound orbits . in this set - up , and with our choice of halo parameters , the two haloes first cross centres of mass at approximately 1.2 gyr , then again at 2.2 gyr .
this leaves ample time after the first encounter for the haloes to interact and evolve within the 5.0 gyr run .
most of our simulations are head - on collisions with an impact parameter @xmath47 . we have also performed two simulations in which the two halo centres are initially offset by @xmath48 in order to test the robustness of our conclusions . in terms of the dimensionless parameter @xmath49 for the total mass @xmath12 , energy @xmath50 , and orbital angular momentum @xmath51
, this corresponds to @xmath52 . in terms of the parameter @xmath53 , defined as the ratio of the angular momentum to that needed for circular orbit at the same velocity and separation , this set - up yields @xmath54 .
as a final test of the generality of our findings , we performed two simulations of collisions involving unequal mass haloes with cores .
.[table - runs ] properties of ten simulations of halo mergers .
all simulations involve two @xmath25 haloes ( except where noted ) , so @xmath55 [ cols= " < , < , < ,
< , < " , ] the final product of the core - core collision does _ not _ show signs of moving towards an nfw profile : the remnant has a core of approximately the same size as in the original haloes . in order to quantify our results ,
we have tried to fit the final profile to the same functional form as used for the initial cored profile ( eqn .
[ hernquist ] ) .
we find the inner region of the remnants well fit by a core radius of @xmath56 @xmath28 kpc ( for run `` core2 '' ) , which is nearly identical to the initial @xmath57 @xmath28 kpc , but the density in the outer region of the remnants falls off more gradually , as @xmath6 instead of the initial exponential cutoff in eqn .
( [ hernquist ] ) .
overall , we find the merged halo of core - core collisions better fit by @xmath58 with @xmath59 and @xmath60 @xmath28 kpc , than by the original profile ( [ hernquist ] ) .
since the remnants have virial radii about 20% larger than their progenitors , this means the relative size of the core in the final product is smaller , a result consistent with those of earlier simulations @xcite .
the high spatial and mass resolution used here allows us to simulate scales more than five times below the core radius and to conclude that the core is preserved during one head - on major merger . in the cusp - cusp collision ,
the product is essentially a new cuspy halo with an nfw - like form , which we find well fit by @xmath61 for radius down to the relaxation scale @xmath62 @xmath28 kpc for the @xmath63 particle run .
however , the best - fitting value for the logarithmic slope of the cusp is shallower than 1 , with @xmath64 , perhaps hinting that collisions of galaxies serve to reduce the cusp at very small radii .
the best - fitting parameters for the density profiles of our two highest resolution cusp - cusp collisions , as well as for the core - cusp collision , are summarized in table [ table - fits ] .
density profiles for the core - core ( dotted ) , cusp - cusp ( solid ) , and cusp - core ( dashed ) runs with @xmath65 particles at 5.0 gyr . the cusp - core collision results in a remnant with a density profile much more similar to the cusp - cusp collision than to the core - core one .
] in the case of the core - cusp merger , any hint of the core is entirely eliminated .
the remnant has an overall density profile much closer to nfw than to the original core model and again well fit by eqn .
( [ eqnnfw ] ) with an inner slope of @xmath66 .
to understand the core - cusp merger in more detail , we calculate the spherically - averaged density profile for particles in each halo separately , first at @xmath67 gyr and again at @xmath68 gyr ; the results are shown in fig .
[ rhomix ] .
interestingly , the particles originating from the cuspy halo maintain a cuspy , nfw - like profile ( left panel ) , while the particles originating from the cored halo retain their original density profile to a remarkable extent as well ( right panel ) .
the final density profile is dominated by the cusp particles in the inner @xmath69 kpc and by the core particles from @xmath70 kpc ; outside of @xmath71 kpc , both types of particles seem to contribute approximately equally ( also see fig . [ track2 ] ) .
the results above are all for head - on collisions with a zero impact parameter . in fig .
[ rho_ang ] we compare the density profiles of the non - zero impact parameter runs with the corresponding zero impact parameter simulations .
the profiles are almost exactly the same , leading us to the conclusion that introducing moderate amounts of orbital angular momentum into a merging system will not influence our results . to ensure that the results are not specific to equal - mass mergers
, we plot the density profiles of our two @xmath72 mass mergers in fig .
[ rho_unequal ] , along with the density profile of the remnant in an equal mass merger for reference .
the remnants of both unequal mass mergers retain cores of @xmath73kpc , the same result we found in the equal mass case .
in fact , the final density profiles in all three are remarkably similar ; this shows that core - core major mergers produce cored remnants .
density profiles for cusp - cusp ( thin lines ) and core - core ( thick lines ) runs with @xmath74 for @xmath47 ( solid ) and @xmath48 ( dashed ) at 3.5 gyr .
the profiles are extremely similar , indicating that introducing a moderate amount of orbital angular momentum does not significantly impact the merger remnant . ]
density profiles for three different core - core mergers at @xmath75 gyr . dashed line : run core2 ( equal mass , both core radii @xmath76 kpc ) .
dotted line : run core4 ( @xmath77 with core radii @xmath78 kpc , @xmath79 kpc ) .
dashed line : run core5 ( @xmath77 with core radii @xmath80 kpc ) .
all three mergers produce similar remnants with cores that are unreduced relative to the larger of the two initial cores , showing that mergers of cored haloes result in cored remnants . ]
although our simulations are collisionless , resulting in a phase space density that must be conserved due to liouville s theorem , we have no _ a priori _ reason to believe the configuration and momentum space distributions will be independently conserved , i.e. that a merger of two cored ( cuspy ) haloes will necessarily result in a cored ( cuspy ) remnant . a system s coarse - grained distribution function @xmath81 is constrained to be non - increasing as a function of time @xcite , and thus @xmath81 in the final state
is required at every point to be less than the maximal initial @xmath81 , but this requirement does not prevent cusps from becoming cores or vice - versa in collisionless simulations .
in fact , violent collisions provide strongly fluctuating potentials that cause particles to rapidly exchange energy with the background , a process that could quickly affect the phase space structure of both initial haloes . since the main objective of our study is to investigate the effects of major mergers on the phase - space distributions of dark matter in the inner regions of galaxy haloes , we examine this region in more detail in this section . in the core - core and cusp - cusp collisions
, we find that as expected , both progenitors contribute essentially equally to the particles at each radius of the merged halo . as figs .
[ densmulti ] and [ densall ] show , the inner density profile of the merger remnants is similar to their progenitors , i.e. , cored haloes remain cored ; cuspy haloes remain cuspy .
fraction of particles starting out within radius @xmath82 at t=0 gyr that also end up within radius @xmath82 at t=5 gyr ( from run mixed ) .
the cuspy particles ( solid line ) are much more likely to stay within the inner 10 kpc than the cored particles ( dashed line ) . ]
the situation is more intriguing for the mixed core - cusp collision . in order to elucidate the inner structure of this type of mergers ,
we first examine where the initial inner particles in each halo end up in the merged halo after 5 gyr .
( the results at 3.5 gyr are very similar . )
[ track1 ] shows the fraction of particles in each halo that start out within radius @xmath82 initially and also remain within @xmath82 at 5 gyr .
( by construction , this fraction approaches 1 at large @xmath82 . )
it shows that in the innermost several kiloparsecs of the merged halo , about 60% of the cusp particles ( solid curve ) that originated from within a given radius stay within that same radius , while very few of the core particles ( dashed curve ) remain .
we believe it is this `` lack of mobility '' of the cusp particles that gives the cuspy halo a stronger influence at the final remnant s centre and forces the final density profile to be cuspy rather than cored .
next we examine the relative contribution of cusp vs core particles to the total density ( or potential ) in the core - cusp collision . fig .
[ track2 ] shows the ratio of core to cusp particles initially ( dashed curve ) and at 5.0 gyr ( solid curve ) .
the figure shows that the cusp particles dominate the core particles number - wise in the centre initially ( if the two galaxies were superimposed with a common centre ) , and that at 5 gyr , the central region of the merger remnant becomes even more dominated by cusp particles .
this result implies that the tightly bound particles in a cusp tend to stay bound together , while a core is much easier to disrupt .
this complements the information from fig .
[ track1 ] by showing that for all particles of a given type , regardless of origin , the cuspy particles tend to take over the central region . moving outward ,
the core particles begin to contribute more to the remnant .
note that the radius where the number of core particles is equal to the number of cusp particles stays almost constant at @xmath83 kpc .
this is one indication that there is not a major reconfiguration of the particles occuring during the course of the simulation .
we have also calculated the quantities in figs .
[ track1 ] and [ track2 ] at 3.5 gyr and find them to be very similar to those found at 5 gyr .
this corroborates the evidence in fig .
[ en_evol ] that the merger is essentially completed by 3.5 gyr .
ratio of the number of core particles within radius r to number of cusp particles within radius r for the cusp - core collision at t=0 gyr ( dashed ) and t=5 gyr ( solid ) .
the remnant is even more centrally dominated by cusp particles than would be the case if the two initial haloes were superimposed , further showing how the core particles dominate the inner structure of the remnant . ]
the particle velocities in our initial haloes are isotropic by construction , but those in our final haloes all exhibit some degree of anisotropy . for the cusp - cusp collision , fig .
[ sigma ] shows the velocity dispersions ( top panels ) and velocity anisotropy ( bottom panels ) as a function of halo radius @xmath82 at @xmath84 , 3.5 , and 5 gyr for the simulation with @xmath2 particles ( run `` cusp3 '' in table 1 ) .
the velocities remain mostly isotropic ( i.e. @xmath85 ) near the centre of the merger remnant , but become mildly anisotropic ( @xmath86 ) with increasing radius due to radial infalls .
we have also examined the velocity distribution function for the same cusp - cusp run .
we find that in the central region of the merged haloes , the velocities are quite gaussian in addition to being isotropic ; fig .
[ fv ] shows @xmath87 vs. @xmath88 for @xmath89 for @xmath90 particles in a radial bin arbitrarily centred on 10.106 kpc .
the distribution function of @xmath91 ( left panel ) is well fit by a gaussian ( dashed line ) with width @xmath92 km s@xmath93 , computed from the particles in that bin .
the distributions of @xmath94 ( middle panel ) and @xmath95 ( right panel ) are also reasonably well fit by the corresponding gaussians : @xmath96 km s@xmath93 and @xmath97 km s@xmath93 . moving away from the centre , we find the radial distribution becomes substantially less gaussian and the angular distributions remain mostly gaussian in the intermediate portion of the halo . near the virial radius ,
however , none of the distributions are gaussian .
the core - core collision exhibits similar trends as in figs .
[ sigma ] and [ fv ] , so we will not show the results here . for the core - cusp collision , we plot in fig .
[ sigmam ] the radial velocity dispersion @xmath98 as a function of radius @xmath82 at @xmath99 , and 3.5 gyr for the core ( crosses ) and cusp ( diamonds ) particles separately .
( little evolution occurs after 3.5 gyr . )
the dispersion of the particles in the cuspy halo changes moderately between 0.0 and 3.5 gyr , with the peak in @xmath98 rising and moving outward but the overall shape maintained . for the particles in the cored halo ,
however , the velocity dispersion near the centre has almost doubled to @xmath100 km s@xmath93 after the merger . to understand why the velocity dispersion of the core particles increases toward the centre of the merged halo upon merging with a cuspy halo , it is instructive to consider a situation where the two haloes are spherically symmetric and share a common centre . in this case
, the potential @xmath101 is a function of the radial coordinate only and can be written as @xmath102 , a sum of the individual potentials @xmath103 and @xmath104 .
for particles in either the cored or cuspy halo , we then have @xmath105 where @xmath106 is the total mass interior to radius @xmath107 . from the point of view of the core particles , adding the cuspy particles has the effect of adding considerable extra mass and thus deepening the potential .
if the density profile of the core particles remains roughly constant , then adding extra mass in the form of cuspy particles essentially heats the core particles to higher a velocity dispersion , as seen in fig .
[ sigmam ] .
the reverse is also true : the core particles also heat the cusp particles .
this effect is actually seen in fig .
[ sigmam ] , as the cuspy particles dispersion increases relative to the original value , especially noticeably near the original peak of @xmath69 @xmath28 kpc .
since the cuspy particles dominate the total mass at radii less than this , however , the core particles are a small perturbation and do not change the cuspy particles dispersion significantly . the change in the peak of the nfw particles dispersion from @xmath69 to @xmath108 @xmath28 kpc
can also be understood in this manner , since the core particles actually contribute more of the interior mass at this radius , as shown in fig .
[ track2 ] .
the previous subsections discussed the density and velocity profiles separately .
it is also instructive to examine the local phase space density @xmath109 , which is shown in fig .
[ psd ] for the three types of mergers studied in this paper .
we find that for both the cusp - cusp ( middle panel ) and core - cusp ( right panel ) mergers , although the individual @xmath33 and @xmath110 are not a pure power - law in @xmath82 ( see figs .
[ densmulti ] , [ sigma ] , and [ sigmam ] ) , the ratio is interestingly well - approximated by a power law @xmath111 , where the best - fitting slope is @xmath112 for both mergers .
the core - core merger ( left panel ) , on the other hand , shows a core in the phase space density since both @xmath33 and @xmath110 are constant at @xmath113 .
our power - law results can be compared with @xcite , who have suggested that the phase space structure of haloes is intimately related to that of self - similar collapse onto spherical density perturbations in an expanding universe .
they solved the jeans equation under the assumption of isotropy and power - law @xmath114 and found that haloes with the nfw profile are well - approximated by the same power law , @xmath115 , as the self - similar solutions in secondary spherical infall models @xcite .
since the phase space density is inversely proportional to the local entropy , taylor & navarro were led to the conclusion that cdm halo structure may be driven toward the most sharply peaked phase space distribution allowable for a monotonically decreasing density profile . for a power law phase space density ,
this solution yields density profiles that are very similar to the nfw profile but with a slightly shallower inner cusp , @xmath116 rather than @xmath117 . using the power - law result @xmath118 in fig .
[ psd ] , we obtain from the jeans equation a density profile @xmath119 , which is consistent with the best - fitting inner slope shown in sec .
one can also look at the question in reverse , i.e. , given a power - law density @xmath120 , what solutions are admitted for the phase space density ?
calculations show that @xmath121 cusps lead to phase space densities well approximated by @xmath122 as @xmath123 , but the slope of @xmath124 is not strictly a constant beyond the central region ( singh & ma , in preparation ) .
our simulations offer insight into the mechanisms that drive a merging system toward an equilibrium structure , e.g. , violent relaxation @xcite and phase mixing .
violent relaxation is due to a time - varying potential which allows the non - conservation of individual particles energies .
the relative motion of the two haloes at the start of our simulations provides such a fluctuating potential . from fig .
[ en_evol ] , it is clear our mergers involve a rapid transfer of orbital energy to internal energy of the individual haloes ( and subsequently the merger remnants ) .
the decay of orbital motion occurs approximately on the crossing timescale , with essentially all energy in the form of the remnants internal energy by 2.5 gyr .
that the orbital motion is so heavily damped and that the remnant haloes do not expand or collapse significantly indicate that the time - varying potential necessary for violent relaxation is eliminated relatively quickly .
one possible exception is that density waves could enable violent relaxation to persist even once the haloes have merged .
phase mixing tends to damp these waves strongly , however , with a decay timescale of @xmath125 for a perturbation of size @xmath126 .
taking @xmath127 km s@xmath93 to be typical for our merged haloes , this means that phase mixing will damp out all perturbations smaller than the virial radius within @xmath128 gyr of the haloes merging . in our simulations , this seemingly short timescale for violent relaxation is sufficient to drive the velocity distribution function in the inner region of the merged halo to a gaussian form , as predicted by statistical reasoning ( e.g. @xcite ) . that the distribution function becomes less gaussian with increasing radius is also in line with expectations because much less mixing is possible in the outer regions of the halo on the timescales we are studying .
average energy change ( in arbitrary units ) versus impact parameter for the cusp - core collision .
squares are particles from the cuspy halo , triangles are particles from the cored halo .
all particles with small impact parameters tend to lose energy , while those with large impact parameters gain energy .
the effect is more dramatic for cuspy particles than cored ones . ]
following the energy distribution through the collision can help in understanding the structure of the final product .
a particularly interesting quantity is the change in energy @xmath129 for a particle from initial to final state as a function of position . @xcite
describe an energy segregation in their collision simulations : particles with higher ( less negative ) energy tend to gain more energy than particles with lower energy .
this means a particle near the centre of an initial halo should , on average , lose energy ( i.e. become more tightly bound ) relative to a particle near the virial radius of a halo .
we follow the energy change of each particle , and plot in fig .
[ impact ] the change in energy as a function of impact parameter for the core - cusp run `` mixed '' . as funato et al .
describe , the particles with small impact parameters tend to lose energy , while the particles with large impact parameters tend to gain it .
the effect is more dramatic for the cuspy particles than for the core particles , perhaps because the inner portion of the cuspy halo is initially more tightly bound but with an essentially equal velocity dispersion profile .
as noted by @xcite , however , this energy segregation does not prevent the system from equilibrating via violent relaxation , at least in the central region .
differential energy distribution for the cusp - core merger .
the solid lines show the cusp particles and the dashed lines show the core particles ; thin lines indicate the initial distributions and thick lines the final distributions .
both distributions are broadened as a result of the merger , and shell - like structures are visible in the final distributions .
the energy units are arbitrary . ]
another interesting quantity is the differential energy distribution @xmath130 , which can be written as a product of the phase - space distribution function @xmath131 and the density of states @xmath132 @xcite : @xmath133 the shape of @xmath130 is therefore _ not _ the same as the shape of @xmath131 . in fig .
[ ehist ] we plot @xmath130 for the core - cusp merger , for both types of particle , at the start of the simulation ( thin curves ) and at 5 gyr ( thick curves ) .
it shows that @xmath130 is broadened , a natural outcome of mergers where the range of energies accessible to an individual particle can change drastically .
the wave - like features in @xmath130 at 5 gyr are indications of shells of particles formed via phase - mixing @xcite .
we note that unlike @xmath131 , which generally decreases with increasing @xmath50 , the shape of @xmath130 is skewed towards higher energies because of the density of states .
both @xmath131 and @xmath130 have been studied analytically for isolated nfw @xcite and king haloes @xcite in equilibrium .
the overall shape of @xmath130 resembles that shown in fig .
[ ehist ] .
we have simulated major mergers of galaxy haloes with both cuspy and cored inner density profiles using gadget , a versatile n - body tree code .
our findings can be summarized as follows : 1 .
mergers of two cored haloes of equal mass result in a cored halo .
the core radius of the merged halo is nearly identical to those of the original haloes , but the core to virial radius ratio is reduced , with the outer density profile of the merger remnant well approximately by @xmath134 .
this result also holds for mergers with parent halo mass ratios of 3:1 , implying all major mergers of two cored haloes result in cored haloes .
mergers of a cuspy halo with either a cored halo or a second cuspy halo result in a cuspy halo .
if the cuspy halo starts out with an nfw form , the merged halo also has an nfw - like form , with a slightly shallower inner cusp of @xmath135 instead of @xmath117 .
3 . the particles from a given halo retain memory of their progenitor s density profile . in major mergers of a cored halo with a cuspy halo , the particles originating from the cuspy halo end up with a cuspy profile , and those from the cored halo end up with a cored profile .
( the cuspy particles dominate the total potential at the centre of the merged halo , so the resulting profile is cuspy . )
the core particles in the inner 10 @xmath28 kpc of the remnant are significantly heated .
the merger remnants relax from the inside outwards , attaining gaussian velocity distributions near their centres but not throughout the halo .
violent relaxation due to time - varying potential is effective only during the initial phase of the mergers ; phase mixing is likely the dominant relaxation process at late times .
5 . including a moderate amount of orbital angular momentum in the form of a non - zero impact parameter has little discernible effect on our findings .
the set of simulations presented in this paper has been performed for mergers of non - spinning haloes in a set bound orbit .
this clearly only probes a restricted range of parameter space .
a possible area for future investigations is to include non - zero initial spins , which can influence tidal interactions significantly but may only affect the final phase space structure weakly @xcite . we have also chosen not to convolute the issues in this paper by ignoring minor mergers with satellite haloes and continuous accretion of diffuse material .
the latter mechanisms may be responsible for generating the first generation of cuspy haloes from initially cored haloes , a process not achieved by major mergers of cored haloes according to our study . although the length and time scales quoted in this paper are for milky - way sized haloes of @xmath25 , there is nothing in the numerical code or in our analysis that is special to this particular mass scale ( except a mildly mass - dependent concentration parameter for nfw haloes ) .
our conclusions in this paper are thus likely to be applicable to major mergers of comparable - mass haloes over a wide mass range .
our finding that cuspy haloes are resilient to major mergers hints that the existence of cuspy haloes is natural in the hierarchical structure formation scenario : the build - up to larger mass haloes via repeated merging may be sufficient to produce cuspy haloes from cored haloes and preserve the general structure of cuspy haloes .
we thank jon arons , andrew benson , andrey kravtsov , and volker springel for useful discussions and jeff filippini for help with the energy analysis .
this research used resources of the national energy research scientific computing center , which is supported by the office of science of the u.s .
department of energy under contract no .
de - ac03 - 76sf00098 .
m is partially supported by an alfred p. sloan fellowship , a cottrell scholars award from the research corporation , and nasa grant nag5 - 12173 .
|
we present the results from a series of collisionless @xmath0-body simulations of major mergers of galaxy dark matter haloes with density profiles having either inner cusps or cores .
our simulations range from @xmath1 to @xmath2 particles , allowing us to probe the phase - space distribution of dark matter particles in the innermost regions ( less than 0.005 virial radii ) of cold dark matter haloes , a subject of much recent debate .
we find that a major merger of two cored haloes yields a cored halo and does not result in a cuspy profile seen in many cosmological simulations .
this result is unchanged if we consider mergers with parent mass ratios of 3:1 instead of 1:1 .
mergers of a cuspy halo with either a cored halo or a second cuspy halo of equal mass , on the other hand , produce cuspy haloes with a slightly reduced inner logarithmic slope .
cuspy haloes , once formed , therefore appear resilient to major mergers .
we find the velocity structure of the remnants to be mildly anisotropic , with a maxwellian velocity distribution near the centre but not in the outer portions of the final haloes .
violent relaxation is effective only during the early phase of mergers , with phase mixing likely to be the dominant relaxation process at late times .
[ firstpage ] galaxies : haloes dark matter methods : @xmath0-body simulations
|
[ sec : motivation ] the relationship between abstract interpretation @xcite and partial evaluation @xcite has received considerable attention ( see for example @xcite and their references ) . in order to motivate and illustrate our proposal for an integration of abstract interpretation and partial evaluation , we use the running example of fig . [ lst : running - ex ] .
it is a simple ciao program which uses peano s arithmetic .
we use the ciao assertion language in order to provide precise descriptions on the initial call patterns . in our case , the ` entry ` declaration is used to inform that all calls to the only exported predicate ( i.e. , ` main/2 ` ) will always be of the form @xmath0 with @xmath1 ground and @xmath2 a variable .
the predicate ` main/2 ` performs two calls to predicate ` formula/2 ` , which contains mode tests ` ground(x ) ` and ` var(w ) ` on its input arguments .
a call ` formula(x , w ) ` returns @xmath3 .
predicate ` two/1 ` returns the natural number 2 in peano s arithmetic .
a call ` minus(a , b , c ) ` returns @xmath4 .
however , if the result becomes a negative number , @xmath5 is left as a free variable .
this indicates that the result is not valid . in turn , a call ` twice(a , b ) ` returns @xmath6 .
prior to computing the result , this predicate checks whether @xmath7 is valid , i.e. , not a variable , and simply returns a variable otherwise . : - module(_,[main/1],[assertions ] ) .
+ : - entry main(s(s(s(l))),r ) : ( ground(l),var(r ) ) .
+ main@xmath8(x , x2):- formula@xmath9(x , x1 ) , formula@xmath10(x1,x2 ) .
+ formula@xmath11(x , w):- ground@xmath12(x),var@xmath13(w),two@xmath14(t),minus@xmath15(t , x , x2),twice@xmath16(x2,w ) . + two@xmath17(s(s(0 ) ) ) .
+ minus@xmath18(0,x , x ) .
+ minus@xmath19(s(x),s(y),r):- minus@xmath20(x , y , r ) .
+ minus@xmath21(s(_x),0,_r ) .
+ twice@xmath22(x,_y):- var@xmath23(x ) .
+ twice@xmath24(x , y):- ground@xmath25(x ) , tw@xmath26(x , y ) .
+ tw@xmath27(0,0 ) .
+ tw@xmath28(s(x),s(s(nx))):- tw@xmath29(x , nx ) . by observing the behaviour of the program
it can be seen that for initial queries satisfying the ` entry ` declaration , all calls to the tests ` ground_{2,1}(x ) ` and ` var_{2,2}(w ) ` will definitely succeed , even if we do not know the concrete values of variable @xmath1 at compile time .
also , the calls to ` ground_{8,1}(x ) ` will succeed , while the calls to ` var_{7,1}(x ) ` will fail .
this shows the benefits of ( 1 ) _ exploiting abstract information in order to abstractly execute certain atoms , which in turn may allow unfolding of other atoms_. however , the use of an abstract domain which captures groundness and freeness information will in general not be sufficient to determine that in the second execution of ` formula/2 ` the tests ` ground_{2,1}(x ) ` and ` var_{2,2}(w ) ` will also succeed .
the reason is that , on success of ` minus_{2,4}(t , x , x2 ) ` , @xmath30 can not be guaranteed to be ground since ` minus_6/3 ` succeeds with a free variable on its third argument position .
it can be observed , however , that for all calls to ` minus/3 ` in executions described by the ` entry ` declaration , such third clause for ` minus/3 ` is useless .
it will never contribute to a success of ` minus/3 ` since such predicate is always called with a value greater than zero on its second argument .
unfolding can make this explicit by fully unfolding calls to ` minus/3 ` since they are sufficiently instantiated ( and as a result the `` dangerous '' third clause is disregarded ) .
it allows concluding that in our particular context , all calls to ` minus/3 ` succeed with a ground third argument .
this shows the importance of ( 2 ) _ performing unfolding steps in order to prune away useless branches , which will result in improved success information_. by the time execution reaches ` twice_{2,5}(x2,w ) ` , we hopefully know that @xmath30 is ground . in order to determine that
, upon success of ` twice_{2,5}(x2,w ) ` ( and thus on success of ` formula_{1,1}(x , w ) ` ) , @xmath31 is ground , we need to perform a fixpoint computation .
since , for example , the success substitution for ` formula_{1,1}(x , x1 ) ` is indeed the call substitution for ` formula_{1,2}(x1,x2 ) ` , the success of the second test ` ground_{2,1}(x ) ` ( i.e. , the one reachable from ` formula_{1,2}(x1,x2 ) ` ) can not be established unless we propagate success substitutions .
this illustrates the importance of ( 3 ) _ propagating ( abstract ) success information , performing fixpoint computations when needed , which simultaneously will result in an improved unfolding_. finally , whenever we call ` formula(x , w ) ` , @xmath31 is a variable , a property which can not be captured if we restrict ourselves to downwards - closed domains .
this indicates ( 4 ) _ the usefulness of having information on non _ downwards - closed _
properties_. throughout the paper we show that the framework we propose is able to eliminate all calls to mode tests ` ground/1 ` and ` var/1 ` , and predicates ` two/1 ` and ` minus/3 ` are both fully unfolded and no longer appear in the residual code .
we have used _ sharing
freeness _ as abstract domain instead of one based on , say regular types , for two reasons . first , to illustrate how non - downwards closed information , including freeness and definite independence , can be correctly exploited by our algorithm in order to optimize the program , and second , to show how unfolding can be of great use in order to improve the accuracy of analyses apparently unrelated to partial deduction , such as the classical _ sharing
[ ex : sp ] the results obtained by ciaopp which implements abstract interpretation with specialized definitions are both the following specialized code and an accurate analysis for such program ( rules are renamed using the prefix @xmath32 ) .
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ sp_main@xmath33(s(s(s(0))),0 ) .
+ sp_main@xmath34(s(s(s(s(b)))),a ) : - sp_tw@xmath35(b , c ) , sp_formula@xmath36(c , a ) .
+ sp_tw@xmath34(0,0 ) .
+ sp_tw@xmath37(s(a),s(s(b ) ) ) : - sp_tw@xmath38(a , b ) .
+ sp_formula@xmath39(0,s(s(s(s(0 ) ) ) ) ) .
+ sp_formula@xmath40(s(a),s(s(s(s(s(s(b ) ) ) ) ) ) ) : - sp_tw@xmath41(a , b ) .
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ in this case , the success information for ` sp_main(x , x2 ) ` guarantees that @xmath30 is definitely ground on success .
note that this is equivalent to proving @xmath42 .
furthermore , our system is able to get to that conclusion even if the ` entry ` only informs about @xmath43 being any possible ground term and @xmath30 a free variable .
the above results can not be achieved unless all four points mentioned before are available in a program analysis / specialization system .
for example , if we use traditional partial deduction @xcite ( pd ) with the corresponding and rules followed by abstract interpretation and _ abstract specialization _ as described in @xcite we only obtain a comparable program after four iterations of the : `` pd + abstract interpretation + abstract specialization '' cycle .
if we keep on adding more calls to formula , every time more iterations are necessary to obtain results comparable to ours . this shows the importance of achieving an algorithm which is able to _ interleave _ pd , with abstract interpretation , extended with abstract specialization , in order to communicate the accuracy gains achieved from one to the other as soon as possible . in any case
, iterating over `` pd + analysis '' is not a good idea from the efficiency point of view .
also , sometimes partially evaluating a partially evaluated program can degrade the quality of the residual program .
the remaining of the paper is organized as follows .
section [ sec : prel - notat ] recalls some preliminary concepts . in sect .
[ sec : unfold - with - abstr ] , we present abstract unfolding which already integrates abstract executability .
section [ sec : spec - defin ] introduces our notion of specialized definition and embeds it within an abstract partial deducer . in sect .
[ sec : effic - prop - abstr ] , we propose our scheme for abstract interpretation with specialized definitions . finally , sect .
[ sec : disc - relat - work ] compares to related work and sect .
[ sec : benefits - or - our ] concludes .
[ sec : prel - notat ] very briefly ( see for example @xcite for details ) , an _ atom _
@xmath7 is a syntactic construction of the form @xmath44 , where @xmath45 , with @xmath46 , is a predicate symbol and @xmath47 are terms .
a _ clause _ is of the form @xmath48 where its head @xmath49 is an atom and its body @xmath50 is a conjunction of atoms .
a _ definite program _ is a finite set of clauses .
a _ goal _ ( or query ) is a conjunction of atoms .
let @xmath51 be a goal of the form @xmath52 , @xmath53 .
the concept of _ computation rule _ , denoted by @xmath54 , is used to select an atom within a goal for its evaluation .
the operational semantics of definite programs is based on derivations @xcite .
let @xmath55 be a renamed apart clause in @xmath56 such that @xmath57 .
then @xmath58 is _ derived _ from @xmath51 and @xmath5 via @xmath54 . as customary , given a program @xmath56 and a goal @xmath51 , an _ sld derivation _ for @xmath59 consists of a possibly infinite sequence @xmath60 of goals , a sequence @xmath61 of properly renamed apart clauses of @xmath56 , and a sequence @xmath62 of mgus such that each @xmath63 is derived from @xmath64 and @xmath65 using @xmath66 .
a derivation step can be non - deterministic when @xmath67 unifies with several clauses in @xmath56 , giving rise to several possible sld derivations for a given goal .
such sld derivations can be organized in _
sld trees_. a finite derivation @xmath68 is called _ successful _ if @xmath69 is empty . in that case @xmath70 is called the computed answer for goal @xmath51 . such a derivation
is called _ failed _ if it is not possible to perform a derivation step with @xmath69 .
given an atom @xmath7 , an _ unfolding rule _
@xcite computes a set of finite sld derivations @xmath71
( i.e. , a possibly incomplete sld tree ) of the form @xmath72 with computed answer substitution @xmath73 for @xmath74 whose associated _ resultants _ ( or residual rules ) are @xmath75 .
the following standard operations are used in the paper to handle keyed - tables : @xmath76 initializes a table @xmath77 .
@xmath78 adds @xmath79 associated to @xmath80 to @xmath77 and deletes previous information associated to @xmath80 , if any .
@xmath81 returns true iff @xmath80 is currently stored in the table . finally , @xmath81 returns the information associated to @xmath80 in @xmath77 . for simplicity ,
we sometimes consider tables as sets and we use the notation @xmath82 to denote that there is an entry in the table t with the corresponding @xmath80 and associated @xmath79 .
abstract interpretation @xcite provides a general formal framework for computing safe approximations of programs behaviour .
programs are interpreted using values in an _ abstract domain _ ( @xmath83 ) instead of the _ concrete domain _ ( @xmath84 ) .
the set of all possible abstract values which represents @xmath83 is usually a complete lattice or cpo which is ascending chain finite .
values in the abstract domain @xmath85 and sets of values in the concrete domain @xmath86 are related via a pair of monotonic mappings @xmath87 : the _ abstraction _ function @xmath88 which assigns to each ( possibly infinite ) set of concrete values an abstract value , and the _ concretization _ function @xmath89 which assigns to each abstract value the ( possibly infinite ) set of concrete values it represents . the operations on the abstract domain @xmath90 that we will use in our algorithms are : * @xmath91 performs the abstract restriction ( or projection ) of a substitution @xmath92 to the set of variables in the expression @xmath93 , denoted @xmath94 ; * @xmath91 extends the substitution @xmath92 to the variables in the set @xmath94 ; * @xmath95 obtains the description which results from adding the abstraction of the unification @xmath96 to the substitution @xmath92 ; * @xmath97 performs the abstract conjunction ( @xmath98 ) of two substitutions ; * @xmath97 performs the abstract disjunction ( @xmath99 ) of two substitutions .
in our algorithms we also use @xmath100 which adapts and projects the information in an abstract atom @xmath101 to the variables in the clause @xmath102 .
an _ abstract atom _ of the form @xmath103 is a concrete atom @xmath51 which comes equipped with an _ abstract substitution _
@xmath104 which is defined over @xmath105 and provides additional information on the context in which the atom will be executed at run - time . can be defined in terms of the operations above as : @xmath100 = @xmath106(@xmath107 .
finally , the most general substitution is represented as @xmath108 , and the least general ( empty ) substitution as @xmath109 .
we now present an extension of sld semantics which exploits abstract information . this will provide the means to overcome difficulties ( 1 ) and ( 2 ) introduced in section [ sec : motivation ] .
the extended semantics handles abstract goals of the form @xmath103 , i.e. , a concrete goal @xmath51 comes equipped with an abstract substitution @xmath104 defined over @xmath105 .
the first rule corresponds to a derivation step .
[ def : der - step - abs ] let @xmath103 be an abstract goal where @xmath110 .
let @xmath54 be a computation rule and let @xmath111@xmath67 .
let @xmath55 be a renamed apart clause in @xmath56 .
then the abstract goal @xmath112 is _ derived _ from @xmath103 and @xmath5 via @xmath54 if @xmath113 @xmath114 , where : @xmath115 an important difference between the above definition and the standard derivation step is that the use of abstract ( call ) substitutions allows imposing further conditions for performing derivation steps , in particular , @xmath116 can not be @xmath109 .
this is because if @xmath117 and @xmath118 then the head of the clause @xmath5 is incompatible with @xmath104 and the unification @xmath119 will definitely fail at run - time .
thus , abstract information allows us to remove useless clauses from the residual program .
this produces more efficient resultants and increases the accuracy of analysis for the residual code .
[ ex : derive ] consider the abstract atom @xmath120 , which appears in the analysis of our running example ( c.f . fig .
[ fig : aog ] ) .
we abbreviate as @xmath121 the successive application of @xmath122 functors @xmath123 to variable @xmath43 . the notation @xmath124 ( resp .
@xmath125 ) indicates that variable @xmath43 is ground ( resp . a free variable ) . after applying a derivation step ,
we obtain the derived abstract goal : + @xmath126 + where the abstract description has been extended with updated information about the freeness of the newly introduced variables .
in particular , both t and x2 are v. the second rule we present makes use of the availability of abstract substitutions to perform _ abstract executability _
@xcite during resolution .
this allows replacing some atoms with simpler ones , and , in particular , with the predefined atoms _ true _ and _ false _ , provided certain conditions hold .
we assume the existence of a predefined _ abstract executability table _ which contains entries of the form @xmath127 which specify the behaviour of external procedures : builtins , libraries , and other user modules .
for instance , for predicate @xmath128 contains the information @xmath129 . for @xmath130
, it contains @xmath131 .
[ def : abs - ex - true ] let @xmath103 be an abstract goal where @xmath110 .
let @xmath54 be a computation rule and let @xmath111@xmath67 .
let @xmath132 be a renamed apart entry in the abstract executability table .
then , the goal @xmath112 is _ abstractly executed _ from @xmath103 and @xmath132 via @xmath54 if @xmath133 and @xmath134 , where @xmath135 [ ex : aexec ] from the derived goal in ex .
[ ex : derive ] , we can apply twice the above rule to abstractly execute the calls to ground and var and obtain : @xmath136 since both calls succeed by using the abstract executability table described above and the information in the abstract substitution .
[ def : aunfold ] let @xmath137 be an abstract atom and @xmath56 a program .
we define @xmath138 as the set of _ resultants _ associated to a finite ( possibly incomplete ) sld tree computed by applying the rules of definitions [ def : der - step - abs ] and [ def : abs - ex - true ] to @xmath137 .
the so - called _ local control _ of pd ensures the termination of the above process . for this purpose
, the unfolding rule must incorporate some mechanism to stop the construction of sld derivations ( we refer to @xcite for details ) .
[ ex : aunfold ] consider an unfolding rule @xmath139 based on homeomorphic embedding @xcite to ensure termination and the initial goal in ex .
[ ex : derive ] .
the derivation continuing from ex .
[ ex : aexec ] performs several additional derivation steps and abstract executions and branches ( we do not include them due to space limitations and also because it is well understood ) .
the following resultants are obtained from the resulting tree : .... formula(s(s(s(s(0),s(s(s(s(0 ) ) ) ) ) .
formula(s(s(s(s(s(a))))),s(s(s(s(s(s(b ) ) ) ) ) ) ) : - tw(a , b ) .... which will later be filtered and renamed resulting in rules 4 and 5 of ex .
[ ex : sp ] .
it is important to note that sld resolution with abstract substitutions is not restricted to the left - to - right computation rule .
however , it is well - known that non - leftmost derivation steps can produce incorrect results if the goal contains _ impure _ atoms to the left of @xmath67 .
more details can be found , e.g. , in @xcite .
also , abstract execution of non - leftmost atoms can be incorrect if the abstract domain used captures properties which are not downwards closed .
a simple solution is to only allow leftmost abstract execution for non - downwards closed domains ( and non - leftmost for derivation steps ) .
[ sec : spec - defin ] [ 1 ] @xmath140 ; @xmath141 ( @xmath142 ) @xmath143 specialized_definition@xmath144 @xmath145 @xmath146 @xmath147 @xmath148 @xmath149 @xmath150 @xmath151 ( @xmath152 ) @xmath153 @xmath154 ( @xmath155 ) @xmath156 @xmath157 @xmath158@xmath159 @xmath160 @xmath161 @xmath162 @xmath163 @xmath164 @xmath165 we now define an abstract partial deduction ( apd ) algorithm whose execution can later be _ interleaved _ in a seamless way with a state - of - the - art abstract interpreter . for this
it is essential that the apd process can generate residual code _
online_. thus , we need to produce a residual , specialized definition for a call pattern as soon as we finish processing it .
this will make it possible for the analysis algorithm to have access to the improved definition .
this may increase the accuracy of the analyzer and addresses the difficulty ( 2 ) described in sect .
[ sec : motivation ] .
typically , pd is presented as an iterative process in which partial evaluations are computed for the new generated atoms until they _ cover _ all calls which can appear in the execution of the residual program .
this is formally known as the _ closedness _ condition of pd @xcite . in order to ensure termination of this global process
, the so - called _ global _ control defines a @xmath166 operator ( see @xcite ) which guarantees that the number of sld trees computed is kept finite , i.e. , it ensures the finiteness of the set of atoms for which partial evaluation is produced .
however , the residual program is not generated until such iterative process terminates .
algorithm [ algo : pd - with - spec - defs ] presents an apd algorithm .
the main difference with standard algorithms is that the resultants computed by ( l26 ) are added to the program during execution of the algorithm ( l30 ) rather than in a later code generation phase . in order to avoid conflicts among the new clauses and the original ones , clauses for specialized definitions
are renamed with a fresh predicate name ( l29 ) prior to adding them to the program ( l30 ) .
the algorithm uses two global data structures .
the _ specialization table _ contains entries of the form @xmath167 .
the atom @xmath168 provides the link with the clauses of the specialized definition for @xmath101 .
the _ generalization table _ stores the results of the function and contains entries @xmath169 where @xmath170 is a generalization of @xmath101 .
computation is initiated by procedure partial_evaluation_with_spec_defs ( l1 - 4 ) which initializes the tables and calls process_call_pattern for each abstract atom @xmath171 in the initial set to be partially evaluated .
the task of process_call_pattern is , if the atom has not been processed yet ( l6 ) , to compute a specialized definition for it ( l7 ) and then process all clauses in its specialized definition by means of calls to process_clause ( l9 - 11 ) .
procedure process_clause traverses clause bodies , processing their corresponding atoms by means of calls to process_call_pattern , in a depth - first , left - to - right fashion .
the order in which pending call patterns ( atoms ) are handled by the algorithm is usually not fixed in pd algorithms .
they are often all put together in a set .
the reason for this presentation is to be as close as possible to our analysis algorithm which enforces a depth - first , left - to - right traversal of program clauses . in this regard ,
the relevant point to note is that this algorithm does not perform success propagation yet ( difficulty 3 ) . in l16
, it becomes apparent that the atom(s ) in @xmath172 will be analyzed with the same call pattern @xmath104 as @xmath173 , which is to their left in the clause .
this , on one hand , may clearly lead to substantial precision loss .
for instance , the abstract pattern @xmath174 which is necessary to obtain the last two resultants of ex .
[ ex : sp ] can not be obtained with this algorithm . in particular
, we can not infer the groundness of @xmath175 which , in turn , prevents us from abstractly executing the next call to @xmath128 and , thus , from obtaining this optimal specialization . on the other hand , this lack of success propagation makes it difficult or even impossible to work with non downwards closed domains , since @xmath104 may contain information which holds before execution of the leftmost atom @xmath173 but which can be uncertain or even false after that .
in fact , in our example @xmath104 contains the info @xmath176 , which becomes false after execution of @xmath177 , since now @xmath175 is ground .
this problem is solved in the algorithm we present in the next section , where analysis information flows from left to right , adding more precise information and eliminating information which is no longer safe or even definitely wrong . for the integration
we propose , the most relevant part of the algorithm comprises l20 - 31 , as it is the code fragment which is _ directly _ executed from our abstract interpreter .
the remaining procedures ( l1-l19 ) will be overridden by more accurate ones later .
the procedure of interest is specialized_definition .
as it is customary , it performs ( l21 ) a generalization of the call @xmath101 using the abstract counterpart of the @xmath178 operator , denoted by @xmath166 , and which is in charge of ensuring termination at the global level .
the result of the generalization , @xmath170 , is inserted in the generalization table @xmath179 ( l22 ) .
correctness of the algorithm requires that @xmath180 .
if @xmath170 has been previously treated ( l23 ) , then its specialized definition @xmath181 is looked up in @xmath182 ( l24 ) and returned .
otherwise , a specialized definition @xmath183 is computed for it by using the @xmath139 operator of def .
[ def : aunfold ] ( l26 ) .
as already mentioned , the specialized definition @xmath183 for the abstract atom @xmath184 is used to extend the original program @xmath56 .
first , the atom @xmath168 is renamed by using @xmath185 which returns an atom with a fresh predicate name , @xmath181 , and optionally filters constants out ( l27 ) .
then , function @xmath186 is applied to rename the clause heads using atom @xmath168 ( l29 ) .
@xmath187 returns @xmath188 where @xmath189 .
finally , the program @xmath56 is extended with the new , _ renamed _
specialized definition , @xmath190 .
three calls to specialized_definition appear ( within an oval box ) during the analysis of our running example in fig .
[ fig : aog ] from the following abstract atoms , first @xmath191 , then @xmath192 and finally @xmath193 .
the output of such executions is used later ( with the proper renaming ) to produce the resultants in ex .
[ ex : sp ] .
for instance , the second clause obtained from the first call to specialized_definition is _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ sp_main@xmath34(s(s(s(s(b)))),a ) : - tw@xmath35(b , c),formula@xmath36(c , a ) . _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ where only the head is renamed .
the renaming of the body literals is done in a later code generation phase as already mentioned , alg .
[ algo : pd - with - spec - defs ] is not able to obtain such abstract atoms due to the absence of success propagation .
[ sec : effic - prop - abstr ] we now present our final algorithm for abstract interpretation with specialized definitions .
this algorithm extends both the apd algorithm [ algo : pd - with - spec - defs ] and the abstract interpretation algorithms in @xcite . w.r.t .
algorithm [ algo : pd - with - spec - defs ] , the main improvement is the addition of success propagation .
unfortunately , this requires computing a global fixpoint .
it is an important objective for us to be able to compute an accurate fixpoint in an efficient way .
w.r.t the algorithms in @xcite , the main improvements are the following .
( 1 ) it deals directly with non - normalized programs .
this point , which does not seem very relevant in a pure analysis system , becomes crucial when combined with a specialization system in order to profit from constants propagated by unfolding .
( 2 ) it incorporates a hardwired efficient graph traversal strategy which eliminates the need for maintaining priority queues explicitly @xcite .
( 3 ) the algorithm includes a widening operation for calls , , which limits the amount of multivariance in order to keep finite the number of call patterns analyzed .
this is required in order to be able to use abstract domains which are infinite , such as regular types .
( 4 ) it also includes a number of simplifications to facilitate understanding , such as the use of the keyed - table adt , which we assume encapsulates proper renaming apart of variables and the application of renaming transformations when needed . in order to compute and propagate success substitutions , algorithm [ algo : ai - with - spec - defs ] computes a _ program analysis graph _ in a similar fashion as state of the art analyzers such as the ciaopp analyzer @xcite .
for instance , the analysis graph computed by algorithm [ algo : ai - with - spec - defs ] for our running example is depicted in fig .
[ fig : aog ] .
the graph has two sorts of nodes .
those which correspond to atoms are called `` or - nodes '' .
for instance , the node @xmath194 indicates that when the atom @xmath195 is called with description @xmath196 the answer ( or success ) substitution computed is @xmath197 . those nodes which correspond to rules are called `` and - nodes '' . in fig .
[ fig : aog ] , they appear within a dashed box and contain the head of the corresponding clause . each and - node has as children as many or - nodes as literals there are in the body . if a child or - node is already in the tree , it is no further expanded and the currently available answer is used .
for instance , the analysis graph in figure [ fig : aog ] contains three occurrences of the abstract atom @xmath192 ( modulo renaming ) , but only one of them has been expanded .
this is depicted by arrows from the two non - expanded occurrences of @xmath198 to the expanded one .
more information on the efficient construction of the analysis graph can be found in @xcite . [ 1 ] @xmath199 ; @xmath200 @xmath140 ; @xmath141 ( @xmath201,j , entry}$ ] ) @xmath202 @xmath203 @xmath204 @xmath205 specialized_definition@xmath206 @xmath207 @xmath147 @xmath148 @xmath208 @xmath209~b_{k},k,1)$ ] @xmath210 @xmath211 ) @xmath212 @xmath213 @xmath214 @xmath148 process_call_pattern(@xmath215,k , i}$ ] ) @xmath216 @xmath148 @xmath217 @xmath218~r , k , i+1)$ ] @xmath219 @xmath220 process_call_pattern(@xmath221,k , i}$ ] ) @xmath216 @xmath148 @xmath217 @xmath222 @xmath148 @xmath223 @xmath224 @xmath225 @xmath226 @xmath227 @xmath228 @xmath229 @xmath230 ( @xmath231 ) @xmath232,k , i}$ ] @xmath233 ( @xmath234~b , k , i$ ] ) @xmath235~b , k , i)$ ] ( @xmath236 ) the program analysis graph is implicitly represented in the algorithm by means of two data structures , the _ answer table _ ( ) and the _ dependency table _ ( ) .
the answer table contains entries of the form @xmath237 which are interpreted as the answer ( success ) pattern for @xmath184 is @xmath238 .
for instance , there exists an entry of the form @xmath239 associated to the atom discussed above .
dependencies indicate direct relations among or - nodes .
an or - node @xmath240 _ depends on _ another or - node @xmath241 iff in the body of some clause for @xmath240 there appears the or - node @xmath241 .
the intuition is that in computing the answer for @xmath240 we have used the answer pattern for @xmath241 . in our algorithm
we store _ backwards _ dependencies , i.e. , for each or - node @xmath241 we keep track of the set of or - nodes which depend on it .
that is to say , the keys in the dependency table are or - nodes and the information associated to each node is the set of other nodes which depend on it , together with some additional information required to iterate when an answer is modified ( updated ) .
each element of a _
dependency set _ for an atom @xmath242 is of the form @xmath243 ~k , i}$ ] .
it should be interpreted as follows : the or - node @xmath244 through the literal at position @xmath245 depends on the or - node @xmath242 .
also , the remaining information @xmath246 $ ] informs that the head of this clause is @xmath247 and the substitution ( in terms of all variables of clause @xmath248 ) just before the call to @xmath242 is @xmath249 .
such information avoids reprocessing atoms in the clause @xmath248 to the left of position @xmath250 .
for instance , the dependency set for @xmath251 is @xmath252 2,2}\}$ ] .
it indicates that the or - node @xmath253 is only used , via literal ( 2,2 ) , in the or - node @xmath191 ( see example [ ex : sp ] ) .
thus , if the answer pattern for @xmath251 is ever updated , then we must reprocess the or - node @xmath254 from position 2,2 .
algorithm [ algo : ai - with - spec - defs ] proceeds as follows .
the procedure abs_int_with_spec_defs initializes the four tables used by the algorithm and calls process_call_pattern for each abstract atom in the initial set .
the function process_call_pattern applies , first of all ( l7 ) , the function to @xmath101 taking into account the set of entries already in .
this returns a substitution @xmath249 s.t .
the most precise function possible is the identity function , but it can only be used with abstract domains with a finite number of abstract values .
this is the case with _ sharing
and thus we will use the identity function in our example .
if the call pattern @xmath256 has not been processed before , it places ( l9 ) @xmath109 as initial answer in for @xmath101 and sets to empty ( l10 ) the set of or - nodes in the graph which depend on @xmath256 .
it then computes ( l11 ) a specialized definition for @xmath256 .
we do not show in algorithm [ algo : ai - with - spec - defs ] the definition of specialized_definition , since it is identical to that in algorithm [ algo : pd - with - spec - defs ] . in the graph
, we show within an oval box the calls to specialized_definition which appear during the execution of the running example ( see the details in sect .
[ sec : spec - defin ] ) .
the clauses in the specialized definition are linked to the box with a dotted arc .
then it launches ( l13 - 15 ) calls to process_clause for the clauses in the specialized definition w.r.t .
which @xmath256 is to be analyzed . only after this , the _ parent _ or - node is added ( l16 - 17 ) to the dependency set for @xmath256 .
the function process_clause performs the success propagation and constitutes the core of the analysis .
first , the current answer ( @xmath257 ) for the call to the literal at position @xmath245 of the form @xmath258 is ( l24 and l29 ) conjoined ( ) , after being extended ( ) to all variables in the clause , with the description @xmath259 from the program point immediately before @xmath50 in order to obtain the description @xmath260 for the program point after @xmath50 .
if @xmath50 is not the last literal , @xmath216 is taken as the ( improved ) calling pattern to process the next literal in the clause in the recursive call ( l25 ) .
this corresponds to left - to - right success propagation and is marked in fig .
[ fig : aog ] with a dashed horizontal arrow .
if we are actually processing the last literal , @xmath260 is ( l30 ) adapted ( ) to the initial call pattern @xmath261 which started process_clause , obtaining @xmath262 .
this value is ( l32 ) disjoined ( ) with the current answer , @xmath263 , for @xmath261 as given by . if the answer changes , then its dependencies , which are readily available in , need to be recomputed ( l36 ) using process_update .
this procedure restarts the processing of all body postfixes which depend on the calling pattern for which the answer has been updated by launching new calls to process_clause .
there is no need of recomputing answers in our example .
the procedure remove_previous_deps eliminates ( l42 ) entries in for the clause postfix which is about to be re - computed .
we do not present its definition here due to lack of space .
note that the new calls to process_clause may in turn launch calls to process_update . on termination of the algorithm a global fixpoint
is guaranteed to have been reached .
note that our algorithm also stores in the dependency sets calls from the initial entry points ( marked with the value @xmath264 in l5 ) .
these do not need to be reprocessed ( l40 ) but are useful for determining the specialized version to use for the initial queries after code generation .
termination of algorithm [ algo : ai - with - spec - defs ] comprises several levels .
first , termination of the algorithm requires the local termination of the process of obtaining a specialized definition .
this corresponds to ensuring termination of function specialized_definition in algorithm [ algo : pd - with - spec - defs ] .
second , we need to guarantee that the number of call patterns for which a specialized definition is computed is finite .
this corresponds to global termination of specialization algorithms . in terms of our algorithm
, this is equivalent to having a finite number of entries in
. the function should be able to guarantee it .
third , it is required that the set of call patterns for which an answer pattern is to be computed be finite .
this corresponds to control of multivariance in context - sensitive analysis . in terms of our algorithm
, this is equivalent to having a finite number of entries in
. the function should be able to guarantee it .
fourth and final , it is required that the computation of the answer pattern for each entry in needs a finite number of iterations .
this is guaranteed since we consider domains which are ascending chain finite .
another way of looking at this problem is that , intuitively , the combined effect of terminating and operators guarantee that the set of specialized definitions which algorithm [ algo : ai - with - spec - defs ] will compute for an initial set of atoms is finite .
these two problems have received considerable attention by the pd community ( see , e.g. , @xcite ) . since algorithm [ algo : ai - with - spec - defs ] performs analysis of the program composed of the set of specialized definitions , once we have guaranteed the finiteness of the program to be analyzed , a terminating together with an abstract domain which is ascending chain finite guarantee termination of the whole process .
[ sec : disc - relat - work ] we have presented a generic framework for the analysis and specialization of logic programs which is currently the basis of the analysis / specialization system implemented in the ciaopp preprocessor .
we argue that , in contrast to other approaches , the fact that our method can be used both as a specializer and analyzer gives us more accuracy and efficiency than the individual techniques .
indeed , the versatility of our framework ( and of our implementation ) can be seen by recasting well - known specialization and analysis frameworks as instances in which the different parameters : unfolding rule , widen call rule , abstraction operator , and analysis domain , take the following values .
[ [ polyvariant - abstract - interpretation ] ] polyvariant abstract interpretation : + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + our algorithm can behave as the analysis algorithm described in @xcite for polyvariant static analysis by defining a @xmath166 operator which returns always the base form of an expression ( i.e. , it loses all constants ) and an @xmath139 operator which performs a single derivation step ( i.e. , it returns the original definition ) .
thus , the resulting framework would always produce a residual program which coincides with the original one and can be analyzed with any abstract domain of interest .
[ [ multivariant - abstract - specialization ] ] multivariant abstract specialization : + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + the specialization power of the framework described in @xcite can be obtained by using the same @xmath166 described in the above point plus an @xmath139 operator which always performs a derive step followed by zero or more abstract execution steps .
it is interesting to note that in the original framework abstract executability is performed as an offline optimization phase while it is performed online in our framework . [
[ classical - partial - deduction ] ] classical partial deduction : + + + + + + + + + + + + + + + + + + + + + + + + + + + + our method can be used to perform classical pd in the style of @xcite by using an abstract domain with the single abstract value @xmath108 and the identity function as rule .
this corresponds to the @xmath265 domain of @xcite in which an atom with variables represents all its instances .
let us note that , in spite of the fact that the algorithm follows a left - to - right computation flow , the process of generating specialized definitions ( as discussed in section [ sec : unfold - with - abstr ] ) can perform _ non - leftmost _ unfolding steps and achieve optimizations as powerful as in pd .
[ [ abstract - partial - deduction ] ] abstract partial deduction : + + + + + + + + + + + + + + + + + + + + + + + + + + + several approaches have been proposed which extend pd by using abstract substitutions @xcite . in essence ,
such approaches are very similar to the abstract partial deduction with call propagation shown in algorithm [ algo : pd - with - spec - defs ] .
though all those proposals identify the need of propagating success substitutions , they either fail to do so or propose means for propagating success information which are not fully integrated with the apd algorithm and , in our opinion , do not fit in as nicely as the use of and or trees . also , these proposals are either strongly coupled to a particular ( downward closed ) abstract domain , i.e. , regular types , as in @xcite or do not provide the exact description of operations on the abstract domain which are needed by the framework , other than general correctness criteria @xcite .
however , the latter allow conjunctive pd , which is not available in our framework . [ [ the - approach - in ] ] the approach in @xcite : + + + + + + + + + + + + + + + + + + + + + + + + was a starting step towards our current framework .
there , the introduction of unfolding steps directly in the and or graph was proposed in order to achieve transformations as powerful as those of pd while at the same time propagating abstract information .
in contrast , we now resort to augmented sld semantics for the specialization side of the framework while using and - or semantics for the analysis side of the framework .
this has both conceptual , the hybrid approach we propose provides satisfactory answers to the four issues raised in section [ sec : motivation ] , and practical advantages , since the important body of work in control of pd is directly applicable to the specialization side of our framework .
[ sec : benefits - or - our ] we have proposed a novel scheme for a seamless integration of the techniques of abstract interpretation and partial deduction .
our scheme is parametric w.r.t . the abstract domain and the control issues which guide the partial deduction process .
existing proposals for the integration use abstract interpretation as a _ means _ for improving partial evaluation rather than as a _ goal _
, at the same level as producing a specialized program .
this implies that , as a result , their objective is to yield a set of atoms which determines a partial evaluation rather than to compute a safe approximation of its success . unlike them ,
a main objective of our work is to improve success information by analyzing the specialized code , rather than the original one .
we achieve this objective by smoothly _ interleaving _ both techniques which improves success information even for abstract domains which are not related directly to partial evaluation .
moreover , with more accurate success information , we can improve further the quality of partial evaluation . the overall method thus yields not only a specialized program but also a safe approximation of its behaviour .
the authors would like to thank john gallagher and michael leuschel for useful discussions on the integration of abstract interpretation and partial deduction .
this work was funded in part by the information society technologies programme of the european commission , future and emerging technologies under the ist-2001 - 38059 _ asap _ project and by the spanish ministry of science and education under the mcyt tic 2002 - 0055 _ cubico _ project .
manuel hermenegildo is also supported by the prince of asturias chair in information science and technology at unm.2 g. puebla , m. hermenegildo , and j. gallagher .
n integration of partial evaluation in a generic abstract interpretation framework . in _ proc . of pepm99
_ , number ns-99 - 1 in brisc series , pages 7585 .
university of aarhus , denmark , 1999 .
|
the relationship between abstract interpretation and partial deduction has received considerable attention and ( partial ) integrations have been proposed starting from both the partial deduction and abstract interpretation perspectives . in this work
we present what we argue is the first fully described generic algorithm for efficient and precise integration of abstract interpretation and partial deduction . taking as starting point state - of - the - art algorithms for context - sensitive , polyvariant abstract interpretation and ( abstract ) partial deduction , we present an algorithm which combines the best of both worlds .
key ingredients include the accurate success propagation inherent to abstract interpretation and the powerful program transformations achievable by partial deduction . in our algorithm ,
the calls which appear in the analysis graph are not analyzed w.r.t . the original definition of the procedure but w.r.t .
_ specialized definitions _ of these procedures .
such specialized definitions are obtained by applying both unfolding and abstract executability .
our framework is parametric w.r.t .
different control strategies and abstract domains . different combinations of such parameters correspond to existing algorithms for program analysis and specialization .
simultaneously , our approach opens the door to the efficient computation of strictly more precise results than those achievable by each of the individual techniques .
the algorithm is now one of the key components of the ciaopp analysis and specialization system .
|
this paper shows that local values of internal energy and enthalpy within a moist atmosphere can be better computed so as to study the energy and enthalpy directly for global / local domains .
difficulties encountered in the past were due to a paradox : the budget equation of temperature was easier to compute than the local values of enthalpy , whereas it was easier to compute the local values of entropy than to solve the budget equation .
one example of this paradox is given by the conservation or imbalance properties of internal energy @xmath2 ( see appendix a ) or for the enthalpy @xmath3 .
it is important to evaluate observed changes in radiative forcing and imbalance properties for energy fluxes , in order i ) to properly assess observed climate change impacts and ii ) to improve our ability to understand energy changes in the climate system at regional scales . in studying the conservation of energy in nwp models and gcms , it is common to assess global energy fluxes from the surface to the top of the atmosphere using re - analysis products .
conservation and/or imbalance properties are thus currently monitored via the computation of fluxes of energy at the surface and the top of the atmosphere rather than by directly computing regional or global integrals of energy throughout the atmosphere . ]
another example of this paradox appears when the enthalpy equation is used in nwp models or in gcms .
the enthalpy equation is never written in terms of @xmath4 , with the specific enthalpy @xmath5 to be evaluated at each grid - point .
it is recalled on the left side of fig.([fig_barycentric ] ) that the enthalpy equation can be written in terms of @xmath6 or @xmath7 , where @xmath8 is the temperature and @xmath9 the moist - air specific heat .
these properties are consequences of material derivatives applied to barycentric motions of open parcels of fluid described by de groot and mazur ( 1962 ) or glansdorff et prigogine ( 1971 ) and applied by marquet ( 1993 ) , zdunkowski and bott ( 2004 ) and catry _ et al . _
( 2007 ) , among others .
it is recalled below that most of the present moist - air formulae for @xmath5 and @xmath10 were derived using the hypothesis of closed parcels of fluid .
the problem concerning the knowledge of reference values for @xmath5 or @xmath10 can be briefly illustrated by starting with the dry air enthalpy written as @xmath11 $ ] .
this is a valid expression provided that the specific heat @xmath12 is assumed to be a constant for @xmath8 close to @xmath13 and @xmath14 is associated with @xmath13 .
the enthalpy of a clear - air parcel of moist air is computed as the sum of @xmath15 , where @xmath16 is computed with @xmath12 and @xmath14 replaced by @xmath17 and @xmath18 .
since the term of @xmath19 in brackets is multiplied by @xmath20 , it can be neglected either if @xmath21 is a constant or if the bracketed term is equal to zero , i.e. if @xmath22 .
but , in fact , this is not valid .
it is only if @xmath21 is a constant that changes in enthalpy or entropy can be easily computed , with no need to determine reference values for enthalpies and entropies .
the above properties are only valid if the parcels of fluid are closed .
there are , however , circumstances in which it is necessary to compare absolute values of enthalpy or entropy .
this is true particularly when open systems are considered and if parcels of fluid are not connected by barycentric motion .
this is illustrated on the right side of fig.([fig_barycentric ] ) .
it is not possible to transform @xmath23 into @xmath24 for these independent parcels since the reference values of @xmath5 and @xmath10 are multiplied by varying terms , all of which depend on @xmath21 . in consequence ,
spatial integrals , vertical fluxes and 3d gradients must be computed by using absolute definitions for @xmath5 and @xmath10 .
accordingly , richardson ( 1922 ) and businger ( 1982 ) , hereafter referred to as r22 and b82 , searched for relevant choices for zero entropy and enthalpy for both dry air and water species . a way to compute the specific moist - air enthalpy
can be found by examining the other aspect of the paradox , concerning the entropy .
it is shown in the following sections that it is impossible to compute all the terms forming the entropy budget equation .
nevertheless , it is possible to use the third law of thermodynamics to compute the moist air entropy in an absolute way , as explained in hauf and hller ( 1987 ) and marquet ( 2011 ) .
this is done by setting the zero of entropies for the most stable crystalline form of all atmospheric species at @xmath0 k. this paper explores new possibilities for computing and analyzing isenthalpic or isentropic processes in open systems , defined by @xmath25constant or @xmath26constant .
this is equivalent to setting @xmath27 and @xmath28 ( using barycentric formalism for the enthalpy and entropy equations ) .
direct local definitions for @xmath5 and @xmath10 may allow the study of new classes of processes associated with open systems .
they include the impact of diffusion of water on @xmath5 or @xmath10 ( entrainment and detrainment at cloud edges ) , and/or the impact of evaporation of water from the surface .
for instance , it is reported in m11 that moist - air entropy is constant within the whole boundary layer of marine stratocumulus , with almost no jump in @xmath10 at the top of the boundary layer .
these properties could only be observed with the open - system formulation derived in m11 , since large vertical gradients and jumps are observed for @xmath21 , @xmath29 and @xmath30 with marine stratocumulus .
other examples of new applications can be found in marquet and geleyn ( 2013 ) and marquet ( 2013 ) , where the moist - air brunt - visl frequency and potential vorticity are defined in terms of absolute values of the specific moist - air entropy derived in m11 .
these moist - air quantities are computed with the help of vertical and 3d gradients of @xmath10 , and thus depend on relevant absolute definitions for the reference entropies of dry air and water vapour .
it is worth noting that the search for absolute and local definitions for @xmath5 and @xmath10 does not correspond to new definitions for the well - known potential temperatures @xmath31 , @xmath30 or @xmath29 .
it is shown in following sections that the absolute formulation of m11 must be understood as being a generalization to open systems of the two specific entropies derived in pauluis _
( 2010 , hereafter referred to as p10 ) , which correspond to the ( pseudo- ) adiabatic potential temperatures @xmath30 or @xmath29 .
the important issue in this paper is to determine which kind of enthalpy might be defined in an absolute way .
it is assumed that this is possible for the `` thermal '' enthalpies , which correspond to those enthalpies for n@xmath32 , o@xmath32 and h@xmath32o substances that are generated by the variations of @xmath33 generated by progressive excitations of the translational , rotational and vibrational states of the molecules , and by the possible changes of phase represented by the latent heats ( with negligible impact of changes of pressure ) . both entropy and thermal enthalpy are thermodynamic state functions .
therefore , both of them can be computed following any possible reversible path that connects the dead state at @xmath0 k and the actual atmospheric states .
the third law of thermodynamics can be used to set the entropy to zero for the dead states corresponding to the most stable cryogenic substances at @xmath0 k. it is assumed that the same hypothesis can be applied for thermal enthalpies at @xmath0 k provided that the fluid is free of chemical or nuclear reactions . the paper is organized as follows
. section [ subsection_dhdt_dtdt ] explains in some detail why the enthalpy equation can be transformed into the temperature equation , with no impact of the reference values for enthalpies .
the importance of these reference values for computing the turbulent enthalpy fluxes is then illustrated in section [ subsection_enthalpy_fluxes ] using results from b82 .
some impacts of reference values on the computations of moist - air entropy are described in section [ subsection_specific_entropy ] , where the essential results of m11 are recalled .
comparisons between the moist - air entropy derived in m11 and those described in p10 are presented in section [ subsection_entropy_p10 ] .
following r22 , it is explained , in section [ subsection_richardson ] , why it is possible to apply nernst s result ( namely the third law of thermodynamics ) to compute the water entropies of moist air . in section [ subsection_unit_mass_dry_air ] , it is shown that the method of computing moist - air enthalpy or entropy `` per unit mass of dry air '' does not obviate the need to determine the reference values of moist - air species in the case of open systems .
a review of different moist static energies is briefly presented in section [ subsection_mse ] in order to list these well - known quantities , which are often considered to be similar to the sum of the moist - air enthalpy and the potential energy .
section [ section_data_method ] corresponds to the data and method part of the paper .
the general formula for the specific moist - air thermal enthalpy is derived in section [ subsection_enthalpy ] .
computations of the standard and reference values are then presented in section [ subsection_std_h ] ( details are given in appendix b ) .
numerical applications are considered in section [ section_results ] .
an analysis of the validity of the approximation @xmath34 is shown in section [ subsection_trouton ] and provides new insights into the questions raised in b82 .
comparisons of the thermal enthalpy with different moist static energies are presented in section [ subsection_comp_h_mse ] .
the properties of the moist - air thermal enthalpy diagrams are analysed in section [ subsection_s_h_diagrams ] .
numerical values of the thermal enthalpy are computed and plotted in section [ subsection_h_tmse_profiles ] for several stratocumulus and cumulus case studies ( see appendix c ) .
the link between the thermal enthalpy diagram and the computation of the wet - bulb temperature are analysed in section [ subsection_tw_psychro ] , where several psychrometric equations are used and compared . in section [ section_conclusion ] ,
other applications are suggested for the various findings presented in this paper .
the enthalpy and temperature equations are described in some detail because reference will be made to them in later sections .
the enthalpy equation can be written as @xmath35 equation ( [ def_barycentric_1 ] ) is the development of the material derivative of the implicit weighted sum @xmath36 , where the index @xmath37 applies to for dry air ( @xmath38 ) , water vapour ( @xmath39 ) , liquid water ( @xmath40 ) and ice ( @xmath41 ) .
it is assumed that it is possible to write @xmath42 for each species , where @xmath43 and @xmath44 are constant terms attributed to each gas .
equation ( [ def_barycentric_2 ] ) @xmath45 is obtained from the properties @xmath46 ( no impact of @xmath44 ) , @xmath47 and @xmath48 .
external changes @xmath49 are created by the divergence of the diffusion fluxes @xmath50 , where @xmath51 represent departures from the barycentric mean @xmath52 .
internal changes @xmath53 are created by phase changes of water . the first law equation ( [ def_barycentric_3bis ] )
is expressed in terms of barycentric motion of an open parcel of fluid . @xmath54
the term involving change in pressure represents the expansion work .
the divergence of the enthalpy flux @xmath55 is separated into two parts in ( [ def_barycentric_3 ] ) , with @xmath56 equal to the external changes @xmath57 .
the heating rate @xmath58 corresponds to other diabatic processes ( mainly radiation ) .
the temperature equation @xmath59 is then obtained by subtracting ( [ def_barycentric_3 ] ) from ( [ def_barycentric_2 ] ) and using the result @xmath60 ( no impact of @xmath44 ) , which leads to ( [ def_barycentric_4 ] ) .
@xmath61 the terms depending on external changes cancel out in the difference ( [ def_barycentric_2 ] ) minus ( [ def_barycentric_3 ] ) .
the implicit sum depending on the diffusion fluxes @xmath50 acts in regions where gradients of temperature exist .
the implicit sum depending on internal changes represents the impact of phase changes of water .
it can be written as @xmath62 , in terms of the vaporization , sublimation and fusion rates . therefore all the terms in ( [ def_barycentric_4 ] ) can be easily computed and there is no need to determine the absolute values for enthalpies to express the first law in terms of the temperature equation . it is also possible to compute @xmath63 , since it is equal to @xmath64 and since the sum @xmath65 is known .
these results correspond to the first part of the paradox . the problem of finding relevant values for the reference enthalpies @xmath66 is explicitly addressed in b82 , where the specific enthalpies are written as @xmath42 ( as in the previous section ) , corresponding to the definitions @xmath67 and @xmath68 .
it is a usual practice in atmospheric science to set reference enthalpies of dry air equal to zero for a given reference temperature , typically for @xmath69c .
the choices of zero - enthalpy for water species are more variable .
it is set to zero either for the water vapour or for the liquid water enthalpies , for the same reference temperature @xmath70c and with the latent heats obviously connecting the other water - enthalpies through @xmath71 or @xmath72 .
it is shown in b82 that the choices of zero - enthalpies for both dry air and liquid water are is in agreement with well - established procedures for computing surface turbulent fluxes .
otherwise , additional fluxes of @xmath21 would have to be added , leading to other definitions of the moist - air enthalpy flux . the same hypothesis @xmath73 is used in the review by fuehrer and friehe ( 2002 ) and for @xmath69c .
it is , however , unlikely that such arbitrary fluxes of @xmath21 may be added to or subtracted from the enthalpy flux , leading to arbitrary closure for the computation of turbulent fluxes of moist air .
the same is true for the vertical integral and the horizontal or vertical gradients of @xmath5 , for which terms depending on @xmath21 could be of real importance if @xmath21 is not constant .
accordingly , one of the aims of the coming sections will be to better justify these assumptions concerning the reference values of partial enthalpies , and possibly to define them more accurately for any value of the reference temperature @xmath13 .
another motivation for the search for accurate definitions of moist - air enthalpy ( or entropy ) is to try to answer the following question illustrated on the right side of fig.([fig_barycentric ] ) : is it possible to determine whether two parcels of fluid have the same local values of @xmath5 or @xmath10 ?
it is easy to answer to this question for the motion of closed parcels of moist - air , for which @xmath21 is a constant and the @xmath44 may be cancelled , or for pseudo - adiabatic motion , for which it is possible to rely on the differential equations given by saunders ( 1957 ) or iribarne and godson ( 1973 ) .
these differential equations are established without the need for precise definitions of partial entropies .
these adiabatic or pseudo - adiabatic processes correspond to the definitions of the well - known liquid - water , equivalent and wet - bulb potential temperatures @xmath29 , @xmath30 and @xmath31 , respectively .
accordingly , the search for absolute and local definitions for @xmath5 and @xmath10 does not correspond to a change in the definitions of @xmath29 , @xmath30 or @xmath31 and the corresponding real temperatures . a new way to answer these questions and to compute the specific moist - air `` thermal enthalpy '' will be found by examining the other aspect of the paradox , concerning the entropy , with the term `` thermal '' explained in the following sections .
the other aspect of the paradox concerns the entropy : it is easier to compute the local values of entropy than to manage its budget equation . like the enthalpy equation ( [ def_barycentric_3 ] ) ,
the entropy equation derived by de groot and mazur ( 1962 ) is suitable for open systems . as in section [ subsection_enthalpy_fluxes ] , the gibbs equation ( [ def_gibbs_1 ] ) is expressed for barycentric motion with diffusion fluxes @xmath50 and with the local gibbs function defined by @xmath74 .
@xmath75 the first law expressed by ( [ def_barycentric_3 ] ) can be introduced in ( [ def_gibbs_1 ] ) , leading to the entropy equation ( [ def_gibbs_2 ] ) .
a term @xmath76 should appear on the right hand side of ( [ def_gibbs_2 ] ) but it is equal to zero if it is assumed that changes of phases of water species are reversible and occur at equal gibbs potential @xmath77 between the two phases .
this excludes the possibility of supercooled water for instance . the first reason why it is difficult to evaluate the entropy equation accurately is due to another term , @xmath78 , which should be added into the right hand side of ( [ def_gibbs_2 ] ) in order to take account of other sources of irreversibility that exist in the real atmosphere .
an entropy equation similar to ( [ def_gibbs_2 ] ) is derived in p10 for the simplified case of moist air composed of dry air , water vapour and liquid water ( no ice ) .
this equation can be rewritten as @xmath79 where the implicit sum @xmath80 in ( [ def_gibbs_2 ] ) is expressed in ( [ def_p10_dsdt ] ) in terms of the rate of change of @xmath81 and @xmath82 due to diffusion of water vapour and precipitation processes , denoted @xmath83 and @xmath84 , respectively .
the impact of the diffusion term depending on @xmath85 is not taken into account in ( [ def_p10_dsdt ] ) .
another reason why it is difficult to evaluate the entropy equations ( [ def_gibbs_2 ] ) or ( [ def_p10_dsdt ] ) accurately is due to the barycentric view , which corresponds to a unit mass of moist air .
the consequence is that external input or output of @xmath81 or @xmath82 must be balanced at the expense of opposite changes in the moist - air value @xmath10 .
this explains the two terms that multiply @xmath83 and @xmath84 in ( [ def_p10_dsdt ] ) , while @xmath10 is approximated by @xmath86 in p10 .
it is thus important to determine the absolute value of @xmath10 in order to compute , from ( [ def_p10_dsdt ] ) , the impact of diffusion of water ( entrainment and detrainment at the edges of clouds ) , or the evaporation of water from the surface .
the same is true if the aim is to determine the impact of these processes on @xmath5 starting from the enthalpy equation ( [ def_barycentric_3 ] ) .
it is possible to by - pass these difficulties .
the aim is to compute @xmath10 directly , without trying to compute @xmath87 via the associated source - sink terms .
following hh87 , it is explained in m11 that it is indeed possible to compute the specific moist - air entropy by determining the reference values from the third law , which states that the entropy of any substance is equal to zero for the most stable crystalline form of the substance and at absolute zero temperature ( the third law can not be applied to liquids or gases ) .
a third - law - based formulation for the specific moist - air entropy @xmath10 is written in m11 as @xmath88 both @xmath89 and @xmath12 are constant terms and the entropy potential temperature @xmath90 depends on the adimensional parameter @xmath91 \:/ c_{pd}$ ] .
this parameter must be evaluated from the reference values of the partial entropies of dry - air and water - vapour components .
it is important to notice that the two notions of `` reference '' and `` standard '' entropies correspond to different numerical values .
the reference values @xmath92 and @xmath93 are associated with the reference temperature @xmath13 , the total pressure @xmath94 , the liquid - water saturation pressure @xmath95 and the water - vapour mixing ratio @xmath96 .
these reference values are determined according to the standard values @xmath97 and @xmath98 which correspond to the same standard values for temperature and pressure for all species , with for instance @xmath99 c ( or @xmath100 c ) and @xmath101 hpa , even for water vapour .
the third - law standard values of specific entropies are available in thermodynamic tables for all atmospheric species .
the values recalled in appendix a and computed with the datasets of appendix b are the same as those used in hh87 and m11 ( see the end of section [ subsection_std_h ] ) .
the aim of the study is thus to arrive at a formulation for the moist - air enthalpy , @xmath5 , that would be similar to results ( [ def_m11_s ] ) and ( [ def_m11_thetas ] ) valid for the moist air entropy , @xmath10 . since the aim of the article is to compute the moist - air enthalpy by determining the reference and standard enthalpies for each species , it is important to explain why the same method is relevant for the moist - air entropy and how it is possible to recover the well - known formulae used in atmospheric science and based on the equivalent and liquid - water potential temperatures @xmath30 and @xmath29 , respectively .
a synthetic view of existing formulations of moist - air entropy derived in iribarne and godson , 1973 , betts , 1973 ( hereafter referred to as b73 ) ; or emanuel , 1994 ( hereafter referred to as e94 ) , is given in p10 .
two moist - air entropies are defined .
the first one is called `` moist entropy '' and is denoted @xmath102 in p10 .
it is written as @xmath103 in ( [ def_p10_se ] ) , since it is associated with @xmath30 .
@xmath104 \ln\left ( \frac{t}{t_r}\right ) + \ : q_v \:\frac{l_{vap}}{t } - \ : q_d \ : r_d \ : \ln\left ( \frac{p - e}{p_r - e_r}\right ) - \ : q_v \ : r_v \ : \ln\left ( \frac{e}{e_{sw}}\right ) .
\label{def_p10_se}\end{aligned}\ ] ] the second one is given by ( [ def_p10_sl ] ) .
it is called `` dry entropy '' and denoted @xmath105 in p10 .
it is associated with the b73 value @xmath29 . @xmath106 \ln\left ( \frac{t}{t_r}\right ) - \ : q_l \:\frac{l_{vap}}{t } - \ : q_d \ : r_d \ : \ln\left ( \frac{p - e}{p_r - e_r}\right ) - \ : q_t \ : r_v \ : \ln\left ( \frac{e}{e_r}\right ) .
\label{def_p10_sl}\end{aligned}\ ] ] it is possible to show that the difference between ( [ def_p10_sl ] ) and ( [ def_p10_se ] ) is equal to @xmath107 .
this is slightly different from the value given in ( a5 ) of p10 because the term @xmath108 was written as @xmath109 in ( a3 ) of p10 .
the important feature is that @xmath110 depends on @xmath21 ( both here and in p10 ) .
therefore , since there is only one physical definition for the moist - air entropy and since @xmath21 is not a constant in the real atmosphere , @xmath103 and @xmath105 can not represent the more general form of moist - air entropy at the same time . more precisely , the comparisons between the third law - based formulation @xmath111 derived in m11 and the two formulations @xmath112 or @xmath113 can be written as @xmath114 + ( s_d)_r \ : , \label{def_p10_s_se } \\ s \ , ( \theta_s ) & = s_l ( \theta_l ) + q_t \left [ \ : ( s_v)_r - ( s_d)_r \ : \right ] + ( s_d)_r \ : .
\label{def_p10_s_sl}\end{aligned}\ ] ] the first result is that , if @xmath21 is a constant , then @xmath103 and @xmath105 can become specialized versions of the moist - air entropy , associated with the use of the conservative variables @xmath30 and @xmath29 , respectively .
however , even if @xmath103 and @xmath105 are equal to @xmath10 up to true constant terms , the constant terms are not equal to zero and they are not the same for @xmath103 and @xmath105 .
moreover , they depend on the value of @xmath21 .
therefore , even if @xmath21 is a constant ( for instance for a given vertical ascent of moist air ) , it is not possible to compare values of @xmath103 or @xmath105 with those for other columns , since the values of @xmath21 and the `` constant '' terms in ( [ def_p10_s_se ] ) and ( [ def_p10_s_sl ] ) differ from one column to another .
this means that it is not possible to compute relevant spatial averages , fluxes or gradients of @xmath112 and @xmath113 , because the link between @xmath30 or @xmath29 and the moist - air entropy must change in space and in time .
the other result is that , if @xmath21 is not a constant , then @xmath115 only if @xmath116 .
similarly , @xmath117 only if @xmath118 . the fact that @xmath112 or @xmath113 may represent the moist air entropy for an open system and varying @xmath21 is thus dependent on these arbitrary choices for the reference entropies .
since the third law is a general thermodynamic property , this study considers that only the third - law - based formulation @xmath119 is general enough to allow relevant computations of spatial average , fluxes or gradients of moist air entropy , and thus of @xmath90 .
this can be taken in relation with the problem studied in b82 for computing the moist - air surface turbulent fluxes , which are performed with arbitrary choices of the reference enthalpies .
it is suggested in appendix c of p10 that the weighted average @xmath120 , where @xmath121 is an arbitrary constant , is a valid definition of the entropy of moist air .
the adiabatic formulation @xmath105 corresponds to @xmath122 and @xmath29 , whereas the pseudo - adiabatic formulation @xmath103 corresponds to @xmath123 and @xmath30 .
the weighted sum @xmath124 applied to ( [ def_p10_s_se ] ) and ( [ def_p10_s_sl ] ) leads to @xmath125 .
\label{def_p10_sa } \end{aligned}\ ] ] this result ( [ def_p10_sa ] ) shows that , if the term by which @xmath21 is multiplied is not equal to zero , @xmath124 will be different from the third law formulation @xmath10 .
if the value @xmath126 of m11 is considered , this term is equal to zero for @xmath127 \ : t_r / l_{vap}(t_r ) \approx 0.356 $ ] , which represents ( and allows the measurement of ) the specific entropy of moist air in all circumstances since no other hypothesis is made concerning the values of the reference entropies , or on adiabatic or pseudo - adiabatic properties , or on constant values for @xmath21 .
this provides another explanation for the result derived in m11 : the third - law potential temperature @xmath90 is in a position about `` @xmath128 '' @xmath129 versus `` @xmath121 '' @xmath130 between @xmath29 and @xmath30 .
the third law can not be by - passed when evaluating the general formula of moist air entropy .
reference values must be set to the standard ones obtained with zero - entropy for the most stable crystalline form at @xmath131 k. if @xmath21 is not a constant , the formulations for @xmath111 , @xmath103 , @xmath105 and @xmath124 are thus different . in m11
, it is claimed that the third - law formulation @xmath111 is the more general one and , in this work , we present results to further support this claim
. the issue of whether the third law can be applied to atmospheric studies or not , and how this can be managed practically , is an old question .
richardson ( r22 , pp.159 - 160 ) already wondered if it could be possible to ascribe a value to energy and entropy for a unit mass of ( water ) substance .
he first proposed taking absolute zero temperature as the zero origin of entropies .
he recognized that the entropy varied as @xmath132 and could cause the integral to take infinite values when @xmath131 k. but he mentioned that nernst had shown that the specific heats tended to zero at @xmath131 k in such a way that the entropy remained finite there .
this was due to debye s law , which is valid for all solids and for which @xmath33 is proportional to @xmath133 .
the entropy defined by @xmath134 is thus proportional to @xmath133 , which is not singular at @xmath135 k. however , richardson did not use the third law and he suggested considering the lowest temperature occurring in the atmosphere ( @xmath136 k ) as the most practical value for the zero origin of entropies .
this is in contradiction with the above conclusions of the comparisons of @xmath10 , @xmath103 and @xmath105 , and it is shown in hh87 and m11 that it is indeed possible to use the third law in atmospheric science . the third law is not used in p10 .
it is explained that the entropy of an ideal gas is fundamentally incompatible with nernst s theorem as it is singular for @xmath8 approaching @xmath0 k. this statement is true for ideal gases , but it does not invalidate the application of the third law in atmospheric science since only the most stable solid states and debye s law should be considered to apply nernst s theorem . according to the advice of richardson and to the conclusion of comparisons between the third law entropy with previous results recalled in @xmath124 of p10 , the aim of the article will be to mimic what is done in m11 for entropy in order to compute the thermal enthalpy ( equal to the internal energy plus @xmath137 ) for a unit mass of moist air .
there is another possibility that could avoid the need to use absolute values for entropies or enthalpies , and thus the third law .
the method is to assume that @xmath20 is a constant and to express moist - air entropy and enthalpy `` per unit mass of dry air '' , and not as specific values expressed `` per unit mass of moist air '' .
this is the choice made , in particular , in normand , 1921 ( hereafter referred to as n21 ) , b73 and e94 , where @xmath138 is expressed as any of @xmath139 with @xmath140 the total water mixing ratio .
similar formulae are valid for the enthalpy expressed per unit of dry air if @xmath10 is replaced by @xmath5
. these formulae are valid for a mixture of dry air , water vapour and liquid water , with ( [ def_s_qd_2 ] ) and ( [ def_s_qd_3 ] ) corresponding to @xmath112 and @xmath113 respectively .
clearly , the term @xmath86 can be transformed into the sum of @xmath141 $ ] plus @xmath93 on the right hand sides of ( [ def_s_qd_2 ] ) and ( [ def_s_qd_3 ] ) .
the reference value @xmath93 is a global offset having no physical meaning .
the bracketed terms can be easily computed , together with @xmath142 and @xmath143 in ( [ def_s_qd_2 ] ) and ( [ def_s_qd_3 ] ) , which are equal to @xmath144 and @xmath145 .
however , the reference values involved in @xmath146 $ ] in ( [ def_s_qd_2 ] ) , or in @xmath147 $ ] , in ( [ def_s_qd_3 ] ) are multiplied by @xmath148 and they can be neglected if and only if @xmath148 is a constant .
conversely , they would acquire physical meanings in the regions and for the processes where @xmath148 is not a constant .
moreover , the study of the term @xmath138 would become irrelevant for the case of open systems and varying @xmath149 , for which it would be necessary to multiply the right hand sides by @xmath20 in order to properly compute @xmath10 , with the terms @xmath150 taking on a physical meaning .
there is thus no real improvement with regard to the study of specific values as in ( [ def_p10_se ] ) and ( [ def_p10_sl ] ) .
the aim of the article is thus to apply the method suggested in r22 and demonstrated in m11 for the entropy , in order to derive a formulation of the specific moist - air thermal enthalpy @xmath5 with a minimum of hypotheses .
the result is intended to be valid for varying @xmath21 and with the reference values of enthalpies defined from physical properties , and not prescribed arbitrarily .
specific values of moist - air enthalpy are often computed by using the well - known mses .
there are , however , several formulations for mses , each of them corresponding to different assumptions for the zero - enthalpies of dry air , water vapour or liquid - water , and to different ways of deriving mse from either the first or second laws .
on the one hand , mse formulations are derived from the second law and are often presented as being similar to the equivalent potential temperature @xmath30 , which is conserved during pseudo - adiabatic ascent or descent of moist - air parcels ( n21 ; madden and robitaille , 1970 ; betts , 1974 ; arakawa and schubert , 1974 ; e94 ; and ambaum , 2010 , hereafter referred to as mr70 , b74 , as74 and a10 ) .
the reason why it is possible to associate the moist - air enthalpy , @xmath5 , with mse quantities ( and thus with the second law ) is to be found in the gibbs equation ( [ def_gibbs_1 ] ) .
if steady state vertical motions are considered , the material derivative reduces to @xmath151 , with all the terms in ( [ def_gibbs_1 ] ) being multiplied by the vertical velocity @xmath152 , which can be omitted hereafter . for vertical hydrostatic motion @xmath153 . if the parcel is closed and undergoes reversible adiabatic processes , then @xmath154 and @xmath155 .
the stationary gibbs equation can then be written as @xmath156 the quantity @xmath157 is called the generalized enthalpy in a10 .
it is thus a quantity that is conserved for vertical motion and provided that all the previous assumptions are valid . on the other hand ,
the mses are often interpreted as generalized forms of energy or enthalpy ( first law ) .
for instance , the mse function is interpreted as the non - kinetic part of the total energy in emanuel ( 2004 ) and peterson _ et al . _
( 2011 ) , with the thermal part representing the specific moist enthalpy in e94 , or the fluctuating part of the total specific enthalpy of a parcel in a10 .
the third - law - based thermal enthalpy , @xmath5 , derived in the next section will be compared with a selection of existing mse formulae , which can be written as @xmath158 t + l_{vap}\:q_v \:+ \phi \ : , \label{def_mse_a10 } \\
\mbox{mse}_m & = \ : c_p\:t + l_{vap}\:q_v \:+ \phi \ : .
\label{def_msem}\end{aligned}\ ] ] these mse quantities are made up of three parts . the first part is the product of the local temperature by a moist - air specific heat , the second part is the product(s ) of latent heats by specific contents , and the third part is the potential energy @xmath159 .
the `` thermal '' counterparts of ( [ def_msed ] ) , ( [ def_msel ] ) , ( [ def_mse_a10 ] ) and ( [ def_msem ] ) , where the potential energy is removed , are noted tmse@xmath160 , tmse@xmath161 , @xmath162 and tmse@xmath163 , respectively
. the quantity mse@xmath160 given by ( [ def_msed ] ) is the most popular .
it is considered in as74 as `` approximately conserved by individual air parcels during moist adiabatic processes '' .
betts , 1975 ( hereafter referred to as b75 ) , mentions that it is an approximate analogue of the equivalent potential temperature @xmath30 .
it is used as a conserved variable for defining saturated updraughts in some deep - convection schemes ( bougeault , 1985 ) .
a liquid - water version mse@xmath161 is defined in b75 by removing the quantity @xmath164 ( assumed to be a constant ) from ( [ def_msed ] ) , leading to ( [ def_msel ] ) .
it is considered in b75 that mse@xmath161 is an analogue of the liquid - water potential temperature @xmath29 .
the formulation mse@xmath161 is generalized in khairoutdinov and randall ( 2003 ) and in bretherton _
( 2005 ) by removing a term @xmath165 from ( [ def_msel ] ) , for the sake of symmetry , which leads to the liquid - ice static energy given by ( [ def_limse ] ) .
limse is used as a conserved variable for defining saturated updraughts in some shallow - convection schemes ( bechtold _ et al .
_ , 2001 ) .
the generalized enthalpy @xmath166 defined by eq.(5.37 ) in a10 is given by ( [ def_mse_a10 ] ) .
the `` moist enthalpy '' ( per unit of dry air ) defined by eq.(4.5.4 ) in e94 is equivalent to @xmath167 .
both @xmath166 and @xmath37 are computed in a10 and e94 with the assumptions @xmath168 .
mse is sometimes defined with @xmath12 replaced in ( [ def_msed ] ) by the moist value @xmath9 , leading to mse@xmath163 given by ( [ def_msem ] ) .
this version , mse@xmath163 , is used in some convective schemes ( e.g. gerard _ et al . _ , 2009 ) , because @xmath169 can be easily computed from @xmath170 .
the concept of mse was not explicitly introduced in dufour et van mieghem ( 1975 , hereafter referred to as dvm75 ) .
these authors were mainly concerned with extensive versions for enthalpy and entropy , computed for a mass @xmath171 of moist air , i.e. with @xmath172 and @xmath173 .
the same method already used in m11 is followed in this section to derive the formulation of the specific moist - air thermal enthalpy @xmath5 in terms of a `` moist enthalpy temperature '' @xmath174 , similar to the moist - static - energy temperature mse@xmath175 introduced in derbyshire _
( 2004 ) .
the moist atmosphere is considered as a mixture of different ideal gases , mainly composed of n@xmath32 , o@xmath32 , ar and co@xmath32 for dry air , plus the three phases of the water species h@xmath32o ( vapour , liquid or solid ) .
it is assumed that the volume of the condensed water species can be neglected , although their impact on the moist definitions of @xmath9 is taken into account .
the dry air is a mixture of 79 % n@xmath32 plus 20 % o@xmath32 , with less than @xmath176 % of ar , co@xmath32 and other gases .
if the chemical reactions ( like the ozone - oxygen cycle in the stratosphere ) are neglected , the thermodynamic properties of dry air are thus determined at @xmath177 % by the observed and constant concentrations of the gases n@xmath32 and o@xmath32 , with no tropospheric sources or sinks for these two gases .
in contrast , the specific contents for the three phases of h@xmath32o are highly variable in time and space , with the evaporation and the precipitation processes acting as tropospheric sources and sinks , respectively .
the moist atmosphere is thus a multi - component mixture of different ideal gases , i.e. dry air plus vapour , and liquid and solid water species .
since specific enthalpy is an additive function ( dalton s law ) , the moist - air formulation is equal to the weighted average of the individual values for the partial specific enthalpies of the dry air , water vapour , liquid water and ice species , leading to @xmath178 the terms can be rearranged ( with the result @xmath179 ) , yielding @xmath180 where the latent heats of vaporization and sublimation are equal to @xmath71 and @xmath181 . as in m11 , the assumption @xmath182 is made , with precipitations considered as equivalent to cloud contents and taken to be at the same temperature as the rest of the parcel .
the result is @xmath183 the next step is to express the enthalpies by linearizing around some reference value @xmath13 , with the hypotheses of constant values for all the specific heats @xmath184 to @xmath185 in the atmospheric range of temperature ( i.e. from @xmath136 to @xmath186 k ) .
this leads to @xmath187 the aim of the next section will be to compute the reference values @xmath14 and @xmath188 , with @xmath18 and @xmath189 determined from @xmath188 by the latent heats @xmath190 and @xmath191 .
if ( [ def_hdr ] ) and ( [ def_hvr ] ) are inserted into ( [ def_h2 ] ) , and after rearrangement of the terms , the moist enthalpy can be written as @xmath192 \nonumber \\ & \quad + \ : { c}_{pd } \ : t \ : \left [ \ : 1 \ : + \ : \lambda \:q_t \ : - \ : \left ( \ : \frac{l_{vap}\:q_l + l_{sub}\:q_i}{{c}_{pd}\ : t } \ : \right ) \ : \right ] \nonumber \\ & \quad \ : + \ : q_t \ : \left [ \ : ( h_v)_r \ : - \ : ( h_d)_r \ : \ : - \ : ( { c}_{pv } \ : - \:{c}_{pd } \ : ) \ : t_r \ : \right ] .
\label{def_h3}\end{aligned}\ ] ] the formulation ( [ def_h3 ] ) is equivalently written by the system ( [ def_h])-([def_upsilon ] ) , in a similar way to the moist entropy formulation ( [ def_m11_s ] ) derived in m11 in terms of @xmath90 , but this time in terms of a moist enthalpy temperature @xmath174 , yielding @xmath193 the first line of ( [ def_h3 ] ) defines @xmath194 by ( [ def_href ] ) . the term @xmath195 and the dimensionless upsilon - term @xmath196 correspond to the last two lines of ( [ def_h3 ] ) , leading to ( [ def_th])-([def_upsilon ] ) . @xmath197
\ : , \label{def_th } \\
t_{il } & = \ : t \ : \left [ \ : 1 - \left ( \ : \frac{l_{vap}(t)\:q_l + l_{sub}(t)\:q_i}{{c}_{pd}\ : t } \ : \right ) \ : \right ] , \label{def_t_l } \\ { \upsilon\!}_r & = \ ; \frac { ( h_{v})_r - ( h_{d})_r } { c_{pd } \ ; t_r } \ : .
\label{def_upsilon}\end{aligned}\ ] ] it is worth noting that the ice - liquid temperature @xmath198 corresponds to a generalization for non - zero @xmath199 of the liquid - water ( potential ) temperature defined in b73 and involved in the first line of @xmath90 recalled in ( [ def_m11_thetas ] ) , in the limit where @xmath200 and @xmath201 for small @xmath202 .
it is more general than the ice - liquid water ( potential ) temperature defined in tripoli and cotton ( 1981 ) ,
in that the constant values @xmath203 and @xmath204 are replaced by the varying values @xmath205 and @xmath206 in ( [ def_t_l ] ) . in ( [ def_upsilon ] ) , the upsilon - term @xmath196 depends on the absolute values for water - vapour and dry - air enthalpies , evaluated at the temperature @xmath13 .
it acts as the dimensionless term @xmath207/c_{pd}$ ] appearing in the entropy formulation ( [ def_m11_thetas ] ) , and it varies with @xmath13 as illustrated in figure [ fig_upsilon_tr ] . of the upsilon term @xmath208 ( solid curve ) , in comparison with the constant value @xmath209 ( dashed straight line ) .
_ [ fig_upsilon_tr ] ] it is important to demonstrate that choosing another value of @xmath13 has no impact on @xmath5 , @xmath194 and @xmath174 as defined by ( [ def_h ] ) to ( [ def_upsilon ] ) respectively .
otherwise , if @xmath194 could be dependent on @xmath13 , the enthalpy temperature @xmath174 would not be equivalent to the moist enthalpy @xmath5 , as is expected , and the moist thermal enthalpy itself would not have a clear physical meaning as a state function of the moist atmosphere .
this result is true for @xmath194 because , from ( [ def_href ] ) and for constant values of @xmath184 , the variations of @xmath210 with @xmath13 are exactly balanced by the term @xmath211 , leading to @xmath212 which corresponds to the definition ( [ def_hdr ] ) applied between @xmath213 and @xmath214 .
the reference enthalpy ( [ def_href ] ) can thus be written as @xmath215 the numerical value @xmath216 kj kg@xmath217 is obtained from the standard value @xmath218 determined in the next section at @xmath219 .
it is shown in figure [ fig_upsilon_tr ] that @xmath220 is not a constant and decreases strongly with @xmath13 . however , the bracketed term in ( [ def_th ] ) is a constant term independent of @xmath13 . from the definitions @xmath221 and ( [ def_upsilon ] ) , the term @xmath222 is equal to @xmath21 times @xmath223 - [ \:(h_{d})_r - { c}_{pd } \ : t_r\:]$ ] .
moreover , from ( [ def_hdr ] ) and ( [ def_hvr ] ) , neither @xmath224 nor @xmath225 varies with @xmath13 , as long as @xmath184 and @xmath226 are assumed to be constant terms .
the result is that the bracketed term in ( [ def_th ] ) is equal to @xmath227 \ , = \
, t_0 \ : ( { \upsilon\!}_0 - \lambda ) \ , \equiv \ , t_{\upsilon } \approx \ : 2362 \mbox{~k } \ , . \label{def_tr_upsilon_cste}\ ] ] @xmath228 is a constant , due to compensating variations with @xmath13 of the two terms @xmath229 and @xmath230 , as illustrated in figure [ fig_h_cste_tr ] . the numerical value @xmath231 k is obtained from the standard values given in the next section , leading to @xmath232 . of the three terms :
`` @xmath233 '' ( dashed , increasing curve ) ; `` @xmath234 '' ( dotted - dashed , decreasing curve ) and `` @xmath235$]''@xmath236 k ( solid , constant straight line ) . _ [ fig_h_cste_tr ] ] an alternative , synthetic and compact expression for @xmath5 defined by ( [ def_h])-([def_tr_upsilon_cste ] ) is given by @xmath237 the constant terms @xmath194 and @xmath228 are given by ( [ def_href0 ] ) and ( [ def_tr_upsilon_cste ] ) .
it is worth noting that the formula ( [ def_h_bis ] ) is symmetric with respect to the condensed water species : it is left unchanged if @xmath82 is replaced by @xmath199 provided that @xmath238 is replaced by @xmath239 , where @xmath238 and @xmath239 connect the water vapour with the condensed phases ( @xmath82 or @xmath199 ) .
one of the goals of this section is to determine the numerical values of the constants @xmath194 and @xmath228 appearing in ( [ def_h_bis ] ) .
they both depend , from ( [ def_href0 ] ) and ( [ def_tr_upsilon_cste ] ) , on the standard values of the partial enthalpies @xmath240 and @xmath241 .
the term `` standard '' means that the values are expressed at the zero celsius temperature @xmath242 k and the conventional pressure @xmath101 hpa .
the aim of this section is thus to provide a general method for computing the partial enthalpies for several components of moist air ( o@xmath243 , n@xmath243 , h@xmath243o ) and at any temperature , in particular for the standard conditions .
the partial enthalpies must be understood as `` thermal enthalpies '' , i.e. the enthalpies generated by the variations of @xmath33 with @xmath8 .
this corresponds to progressive excitations of the translational , rotational and vibrational states of the molecules , and by the possible changes of phase represented by the latent heats ( with negligible impact of changes of pressure ) .
the consequence is that the concept of thermal enthalpy does not correspond to the concept of `` standard enthalpies of formation or reaction '' which are denoted by @xmath244 and @xmath245 and are available in chemical tables for most species .
the concepts of standard enthalpies of formation or reaction are useful in case of chemically reacting species , with the consequence that the relative concentrations of these species are determined by the equilibrium values depending on the law of mass action , depending on the reaction considered . in that case , the concentrations depend on local pressure and temperature ( as in the o@xmath246 chemistry regions ) .
but , except in the stratosphere , the concentrations of the species are free parameters with , for instance , tropospheric water contents that are modified by diffusion , convection or precipitation processes .
the standard enthalpies of formation or reaction thus can not be relevant starting points for the determination of the tropospheric values of standard thermal enthalpies @xmath240 and @xmath241 , and thus @xmath247 .
moreover , in chemical thermodynamics , it is assumed that the hypothesis @xmath248 is valid for all gaseous forms of pure species at the conventional pressure of @xmath249 hpa and the standard temperature of @xmath250 k. this is true , for instance , for o@xmath32 and n@xmath32 and thus for @xmath177 % of dry air .
however , since the thermal enthalpies must depend on the variations of @xmath33 with @xmath8 and on the latent heats , and since these quantities are different for each gas , the thermal enthalpies of o@xmath32 and n@xmath32 must be different in the standard conditions and @xmath251 .
the method for computing the thermal enthalpies of n@xmath32 , o@xmath32 and h@xmath32o makes use of the important result that enthalpy is a state function : it can be evaluated by following any of the reversible paths that connect the dead state at @xmath0 k and the actual atmospheric state at temperature @xmath8 .
the challenge is thus to compute the integral of @xmath33 and the sum of all latent heats for a given substance from @xmath0 k to the standard temperature @xmath219 , by following a reversible path involving the most stable forms of this substance at each temperature , first for the solid state(s ) , then for the liquid state , and finally for the gaseous form of the substance .
all these processes are represented by the various terms of the mathematical formula @xmath252 other forms of enthalpy are associated with the potential ( @xmath253 ) , chemical or nuclear energies . corresponding to table [ table_cp_o2 ] in appendix b. units of @xmath9 are j k@xmath217 kg@xmath217 .
the latent heats are in units of kj kg@xmath217 . _ [ fig_cp_o2 ] ] corresponding to table [ table_cp_n2 ] in appendix b. units of @xmath9 are j k@xmath217 kg@xmath217 .
the latent heats are in units of kj kg@xmath217 . _ [ fig_cp_n2 ] ] in appendix b , and the usual constant value @xmath254 ( dashed line ) .
units are j k@xmath217 kg@xmath217 . _
[ fig_cp_ice ] ] the arbitrary reference level @xmath255 needed for defining @xmath256 is the same for all the molecules at a given point .
the impact of @xmath255 is thus to add or to retrieve a true constant term independent of @xmath81 , @xmath82 and @xmath199 .
therefore @xmath255 has no physical impact on the energy or on the enthalpy functions .
this is true even from the barycentric standpoint of fluid dynamics , where differential fluxes of matter occur . nuclear reactions and the formation of atoms from elementary particles form a huge but dead state of energy in the real atmosphere .
however , these forms of unavailable energy are not taken into account in this study .
it is assumed that the main gases , n@xmath32 , o@xmath32 and h@xmath32o , that compose the moist atmosphere only interact with the kinetic and potential energies via the standard thermal enthalpies defined by ( [ def_abs_h ] ) .
( bold solid line ) and n@xmath32 ( thin solid line ) as a function of the absolute temperature , with units of kj kg@xmath217 .
the dry - air curve ( dashed line ) represents the dry - air enthalpy ( @xmath19 ) , computed and plotted from @xmath257 to @xmath186 k with the `` constant specific heat capacity '' assumption ( @xmath258 j k@xmath217 kg@xmath217 ) .
_ [ fig_all_enthalpies2 ] ] .
the enthalpy of ice - ih is plotted in the range of temperature from @xmath0 to @xmath259 k ( thin solid line ) .
the three dashed lines represent the ice , liquid - water and water - vapour enthalpies computed with `` constant specific heat capacity '' assumptions ( @xmath260 , @xmath261 and @xmath262 j k@xmath217 kg@xmath217 ) .
the three bold dots represent the standard values @xmath263 , @xmath264 and @xmath265 j kg@xmath217 .
latent heat curves are described in the text .
_ [ fig_all_enthalpies1 ] ] the integral of @xmath33 in the thermal enthalpy formulation ( [ def_abs_h ] ) can be computed by using the specific heats described in the tables [ table_cp_o2 ] to [ table_h2o_ice ] for o@xmath32 , n@xmath32 and ice - ih ( see appendix b ) .
the latent heats are indicated in figures [ fig_cp_o2 ] , [ fig_cp_n2 ] and [ fig_all_enthalpies1 ] .
the remaining problem is thus to determine the value of the thermal enthalpy at @xmath0 k. the standard entropies used in h87 and m11 are based on the third law .
it is assumed that a kind of third law can apply to the thermal enthalpy for all substances with no chemical reaction , i.e. zero - enthalpy for @xmath131 k. this means that the thermal enthalpies of n@xmath32 , o@xmath32 and h@xmath32o are equal to zero for the most stable solid states at @xmath0 k ( thermally dead states ) .
otherwise , if any indeterminacy existed for the values of thermal enthalpies at @xmath0 k , except a possible identical global shift valid for all species , the thermal enthalpy of moist air could not be defined locally and would not have a clear physical meaning for all values of @xmath219 in ( [ def_abs_h ] ) .
the curves for the thermal enthalpies of dry - air species are shown in figure [ fig_all_enthalpies2 ] .
they were computed with the values of @xmath33 and the latent heats given in appendix b for o@xmath32 and n@xmath32 ( tables [ table_cp_o2 ] and [ table_cp_n2 ] ) .
the thermal enthalpies of o@xmath32 and n@xmath32 have about the same value of @xmath266 kj kg@xmath217 at @xmath267 k , with about the same value for dry air ( composed at 99% of o@xmath32 plus n@xmath32 ) .
a thin horizontal line has been added to represent this common value @xmath266 j kg@xmath217 . according to figures [ fig_cp_o2 ] and [ fig_cp_n2 ] , the variation with @xmath8 of specific heat for o@xmath32 and n@xmath32 is small above @xmath268
k. this corresponds to constant values for @xmath12 and therefore the dry - air thermal enthalpy can be approximated by @xmath269 for @xmath270 k , leading to a standard value of @xmath271 of @xmath272 kj kg@xmath217 at @xmath259 k represented by the `` crossed circle '' symbol on figure [ fig_all_enthalpies2 ] .
the curves for the thermal enthalpies of water species are depicted in figure [ fig_all_enthalpies1 ] .
they were computed with the values of @xmath33 plotted in figure [ fig_cp_ice ] and the specific heats given in appendix c for phase ih of ice ( table [ table_h2o_ice ] ) , plus the values of latent heats of fusion and vaporization @xmath273 and @xmath274 given in appendix a. the latent heats of sublimation ( @xmath275 ) and vaporization ( @xmath276 ) are depicted by solid bold lines .
the maximum at @xmath277 k for @xmath239 corresponds to a known thermodynamic property .
the same dry - air dashed line as in figure [ fig_all_enthalpies2 ] is shown in figure [ fig_all_enthalpies1 ] from @xmath268 to @xmath186 k , with the same `` crossed circle '' symbol at @xmath278 k. the thin horizontal line represents the same constant value of @xmath272 j kg@xmath217 as plotted in figure [ fig_all_enthalpies2 ] , in order to facilitate the comparison of the vertical scales in figures [ fig_all_enthalpies1 ] and [ fig_all_enthalpies2 ] . finally , the values of the standard thermal enthalpies for dry - air and water species are given by @xmath279 the standard values @xmath240 and @xmath241 lead to @xmath280 in ( [ def_upsilon ] ) , @xmath281 kj kg@xmath217 in ( [ def_href0 ] ) and @xmath282 k in ( [ def_tr_upsilon_cste ] ) .
the term @xmath283 is close to @xmath284 and it is about ten times larger than @xmath285 .
this means that @xmath286 is a dominant term in the last term of ( [ def_h_bis ] ) , which is thus almost equal to @xmath164 .
thus the bracketed term in ( [ def_th ] ) can not be neglected , because it is of the same order of magnitude as the other moist terms depending on @xmath21 , @xmath82 or @xmath199 in ( [ def_t_l ] ) .
if the specific heat for dry air , ice , liquid water and water vapour are assumed to be constant with @xmath8 above @xmath268 k , the thermal enthalpies can be approximated by @xmath287 reference values are computed with @xmath288 in ( [ def_hd_term])-([def_hi_term ] ) .
the important result is that hypotheses of linear laws ( @xmath289 ) for enthalpies of dry - air or water species are clearly invalidated since the straight lines @xmath290 intersect at @xmath0 k in figure [ fig_all_enthalpies1 ] , at the non - zero values @xmath291 .
it is worth noting that the same dataset used to compute the standard values of enthalpies can be used to compute the standard values of entropies for o@xmath32 , n@xmath32 and h@xmath32o . it is important to check this so as to demonstrate the good level of accuracy of the dataset and method described in this section .
the results ( not shown ) indicated that the standard entropies are close to the values published in hh87 and used in m11 : @xmath292% deviation for n@xmath32 vapour , @xmath293% for o@xmath32 vapour and @xmath294% for ice , including a residual entropy of @xmath295 j k@xmath217 kg@xmath217 for ice - ih at @xmath0 k due to proton disorder and to the remaining randomness of hydrogen bonds at @xmath0 k ( pauling , 1935 , nagle , 1966 ) .
it is assumed in many studies ( b82 , e94 or a10 ) that @xmath296 at @xmath99 c and that this common value for the enthalpies is equal to zero .
the accuracies of these two approximations are analyzed in this section .
the special temperature for which @xmath297 can be computed by setting ( [ def_hd_term ] ) and ( [ def_hl_term ] ) equal .
the result @xmath298 k @xmath299 c is precisely the temperature for which the curves for dry - air and liquid - water enthalpies intersect in figure [ fig_all_enthalpies1 ] .
however , the difference @xmath300 is positive and becomes progressively larger for increasing temperatures and in particular for temperatures that are positive on the celsius scale .
the first hypothesis @xmath296 is thus an approximation based on a coincidence .
it is not a fundamental physical property and it is not valid for all temperatures .
this means that it could be worthwhile to avoid this approximation and to compute the moist - air enthalpy by ( [ def_h_bis ] ) and with the more accurate observed values for @xmath301 and @xmath302 given by ( [ def_hd_term ] ) and ( [ def_hl_term ] ) .
the second hypothesis is that both @xmath303 and @xmath304 could be equal to zero , or that they could be small terms in comparison with the latent heat @xmath305 .
a large vertical step is indeed observed at @xmath219 in figure [ fig_all_enthalpies1 ] between the water vapour and both liquid - water and dry - air enthalpies .
this means that @xmath274 is the dominant term in this diagram .
the reason why the latent heat of vaporization for h@xmath32o is much larger than those for n@xmath32 or o@xmath32 ( @xmath306 versus @xmath307 and @xmath308 kj kg@xmath217 ) is due to trouton s rule ( wisniak , 2001 ) .
this rule states that almost all substances follow the general result @xmath309 j mol@xmath217 k@xmath217 at their boiling temperature and at normal pressure of @xmath249 hpa .
the dominant feature for @xmath274 is thus a consequence of the boiling temperature of h@xmath32o ( 373.15 k ) , which is about @xmath310 times greater than those for n@xmath32 ( 77.4 k ) and o@xmath32 ( 90 k ) .
the consequence of the coincidence @xmath311 and of trouton s rule is that @xmath312 is valid to within @xmath313 % .
the last term of ( [ def_h_bis ] ) can thus be approximated by @xmath314 , leading to the approximate formula for the thermal enthalpy of moist air @xmath315 unlike ( [ def_h_bis ] ) , this approximate formula is not symmetric if @xmath316 is replaced by @xmath317 , since it is the term @xmath318 that is involved in ( [ def_h_ter ] ) .
this means that the internal liquid - ice symmetry of the enthalpy equation ( [ def_h_bis ] ) may not be observed in approximate versions of @xmath5 .
the results presented in this section show the extent to which the thermodynamic properties of water are special and different from the properties of dry air .
the coincidence @xmath311 and trouton s rule associated with the high boiling temperature of h@xmath32o are empirical results . since it is possible to avoid these approximations , and since they can lead to systematic errors , theoretical investigation dealing with moist - air thermal enthalpy should be based on the more general formula ( [ def_h_bis ] ) .
otherwise , some physical properties might be poorly represented or misinterpreted .
the structures of @xmath319 , with @xmath5 given by ( [ def_h_bis ] ) or by the approximation ( [ def_h_ter ] ) , are clearly similar to the mse quantities recalled in section [ subsection_mse ] , except for the constant term @xmath194 , which has no physical meaning since it corresponds to a global shift in the moist enthalpy units
. it will be shown in the following sections that @xmath319 is especially close to mse@xmath160 and mse@xmath163 given by ( [ def_msed ] ) and ( [ def_msem ] ) . accordingly
, in this section , comparisons are made between the analytical formulations of the thermal enthalpy @xmath320 computed with ( [ def_h_bis ] ) and the quantities tmse@xmath160 and tmse@xmath163 given by ( [ def_msed ] ) and ( [ def_msem ] ) .
the relative accuracies of the approximation of @xmath5 by tmse@xmath160 or tmse@xmath163 are evaluated by the terms @xmath321/(c_{pd}\:t)$ ] and @xmath322/(c_{pd}\:t)$ ] .
after some algebra , they can be written as @xmath323 since the specific contents @xmath21 , @xmath82 and @xmath199 are small terms ( less than @xmath324 kg kg@xmath217 ) , @xmath325 and @xmath325 are less than @xmath326 % if absolute values of the terms in parentheses in ( [ def_deltah3 ] ) and ( [ def_deltah2 ] ) are lower than @xmath327 . the last two terms in parentheses in ( [ def_deltah2 ] ) can be approximated by @xmath328 for @xmath82 and @xmath329 for @xmath199 . the first term in parentheses in ( [ def_deltah2 ] ) can be evaluated from the results @xmath282 k and @xmath330 , leading to a positive value of about @xmath292 for the first quantity in factor @xmath21 .
similarly , the first term in parentheses in ( [ def_deltah3 ] ) is about @xmath331 at @xmath242 k , and the second term in parentheses in ( [ def_deltah3 ] ) is equal to @xmath332 . if @xmath21 is replaced by @xmath333 , @xmath334 and @xmath335 can be approximated at @xmath219 by @xmath336 @xmath334 is thus a positive difference indicating that the larger the water species contents @xmath81 and @xmath82 are , the more tmse@xmath163 overestimates the moist thermal enthalpy @xmath320 . except for very high and unrealistic values of @xmath199 ,
@xmath335 is negative and tmse@xmath160 underestimates @xmath320 .
the specific ice content @xmath199 further increases the overestimation of tmse@xmath163 , while it decreases the underestimation of tmse@xmath160 .
the approximations of @xmath337 by tmse@xmath163 or tmse@xmath160 can thus be considered as accurate for under - saturated conditions ( @xmath338 ) , with @xmath334 and @xmath339 lower than @xmath176% for @xmath81 lower than @xmath340 g kg@xmath217 .
the differences become larger within clouds , and in particular for tmse@xmath163 , with @xmath334 increasing by about @xmath341% for each @xmath176 g kg@xmath217 of liquid - water content .
possible systematic differences of about @xmath176% correspond to differences in @xmath174 and @xmath8 of about @xmath327 k. this can be of some importance for the accurate determination of the moist thermal enthalpy within clouds or moist regions . such large differences may modify the analysis of the impact of drying or moistening processes on @xmath5 and on the local temperature . this is another justification for the use of the complete formula ( [ def_h_bis ] ) when evaluating @xmath5 , with no approximation for the second line depending on @xmath21 .
the possibility of computing specific moist values for the thermal enthalpy @xmath5 offers the opportunity to plot the enthalpy diagram shown in figure [ fig_h_qt_diagrams ] , where @xmath21 is plotted as a function of @xmath5 .
this diagram is the enthalpy counterpart of the entropy diagram ( @xmath21 , @xmath10 ) given in marquet and geleyn ( 2013 ) . in this enthalpy diagram ,
the moist thermal enthalpy @xmath5 given by ( [ def_h_bis ] ) is compared with the quantity @xmath342 , where tmse@xmath343 is the thermal part of ( [ def_msem ] ) .
the use of tmse@xmath163 instead of @xmath5 generates an error @xmath334 measured by ( [ def_deltah2 ] ) and represented by the thin solid lines ( values @xmath344 and @xmath313 % ) .
positive values of @xmath334 correspond to dashed lines located to the right of the solid lines , or equivalently to tmse@xmath345 .
the water - vapour saturation pressure is computed with the liquid - water formulation if @xmath346 c and with the ice formulation otherwise .
this creates zigzag features for @xmath334 in the saturated regions and around @xmath0 c ( about @xmath347 kj kg@xmath217 ) .
the total specific water content is plotted against the moist enthalpy @xmath5 or against @xmath348tmse@xmath163 , with @xmath349 kj kg@xmath217 .
the saturation curves are plotted in the centre of the diagram ( bold solid lines for @xmath5 and dashed lines for tmse@xmath163 ) .
the saturated regions are located above the saturation curves .
isotherms are labelled every @xmath340 k , with bold solid lines for @xmath5 and dashed lines for tmse@xmath163 .
other elements are described in the text .
[ fig_h_qt_diagrams ] _ ] the isotherms are represented on figure [ fig_h_qt_diagrams ] for either @xmath5 ( solid lines ) or @xmath348tmse@xmath163 ( dashed lines ) as abscissae ( from @xmath350 to @xmath351 k by steps of @xmath340 k ) .
the two sets of isotherms are almost superimposed in the non - saturated domain .
this means that the approximation @xmath352 is accurate in the non - saturated region , where @xmath334 increases gradually from @xmath0 % for the dry - air case to less than @xmath353 % to @xmath292 % , depending on the temperature .
it is , however , worth noting that a value of @xmath294% for @xmath334 corresponds to large differences in @xmath8 of about @xmath354 k. ( a ) is depicted as a dotted line for clear - air conditions and as dark circles for in - cloud regions .
the enthalpy temperature @xmath174 and tmse@xmath355 are represented by a solid line and dark squares for @xmath174 ( clear - air and in - cloud regions ) and by a bold dashed line and open diamonds for tmse@xmath355 ( clear - air and in - cloud regions ) .
( b ) the epic stratocumulus vertical profile and ( c ) the bomex shallow cumulus vertical profile , where the profiles of tmse@xmath356 are plotted .
thin vertical solid lines highlight differences between @xmath174 , tmse@xmath355 and tmse@xmath356 .
_ [ fig_hm_mse_sc],title="fig : " ] ( b ) is depicted as a dotted line for clear - air conditions and as dark circles for in - cloud regions .
the enthalpy temperature @xmath174 and tmse@xmath355 are represented by a solid line and dark squares for @xmath174 ( clear - air and in - cloud regions ) and by a bold dashed line and open diamonds for tmse@xmath355 ( clear - air and in - cloud regions ) .
( b ) the epic stratocumulus vertical profile and ( c ) the bomex shallow cumulus vertical profile , where the profiles of tmse@xmath356 are plotted .
thin vertical solid lines highlight differences between @xmath174 , tmse@xmath355 and tmse@xmath356 . _
[ fig_hm_mse_sc],title="fig : " ] ( c ) is depicted as a dotted line for clear - air conditions and as dark circles for in - cloud regions .
the enthalpy temperature @xmath174 and tmse@xmath355 are represented by a solid line and dark squares for @xmath174 ( clear - air and in - cloud regions ) and by a bold dashed line and open diamonds for tmse@xmath355 ( clear - air and in - cloud regions ) .
( b ) the epic stratocumulus vertical profile and ( c ) the bomex shallow cumulus vertical profile , where the profiles of tmse@xmath356 are plotted .
thin vertical solid lines highlight differences between @xmath174 , tmse@xmath355 and tmse@xmath356 . _
[ fig_hm_mse_sc],title="fig : " ] according to the formula ( [ def_xm_approx ] ) , the approximations are greater in the saturated region ( above the saturation curves ) , reaching @xmath176 % for @xmath357 g kg@xmath217 or @xmath358 g kg@xmath217 .
larger differences between moist thermal enthalpy and tmse@xmath163 values are indeed observed in the saturated region of the enthalpy figure [ fig_h_qt_diagrams ] .
the black spots represent moistening processes associated with large isothermal increases of @xmath21 which originate at the saturation line at @xmath359 and @xmath360 k. these processes correspond to a decrease in @xmath5 in the ice region and to almost constant values in the liquid - water domain .
conversely , tmse@xmath163 exhibits a clear increase with @xmath21 in the saturated domains and for the whole range of temperatures .
although increases of more than @xmath361 g kg@xmath217 for @xmath21 are large and unrealistic , this is only for the sake of clarity and the same kind of changes in @xmath5 or tmse@xmath163 are observed for smaller and more realistic increases of @xmath21 .
these differences suggest that tmse@xmath163 does not give an accurate measurement of the thermal enthalpy for saturated moist air , because they correspond to different physical properties .
such differences may be observed within stratocumulus or cumulus clouds , implying possible systematic biases for the impacts of entrainment and detrainment processes on the budget of moist enthalpy ( and thus on the local temperature ) . and of several mse quantities defined in ( [ def_msed])-([def_msem ] ) for ( a ) the fire - i ( rf03b ) stratocumulus and ( b ) the bomex shallow cumulus .
units are in kj kg@xmath217 . _
[ fig_mse_deficit],title="fig : " ] and of several mse quantities defined in ( [ def_msed])-([def_msem ] ) for ( a ) the fire - i ( rf03b ) stratocumulus and ( b ) the bomex shallow cumulus .
units are in kj kg@xmath217 . _
[ fig_mse_deficit],title="fig : " ] + ( a ) ( b ) + vertical profiles of @xmath174 , tmse@xmath163 and tmse@xmath160 are compared in figures [ fig_hm_mse_sc](a)-(c ) for two stratocumulus cases ( fire - i and epic ) and one shallow cumulus case ( bomex ) .
they correspond to the datasets given in appendix c. cloud liquid - water contents are small , but the values of @xmath81 in the pbl are large enough to allow clear evaluations of impacts of water vapour . for the sake of readability ,
vertical profiles of tmse@xmath160 are not plotted for fire - i in ( a ) .
a large positive top - of - pbl jump in @xmath8 is associated with large negative jumps in both @xmath174 and tmse@xmath355 in ( a ) . moreover , in - cloud and clear - air values are different in the entrainment region represented by the thin solid line boxes .
these results are different from the important properties observed in m11 with the same dataset but with moist entropy , where it is shown that in - cloud and clear - air values of @xmath90 are almost equal and almost constant in the whole pbl , including in the entrainment region at the top .
it can be concluded that the specific moist enthalpy represented by @xmath174 is not conserved ( nor well - mixed ) by the moist turbulence within the pbl of fire - i stratocumulus , contrary to what happens for the specific moist entropy .
the approximation of ( [ def_xm_approx ] ) suggests that tmse@xmath355 is warmer than the enthalpy temperature @xmath174 , due to the term @xmath362 , whereas ( [ def_xd_approx ] ) suggests that tmse@xmath356 is cooler , due to the term @xmath363 .
this is observed for all profiles ( a)-(c ) , where tmse@xmath163 overestimates @xmath174 ( and thus the moist enthalpy ) by about @xmath176 k in the warm and moist lower troposphere , whereas tmse@xmath160 underestimates @xmath174 by about @xmath364 k for epic and bomex .
according to ( [ def_deltah3 ] ) to ( [ def_xd_approx ] ) , the three formulations for @xmath5 , tmse and tmse@xmath160 are the same in the dry - air limit ( @xmath365 ) .
this is confirmed by the vertical profiles depicted on figures [ fig_hm_mse_sc](a)-(c ) : the curves converge above the top of the pbl , where @xmath81 is small . in order to investigate the comparison between enthalpy and mse quantities differently ,
vertical profiles of deficit in surface values are shown in figure [ fig_mse_deficit](a ) for the fire - i stratocumulus case and in ( b ) for the bomex shallow cumulus case .
this kind of surface - deficit chart is commonly used in studies of convective processes .
it is shown that mse@xmath163 , @xmath319 and mse@xmath160 remain close to each other at each level and for the two cases , with the generalized enthalpy located between the others and with mse@xmath163 being a better approximation for @xmath319 .
this confirms the results observed in figures [ fig_hm_mse_sc](a ) and ( c ) .
in figure [ fig_mse_deficit](a)-(b ) , limse and @xmath166 are more different from @xmath319 than mse@xmath163 and mse@xmath160 .
this means that the thermal part of limse and @xmath166 can not represent the moist - air thermal enthalpy @xmath5 accurately .
large jumps in all variables are observed close to the surface for bomex , except for limse , due to the impact of large values of surface specific humidity that are not taken into account in limse .
the impact of surface values are not observed for fire - i , since the in - flight measurements were taken at altitude .
the conclusion of this section is that mse@xmath163 is probably the best candidate for approximating @xmath319 .
however , the fact that systematic differences exist between @xmath5 and mse quantities in the moist lower pbl only , and not in the dry air above , may have significant physical implications if the purpose is to accurately analyse moist enthalpy budgets or differential budgets , or to understand the convective processes ( entrainment and detrainment ) , or to validate the long - term budgets for nwp models and gcms by comparing them with climatology or reanalyses .
the differences between @xmath319 and mse quantities observed in figures [ fig_h_qt_diagrams ] to [ fig_mse_deficit ] can be interpreted in a different way .
if enthalpy is replaced in abscissa by temperature , the enthalpy of figure [ fig_h_qt_diagrams ] is transformed into the psychrometric chart plotted on figure [ fig_t_qt_h_psychro ] , where the isotherms are vertical .
the aim of psychrometric charts is to determine and plot the lines of constant wet - bulb temperature @xmath366 .
it is assumed in n21 or dvm75 that @xmath366 is the temperature attained by a mass of moist air brought to saturation by water evaporating into it , with a latent heat continuously supplied by the wet bulb .
it is thus an isenthalpic process at constant pressure for the whole system of the moist air plus the wet bulb , with the mass of dry air assumed to be constant .
the wmo ( 2008 ) psychrometric equation is derived in dvm75 .
it can be written as @xmath367 the wet - bulb temperature may be defined differently .
the left hand side is approximated by @xmath368 in n21 .
the definition ( 6.67 ) or ( 6.78 ) published in bohren and albrecht ( 1998 ) corresponds to @xmath369 being replaced by @xmath370 .
psychrometric equations are derived in e94 ( 4.6.4 ) and a10 ( 5.41 ) by assuming conservation of @xmath371 expressed per unit mass of dry air .
they correspond to @xmath372 being replaced by @xmath373 .
all these terms depending on mixing ratios are expected to be small in comparison with @xmath12 in ( [ def_psychro_dvm75 ] ) .
the wet - bulb temperature may be derived from the second law and the conservation of the pseudo - adiabatic potential temperature @xmath31 .
it is the temperature attained by a parcel of fluid brought adiabatically to saturation by upward displacement and at a lower pressure , then carried adiabatically back to the original pressure , with water assumed to be continuously supplied to maintain saturation .
hpa and for @xmath374 . vertical isotherms and lines of constant @xmath366 ( thin solid lines ) are for @xmath375c , @xmath376c and @xmath377c in a non - saturation region ( where @xmath338 ) .
the two versions derived from dvm75 and n21 can be compared using lines of constant @xmath31 .
bold lines for constant enthalpy ( @xmath5 ) or mse quantities ( mse@xmath160 , mse@xmath163 and @xmath166 ) decrease with @xmath8
. lines of constant relative humidity curve upwards ( from @xmath378% to @xmath326% ) , the saturation curve corresponding to @xmath257% .
[ fig_t_qt_h_psychro ] _ ] psychrometric lines of equal @xmath366 and lines of constant @xmath31 are plotted on figure [ fig_t_qt_h_psychro ] .
the figure shows that the versions of n21 and dvm75 are very close to each other .
this is a confirmation that @xmath368 is the leading - order term in ( [ def_psychro_dvm75 ] ) .
larger differences are observed with lines of constant @xmath31 , in particular for @xmath379 % .
the wet - bulb temperatures computed from the methods described in n21 or dvm75 and the one using @xmath31 are thus different . it is common practice to plot isenthalpic lines on the same psychrometric charts
. these lines of constant specific enthalpy expressed per unit mass of dry or moist air correspond to open systems , since values of @xmath380 vary from @xmath0 to the saturating value @xmath381 .
we thus have an opportunity to compare the properties of @xmath5 given by ( [ def_h_bis ] ) with those from mse formulations , because mse quantities are derived for closed systems only and with arbitrary assumptions like @xmath382 at @xmath70 c , whereas the definition of @xmath5 is valid for both closed and open systems _ a priori _ and is derived from a third law .
this study is the continuation of the comparison of @xmath5 and tmse@xmath163 in the figure [ fig_h_qt_diagrams ] .
all isenthalpic lines are oriented in a downward direction on figure [ fig_t_qt_h_psychro ] .
they are almost parallel to psychrometric lines of constant values of @xmath366 , but do not coincide with them .
this probably invalidates the possibility that psychrometric lines might be computed as constant values of @xmath5 , although additional measurements are required to determine which curve is relevant for low rh .
moreover , systematic differences exist on figure [ fig_t_qt_h_psychro ] : the curves of constant @xmath5 are located between the curves of constant mse@xmath160 and mse@xmath163 , whereas the curves of constant @xmath166 are more different and are located above the others . the closed system assumption and the use of arbitrary definitions for the reference enthalpies explain the differences observed between the curves plotted with @xmath5 and with other mse quantities .
in the same way that moist entropy is expressed in m11 in terms of @xmath383 up to a constant entropy reference value , moist - air thermal enthalpy is expressed in terms of @xmath384 up to a constant enthalpy reference value .
the potential temperature @xmath90 and the enthalpy temperature @xmath174 are thus equivalent to the specific moist - air entropy ( @xmath10 ) and thermal enthalpy ( @xmath5 ) .
thermal enthalpies are generated by variations of @xmath33 with @xmath8 corresponding to progressive excitations of the translational , rotational and vibrational states of the molecules , and by possible changes of phase represented by latent heats ( with negligible impact of changes of pressure ) .
the original feature is that both @xmath174 and @xmath90 depend on standard values of enthalpies and entropies , via the terms @xmath385 and @xmath386 , respectively . in this paper , standard values
are determined with an error of a few percent for the cryogenic properties of the dry - air and water - components of moist air .
different choices for these standard values would generate other values for @xmath174 and @xmath90 .
the use of cryogenic values for the specific heats of solid and liquid phases , in addition to latent heats when changes of phases occur , eliminate the issue of referring to the perfect gas law when the temperature is too low .
the important feature in the definition of @xmath5 and @xmath10 is that the specific contents of the moist - air components multiply several constant terms which depend on the reference values @xmath13 or @xmath94 .
this leads naturally to additional varying terms depending on @xmath21 .
it is for this reason that the standard values @xmath240 , @xmath241 , @xmath98 and @xmath97 must determined for varying @xmath21 . this explains why the third law versions for @xmath5 and @xmath10 are different from the ones obtained in dvm75 , e94 or p10 , where only quantities per unit of dry - air are considered and where hypotheses like @xmath387 , @xmath388 , @xmath389 or @xmath382 are made at a temperature different from @xmath0 k. this suggests that the third law of thermodynamics may have an important and somewhat surprising physical meaning when moist - air thermodynamics is considered
. the constant volume or constant dry - air point of view may be useful for laboratory experiments , but it is not relevant to isolate the dry - air part from a moving open moist - air parcel with multiple components .
the specific value approach is more suitable for applications to a barycentric view and to equations of compressible fluids .
this may explain some of the differences between the present third law results and previous ones .
it has been shown that the enthalpy temperature is expressed in terms of the quantity @xmath228 , which depends on the difference between the standard values of dry - air and water - vapour enthalpies ( in the same way as the moist entropy depends , via @xmath386 , on the difference between the standard values of dry - air and of water - vapour entropies ) .
it should be noted that @xmath228 is a constant term independent of the reference temperature @xmath13 .
this means that the same property observed by entropy in m11 is also valid for the thermal enthalpy , i.e. they are both independent of the choice made for the reference temperature and pressure @xmath13 and @xmath94 .
since they are well - defined local quantities , it becomes possible to use specific values of thermal enthalpy or entropy to compute local budgets , integral or mean values of @xmath5 and @xmath10 , either in space or time .
a more surprising result concerns the assumption that generalized enthalpy might be roughly represented by mse@xmath163 or mse@xmath160 .
it is shown that it is serendipitous since : i ) the numerical values of standard enthalpies of dry air and liquid water are close to each other ; and ii ) @xmath274 is a large dominant term of @xmath228 .
it is shown in m11 that no such coincidence exists for the entropy , thus explaining the large impact of the new term @xmath386 introduced for the entropy computations . however , even if mse@xmath163 or mse@xmath160 might be considered as relevant approximations of @xmath319 ( at least for non - saturated conditions ) , they do not follow the same internal liquid - ice symmetry as in the case of the specific moist - air enthalpy formulation .
the enthalpy diagram illustrates the possibility of representing all the thermodynamic properties of moist air as a function of @xmath5 , @xmath390 and @xmath21 alone , as in pauluis and schumacher ( 2010 ) who suggest using the set ( @xmath10 , @xmath390 , @xmath21 ) to compute any of the basic variables ( @xmath8 , @xmath390 , @xmath81 , @xmath82 , @xmath199 ) .
it is thus possible to use the first- or second - law set of variables ( @xmath5 , @xmath390 , @xmath21 ) or ( @xmath10 , @xmath390 , @xmath21 ) as prognostic variables .
this is true for les models where each grid cell represents a homogeneous parcel of moist air ( as long as the pressure is known , in spite of non - hydrostatic processes ) .
this result is no longer valid for a parcel subject to sub - grid variability .
all properties of a parcel in nwp models or gcms are understood as weighted sums of properties observed for a given fraction of unsaturated air plus the other part from saturated air ( with weighting factors depending on cloud fractions ) .
however , the method of representing moist - air properties by ( @xmath10 , @xmath390 , @xmath21 ) or by ( @xmath5 , @xmath390 , @xmath21 ) may be used even in nwp models or gcms , provided that : it is i ) applied separately for each of the unsaturated and saturated fractions of moist air , and ii ) the values of cloud fraction can be determined by some other processes .
it thus becomes possible to compute and study the tendencies of @xmath5 or @xmath10 generated by various processes ( dynamics , radiation , convection , turbulence ) and to compare the results with reanalysis outputs .
an alternative method could consist of evaluating the finite differences @xmath391 at each point , or @xmath392 , with the specific thermal enthalpy or entropy computed with zero values at @xmath0 k. another application concerns pure isenthalpic processes , as plotted on psychrometric charts .
it is shown in this paper that , depending on how the specific enthalpy is defined , the isenthalpic lines are different .
since there is only one kind of isenthalpic process in the real world , it might be valuable to compare the new thermal enthalpy versions involving real processes and/or observations .
the differences between mse quantities and @xmath319 are analysed for several cumulus and stratocumulus vertical profiles .
mse@xmath163 is greater , and mse@xmath160 smaller , than the generalized enthalpy @xmath319 in moist conditions .
these systematic differences of more than @xmath176% may have significant physical implications for the accurate determination of the moist thermal enthalpy within clouds .
they may modify results from analyses of the impact of drying or moistening processes on @xmath5 , and therefore on local temperature .
this confirms the striking results obtained when referring to the enthalpy diagrams , and suggests that the formulation of thermal enthalpy ( [ def_h_bis ] ) should be used to compute moist enthalpy or energy budgets accurately .
this may be important for the validation of nwp models and/or gcms , when the budget of thermal enthalpy simulated by these models is compared with those obtained from climatology or reanalyses .
all of the above suggests new possible applications for the third law formulation of entropy and thermal enthalpy .
it may be used to study dry - air entrainment and moist - air detrainment occurring along cloud edges , as well as the conservative properties used when parameterizing shallow or deep - convection . for the latter phenomena ,
mse quantities should be replaced by the specific generalized enthalpy @xmath319 .
the author is most grateful to jean - franois geleyn and maarten ambaum for stimulating discussions .
he would like to thank the anonymous referees for their constructive comments , which helped to improve the manuscript .
thanks to s. becker and y. tourre for the improvements of the written english .
s. r. de roode and q. wang kindly provided the validation data from nasa flights during the fire i experiment .
= = @xmath12 specific heat of dry air ( @xmath393 j k@xmath217
kg@xmath217 ) + @xmath17 spec .
heat of water vapour ( @xmath394 j k@xmath217 kg@xmath217 ) + @xmath395 spec .
heat of liquid water ( @xmath396 j k@xmath217 kg@xmath217 ) + @xmath397 spec .
heat of ice ( @xmath398 j k@xmath217 kg@xmath217 ) + @xmath9 specific heat at constant pressure for moist air ( @xmath399 ) + @xmath400 @xmath401 + @xmath402 @xmath403 + @xmath404 @xmath405 + @xmath406 @xmath407 + @xmath408 @xmath409 + @xmath410 @xmath411 + @xmath412 water - vapour partial pressure + @xmath413 partial saturating pressure over liquid water + @xmath95 water - vapour reference partial pressure : @xmath414 hpa + @xmath415 internal energy : @xmath416 + @xmath417 magnitude of earth s gravity : @xmath418 m@xmath419 s@xmath420 + @xmath5 specific enthalpy + @xmath210 reference enthalpy of dry air at @xmath13 + @xmath421 reference enthalpy of water vapour at @xmath13 + @xmath240 standard specific enthalpy of dry air at @xmath219 ( @xmath272 kj kg@xmath217 ) + @xmath241 standard specific enthalpy of water vapour at @xmath219 ( @xmath422 kj
kg@xmath217 ) + @xmath423 standard specific enthalpy of liquid water at @xmath219 ( @xmath424 kj kg@xmath217 ) + @xmath425 standard specific enthalpy of ice water at @xmath219 ( @xmath250 kj kg@xmath217 ) + @xmath166 a specific moist static energy ( a10 ) + @xmath426 @xmath427 / c_{pd } \approx 5.87 $ ] + @xmath238 @xmath428 : latent heat of vaporization + @xmath274 @xmath429 @xmath430 j kg@xmath217 at @xmath219 + @xmath431 @xmath432 : latent heat of fusion + @xmath433 @xmath434 @xmath430 j kg@xmath217 at @xmath219 + @xmath239 @xmath435 : latent heat of sublimation + @xmath436 @xmath437 @xmath430 j kg@xmath217 at @xmath219 + @xmath171 a mass of moist air + @xmath390 @xmath438 : local value of pressure + @xmath94 @xmath439 : reference pressure ( @xmath440 ) + @xmath441 local dry - air partial pressure + @xmath442 reference dry - air partial pressure ( @xmath443 ) + @xmath444 @xmath445 hpa : conventional pressure + @xmath446 @xmath447 : specific dry air content + @xmath448 @xmath449 : specific water vapour content + @xmath450 @xmath451 : specific liquid water content + @xmath452 @xmath453 : specific ice content content + @xmath454 @xmath455 : total specific water content + @xmath456 specific for saturating water vapour content + @xmath457 @xmath458 : mixing ratio for water vapour + @xmath459 @xmath460 : mixing ratio for liquid water + @xmath461 @xmath462 : mixing ratio for ice + @xmath463 reference mixing ratio for water species : @xmath464 and @xmath465 g kg@xmath217 + @xmath466 mixing ratio for saturating water vapour + @xmath467 @xmath468 : mixing ratio for total water + @xmath469 specific mass of dry air + @xmath470 specific mass of water vapour + @xmath471 specific mass of liquid water + @xmath472 specific mass of ice + @xmath473 specific mass of moist air @xmath474 + @xmath475 relative humidity ( @xmath476 for liquid water ) + @xmath477 water - vapour gas constant = ( @xmath478 j k@xmath217
kg@xmath217 ) @xmath479 dry - air gas constant ( @xmath480 j k@xmath217
kg@xmath217 ) + @xmath477 water - vapour gas constant
( @xmath481 j k@xmath217
kg@xmath217 ) + @xmath482 @xmath483 : gas constant for moist air + @xmath10 specific entropy + @xmath484 reference values for the entropy of dry air , + @xmath485 reference values for the entropy of water vapour , + @xmath98 standard specific entropy of dry air at @xmath219 and @xmath444 : @xmath486 j k@xmath217
kg@xmath217 + @xmath97 standard specific entropy of water vapour at @xmath219 and @xmath444 : @xmath487 j k@xmath217 kg@xmath217 + @xmath488 standard specific entropy of liquid water at @xmath219 and @xmath444 : @xmath489 j k@xmath217 kg@xmath217 + @xmath490 standard specific entropy of solid water at @xmath219 and @xmath444 : @xmath491 j k@xmath217 kg@xmath217 + @xmath492 entropies in p10 ( @xmath124 is their weighted sum ) + @xmath8 local temperature + @xmath13 reference temperature ( @xmath493 ) + @xmath366 isenthalpic wet - bulb temperature + @xmath198 ice - liquid water temperature + @xmath494 zero celsius temperature ( @xmath259 k ) + @xmath228 a constant temperature ( @xmath495 k ) + @xmath496 @xmath497 : potential temperature + @xmath498 equivalent potential temperature + @xmath499 liquid - water potential temperature + @xmath500 pseudo - adiabatic wet - bulb potential temperature + @xmath501 moist entropy potential temperature ( m11 ) + @xmath196 @xmath502 / ( c_{pd}\:t_r)$ ] , @xmath503 + @xmath504 , @xmath505 individual and barycentric mean velocities + @xmath50 individual barycentric diffusion flux + @xmath152 vertical component of the velocity + @xmath506 3d - gradient operator + @xmath507 gibbs function ( @xmath508 ) + @xmath44 @xmath509 for species @xmath37 ( b82 ) + @xmath510 , @xmath511 external and internal changes + @xmath58 diabatic heating rate + @xmath83 , @xmath84 rate of change of @xmath81 and @xmath82 + @xmath78 irreversible source of entropy + @xmath256 gravitational potential energy ( @xmath512 ) + gcm general circulation model + nwp numerical weather prediction + mse moist static energy + les large eddy simulation + tmse thermal moist static energy .values of the specific heat for o@xmath32 are given as a function of the absolute temperature .
the first table corresponds to the solid-@xmath1 form .
units are k for @xmath8 and j k@xmath217 mol@xmath217 for @xmath9 , to be divided by @xmath513 kg mol@xmath217 to obtain units of j k@xmath217 kg@xmath217 .
data were obtained up to @xmath340 k from table - i of fh69 , with some interpolation performed from their figures 1 and 2 for the range @xmath514 to @xmath515 k. the second table corresponds to the solid-@xmath516 form .
data were obtained from table i of fh69 for the range @xmath517 to @xmath518 k , with interpolation performed from their figures 1 and 2 for the range @xmath519 to @xmath520 k. the third table corresponds to the solid-@xmath408 form ( @xmath518 to @xmath521
k ) and the fourth table to the liquid form ( @xmath521 to @xmath522 k ) .
data were obtained from fh69 , with linear interpolation from their figure 1 .
the fifth table corresponds to the vapour form at @xmath523 hpa ( above @xmath522 k ) and with @xmath9 directly expressed in j k@xmath217 kg@xmath217 .
data were obtained from jacobsen _ et al . _
( 1997 , table 5.79 ) .
[ table_cp_o2 ] [ cols="^,^,^,^,^,^",options="header " , ] @xmath524 chase mw jr .
journal of physics and chemical reference data .
monograph no.9 .
nist - janaf thermochemical tables .
_ american chemical society and american institute of physics .
_ vol.1 , pp .
vol.2 , pp . 959 - 1951
( c98 ) .
@xmath524 derbyshire sh , beau i , bechtold p , grandpeix j - y , piriou j - m , redelsperger j - l , soares pmm .
sensitivity of moist convection to environmental humidity
. _ q. j. r. meteorol .
_ * 130 * ( 604 ) : 178196 .
@xmath524 emanuel ka .
tropical cycle energetics and structure .
chapter 8 in `` atmospheric turbulence and mesoscale meteorology . scientific research inspired by doug lilly . '' edited by e.e .
fedorovich , r. rotuno and b. stevens .
p:165 - 192 .
cambridge university press .
@xmath524 gerard l , piriou j - f , brozkova r , geleyn j - f , banciu d. 2009 .
cloud and precipitation parameterezization in a meso - gamma - scale operational weather prediction model . _ mon .
* 137 * ( 11 ) : 39603977 .
@xmath524 glansdorff p , prigogine i. 1971 .
structure stabilit et fluctuations .
_ masson ed_. paris ( also available in english : thermodynamic theory of structure , stability and fluctuations , wiley - interscience ) .
@xmath524 kudryavtsev in , nemchenko ke .
lattice dynamics and heat capacity of solid nitrogen .
proceeding of the 10th international conference on phonon scattering in condensed matter .
august 12 - 17 , 2001 .
dartmouth , usa .
@xmath524 marquet p. 2011 .
definition of a moist entropic potential temperature .
application to fire - i data flights .
_ q. j. r. meteorol .
* 137 * ( 656 ) : 768791 ( m11 ) .
arxiv:1401.1097 [ ao - ph ] @xmath524 marquet p , geleyn j - f .
2013 . on a general definition of the squared brunt - visl frequency associated with the specific moist entropy potential temperature .
_ q. j. r. meteorol .
* 139 * ( 670 ) : 85100 .
arxiv:1401.2379 [ ao - ph ] @xmath524 marquet p. 2013 . on the definition of a moist - air potential vorticity .
_ q. j. r. meteorol .
_ accepted in april , 2013 .
early view stage + http://arxiv.org/abs/1401.2006 .
arxiv:1401.2006 [ ao - ph ] @xmath524 siebesma ap , bretherton cs , brown a , chlond a , cuxart j , duynkerke pg , jiang h , khairoutdinov m , lewellen d , moeng ch , snchez e , stevens b , stevens de .
2003 . a large eddy simulation intercomparison study of shallow cumulus convection .
_ j. atmos .
* 60 * ( 10 ) : 18701891 .
|
the specific thermal enthalpy of a moist - air parcel is defined analytically following a method in which specific moist entropy is derived from the third law of thermodynamics .
specific thermal enthalpy is computed by integrating specific heat content with respect to absolute temperature and including the impacts of various latent heats ( i.e. , solid condensation , sublimation , melting , and evaporation ) .
it is assumed that thermal enthalpies can be set to zero at @xmath0 k for the solid form of the main chemically inactive components of the atmosphere ( solid-@xmath1 oxygen and nitrogen , hexagonal ice ) .
the moist thermal enthalpy is compared to already existing formulations of moist static energy ( mse ) .
it is shown that the differences between thermal enthalpy and the thermal part of mse may be quite large .
this prevents the use of mse to evaluate the enthalpy budget of a moist atmosphere accurately , a situation that is particularly true when dry - air and cloud parcels mix because of entrainment / detrainment processes along the edges of cloud .
other differences are observed when mse or moist - air thermal enthalpy is plotted on a psychrometric diagram or when vertical profiles of surface deficit are plotted .
_ paper accepted for publication in the .
_ + _ last revised version in january 2014 . _
+ _ : [email protected]_
|
negatively curved surfaces have been studied experimentally including lattice dislocations of solid - state crystals with non - euclidean properties , e.g. , in various magnetic nanostructures @xcite , or in materials showing a conical geometry @xcite . on the other hand ,
theoretical interest has been focused on the geometry of anti de sitter ( ads ) space @xcite and the complex ( e.g. neural ) networks @xcite .
the nontrivial boundary structure of finite hyperbolic spaces plays an essential role in the analysis of ads space by means of entanglement entropy @xcite .
the mutual relations among condensed - matter physics , the general theory of relativity , and the conformal field theory ( cft ) enrich the interdisciplinary research , such as ads - cft correspondence @xcite . in 2007 , we proposed a way to generalize the corner transfer matrix renormalization group ( ctmrg ) algorithm to a simple hyperbolic surface @xcite .
the current study extends our recent generalization to arbitrary regular hyperbolic geometry @xcite , and we focus on the complexity of the boundary effects of hyperbolic lattice geometry . because no satisfactory analytic studies are available yet , we analyze multistate spin models on hyperbolic lattices numerically . in particular , we focus on phase transitions , which require special treatment because of the non - negligible boundary effect , as they have been suppressed thus far .
the ctmrg algorithm enables us to calculate the free energy accurately , including all thermodynamic functions in thermal equilibrium .
the free energy approach of the current work has been chosen because it incorporates all the non - negligible boundary effects of the models on hyperbolic lattices .
the current study is intended to provide complementary and missing information with respect to the boundary effects performed by monte carlo ( mc ) simulations @xcite .
hyperbolic lattice geometry is known for exhibiting strong boundary effects , which prevent the mc simulations from performing sufficiently precise calculations to classify the phase transitions by means of the free energy .
the reason lies in the large fluctuations coming from the rich boundary structure : the boundary size grows exponentially , as the diameter ( the shortest distance from the lattice center to the boundary ) increases linearly .
_ studied the ising model via mc simulations , and they determined the phase transition by gradually contracting a couple of the boundary layers @xcite .
the aim of this paper is to perform a thorough numerical analysis of the multistate spin models by ctmrg with respect to the boundary layers , which are gradually contracted from the entire lattice in the thermodynamic limit .
the paper is organized as follows : in sec .
ii we briefly classify an infinite set of hyperbolic lattices by a pair of two integers @xmath0 .
we consider the multistate potts model on the negatively curved lattice surfaces .
iii a is devoted to the numerical ctmrg analysis of the first- and second - order phase transitions on square and bethe lattices .
these two lattices are chosen because they are exactly solvable and can serve as the benchmarks for the future studies .
this section can be skipped if the reader is familiar with numerical analysis of phase transitions using thermodynamic functions .
section iii b contains the core of this work , where the dependence of the free energy on the gradually contracted layers is shown .
we generalize our study to classify @xmath1-state potts models on regular hyperbolic lattice geometries .
a concise study of the phase transition deeply inside the infinite bulk is given when contracting the infinite number of boundary layers .
a phase - coexistence region is specified as a temperature interval , within which the first - order phase transition can be controlled by varying the magnetic field on the boundary layer .
section iv is devoted to discussions and concluding remarks .
consider a regular two - dimensional lattice ( a curved surface ) formed by the tessellation of congruent polygons with @xmath2 sides ( referred to as the @xmath2-gons ) , and the coordination number @xmath3 is fixed for the entire lattice .
such a regular lattice can be described by a pair of integers , @xmath0 , known as the schlfli symbol @xcite .
depending on the choice of the two integers @xmath4 and @xmath5 , the schlfli symbol can describe either euclidean ( flat ) or curved ( spherical or hyperbolic ) lattice geometries .
if the condition @xmath6 holds , only the following three regular euclidean two - dimensional lattices exist : the triangular @xmath7 lattice , the square @xmath8 lattice , and the honeycomb @xmath9 lattice .
the inequality @xmath10 describes five spherically curved lattices of finite size .
however , if @xmath11 , an infinite set of hyperbolic lattice geometries can be constructed by tessellation .
each hyperbolic lattice has a particular constant and negative gaussian curvature , provided that the size of any @xmath2-gon side is identical @xcite .
although an arbitrary hyperbolic lattice @xmath0 of infinite size forms a negatively curved surface , none of the lattices can be embedded in the three- or finite - dimensional space .
hence , the hausdorff dimension is infinite for an arbitrary hyperbolic lattice @xmath0 , provided that the thermodynamic limit is taken . to visualize the hyperbolic lattice @xmath0 ,
it is useful to project each lattice onto a unitary circle ; this mapping is known as the poincar disk representation @xcite . in such a representation , the congruent @xmath2-gons ( all of the polygons have an identical size and shape on the entire lattice ) are deformed if displayed graphically on the disk .
the size of the @xmath2-gons decreases toward the circumference of the disk ; the circumference represents the lattice boundary ( in the infinity ) . in the poincar disk representation .
the entire hyperbolic lattice is composed of @xmath12 concentric layers , where @xmath12 is referred to as the lattice size ( the radius ) , which enumerates the layers from the center outward .
the inner sublattice , the bulk , has @xmath13 sites with the diameter @xmath14 of the layers .
the outer layers , the boundary , are enumerated by @xmath15 and consists of @xmath16 sites.,scaledwidth=49.5% ] at first we specify the size of the lattice in terms of the radius @xmath12 , which is well - visible if the hyperbolic lattice is graphically represented in the poincar disk .
let the integer @xmath12 denotes the actual radius of the lattice beginning in a central lattice site .
an example of a hyperbolic lattice geometry @xmath17 is shown in fig .
[ fig1 ] ( on the left ) , where the smallest lattice radius with @xmath18 contains four regular pentagons . if @xmath19 , the total number of the pentagons is @xmath20 , etc . to enhance the @xmath12-dependence , i.e. , the layer structure , the inner layers of the lattice are depicted by the distinct intensity of the gray color .
now it is obvious how the total number of the @xmath2-gons grows exponentially with respect to the linear increase of the radius @xmath12 ( generalization to any hyperbolic lattice @xmath0 is straightforward ) . to characterize the boundary layers as a hyperbolic lattice
, we ascribe the boundary layers to an integer variable @xmath15 , which enumerates the number of the outermost layers , @xmath21 ( provided that @xmath22 ) . whenever we refer to the bulk properties of the ( hyperbolic ) lattice
, one should consider such an innermost sublattice , which has the radius @xmath14 infinite ; or , numerically , @xmath14 is sufficiently large so that all thermodynamic functions ( normalized to the lattice site ) has completely converged .
let @xmath23 be the total number of the lattice vertices ( the sites ) for a given diameter @xmath12 . if @xmath24 , the entire lattice can be thought of as a system composed of two parts : the bulk subsystem containing @xmath14 inner layers and the @xmath15 outer boundary layers .
the total number of the sites is an integer function of @xmath12 and is denoted by @xmath25 , which depends on the lattice geometry .
the inner @xmath14 layers and the outer @xmath15 layers , respectively , consist of @xmath13 and @xmath16 sites , i.e. , @xmath26 , as schematically sketched in fig .
[ fig1 ] ( on the right ) . between the boundary and the bulk layers
is calculated for the three selected hyperbolic lattices @xmath27 , @xmath28 , and @xmath17 in the thermodynamic limit ( @xmath29 ) . the inset shows the detail for the outermost layer when @xmath30.,scaledwidth=49.5% ] if we evaluate the ratio between the boundary sites @xmath16 and the inner bulk sites @xmath13 , we aim to quantify the significance of the boundary sites , particularly , @xmath31 recall that @xmath32 for the euclidean lattice geometries .
however , the ratio @xmath33 is always positive for the hyperbolic lattices , as shown in fig .
it is instructive to stress that @xmath34 for the three selected lattices , which means that the number of the boundary sites is at least twice as large as the inner @xmath35 sites , even in the thermodynamic limit @xmath29 ( the inset ) . or , to be more specific , @xmath36 , @xmath37 , and @xmath38 on the lattices @xmath17 , @xmath28 , and @xmath27 , respectively .
this clearly implies that an enormous number of sites lies on the outermost boundary layer , which significantly influence the inner ( bulk ) properties that we study in this paper .
having defined the lattice geometry @xmath0 , now we consider a relevant multistate spin system on those underlying lattices .
let each lattice vertex ( site ) be a multistate spin variable interacting with the @xmath3 nearest neighboring spins in each @xmath2-gon .
we study the @xmath1-state potts model @xcite with the nearest - neighbor coupling @xmath39 ( preferably the ferromagnetic one , @xmath40 , to avoid frustration for odd @xmath2 ) , whose hamiltonian has the form @xmath41 the kronecker delta @xmath42 acts on the two adjacent ( nearest - neighbor ) @xmath1-state spin variables @xmath43 and @xmath44 positioned on the vertices @xmath45 and @xmath46 . a uniform constant magnetic field @xmath47 acting on every spin variable @xmath43
is imposed , whereas an independent magnetic field @xmath48 is applied on the boundary spins only .
each @xmath1-state spin variable @xmath43 possesses @xmath1 degrees of freedom , @xmath49 , where @xmath50 .
the summation @xmath51 runs over all the pairs of the nearest - neighboring spins on the entire lattice in the first term .
the summation in the second term , @xmath52 , specifies the interaction of the magnetic field @xmath47 with the spin site on the lattice in a preferred spin level @xmath53 ( we selected the zeroth spin level , i.e. , @xmath54 ) .
the third term @xmath55 denotes the interaction of the magnetic field @xmath48 with the @xmath56 spin sites located on the outermost boundary layer . when the number of the spin degrees of freedom is set to be as small as @xmath57 ,
the model is equivalent to the ising model . for the calculation of the thermodynamic functions , we apply the generalized ctmrg algorithm @xcite , which has been proven to be a powerful numerical method for classical multistate spin lattice models @xcite .
the ctmrg is a classical counterpart of the density matrix renormalization group @xcite .
it has been successfully applied to various spin models on euclidean and the hyperbolic lattices @xcite .
the algorithm is an efficient and accurate tool to calculate the partition function @xmath58 , or eventually , the free energy @xmath59 , and the related thermodynamic functions . here ,
@xmath60 and @xmath61 , respectively , are the boltzmann constant and temperature . in the following we use the dimensionless units , and we consider @xmath62 . more specific details of the ctmrg algorithm modified for the hyperbolic lattices are summarized in refs . .
to summarize the ctmrg idea in brief , the partition function is considered in terms of the tensor product of local boltzmann weights @xmath63 of the @xmath2-gonal shape , where the hamiltonian @xmath64 is a cyclic spin chain of interacting spins within the @xmath2-gon for a given @xmath0 lattice geometry .
the algorithm is designed to cover the entire lattice with the identical boltzmann weights so as to satisfy the constant coordination number @xmath3 .
the lattice expands iteratively in the algorithm , starting from the @xmath3 boltzmann weight sharing the central site .
the lattice grows linearly with the diameter @xmath12 within the renormalization group procedure until the renormalized corner transfer tensors converge .
the full convergence is associated with the thermodynamic limit , which is a necessary condition to analyze the phase transitions .
the @xmath1-state potts model is exactly solvable @xcite on the square lattice @xmath8 , resulting in the phase transition temperature @xmath65 for any @xmath50 . in a special case ,
the two - state potts model is equivalent to the ising model , which has an exact solution on the bethe lattices @xcite for arbitrary coordination number @xmath5 , and the phase - transition temperature is @xmath66 $ ] .
in the following , we refer to these exact phase transition temperatures , which we use for comparison with the numerical results .
as we have shown earlier @xcite , the thermodynamic properties of the spin models in the bulk on the hyperbolic lattices @xmath67 are , in fact , numerically indistinguishable from those on the bethe @xmath68 lattices @xcite , regardless of the coordination number @xmath3 ; cf .
therefore , we use the hyperbolic lattice @xmath69 as the bethe lattice @xmath68 .
figure [ fig3 ] shows the lattice structure in the poincar representation of the hyperbolic lattice @xmath27 ( left ) being numerically equivalent to the bethe lattice @xmath70 ( right ) .
the ctmrg algorithm applied to an arbitrary lattice @xmath0 is designed to join @xmath3 corner transfer tensors around a central site @xmath71 ; for better visibility , fig .
[ fig3 ] does not display the spin @xmath71 at the center of the unitary poincar circle for the two lattices @xcite .
having compared the relative error @xmath72 at the phase - transition temperature by the ctmrg algorithm with respect to the analytic result , we obtained high accuracy , @xmath73 for @xmath57 and @xmath69 .
( left ) and @xmath70 ( right ) are numerically indistinguishable if the bulk properties of the spin models are considered . ]
there are two independent calculations in ctmrg to specify the phase transitions .
\(i ) the simplest one is to calculate the expectation value for the spin ( the spontaneous magnetization ) @xmath74 - 1 } { q-1}\ , , \label{mg}\ ] ] where @xmath71 is the @xmath1-state spin variable located in the central site of the lattice and @xmath75 is a reduced density matrix that is the partial configuration sum of the product of @xmath3 corner transfer tensors @xcite .
this approach leads to an accurate determination of the phase transition temperature in the bulk @xcite .
the effects of the lattice boundaries are strongly suppressed in this type of calculation .
therefore , the spontaneous magnetization in eq . has been used as a reliable criterion to determine the phase transition in the bulk .
analogously , the nearest - neighbor ( nn ) correlation energy , @xmath76 can also be evaluated between two neighboring spins in the center of the lattice . here
, the phase transition can be extracted from the singularity ( a diverging peak ) of the specific heat , which is proportional to taking the numerical derivative of @xmath77 with respect to temperature .
\(ii ) the second , and independent way to find the phase transition is to evaluate the free energy of the entire system , and the boundary effects contribute significantly to the free energy .
we have recently derived recursive expressions for the free energy per spin site , @xmath78 for any multistate spin system including the boundary magnetic field @xmath48 .
a detailed analysis of the free energy per site is given in ref . .
thus , the free energy per spin site ( expressed in the thermodynamic limit ) has the form @xmath79 we can directly use eq .
since the partition function @xmath80 can be calculated to a high numerical accuracy by the ctmrg algorithm , and the total number of the spin sites @xmath81 has an analytic expression for any @xmath0 geometry @xcite .
recall , that the numerical derivatives of the free energy with respect to temperature @xmath61 or the magnetic field @xmath47 , yield the correct thermodynamic functions with the singular behavior at the phase transition .
below , we refer to the following thermodynamic functions : the normalized entropy per spin , @xmath82 the internal energy , @xmath83 and the specific heat , @xmath84 the normalization factor @xmath85 in eq . is chosen to restrict the entropy to the interval @xmath86 ( note that @xmath87 at @xmath88 and @xmath89 at @xmath90 in this work ) . for tutorial purposes , we select the @xmath91-state potts model to represent a second - order ( continuous ) phase transition , and the @xmath92-state potts model , which is known for exhibiting a first - order ( discontinuous ) phase transition on the square lattice @xmath8 . the @xmath1-state potts models results in the second - order phase transitions if @xmath93 , whereas the first - order phase transitions arise if @xmath94 on the two - dimensional euclidean lattices @xcite .
we first concentrate on the simplest case : no boundary layers are contracted ( @xmath95 ) and no magnetic field is imposed on the boundary layer ( @xmath96 ) , whereas the thermodynamic limit is taken ( @xmath29 ) .
first , we investigate the exactly solvable ising model on the square lattice @xmath97 and the bethe @xmath98 lattice , and we compare the correctness of the phase - transition temperature with the exact one .
the ising model exhibits an order - disorder phase transition at critical temperature @xmath99 on the square lattice and @xmath100 on the bethe lattice .
the four typical thermodynamic functions with respect to temperature for the ising model are shown in fig .
the phase transition temperature @xmath101 is evident from the typical behavior at zero magnetic field @xmath47 .
the dependence of the four functions at the nonzero field , @xmath102 , is also included .
the second - order phase transition temperature is well - visible as the diverging peak in the specific heat [ cf .
eq . ] , and the free energy does not exhibit any non - analytic behavior .
, the free energy per site @xmath103 , the internal energy @xmath104 , and the specific heat @xmath105 calculated on the square lattice @xmath8 ( black circles ) and the bethe lattice @xmath70 ( red triangles ) .
response of the model to the external magnetic fields @xmath106 ( green symbol ` @xmath107 ' ) and @xmath102 ( blue symbol ` @xmath108 ' ) is shown .
the vertical dot - dashed lines help in identifying the correct phase transition temperature @xmath101 including the _ false _ one located at @xmath109.,scaledwidth=49.5% ] since the spontaneous magnetization of the ising model is calculated as the expectation value of the spin in the center of the lattice ( where all the boundary effects are suppressed ) , the phase - transition temperature agrees with the analytic result for both the square and the bethe lattices .
it is located at the temperature @xmath101 , which separates the ordered phase with @xmath110 from the disordered one , where @xmath111 at @xmath106 .
the calculation of the remaining three thermodynamic functions @xmath103 , @xmath104 , and @xmath105 includes the boundaries .
therefore , the ising model on the bethe lattice shows no evident singularity at @xmath112 if the specific heat @xmath113 for @xmath106 is calculated ( cf .
@xmath114 ) . there is , however , a remarkable ( but non - diverging ) maximum in @xmath113 at significantly lower temperature , which we do not associate with a phase transition . in the following ,
we refer to that non - diverging maximum in the specific heat as the _ false _ phase transition , which is located at temperature @xmath109 .
the non - existence of the phase transition on the entire bethe lattice is not surprising , as was pointed out in ref . .
our numerical analysis using the free energy , thus , reflects such a feature on the hyperbolic lattice , and it will be resolved below .
-state potts model : the spontaneous magnetization @xmath115 , the free energy per site @xmath103 , the nearest - neighbor correlation energy @xmath116 , and the normalized entropy per site @xmath117 are plotted with respect to temperature on the square and the bethe lattices for @xmath106 and @xmath102 .
false _ phase transition temperature is located at @xmath118 on the bethe lattice at @xmath106.,scaledwidth=49.5% ] the first - order phase transition is associated with a sudden ( discontinuous ) change of some thermodynamic functions at the phase transition .
specifically , a nonzero latent heat @xmath119 is associated with the discontinuous phase transition at @xmath120 , in particular , @xmath121 .
the nonzero latent heat is the typical signature of the first - order phase transition , which originates in the nonanalytic behavior of the free energy at the phase transition .
there is a typical kink ( crossover ) in the free energy per site , which can be precisely determined by ctmrg after introducing free and fixed boundary conditions on the euclidean lattices @xcite . to visualize this effect
, we plotted a couple of the thermodynamic functions for the 5-state potts model in fig .
[ fig5 ] . to avoid showing a qualitative similarity of the internal energy @xmath104 and the specific heat @xmath105 ( as shown in fig .
[ fig4 ] ) , we replace them by the nn correlation energy @xmath116 and the normalized entropy @xmath117 , respectively .
the evident discontinuity at the phase transition occurs in @xmath122 , @xmath123 , and @xmath124 at @xmath106 [ including the less visible kink at @xmath125 in @xmath126 , for details , cf .
ref . ] .
obviously , the nonzero latent heat is present , and the method correctly captures the first - order phase transition temperature , which has an analytic solution @xcite @xmath127 on @xmath8 . on the other hand ,
the 5-state potts model on the bethe lattice shows different features .
if the boundary effects are neglected , a large discontinuous jump emerges in the magnetization @xmath114 .
there is a jump in the nn correlation energy @xmath128 , which is proportional to a nonzero latent heat , that is the sign of the first - order phase transition in the bulk ) .
however , when analyzing the free energy @xmath129 and the normalized entropy @xmath130 on the entire lattice system with the boundary effects , both the free energy and the entropy , remain continuous ( analytic ) in the entire temperature interval .
again , we have localized the _ false _ phase transition , which is located at @xmath131 . the vertical dot - dashed lines in fig .
[ fig5 ] serve as guides for the eye to denote the correct and the _ false _ phase transitions . and the entropy @xmath132 .
the two bottom panels show the entropy @xmath132 versus the nn correlation energy @xmath133 .
the second - order phase transition is manifested by the continuous behavior of the thermodynamic functions , whereas the first - order phase transition opens the nonzero gaps in the thermodynamic functions.,scaledwidth=49.0% ] to emphasize the differences in the thermodynamic properties of the potts models on the square and bethe lattices , we plot mutual functional relations of the thermodynamic functions in fig .
the graphs show the continuous ( @xmath57 ) and the discontinuous ( @xmath134 ) phase transitions from a different viewpoint .
the two top panels depict the functional dependence of the spontaneous magnetization per site @xmath115 on the normalized entropy @xmath117 on the square and the bethe lattices in the thermodynamic limit .
the two bottom graphs show the normalized entropy per site @xmath117 with respect to the nn correlation function @xmath116 .
all the graphs are evaluated for @xmath135 , @xmath106 , and @xmath96 within the temperature interval @xmath136 .
the ising model on the both lattices reproduces the second - order phase transition as the continuous functions , whereas the first - order phase transition ( ascribed to the case @xmath134 ) exhibits the discontinuities in the thermodynamic functions . for the 5-state potts model on the bethe lattice @xmath70 forms a two - dimensional surface of a shell - like object .
each thick or thin surface curves correspond to the particular flows of @xmath137 with respect to the varying temperature @xmath61 for given fixed values @xmath39 and @xmath47 .
the data shown are calculated for @xmath138 in the thermodynamic limit ( @xmath29).,scaledwidth=49.0% ] if one is interested in understanding the global phase structure of various systems , a convex set @xcite is a useful tool for such a demonstration .
the convex set is usually constructed for three typical thermodynamic functions to reveal the details of various types of the phase transitions if plotting their mutual functional dependences . since the convex sets have been constructed neither for the first - order phase transitions nor for the non - euclidean lattices , we plot fig . [ fig7 ] to display the convex set @xmath137 for the 5-state potts model on the bethe lattice .
we chose the nn correlation function @xmath139 , the normalized entropy @xmath140 , and the spontaneous magnetization @xmath141 to construct @xmath137 after having scanned sufficiently wide intervals of the parameters @xmath61 , @xmath39 , and @xmath47 .
in particular , we calculated the thermodynamic functions in the intervals of the temperatures @xmath142 , the coupling constants @xmath143 , and the uniform magnetic field @xmath144 .
the constructed convex set has to be understood as the functional set @xmath145 .
the convex set @xmath137 of the 5-state potts model forms a two - dimensional , ] semi - open , and curved surface with a shell - like structure .
the surfaces of the upper and the lower shells , respectively , correspond to @xmath146 and @xmath147 .
it is worth mentioning that the present convex set differs remarkably from those obtained for the models on the euclidean surfaces .
for instance , if the ising model on the @xmath8 lattice is considered ( not shown ) , the convex set covers a smaller surface region and is identical to that evaluated in ref . .
the full thick red curves ( bordering the front edges ) correspond to the case @xmath148 and @xmath106 when the entire temperature interval is scanned .
the red dashed curves connecting the full curves serve only as guides for the eyes .
the dashed curves denote a discontinuous jump ( it is used to connect the two full thick red lines ) , and the discontinuity can not by reached by any @xmath61 , @xmath39 , and @xmath47 .
the unique jump is , therefore , explicitly shown in the two panels in fig .
[ fig6 ] ( on the right ) .
the discontinuity can be asymptotically approached for small magnetic fields @xmath47 .
the point , at which the full and the two dashed red curves meet is a kind of _ repeller _ point at @xmath149 .
no discontinuity of that kind has been reported yet .
the thin black dashed curves refer to the zero nn spin coupling ( @xmath150 ) ; they divide the top ( bottom ) shell into two regions ( the green one with @xmath40 and the blue one with @xmath151 ) .
the sign of the spontaneous magnetization @xmath152 coincides with the sign of the magnetic field @xmath47 .
each thin curve on the surface of @xmath137 is parameterized by the temperature @xmath61 for the fixed values @xmath39 and @xmath47 .
when restricting to the upper shell only , there are three important _ attractors _ ( fixed points ) in the graph : @xmath153 , @xmath154 , and @xmath155 .
the attractor @xmath153 is connected either with @xmath154 or @xmath155 while varying temperature @xmath156 . the attractor @xmath153 is positioned at @xmath157 at @xmath90 , and it can be reached for arbitrary @xmath39 and @xmath47 . the two distinct attractors @xmath158 and @xmath159 can be accessed at @xmath88 : the first one at @xmath160 and the second one at @xmath161 ( not shown due to strong frustration effects caused by the competing antiferromagnetic interaction @xmath151 and the uniform magnetic field , @xmath162 ; the numerical data are not reliable if the frustration becomes strong , i.e. if @xmath163 ) . the two attractors @xmath158 and @xmath159 located at @xmath88 , which is equivalent to @xmath164 on the graph , are mutually inaccessible by varying the temperature @xmath61 .
this is due to a bifurcation point @xmath165 that separates them ( cf . fig .
[ fig7 ] ) .
the bifurcation point is located at @xmath166 on the upper shell .
the top and the bottom shells do not show a mirror symmetry with respect to the plane at @xmath167 .
the bottom shell surface ( for negative @xmath152 ) is less shallow if compared to the upper one .
this is due to the potts model spin levels , for which the spontaneous magnetization of the 5-state potts model exhibits five - fold degeneracy of the free energy below the phase transition .
the free energy minima at @xmath168 are associated with @xmath134 projections of the spontaneous magnetization in eq .
being polarized at @xmath88 either to the spin level with the magnetization @xmath169 or to the four identical levels with @xmath170 ( cf . ref . ) , provided that the spontaneous symmetry - breaking has occurred in the thermodynamic limit [ because the magnetic field @xmath47 is in effect up to the spin level @xmath171 , the term @xmath172 is used in hamiltonian ] . apart from the atypical discontinuous region denoted by the thick dashed red curves , there is another interesting feature : both the top and the bottom shells asymptotically reach the zero entropy limit @xmath173 as @xmath174 for the entire spectrum of @xmath133 , which is not accessible on the euclidean lattice geometry @xcite , where the entropy of the bifurcation point is nonzero ( i.e. , @xmath175 for the ising model on @xmath8 ) .
as we have mentioned earlier , the boundary of the hyperbolic lattices affect the bulk properties , such as the suppression of the phase transition . at the same time
, a _ false _ phase transition appears , which has an obvious origin in the rich boundary structure .
therefore , we propose a thermodynamic function for any underlying geometry @xmath0 @xmath176 where @xmath78 is the free energy per site [ cf .
. ] of the entire system @xmath12 in the thermodynamic limit and @xmath177 is the free energy of the @xmath15 outermost boundary layers . to avoid confusions
, we stress that @xmath178 is not identical to an isolated free energy of the @xmath14 layers .
instead , it is meant to describe a thermodynamic function , that represents _ bulk excess _ free energy , @xmath178 .
it contains all the contributions coming from the @xmath15 contracted boundary layers ( via normalization factors within the bulk , which depends recursively on each of the @xmath15 boundary factors @xcite ) . such a specific definition of the _ bulk excess _
free energy in eq . has been chosen as the only numerically feasible quantity , that can determine the correct phase transition , including the boundary effects .
( for completeness , @xmath179 if @xmath95 only . ) first , the _ bulk excess _
free energy is tested in two asymptotic regimes : @xmath95 and @xmath180 .
we focus on the phase transition analysis in the ising model ( @xmath57 ) on the bethe lattice @xmath70 .
if @xmath180 , we consider the case when both the number of bulk layers , @xmath14 , and the number of boundary layers , @xmath15 , are infinite . to achieve this , one can , e.g. , take @xmath181 and perform the thermodynamic limit @xmath29 .
whenever we refer to the two asymptotic cases ( @xmath29 and @xmath180 ) , we mean sufficiently large ( finite ) values @xmath12 and @xmath15 , for which any further increase of either of them has not affected the thermodynamic functions in eqs . - at all .
since these thermodynamic functions are normalized per spin site , they always have to converge completely . typically , we set @xmath182 for @xmath183 ( which is the case of @xmath184 ) .
we have checked that the numerical results remain identical if @xmath15 is chosen arbitrarily , regardless of the order in which the two limits @xmath15 and @xmath12 are taken . , regardless of @xmath48 .
the two cases are shown : ( 1 ) @xmath95 , @xmath29 ( the circles ) , i.e. no contracted boundary layers , and ( 2 ) @xmath180 , @xmath29 ( the triangles ) , i.e. infinitely many boundary layers are contracted while the inner lattice size remains infinite , too .
the two dot - dashed vertical lines denote the correct and the _ false _ phase transition temperatures.,scaledwidth=49.5% ] the upper graph of fig .
[ fig8 ] shows the _ bulk excess _
free energy per site , @xmath185 , for the ising model on the bethe lattice when @xmath95 and @xmath180 .
there is an evident singularity in the _ bulk excess _
free energy in the correct phase transition , @xmath186 .
the lower graph shows the second derivative of the _ bulk excess _
free energy with respect to temperature , @xmath187 for both @xmath95 and @xmath180 , which we refer to as the ` specific heat ' in the following ( where the term `` specific heat '' is true only if @xmath95 ) .
the singularity of the ` specific heat ' for @xmath180 occurs exactly at the phase - transition temperature @xmath188 , as is analytically derived in the bulk of the bethe lattice @xcite .
the singularity is identical to that in the spontaneous magnetization in fig .
[ fig4 ] . if an arbitrary boundary magnetic field @xmath48 is imposed on the boundary layer , both the phase transition temperature @xmath112 and the _ bulk excess _
free energy @xmath189 remain unaffected by setting any @xmath190 ( not shown ) .
an additional remark is inevitable : the square lattice results in @xmath191 , regardless of whether we took @xmath95 or @xmath180 in the thermodynamic limit .
moreover , if an arbitrary boundary magnetic field @xmath48 is imposed , all thermodynamic functions in the bulk are unaffected by it .
this is true for any spin model exhibiting a second - order phase transition on euclidean lattices , since the boundaries are negligible [ @xmath192 ( for any @xmath193 ) ] . .
notice the specific region located around the _ false _ phase transition @xmath195 , where the _ bulk excess _
free energy grows , is denoted by the dot - dashed arrows.,scaledwidth=49.0% ] from the previous analysis we came to understand the importance of contracting an infinite number of the outermost layers in order to detect the correct phase transition in the bulk by means of @xmath189 .
now , we analyze _ bulk excess _
free energy if we gradually contract the individual boundary layers ( indexed by @xmath15 ) .
the aim is to reveal how the _ false _ phase transition at @xmath95 gets suppressed , and the correct phase transition appears at @xmath180 . as an instructive example
, the ising model on the bethe lattice is again considered .
figure [ fig9 ] shows the @xmath15-dependence of the _ bulk excess _
free energy with respect to temperature for @xmath196 . for comparison ,
the two limiting cases @xmath95 and @xmath180 are also plotted .
as the number of the contracted boundary layers @xmath15 increases , the _ bulk excess _
free energy for @xmath197 undergoes a specific regime .
we also observe dramatic changes of @xmath185 toward the asymptotics @xmath198 ( shown by the thick blue line in the limit @xmath180 in accord with the upper graph of fig . [ fig8 ] ) .
we stress an abrupt ( continuous ) decrease of the _ bulk excess _
free energy for @xmath199 .
the higher @xmath15 is , the closer the abrupt decrease in the asymptotic correct phase transition occurs .
the _ false _ phase transition at @xmath200 ( depicted in figs .
[ fig4 ] and [ fig8 ] ) is denoted by a thin vertical dot - dashed line and is relevant for @xmath201 in this particular case .
it is clear that the _ bulk excess _
free energy remains significantly influenced by the lattice boundaries , and the _ false _ phase transition @xmath195 prevails over the correct phase transition @xmath120 for @xmath202 .
if @xmath30 , there is a surprisingly distinct asymptotic behavior of the _ bulk excess _
free energy @xmath203 at @xmath204 compared to @xmath205 .
this is also manifested in the fact that after contracting the outermost boundary layer , the _ bulk excess _
free energy changes dramatically , despite the fact that the majority of lattice sites were disregarded ( @xmath206 ) . for illustration purposes , the total number of outermost boundary spins @xmath207 is approximately 33 times larger on the lattice geometry @xmath27 than the spins inside , cf .
( recall that the regular bethe lattice @xmath70 gives @xmath208 . )
now , we focus on the interval @xmath201 , where @xmath195 dominates .
the _ bulk excess _
free energy increases locally around @xmath195 , which is not allowed for the free energy in eq . .
there is an inflection point , which is exactly positioned at @xmath209 . to investigate this feature in detail ,
we calculate the ` specific heat ' for each @xmath15 ( @xmath201 ) separately , as shown in fig .
[ fig10 ] . after the boundary layers @xmath210 are gradually contracted .
the inset shows the maximum for @xmath95 and @xmath30.,scaledwidth=49.0% ] let us point out that the ` specific heat ' for @xmath30 exhibits a maximum at the identical @xmath200 , as plotted in the inset of fig .
[ fig10 ] .
as @xmath15 consequently increases ( @xmath211 ) , there is no maximum anymore .
however , the crossover in @xmath212 is a testament to the uniqueness of @xmath200 [ notice the inflection point in @xmath213 which causes the crossover ] . for @xmath214 , the _ false _ phase transition is completely suppressed , and the abrupt decrease of @xmath215 accompanies the singularity , which converges to correct phase transition in the bulk , @xmath216 .
we reached the complete numerical convergence of the _ bulk excess _
free energy after contracting @xmath217 layers .
an analogous dependence of the _ bulk excess _
free energy @xmath178 on @xmath15 was confirmed numerically on different lattice geometries for the ising model .
figure [ fig11 ] depicts @xmath178 for the ising model on the several representative lattice geometries in the thermodynamic limit when @xmath180 . imposing @xmath162 on the ising model on various hyperbolic @xmath0 shifts the _ bulk excess _
free energy as in fig .
the profile of the _ bulk excess _
free energy at @xmath180 does not depend on @xmath48 when the ising model is considered , provided that @xmath106 and @xmath29 . at sufficiently high temperatures , @xmath218 ,
the slopes of @xmath178 became identical for any @xmath193 .
in other words , if we evaluate the normalized entropy per site in eq . , then @xmath219 for @xmath90 . this is a numerical confirmation that the entropy of the @xmath1-state spin models has to saturate at @xmath220 at high temperatures regardless of @xmath15 and the lattice geometry @xcite . when taking both the limits @xmath29 and @xmath180 for @xmath106 .
the _ bulk excess _
free energy remains unchanged for arbitrarily chosen @xmath221.,scaledwidth=49.0% ] the @xmath1-state potts model with @xmath222 exhibits a discontinuous ( first - order ) phase transition on hyperbolic lattices @xmath0 .
we have chosen the @xmath92-state potts model on the bethe lattice as a typical example .
discontinuous phase transitions characteristically exhibit a small temperature region , where two ( or more ) phases can coexist .
for instance , a simple order - disorder phase transition occurs for the @xmath1-state potts models , where the discontinuous phase transition @xmath120 can be located .
in such a case , the free energy exhibits a kink at @xmath120 on the euclidean lattices @xcite .
the phase transition is not critical since the correlation length does not diverge .
the kink in the free energy means that taking its first derivative with respect to temperature results in a discontinuous function , a nonzero gap in the internal energy , at the phase transition .
the gap @xmath223 is proportional to the latent heat , which is a clear sign of the first - order phase transition .
we do not study the @xmath15 dependence of the _ bulk excess _
free energy for the 5-state potts model , as it is analogous to the 2-state potts model . instead , we focus on the special features of the first - order phase transition on the hyperbolic geometries .
if a model shows a discontinuous phase transition on the hyperbolic lattices , no kink in the free energy or the _ bulk excess _
free energy is found . the top graph in fig .
[ fig12 ] shows the temperature dependence of the _ bulk excess _
free energy for the @xmath92-state potts model on the bethe lattice when @xmath29 and @xmath180 at zero magnetic field @xmath47 .
we have considered a large variety of boundary conditions , which are determined by the boundary magnetic field @xmath48 .
we found out that the singular behavior of the _ bulk excess _
free energy for @xmath180 , yields many phase transitions @xmath224 , which are sensitive to the imposed magnetic field @xmath48 . within the interval @xmath225
, we locate a phase - coexistence region on the finite - temperature interval @xmath226 .
the singular behavior in @xmath189 appears at @xmath120 and is accompanied by a discontinuous jump ( denoted by the dotted line ) .
the bottom graph zooms - in on the relevant part of the top graph , namely , the phase - coexistence region , and it shows , for which @xmath48 the phase transition @xmath224 occurs .
( we remark that there were no such discontinuities of the _ bulk excess _
free energy for the ising model : the abrupt decrease with respect to @xmath15 was always continuous . )
-state potts model on the bethe lattice @xmath70 after contracting the infinite number of the boundary layers ( @xmath180 ) .
the phase transitions are shown as the discontinuities in @xmath189 for the particular boundary fields @xmath48 .
the discontinuities are marked by the dotted vertical lines .
the graphs were constructed for @xmath29 , @xmath180 , and @xmath106.,scaledwidth=49.0% ] thus , the most emerging feature is the discontinuity of the _ bulk excess _
free energy while tuning the magnetic fields in the interval @xmath227 .
the finite - temperature interval determining the phase - coexistence region @xmath228 has a lower bound at @xmath229 , which is stable for any boundary magnetic fields @xmath230 ( including the negative fields ) , and an upper bound @xmath231 , which is saturated at @xmath232 .
figure [ fig13 ] extends our findings to more examples of the @xmath1-state potts models on various hyperbolic lattices .
since the hausdorff dimension of the hyperbolic geometries @xmath0 is infinite for arbitrary @xmath233 , the first - order phase transition has to be realized for @xmath234 .
this is related to the fact that the @xmath1-state potts model on the three - dimensional cubic lattice @xcite exhibits first - order phase transitions if @xmath234 , whereas the potts model on the two - dimensional euclidean lattices shows this feature if @xmath94 .
-state potts model for the selected potts states @xmath234 and lattices @xmath0 .
the vertical dotted lines border the discontinuous phase coexistence regions @xmath226 for @xmath29 , @xmath180 , and @xmath106.,scaledwidth=49.0% ] in general , the phase - coexistence region @xmath226 has a lower bound at @xmath235 and a finite upper bound at @xmath236 .
the width of the phase coexistence region @xmath237 is primarily proportional to the lattice coordination number @xmath3 and the number of spin states @xmath1 , whereas there is a weak dependence on the @xmath2 ( the larger the @xmath3 and/or @xmath1 , the wider @xmath226 ) .
we , therefore , conjecture that the first - order phase transition temperature @xmath120 is continuously adjustable within the interval @xmath226 by varying the boundary magnetic field @xmath48 , which is achieved for the potts models with @xmath234 and for an arbitrary hyperbolic geometry @xmath0 .
the reason for studying the complex behavior of the multistate spin models on the hyperbolic geometries lies in the potential to mimic the thermodynamic properties of various neural structures , social networks , anti de sitter spaces , etc .
we have introduced a way of specifying the novel function in the thermodynamic limit , the so - called _ bulk excess _
free energy per spin site , which is correlated with the number of contracted boundary spin layers of the entire lattice . at the same time , the _ bulk excess _
free energy remains sensitive to the magnetic field imposed on the outermost boundary layer .
recall that the phase transitions on the euclidean lattices are not influenced by choosing any type of the boundary conditions in the thermodynamic limit .
we have studied the boundary response to the bulk properties and the phase transitions in the classical multistate spin models on various hyperbolic lattices in the thermodynamic limit .
we applied ctmrg to the @xmath1-state potts models and calculated the free energy per site with high accuracy .
the potts model has been selected for its availability to describe both second- and first - order phase transitions . with respect to our results , there is no phase transition present on any hyperbolic lattice if the entire lattice in the thermodynamic limit is considered , provided that no boundary layers were contracted . to be more specific
, there is a _
false _ phase transition induced by the huge boundary structure , which is strong enough to prevail over the correct phase transition observed deep inside the system if suppressing the boundaries .
however , if a sufficient number of boundary layers are contracted ( while keeping the infinite number of inner layers ) , the correct phase transition can appear in the bulk ( known from the analytic solutions ) .
our results agrees with the only available analytic solution of spin models on bethe lattices .
we have conjectured that the multistate spin systems on hyperbolic lattices , which exhibit a continuous ( second - order ) phase transition , exhibit a firm phase - transition point irrespective of the imposed boundary magnetic field @xmath48 .
note that such spin models on hyperbolic lattice geometries belong to the mean - field universality class @xcite .
if , however , a multistate spin system results in a discontinuous ( first - order ) phase transition in the bulk , the phase - transition point depends on the magnetic field @xmath48 .
in particular , if @xmath48 is varied on the outermost boundary layer , the phase - transition temperature @xmath224 changes continuously within the temperature region @xmath237 for any @xmath234 potts models on the hyperbolic geometries @xmath0 . with respect to our recent studies of quantum ising and heisenberg models on hyperbolic lattices @xcite , which exhibit analogous physical properties to those of the classical spins model , we expect the presence of identical features . to prove this conjecture , we intend to perform extensive numerical studies on quantum systems .
the multistate potts models on non - euclidean lattices were studied in this paper for their capability to mimic realistic interacting systems , such as neural systems , social internet networks , etc .
, in which the boundary influences the behavior in the central part .
the current remarkable physical feature opens leads to an alternative view of the importance of boundary effects , as we have shown that one can continuously control the thermodynamic properties of certain types of complex systems on hyperbolic geometries . a condensed - matter view point on the ads - cft correspondence
is known to restrict a negatively curved space geometry ( ads ) to a preferred coordinate system , i.e. , a lattice @xcite .
we reduced the problem to an infinite set of two - dimensional curved hyperbolic surfaces , where the underlying lattice geometry @xmath0 varies by changing the two integer lattice parameters @xmath2 and @xmath3 .
the impact of the @xmath0 geometry on the phase transitions with respect to the boundary effects has been studied .
making use of the free - energy language , the boundary structure of the hyperbolic geometry is naturally incorporated into the solution to carry the essential feature of ads space .
our intention for future studies is to calculate the von neumann entanglement entropy of a subsystem , e.g. a bulk , for the quantum heisenberg and transverse field ising models on @xmath238 lattice geometries @xcite .
the impetus for this calculation lies in the concept of holographic entanglement entropy @xcite , in which a nongravitational theory is expected to exist on the boundary of the ( bulk ) subsystem of ( @xmath239)-dimensional hyperbolic spaces .
the entanglement entropy , which is related to a reduced density matrix of the bulk subsystem , provides an appropriate measure of the amount of information within the ads - cft correspondence .
hence , the entropy is proportional to the boundary ( minimal area surface ) within ads space , which links the duality with the corresponding @xmath240-dimensional bulk region defined in cft .
99 h. yoshikawa , k. hayashida , y. kozuka , a. horiguchi , and k. agawa , appl .
lett . * 85 * , 5287 ( 2004 ) .
f. liang , l. guo , q. p. zhong , x. g. wen , c. p. chen , n. n. zhang , and w. g. chu , appl .
89 * , 103105 ( 2006 ) .
a. cabot , a. p. alivisatos , v. f. puntes , l. balcells , o. iglesias , and a. labarta , phys .
b * 79 * , 094419 ( 2009 ) .
w. a. moura - melo , a. r. pereira , l. a. s. mol , and a. s. t. pires , phys .
a * 360 * , 472 ( 2007 ) . v. a. kazakov , phys .
a * 119 * , 140 ( 1986 ) . c. holm and w. janke , phys
b * 375 * , 69 ( 1996 ) .
d. krioukov , f. papadopoulos , a. vahdat , and m. bogu , phys .
e * 80 * , 035101 ( 2009 ) .
d. krioukov , f. papadopoulos , m. kitsak , a. vahdat , and m. bogu , phys .
e * 82 * , 036106 ( 2010 ) .
g. t hooft , in _ world scientific series in 20th century physics _ , vol . 4 , edited by a. ali , j. ellis , and s. randjbar - daemi ( world scientific , 1993 )
. l. susskind , j. math .
( n.y . ) * 36 * , 6377 ( 1995 ) . s. ryu and . t. takayanagi , phys .
96 * , 181602 ( 2006 ) .
j. maldacena , adv .
* 2 * , 231 ( 1998 ) ; int .
. phys . * 38 * , 1113 ( 1999 ) .
k. ueda , r. krcmar , a. gendiar , and t. nishino , j. phys .
soc . jpn .
* 76 * , 084004 ( 2007 ) .
t. nishino and k. okunishi , j. phys .
soc . jpn . *
65 * , 891 ( 1996 ) ; t. nishino and k. okunishi , j. phys .
. jpn . * 66 * , 3040 ( 1997 ) .
m. serina , j. genzor , y. lee , and a. gendiar , phys .
e * 93 * , 042123 ( 2016 ) .
h. shima and y. sakaniwa , j. stat .
p08017 ( 2006 ) .
h. shima and y. sakaniwa , j. phys .
a : math . gen . * 39 * , 4921 ( 2006 ) . s. k. baek , p. minnhagen , and b. j. kim , epl 79 , 26002 ( 2007 ) .
y. sakaniwa and h. shima , phys .
e * 80 * , 021103 ( 2009 ) .
i. hasegawa , y. sakaniwa and h. shima : surf .
* 601 * , 5232 ( 2007 ) .
r. mosseri and j. f. sadoc , j. physique - lettres * 43 * , l249 ( 1982 ) .
j. w. anderson , hyperbolic geometry ( 2nd edition , springer , london , 2005 ) .
f. w. wu , rev .
mod . phys .
* 54 * ( 1982 ) 235 .
s. r. white , phys .
* 69 * , 2863 ( 1992 ) .
u. schollwck , rev .
* 77 * , 259 ( 2005 ) .
r. krcmar , a. gendiar , k. ueda , and t. nishino , j. phys .
a * 41 * , 125001 ( 2008 ) .
r. j. baxter , exactly solved models in statistical mechanics ( academic press , london , 1982 . )
a. gendiar , r. krcmar , s. andergassen , m. danika , and t. nishino , phys .
e * 86 * , 021105 ( 2012 ) .
a. gendiar , m. danika , r. krcmar , and t. nishino , phys .
e * 90 * , 012122 ( 2014 ) .
m. danika and a. gendiar , j. phys .
a : math . theor . * 48 * , 435002 ( 2015 ) .
m. danika and a. gendiar , j. phys .
a : math . theor . * 49 * , 145003 ( 2016 ) .
j. genzor , v. buek , and a. gendiar , physica a * 420 * , 200 ( 2015 ) . f. verstraete and j. i. cirac , phys .
b * 73 * , 094423 ( 2006 ) . c. a. schwerdtfeger and d. a. mazziotti , j. chem . phys .
* 130 * , 224102 ( 2009 ) .
v. zauner , d. draxler , l. vanderstraeten , j. haegeman , and f. verstraete , arxiv:1412.7642 .
a. gendiar and t. nishino , phys .
e * 65 * , 046702 ( 2002 ) .
p. w. anderson , phys . today *
66*(4 ) , 9 ( 2013 ) . m. danika and a. gendiar ( unpublished ) .
|
the multistate potts models on two - dimensional hyperbolic lattices are studied with respect to various boundary effects .
the free energy is numerically calculated using the corner transfer matrix renormalization group method .
we analyze phase transitions of the potts models in the thermodynamic limit with respect to contracted boundary layers .
a _
false _ phase transition is present even if a couple of the boundary layers are contracted .
its significance weakens , as the number of the contracted boundary layers increases , until the correct phase transition ( deep inside the bulk ) prevails over the _
false _ one . for this purpose
we derive a thermodynamic quantity , the so - called _ bulk excess _
free energy , which depends on the contracted boundary layers and memorizes additional boundary effects .
in particular , the magnetic field is imposed on the outermost boundary layer . while the boundary magnetic field does not affect the second - order phase transition in the bulk if suppressing all the boundary effects on the hyperbolic lattices , the first - order ( discontinuous ) phase transition is significantly sensitive to the boundary magnetic field .
contrary to the phase transition on the euclidean lattices , the discontinuous phase transition on the hyperbolic lattices can be continuously controlled ( within a certain temperature coexistence region ) by varying the boundary magnetic field .
|
in a recent work , ken brown used algebraic methods to analyse random walks on a class of semigroups called `` left - regular bands '' . these walks include the hyperplane chamber walks of bidigare , hanlon , and rockmore . in this paper , we look at the special case of reflection arrangements that arise in the study of coxeter groups .
the random walks that we look at can be thought of as shuffles on the coxeter group .
the motivating examples are the riffle shuffle and the random - to - top shuffle on a deck of @xmath5 cards .
these correspond to the symmetric group @xmath0 , i.e. , the coxeter group of type @xmath1 .
the only coxeter groups that we deal with are the ones of type @xmath1 , @xmath2 and @xmath3 .
to make this paper accessible to readers unfamiliar with coxeter groups we have included two appendices .
appendix [ app : hyperplane ] reviews the facts we need about hyperplane arrangements . in appendix
[ app : coxeter ] , we give a brief review of coxeter groups and then explain everything in concrete terms for the three cases mentioned above . for the general theory of coxeter groups ,
we refer the reader to .
let @xmath6 be the simplicial ( coxeter ) complex associated to a coxeter group @xmath7 .
also let @xmath8 be the set of chambers ( or maximal simplices ) of @xmath6 .
then @xmath6 is a semigroup containing @xmath8 as an ideal . the product in @xmath6 is given by the projection maps as explained in appendix [ app : hyperplane ] .
we now describe the walk .
let @xmath9 be a probability distribution on @xmath6 .
if the walk is in chamber @xmath10 , then move to the chamber @xmath11 , where @xmath12 is chosen with probability @xmath13 .
the product @xmath11 is again a chamber because @xmath8 is an ideal in @xmath6 .
next we describe the walk more algebraically .
fix a commutative ring @xmath14 and consider the semigroup algebra @xmath15 .
the @xmath14-module @xmath16 spanned by the chambers is an ideal in @xmath15 . in particular , it is a module over @xmath15 ; we therefore obtain a homomorphism @xmath17 the latter being the ring of @xmath14-endomorphisms of @xmath16 .
this map is in fact an inclusion ; so we can regard elements of @xmath15 as operators on @xmath16 .
let @xmath18 be such that @xmath19 .
it is straightforward to check that the transition matrix of the random walk determined by @xmath20 is simply the matrix of the operator `` left multiplication by @xmath21 '' . in other words , it is the image of @xmath21 under the inclusion @xmath22 . a way to analyse
the random walk is to focus on @xmath15 and understand the structure of the subalgebra @xmath23 $ ] that @xmath21 generates in @xmath15 . because of the inclusion map above , we can also analyse the structure of @xmath23 $ ] in @xmath24 .
we will make use of both viewpoints in our examples
. the reader may be more familiar with this algebraic approach in the context of groups rather than semigroups . in our examples ,
the element @xmath21 will never be normalised .
in other words , if we write @xmath18 then @xmath25 .
so , strictly speaking , we will not have a probability distribution on @xmath6 .
the main reason for doing this is to simplify the algebra .
so everytime we describe the random walk associated to @xmath21 , the description will be correct upto a normalisation factor
. this will also be true for all other objects that will show up in the analysis . to get rid of this anamoly
, we adopt the following convention . if the random walk description says `` do xxx at random . ''
then the normalisation factor is `` the number of ways of doing xxx '' . in general
, the random walk description will involve a sequence of independent random acts in which case the normalisation factor will be the product of the individual factors . in most examples , we will prefer to first define all objects of interest as elements of @xmath15 and later to provide motivation by interpreting the associated random walks .
making this translation will always involve the normalisation factor defined above .
next we mention that in our examples , the element @xmath21 will always lie in the subalgebra of @xmath7-invariants of @xmath15 .
we now discuss this subalgebra .
the @xmath7-invariants of @xmath15 under the natural @xmath7-action form a @xmath14-algebra @xmath26 . as a @xmath14-module , @xmath26 is free with one basis element for each @xmath7-orbit in @xmath6 , that basis element being the sum of the simplices in the orbit .
since orbits correspond to types of simplices , we get a basis vector @xmath27 for each @xmath28 , where @xmath29 is the set of simplices of type @xmath30 and @xmath31 is the set of all labels or types of vertices of @xmath6 . also define @xmath32 to be the set of simplices of rank @xmath33 .
as already mentioned , in our examples @xmath34 and hence @xmath23 $ ] is a certain commutative subalgebra of @xmath26 .
since the element @xmath21 is always motivated by some shuffle considerations , we will call @xmath23 $ ] a shuffle algebra .
bidigare proved that @xmath26 is anti - isomorphic to solomon s descent algebra , which is a certain subalgebra of the group algebra @xmath35 .
so shuffle algebras are anti - subalgebras of solomon s descent algebra . in section [ s : mt ]
we define shuffle algebras and provide a semi - simplicity criterion for their analysis . in the next three sections , we give examples of shuffle algebras of type @xmath36 and @xmath3 respectively .
these sections should be read in conjunction with appendices [ subs : a ] , [ subs : b ] and [ subs : d ] respectively .
our examples are motivated by shuffle considerations on the one hand and the geometry of the coxeter complex on the other .
the striking similarity among the examples can be traced to maps among the three coxeter complexes .
these maps which give a unified framework for our examples are explained in section [ s : maps ] .
this section gives us a good overall picture of how things fit together and can play an important role in the further development of the theory .
we see an example of this in section [ s : q ] , where we briefly discuss the generalisation of the random walks to buildings .
this section is somewhat technical and it is better to read it lightly at present .
you may want to go over the details later with a concrete example in mind .
we first abstract the notion of a shuffle algebra and then develop the main tool for the analysis of these algebras .
we first need a preliminary definition .
an additive ( resp .
multiplicative ) _ shuffle semigroup _ is a subsemigroup of the semigroup of non - negative ( resp .
positive ) integers under addition ( resp .
multiplication ) .
note that there is only one additive shuffle semigroup upto isomorphism .
however in the multiplicative case , there could be many .
now we are ready to define a shuffle algebra .
we say that a subalgebra @xmath37 is an additive ( resp .
multiplicative ) _ shuffle algebra _ if it satisfies the following conditions : 1 .
2 . @xmath39 has a basis of the form @xmath40,@xmath41,@xmath42, where @xmath43 is an element of @xmath44 , the span of simplices of rank @xmath33 .
@xmath45\subseteq k\sigma$ ] .
4 . @xmath39 contains an additive ( resp .
multiplicative ) shuffle semigroup @xmath46 as a spanning set .
note that the definition of a shuffle algebra @xmath39 depends on the coxeter complex @xmath6 under consideration .
in fact , conditions @xmath47 and @xmath48 say that @xmath39 is in tune with the geometry of @xmath6 . hence it is more correct to write a shuffle algebra as a pair @xmath49 .
but we will not bother with this since in our examples the coxeter complex @xmath6 will always be clear from context . an alternate way to express condition
@xmath50 is to say that there exists an injective semigroup map @xmath51 such that the image spans @xmath39 . in all our examples , the semigroup @xmath46 will have an interpretation in terms of card shuffles .
this explains our terminology .
the above definition is mainly motivated by the infinite families @xmath36 and @xmath3 .
the significance of this definition for the sporadic coxeter groups of type @xmath52 , @xmath53 , etc .
is not clear .
it is also natural to ask whether shuffle algebras can be classified .
we do not know the answer .
we would also like to point out that we have imposed the strictest possible conditions that our examples led us to . as a result ,
some of the algebras we consider are not shuffle algebras ; for example , the shuffle double constructions in sections [ subs : brs ] and [ subs : drs ] . hence a more flexible definition might work better , say for the classification problem .
ken brown shows that if @xmath14 has characteristic 0 and @xmath21 is a non - negative integral linear combination in the canonical basis of @xmath15 consisting of all simplices then @xmath23 $ ] is _ split - semisimple _ , that is , @xmath23 { \cong}k^n$ ] for @xmath54 . to guarantee this result for an arbitrary field he introduces an additional condition on @xmath41 . in all our examples , we assume that @xmath14 has characteristic 0 .
also @xmath41 will always be a non - negative integral linear combination in the canonical basis of @xmath15 .
hence it follows by condition @xmath48 that the shuffle algebras we consider are split - semisimple .
we will not rely on brown s result however .
instead we will use a semisimplicity criterion which we now state .
[ p : ss ] let @xmath39 be an @xmath5-dim algebra containing a spanning semigroup @xmath46 .
suppose that there exist characters @xmath55,
,@xmath56 of @xmath46 and elements @xmath57, ,@xmath58 of @xmath39 such that for every @xmath59 , we have @xmath60 then @xmath61 with @xmath62 as the primitive idempotents .
also the @xmath63 s extend to characters of @xmath39 .
the elements @xmath57, ,@xmath58 of @xmath39 span @xmath46 which in turn spans @xmath39 . since @xmath64 this implies that @xmath57, ,@xmath58 form a basis for @xmath39 .
this yields a vector space isomorphism @xmath65 , where @xmath66 .
hence for @xmath59 , @xmath67 .
since @xmath68 are characters of @xmath46 , we have @xmath69 for @xmath70 . the fact that @xmath71 respects the algebra structures on a spanning set @xmath46 of @xmath39 implies that it is in fact an algebra homomorphism .
as mentioned before , this proposition will apply to all our examples .
the algebra @xmath39 will be a shuffle algebra and @xmath46 the shuffle semigroup contained in @xmath39 .
the characters @xmath72 of @xmath46 that we will use are @xmath73 if @xmath46 is additive and @xmath74 if @xmath46 is multiplicative . here
@xmath75 is a fixed non - negative integer .
the specific values of @xmath75 that we need to choose depend on the example at hand .
the advantage of proposition [ p : ss ] is that it is tailor - made for shuffle algebras and it gives an explicit isomorphism of @xmath39 with @xmath76 .
the coxeter complexes @xmath6 of type @xmath36 and @xmath3 can be described using the language of `` ordered partitions '' .
also , chambers of @xmath6 can be described as a `` deck of cards '' .
this is explained in appendices [ subs : a ] , [ subs : b ] and [ subs : d ] respectively .
we follow the convention that `` left to right '' for an ordered partition is `` top to bottom '' for a deck of cards .
it is essential to understand this language before reading the examples .
the side shuffle and riffle shuffle of type @xmath1 considered in sections [ subs : ass ] and [ subs : ars ] are the motivating examples of this paper .
it is also a good idea to first understand these two basic examples before proceeding to the rest .
we consider three examples ; the side shuffle , the two sided shuffle and the riffle shuffle .
the first two are additive in nature while the third is multiplicative .
the analysis for the first and third example parallels that of diaconis , fill and pitman and bayer and diaconis respectively .
the difference is that we prefer to work with @xmath26 , which is the more geometric description of solomon s descent algebra .
elements of the label set @xmath31 will be written as @xmath77 ; see figure 2.1 in appendix [ subs : cd ] . recall that @xmath78 , the sum of simplices of type @xmath30 , for @xmath79 form a basis for @xmath26 .
hence these elements will play a key role in the analysis .
this is more commonly known as the random - to - top shuffle or the tsetlin library .
the element of interest is @xmath80 , that is , it is the sum of all vertices of type @xmath81 . in terms of ordered partitions , @xmath82 is the sum of all ordered two block partitions of @xmath83 $ ] such that the first block is a singleton . as explained in section [ subs : rw ] , there is a random walk on a deck of @xmath5 cards associated to @xmath21 .
it consists of removing a card at random and replacing it on top . when @xmath84 , for example , @xmath85 is a typical summand of @xmath21 .
its product with the deck @xmath86,@xmath87,@xmath88,@xmath89 gives @xmath90,@xmath91,@xmath88,@xmath89 , that is , the overall effect is to remove the card labelled @xmath92 and put it on top .
note that the normalisation factor in this case is @xmath5 , consistent with the convention of section [ subs : ne ] .
in addition to @xmath93 , we define @xmath94 where @xmath95 for @xmath96 .
to explain in words , @xmath41 is the sum of all the vertices of type @xmath81 , @xmath42 is the sum of all the edges of type @xmath97 and so on till @xmath98 which is the sum of all chambers of @xmath6 .
as ordered partitions , @xmath99 is the sum of all ordered @xmath100 block partitions of @xmath83 $ ] such that the first @xmath33 blocks are singletons . also for convenience ,
we define @xmath101 . just like @xmath41
, we can describe the random walk on a deck of @xmath5 cards associated to @xmath43 .
@xmath102 choose @xmath33 distinct cards at random and put them on top in a random order .
note that @xmath103 was previously defined in an artificial way .
but the random walk description for @xmath103 makes perfect sense and we now see the motivation in setting @xmath101 .
the two are identical as random walks and by our convention have the same normalisation factor of @xmath104 . from now on , we will leave out the discussion on normalisation .
we now present two methods to analyse @xmath105 $ ] . *
the shuffle method*. put @xmath106 .
define @xmath107-shuffles @xmath108 for @xmath109 by @xmath110 where @xmath111 are the stirling numbers of the second kind .
they count the number of ways in which @xmath107 elements can be divided into @xmath33 non - empty subsets .
we make the convention that @xmath112 for @xmath113 .
also @xmath114 and @xmath115 for @xmath116 .
note that @xmath117 and @xmath118 .
the motivation behind this definition will soon become clear .
unlike for @xmath43 , the description of @xmath108 in terms of ordered partitions is not clear .
however , the associated random walk on a deck of @xmath5 cards can be readily described .
@xmath119 pick a card at random and mark it .
repeat this process @xmath107 times .
there is no restriction on the number of times a given card may get marked .
move all the marked cards to the top in a random order . to see the equivalence of the two definitions , note that @xmath108 can be expressed as a sum indexed by the number of distinct cards that get marked .
suppose that @xmath33 distinct cards get marked . by our marking scheme for @xmath108
, there are exactly @xmath111 ways in which the same set of @xmath33 cards gets marked .
this leads to equation . with the probabilistic interpretation for @xmath108
, one can also convince oneself that @xmath120 .
this relation was the main motivation behind the definition of @xmath108 .
thus we get an additive shuffle semigroup @xmath121 contained in @xmath45 = k[s_1]$ ] .
the additivity says that the semigroup @xmath46 spans @xmath39 .
we can deduce from equation that @xmath122,@xmath41, ,@xmath98 lie in @xmath39 .
also they span @xmath46 and hence @xmath39 .
since we know that they are linearly independent in @xmath26 , it now follows that they form a basis for @xmath39 .
hence @xmath39 is a shuffle algebra .
the analysis so far shows that the element @xmath41 determines what @xmath123 should be , assuming that we want @xmath124 .
we will see this more directly in the second method .
now we show that @xmath125 .
we do this by moulding equation in the shape of equation ( * ) of proposition [ p : ss ] .
the key step is to use the following explicit formula for the stirling numbers @xmath111 ; see . @xmath126 with the convention that @xmath127 .
substituting the above expression in the formula for @xmath108 and rearranging terms , we get for @xmath109 @xmath128 observe that @xmath129 because @xmath130 .
the remaining @xmath62 s are clearly non - zero .
now apply proposition [ p : ss ] to @xmath46 and @xmath39 along with the @xmath5 characters of @xmath46 given by @xmath131 for @xmath132 .
this gives @xmath133 .
the isomorphism maps @xmath108 to @xmath134 .
these give the possible set of eigenvalues of @xmath108 considered as an operator on any @xmath39-module .
the formulas for the @xmath62 s are identical to those obtained in .
the minimal polynomial for @xmath108 is given by @xmath135 .
if we expand this polynomial , substitute @xmath136 and use equation then we get some identities involving the stirling numbers .
* the direct method*. we first claim that @xmath137 if @xmath138 and @xmath139 .
we see this from the description of the @xmath21 s as ordered partitions as follows .
a typical summand of @xmath43 is @xmath86,@xmath87,@xmath140,@xmath141,@xmath142 .
this term appears @xmath33 times in the product @xmath143 ; the @xmath33 summands of @xmath41 that contribute being the ones in which the element in the singleton block is one of @xmath144 .
this yields @xmath145 .
the remaining terms yield @xmath146 and the claim follows .
for @xmath147 , the claim says that @xmath148 .
thus @xmath149 is determined by @xmath41 and the same is true for @xmath150 and so on .
more precisely , by induction we get @xmath151 this along with @xmath139 implies that both @xmath122,@xmath41, ,@xmath98 and @xmath152,@xmath153, ,@xmath154 form a basis for @xmath39 .
the basic relation satisfied by @xmath41 is @xmath155 , that is , @xmath156 .
hence @xmath157}{x(x-1)\ldots(x - n+2)(x - n)}$ ] , where the map sends @xmath41 to @xmath158 .
this shows that @xmath39 is split semisimple .
note that this time we did not use proposition [ p : ss ] to arrive at this conclusion .
we now give algebraic motivation for equation by rederiving it here . right now
@xmath43 is defined only for @xmath159 .
but we may extend the definition of @xmath43 to any @xmath33 using equation . with this extension
, one may check that @xmath101 and @xmath160 for @xmath161 .
we now find a formula for @xmath162 by inverting equation formally .
write @xmath163 for some constants @xmath111 .
multiply both sides on the right by @xmath41 and use the relation @xmath137 to obtain the recursion : @xmath164 with @xmath165 .
the recursion and the initial conditions show that @xmath111 are the stirling numbers of the second kind .
we now define the @xmath107-shuffle @xmath108 to be @xmath166 .
this gives equation . by our definition
, it follows directly that @xmath120 .
note that this time we did not rely on any probabilistic interpretation of @xmath108 to derive this additive relation .
consider the map @xmath167 { \twoheadrightarrow}a$ ] which sends @xmath158 to @xmath41 .
along with @xmath168 , the other sequence which played a prominent role in our analysis was @xmath169 , where @xmath170 . both these sequences are polynomial sequences of binomial type .
this kind of structure seems to be common to all the additive examples .
the first sequence is always the same but the second sequence varies .
and the relation between the two gives us various analogues of the stirling numbers .
it would be worthwhile to understand this better .
it is possible to write down an explicit formula for @xmath171 .
we state it here for completeness .
@xmath172 there is also a @xmath4-analogue of the side shuffle which we will explain in section [ s : q ] .
this example is similar to the side shuffle and it will be useful to keep the analogy in mind .
the element of interest is @xmath173 . in terms of ordered partitions , @xmath21 is the sum of all ordered two block partitions of @xmath83 $ ] such that either the first block or the second block is a singleton .
the associated random walk on a deck of cards consists of removing a card at random and replacing it either on top or at the bottom ; whence the name .
in addition to @xmath93 , we define @xmath174 , where @xmath175 for @xmath96 .
note that @xmath176 is the union of the first @xmath14 and last @xmath177 elements of the label set .
hence the cardinality of @xmath176 is always @xmath33 .
also as ordered partitions , @xmath178 is the sum of all ordered @xmath100 block partitions of @xmath83 $ ] such that the first @xmath14 and the last @xmath177 blocks are singletons . as is evident from the formula , expressing @xmath43 in this language is not pleasant .
however the probability description will be simple as we will see .
also put @xmath106 and @xmath179 .
now define @xmath107-shuffles @xmath108 for @xmath109 by @xmath180 where @xmath111 counts the number of ways in which a set of @xmath107 elements can be divided into @xmath33 non - empty subsets , where in each subset the elements are further divided into two subsets .
we will call the @xmath111 s the _ signed _ stirling numbers .
they satisfy the recursion : @xmath181 with @xmath182 and @xmath183 .
an explicit formula is as under : @xmath184 note that @xmath117 and @xmath118 .
these definitions are motivated by the following interpretations of the random walks associated to @xmath43 and @xmath108 .
@xmath102 choose @xmath33 cards at random . split these into two distinct piles such that every split is equally likely .
move the cards in the first pile to the top in a random order and those in the second pile to the bottom in a random order .
@xmath119 pick a card at random and mark it @xmath185 or @xmath186 .
repeat this process @xmath107 times .
if at any time we choose a card that has already been marked then we overwrite that mark . now move all cards marked @xmath185 to the top and those marked @xmath186 to the bottom in a random order .
first check that @xmath179 .
next note that @xmath43 can be expressed as a sum depending on the number of cards @xmath14 that are moved to the top .
there are @xmath187 ways for this to occur .
this leads us to the equation for @xmath43 in terms of the @xmath178 s that we wrote earlier .
the interpretation of @xmath108 is slightly subtle . as for the side shuffle ,
express @xmath108 as a sum indexed by the number of distinct cards that get marked .
suppose that @xmath33 distinct cards get marked . by our marking scheme for @xmath108 ,
there are exactly @xmath188 ways in which the same set of @xmath33 cards gets marked .
this is because we use distinct labels @xmath185 and @xmath186 .
however , this extra factor is compensated in the next step where we use the last label on a marked card to decide whether it goes to the top or the bottom .
now observe that @xmath120 .
this gives us an additive shuffle semigroup @xmath121 contained in @xmath45 = k[s_1]$ ] .
we repeat the analysis for the side - shuffle to show that @xmath133 , where @xmath108 maps to @xmath189 .
the formula for the primitive idempotents is given by @xmath190 this time the relation @xmath191 forces @xmath129 . also equation
can be rewritten as @xmath192 as for the side - shuffle , the direct method also works . using the description of the @xmath21 s as ordered partitions , it is easy to see that @xmath193 if @xmath138 and @xmath194 . by induction
we get @xmath195 this along with @xmath194 implies that both @xmath122,@xmath41, ,@xmath98 and @xmath196,@xmath153, ,@xmath154 form a basis for @xmath39 .
the basic relation satisfied by @xmath41 is @xmath197 .
hence @xmath157}{x(x-2)\ldots(x-2n+4)(x-2n)}$ ] .
this shows that @xmath39 is split semisimple .
inverting equation formally and using the relation @xmath193 leads to the recursion of the signed stirling numbers written earlier and hence to equation .
a _ riffle shuffle _ is a common method people use for shuffling a deck of cards .
the gilbert - shannon - reeds mathematical model for riffle shuffling consists of first cutting the deck into two parts , making the cut according to a binomial probability distribution , and then interleaving the two parts in such a way that every interleaving is equally likely .
a natural generalisation is the @xmath107-shuffle in which we cut the deck into @xmath107 parts ( rather than just 2 ) and then interleave them . from our point of view , it is more natural to consider the inverse riffle shuffle @xmath198 and more generally the inverse @xmath107-shuffle @xmath108
. it has the following description as a random walk on a deck of @xmath5 cards .
@xmath108 : label each card randomly with an integer from @xmath199 to @xmath107 .
move all the cards labelled @xmath199 to the bottom of the deck , preserving their relative order .
next move all the cards labelled @xmath92 above these again preserving their relative order and so on .
we can understand the inverse @xmath107-shuffle @xmath108 as an element of @xmath26 .
we begin with the element @xmath200 ; that is , @xmath41 is the sum of all the vertices in @xmath6 . in terms of ordered partitions
, it is the sum of all ordered two block partitions of @xmath83 $ ] .
this is closely related to the inverse riffle shuffle , in fact , @xmath201 , where @xmath106 is the one block partition .
we will show this in more generality .
let @xmath202 be the sum of all the simplices of rank @xmath203 in @xmath6 for @xmath204 .
then define shuffles @xmath108 for @xmath205 by @xmath206 in terms of ordered partitions , @xmath207 is the sum of all ordered @xmath33 block partitions of @xmath83 $ ] .
to express @xmath108 in this language , we need the notion of a weakly ordered partition of @xmath83 $ ] .
it is an ordered partition of @xmath83 $ ] in which the blocks are allowed to be empty .
note that there are exactly @xmath208 weakly ordered @xmath107 block partitions that give the same ordered @xmath33 block partition of @xmath83 $ ] .
it follows that @xmath108 is the sum of all weakly ordered @xmath107 block partitions of @xmath83 $ ] .
if we assign labels @xmath209 to these weakly ordered @xmath107 blocks such that the leftmost block is labelled @xmath107 and the rightmost is labelled @xmath199 , then we get the random walk description for @xmath108 that we started with .
it is clear from both viewpoints that @xmath210 .
this gives us a multiplicative shuffle semigroup @xmath211 contained in @xmath45 = k[s_2]$ ] .
it now follows that @xmath122,@xmath41,
,@xmath98 is a basis for @xmath39 and that @xmath39 is a shuffle algebra .
now we show that @xmath125 .
expand each @xmath208 as a polynomial in @xmath107 ( with no constant term ) : @xmath212 for some constants @xmath213 . to be precise
, @xmath213 is the coefficient of @xmath214 in the polynomial @xmath215 . substituting the above expression in the formula for @xmath108 and rearranging terms
, we get for @xmath205 @xmath216 all the @xmath62 s are clearly non - zero .
now apply proposition [ p : ss ] to @xmath46 and @xmath39 along with the @xmath5 characters of @xmath46 given by @xmath217 for @xmath218 .
this gives @xmath133 .
the isomorphism maps @xmath108 to @xmath219 .
these give the possible set of eigenvalues of @xmath108 considered as an operator on any @xmath39-module .
they are distinct for @xmath220 and hence @xmath108 generates @xmath39 for all @xmath221 .
also the minimal polynomial for @xmath108 on any faithful @xmath39-module is given by @xmath222 .
write it as @xmath223 . now if we substitute @xmath136 and use equation then we get identities involving binomial coefficients , namely @xmath224 for @xmath225
. it would be nice to have an explicit formula for the coefficients @xmath226 also .
the analysis that we gave for this example was the analogue of the shuffle method of the previous two examples .
the direct computational method of those examples is not feasible here .
writing a formula for @xmath143 is not easy because of the multiplicative nature of the example .
we consider two examples , the side shuffle and the riffle shuffle . they should not be regarded as mere generalisations of the ones discussed so far . in fact
, we will see in section [ s : maps ] that they are in a sense more fundamental and have a right to exist on their own . in this section
, the term partition always refers to a partition of type @xmath2 ; see appendix [ subs : b ] .
also elements of the label set @xmath31 will be written as @xmath227 ; see figure 2.2 in appendix [ subs : cd ] .
the element of interest is @xmath80 . in terms of partitions , @xmath82 is the sum of all three block partitions such that the first ( and hence the last ) block is a singleton .
the associated random walk on a deck of @xmath5 signed cards consists of removing a card at random and replacing it on top with either the same sign or its reverse both with equal probability .
note that the normalisation factor is @xmath228 .
in addition to @xmath93 , we define @xmath94 where @xmath95 for @xmath229 .
to explain in words , @xmath41 is the sum of all the vertices of type @xmath81 , @xmath42 is the sum of all the edges of type @xmath97 and so on till @xmath230 which is the sum of all chambers of @xmath6 . as partitions
, @xmath99 is the sum of all @xmath231 block partitions such that the first ( and hence the last ) @xmath33 blocks are singletons .
define @xmath107-shuffles for @xmath109 by @xmath232 where @xmath111 are the signed stirling numbers defined in section [ subs : atss ] .
the associated random walks on a deck of @xmath5 signed cards are as follows : @xmath102 choose @xmath33 cards at random .
move them to the top in a random order . now with equal probability either flip or do not flip the sign of the chosen cards .
@xmath119 pick a card at random and with equal probability either flip or do not flip its sign .
repeat this process @xmath107 times .
there is no restriction on the number of times the same card may get picked .
move all the picked cards to the top in a random order .
the interpretation of @xmath108 involves the same kind of subtlety as the two sided shuffle of type @xmath1 discussed in section [ subs : atss ] .
we observe that @xmath120 . after this
, the analysis works exactly like the two sided shuffle for type @xmath1 .
the only difference is that @xmath233 since we do not have any relation like @xmath234 .
this gives @xmath235 , where @xmath108 maps to @xmath236 . like the additive examples of type @xmath1
, we can also use the direct method here . using the partition description of the @xmath21 s , it is easy to see that @xmath193 if @xmath237 and @xmath238 and the rest is similar .
this example is motivated by the riffle shuffle of type @xmath1 . in this case , the shuffles @xmath108 split into two cases depending on the parity of @xmath107 .
they can be described using the inverse @xmath107-shuffles of section [ subs : ars ] and a special shuffle @xmath198 .
we will call @xmath198 the inverse _ signed _ riffle shuffle or the inverse riffle shuffle of type @xmath2 .
@xmath198 : for every card , we either flip or do not flip its sign with equal probability . the cards with unchanged signs move to the top in the same relative order and the rest move to the bottom in the reverse relative order .
@xmath239 : we do a usual inverse @xmath107-shuffle with labels @xmath209 .
then within each of the @xmath107 blocks with a fixed label we do an inverse signed riffle shuffle @xmath198 .
@xmath240 : we do a usual inverse @xmath241-shuffle with labels @xmath242 .
then we do an inverse signed riffle shuffle @xmath198 on each block except the one labelled 0 .
these shuffles are multiplicative .
one can check directly that @xmath243 for any @xmath244 irrespective of parity .
however , we will derive this relation by considering the shuffles @xmath108 as elements of @xmath26 .
it is natural to first analyse the even and odd shuffles separately and then to put them together later . *
the even part*. the element of interest is @xmath245 . in terms of partitions
, @xmath246 is the sum of all three block partitions such that the second ( zero ) block is empty .
if we try to analyse @xmath247 , @xmath248 and so on then we observe that they involve only those partitions that have an empty zero block .
in geometric language , they involve only those faces whose type contains the letter @xmath249 .
this motivates the definition of @xmath43 which we now give .
put @xmath250 for @xmath251 . in terms of partitions , @xmath99 is the sum of all @xmath231 block partitions such that the zero block is empty . in the spirit of the riffle shuffle of type @xmath1 ,
we define shuffles @xmath239 for @xmath205 by @xmath252 note that @xmath253 . also put @xmath254 . to express @xmath239 in words ,
we use the notion of a weak partition that we defined in appendix [ subs : b ] .
it is a partition in which the signed blocks are also allowed to be empty .
since the partitions are anti - symmetric , there are exactly @xmath208 weak @xmath255 block partitions that give the same @xmath231 block partition .
it follows that @xmath239 is the sum of all weak @xmath255 block partitions such that the zero block is always empty ; so we have upto @xmath256 non - empty blocks .
this explains the term `` @xmath256-shuffle '' .
if we assign labels @xmath209 in decreasing order to the first @xmath107 blocks with the leftmost ( signed ) block labelled @xmath107 , then we get the random walk description for @xmath239 that we started with . with the partition description it is immediate that @xmath257 .
this gives us a multiplicative shuffle semigroup @xmath258 contained in @xmath105 = k[s_2]$ ] .
it now follows that @xmath122,@xmath41,
,@xmath230 is a basis for @xmath105 $ ] and that it is a shuffle algebra . *
the odd part*. this case is completely analogous to the even case , the only difference being that now we put no restriction on the zero block .
we begin with the element @xmath259 ; that is , @xmath260 is the sum of all the vertices in @xmath6 .
we also define @xmath261 as the sum of all the simplices of rank @xmath33 in @xmath6 for @xmath229 . in terms of partitions , @xmath262 is the sum of all @xmath231 block partitions . note that @xmath263 . also put @xmath264 .
next define shuffles @xmath240 for @xmath109 by @xmath265 to express @xmath240 in words , we again use the notion of a weak partition .
exactly as in the even case , it follows that @xmath240 is the sum of all weak @xmath255 block partitions .
the random walk description is obtained by assigning labels @xmath266 in decreasing order to the first @xmath241 blocks with the leftmost ( signed ) block labelled @xmath107 . with the partition description it is immediate that @xmath267 .
this gives us a multiplicative shuffle semigroup @xmath268 contained in @xmath269 = k[s_3]$ ] .
it now follows that @xmath270,@xmath260,
,@xmath271 is a basis for @xmath269 $ ] and that it is a shuffle algebra . *
the shuffle double . *
we now put the even and odd parts together .
define @xmath272 .
it is immediate from the descriptions of @xmath239 and @xmath240 in terms of partitions that @xmath243 for all @xmath273 irrespective of parity .
thus we get a multiplicative shuffle semigroup @xmath46 as a spanning set in @xmath274 $ ] .
it follows that @xmath39 is a @xmath228-dimensional algebra with basis @xmath275 .
we call it a shuffle `` double '' algebra .
it is not a shuffle algebra because the first three conditions are each violated by a factor of two . for example , in condition @xmath48 , instead of a single generator @xmath41 , we have two generators @xmath41 and @xmath260 . we now show that @xmath276 .
following the example of the riffle shuffle of type @xmath1 , we write equation as @xmath277 here @xmath213 is the coefficient of @xmath214 in the polynomial @xmath278 .
it is @xmath279 times the @xmath213 that occurred in the riffle shuffle of type @xmath1 . for notational convenience , we set @xmath280 .
similarly , we rewrite equation as @xmath281 here @xmath282 is the coefficient of @xmath214 in the polynomial @xmath283 .
note that @xmath284 .
we next define two different kinds of characters of @xmath46 . for any non - negative integer @xmath203 ,
define @xmath285 for @xmath286 .
also define @xmath287 for @xmath286 and @xmath288 for @xmath289 .
now observe that for any @xmath59 @xmath290 this can be checked separately for @xmath291 using equations and .
the second summation actually goes only till @xmath292 , since @xmath293 .
now apply proposition [ p : ss ] to @xmath46 and @xmath39 along with the @xmath228 characters @xmath63 for @xmath294 and @xmath295 for @xmath296 to conclude that @xmath276 .
it is clear that @xmath105 { \stackrel{\cong}{\longrightarrow}}k^{n+1}$ ] , where the @xmath297 factors correspond to the idempotents @xmath298 .
the isomorphism maps @xmath239 to @xmath299 . to see that @xmath269 { \cong}k^{n+1}$ ] , first note that @xmath300 for @xmath301 .
hence we need to lump together the factors corresponding to @xmath62 and @xmath302 .
thus we get @xmath269 { \stackrel{\cong}{\longrightarrow}}k^{n+1}$ ] , with @xmath303 as the primitive idempotents . the isomorphism maps @xmath240 to @xmath304 .
we consider two examples , the side shuffle and the riffle shuffle . in this section
, the term partition always refers to a partition of type @xmath3 ; see appendix [ subs : d ] .
elements of the label set @xmath31 will be written as @xmath305 ; see figure 2.3 in appendix [ subs : cd ] .
we also recall that the random walk operates on an almost signed deck of cards or a deck of type @xmath3 .
it is a deck in which every card , except the bottomost , is signed . also ,
one of the two basic algebras of section [ subs : drs ] does not strictly satisfy condition @xmath306 in the definition of a shuffle algebra .
this is relevant to the remark in section [ subs : defn ] .
the element of interest is @xmath80 . in terms of partitions
, @xmath82 is the sum of all three block partitions such that the first ( and hence the last ) block is a singleton .
the associated random walk on an almost signed deck of @xmath5 cards consists of removing a card at random and replacing it on top with either the same sign or its reverse both with equal probability .
if the bottomost unsigned card is chosen then we give it a sign and then erase the sign on the new bottomost card .
in addition to @xmath93 , we define @xmath94 where @xmath95 for @xmath307 .
note that the choice for @xmath98 is not immediately obvious .
we set @xmath308 . the motivation becomes clearer from the following . as partitions
, @xmath99 is the sum of all @xmath231 block partitions such that the first ( and hence the last ) @xmath33 blocks are singletons . observe that for @xmath309 , this does give us the above definition of @xmath98 .
we also set @xmath179 .
define @xmath107-shuffles for @xmath109 by @xmath232 where @xmath111 are the signed stirling numbers defined in section [ subs : atss ] .
the associated random walks on a deck of type @xmath3 are as follows .
@xmath102 choose @xmath33 cards at random .
move them to the top in a random order . now with equal probability either flip or do not flip the sign of the chosen cards .
if the bottomost unsigned card is one of the chosen cards then we give it a sign and then erase the sign on the new bottomost card .
@xmath119 assign the sign @xmath310 to the bottomost card .
now do the side shuffle @xmath108 of type @xmath2 .
drop the sign of the bottomost card .
observe that @xmath179 .
we did not have any such relation for the side shuffle of type @xmath2 .
it is interesting how this relation emerges by just making the bottomost card unsigned .
also observe that @xmath120 . at this point
, it is clear that the rest is identical to the two sided shuffle for type @xmath1 and we omit it .
the shuffles @xmath108 in this example can be described in the same way as the inverse @xmath107-shuffles of type @xmath311 with a minor modification .
this is because we are now operating on a deck of type @xmath3 rather than of type @xmath2 .
the difference between the two is that in the former the bottomost card is unsigned .
we now describe @xmath108 , the inverse @xmath107-shuffle of type @xmath3 .
@xmath119 assign the sign @xmath310 to the bottomost card .
now do an inverse @xmath107-shuffle of type @xmath2 .
drop the sign of the bottomost card .
it is useful to describe these shuffles in another equivalent way using the analogy with the previous situation rather than the end result .
for that , we first define the inverse riffle shuffle of type @xmath3 . for every signed card , we either flip or do not flip its sign with equal probability .
the cards whose signs were flipped move below the unsigned card ( in the reverse relative order ) .
the rest stay on top .
note that after this shuffle , we may not have a deck of type @xmath3 , since the unsigned card is not necessarily at the bottom .
hence we define a _ correction _ operation which assigns a sign to the unsigned card and then erases the sign of the bottomost card .
we now give a description of our shuffles as operators on a deck of type @xmath3 .
@xmath239 : we do a usual inverse @xmath107-shuffle with labels @xmath209 .
then within each block with a fixed label we do an inverse riffle shuffle of type @xmath3 or @xmath2 depending on whether the block contains the unsigned card or not .
lastly , we do the correction .
@xmath240 : we do a usual inverse @xmath241-shuffle with labels @xmath242 .
then we repeat the above on each block except the one labelled 0 and then do the correction .
note that @xmath198 is just an inverse riffle shuffle of type @xmath3 followed by the correction . *
the even part*. the element of interest is @xmath312 . in terms of partitions
, @xmath313 is the sum of all three block partitions such that the central block is empty .
put @xmath314 for @xmath251 . explicitly for @xmath315 , we have @xmath316 .
note that we consider only those faces whose type contains either @xmath317 or @xmath318 .
if it contains both @xmath317 and @xmath318 then we put a coefficient of @xmath92 in front . in terms of partitions
, @xmath99 is the sum of all @xmath231 block partitions such that the central block is empty . in this description
we follow the convention that the partitions @xmath319 and @xmath320 are counted separately , even though they are both equal to @xmath321 .
this explains the coefficient of @xmath92 in front of faces whose type contains both @xmath317 and @xmath318 . *
the odd part*. the element of interest is @xmath322 .
note that in contrast to all our examples so far , @xmath260 is a combination of elements of @xmath323 and @xmath324 .
next we define @xmath325 for @xmath251 . here
@xmath43 is as defined in the even case .
thus we see that @xmath261 is a combination of elements of @xmath44 and @xmath326 .
this is the only example where the geometric description of objects of interest is somewhat complicated . for better motivation
we now consider partitions .
we have already interpreted @xmath43 in terms of partitions .
note that the two summation terms that we added give the sum of all @xmath231 block partitions with a non - empty central block .
hence we see that @xmath261 is the sum of all @xmath231 block partitions . while counting partitions , we adopt the same convention as in the even case .
* the shuffle double . *
we define shuffles @xmath239 and @xmath240 by equations and and the analysis is identical word for word to the case of @xmath2 .
we only point out that partition now means partition of type @xmath3 and @xmath43 and @xmath261 refer to the definitions that we made above .
the shuffles can be described using weak partitions and this leads to the two random walk interpretations that we gave earlier .
they are slightly complicated now because of the more involved nature of the partitions of type @xmath3 . a clear conceptual explanation of the close connection between the riffle shuffles of type @xmath2 and @xmath3 is given in the next section .
the riffle shuffle examples in sections [ subs : ars ] and [ subs : brs](odd part ) show that for types @xmath1 and @xmath2 , @xmath327 , where @xmath202 is the sum of all the simplices of rank @xmath203 .
this is because @xmath202 and @xmath43 are elements of a commutative ( shuffle ) algebra .
the fact that @xmath327 can also be proven by a direct geometric argument .
the key property is that the simplicial complex induced on the support of any face of the coxeter complex of type @xmath1 or @xmath2 is again of the same type .
this property fails for @xmath3 .
it is incomplete in this geometric sense .
note that the @xmath202 s as defined above did not play any role in the riffle shuffle for @xmath3 .
in fact , their role for type @xmath3 in this theory is far from being clear .
recall that the reflection arrangements of type @xmath2 , @xmath3 and @xmath1 are given by the hyperplanes @xmath328 ; @xmath329 and @xmath330 @xmath331 respectively . let @xmath332 , @xmath333 and @xmath334 be the corresponding coxeter complexes .
observe that the arrangement for @xmath3 ( resp .
@xmath1 ) is obtained from the one for @xmath2 ( resp .
@xmath3 ) by deleting some hyperplanes . following the discussion in appendix [ subs : map ] ,
we get maps @xmath335 . these are semigroup homomorphisms and they induce algebra homomorphisms @xmath336 .
we will use the language of partitions to make these maps more explicit .
c@c figure [ maps ] shows the intersection of the hyperplane arrangements for @xmath337 , @xmath338 and @xmath339 with the boundary of a cube centred at the origin . for @xmath337 and @xmath338 ,
this gives us the corresponding coxeter complex .
for @xmath339 , we do not quite get @xmath340 because the braid arrangement is not essential .
so we cut the braid arrangement by the hyperplane @xmath341 .
the complex @xmath340 can be seen in figure [ maps ] as the dotted hexagon with three vertices each of type @xmath81 and @xmath342 .
also in figure [ maps ] , we have labelled only some of the vertices since they uniquely determine the remaining labels . for the case @xmath2 ( resp .
@xmath3 ) we have split the partition description for a face @xmath344 into 4 ( resp .
3 ) cases ; see appendix [ subs : b ] and [ subs : d ] . from those descriptions , it is straightforward to describe the map
. let @xmath345 . *
a face of type @xmath346 in @xmath332 maps to a face of type @xmath346 in @xmath333 . * a face of type @xmath347 in @xmath332 maps to a face either of type @xmath348 or @xmath349 depending on a parity condition . * a face of type @xmath350 in @xmath332 maps to a face of one higher dimension of type @xmath351 . for @xmath315 ,
this is also clear from figure [ maps ] , where we see that @xmath352 and @xmath353 . *
a face of type @xmath354 in @xmath332 maps to a face of type @xmath351 .
the maps in the first three cases are bijective .
more precisely , the faces of type @xmath346 in @xmath332 are in bijection with the faces of type @xmath346 in @xmath333 and so on .
however in the last case , the map is two to one .
this is again illustrated in figure [ maps ] , where we see that there are two faces of type @xmath355 which map to a face of type @xmath356 .
recall that @xmath78 is the sum of all faces of type @xmath30 . from the above discussion
it follows that for the map @xmath357 , we have @xmath358 , @xmath359 , @xmath360 and @xmath361 .
it is easier to describe this composite map than the map from @xmath333 to @xmath334 .
in contrast to the previous map , this composite map is hard to describe using face types but easy to describe using partitions . to get an ordered partition of @xmath83 $ ] , starting with an anti - symmetric ordered partition of @xmath363 $ ]
, we simply forget the set @xmath364 $ ] .
for example , @xmath365 maps to @xmath366 .
we explain some special cases .
observe that a face of type @xmath81 maps to a face either of type @xmath81 or @xmath367 depending on whether the element in the singleton first block has a plus sign or minus .
it follows that for the map @xmath368 , @xmath82 maps to @xmath369 .
this is the primary reason why the side shuffle of type @xmath2 is more closely related to the two - sided shuffle ( rather than the side shuffle ) of type @xmath1 .
next we show that inverse @xmath107-shuffles map to inverse @xmath107-shuffles . for
that observe that weak @xmath255 block partitions of type @xmath2 correspond to weak @xmath255 block partitions of type @xmath1 .
for example , @xmath370 .
similarly , weak @xmath255 block partitions of type @xmath2 with an empty zero block correspond to weak @xmath256 block partitions of type @xmath1 .
for example , when @xmath371 , we get that @xmath246 maps to @xmath372 . in other words , the inverse riffle shuffle of type @xmath2 maps to the inverse riffle shuffle of type @xmath1 .
we recall the definitions of @xmath43 for the side shuffles of type @xmath2 and @xmath3 and the two - sided shuffle of type @xmath1 .
* @xmath373 where @xmath95 for @xmath229 . * @xmath373 where @xmath95 for @xmath307 and @xmath374 .
also @xmath179 .
* @xmath375 where @xmath175 for @xmath96 and @xmath179 .
we claim that @xmath376 under the maps @xmath336 . from the discussion in section [ subs : map1 ] , it is easy to see the claim for the first map .
a point worth noting is that @xmath377 and the claim holds for @xmath378 because of the relation @xmath179 for @xmath3 .
similarly , using the discussion in section [ subs : map2 ] , we can show the claim for the composite map . in earlier sections ,
the introduction of @xmath103 and the relation @xmath179 for types @xmath3 and @xmath1 was somewhat artificial and justified only from the random walk perspective .
however we now also have a clear geometric perspective .
recall that the shuffles are given by the same formula @xmath379 in all three cases with @xmath111 being the signed stirling numbers .
hence we get @xmath380 . to summarise
, we have : @xmath381 recall the definitions of @xmath43 and @xmath261 for the riffle shuffles of type @xmath2 and @xmath3 from sections [ subs : brs ] and [ subs : drs ] .
following the discussion in section [ subs : map1 ] , it is clear that @xmath343 maps @xmath43 to @xmath43 and @xmath261 to @xmath261 .
so the map restricted to the algebras @xmath274 $ ] is an algebra map that maps a basis to a basis .
hence it is an isomorphism .
it also clearly maps @xmath108 to @xmath108 .
again we see that the somewhat unmotivated definitions of @xmath43 and @xmath261 for @xmath3 have a more natural meaning as the images of the corresponding elements in @xmath332 . for the map @xmath362 , the discussion at the end of section [ subs : map2 ] and the interpretation of the inverse @xmath107-shuffles @xmath108 as weak partitions shows that @xmath382 .
this gives us a surjective map @xmath274 = k[s_2,s_3 ] \rightarrow a = k[s_2]$ ] .
however we do not know a good way to describe the images of @xmath43 and @xmath261 under this map .
to summarise , we have : @xmath383 the coxeter complex of type @xmath2 is probably the correct place to look for shuffle algebras
. the images of these shuffle algebras under our maps would then yield shuffle algebras ( in a suitably generalised sense ) of type @xmath3 and @xmath1 . among the examples in this paper ,
the only one that we failed to describe in this manner was the side shuffle of type @xmath1 .
we first give a brief review of buildings .
for the general theory see .
let @xmath7 be a coxeter group and @xmath384 its coxeter complex . roughly , a building @xmath385 of type @xmath7 is a union of subcomplexes @xmath6 ( called apartments ) which fit together nicely .
each apartment @xmath6 is isomorphic to @xmath384 . as a simplicial complex
, @xmath385 is pure and labelled .
the term _ pure _ means that all maximal simplices ( chambers ) have the same dimension .
for any two simplices in @xmath385 , there is an apartment @xmath6 containing both of them . using this fact
, we can define a product on @xmath385 as follows . for @xmath386 ,
we choose an apartment @xmath6 containing @xmath158 and @xmath387 and define @xmath388 to be their product in @xmath6 . since @xmath6 is a coxeter complex , we know how to do this .
furthermore , it can be shown that the product does not depend on the choice of @xmath6 .
so this defines a product on @xmath385 with the set of chambers @xmath8 contained as an ideal . as in section
[ subs : rw ] , we can now define a random walk on @xmath8 starting with a probability distribution on @xmath385 . unfortunately ,
the product on @xmath385 is not associative , that is , @xmath385 is no longer a semigroup . as a result @xmath16 is not a module over @xmath389 and so our algebraic methods break down
however , we know three examples where they do work
. this is because in these cases , the subalgebra of interest , namely @xmath23 $ ] , turns out to be associative .
a theorem of tits says that if @xmath385 is a finite , irreducible moufang building then it is the building of an absolutely simple algebraic group @xmath390 over a finite field @xmath391 .
the buildings we consider satisfy this hypothesis ; hence in each example we will also specify the algebraic group @xmath390
. the group @xmath390 acts by simplicial type - preserving automorphisms on @xmath385 . to avoid getting into details
, we just mention that the action of @xmath390 is very closely related to the geometry of the building .
the three examples we consider can be thought of as @xmath4-analogues of the side shuffles of type @xmath1 , @xmath2 and @xmath3 .
we use the same definitions for the @xmath43 s as before , except that we now apply them to the building @xmath385 rather than the coxeter complex @xmath6 .
we will use the notation @xmath392 = 1 + q + \ldots + q^{j-1}$ ] to denote the @xmath4-numbers .
they will show up a lot in our analysis .
we first briefly describe the building of type @xmath1 associated to the algebraic group @xmath393 ( also @xmath394 ) .
let @xmath395 be the @xmath5-dimensional vector space over @xmath391 and @xmath396 be the lattice of subspaces of @xmath395 .
the building of type @xmath1 associated to @xmath393 is simply the flag ( order ) complex @xmath397 .
it is a labelled simplicial complex .
a vertex of @xmath397 is by definition a proper subspace of @xmath395 and we label it by its dimension . to be consistent with earlier notation , we label the vertices @xmath398 instead of just @xmath399 .
in particular , vertices of type @xmath81 are the one dimensional subspaces of @xmath395 .
let @xmath400 be the boolean lattice of rank @xmath5 consisting of all subsets of an @xmath5-set ordered under inclusion .
note that the order complex @xmath401 is the coxeter complex of type @xmath1 , which we earlier denoted by @xmath334 .
also note that a choice of a basis for @xmath395 gives an embedding of @xmath402 into @xmath403 .
the subcomplexes @xmath401 , for various embeddings @xmath404 , play the role of apartments .
alternatively , it is also useful to think of @xmath402 as a specialisation of @xmath403 for the degenerate case @xmath405 . in this sense @xmath397
may be regarded as a @xmath4-analogue of @xmath406 . for the rest of the section
we will denote the building @xmath397 by @xmath385 .
the product in @xmath385 can be made more explicit .
a face of @xmath385 is a chain in @xmath403 and a chamber is a maximal chain in @xmath403 . for @xmath407 ,
the face @xmath388 is the chain in @xmath403 obtained by refining the chain @xmath158 by the chain @xmath387 ( using meets and joins as in a jordan hlder product ) .
this is a generalisation of the product in @xmath334 which we defined in terms of refinement of one ordered partition by another . the side shuffle of type @xmath1 ( section [ subs : ass ] ) generalises in a straightforward way to this setting .
let @xmath41 be the sum of all the vertices of type @xmath81 , @xmath42 be the sum of all the edges of type @xmath97 and so on till @xmath98 which is the sum of all chambers of @xmath385 .
the number of summands in @xmath99 , which is same as the number of faces of type @xmath408 , is @xmath83[n-1 ] \ldots [ n - j+1]$ ] .
it follows directly from the definition that @xmath409 \sigma_j + q^j \sigma_{j+1}$ ] if @xmath138 and @xmath410 \sigma_{n-1}$ ] . as a consistency check , we can also verify that the number of terms on each side is the same .
these facts imply that @xmath41 is power associative .
hence we may write the following without ambiguity . @xmath411)\ldots(\sigma_1-[j]).\ ] ] this along with @xmath410 \sigma_{n-1}$ ] implies that both @xmath122,@xmath41, ,@xmath98 and @xmath196,@xmath153, ,@xmath154 form a basis for the associative algebra @xmath45 $ ] . the basic relation satisfied by @xmath41 is @xmath412)\ldots(\sigma_1-[n-2])(\sigma_1-[n ] ) = 0 $ ] .
hence @xmath157}{x(x-[1])\ldots(x-[n-2])(x-[n])}$ ] , where @xmath41 maps to @xmath158 .
since the roots are distinct we conclude that @xmath39 is split semisimple . as before ,
we extend the definition of @xmath43 to any @xmath33 using to obtain @xmath101 and @xmath160 for @xmath161 .
inverting equation formally , we get @xmath163 where @xmath111 satisfies the recursion : @xmath413 s(a-1,j ) + q^{j-1 } s(a-1,j-1)$ ] with @xmath414 and @xmath415 .
the numbers @xmath111 give a @xmath4-analogue to the stirling numbers of the second kind .
for more information on these numbers , see and the references therein .
the analysis we gave for this example was the analogue of the direct method of section [ subs : ass ] .
the difficulty with the shuffle method is that we do not know of a good random walk interpretation of @xmath166 .
however as the direct method shows , the algebra @xmath39 satisfies all the properties of an additive shuffle algebra ( except that now we are in the more general context of buildings ) .
there are also natural ways to define the q - analogue of the two - sided shuffle and the riffle shuffle .
however the analysis breaks down because the basic elements of interest are not power associative .
there are two possibilities for the building depending on whether the associated algebraic group is the symplectic group @xmath416 or the orthogonal group @xmath417 .
we consider them separately .
they may both be regarded as @xmath4-analogues of @xmath332 . the side shuffle for @xmath311 generalises to both cases .
as expected , the computations have the same spirit as the previous section . * the symplectic case .
* let @xmath395 be a ( even dimensional ) vector space over @xmath391 with a skew - symmetric non - degenerate bilinear form @xmath418 .
more explicitly , @xmath395 has a basis @xmath419 such that @xmath420 , @xmath421 and all other pairings between basis vectors are zero .
the symplectic group @xmath416 is the group of automorphisms of @xmath395 that preserve the bilinear form @xmath418 .
the building of type @xmath311 associated to the symplectic group is the flag ( order ) complex of all isotropic subspaces of @xmath395
. we will denote it by @xmath385 .
the definition of the product in @xmath385 is similar to that in section [ subs : qass ] .
the difference occurs due to the fact that the sum of two isotropic subspaces may not be isotropic .
let @xmath41 be the sum of all the vertices of type @xmath81 , @xmath42 be the sum of all the edges of type @xmath97 and so on till @xmath230 which is the sum of all chambers of @xmath385 .
it follows from the definition that @xmath422 \sigma_j + q^j \sigma_{j+1}$ ] if @xmath237 and @xmath423 \sigma_{n}$ ] .
an easy way to check this is to first figure out the coefficient of @xmath146 .
it is @xmath424 , the same answer that we got for the @xmath1 case .
there is no difference in the two situations with respect to this coefficient . with this information
, we can find the coefficient of @xmath99 by simply counting the number of terms involved on both sides .
the number of summands in @xmath99 , which is same as the number of faces of type @xmath408 , is @xmath425[2n-2 ] \ldots [ 2n-2j+2]$ ] .
the identity is then a consequence of the simple relation @xmath425 = ( 1+q^{2n - j})[j ] + q^j [ 2n-2j]$ ] .
it is a good exercise to compute the coefficient of @xmath43 directly .
these facts imply that @xmath41 is power associative .
hence we may write the following without ambiguity .
@xmath426)\ldots(\sigma_1-(1+q^{2n - j})[j]).\ ] ] this along with @xmath427 \sigma_{n}$ ] implies that both @xmath122,@xmath41, ,@xmath230 and @xmath196,@xmath153, ,@xmath428 form a basis for the associative algebra @xmath45 $ ] .
the basic relation satisfied by @xmath41 is @xmath429)\ldots(\sigma_1-(1+q^{n+1})[n-1])(\sigma_1-[2n ] ) = 0 $ ] .
hence @xmath157}{x(x-(1+q^{2n-1})[1])\ldots(x-(1+q^{n+1})[n-1])(x-[2n])}$ ] .
this shows that @xmath39 is split semisimple . as before
, we extend the definition of @xmath43 to any @xmath33 using to obtain @xmath160 for @xmath161 . inverting equation formally , we get @xmath163 where @xmath111 satisfies the recursion : @xmath430 s(a-1,j ) + q^{j-1 } s(a-1,j-1)$ ] with @xmath431 and @xmath415 . the numbers @xmath111 give a @xmath4-analogue to the signed stirling numbers .
note that they now depend on @xmath5 .
we do not know whether they have been considered before or whether they have an explicit formula . * the orthogonal case .
* let @xmath395 be a ( even dimensional ) vector space over @xmath391 with a symmetric non - degenerate bilinear form @xmath418 .
we also assume that @xmath395 has an isotropic subspace of dimension @xmath5 .
more explicitly , @xmath395 has a basis @xmath419 such that @xmath432 and all other pairings between basis vectors are zero .
the orthogonal group @xmath417 is the group of automorphisms of @xmath395 that preserve the bilinear form @xmath418 .
the building of type @xmath311 associated to the orthogonal group is the flag ( order ) complex of all isotropic subspaces of @xmath395
. we will denote it by @xmath433 .
the analysis parallels the symplectic case . with the same definitions of the @xmath43 s
, we get @xmath434 \sigma_j + q^j \sigma_{j+1}$ ] if @xmath237 and @xmath435 \sigma_{n}$ ] .
the reason for the difference is that now the number of summands in @xmath99 is @xmath436(1+q^{n-2})[n-1 ] \ldots ( 1+q^{n - j})[n - j+1]$ ] .
the rest is similar to the symplectic case .
we only point out that the stirling numbers now satisfy the recursion : @xmath437 s(a-1,j ) + q^{j-1 } s(a-1,j-1)$ ] with @xmath438 and @xmath415 .
the building @xmath433 that we considered here is not thick .
classically , the group @xmath417 is considered to be of type @xmath3 since there is a thick building @xmath439 associated to it .
we will discuss this next . for more information on this ,
see .
let @xmath395 and @xmath418 be as in the orthogonal case of section [ subs : qbss ] .
the building of type @xmath3 , which we denote by @xmath439 , is the flag complex of the following so - called `` oriflamme geometry '' .
the vertices in @xmath439 are the non - zero isotropic subspaces of @xmath395 of dimension @xmath440 .
two such subspaces are called incident if one is contained in the other or if both have dimension @xmath5 and their intersection has dimension @xmath292 . a simplex in @xmath439 is a set of pairwise incident vertices . there is a natural product preserving map @xmath441 , which generalises the map @xmath343 in section [ subs : map1 ] . to describe this map ,
we identify the vector space @xmath395 and the bilinear form @xmath418 in the two cases .
a vertex in @xmath433 is a non - zero isotropic subspace of @xmath395 .
if it has dimension @xmath442 then we map it to itself .
otherwise there are exactly two maximal isotropic subspaces of dimension @xmath5 in @xmath395 that contain it .
and we map it to the edge joining the two vertices in @xmath439 representing these two subspaces .
this again explains how a vertex of type @xmath367 maps to an edge of type @xmath443 . in the orthogonal case of section [ subs : qbss ]
, we have the shuffle algebra @xmath105 $ ] in @xmath433 . the image of this shuffle algebra under the above map yields a shuffle algebra in @xmath439 .
as a @xmath4-analogue of the map between side shuffles of type @xmath2 and @xmath3 , we get a map @xmath444 the @xmath43 s on the right are defined as for the side shuffle of type @xmath3 and we again get the relation @xmath179 .
the basic relation satisfied by @xmath41 is @xmath445)\ldots(\sigma_1-(1+q^{n+1})[n-2])(\sigma_1-(1+q^{n-1})[n ] ) = 0 $ ] .
note that the term corresponding to @xmath292 is absent .
the rest of the analysis is routine and we omit it . in light of the discussion in section [ s : maps ] , we may say the following .
the failure to get a @xmath4-two sided shuffle of type @xmath1 is the result of our failure to define a product preserving map @xmath446 .
also we failed to get a @xmath4-riffle shuffle of type @xmath2 and as a result also failed on the other two fronts .
we conclude by suggesting some problems for future consideration . one problem is to classify them with minor modifications of the definition if necessary ( see remark in section [ subs : defn ] ) .
also it is natural to ask whether there is a sensible generalisation to the case of infinite coxeter groups .
there is no good theory for random walks on buildings the major difficulty being the non - associativity of the product .
our methods in section [ s : q ] suggest that it would be worthwhile to systematically understand the power associative elements of a building and also to classify product preserving maps between buildings . in this paper , we considered only real reflection groups .
the difficulty in passing to the complex case is the absence of an analogue of the coxeter complex . as an example , consider the complex reflection group @xmath447 . for @xmath448 , it specialises to the coxeter group of type @xmath2 .
the first step would be to generalise the semigroup of ordered partitions of type @xmath2 .
however , it is not clear how to do this . in ,
brown defines a random walk associated to matroids . and
he relates the eigenvalue multiplicities to invariants of the lattice of flats .
he calls these invariants the generalised derangement numbers .
the motivating examples are the side and @xmath4-side shuffle of type @xmath1 and the multiplicities then are the usual and @xmath4-derangement numbers . in this sense , we may think of ordinary matroids as related to the coxeter group of type @xmath1 .
there is a notion of a @xmath7-matroid for any coxeter group @xmath7 .
we ask whether it is possible to generalise the above to this setting . as a positive result in this direction
, bidigare shows that for the side shuffle of type @xmath2 , the multiplicities are the signed derangement numbers .
most of this section and a part of the next is taken directly from .
more details concerning the material reviewed here can be found in . throughout this section
@xmath449 denotes a finite set of affine hyperplanes in @xmath450 .
let @xmath451 and @xmath452 be the two open halfspaces determined by @xmath453 ; the choice of which one to call @xmath451 is arbitrary but fixed .
the hyperplanes @xmath453 induce a partition of @xmath395 into convex sets called _ faces _ ( or _ relatively open faces _ ) .
these are the nonempty sets @xmath454 of the form @xmath455 where @xmath456 and @xmath457 .
equivalently , if we choose for each @xmath203 an affine function @xmath458 such that @xmath453 is defined by @xmath459 , then a face is a nonempty set defined by equalities and inequalities of the form @xmath460 , @xmath461 , or @xmath459 , one for each @xmath462 .
the sequence @xmath463 that encodes the definition of @xmath344 is called the _ sign sequence _ of @xmath344 and is denoted @xmath464 . the faces such that @xmath465 for all @xmath203 are called _
chambers_. they are convex open sets that partition the complement @xmath466 . in general , a face @xmath344 is open relative to its _ support _ , which is defined to be the affine subspace @xmath467 since @xmath344 is open in @xmath468 , we can also describe @xmath468 as the affine span of @xmath344 .
the _ face poset _ of @xmath469 is the set @xmath470 of faces , ordered as follows : @xmath471 if for each @xmath462 either @xmath472 or @xmath473 . in other words
, the description of @xmath344 by linear equalities and inequalities is obtained from that of @xmath390 by changing zero or more inequalities to equalities .
we say that @xmath344 is a face of @xmath390 .
note that the chambers are precisely the maximal elements of the face poset .
we denote the set of chambers by @xmath8 .
the set @xmath470 of faces is also a semigroup .
given @xmath474 , their _ product _ @xmath475 is the face with sign sequence @xmath476 this has a geometric interpretation : if we move on a straight line from a point of @xmath344 toward a point of @xmath390 , then @xmath475 is the face we are in after moving a small positive distance .
hence , we may think of @xmath475 as the projection of @xmath390 on @xmath344 .
notice that the face relation can be described in terms of the product : one has @xmath477 also note that the set of chambers @xmath8 is an ideal of @xmath470 .
a second poset associated with the arrangement @xmath469 is the _ semilattice of flats _ , also called the _ intersection semilattice _ , which we denote by @xmath478 .
it consists of all nonempty affine subspaces @xmath479 of the form @xmath480 , where @xmath481 is an arbitrary subset ( possibly empty ) .
we order @xmath478 by inclusion .
[ warning : many authors order @xmath478 by reverse inclusion . ]
notice that any two elements @xmath482 have a least upper bound @xmath483 in @xmath478 , which is the intersection of all hyperplanes @xmath484 containing both @xmath485 and @xmath486 ; hence @xmath478 is an _ upper semilattice _ ( poset with least upper bounds ) .
it is a lattice if the arrangement @xmath469 is _ central _ ,
i.e. , if @xmath487 .
indeed , this intersection is then the smallest element of @xmath478 , and a finite upper semilattice with a smallest element is a lattice .
the support map gives a surjection @xmath488 which preserves order and also behaves nicely with respect to the semigroup structure .
namely , we have @xmath489 and @xmath490 let @xmath449 and @xmath491 be two hyperplane arrangements such that @xmath492 , that is , the arrangement @xmath493 is obtained from @xmath469 by deleting some of the hyperplanes .
let @xmath494 and @xmath495 be the respective hyperplane face semigroups .
then there is a natural map @xmath496 .
an element of @xmath494 is a face @xmath344 with a sign sequence , say @xmath497 .
we map it to the face of @xmath493 with the sign sequence @xmath498 . in other words , we forget the signs of the elements of @xmath31 that are not in @xmath499 .
it is clear that this map is a semigroup homomorphism and preserves the face relation .
the second fact is implied by the first in view of equation .
suppose now that @xmath469 is a _
arrangement , i.e. , that the hyperplanes have a nonempty intersection .
we may assume that this intersection contains the origin .
suppose further that @xmath500 , in which case @xmath469 is said to be _
essential_. ( there is no loss of generality in making this assumption ; for if it fails , then we can replace @xmath395 by the quotient space @xmath501 . ) the hyperplanes then induce a cell - decomposition of the unit sphere , the cells being the intersections with the sphere of the faces @xmath502 .
thus @xmath470 , as a poset , can be identified with the poset of cells of a regular cell - complex @xmath6 , homeomorphic to a sphere .
note that the face @xmath503 , which is the identity of the semigroup @xmath470 , is not visible in the spherical picture ; it corresponds to the empty cell .
the cell - complex @xmath6 plays a crucial role in , to which we refer for more details .
thus the cell - complex @xmath6 is a semigroup which we call the hyperplane face semigroup
. the maximal cells ( which we again call chambers and denote @xmath8 ) is an ideal in @xmath6 .
we work with an arbitrary finite coxeter group @xmath7 and its associated hyperplane face semigroup @xmath6 ( the coxeter complex of @xmath7 ) . but we will explain everything in concrete terms for the cases @xmath504 . this should make the discussion accessible to readers unfamiliar with coxeter groups .
we begin with a very quick review of the basic facts that we need about finite coxeter groups and their associated simplicial complexes @xmath6 .
details can be found in many places , such as .
a _ finite reflection group _ on a real inner - product space @xmath395 is a finite group of orthogonal transformations of @xmath395 generated by reflections @xmath505 with respect to hyperplanes @xmath506 through the origin .
the set of hyperplanes @xmath506 such that @xmath507 is the _ reflection arrangement _ associated with @xmath7 .
its hyperplane face semigroup @xmath6 is called the _ coxeter complex _ of @xmath7 . geometrically ,
this complex is obtained by cutting the unit sphere in @xmath395 by the hyperplanes @xmath506 , as in section [ sub : cell ] .
( as explained there , one might have to first pass to a quotient of @xmath395 . )
it turns out that the coxeter complex @xmath6 is always a simplicial complex .
furthermore , the action of @xmath7 on @xmath395 induces an action of @xmath7 on @xmath6 , and this action is simply - transitive on the chambers .
thus the set @xmath8 of chambers can be identified with @xmath7 , once a `` fundamental chamber '' @xmath508 is chosen .
the number @xmath509 of vertices of a chamber of @xmath6 is called the _ rank
_ of @xmath6 ( and of @xmath7 ) ; thus the dimension of @xmath6 as a simplicial complex is @xmath510 . it is known that one can color the vertices of @xmath6 with @xmath509 colors in such a way that vertices connected by an edge have distinct colors .
the color of a vertex is also called its _ label _ , or its _ type _ , and we denote by @xmath31 the set of all types .
we can also define @xmath511 for any @xmath512 ; it is the subset of @xmath31 consisting of the types of the vertices of @xmath344 .
for example , every chamber has type @xmath31 , while the empty simplex has type @xmath513 .
the action of @xmath7 is type - preserving ; moreover , two simplices are in the same @xmath7-orbit if and only if they have the same type .
choose a fundamental chamber @xmath508 .
it is known that the reflections @xmath514 in the facets of @xmath508 generate @xmath7 .
in fact , @xmath7 has a presentation of the form @xmath515 with @xmath516 and @xmath517 .
this data is conveniently encoded in a picture called the coxeter diagram of @xmath7 .
this diagram is a graph , with vertices and edges , defined as follows : there are @xmath509 vertices , one for each generator @xmath518 , and the vertices corresponding to @xmath203 and @xmath33 are connected by an edge if and only if @xmath519 .
if @xmath520 then we simply label the edge with the number @xmath521 .
the figures show the coxeter diagrams which are of interest to us , namely the ones of type @xmath1 , @xmath2 and @xmath3 .
it is customary to use the generators of @xmath7 , or the vertices of the coxeter diagram to label the vertices of its coxeter complex @xmath6 .
a vertex of the fundamental chamber @xmath508 is labelled @xmath514 if it is fixed by all the fundamental reflections except @xmath514 .
since @xmath7 acts transitively on @xmath8 and the action is type - preserving , this determines the type of all the vertices of @xmath6 .
units < 1 mm , 1 mm > x from 0 to 80 , y from -10 to 20 at 40 -5 360 degrees from 11 10 center at 10 10 at 10 5 11 10 29 10 / 360 degrees from 31 10 center at 30 10 at 30 5 31 10 45 10 / at 50 10 55 10 69 10 / 360 degrees from 71 10 center at 70 10 at 70 5 units < 1 mm , 1 mm > x from 0 to 80 , y from -10 to 20 at 50 -5 360 degrees from 11 10 center at 10 10 at 10 5 11 10 29 10 / 360 degrees from 31 10 center at 30 10 at 30 5 31 10 45 10 / at 50 10 55 10 69 10 / 360 degrees from 71 10 center at 70 10 at 70 5 71 10 89 10 / at 80 15 360 degrees from 91 10 center at 90 10 at 90 5 units < 1 mm , 1 mm > x from 0 to 80 , y from -10 to 25 at 50 -10 360 degrees from 11 10 center at 10 10 at 10 5 11 10 29 10 / 360 degrees from 31 10 center at 30 10 at 30 5 31 10 45 10 / at 50 10 55 10 69 10 / 360 degrees from 71 10 center at 70 10 at 70 5 360 degrees from 86 20 center at 85 20 360 degrees from 86 0 center at 85 0 71 10.5 84 19.5 / 71 9.5 84 0.5 / at 87 16 at 87 -4 the coxeter group @xmath522 acts on @xmath523 by permuting the coordinates .
the arrangement in this case is the _ braid arrangement _ in @xmath523 .
it is discussed in detail in .
it consists of the @xmath524 hyperplanes @xmath525 defined by @xmath330 , where @xmath526 .
each chamber is determined by an ordering of the coordinates , so it corresponds to a permutation .
the faces of a chamber are obtained by changing to equalities some of the inequalities defining that chamber .
we fix @xmath527 to be the fundamental chamber @xmath508 .
the supports of the facets of @xmath508 are hyperplanes of the form @xmath528 , where @xmath529 .
the reflection in the hyperplane @xmath528 correponds to the generator @xmath514 that interchanges the coordinates @xmath530 and @xmath531 .
the chamber @xmath508 has @xmath292 vertices , namely the labels @xmath81,@xmath342,@xmath534,@xmath367 are assigned by the rule mentioned in subsection [ subs : cd ] .
applying the action of @xmath7 we see , for example , that @xmath536 gives all vertices of type @xmath81 as @xmath537 varies over all permutations of @xmath83=\{1,\dots , n\}$ ] .
we encode the system of equalities and inequalities defining a face @xmath344 by an ordered partition @xmath538 of @xmath83 $ ] . here
@xmath539 are disjoint nonempty sets whose union is @xmath83 $ ] , and their order counts .
thus the simplices of @xmath6 are ordered partitions @xmath540 of the set @xmath83 $ ] .
for example , the vertices of type @xmath81 are ordered two block partitions such that the first block is a singleton .
the product in @xmath6 is also easy to describe .
we multiply two ordered partitions by taking intersections and ordering them lexicographically ; more precisely , if @xmath540 and @xmath541 , then @xmath542 where the hat means `` delete empty intersections '' .
the 1-block ordered partition is the identity .
note that @xmath186 is a face of @xmath508 if and only if @xmath508 consists of an ordered partition of @xmath543 followed by an ordered partition of @xmath544 , and so on , that is , if and only if @xmath508 is a refinement of @xmath186 .
the chambers are the ordered partitions into singletons , so they correspond to the permutations of @xmath83 $ ] or a deck of @xmath5 cards .
the group of signed permutations @xmath545 acts on @xmath523 with the subgroup @xmath0 permuting the coordinates and the subgroup @xmath546 flipping the signs of the coordinates .
the reflection arrangement in this case consists of the hyperplanes defined by @xmath329 and @xmath547 , where @xmath526 .
a chamber is given by an ordering of the coordinates , their negatives and zero .
for example , for @xmath548 , @xmath549 specifies a chamber .
note that the inequalities that appear on the left of @xmath550 completely determine a chamber ( and the same is true for any face since it is obtained by changing to equalities some of the inequalities defining a chamber ) .
thus we see that a chamber corresponds to a signed permutation .
we fix @xmath551 to be the fundamental chamber @xmath508 .
the supports of the facets of @xmath508 are hyperplanes of the form @xmath528 , where @xmath529 and @xmath552 .
the generators @xmath514 interchange the coordinates @xmath530 and @xmath531 and the generator @xmath249 flips the sign of @xmath553 .
the @xmath5 vertices of @xmath508 along with their labels are as follows .
it is convenient to describe a face @xmath344 by an anti - symmetric ordered partition @xmath560 of @xmath363=\{1,\dots , n,\overline{1},\ldots,\overline{n}\}$ ] .
we will call this a _ partition of type @xmath2_. for example , for @xmath548 , the face @xmath561 is written @xmath365 .
the set @xmath562 satisfies @xmath563 and is allowed to be empty .
we call it the zero block .
it is the only set that is allowed to contain both a number and its negative .
also , the sets @xmath539 are necessarily non - empty .
we also define a _ weak partition of type @xmath2 _ to be a partition as above but where the sets @xmath539 are allowed to be empty .
note that for a face @xmath344 , the zero block of its partition is empty if and only if the type of @xmath344 contains the letter @xmath249 .
we split the description for a face @xmath344 into four cases , depending on whether the type of @xmath344 contains * neither @xmath367 nor @xmath249 : + an ordered partition @xmath560 of @xmath363 $ ] of type @xmath2 with the added restriction that the zero block @xmath562 has at least @xmath564 elements .
* @xmath249 but not @xmath367 : + an ordered partition @xmath560 of @xmath363 $ ] , with the restriction that the zero block is empty and @xmath565 has at least @xmath92 elements . *
@xmath367 but not @xmath249 : + an ordered partition @xmath560 of @xmath363 $ ] , with the restriction that the zero block has exactly @xmath92 elements . *
both @xmath367 and @xmath249 : + an ordered partition @xmath560 of @xmath363 $ ] , with the restriction that the zero block is empty and @xmath565 has at exactly @xmath199 element .
the product in @xmath6 is defined exactly as for the @xmath1 case , where we refine the first partition by the second .
note that @xmath186 is a face of @xmath508 if and only if @xmath508 is a refinement of @xmath186 .
as mentioned earlier , the second half of the partition following @xmath562 contains no new information .
hence we may describe a @xmath566 dimensional face by a @xmath567 block partition @xmath568 of @xmath83 $ ] .
we think of @xmath539 as signed sets and @xmath562 ( possibly empty ) as an unsigned set . the chambers are then @xmath569 block partitions into @xmath5 singletons and an empty zero block , so they correspond to signed permutations of @xmath83 $ ] or a deck of @xmath5 signed cards
. we will call such a deck as a _ deck of type @xmath2_. however , the product and the face relation is not so natural now .
so we will mainly use the first description .
the group of even signed permutations @xmath570 is an index 2 subgroup of the group of signed permutations @xmath571 . here
@xmath390 is the index 2 subgroup of @xmath546 consisting of @xmath5-tuples that have an even number of negative signs .
the group @xmath7 acts on @xmath523 with the subgroup @xmath0 permuting the coordinates and the subgroup @xmath390 flipping the signs of the coordinates .
the reflection arrangement in this case consists of the hyperplanes defined by @xmath329 , where @xmath526 .
it is obtained from the reflection arrangement of type @xmath2 by deleting the coordinate hyperplanes @xmath572 for @xmath573 . on the level of chambers ,
this has the effect of merging together pairs of adjacent chambers of the coxeter complex of type @xmath2 ; see figure [ maps ] .
for example , the chamber of the coxeter complex of type @xmath3 , @xmath574 ( which we fix as our fundamental chamber @xmath508 ) is the union of the two chambers of the coxeter complex of type @xmath2 , namely , @xmath551 and @xmath575 .
the supports of the facets of @xmath508 are hyperplanes of the form @xmath528 , where @xmath529 and @xmath576 .
the generators @xmath514 interchange the coordinates @xmath530 and @xmath531 , the generator @xmath317 interchanges the coordinates @xmath577 and @xmath578 and the generator @xmath318 interchanges @xmath577 and @xmath553 and flips the signs of both .
the @xmath5 vertices of @xmath508 along with their labels are as follows . applying the group action ,
we see , for example , that the vertices of type @xmath317 and @xmath318 both have the form @xmath582 , where @xmath559 and they are distinguished by the parity of the product @xmath583 .
also observe that the edge of the fundamental chamber @xmath508 of type @xmath584 is given by @xmath585 . as in the @xmath2 case
, we describe a face @xmath344 by an anti - symmetric ordered partition @xmath586 of @xmath363 $ ] .
we repeat that the sets @xmath539 are non - empty and can not contain both a number and its negative .
we call the block @xmath587 in the middle , the central block rather than the zero block
. it always has an even number of elements .
furthermore , we impose the relation @xmath588 where @xmath589 is empty , @xmath508 has exactly two elements , namely , a number and its negative , and @xmath590 .
we split the description for a face @xmath344 into three cases , depending on whether the type of @xmath344 contains * neither @xmath317 nor @xmath318 : + an ordered partition @xmath586 of @xmath363 $ ] , with the restriction that the central block @xmath508 ( in this case , we may think of it as the zero block ) has at least @xmath564 elements . *
exactly one of @xmath317 and @xmath318 : + an ordered partition @xmath586 of @xmath363 $ ] , with the restriction that @xmath508 is empty and @xmath565 has at least @xmath92 elements . * both @xmath317 and @xmath318 : + an ordered partition @xmath591 of @xmath363 $ ] , with the restriction that @xmath508 has exactly @xmath92 elements , namely , a number and its negative . + an ordered partition @xmath592 where @xmath589 is empty and @xmath593 has exactly @xmath199 element . here
the number @xmath14 is always the rank of the face .
the first three descriptions are obtained directly from the equalities and inequalities that define a face .
the reason why we call @xmath508 the central block rather than the zero block should be clear now .
the motivation for the second description in case @xmath594 is as follows . to give an example
, the product of @xmath595 with @xmath596 , must be @xmath597 .
but if we refine the first partition by the second then we get @xmath598 .
this partition has two singletons in the middle and does not fit any of the first three descriptions given above .
so we allow ourselves to merge the two singletons in the middle .
this identification allows us to multiply two partitions @xmath186 and @xmath508 in the same way as before ; that is , we refine @xmath186 using @xmath508 . as in the @xmath2 case , if we throw off the ( redundant ) second half of the partition then we see that chambers correspond to an _ almost _ signed deck of cards or a _ deck of type @xmath3_. it is a deck in which every card , except the bottomost , is signed .
we also define a _ weak partition of type @xmath3 _ to be a partition of type @xmath3 where the sets @xmath539 are allowed to be empty .
it is a pleasure to thank my advisor ken brown for introducing me to this area of mathematics and sharing his insights with me .
i would also like to thank marcelo aguiar for giving valuable suggestions .
t. patrick bidigare , phil hanlon , and daniel n. rockmore , _ a combinatorial description of the spectrum for the tsetlin library and its generalization to hyperplane arrangements _ , duke math .
j. * 99 * ( 1999 ) , no . 1 , 135174 .
anders bjrner , michel las vergnas , bernd sturmfels , neil white , and gnter m. ziegler , _ oriented matroids _
, encyclopedia of mathematics and its applications , vol .
46 , cambridge university press , cambridge , 1993 .
persi diaconis , _ group representations in probability and statistics _ , institute of mathematical statistics lecture notes monograph series , vol . 11 , institute of mathematical statistics , hayward , ca , 1988 . richard p. stanley , _ enumerative combinatorics .
vol . 1 _ , cambridge studies in advanced mathematics , vol . 49 , cambridge university press , cambridge , 1997 , with a foreword by gian - carlo rota , corrected reprint of the 1986 original
|
the random - to - top and the riffle shuffle are two well - studied methods for shuffling a deck of cards .
these correspond to the symmetric group @xmath0 , i.e. , the coxeter group of type @xmath1 . in this paper
, we give analogous shuffles for the coxeter groups of type @xmath2 and @xmath3 .
these can be interpreted as shuffles on a `` signed '' deck of cards . with these examples as motivation , we abstract the notion of a shuffle algebra which captures the connection between the algebraic structure of the shuffles and the geometry of the coxeter groups .
we also briefly discuss the generalisation to buildings which leads to @xmath4-analogues .
|
precision tests of the equivalence principle ( ep ) in different physical phenomena is of fundamental interest .
this statement is especially valid for testing ep in case of a quantum motion of antiatoms .
detailed studies of gravitational properties of antimatter are planned by most groups involved in experiments with antihydrogen ( @xmath0 ) @xcite . here
we discuss a possibility of exploring gravitational states of @xmath0 @xcite with potentially very precise spectroscopic methods .
we study the behaviour of @xmath0 bounded in the gravitational field near a material surface under the influence of alternating magnetic field , with the frequency adjusted to induce resonance transitions between lowest gravitational states .
this approach gives access to the gravitational energy level spacing and thus to the gravitational mass of @xmath0 .
in this section we will study a motion of an @xmath0 atom , localized in the gravitational state near a horizontal plane mirror , under the influence of the alternating magnetic field .
the interaction of magnetic field with _ a moving through the field _ ground state @xmath1 atom @xcite is dominated by interaction of an average magnetic moment of an atom @xcite in given hyperfine state with the magnetic field .
we will focus on the alternating magnetic field with a gradient in the vertical direction .
this condition is needed for coupling the field and the center of mass ( c.m ) @xmath0 motion in the gravitational field of the earth .
it allows one to induce resonant transitions between quantum gravitational states of @xmath0 @xcite .
such states have similar properties with those , discovered for neutrons @xcite , in particular they are characterised by the following energy ( @xmath2 ) and spatial ( @xmath3 ) values : @xmath4{\frac{\hbar^2m^2g^2}{2 m } } \left(\lambda_n+\delta \right),\\ & h_n&=\sqrt[3]{\frac{\hbar^{2}}{2mmg } } \lambda_n , \label { hn } \\ & \mathop{\rm ai}(&-\lambda_n)=0.\end{aligned}\ ] ] here @xmath5 is the gravitational mass of @xmath0 , @xmath6 is the inertial mass of @xmath0 ( we distinguish between @xmath5 and @xmath6 in view of discussing ep test ) , @xmath7 is a gravitational field intensity near the earth surface , @xmath8 is an airy function @xcite , @xmath9 is a universal complex shift of gravitational levels due to the interaction with a material ( perfectly conducting ) wall @xcite .
all the states acquire equal width depending on material surface substance @xmath10{\frac{\hbar^2m^2g^2}{2m}}|\delta|$ ] .
this width corresponds to the lifetime of @xmath11 s in case of a perfectly conducting surface and is twice larger for silica @xcite .
we will consider the magnetic field in the form : @xmath12 here @xmath13 is an amplitude of a constant , vertically aligned , component of magnetic field , @xmath14 is the value of magnetic field gradient , @xmath15 is a distance measured in a vertical direction , @xmath16 is a distance measured in horizontal direction , parallel to a surface of a mirror . time - varying magnetic field ( [ magn1 ] ) is accompanied with an electric field ( @xmath17=-\frac{1}{c } \partial \vec{b}/\partial t$ ] ) .
however , for the ultracold atom velocities , corresponding interaction terms are small and thus will be omitted .
inhomogeneous magnetic field couples the spin and spatial degrees of freedom .
@xmath0 wave - function in this case is described by a four - component column ( in a nonrelativistic treatise ) in the spin space , each component being a function of the c.m .
coordinate @xmath18 , relative @xmath19 coordinate @xmath20 and time @xmath21 .
the corresponding schrdinger equation is : @xmath22_{\alpha , \alpha ' } \phi_{\alpha'}(\vec{r},\vec{\rho},t).\ ] ] subscript @xmath23 indicates one of four spin states of @xmath19 system .
the meaning of the interaction terms is the following .
@xmath24 is a atom - mirror interaction potential , which at asymptotic atom - mirror distance turns into the van der waals/ casimir - polder potential ( see @xcite and references there in ) .
@xmath25 is the hamiltonian of interatomic motion , which includes hyperfine interaction .
@xmath26 here @xmath27 , @xmath28 is the antiproton mass , @xmath29 is the positron mass , @xmath30 , @xmath31 is the hyperfine constant , @xmath32 is the operator of the total spin of antiproton and positron .
we will treat only @xmath0 atoms in a @xmath33-state ( below we will show that excitation of other states in a studied process is improbable ) .
the term @xmath34 is a model operator , which effectively accounts for the hyperfine interaction and reproduces the hyperfine energy splitting correctly .
the term @xmath35 describes the field - magnetic moment interaction : @xmath36 here @xmath37 and @xmath38 are magnetic moments of positron and antiproton respectively .
as far as the field @xmath39 changes spatially and temporally , this term couples the spin and the c.m . motion .
we will assume that in typical conditions of a spectroscopy experiment @xmath0 velocity @xmath40 parallel to the mirror surface ( directed along @xmath16-axis ) is of order of a few @xmath41 and is much larger than a typical vertical velocity in lowest gravitational states ( which is of order of @xmath42 ) .
we will treat motion in a frame moving with the velocity @xmath40 of the @xmath0 atom along the mirror surface .
thus we will consider the x - component motion as a classical motion with a given velocity @xmath40 and will substitute @xmath16-dependence by @xmath21-dependence .
we will also assume that @xmath43 , where @xmath44 cm is a typical size of an experimental installation of interest .
this condition is needed for `` freezing '' the magnetic moment of an atom along the vertical direction ; it provides the maximum transition probability
. we will be interested in the weak field case , such that the zeeman splitting is much smaller than the hyperfine level spacing @xmath45 .
the hierarchy of all mentioned above interaction terms can be formulated as follows : @xmath46 and thus it assumes using the adiabatic expansion for solving eq.([schr ] ) ; it is based on the fact that an internal state of an @xmath0 atom follows adiabatically the spatial and temporal variations of external magnetic field . neglecting non - adiabatic coupling , an equation system for the amplitude @xmath47 of a gravitational state @xmath48 has the form : @xmath49 transition frequency @xmath50 is determined by gravitational energy level spacing .
this fact is used in the proposed approach to access gravitational level spacing by scanning the applied field frequency , as will be explained in the following . in this formalism
the role of the coupling potential @xmath51 is played by the energy of an atom in a fixed hyperfine state thought of as a function of ( slowly varying ) distance @xmath15 and time @xmath21 . @xmath52
the gravitational state wave - function is given by the airy function @xcite .
energy @xmath53 is the eigenvalue of internal and magnetic interactions @xmath54 , where the c.m .
coordinate @xmath18 and time @xmath21 are treated as slow - changing parameters .
corresponding expressions for eigen - energies of a @xmath33 manifold are : @xmath55 subscripts @xmath56 are standard notations for hyperfine states of a @xmath33 manifold in the magnetic field .
the presence of constant field @xmath13 produces the zeeman splitting between states @xmath57 and @xmath58 .
as far as energy of states @xmath59 depend on magnetic field linearly , while for states @xmath60 it depends quadratically , only transition between @xmath59 states take place in case of a week field . in the following
we will consider only transitions between gravitational states in a @xmath61 manifold .
qualitative behavior of the transition probability is given by the rabbi formula , which can be deduced by neglecting high frequency terms compared to resonance couplings of only two states , initial @xmath62 and final @xmath63 , in case the field frequency @xmath64 is close to the transition frequency @xmath65 : @xmath66 the factor @xmath67 appears in front of the right - hand side of the above expression due to the fact that only two @xmath68 of four hyperfine states participate in magnetically induced transitions .
it is important that the transition frequencies @xmath69 do not depend on the antiatom - surface interaction up to the second order in @xmath70 .
this is a consequence of the already mentioned fact that all energies of gravitational states acquire equal ( complex ) shift due to the interaction with a material surface . a resonant spectroscopy of @xmath0 gravitational states could consist in observing @xmath0 atoms localized in the gravitational field above a material surface at a certain height as a function of the applied magnetic field frequency .
a `` flow - throw type '' experiment , analogous to the one , discussed for spectroscopy of neutron gravitational states @xcite , includes three main steps .
a sketch of a principal scheme of an experiment , proposed in @xcite , is shown in fig.[figsketch ] . gravitational states . 1 - a source of ultracold antihydrogen , 2 -a mirror , 3 -an absorber , 4 -a magnetic field , 5 -a detector.,width=377 ] first
, @xmath0 is prepared in a ground gravitational state .
this is achieved by passing @xmath0 through a slit , formed by a mirror and an absorber placed above it at a given height @xmath71 . a mirror and an absorber form a waveguide with a state dependent transmission @xcite .
the choice of @xmath72 @xmath73 provides that only @xmath0 atoms in a ground gravitational state pass through the slit .
second , @xmath0 atoms are affected by alternating magnetic field ( [ magn1 ] ) while moving parallel to the mirror . an excited gravitational state is resonantly populated .
third , the number of @xmath0 atoms in an excited state is measured by means of counting annihilation events in a detector placed at a height @xmath74 above the mirror .
the value of @xmath74 is chosen to be larger , than the spatial size of the gravitational ground state and smaller than the spatial size of the final state ( [ hn ] ) , @xmath75 , so that the ground state atoms pass through , while atoms in the excited state are detected .
we present a simulation of the number of detected annihilation events as a function of the field frequency in fig.[figtrans ] for the transition from the ground to the @xmath76-th excited state , based on numerical solution of equation system eq.([adiab ] ) .
corresponding resonance transition frequency is @xmath77 hz . the value of the field gradient , optimized to obtain the maximum probability of @xmath78 transition during the time of flight @xmath79 s , turned to be equal @xmath80 gs / m , the corresponding guiding field value , which guarantees adiabaticity of the magnetic moment motion , is @xmath81 gs . -th gravitational state.,width=377 ] it follows from ( [ egrav ] ) that the @xmath0 gravitational mass could be deduced from the measured transition frequency @xmath82 as follows : @xmath83 let us mention that in the above formula @xmath7 means the _ gravitational field intensity _ near the earth surface , a value which characterizes properties of field and is assumed to be known with a high precision . at the same time all the information about gravitational properties of @xmath0 is included in the gravitational mass @xmath5 .
equality of the gravitational mass @xmath5 and the inertial mass @xmath6 , imposed by the equivalence principle , results in the following expression : @xmath84 assuming that the spectral line width is determined by the lifetime @xmath85 s of gravitational states , we estimate that the gravitational mass @xmath5 can be deduced with the relative accuracy @xmath86 for @xmath87 annihilation events for the transition to the @xmath76-th state .
we proposed a novel approach to study gravitational properties of antiatoms based on spectroscopy of quantum states of @xmath0 in the gravitational field of the earth near a material surface .
we showed that the gravitational mass @xmath5 of the @xmath0 atom can be deduced from measuring the level spacing between gravitational states by means of resonance transitions , induced by alternating gradient magnetic field .
we study main properties of the interaction of gravitationally bound @xmath0 with magnetic field and find the probabilities of resonance transitions from ground to excited states . in the proposed approach
the number of annihilation events is countered as a function of the applied field frequency .
the spatial positioning of the annihilation detector at a height , corresponding to the classical turning point of the second gravitational state ensures that only @xmath0 atoms in the final state are detected .
the field gradient required for such a transition , @xmath80 gs / m , as well as the guiding field @xmath81 gs are easily accessible in experiments .
the width of the spectral line is determined by the time of life of gravitational states @xmath88 s. using transition from the ground to fifth excited state and assuming that the number of detected annihilation events is @xmath87 , the gravitational mass of @xmath0 can be deduced with the precision @xmath89 . though systematic effects are assumed to be small , their accurate estimation is underway .
|
we study a method to induce resonant transitions between antihydrogen ( @xmath0 ) quantum states above a material surface in the gravitational field of the earth .
the method consists in applying a gradient of magnetic field which is temporally oscillating with the frequency equal to a frequency of a transition between gravitational states of antihydrogen .
corresponding resonant change in a spatial density of antihydrogen atoms can be measured as a function of the frequency of applied field .
we estimate an accuracy of measuring antihydrogen gravitational states spacing and show how a value of the gravitational mass of the @xmath0 atom can be deduced from such a measurement .
|
[ fig : selfx ] shows the current sample of 1300 pulsars in the atnf on - line catalogue ( hobbs et al .
these proceedings ) projected onto the galactic plane . rather than being distributed about the galactic centre , the majority of pulsars are local objects . far from being representative of the true population , this sample is heavily biased by a number of selection effects which we now outline below . * the inverse square law . *
like all astronomical sources , observed pulsars of a given luminosity @xmath0 are strongly selected by their apparent flux density , @xmath1 . for pulsars , which beam to a certain fraction @xmath2 of @xmath3 sr .
] , @xmath4 , where @xmath5 is the distance to the pulsar . since all pulsar surveys have some limiting flux density only those objects bright or close enough will be detectable . * the radio sky background .
* one limit to pulsar search sensitivities is the thermal noise in the receiver , i.e. the `` system temperature '' , @xmath6 . while every effort is made to minimize @xmath6 at the telescope , synchroton radiating electrons in the galactic magnetic field contribute significantly with a `` sky background '' component , @xmath7 . at observing frequencies @xmath8 ghz
, @xmath7 dominates @xmath6 along the galactic plane .
fortunately , @xmath9 so this is effect is significantly reduced at higher frequencies ; e.g. for the pm system , @xmath10 ghz . * propogation effects in the intersellar medium ( ism ) . *
the ism is a mixed blessing for pulsar astronomers . on one hand ,
the dispersion of pulses caused by the ionized component of the ism permits an estimate of @xmath5 through the dispersion measure .
conversely , dispersion and scatter - broadening of the pulses conspire against detection of short period and/or distant objects .
the effects of scattering are shown in fig .
[ fig : scatt ] .
fortunately , like @xmath7 , the scatter - broadening time @xmath11 has a strong frequency dependence , scaling roughly as @xmath12 . as shown in fig .
[ fig : scatt ] , there is a transition frequency of about 1 ghz , below which scattering can hide a large fraction of the population .
another factor is scintillation , the modulation of apparent flux densities by refractive or diffractive `` screens '' of material along the line of sight ( rickett 1970 ) .
this is particularly important for nearby pulsars , where the apparent flux densities can vary significantly .
for example , two northern sky surveys carried out 20 years apart with comparable sensitivity ( damashek , taylor & hulse 1978 ; sayer , nice & taylor 1997 ) detected a number of pulsars above and below the nominal search thresholds of one experiment but not the other .
ideally , surveying the sky multiple times minimizes the effects of scintillation against pulsars nominally above the threshold , and maximizes the detections of faint pulsars through favourable scintillation . * finite size of the emission beam . * as mentioned above , the fact that pulsars do not beam to @xmath3 sr means that we see only a fraction @xmath2 of the total active population . for a circular pencil beam ,
gunn & ostriker ( 1970 ) estimated @xmath13 .
a consensus on the precise shape and evolution of the emission beam , however , has yet to be reached .
narayan & vivekanand ( 1983 ) argued that the beam shape is elongated in the meridonial direction .
lyne & manchester ( 1988 ) , on the other hand , favour a circular beam . using the same database , biggs ( 1990 ) presented evidence in favour of meridonal compression !
all of these studies do agree that the beam size is period dependent , with shorter period pulsars having larger beaming fractions .
for example , tauris & manchester ( 1998 ) found that @xmath14 ^ 2 + 0.03 $ ] , where @xmath15 is the period .
undoubtedly , a complete model for @xmath2 needs to account for other factors , such as evolution of the inclination angle between the spin and magnetic axes .
given the uncertainties , most authors quote results with and without a beaming correction . * pulse nulling . *
the abrupt cessation of the pulsed emission for many pulse periods , was first identified by backer ( 1970 ) .
ritchings ( 1976 ) presented evidence that the incidence of nulling became more frequent in older long - period pulsars , suggesting that it signified the onset of the final stages of the neutron star s life as an active radio pulsar .
since most pulsar surveys have short ( @xmath16 few min ) integration times , there is an obvious selection effect against nulling objects .
means of combatting this effect are to look for individual pulses in search data ( see e.g. nice 1999 ) , survey the sky many times , or use longer integrations .
the pm survey , which employs 35-min pointings , is proving particularly effective at detecting nulling pulsars and should soon be able to better quantify this population and provide a more satisfactory understanding of nulling pulsars . *
orbital motion . * in standard pulsar searches , where time series are fourier transformed ,
the signal from a binary pulsar can be doppler shifted over several bins in the fourier domain .
in extreme cases , where the survey integration time is a significant fraction of the orbital period , this results in a loss of sensitivity .
for example the loss of signal - to - noise of the original binary pulsar b1913 + 16 during a 35-min observation of the pm survey can be as much as 90% . correcting for this effect using `` acceleration searches '' , is now becoming more routine ,
thanks to the ever - increasing availability of high - speed computer resources . since it is quite possible that young pulsars in tight binary orbits exist , the very deep ( @xmath17 210 hr ) searches for young pulsars described by camilo in this volume are now being re - analysed with full acceleration searches .
to get an idea of how biased the sample is due to the above effects , fig .
[ fig : selfx ] shows the cummulative distribution of pulsars as a function of distance from the sun projected onto the galactic plane .
also shown is the expected distribution for a simulated population in which there are no selection effects .
as can be seen , the two samples are closely matched only out to a kpc or so before the selection effects become significant . from these curves
, we deduce that _ less than 10% of the potentially observable population in the galaxy are currently detectable .
_ rigorous conclusions about the true pulsar population can only be made after properly accounting for these selection effects .
shortly after the discovery of pulsars , a number of authors began to consider their implications for the galactic population of neutron stars .
a widely cited paper from that era is the work of gunn & ostriker ( 1970 ; hereafter go ) . to tackle the problem analytically ,
go made two simplifying assumptions based on the sample of 52 pulsars known at that time : ( a ) the relationship @xmath18 , where @xmath0 is the radio luminosity , and @xmath19 is the dipole magnetic field strength ; ( b ) the evolution of @xmath19 being an exponential decay with a time constant @xmath20 . with these in hand , and using the dipolar spin - down expression @xmath21 , go derived expected distributions for the observed population . in particular , they showed that the observed number of pulsars @xmath22 , where @xmath23 is the mean distance of the observed sample , @xmath24 is a parameter relating to the completeness of the surveys , @xmath25 is the local birth rate of potentially observable pulsars and @xmath26 is derived from their model fit to the observed period distribution . assuming @xmath27 , and extrapolating the local birth rate over the galaxy
, they arrived at a galactic birth rate @xmath28 , in good agreement with the best estimates of the rate of type - ii supernovae at that time ( e.g. blaauw 1961 ) .
although mathematically appealing , go s analytical approach required a number of simplifying assumptions about the pulsar population and the galaxy itself .
one of these assumptions , the spontaneous decay of the magnetic field , continues to be extremely controversial . since go s original study
, many papers have been written presenting arguments for and against field decay ( see e.g. lyne , manchester & taylor 1985 ; bailes 1989 ; narayan & ostriker 1990 ; bhattacharya et al .
1992 ; see also the contribution by van leeuwen et al . in this volume ) .
a less model - dependent approach to the problem , first developed by large ( 1971 ) , can be summarized by the following expression : @xmath29 here @xmath30 is the observed population of pulsars as a function of period , @xmath15 , distance from the galactic plane , @xmath31 , galactocentric radius , @xmath32 and luminosity , @xmath0 .
the quantity @xmath33 represents the volume of the galaxy effectively searched and @xmath34 is the underlying ( true ) distribution of the population . since we know @xmath35 and can estimate @xmath33 on the basis of ( hopefully ) well - understood survey sensitivities , we can invert the above expression to solve for @xmath34 . the only simplification required to do this is to assume and @xmath31 ,
there are no significant relationships between any of these parameters .
the assumption of independent distributions seems to be well founded . ]
that @xmath15 , @xmath31 , @xmath32 and @xmath0 are independent quantities .
the problem then reduces to four equations which can be solved for the underlying distributions of interest : @xmath36 , @xmath37 , @xmath38 and @xmath39 .
large s method was somewhat ahead of its time : back in 1971 , pulsar surveys of the galaxy were still in their early stages so that @xmath33 was not well determined . by the late 1970s , however , a number of large - scale searches had been carried out and taylor & manchester ( 1977 ; hereafter tm ) applied the above technique to the sample of @xmath40 pulsars then known ( see also davies , lyne & seiradakis 1977 ) . integrating the derived distribution functions over the galaxy , and assuming @xmath41 , tm estimated the total number of active pulsars in the galaxy to range between 60,000 and 850,000 . here
the principle uncertainty is the assumed distance model .
tm considered a uniform electron content with a mean electron density in the range @xmath42 @xmath43 . to calculate the galactic birth rate , @xmath44 ,
tm required an estimate of the mean pulsar lifetime , @xmath45 , which they obtained from an analysis of the @xmath31 distribution as a function of characteristic age @xmath46 . in the @xmath47 diagram
( see their fig .
7 ) , tm argued that the discrepancy between the expected and observed @xmath31 for characteristic ages larger than a few myr set a limit to @xmath48 myr .
this leads to an implied birthrate @xmath49 !
davies et al .
( 1977 ) reached similar conclusions .
the rather high birth rates from these and other analyses prompted phinney & blandford ( 1981 ) and in particular vivekanand & narayan ( 1981 ; hereafter vn ) to consider a less model - dependent approach to estimating @xmath44 .
the method involves binning the observed sample as a function of spin period , @xmath15 , and in each bin computing the flow or `` current '' of pulsars , @xmath50 , through the bin : @xmath51 here , @xmath52 is the number of pulsars in a period bin of width @xmath53 , @xmath54 and @xmath55 are the `` scale factor '' and beaming fraction of the @xmath56 pulsar respectively .
as discussed earlier , @xmath55 is based on some beaming model . for a given pulsar ,
its scale factor @xmath54 represents the number of pulsars with similar parameters in the galaxy . in principle , this is very similar to the @xmath57 first used to correct observationally biased samples of quasars ( schmidt 1968 ) . in practice @xmath58
is computed using a monte carlo simulation of @xmath30 pulsars with identical periods and luminosities .
using accurate models for the various pulsar surveys , it is relatively straightforward to calculate the number of pulsars @xmath35 that are detectable from that population . as a result , @xmath59 .
detailed simulations to test this approach show that the scale factors give reliable results about the population of pulsars with luminosities above @xmath60 , the minimum luminosity in the sample ( see lorimer et al .
we note in passing that it is possible to form the luminosity function for @xmath61 by weighting the observed luminosities by the appropriate scale factors . integrating this distribution
then yields @xmath62 .
the beauty of the pulsar - current analysis is that it makes just two fundamental assumptions about pulsars : ( i ) they are a steady - state population ; ( ii ) they are spinning down steadily from short to long periods .
the first of these is justified since the ages of pulsars , while not well known ( @xmath63 yr ) , are certainly less than the age of the galaxy , @xmath64 yr .
the second is , of course , well in accord with timing observations .
the birth - rate can be computed from this analysis by simply plotting @xmath50 as a function of @xmath15 . in the standard model where pulsars are born spinning rapidly
, there should be a peak in the current at short periods followed by a decline in the current as pulsars end their life with longer periods .
the birth rate is then just the height of this peak .
vn derived @xmath65 , in much better agreement with the supernova rate .
a by - product of vn s analysis was their conclusion that a significant fraction of pulsars are `` injected '' into the population with relatively long initial spin periods ( @xmath66 s ) , rather than the standard picture of birth with @xmath67 ms .
rather like field decay , arguments for and against injection have been presented ever since .
following criticism by lyne , manchester & taylor ( 1985 ; hereafter lmt ) that vn s analysis had not properly taken account of selection effects such as scattering , narayan ( 1987 ) performed a more detailed analysis which provided further support for injection .
lorimer et al .
( 1993 ) used monte carlo simulations to investigate the validity of the pulsar current approach and found that it was prone to systematic errors induced by faint , nearby pulsars . excluding these objects from the analysis
significantly reduces the case for injection .
perhaps the final word on this issue will be a detailed pulsar current analysis of the pm survey sample ( vranesevic et al .
this volume ) .
a major effort to quantify the galactic distribution was the analysis of lmt who combined the theoretical framework of go with the numerical approach of large to the sample of 316 pulsars detected in four major 400-mhz surveys .
two key results from their paper are shown in fig .
[ fig : lmt ] . as a result of their numerical analysis ,
lmt were able to place better constraints on the radial distribution than tm .
however , due to the extreme selection effects on 400-mhz surveys towards the inner regions of the galaxy , the corrected radial distribution becomes very uncertain below 4 kpc ( see fig . [
fig : lmt ] ) . to derive the birth rate , @xmath44 , rather than applying a pulsar current anlalysis ,
lmt followed go s approach to model the observed distributions and concluded that the best - fit timescale for exponential field decay to be 9.1 myr .
since their analysis assumed go s @xmath18 luminosity law , magnetic field decay provided a mechanism for luminosity decay . by considering their corrected luminosity function as a steady - state population where pulsars flowed into successively fainter luminosity bins , lmt were able to derive @xmath44 as a function of @xmath0 .
the result of this analysis is shown in fig . [
fig : lmt ] , from which lmt concluded @xmath68 . while an elegant approach to the problem , lmts method relies on the field decay hypothesis being correct .
as mentioned previously , the issue of magnetic field decay remains unresolved .
since magnetic field evolution underpins much of pulsar statistics , perhaps the single most important breakthrough in this area would be a comprehensive reassesment of this issue .
although much effort has gone into improving the monte carlo simulations of pulsar population modeling since lmt , relatively little progress has been made in improving our knowledge of the galactic distribution .
of particular interest is the underlying density @xmath69 of pulsars as a function of galactocentric radius , @xmath32 .
lmt and others were only able to place poor constraints on this function ( fig .
[ fig : lmt ] ) and most subsequent work has assumed a gaussian distribution for @xmath69 ( e.g. narayan 1987 ) .
in fact , as pointed out by bailes & kniffen ( 1991 ) , there is no reason to prefer this function over one where @xmath70 as @xmath71 .
improving our understanding of @xmath69 at small @xmath32 requires better statistics of the inner - galaxy pulsars . due to the propogation and sky - background selection effects mentioned earlier ,
low - frequency ( 0.4 ghz ) surveys are very poor probes of this population .
one of the main motivations for high - frequency ( 1.4 ghz ) surveys of the galactic plane is that they are less prone to these effects .
johnston ( 1994 ) analysed two such surveys carried out in the late 1980s ( clifton et al .
1992 ; johnston et al .
1992 ) and found that models with a deficit of pulsars in the inner galaxy were strongly preferred over a simple gaussian profile for @xmath69 .
the sample of @xmath40 pulsars detected by these two early surveys has now been completely surpassed by the phenomenal success of the pm survey : over 700 new pulsars have been found in the galactic plane search so far .
together with re - detections of known pulsars , the pm sample amounts to 914 pulsars .
i have applied large s numerical approach to this new sample to derive new distribution functions in @xmath32 , @xmath0 , @xmath31 and @xmath15 .
preliminary results are shown in fig . [
fig : new ] .
the corrected radial density function clearly supports johnston s conclusion for a deficit of pulsars in the inner galaxy .
more work is required in quantifying the significance of this result , particularly in the @xmath72 kpc bin where only one pulsar is presently known !
while a population of inner - galaxy pulsars could be masked from the pm survey by severe scattering not currently taken into account , it is interesting that the now completed effelsberg 5-ghz galactic centre survey ( klein et al .
these proceedings ) has not found a single pulsar .
although the pm survey has been the most prolific probe of the galactic population to date , we are still only scratching the surface .
one of many uncertain areas about the pulsar population is the shape of the luminosity function . from fig .
[ fig : new ] we see a clear departure from the @xmath73 relationship at low @xmath0 . integrating fig .
[ fig : new ] results in a galactic population of @xmath74 potentially observable pulsars with @xmath75 mjy kpc@xmath76 . below this limit , the population is essentially unknown .
future surveys with the ska , which should easily be able to detect @xmath77 pulsars with @xmath75 mjy kpc@xmath76 should also detect many fainter objects and truly constrain the pulsar luminosity function .
backer d. c. , 1970 , nature , 228 , 42 bailes m. & kniffen d. a. , 1992 , , 391 , 659 bailes m. , 1989 , , 342 , 917 bhattacharya d. , et al .
1992 , , 254 , 198 biggs j. d. , 1990 , , 245 , 514 blaauw a. , 1961 , bull.astr.inst.netherlands , 15 , 265 clifton t. r. , et al . 1992 , , 254 , 177 damashek m. , taylor j. h. , & hulse r. a. , 1978 , , 225 , l31 davies j. g. , lyne a. g. , & seiradakis j. h. , 1977 , , 179 , 635 gunn j. e. & ostriker j. p. , 1970
, , 160 , 979 ( go ) johnston s. , 1994 , , 268 , 595 johnston s. , et al .
1992 , , 255 , 401 large m. i. , 1971 .
iau symposium no .
46 , dordrecht , reidel , p. 165
lorimer d. r. , et al .
1993 , , 263 , 403 lyne a. g. & manchester r. n. , 1988 , , 234 , 477 lyne a. g. , manchester r. n. , & taylor j. h. , 1985 , , 213 , 613 ( lmt ) narayan r. & ostriker j. p. , 1990
, , 352 , 222 narayan r. & vivekanand m. , 1983 , , 122 , 45 narayan r. , 1987 , , 319 , 162 nice d. j. , 1999 , , 513 , 927 phinney e. s. & blandford r. d. , 1981 , , 194 , 137 rickett b. j. , 1970 , , 150 , 67 ritchings r. t. , 1976 , , 176 , 249 sayer r. w. , nice d. j. , & taylor j. h. , 1997 , , 474 , 426 schmidt m. , 1968 , , 151 , 393 tauris t. m. & manchester r. n. , 1998 , , 298 , 625 taylor j. h. & manchester r. n. , 1977 , , 215 , 885 ( tm ) vivekanand m. & narayan r. , 1981 , ja&a , 2 , 315 ( vn )
|
we review current understanding of the underlying , as opposed to the observed , pulsar population . the observed sample is heavily biased by selection effects , so that surveys see less than 10% of all potentially observable pulsars .
we compare various techniques used to correct the sample for these biases . by far
the most significant recent development has been the discovery of over 700 pulsars in the parkes multibeam ( pm ) survey .
this new sample is far less affected by selection effects and we use it to make a preliminary analysis of the galactic pulsar distribution , finding further evidence for a deficit of pulsars in the inner galaxy .
# 1_#1 _ # 1_#1 _ = # 1 1.25 in .125 in .25 in
|
the motivations for studying jet observables instead of single - inclusive observables are manifold and shall not be discussed extensively here . instead , only two important points will be mentioned .
firstly , it has turned out that single - inclusive observables do not fully constrain the analytical models for partonic energy loss .
the nuclear modification factor for instance is described by all models equally well albeit with very different transport coefficients@xcite .
the benefit of studying sub - leading fragments and jet observables is that these are much more discriminating .
the downside is that they are not well modelled by the existing analytical calculations and new theoretical tools are needed .
secondly , on the experimental side there is need for reliable tools that allow to disentangle jets from the background .
this requires a quantitative understanding of jet areas , background and background fluctuations@xcite and the response of jet finding algorithms to quenched jets@xcite .
further complications arise from the fact that the implicit assumption that jets and background are uncorrelated is strictly speaking not justified .
it is to be expected on general grounds that there is a backreaction of the jets to the medium , but this is very difficult to quantify with the currently available tools .
it is , however , clear that a quantitative understanding of jets , background and their correlations requires running jet finders on the theory predictions . in particle physics monte carlo ( mc )
event generators have emerged as extremely powerful and versatile tools linking theory and experiment .
it is the aim of this talk to make an assessment of the status of mc models for jet quenching by highlighting conceptual issues and to what extent they are solved or solvable using mc tools .
it focusses on dedicated jet quenching mc models and does not discuss parton cascade codes .
the latter should in principle also be capable of describing jet quenching , but as the approach and focus are quite different from specialised jet quenching mc codes they will not be dicussed here .
jet evolution in vacuum arises due to collinear divergences in real emission matrix elements .
the singularity structure is universal , leading to a factorisation in the collinear region .
the differential cross section for a process with @xmath0 partons can thus be written as the product of the differential cross section for the process with @xmath1 partons , the single particle phase space and the differential radiation probability characterised by the altarelli - parisi splitting function : @xmath2 the variable @xmath3 quantifies the hardness of the splitting and can be the transverse momentum squared @xmath4 , the virtuality @xmath5 or the radiation angle squared @xmath6 all these choices are to leading logarithmic accuracy equivalent .
the regularisation of the @xmath7-singularity gives rise to large logarithms that can compensate the smallness of the strong coupling @xmath8 and thus have to be resummed to all orders .
this proceeds via a dglap evolution equation , which for only one parton species can be written as @xmath9 \,.\ ] ] here , @xmath10 is the parton density at energy fraction @xmath11 and scale @xmath3 .
the change in @xmath10 when increasing the scale is given by an increase due to splittings of partons at higher energy fraction @xmath11 and a decrease due to partons at @xmath11 that split to populate smaller energies .
this equation can be integrated with the help of the sudakov form factor defined as @xmath12 the integrated evolution equation then becomes @xmath13 indicating that @xmath14 can be interpreted as the probability for having no emission between @xmath3 and @xmath15 . here , @xmath15 is the infra - red cut - off scale , at which the dynamics becomes dominated by non - perturbative physics . as it regularises a soft and collinear divergence it is important that the observables that are to be calculated are insensitive to the exact value of the cut - off , i.e. are infra - red and collinear safe .
the sudakov form factor is the basis for an iterative monte carlo ( mc ) algorithm that generates the radiative corrections to any given hard process and is called the parton shower .
it resums real emissions to all orders to leading logrithmic accuracy and via unitarity also includes the corresponding virtual corrections .
the elements of monte carlo event generators that are relevant for jet production and evolution are matrix elements , parton showers and the hadronisation .
the matrix elements describe hard scattering processes and are calculated at fixed order in perturbation theory .
considerable effort has been made in the last years to provide the matrix elements for all phenomenologically relevant processes at next - to - leading order in the mc generators and to interface them with the parton showers .
the initial and final state parton showers generate radiative corrections to the matrix elements by resumming collinear or soft and collinear logarithms to leading logarthmic approximation .
these parts of the event generators ( i.e. matrix elements and parton showers ) are nothing but a faithful representation of perturbation theory with controlled systematic accuracy .
in contrast to this , the hadronisation is governed by non - perturbative physics and is not accessible with perturbative methods .
it is instead simulated with the help of phenomenological models , the most commonly used and best tested models are the lund string model and the cluster hadronisation .
the model parameters are tuned to data ( mainly from lep ) and assumed to be universal .
in heavy ion collisions the hard matrix elements remain unchanged , due to the high scale they involve .
they thus happen on very short time and length scales early in the reaction and are not affected by the formation of a system of considerable density .
the final state parton shower , on the other hand , evolves over a considerably longer time scale and is most likely modified by the formation of a dense and hot medium .
there is , however , no general theory for jet evolution in a dense medium , but only results for special cases ( for instance the single gluon radiation spectrum in the eikonal limit ) .
the initial state parton shower was found to be unmodified by the rhic experiments .
the hadronisation , finally , is likely to experience modifications , but due to its non - perturbative nature there is only very little theoretical guidance .
the parton shower as introduced in section [ sec_vacjets ] is formulated in momentum space .
the theoretical description of jet evolution in a medium requires a simultaneous formulation in momentum and configuration space .
this can be achieved by estimating the time scale for gluon radiation from the uncertainty principle : @xmath16 the first expression is simply the inverse transverse energy of the radiated gluon , while the second is the lifetime @xmath17 of an unstable state decaying into two partons characterised by a scale @xmath3 boosted to the laboratory frame .
the two arguments are parametrically equivalent . apparently the formation or decoherence of energetic fragments in the vacuum parton shower is delayed by time dilation to times typically larger than the medium length . in the case of medium induced radiation
the transverse momentum comes predominantly from scattering in the medium : @xmath18 .
one thus finds for the lifetime of medium induced radiation @xmath19 this means that soft emissions decohere first and at large angles . at face value
these qualitative estimates indicate that jets emerging from heavy ion collisions should have a hard core that fragments in vacuum accompanied by soft medium induced radiation at large angles@xcite .
this is in qualitative agreement with early jet measurements by atlas@xcite and cms@xcite and can be further clarified and quantified by measurements of intra - jet distributions ( for instance fragmentation functions ) .
the currently available mc models for jet quenching follow rather different approaches .
they shall only be briefly mentioned here , for details the reader is referred to the original publications .
hijing : : introduces a medium induced parton splitting process , collisional energy loss is neglected@xcite hydjet++/pyquen : : simulates radiative energy loss by sampling a bdmps gluon spectrum and includes perturbative elastic scattering@xcite jewel : : aims for a unified description of all scattering processes in terms of matrix elements and parton showers ( work in progress)@xcite q - pythia / q - herwig : : adds a term derived from bdmps to the splitting function and thus contains no elastic scattering@xcite yajem : : assumes that medium interations increase the virtuality of partons in the parton shower ( leading to enhanced radiation ) and can subtract energy and momentum to simulated collisional energy loss@xcite martini : : is based on the amy transition rates and also contains a transition rate for elastic scattering@xcite going from single - inclusive to jet observables raises conceptual issues , that will be discussed in the following .
the analytical calculations of radiative energy loss operate in the eikonal approximation where @xmath20 consequently , energy and momentum are not conserved and the scattering centres are recoilless , i.e. there is no collisional energy loss .
in contrast to this , phenomenology at rhic and lhc requires that all quantities in the above equation can be of the same order , but are subject to energy - momentum conservation .
there are thus large uncertainties in the predictions of analytical calculations due to kinematical ambiguities .
in general , i.e. non - eikonal , kinematics phase space restrictions due to energy - momentum conservation have to be taken into account .
this is not necessarily a small effect , but can have rather dramatic effects on the gluon spectrum .
furthermore , the scattering centres become dynamical and recoil against the jet giving rise to collisional energy loss ( also in inelastic interactions ) and induced radiation off the scattering centre .
this already hints at an additional complication , namely that elastic and inelastic scatterings can not be unambiguously separated from each other .
they are two possible outcomes ( depending on a definition ) of the same process rather than different processes .
there is also no clear separation between vacuum and medium induced radiation any more .
incorporating exact energy - momentum conservation is for mc models typically straight forward .
the ambiguity between elastic and inelastic scattering , on the other hand , requires a unified description of both processes , which is currently under development .
dynamical scattering centres are problematic for models based on effective descriptions for the medium interactions but are less challenging for models incorporating microscopical interaction models .
therefore , collisional energy loss is typically either neglected or added as a separate process .
radiation off the scattering centres requires model dependent assumptions and has not been included yet , but first steps are being taken .
concerning the ambiguity between vacuum and medium induced radiation the models either build on an unified description and do nt distinguish at all between the two , or they assume a complete factorisation where all vacuum radiation happens first and the medium induced emissions follow afterwards .
the analytical models only compute single gluon radiation , which is then iterated probabilistically .
this leads to complications due to the non - conservation of energy and momentum in the eikonal limit . concerning the interplay between multiple gluon emissions
first theoretical progress was made recently through the calculation of gluon radiation off a colour dipole@xcite .
the results indicate , that interference between emissions off the two legs occurs only in restricted regions of phase space so that the dominant process is independent radiation off both legs of the dipole@xcite .
another important effect is that radiated gluons themselves radiate requiring a democratic treatment of all partons .
this can have significant effects in particular on @xmath21- and angular distributions .
this also means that the energy loss is not a meaningful quantity any more and that rather the entire fragmentation pattern has to be considered .
energy - momentum conservation is already important for single gluon radiation , for the description of multi - parton final states it is crucial .
these considerations allow for the conclusion , that theories without democratic radiation are not suitable for jet phenomenology . treating the non - abelian landau - pomerantchuk - migdal - effect , which is a quantum - mechanical interference , is notoriously difficult for mc models .
therefore , most of them use effective descriptions of the single gluon radiation process .
however , in@xcite a local and probabilistic formulation of the lpm - effect was derived , that will also be included in future mc models .
the mc models also iterate single gluon radiation , which always involves model dependent assumptions .
the common assumption that all partons radiate independently has received support from the recent results on gluon radiation off colour dipoles .
democratic treatment is by construction easily achieved ( except for the scattering centre ) .
understanding @xmath21-broadening is important as it affects the response of jet finders to quenched jets . in the analytical models it is governed by brownian motion in the transverse space .
it is , however , sensitive to energy - momentum conservation , democratic multiple gluon radiation and contamination by energetic recoils .
the assumptions in the mc models vary from collinear gluon emission to parton shower kinematics and generally leave room for improvements of the microscopic dynamics .
the propagation of highly energetic jets leads to modifications of the medium in the vicinity of the jets , which is likely to survive into the hadronic stage . a quantitative understanding of jet - induced modifications of the medium is thus important for the experimental background subtraction .
but it is also interesting in its own right , as it for instance gives access to the interplay between weakly and strongly coupled regimes .
a satisfactory level of understanding of the jet backreaction probably requires a unified description of jet and medium evolution . while this is still some way to go first attempts to characterise the reaction of the medium to jets
have been made from the jet quenching side by tracking the recoiling scattering centres@xcite and from the medium modelling side by solving hydrodynamics with source terms that describe the energy and momentum deposition of a jet@xcite .
the fact that most mc models use hydrodynamic calculations as model for the medium makes it difficult for them to quantify the backreaction .
a conversion of hydrodynamical results into a population of scattering centres is possible , but involves model dependent assumptions .
even then , the information about the distortion due to the passage of a jet can not be propagated back to the hydrodynamical calculation .
event - by - event hydrodynamics is a development that may allow for a simultaneous solution in the future , but the problem of having two vastly different approaches for the jet and the medium evolution remains .
this is partly a problem of very different scales , as the jets are described in perturbation theory requiring a high scale , whereas the bulk medium evolution is governed by non - perturbative processes at much lower scales .
consequently , also the language is vastly different with the jets being formulated in terms of individual partons while the medium is usually treated using continuum dynamics .
bringing both regimes together is thus far from trivial .
parton cascades that are built on a partonic language also for the medium have so far not been able to consistently include the parton shower .
medium modifications to the hadronisation phase are likely and raise conceptually both interesting and difficult questions . due to the non - perturbative nature of the problem
there is at best very little theoretical guidance .
therefore , it is commonly assumed that hadronisation happens at late times and therefore in vacuum .
this is a reasonable assumption for energetic fragments due to the large boost factors , but not necessarily for soft and semi - hard fargments .
even if one assumes hadronisation in vacuum there are complications due to the medium : the hadronisation is sensitive to the colour topology of the event , but strong interactions in a medium inevitably alter the colour configuration .
furthermore , it is unclear how jet and medium hadronisation interplay , in particular in regions of phase space where soft fragments of the jets overlap with relatively hard fluctuations or recoils from the medium .
this leads to potentially large uncertainties even in a factorised approach .
mc models assume like most other models that hadronisation happens in vacuum .
some allow for modifications of the colour toplogy , but with the exception of they all rely on the lund string model . as it is not clear how to sytematically improve the existing approaches , the currenly most promising strategy is to understand and quantify the systematics and uncertainties of the models , for instance by varying assumptions about the colour topology , use of different hadronisation models etc .
some of the implemented prescriptions have a potentially dangerous shortcoming in that they are infra - red and/or collinear sensitive .
this is a point that could be improved upon .
finally , there are alternative ideas ( like pre - hadron formation ) that should in principle be suitable for a mc implementation . comparing new ideas to the traditional approaches
could perhaps also lead to new insights .
there is strong motivation from both the theoretical and experimental side for studying jets in heavy ion collisions , although this implies a considerable increase of complexity in the theoretical description . among the conceptual issues raised by the transition from single - inclusive quantities to jets non - eikonal kinematics , multi gluon emission , @xmath21-broadening , jet - induced medium modifications and hadronisation are the most important ones .
mc generators are powerful and versatile theory tools that allow to explore all these issues .
they are designed to describe jets on the basis of multi - particle final states and allow to account dynamically for the interactions between the jet and the medium . however , as long as there is no general theory of jet quenching , also mc generators will have to rely on phenomenological models .
it is to be expected that there will be considerable progress in the next years , driven by the jet data from rhic and lhc and fruitful interaction between theorists and experimentalists .
the author would like to thank urs wiedemann for many enlightening discussions .
10 s. a. bass , c. gale , a. majumder , c. nonaka , g. y. qin , t. renk and j. ruppert , phys .
c * 79 * ( 2009 ) 024901 [ arxiv:0808.0908 [ nucl - th ] ] .
m. cacciari , g. p. salam and g. soyez , eur .
j. c * 71 * , 1692 ( 2011 ) [ arxiv:1101.2878 [ hep - ph ] ] .
e. bruna [ star collaboration ] , j. phys .
ser . * 230 * , 012009 ( 2010 ) [ arxiv:1006.0952 [ nucl - ex ] ] .
j. casalderrey - solana , j. g. milhano and u. a. wiedemann , j. phys .
g * 38 * , 035006 ( 2011 ) [ arxiv:1012.0745 [ hep - ph ] ] .
g. aad _ et al . _ [ atlas collaboration ] , phys .
lett . *
105 * , 252303 ( 2010 ) [ arxiv:1011.6182 [ hep - ex ] ] . s. chatrchyan _ et al . _
[ cms collaboration ] , arxiv:1102.1957 [ nucl - ex ] .
x. n. wang and m. gyulassy , phys .
d * 44 * ( 1991 ) 3501 .
w. t. deng , x. n. wang and r. xu , phys .
c * 83 * , 014915 ( 2011 ) [ arxiv:1008.1841 [ hep - ph ] ] .
i. p. lokhtin and a. m. snigirev , eur .
j. c * 45 * ( 2006 ) 211 [ arxiv : hep - ph/0506189 ] .
i. p. lokhtin , l. v. malinina , s. v. petrushanko , a. m. snigirev , i. arsene and k. tywoniuk , comput . phys .
commun . * 180 * , 779 ( 2009 ) [ arxiv:0809.2708 [ hep - ph ] ] .
k. zapp , g. ingelman , j. rathsman , j. stachel and u. a. wiedemann , eur .
j. c * 60 * , 617 ( 2009 ) [ arxiv:0804.3568 [ hep - ph ] ] .
k. zapp , j. stachel and u. a. wiedemann , phys .
lett . * 103 * ( 2009 ) 152302 [ arxiv:0812.3888 [ hep - ph ] ] .
n. armesto , l. cunqueiro and c. a. salgado , eur .
j. c * 63 * , 679 ( 2009 ) [ arxiv:0907.1014 [ hep - ph ] ] .
n. armesto , g. corcella , l. cunqueiro and c. a. salgado , jhep * 0911 * , 122 ( 2009 ) [ arxiv:0909.5118 [ hep - ph ] ] .
t. renk , phys .
c * 78 * , 034908 ( 2008 ) [ arxiv:0806.0305 [ hep - ph ] ] .
t. renk , phys .
c * 79 * , 054906 ( 2009 ) [ arxiv:0901.2818 [ hep - ph ] ] .
b. schenke , c. gale and s. jeon , phys .
c * 80 * , 054913 ( 2009 ) [ arxiv:0909.2037 [ hep - ph ] ] .
b. schenke , s. jeon and c. gale , phys .
c * 83 * , 044907 ( 2011 ) [ arxiv:1101.0425 [ hep - ph ] ] .
y. mehtar - tani and k. tywoniuk , arxiv:1105.1346 [ hep - ph ] .
j. casalderrey - solana and e. iancu , jhep * 1108 * ( 2011 ) 015 [ arxiv:1105.1760 [ hep - ph ] ] .
k. c. zapp , j. stachel and u. a. wiedemann , jhep * 1107 * ( 2011 ) 118 [ arxiv:1103.6252 [ hep - ph ] ] .
r. b. neufeld and t. renk , phys .
c * 82 * , 044903 ( 2010 ) [ arxiv:1001.5068 [ nucl - th ] ] .
|
a thorough understanding of jet quenching on the basis of multi - particle final states and jet observables requires new theoretical tools .
this talk summarises the status and propects of the theoretical description of jet quenching in terms of monte carlo generators .
|
at the time einstein decided to apply his equations to the universe , observational data in cosmology was rather scarce , if not inexistent .
hence , instead of relying on observations to constrain the spacetime metric , he postulated the
cosmological principle , " the isotropy and homogeneity of the universe . similarly , guided by theoretical prejudice
, einstein assumed that the universe was static , which , incidentally , is what forced him to introduce a cosmological constant in his equations .
it turns out that the universe is actually homogeneous and isotropic on large scales , as confirmed , for example , by the homogeneity of the distribution of luminous red galaxies @xcite and the isotropy of the cosmic microwave background @xcite .
however , the universe is not static , and einstein s staticity assumption prevented him from predicting the expansion of the universe later discovered by hubble .
the cosmological principle has a similar , but formally independent counterpart for cosmological perturbations .
it states that perturbations are statistically homogeneous and isotropic , that is , that their correlation functions are invariant under translations and rotations .
these assumptions are so ingrained and integrated into our treatment of cosmological perturbations , that often we are unaware of them . as a consequence ,
the statistical homogeneity and isotropy of cosmological perturbations has received little attention in the literature , and only recently the large scale anomalies observed in the cosmic microwave background radiation ( cmb ) have led to a spurt of interest in statistical anisotropy .
in fact , the data provided by the wmap cosmic microwave background experiment @xcite , shows hints of violations of statistical isotropy .
these hints include the alignment of the quadrupole and octopole @xcite , unlikely correlations between `` multipole vectors '' @xcite , a north - south asymmetry in the cmb sky @xcite and evidence for the existence of a symmetry plane @xcite . on the other hand ,
a measure of statistical anisotropy proposed in @xcite shows no evidence for statistical anisotropy , and it has been also suggested that the previous hints are artifacts of the heavily processed maps used to analyze the data @xcite . in any case , regardless of whether the cmb sky is truly statistically anisotropic or not , the statistical isotropy of cosmological perturbations should be subject to empirical test , rather than taken on faith .
whereas most studies have focused on the statistical isotropy of the cmb sky , in this work we analyze the isotropy of the primordial perturbations themselves . verifying whether cosmological perturbations are statistically anisotropic
is a way to confirm whether our basic assumptions about structure formation , and ultimately our universe , are correct .
it provides a novel way to test our current understanding of the origin of structure , embodied in inflationary models , and it might provide information about the universe at high energies and long wavelengths . in this work
we are not concerned with the mechanism responsible for the generation of statistically anisotropic primordial perturbations , although if it were impossible to seed them our analysis would be perhaps an academic exercise .
the most straightforward mechanism to explain statistical anisotropies involves isotropy violations in the background spacetime itself .
these departures from isotropy can arise from non - trivial spatial topologies @xcite or departures from the _ background _ friedmann - robertson - walker metric @xcite .
however , these explanations require the breaking of isotropy from the onset , and they also clash with our understanding of the origin of structure . alternatively
, statistical anisotropies might arise from coherent magnetic fields in the universe @xcite .
if these magnetic fields can be regarded as perturbations in an isotropic spacetime , their contribution to the metric perturbations is quadratic in the fields , and hence presumably highly suppressed . on the other hand ,
if they significantly contribute to the energy density of the universe , one expects deviations from isotropy at the level of the background spacetime itself , violating again isotropy from the onset . in this paper
we assume that the universe is homogeneous , isotropic , spatially flat and practically infinite . but even homogeneous and isotropic universes can be dominated by sets of non - vanishing non - scalar background fields if the latter appear in special configurations , like for example , vector field triads . in these universes , perturbations are expected to be statistically anisotropic @xcite . in general the existence of these non - vanishing fields
signals the spontaneous breaking of lorentz - invariance @xcite ( rotational invariance to be more precise ) , so the study of statistical anisotropies in the primordial perturbations opens a new window to test the basic symmetries of nature .
the paper is organized as follows . in section [ sec : formalism ] we precisely define what we mean by statistically isotropic or anisotropic primordial perturbations .
it is in this section where we introduce a quantitative characterization of primordial statistical anisotropies . in section
[ sec : temperature ] we connect statistical anisotropies in the primordial perturbations to statistical anisotropies in the cosmic microwave background temperature fluctuations .
we define statistics that capture the amount of statistical anisotropy in the cmb and at the same time mirror the amount of anisotropy in the primordial density ripples , and we also discuss the relation between our statistics and the bipolar spectrum of hajian and souradeep @xcite .
section [ sec : looking ] contains a description of strategies to test and constrain the statistical isotropy of the primordial perturbations .
this task is simplified if there is a cut - off in the anisotropies at a particular arbitrary multipole , which is the case we mainly consider . as an illustration of the formalism ,
we look for the simplest departure from statistical isotropy : a preferred direction in the primordial perturbations .
we apply our statistics to the full sky cmb maps and do not find evidence of statistical anisotropy .
finally , in section [ sec : conclusions ] we draw our conclusions .
we have also included an appendix where we summarize most of the formulae that we need in our derivations .
consider a real gaussian random field in euclidean ( flat ) three - dimensional space , @xmath0 . at this point
it is not important to know what @xmath1 actually is , although the reader can think of @xmath1 as the newtonian potential ( in longitudinal gauge @xcite ) .
the isometries of euclidean space consist of translations , rotations and reflections @xcite .
it is therefore useful to analyze how random fields behave under these transformations .
the gaussian random field @xmath1 is _ statistically homogeneous _ if its momenta are invariant under translations , @xmath2 where @xmath3 denotes ensemble average .
is real by assumption , @xmath4 .
we include the complex conjugation in the two - point function for convenience .
] it is not essential to assume that the field is gaussian .
this assumption just allows us to concentrate on the average @xmath5 and the correlation @xmath6 , which uniquely characterize the statistical properties of the random field in this case .
a non - gaussian field that does not satisfy equations ( [ eq : homogeneous ] ) is not statistically homogeneous either .
the average @xmath5 of a statistically homogeneous random field is a constant in space , and hence is trivially invariant under rotations .
a gaussian random variable @xmath1 is _ statistically isotropic _ ( around @xmath7 ) if its momenta are invariant under rotations ( around @xmath7 ) , @xmath8 if we assume that the field is statistically homogeneous , equations ( [ eq : isotropic ] ) are satisfied for all points @xmath9 if they are satisfied for a single @xmath7 ( say , the origin @xmath10 . )
conversely , one can also show that if a random field @xmath1 fulfills equations ( [ eq : isotropic ] ) for all possible points @xmath7 , then it also fulfills equations ( [ eq : homogeneous ] ) .
in other words , if a random field is statistically isotropic around all points , it is statistically homogeneous . note that it is not necessary to discuss the statistical properties of @xmath1 under inversions
. the inversion of a single point can be accomplished by a translation , and the inversion of two points can be accomplished by a rotation .
often it is convenient to decompose spatial functions in eigenvectors of the translation operator . in flat space
these are the plane waves @xmath11 .
let us hence consider a fourier decomposition of the random field , @xmath12 the components @xmath13 ) are random fields in fourier space . in this space , the requirement of statistical homogeneity ( [ eq : homogeneous ] ) translates into @xmath14 where @xmath15 is the ( constant ) mean of @xmath1 and @xmath16 denotes the power spectrum , which is an arbitrary positive definite function .
] of @xmath17 .
on the other hand , the requirement of statistical isotropy ( [ eq : isotropic ] ) implies @xmath18 that is , that the power spectrum only depends on the magnitude @xmath19 of the wave vector @xmath17 .
because any positive function can be written as a square , we can express the power spectrum as @xmath20 where @xmath21 is a real function by definition .
then , the random variable @xmath1 is statistically isotropic if and only if @xmath21 only depends on @xmath19 . in this paper , we consider statistically anisotropic primordial perturbations , that is , primordial perturbations whose power spectrum does not only depend on the magnitude @xmath19 .
it is hence going to be useful to decompose @xmath22 into components that transform differently under rotations of the wave vector , @xmath23 where @xmath24 and the @xmath25 are spherical harmonics . because @xmath22 is real , the complex functions @xmath26 satisfy @xmath27 .
this decomposition of @xmath21 carries over to the power spectrum itself . inserting equation ( [ eq : phi_expansion ] ) into equation ( [ eq : phi_definition ] ) and using relation ( [ eq : three_spherical ] )
we find that @xmath28 and the real coefficient @xmath29 is given by equation ( [ eq : d ] ) .
if the random field is statistically isotropic , @xmath30 for @xmath31 . on the contrary ,
if @xmath1 is statistically anisotropic , there exists at least one @xmath31 and @xmath32 such that @xmath33 . for simplicity
we shall express our results mostly in terms of @xmath34 , though the reader should be aware that the actual free parameters that characterize the statistical anisotropies are the @xmath35 .
observe that because @xmath36 is real , @xmath37 , which also follows from the relation ( [ eq : p_lm ] ) .
sometimes , one is interested in random fields @xmath38 that are defined not on @xmath39 , but rather on a two - sphere , @xmath40 .
the standard case is the temperature fluctuations on the cosmic microwave background sky .
because the isometries of a sphere consist only of rotations and inversions , for gaussian fields it suffices to consider the properties of the random field under rotations ( the behavior of the cmb under inversions has been studied in @xcite . ) a gaussian random field is _ statistically isotropic _ if the average @xmath41 and the two - point function @xmath42 are invariant under rotations , @xmath43 in the applications we have in mind , random fields on a sphere receive weighted contributions from an infinite number of fourier modes .
for instance , the temperature fluctuations measured by an observer at @xmath44 in the direction @xmath45 can be written as @xcite @xmath46 where @xmath1 is a random field in fourier space and @xmath47 is a real function .
we shall specify what @xmath1 and @xmath47 represent below . at this point , we just want to know what constraints statistically isotropy of @xmath48 imposes on the quantities @xmath47 .
first , note that it follows from equation ( [ eq : delta_expansion ] ) that if @xmath1 is statistically homogeneous , @xmath49 is invariant under rotations if @xmath50 .
furthermore , inserting the expansion ( [ eq : delta_expansion ] ) into the second equation in ( [ eq : sphere_isotropic ] ) we arrive at the condition @xmath51=0.\ ] ] therefore , @xmath48 is statistically isotropic if @xmath52 note that the two - point function @xmath53 does not depend on @xmath44 if the random field @xmath1 is statistically homogeneous .
our main concern here are the contribution of scalar perturbations to the temperature fluctuations in the cosmic microwave background .
let us denote by @xmath54 the temperature fluctuations measured by an observer at @xmath55 looking at direction @xmath45 in the sky . as hinted above , it is more convenient to study the fourier transform of the previous fluctuations , @xmath56 . in linear perturbation theory ,
the evolution of @xmath56 is described by a set of decoupled differential equations for each mode @xmath17 .
the linearity of these equations allows us to separate the initial conditions from the evolution .
in particular , for adiabatic initial conditions one can write @xmath57 where @xmath47 is the temperature fluctuation one obtains by evolving the perturbation equations from initial conditions where @xmath58 , and @xmath59 is the value of the newtonian potential at a sufficiently early time , well into the radiation dominated era . note that this split between evolution and initial conditions is arbitrary to some extent : what we consider as initial conditions here might be regarded for instance as the outcome of an inflationary stage .
recall now our discussion of statistical isotropy of a random field defined on a sphere .
it follows from equation ( [ eq : origins ] ) that violations of statistical isotropy in the cmb sky might originate from two sources : from the initial conditions ( encoded in the power spectrum @xmath36 ) , or from the evolution ( encoded in the transfer function @xmath47 ) . in this work
we assume that the evolution does not generate statistically anisotropic perturbations , that is , we assume that @xmath60 for all rotation matrices @xmath61 in @xmath62 .
this is the case in cosmological models where the universe does not contain any non - scalar background field , like in the standard @xmath63cdm scenario .
@xcite . for example , if there exists a background vector field @xmath64 in the universe , the scalar @xmath65 is not invariant under @xmath66 . ] as a consequence , the differential equations that describe the evolution of the perturbations are invariant under rotations of @xmath17 and @xmath45 , and it follows that @xmath67 ) is a function of the two scalars one can construct out of @xmath17 and @xmath45 , namely @xmath19 and @xmath68 , @xmath69 .
this property then allows us to expand @xmath70 in legendre polynomials @xmath71 , @xmath72 this expansion will turn to be useful to compute the two - point function of the temperature anisotropies .
it is customary to characterize the properties of the temperature fluctuations by the angular two - point function @xmath73 , where the @xmath74 are the coefficients in the spherical harmonic decomposition @xmath75 and @xmath44 is our position in space ( because of statistical homogeneity , the statistical properties of the temperature fluctuations do not depend on @xmath44 . )
we would like to express this angular two - point function in terms of the initial power spectrum . using the expansion ( [ eq : expansion_legendre ] ) and the addition theorem ( [ eq : addition - theorem ] ) we find first @xmath76 then , using the definition of the power spectrum ( [ eq : power - spectrum ] ) and its expansion in spherical harmonics ( [ eq : spherical_power ] ) we arrive at our sought result @xmath77 and we have used equation ( [ eq : three_spherical ] ) in the appendix .
the structure of the two - point function can be understood in the language of addition of angular momenta .
the @xmath78 multipole of the primordial anisotropies can be thought of as a state with angular momentum quantum numbers @xmath79 and @xmath80 .
the temperature perturbation @xmath81 creates " angular momentum with quantum numbers @xmath82 , and , similarly , @xmath83 destroys angular momentum with quantum numbers @xmath84 .
therefore , the contribution of @xmath34 to the two - point function vanishes if the angular momenta @xmath85 and @xmath78 can not add to @xmath84 .
in fact , @xmath29 is zero unless @xmath86 and @xmath87 .
in general it is only possible to compute the transfer functions @xmath88 in equation ( [ eq : k ] ) numerically . in order to provide a better handle on statistical anisotropies , in the following
we consider a limit where it is possible to analyze them analytically , namely , on large angular scales
. as it is well known , on these scales , neglecting the integrated sachs - wolfe effect , the sachs - wolfe relation @xmath89 holds @xcite , where @xmath90 denotes the comoving distance to last scattering .. ] it is important to realize that this result is an approximation with a limited range of validity . in particular , at @xmath91 the integrated sachs - wolfe effect modifies .
since we do not know what is causing cosmic acceleration , this is just a guess .
other forms of dark energy lead to different isw effects . ]
the relation ( [ eq : delta_approximation ] ) @xcite , and at @xmath92 , the baryon velocity significantly contributes to the temperature anisotropies @xcite .
therefore , the approximation ( [ eq : delta_approximation ] ) is only valid around @xmath93 . just as one parametrizes the primordial spectrum in terms of an amplitude and a spectral index ,
let us parameterize the different components of the anisotropies @xmath26 as power laws .
it follows then from equation ( [ eq : p_lm ] ) that @xmath34 is a superposition of different power - laws , with different amplitudes and spectral indices .
let us assume for simplicity , and without loss of generality that @xmath94 in this case , the integral in equation ( [ eq : k ] ) can be explicitly evaluated . using equations ( 11.4.33 ) and ( 15.1.20 ) in @xcite
we arrive at @xmath95\gamma\left[(l_2+l_1+n-1)/2\right ] } { 9\,\gamma\left[(l_1-l_2 + 4-n)/2\right ] \gamma\left[(l_2-l_1 + 4-n)/2\right ] \gamma\left[(l_2+l_1 + 5-n)/2)\right ] } \frac{a_{lm}}{(k _ * d_s)^{n-1}}.\ ] ] hence , given the amplitude @xmath96 and the spectral index in any given anisotropy multipole , one can exactly predict what the two - point function on large angular scales should be . in general
however , we do not know what @xmath96 or @xmath97 , are . under these circumstances ,
equation ( [ eq : explicit_k ] ) in combination with ( [ eq : two - point ] ) could be used to fit for the values of @xmath96 and @xmath97 ( see below . ) if primordial perturbations are statistically isotropic ( @xmath30 for @xmath31 ) , one can easily verify that the two point function ( [ eq : two - point ] ) is diagonal . in this case
it is customary to define the angular power spectrum @xmath98 through @xmath99 but if perturbations are statistically anisotropic the two - point function is not diagonal , and the definition of the angular power spectrum above is meaningless . in any case , one never measures the @xmath98 s directly .
instead , one uses an appropriate estimator to infer their values . here
we shall define the @xmath98 s as the expectation value of the estimator that is commonly used to compute the latter , @xmath100 the angular power spectrum defined this way is invariant under rotations .
it reduces to ( [ eq : conventional_c ] ) if perturbations are statistically isotropic and it is also closer to the quantities that we actually measure .
note however that in the presence of statistical anisotropies , the @xmath98 s defined in this way do not completely characterize the statistical properties of the temperature fluctuations , since the two - point function is not diagonal .
we want to find out now how the angular power spectrum is related to the primordial power spectrum .
to this end , we substitute equation ( [ eq : two - point ] ) into the definition ( [ eq : c_l ] ) . using then the identity equation ( [ eq : d_identity ] ) we get @xmath101 therefore , @xmath102 acts like an effective power spectrum .
if primordial perturbations are statistically isotropic , then the effective power spectrum reduces to the conventional power spectrum , @xmath103 .
otherwise , the effective power spectrum receives contributions from all anisotropy components , @xmath104 where we have used equations ( [ eq : orthonormality ] ) and ( [ eq : complex - conjugation ] ) . let us stress that the angular power spectrum defined in equation ( [ eq : c_l ] ) does not differentiate between the different anisotropy multipoles .
a scale invariant monopole perturbation ( @xmath105 ) leads to the same set of @xmath98 s as a scale invariant anisotropy in any other multipole ( @xmath26 ) .
hence , conventional numerical codes , such as cmbfast @xcite , or cmbeasy @xcite can be used to determine the imprints of statistical anisotropies on the angular power spectrum .
alternatively , one can use equation ( [ eq : explicit_k ] ) to explicitly evaluate the @xmath98 s on large but not too large angular scales .
the power in the different multipoles of the primordial perturbations @xmath106 uniquely determines the different two - point functions of the temperature anisotropies .
the latter are observables , provided that sufficient realizations of the random field are available .
therefore , one can hope to determine what the @xmath34 are , or at least , place upper limits on their amplitude .
this section is devoted to the ultimate goal of finding ways to measure @xmath34 .
the first obstacle we have to find in our quest to measure @xmath34 is that the power is hidden behind a convolution , equation ( [ eq : k ] ) . even if primordial perturbations are statistically isotropic it is difficult to invert that integral in order to find the primordial power @xcite .
therefore , as a first step towards our goal we shall concentrate in determining the integral @xmath107 in equation ( [ eq : k ] ) . because @xmath108 vanishes when @xmath30 , we can think of @xmath107 as a measure of the anisotropic power ; a non - vanishing @xmath107 for @xmath109 would automatically imply the existence of statistically anisotropic primordial perturbations .
the advantage of focusing on @xmath107 is that we do not have to know what the transfer functions @xmath110 are .
therefore we can analyze all angular scales without having to make any assumptions about the underlying cosmology .
the task of determining some of the values of @xmath107 is greatly simplified when the anisotropies have a cut - off at at multipole @xmath111 , @xmath112 for @xmath113 . in this case , it follows from equation ( [ eq : spherical_power ] ) that the power @xmath34 is cut - off for multipoles @xmath114 .
let us drop the factor of two and assume that @xmath30 for @xmath113 , with @xmath115 even . in this case ,
@xmath108 vanishes for @xmath113 and
we can use the two - point function to directly determine the values of @xmath107 for @xmath111 . at the end of this subsection
, we shall comment on how to iteratively establish whether there exists a putative cut - off at @xmath111 .
consider equation ( [ eq : two - point ] ) with @xmath116 and @xmath117 .
then , due to the properties of angular momentum addition @xmath118 which allows to directly solve for @xmath107 in terms of the two - point function .
however , the two - point function is not directly observable . in order to measure @xmath107
we need a sufficient number of realizations of the random process . at this point , note that the value of @xmath107 does not depend on @xmath80 .
hence we can define the following unbiased estimator , @xmath119 for @xmath120 , our estimator @xmath121 gives the @xmath98 s , equation ( [ eq : c_l ] ) .
therefore , @xmath121 can be regarded as a generalization of the angular power spectrum .
however , note that for @xmath122 the variable @xmath121 is not a scalar , and , in particular , does not transform under an irreducible representation of the rotation group .
let us stress here an additional property of @xmath121 .
namely , because @xmath121 is quadratic in the temperature fluctuations , our study of isotropy is decoupled of issues about non - gaussianity .
if the expectation value of @xmath121 does not vanish for @xmath123 , primordial perturbations are statistically anisotropic , irrespective of whether temperature anisotropies are gaussian or not .
however , we can not measure @xmath124 directly , and the different realizations of the random process @xmath121 are expected to fluctuate around its expectation value .
it is here where we have to make additional assumptions about the nature of the fluctuations . in order to estimate the variance of @xmath121 ,
let us assume for simplicity that perturbations are gaussian and statistically isotropic .
this assumption is justified because we are attempting to detect the existence of statistical anisotropies .
then , under these conditions , we find that the real and imaginary parts of @xmath121 are uncorrelated , and for @xmath122 @xmath125 we encounter here the same cosmic variance problem one faces when trying to measure the angular power spectrum . because for a given @xmath79 we only have a finite number of samples of the two - point function , @xmath126 , we can not measure @xmath127 with infinite precision . by the way , note that if the value of the estimator happens to be consistent with zero , then we guessed incorrectly , and primordial anisotropies are not cut - off at @xmath111 , but eventually at @xmath128 .
as we shall see , the orientation - dependence of our statistic @xmath121 makes the analysis of temperature anisotropies rather cumbersome .
therefore , it may be advantageous to use an estimator of @xmath107 that possesses well - defined properties under rotations . in order to construct it ,
let us consider the quadratic combination @xmath129 where @xmath130 is a clebsch - gordan coefficient .
the expectation value of @xmath131 is nothing else but the bipolar spherical harmonic of hajian and souradeep @xcite .
it can be shown @xcite that under rotations @xmath131 transforms under an irreducible representation of the rotation group , @xmath132 where @xmath29 is the wigner matrix of the rotation .
using equation ( [ eq : two - point ] ) , we can express the expectation value of @xmath131 ( the bipolar coefficients ) in terms of our quantity @xmath107 , @xmath133 again , the previous expression simplifies if there is an anisotropy cut - off at multipoles @xmath113 . in that case
, one can use the previous expression as a guidance to derive an unbiased estimator of @xmath107 , @xmath134 which is guaranteed to transform appropriately under rotations .
however , as opposed to the previous case , we only have one realization of the quantity we are trying to estimate , so our estimator suffers from a larger cosmic variance . as a matter of fact , the mean square fluctuations of @xmath135 are given by @xmath136 where we have used equations ( [ eq : d ] ) and ( [ eq : symmetry ] ) and , once again , we have assumed that perturbations are statistically anisotropic .
it is apparent from the last equation , that the variance of @xmath121 , equation ( [ eq : k_variance ] ) , is smaller that the one of @xmath135 . in this paper
we shall concentrate on the lower variance estimator @xmath121 ; we reserve the study of @xmath135 and related quantities to a hopefully forthcoming publication . to conclude this section ,
let us comment on the relation between our integral @xmath107 and the bipolar spectrum we briefly mentioned during our previous discussion .
in order to characterize the statistical anisotropy of the cmb sky , hajian and souradeep @xcite define the bipolar spherical coefficient to be , up to a sign , the expectation value of @xmath131 , equation ( [ eq : bipolar ] ) . in that respect @xmath108 , equation ( [ eq : two - point ] ) , and
the bipolar spherical coefficients are similar : both vanish if the cmb sky is statistically isotropic and viceversa . to reduce the errors of their estimates , hajian and souradeep also define the bipolar spectrum , which is quadratic in the bipolar spectrum coefficients , and hence quartic in the @xmath74 s .
we have not followed this path here , but have defined instead suitable estimators that are quadratic in the temperature fluctuations .
the main difference between out approach and the one of @xcite though is the physical focus . whereas the authors of @xcite are concerned with the statistical properties of the cmb by itself , we are ultimately interested in the statistical properties of the primordial perturbations .
hence , our statistics @xmath121 and @xmath135 ( and as we shall see also @xmath137 and @xmath138 ) mirror the properties of the primordial perturbations and owe its definition to a possible origin of statistical anisotropies in the cmb .
our analysis so far has been fairly general .
we have not made any strong assumptions about the nature of the primordial perturbations , the content of the universe or the angular scales involved .
this is why our handle on statistical anisotropies is not as strong as it could be . at this point
we shall trade generality for constraining power . on large angular scales ,
the integral @xmath107 can be explicitly evaluated , equation ( [ eq : explicit_k ] ) .
therefore , we can use large angular scales to explicitly determine the power in the primordial anisotropy multipoles . for simplicity ,
let us assume that the power @xmath34 is scale invariant in the scales of interest , @xmath139 . in a more sophisticated analysis we would try to fit for the value of the spectral index @xmath97 . setting accordingly @xmath140 in equation ( [ eq : explicit_k ] ) we find that the integral is proportional to @xmath141 \gamma[(l_2+l_1)/2 ] } { \gamma[(l_1-l_2 + 3)/2]\gamma[(l_2-l_1 + 3)/2]\gamma[(l_2+l_1 + 4)/2]},\ ] ] which does not depend on @xmath79 or @xmath80 .
therefore , substituting its value into equation ( [ eq : k_estimator ] ) we arrive at the following estimator for the amplitude @xmath142 @xmath143 where @xmath115 is again the multipole above which the power in the multipoles is cut - off .
note that we have actually found a set of estimators ( one for each @xmath79 ) of the very same amplitude @xmath142 .
hence , we can use this degeneracy to improve the errors of our measurements . define the averaged estimator @xmath144 in order to improve our measurement of @xmath142 , we would like the difference @xmath145 to be as large as possible .
the values of @xmath146 and @xmath147 are essentially determined by the regime where our large - scale approximation is valid .
therefore , following our previous discussion , we will set @xmath148 and @xmath149 in future evaluations . assuming again that primordial perturbations are gaussian and statistically isotropic we find that the mean square fluctuations of our estimator are @xmath150 where the variances of the individual estimators essentially follow from equation ( [ eq : k_variance ] ) @xmath151 so far we have concentrated on the amplitude of the primordial multipoles right below an eventually existing cut - off above @xmath111 . as opposed to our previous discussion of the integrals @xmath107 , in the case at hand
it is possible to easily determine @xmath152 for all values of @xmath153 by a recursive procedure .
imagine we know @xmath154 with sufficient accuracy for @xmath155 .
then , using equation ( [ eq : two - point ] ) we can solve for @xmath152 , @xmath156 because the left hand side of the equation does not depend on @xmath157 , while the right hand side does , we could use it to construct an unbiased estimator for @xmath152 and compute its variance , just as we did before . however , since we shall not need these estimators here , we shall not pursue this venue any further . for completeness , let us mention that in full analogy with the definition of @xmath158 , one can also derive an estimator @xmath159 of the amplitude @xmath142 that transforms under an irreducible representation of the rotation group , @xmath160 its variance trivially follows from equation ( [ eq : cal_k_variance ] ) ; as in the case of @xmath161 the variance of @xmath138 is bigger than the one of @xmath137 . as mentioned earlier
, we shall not further consider the estimator @xmath138 here .
let us finally apply the formalism we have developed to the anisotropies we observe in the cmb sky .
imagine that , for whatever reason , primordial perturbations display a preferred direction .
more precisely , suppose that primordial perturbations contain a dipolar component @xmath162 .
this dipole gives rise to a dipole @xmath163 and a quadrupole @xmath164 in the primordial spectrum . in the following
we shall test for the existence of the quadrupole component .
a discussion of the impact of a preferred direction on the cmb sky can be also found in @xcite .
assume hence that perturbations in our universe are statistically anisotropic , and @xmath34 only vanishes for @xmath165 .
we can then use the estimator defined in equation ( [ eq : k ] ) to measure @xmath166 . if there is no power in the anisotropy multipole @xmath167 , then @xmath168 vanishes .
therefore , the statistic @xmath121 probes the existence of preferred direction in the primordial perturbations .
( data points ) and its root mean square fluctuation in a statistically isotropic universe ( continuous line ) .
[ fig : k],title="fig:",width=302 ] ( data points ) and its root mean square fluctuation in a statistically isotropic universe ( continuous line ) . [
fig : k],title="fig:",width=302 ] ( data points ) and its root mean square fluctuation in a statistically isotropic universe ( continuous line).[fig : a],title="fig:",width=302 ] ( data points ) and its root mean square fluctuation in a statistically isotropic universe ( continuous line).[fig : a],title="fig:",width=302 ] we have numerically evaluated @xmath169 using the foreground cleaned maps described in references @xcite ( henceforth toh ) and @xcite ( henceforth lilc ) .
any conclusion derived from these maps should be interpreted with care , as it has been argued that they are contaminated by residual foregrounds and their noise properties are ill - understood @xcite . however , because sky cuts in cmb maps introduce artificial anisotropies @xcite , at this point we are forced to proceed with these all - sky maps , in the hope that they accurately capture the real temperature fluctuations @xcite .
for illustration , in figure [ fig : k ] we plot the real and imaginary parts of @xmath169 for values of @xmath79 ranging between @xmath170 and @xmath171 and @xmath172 .
our choice of the upper @xmath79 limit is dictated by our neglect of instrumental noise in the error budget .
indeed , the errors in wmap s measurement of the power spectrum @xcite are cosmic variance dominated ( they contribute more than @xmath173 of the total error ) only up to a multipole of about @xmath174 .
the error contours delimit the mean square fluctuations of the estimator under the assumption of isotropy , equation ( [ eq : k_variance ] ) . in order to evaluate the latter quantities , we use the @xmath63cdm model that best fits the wmap data only @xcite .
it is evident from the figure that both toh and lilc maps lead to similar but not quite identical results .
we can determine whether data is well - fit by statistical isotropic perturbations by computing the value of @xmath175 for the statistic @xmath121 . in order to determine how @xmath175 is distributed ,
we have generated @xmath176 random skies drawn from a statistically isotropic distribution with wmap s best fit angular power spectrum .
the fraction of realizations @xmath177 that have an overall @xmath175 smaller than the one of the cmb sky in galactic coordinates is @xmath178 for the toh map and @xmath179 for the lilc map .
however , these results are misleading , as it turns out that the values of @xmath175 strongly depend on the orientation of the map .
this is due to the non scalar nature of @xmath180 which is a consequence of the peculiar transformation properties of our estimator under rotations . in figure
[ fig : orientation ] we plot how @xmath181 changes as a function of the map orientation . as seen in the figure , the value of @xmath181 widely varies with the orientation , so it is unclear at this point whether the map is consistent with statistical isotropy . as a function of rotation angle .
the cmb map is rotated along the @xmath182-axis ( left ) and @xmath183-axis ( right ) of the galactic coordinate system .
note that the plots have a period of @xmath184 .
chi square is invariant under rotations around the @xmath185-axis ( not shown ) .
[ fig : orientation],title="fig:",width=302 ] as a function of rotation angle .
the cmb map is rotated along the @xmath182-axis ( left ) and @xmath183-axis ( right ) of the galactic coordinate system .
note that the plots have a period of @xmath184 .
chi square is invariant under rotations around the @xmath185-axis ( not shown ) .
[ fig : orientation],title="fig:",width=302 ] if the cmb sky is statistically isotropic , as we rotate the temperature anisotropy map into all possible orientations , we expect the corresponding values of @xmath181 to sample the distribution of random skies @xmath186 we have considered earlier . in figure
[ fig : comparison ] we compare the probability distribution function of @xmath175 for an isotropic sky with the distribution @xmath187 we obtain by randomly orienting the actual cmb map . at first sight
, it seems that both distributions are quite similar . in order to quantify the significance of their agreement or eventual disagreement ,
let us consider the probability @xmath188 that the @xmath175 of a statistically isotropic sky is less than the one of the randomly oriented actual cmb map , @xmath189 where , again , @xmath190 is the fraction of statistically isotropic random skies with a chi square less than @xmath191 , and @xmath192 is the probability that the randomly oriented actual cmb mas has chi square equal to @xmath181 .
if the chi square values of a map happened to be invariant under rotations and equaled @xmath191 , we would have @xmath193 , so @xmath188 would simply be @xmath177 . on the other hand , if both @xmath194 and @xmath195 represented the same distribution we would get @xmath196 . integrating equation ( [ eq : p ] )
numerically we find @xmath197 that is , there is no significant evidence for statistical anisotropy .
values of @xmath121 for statistically isotropic random skies and random orientations of the actual cmb map .
the distributions have been binned to reduce noise.[fig : comparison],title="fig:",width=302 ] values of @xmath121 for statistically isotropic random skies and random orientations of the actual cmb map .
the distributions have been binned to reduce noise.[fig : comparison],title="fig:",width=302 ] we can try to strengthen these results by considering large angular scales . in figure
[ fig : a ] we plot the values of @xmath198 for @xmath172 and @xmath199 .
again , as shown in figure [ fig : orientation ] , the values of @xmath175 depend on the map orientation . in order to quantify up to
what extent the maps are statistically isotropic on large angular scales , we have computed the probability @xmath188 in equation ( [ eq : p ] ) , applied in this case to the chi square of @xmath158 .
the probabilities are @xmath200 which agree quite well with the previous values .
this is to be expected to some extent , since @xmath121 and @xmath137 are proportional to each other .
however , note that the @xmath175 of @xmath121 and @xmath137 involve sums over different ranges of @xmath79 . finally , to confirm our conclusions ,
let us evaluate the estimator for the amplitude of an eventual quadrupole anisotropy component , equation ( [ eq : a ] ) . in figure
[ fig : a_tot ] we plot how the corresponding @xmath175 varies as a function of the map orientation and compare , as in the previous cases , randomly simulated skies with random orientations of the actual cmb sky . in this case , signs of statistical anisotropies are still absent .
indeed , the probability ( [ eq : p ] ) for the statistic @xmath201 is @xmath202 as a function of rotation angle .
the cmb map is rotated along the @xmath183-axis in galactic coordinates . in the right panel
we plot the distribution of values of @xmath175 for randomly generated skies and randomly chosen orientations of the lilc map .
[ fig : a_tot],title="fig:",width=302 ] as a function of rotation angle .
the cmb map is rotated along the @xmath183-axis in galactic coordinates . in the right panel
we plot the distribution of values of @xmath175 for randomly generated skies and randomly chosen orientations of the lilc map .
[ fig : a_tot],title="fig:",width=302 ]
in this paper we have proposed a model - independent parametrization of the primordial spectrum that accommodates statistical anisotropies .
we have determined how these anisotropies impact the temperature fluctuations in the cosmic microwave background , and what type of footprints are left behind by the former .
remarkably , the angular power spectrum does not differentiate between statistically isotropic and anisotropic perturbations .
as far as the @xmath98 s are concerned , the angular power spectrum is as well fit by an isotropic primordial spectrum than by an anisotropic one . in order to find out whether primordial perturbations are statistically isotropic
, one has to study the non - diagonal components of the two - point correlation function . in this paper , we have considered a set of statistics that measure the amplitude of the different multipoles in the primordial perturbations .
the first set , @xmath121 and @xmath137 , does not transform under an irreducible representation of the rotation group , but has a lower variance than the second set , @xmath135 and @xmath138 , which does possess well - defined transformation properties under rotations . in this paper , we have concentrated on the first set , leaving the second for future work .
when applied to the full - sky maps of the cmb temperature fluctuations currently available , the former statistics do not lead to evidence for statistical anisotropy , although the analysis is obscured by the strong variations of our estimators with the map orientation .
the statistical isotropy of the primordial perturbations is an observational issue . as with gaussianity , scale invariance or adiabaticity
, it should be constrained rather than postulated or automatically implied . in this paper
we have taken the first steps in that direction . if at the end of the journey it turns out that observational data is inconsistent with isotropic primordial perturbations , our current ( inflationary ) models for the origin of structure will have to be discarded and replaced by radically different ones .
some of the results in this paper have been derived using the gnu gsl numerical libraries and the healpix @xcite and @xcite software packages .
the author thanks simon catterall , wayne hu , dragan huterer , levon pogosian and mark trodden for valuable feedback and comments , and benjamin wandelt for useful conversations during the embrionary stages of this project .
i am particularly grateful to wayne hu for inquiring about the behavior of our estimators under rotations .
in this appendix we list a series of useful formulae , mainly concerning spherical harmonics .
we follow the conventions in @xcite ( which agree with the ones of @xcite . ) * expansion of plane wave in legendre polynomials @xmath203 * addition theorem @xmath204 * orthonormality of spherical harmonics @xmath205 * behavior under complex conjugation @xmath206 * addition of angular momenta @xmath207 where @xmath208 the @xmath209 are ( real ) clebsch - gordan coefficients . * symmetry properties of clebsch - gordan coefficients @xmath210 * integral of three spherical harmonics @xmath211 note that @xmath212 is invariant under spatial rotations .
hence , it follows from equations ( [ eq : three_spherical ] ) and ( [ eq : orthonormality ] ) that @xmath213 * relation between clebsch - gordan coefficients and wigner s 3-j symbol @xmath214 d. w. hogg , d. j. eisenstein , m. r. blanton , n. a. bahcall , j. brinkmann , j. e. gunn and d. p. schneider , `` cosmic homogeneity demonstrated with luminous red galaxies , '' astrophys . j. * 624 * , 54 ( 2005 ) [ arxiv : astro - ph/0411197 ] . c. l. bennett _ et al .
_ , `` first year wilkinson microwave anisotropy probe ( wmap ) observations : preliminary maps and basic results , '' astrophys . j. suppl .
* 148 * , 1 ( 2003 ) [ arxiv : astro - ph/0302207 ] .
d. n. spergel _ et al . _
[ wmap collaboration ] , `` first year wilkinson microwave anisotropy probe ( wmap ) observations : determination of cosmological parameters , '' astrophys .
j. suppl .
* 148 * , 175 ( 2003 ) [ arxiv : astro - ph/0302209 ] .
a. de oliveira - costa , m. tegmark , m. zaldarriaga and a. hamilton , `` the significance of the largest scale cmb fluctuations in wmap , '' phys .
d * 69 * , 063516 ( 2004 ) [ arxiv : astro - ph/0307282 ] .
ralston and p. jain , `` the virgo alignment puzzle in propagation of radiation on cosmological scales , '' int .
j. mod .
d * 13 * , 1857 ( 2004 ) [ arxiv : astro - ph/0311430 ] .
d. j. schwarz , g. d. starkman , d. huterer and c. j. copi , `` is the low - l microwave background cosmic ?
, '' phys .
lett . * 93 * , 221301 ( 2004 ) [ arxiv : astro - ph/0403353 ] .
c. j. copi , d. huterer and g. d. starkman , `` multipole vectors a new representation of the cmb sky and evidence for statistical anisotropy or non - gaussianity at 2<=l<=8 , '' phys .
d * 70 * , 043515 ( 2004 ) [ arxiv : astro - ph/0310511 ] .
j. copi , d. huterer , d. j. schwarz and g. d. starkman , `` on the large - angle anomalies of the microwave sky , '' arxiv : astro - ph/0508047 . h. k. eriksen , f. k. hansen , a. j. banday , k. m. gorski and p. b. lilje , `` asymmetries in the cmb anisotropy field , '' astrophys .
j. * 605 * , 14 ( 2004 ) [ erratum - ibid .
* 609 * , 1198 ( 2004 ) ] [ arxiv : astro - ph/0307507 ] .
a. hajian and t. souradeep , `` the cosmic microwave background bipolar power spectrum : basic formalism and applications , '' arxiv : astro - ph/0501001 .
a. hajian , t. souradeep and n. j. cornish , `` statistical isotropy of the wmap data : a bipolar power spectrum analysis , '' astrophys .
j. * 618 * , l63 ( 2004 ) [ arxiv : astro - ph/0406354 ] .
a. hajian and t. souradeep , `` measuring statistical isotropy of the cmb anisotropy , '' astrophys .
j. * 597 * , l5 ( 2003 ) [ arxiv : astro - ph/0308001 ] .
j. levin , `` topology and the cosmic microwave background , '' phys .
* 365 * , 251 ( 2002 ) [ arxiv : gr - qc/0108043 ] .
a. riazuelo , j. weeks , j. p. uzan , r. lehoucq and j. p. luminet , `` cosmic microwave background anisotropies in multi - connected flat spaces , '' phys .
d * 69 * , 103518 ( 2004 ) [ arxiv : astro - ph/0311314 ] .
e. f. bunn , p. ferreira and j. silk , `` how anisotropic is our universe ?
, '' phys .
* 77 * , 2883 ( 1996 ) [ arxiv : astro - ph/9605123 ] .
t. r. jaffe , a. j. banday , h. k. eriksen , k. m. gorski and f. k. hansen , `` evidence of vorticity and shear at large angular scales in the wmap data : a violation of cosmological isotropy ?
, '' arxiv : astro - ph/0503213 .
d. grasso and h. r. rubinstein , `` magnetic fields in the early universe , '' phys .
rept . *
348 * , 163 ( 2001 ) [ arxiv : astro - ph/0009061 ] .
m. giovannini , `` magnetized cmb anisotropies , '' arxiv : astro - ph/0508544 .
v. a. kostelecky , `` gravity , lorentz violation , and the standard model , '' phys .
d * 69 * , 105009 ( 2004 ) [ arxiv : hep - th/0312310 ] .
t. jacobson , s. liberati and d. mattingly , `` lorentz violation at high energy : concepts , phenomena and astrophysical constraints , '' arxiv : astro - ph/0505267 .
v. f. mukhanov , h. a. feldman and r. h. brandenberger , `` theory of cosmological perturbations .
classical perturbations .
quantum theory of perturbations .
extensions , '' phys .
rept . * 215 * , 203 ( 1992 ) .
l. kofman and a. a. starobinsky , `` effect of the cosmological constant on large scale anisotropies in the microwave backbround , '' sov .
* 11 * , 271 ( 1985 ) [ pisma astron .
11 * , 643 ( 1985 ) ] .
h. k. eriksen , a. j. banday , k. m. gorski and p. b. lilje , `` foreground removal by an internal linear combination method : limitations and implications , '' astrophys . j. * 612 * , 633 ( 2004 ) [ arxiv : astro - ph/0403098 ] .
a. slosar and u. seljak , `` assessing the effects of foregrounds and sky removal in wmap , '' phys .
d * 70 * , 083002 ( 2004 ) [ arxiv : astro - ph/0404567 ] .
|
we propose and develop a formalism to describe and constrain statistically anisotropic primordial perturbations . starting from a decomposition of the primordial power spectrum in spherical harmonics
, we find how the temperature fluctuations observed in the cmb sky are directly related to the coefficients in this harmonic expansion .
although the angular power spectrum does not discriminate between statistically isotropic and anisotropic perturbations , it is possible to define analogous quadratic estimators that are direct measures of statistical anisotropy . as a simple illustration of our formalism we test for the existence of a preferred direction in the primordial perturbations using full - sky cmb maps .
we do not find significant evidence supporting the existence of a dipole component in the primordial spectrum .
|
quantum key distribution ( qkd ) harnesses the laws of quantum mechanics to distribute a secret key with a security level unachievable by classical means @xcite . in a qkd protocol
two parties , commonly called alice ( message sender ) and bob ( message receiver ) , want to establish a secret key between them , by sending quantum and classical information through an insecure channel .
the communication channel can be spied by a powerful eavesdropper , typically called eve , who is assumed to be technologically much more advanced than alice and bob .
thus , eve can listen all the transmitted classical messages and manipulate the quantum information at her own will , in principle , being only limited by the laws of quantum physics .
however , this manipulation unavoidably introduces perturbations of quantum nature , which may be detected by the communication partners ( alice and bob ) , depending on the features of the communication protocol they are employing . in such a way , they are able to determine how much information was leaked and keep only the secure part in the final secret key . standard qkd protocols ( one - way protocols ) , such as bb84 @xcite , six - state @xcite and and sarg04 @xcite , use an encoding method that prevents bob to decode the information in a deterministic way .
this means that a fraction of the transmitted key bits must be discarded , due to the fact that bob s measurements do not allow him to deduce with certainty the corresponding values .
on the other hand , two - way quantum key distribution ( twqkd ) protocols @xcite can provide a solution to this flaw . in this kind of protocols the preparation of the quantum states that will be encoded
is carried out by bob .
therefore , with this additional knowledge , he can always perform the right decoding measurements and no bits must be discarded . an important factor to assess
the performance of a given qkd protocol is the key rate , that is , the number of secret key bits distributed per unit of time . besides the improvements related to technological developments @xcite ,
the use of qudits instead of two - level systems has been proposed as an option to achieve larger secret key rates @xcite .
twqkd protocols have also the potential to fulfill this goal , given that the key bits can be decoded more efficiently than with one - way schemes .
regarding security , it was recently shown that the performance of some twqkd protocols is comparable to that of their one - way counterparts and can even surpass it @xcite .
the two paradigmatic twqkd schemes proposed to date are lm05 ( based on non - orthogonal states ) @xcite and the ping - pong protocol ( based on entanglement ) @xcite , both of them being deterministic . in spite
that the original ping - pong protocol is vulnerable to zero error attacks @xcite , there exists a non - deterministic version that overcomes this shortcoming and has been proven secure @xcite . on the other hand , security proofs for lm05 and a deterministic twqkd protocol inspired on it ( lm05 ) are given in @xcite and @xcite , respectively . in this work
we investigate the security of several twqkd schemes based on non - orthogonal states , including deterministic and non - deterministic ones .
the proofs that we construct guarantee security against collective attacks , when the classical post - processing is implemented with `` direct reconcilliation '' @xcite . in section
ii we describe the basic structure of the protocols studied here . in section
iii we present a simple twqkd protocol , whose security proof constitutes the basis of some subsequent results .
next , we prove in section iv the security of the lm05 protocol , by introducing a slight modification that reduces the corresponding analysis to that of the scheme described in section iii . finally , section v is focused on the study of two new deterministic twqkd protocols .
one of them can be seen as a generalization of the lm05 protocol , due to the use of more than two preparation bases .
we prove that the security of this generalized protocol is the same of lm05 .
for the latter protocol , which we term `` twqkd six - state '' , we provide a lower bound for the eavesdropping information that coincides with the corresponding to the six - state protocol @xcite .
the security of the novel twqkd six - state protocol is at least as good as that of lm05 and probably even bigger ( considering the tightness of the bound tracked down ) .
all the calculations are performed analytically , showing explicitly the attacks that allow eve to get a given amount of information .
we take advantage of the so called gram matrix formulation @xcite in our analysis .
the basic modus operandi of the protocols considered here is the following ( see fig . [ fig:1 ] for a general description ) : 1 .
bob ( the message receiver ) prepares the @xmath0-th qubit state from a basis set @xmath1 , according to some probability distribution , and sends it to alice through the _ quantum forward channel _ ( qfc ) .
alice ( the message sender ) performs the encoding mode ( em ) with probability @xmath2 and control mode ( cm ) with probability @xmath3 ( @xmath4 ) . in em
she encodes one bit of information , by randomly applying unitary operations from a predetermined set @xmath5 , and sends back the qubit to bob through the _ quantum backward channel _ ( qbc ) . instead , in cm
, she measures the received qubit in a given basis , by choosing randomly among the possible preparation bases used by bob , to get information about the prepared state .
it is important to note that the qfc and the qbc are the same physical channel and the distinction is used to facilitate the description of the eavesdropping attacks .
3 . bob measures the qubit in the @xmath0-th preparation basis to decode the information .
4 . after all the qubits have been transmitted and measured , a leftover part of the protocol is carried out throughout a classical channel .
alice publicly reveals which qubits were used for cm and which ones for em . in the case of cm alice and bob
compare the prepared states and the measurement outcomes , whenever alice s measurements had been performed properly in the preparation basis .
this procedure allows them to establish the _ forward noise _ @xmath6 , defined as the probability that such states do nt match up . on the other hand ,
a fraction of the systems used for em is randomly chosen to determine the probability that encoded and decoded symbols be different , @xmath7 , the overall noise .
the remaining symbols are not disclosed and constitute what is known as the `` raw key '' .
_ classical post - processing _ : alice and bob interchange additional classical information in order to obtain identical and completely secret bit strings starting from their raw keys . to this aim
two sub - protocols are required .
the first is called `` error correction '' @xcite and it is used to amend the errors associated to the overall noise @xmath7 . in this way , the encoded and decoded keys become perfectly correlated , i.e. , alice s and bob s bits are identical . in the following sub - protocol information potentially leaked during the quantum transmission
is removed from the key , which is known as `` privacy amplification '' @xcite . here
we will consider the usual approach for the classical post - processing , where the involved sub - protocols are implemented with one - way classical communication ( see , e.g. , @xcite for a different approach ) . and the unitary operations @xmath5 are chosen probabilistically by bob and alice , respectively .
bob performs the decoding measuring the qubit in the preparation basis and for cm alice measures the received qubit in any of the possible preparation bases ( randomly chosen ) .
alice also chose randomly between em and cm with probabilities @xmath2 and @xmath3 , respectively .
see the general protocol description in the main text for more details . ]
let us consider a generic twqkd protocol where alice uses the encoding operations set @xmath8 ( the bits `` 0 '' and `` 1 '' are encoded with the application of the unitary operation @xmath9 and @xmath10 , respectively ) and bob s preparation is such that eve observes the state @xmath11 in the qfc . in what follows we will denote operators acting on the encoded qubit hilbert space with the superscript @xmath12 .
other details , such as the preparation bases used by bob , are irrelevant by now . for keys of infinite size
, the secret fraction , @xmath13 , is given by @xcite @xmath14 where @xmath15 is the mutual information between alice s and bob s data , which measures the correlation of the encoded and decoded keys .
if alice sends one bit , i.e. , she uses @xmath9 and @xmath10 with equal probability , @xmath16 , being @xmath17 the binary entropy and @xmath7 the probability of miss - matching between the encoded and decoded bit . here ,
we assume ( as usual ) that the quantum channel is a depolarizing channel @xcite and hence @xmath7 does not depend on the specific states carrying the information .
moreover , @xmath18 represents the maximum information that the eavesdropper ( eve ) can obtain using her best strategy ( the expression for @xmath18 will be presented latter ) .
we consider the case of collective attacks and classical post - processing implemented with `` direct reconciliation '' , implying that @xmath18 corresponds to eve s knowledge about the encoded key @xcite .
accordingly , the most general attack that eve can perform in the qfc consists of a joint interaction between the sent qubit and some auxiliar ancilla . without loss of generality ,
this interaction may be represented as @xmath19 where @xmath20 is a unitary operation acting on the joint hilbert space of the ancilla and the qubit , and @xmath21 is the ancilla state before the attack .
the set @xmath22 contains the eigenstates of the pauli operator @xmath23 , which corresponds to the @xmath24 direction in the bloch sphere .
unitarity is guaranteed whenever the ancilla states @xmath25 fulfil the conditions : @xmath26 after the forward attack , eve resends the qubit to alice and stores her ancilla until the qubit is sent back through the qbc . in this way ,
the joint encoded states that she can access in the qbc are @xmath27 and @xmath28 , with @xmath29 ( @xmath30 being the state conditioned to the alice s encoded bit `` 0 '' ( `` 1 '' ) . following the definition of collective attack
, eve can extract the encoded information by performing coherent measurements on any number of qubits and ancillae that she considers convenient . to this aim
, it is assumed that she has an unlimited resource of quantum memory at her disposition @xcite .
accordingly , the expression for eve s information is given by @xcite : @xmath31\right\},\label{eq:3}\end{aligned}\ ] ] where @xmath32 is the von neumann entropy and @xmath33 is the holevo quantity @xcite for the alphabet encoded into @xmath34 .
the maximization is taken over all possible eve s unitaries .
since the conditional density operator @xmath35 only differ of @xmath36 by unitary transformations , we can write @xmath37 . hence , @xmath38,\label{eq:4}\ ] ] with @xmath39 . to compute eve s information , in eq . , we need first to obtain the eigenvalues of the alice - eve joint state @xmath40 .
we will perform this task employing the gram matrix representation @xcite . for a mixed state ( written as a mixture of non - orthogonal pure states )
@xmath41 , @xmath42 , the elements of the gram matrix , @xmath43 , are defined as @xmath44 we take advantage of the fact that the eigenvalues of @xmath43 and @xmath45 are the same ( including their multiplicities ) @xcite , and that @xmath40 is by construction a mixture of pure states . in the appendix a we write explicitly this matrix , which is of dimension @xmath46 , and compute its exact eigenvalues .
note that the a priori hilbert space for the joint state @xmath40 is of dimension 8 , given that 4 ancilla states are required to fully describe the eve s forward attack , cf .
this entails a clear advantage of using the gram matrix over the standard method ( in which @xmath40 is written in some orthonormal basis ) , i.e. : first , finding a suitable orthonormal basis is not obvious ; and second , even if that were the case , computing the eigenvalues of a @xmath47 matrix is a considerably more difficult task than doing it for @xmath46 matrix . from eq .
, we deduce that the perturbation that eve s unitary , @xmath20 , causes on the qubit state @xmath48 ( @xmath49 ) is given by @xmath50 ( @xmath51 ) . considering the generic form of a twqkd protocol ( described in section [ section : twqkd ] and fig .
[ fig:1 ] ) the set of bob s prepared states @xmath1 can be written as @xmath52 , where @xmath53 is the set of bloch sphere directions used in the bob s preparation .
the fact that eve has access to the state @xmath11 in the forward channel means that bob prepared @xmath48 and @xmath49 with the same probability , regardless of the preparation basis direction choice , @xmath24 .
as aforementioned , we are assuming that the qubits are sent through a depolarizing channel and consequently @xmath54 for any choice @xmath55 . here
@xmath6 is the natural noise of the forward channel , which is simply the probability that alice receives @xmath48 ( @xmath49 ) if bob prepared @xmath49 ( @xmath48 ) . in this way , eq . tells us that , in order to pass unnoticed , eve is restricted to interactions whose perturbation equals @xmath6 for all the transmitted states .
considering this constraint , we obtain in appendix a the following expressions for the eigenvalues of @xmath40 : @xmath56\nonumber \\ & = \frac{1}{4}\left[1\pm\sqrt{(1 - 2q_{f})^{2}+4|\langle\epsilon_{00}^{\hat{z}}|\epsilon_{10}^{\hat{z}}\rangle|^{2}}\right],\label{eq:7}\end{aligned}\ ] ] each one with multiplicity 2 .
equation is then translated into @xmath57 , for fixed @xmath6 . a simple argument
( see appendix a ) leads to conclude that the von neumann entropy , @xmath58,\ ] ] takes its maximum when @xmath59 .
thereby , @xmath60 and @xmath61 in this case , eq . allows to determine eve s information in terms of the noise introduced along the @xmath62 direction , @xmath63 .
this implies that bob must prepare eigenstates of @xmath10 and alice must measure such an observable in cm .
this kind of protocol is clearly non - deterministic since the encoding operations , @xmath8 , have no effect on the states @xmath64 . in the next subsection
we will see that , regardless of the number of preparation bases used , eve s information can not be less than the given by eq . .
here we show that , if the previous protocol is implemented in a depolarizing channel , no preparation strategy can reduce @xmath18 below the value corresponding to eq . . to this aim
we present an explicit eavesdropping attack that simulates a depolarizing channel with noise @xmath6 and provides eve @xmath65 bits of information .
besides the condition @xmath66 , this attack satisfies the following equations : @xmath67 @xmath68 @xmath69 a general qubit pure state may be written as @xmath70 where @xmath71 and @xmath72 are the azimuthal and polar angles in the bloch sphere , respectively .
the effect of the forward attack is then @xmath73 where @xmath74 is the state orthogonal to @xmath75 and @xmath76 are ancilla states associated to the eve s attack .
the disturbance generated on @xmath75 is @xmath77 , cf .
taking into account eqs . - , we show in the appendix b that @xmath78 .
thus , we have proved that the attack described by eqs .
- fully simulates a depolarizing channel with characteristic noise @xmath6 . on the other hand , eq .
implies that @xmath79 . although this result could suggest that it is sufficient for bob to prepare only eigenstates of @xmath10 , indeed at least one additional basis must be incorporated into the preparation . as already mentioned , this is necessary because the encoding operations @xmath8 do not affect these prepared states ( the eigenstates of @xmath10 ) .
furthermore , the transmission of non - orthogonal states is the key ingredient to guarantee the protection of the information .
accordingly , the simplest protocol whose security is established by eq . could be one where the encoding is performed with the set @xmath8 and bob prepares the states @xmath80 .
the results obtained in this section will be employed to prove the security of the lm05 protocol in what follows .
the lm05 protocol corresponds specifically to version 2 of the `` qubit lm05 protocol , implemented with reverse reconciliation '' , proposed in ref . @xcite . in this scheme
the encoding operations are @xmath81 and bob s preparation is carried out with the states @xmath80 .
moreover , bob measures the received state in the preparation basis , interpreting an unchanged state as the bit `` 0 '' and one flipped as the bit `` 1 '' .
this implies that the protocol is deterministic , since he can always perfectly distinguish the encoded bit with his measurements . on the other hand , the operations @xmath82 can cause a bit flip or act similarly as the identity , depending on the preparation basis , @xmath83 .
therefore , at the end of the quantum transmission bob must reveal the preparation bases in order to alice can infer all the encoded bits .
it is the disclosing of such information what the authors term `` reverse reconciliation '' in @xcite . here
we will consider that bob s choices and the corresponding to alice in each mode of the protocol ( em and cm ) are performed with equal probability . in cm
she measures randomly the observables @xmath84 and @xmath10 to determine the forward noise . according to the the description of lm05 ,
there is not a unique holevo quantity to be maximized for obtaining eve s information ( eq . ) . instead of that , the states from which eve can extract the encoding depend on the associated preparation direction .
suppose for instance that the bob s chosen preparation direction was @xmath62 .
then , @xmath85 $ ] and @xmath86 $ ] , given that the operations sub - set @xmath87 produce a bit flip while the sub - set @xmath8 leave the state unchanged . in this case
the state @xmath40 turns out to be @xmath88 and the corresponding holevo quantity , @xmath89 equation follows from the fact that @xmath90 and the von neumann entropy is invariant under unitary transformations @xcite .
the same kind of reasoning leads to the holevo quantity corresponding to the bob s preparation in the @xmath91 direction : @xmath92 where @xmath93 $ ] .
in this way , eve s information is given by @xmath94 where the factor @xmath95 stems from the fact that bob uses each preparation basis with equal probability .
the upper bound is due to the general relation @xmath96\leq\mathrm{max}f(x)+\mathrm{max}g(x)$ ] , for @xmath97 and @xmath98 arbitrary functions .
the optimization problem involved in eq . could , in principle , be a considerable task .
this is so basically due to the fact that the holevo quantities are subtractions of two terms , each one depending on the general unitary @xmath20 employed in the eavesdropping attack .
moreover , the gram matrix associated to @xmath40 would be of dimension @xmath47 , which prevents a possible analytical advantage associated to the use of such a representation . to circumvent this hindrance
, we introduce a slight modification to the protocol , thereby reducing its security proof to that of the protocol considered in the previous section .
such a modification consists in additional classical information that alice publicly reveals at the end of the quantum transmission , such as follows : * if the preparation direction was @xmath91 and alice used @xmath9 or @xmath10 , she announces that @xmath84 and @xmath99 were discarded .
similarly , if alice used @xmath84 or @xmath99 , she announces that @xmath9 and @xmath10 were discarded . *
if the preparation direction was @xmath62 and alice used @xmath9 or @xmath84 , she announces that @xmath99 and @xmath10 were discarded .
similarly , if alice used @xmath99 or @xmath10 , she announces that @xmath9 and @xmath84 were discarded .
notice that in this scenario eve is unable to obtain directly any information about the encoded bit .
however , the encoded states that she can access in the qbc take a simpler form now .
consider the case when bob s preparation was along the @xmath91 direction and alice used @xmath9 .
thus , we have @xmath100 @xmath101 where we have designed the density operator by @xmath102 instead of @xmath45 to emphasize that these states correspond to the modified protocol .
the corresponding holevo quantity , @xmath103 , takes exactly the same form when alice performed @xmath84 or @xmath99 , due to the invariance of the von neumann entropy under unitary transformations . using the same reasoning , it is easy to obtain the holevo quantity corresponding to the case when the preparation direction is @xmath62 , whose the general expression is given by @xmath104 with @xmath105 and @xmath106 $ ] ; @xmath107 being the complementary direction of @xmath108 , i.e. , @xmath109 if @xmath110 and vice - versa .
the associated expression for eve s information , @xmath111 , is the same of eq . with the modified holevo quantities .
the bound appearing in that equation can be written as @xmath112 where @xmath113 .
it is worthwhile to note that @xmath114 corresponds exactly to @xmath18 in eq . and hence @xmath115 .
to find the expression for @xmath116 , we can proceed in a way completely analogous to the one leading to eq . , writing the gram matrix with the states @xmath117 instead of @xmath118 ( see eq . and appendix a ) .
it is thus not difficult to grasp that the result is @xmath119 . in this way
, the estimation of the noise @xmath63 ( @xmath120 ) allows to determine eve s information when the encoded states were prepared along the @xmath91 ( @xmath62 ) direction . assuming a depolarizing channel we get @xmath121=h(q_{f}).\label{eq:17}\ ] ] at this point
, we remark that the objective of the introduced modification is to simplify the security proof and by no means is necessary in practice .
actually , it potentially reduces eve s uncertainty about the encoding and therefore @xmath111 ( eq . ) is an upper bound to the eve s information , @xmath18 , for the lm05 protocol .
furthermore , such a bound coincides with the one calculated in @xcite employing other methods . on the other hand ,
we show in appendix c that the attack represented by eqs .
- makes such bound in eq .
tight , which implies that for the modified version eve can in fact get @xmath65 bits of information .
considering the lm05 protocol , we note that if besides @xmath62 and @xmath91 one includes the preparation direction @xmath122 and bob always measures in the preparation basis , the resulting scheme is still deterministic .
the reason is that , as in the case of lm05 , the encoding operations @xmath81 always flip or leave unchanged the prepared state depending on the preparation basis .
now let us suppose that alice and bob perform their choices according to a uniform probability distribution .
the resulting scheme , which we term `` twqkd six - state '' ( given the use of the same preparation strategy of the six - state protocol @xcite ) , had not been proposed to date , to the best of our knowledge .
we leave the remaining details of the protocol identical to those of lm05. regarding the decoding , for example , alice and bob agree that a unchanged ( flipped ) state corresponds to the bit `` 0 '' ( `` 1 '' ) . to that aim bob
must publicly disclose his basis preparation choices after the quantum transmission , so that alice can know the effect of her operations in each case .
furthermore , in cm alice should measure any of the observables @xmath123 with equal probability , in order to determine the noise introduced along each direction .
figure [ fig:2 ] illustrates the quantum part of this protocol . for the state preparation , ( in em ) alice s encoding operation , and ( in cm ) alice s measurement basis . ]
similar to the lm05 protocol case , for the twqkd six - state protocol there is a holevo quantity associated to each possible direction chosen by bob : @xmath124 where the alice - eve joint state @xmath40 is given by eq . with the conditional joint state
@xmath125 $ ] , @xmath126 .
likewise , the optimization to obtain eve s information , @xmath127 , is at least as difficult as doing it for lm05 ( eq . ) .
nevertheless , the security of the twqkd six - state scheme immediately follows from that of lm05 , since the only difference is the usage of the additional preparation direction @xmath122 and it constrains more eve s strategies ( she must now cause the same perturbation not only along @xmath91 and @xmath62 but also along @xmath122 ) .
this implies that eve s information is also upper bounded by @xmath65 , according to eq . . in appendix c
we compute the von neumann entropies , @xmath128 and @xmath129 , for the unitary transformation described by eqs . - .
the corresponding eavesdropping information is given by @xmath130 the right hand side of eq .
is identical to the expression for eve s information corresponding to the six - state protocol @xcite .
this is remarkable if one considers that the same kind of states are prepared in both the six - state scheme and its twqkd version .
moreover , the analysed attack is in fact one of the most powerful since it makes tight the bound in eq .
, as we also prove in appendix c. based on these facts , we conjecture that eq
. actually corresponds to @xmath18 for the twqkd six - state protocol .
we can summarize the results concerning the lm05 and twqkd six - state schemes with the following inequallity : @xmath131 these bounds are plotted in fig .
[ fig:5.2 ] , as well as the mutual information between alice and bob ( @xmath15 ) , and the eavesdropping information ( @xmath132 ) for the lm05 protocol . ( blue solid line ) and eve s information ( @xmath132 ) for different protocols , as function of the forward noise ( @xmath133 ) .
it is assumed that the overall noise ( q ) and the forward noise coincide , which is the case if the qfc and the qbc are correlated ( see , e.g. , in ref . @xcite some comments at this respect).@xmath132 for the generalized lm05 protocol
is plotted as a ( dark - green ) dotted line .
the upper bound to @xmath132 for lm05 protocol is represented by a ( dark - yellow ) dashed line .
the lower bound to @xmath132 for the twqkd six - state is drawn as the ( dark - red ) dash - dotted line . ]
now let us consider a twqkd protocol in which bob prepares states from the set @xmath134 , i.e. , any state on the perpendicular plane to the @xmath135 direction of the bloch sphere , and alice performs the encoding with @xmath136 .
such a protocol is clearly deterministic and can be considered as a generalization of the lm05 protocol @xcite , which only uses two bases for the preparation , e.g. @xmath137 and @xmath138 .
we will show that this generalized protocol does not have any advantage over the standard lm05 , within the assumption that the channel used to transmit the states is depolarizing .
we start computing the eavesdropping information for the attack described by the following conditions : @xmath139 @xmath140 @xmath141 this attack only differs from the one represented by eqs . - concerning the value of the element @xmath142 .
in addition , we write @xmath63 instead of @xmath6 because @xmath63 is not the disturbance measured in this protocol . considering that the state @xmath143 is equal to @xmath75 ( defined in section iii b ) for @xmath144
, we can use eq . from appendix b to compute the perturbation that the attack defined by eqs .
- introduces on such a state . expressing the attack unitary operation as @xmath145 , with @xmath146 being the state perpendicular to @xmath147
, we have @xmath148 where eqs . - have been employed .
thus , using the unitarity condition , we have @xmath149 . using this result and the fact that , from eq .
, this attack provides eve @xmath150 bits of information ( it fulfills the condition to get eq . )
, we get @xmath151 we have taken into account that @xmath152 must coincide with the forward noise . the expression in eq .
is exactly the found for the lm05 protocol in @xcite and therefore we have proved that it is not possible to improve the protocol security with this extended version on the number of preparation basis .
we have studied the performance of twqkd protocols that distribute a secret key by means of non - orthogonal qubit states . for the security analysis
we have used techniques whose application is new in the context of secure communications , to the best of our knowledge .
specifically , we employed a matrix representation @xcite that allows to easily compute the eigenvalues of certain density operators , involved in the determination of the amount of the eavesdropping information . in this way , a simple calculation leads to eq . , which is the more important expression for the posterior analysis .
with this equation we provided a new ( alternative ) security proof for a protocol proposed in ref .
@xcite ( that we have named lm05 ) , obtaining an upper bound to the eavesdropping information that coincides with the previous result in @xcite .
moreover , we proposed a novel protocol ( twqkd six - state ) that is at least as secure as lm05 and computed an analytical lower bound to the eavesdropping information .
such a bound equals the maximum eavesdropping information corresponding to the six - state scheme @xcite .
we remark that the security of the twqkd six - state protocol is possibly better than that of the lm05 protocol , since an additional preparation basis is used by bob , and therefore our lower bound could be tight .
finally , we showed that the inclusion of more than two preparation bases does not improve the security of the lm05 protocol @xcite .
the construction of a full security proof for the new twqkd six - state protocol proposed here as well as its experimental analysis and tests are interesting directions for a future research .
* acknowledgments . *
we thank t. b. batalho , s. p. walborn , and l. s. cruz for valuable discussions and suggestions .
we acknowledge financial support from ufabc , cnpq , capes , and fapesp .
rms acknowledges support from the royal society through the newton advanced fellowship scheme ( grant n. r1660101 ) .
this work was performed as part of the brazilian national institute of science and technology for quantum information ( inct - iq ) .
in this appendix we maximize the von neumann entropy for the alice - eve joint state @xmath153 $ ] , assuming that @xmath154 is fixed . using eq . , with @xmath155 , and eq .
, the corresponding gram matrix is explicitly written as @xmath156 the eigenvalues of @xmath43 are given by @xmath157,\ ] ] which implies that the von neumann entropy @xmath158 $ ] will be a function of @xmath6 and @xmath159 . for a fixed @xmath6
we obtain @xmath160,\ ] ] with @xmath161 . taking the derivative with respect @xmath162
, it is easy to see that @xmath128 is monotonically decreasing in the interval @xmath163 and therefore @xmath128 takes its maximum at @xmath164 .
using the expression for a general qubit pure state , @xmath165 , given in section iii b , we can write the action of an eavesdropping interaction @xmath20 as @xmath166 were we have considered the linearity of @xmath20 and eq . has been used , with @xmath155 .
taking into account that @xmath75 and its orthonormal state , @xmath167 , are defined in terms of the states @xmath64 , we have @xmath168 @xmath169 therefore , using eq . , we can rewrite eq .
as @xmath73 with @xmath170 and a similar expression holds for @xmath171 . if the states @xmath118 satisfy eqs
- , then @xmath172 hence , by the unitarity of @xmath20 , @xmath173
here we compute the eavesdropping information for the twqkd six - state protocol , when the attack is given by eqs . - . the average over the associated holevo quantities , eq .
, is : @xmath174-s\left[\rho_{w}^{ae|0}\right]\right\ } .\ ] ] the elements of the gram matrix corresponding to @xmath40 result from the scalar products between the states @xmath175 , being @xmath20 the attack described by eqs . - .
for this attack it turns out that the sets @xmath176 and @xmath177 are orthogonal .
we denote the gram matrices associated to the first and second sets as @xmath178and @xmath179 , respectively , and their elements are ( see eq . ) @xmath180 , @xmath181 .
the orthogonality property implies that the eigenvalues of @xmath40 are derived from the corresponding to @xmath178and @xmath179 . using eqs .
- and the usual assumption of the depolarizing channel , we get @xmath182 @xmath183 one finds that the eigenvalues of @xmath178 and @xmath179 are @xmath184 , the first one having multiplicity 3 .
therefore , @xmath185 for the analysed attack we already know that @xmath186 , according to eq . .
on the other hand , the sets @xmath187 and @xmath188 determine the gram matrices needed to compute @xmath189 $ ] and @xmath190 $ ] .
the obtained expressions are @xmath191 corresponding to @xmath192 , and @xmath193 for @xmath194 .
the eigenvalues of @xmath195 and @xmath196 are @xmath197 , each one with multiplicity 2 .
therefore , @xmath198
|
within the broad research scenario of quantum secure communication , two - way quantum key distribution ( twqkd ) is a relatively new proposal for sharing secret keys that is not fully explored yet .
we analyse the security of twqkd schemes that use qubits prepared in non - orthogonal states to transmit the key . investigating protocols that employ an arbitrary number of bases for the channel preparation , we show , in particular , that the security of the lm05 protocol can not be improved by the use of more than two preparation bases . we also provide
a new proof of unconditional security for a deterministic twqkd protocol recently proposed [ phys .
rev .
a * 88 * , 062302 ( 2013 ) ] .
in addition , we introduce a novel deterministic protocol named `` twqkd six - state '' and compute an analytical lower bound ( which can be tightened ) for the maximum amount of information that an eavesdropper could extract in this case .
an interesting advantage of our approach to the security analysis of twqkd is the great simplicity and transparency of the derivations .
|
in recent years the discovery and characterization of multiple stellar populations in globular clusters ( gcs ) has given a new impetus to gc studies , opening entirely new scenarios for their formation .
first came the discovery that the main sequence ( ms ) of @xmath2 cen splits into two , distinct parallel sequences ( bedin et al .
subsequent spectroscopic analysis showed that , instead of a lower metal abundance , the blue sequence has a _ higher _ metallicity compared to the red sequence , inescapably demanding that blue ms stars must be greatly enriched in helium ( piotto et al .
2005 : see also norris 2004 ) .
this is now quantified in @xmath3 ( king et al .
2012 ) , compared to the metal - poor population that is presumed to have near - primordial helium abundance ( @xmath4 ) . since then
, evidence for multiple populations in gcs has rapidly accumulated , mainly thanks to the exquisite photometric accuracy of the wide - field channel of the advanced camera for surveys ( acs / wfc ) .
ngc 2808 was found to have three main sequences ( piotto et al .
2007 ) , in spite of showing no dispersion in the abundance of iron - peak elements , again demanding that each ms have a different helium content , up to @xmath5 .
in addition , multiple sub - giant branches were found in several clusters ( milone et al .
2008 ; piotto 2009 ; piotto et al .
2012 ; bellini et al .
2013 ) , as well as a multiple main sequences ( milone et al .
2010 , 2012ab , 2015 ; bellini et al . 2013 : piotto et al .
2015 ) .
besides shaking the old paradigm of gcs as simple stellar populations , these findings raise three fundamental questions : 1 ) is the multiple - population phenomenon a general property of gcs , no matter in which galaxy they are hosted today ?
2 ) how were such multiple populations generated ?
3 ) what is the origin of the helium enrichment ? the answers to these questions have deep implications for our understanding of the formation of gcs , i.e. , of objects that are still playing a pivotal role in so many astrophysical areas and especially for the early star formation in the universe .
indeed , gc ages place their formation tantalizingly close to the epoch of re - ionization , to speculate they ( or their precursors ) may have had a part in it .
meanwhile , the discovery of multiple populations with large helium differences has shed new light on one long - standing mystery : the puzzling morphology of the horizontal branch ( hb ) and its poor correlation with cluster metallicity , an issue that , after van den bergh ( 1967 ) , is traditionally referred to as the _ second parameter _ problem .
indeed , the helium abundance has strong control over the range of effective temperatures covered by stars during their hb phase , with higher helium leading to higher effective temperatures for hb stars , hence brighter cluster uv luminosities .
thus , high - helium multiple main sequences are associated with multimodal distributions of stars on the hb , with prototypical examples being offered by @xmath2 cen and ngc 2808 .
multimodal hb distributions offered in fact the first hint for high - helium sub - populations in ngc 2808 ( dantona & caloi 2004 ) , which was then confirmed by the _ hst _ discovery of multiple main sequences in the same cluster ( piotto et al . 2007 ) .
most clusters with multiple populations are chemically homogeneous in iron and other heavy elements , which implies that helium enrichment and cno processing have proceeded without being contaminated by supernova products .
this excludes in most cases massive stars ( @xmath6 m@xmath7 ) from having been involved in the chemical enrichment of secondary populations , in either their fast rotating ( e.g. , krause et al .
2013 ) or binary ( de mink et al .
2009 ) versions ( e.g. , renzini 2008 , 2013 , for critical reviews ) .
this leaves intermediate - mass agb stars as possible sources of the helium - enriched and p - capture processed material out of which second generations would have been formed .
however , the agb scenario is also far from providing a detailed description of the sequence of events leading to gcs and their multiple populations as we see them today . particularly intriguing is the mass budget requirement in this scenario , as the first generation delivers only @xmath8 of its mass as agb ejecta , whereas secondary populations in gcs tend to be nearly as massive as or even more massive than the first generation . thus , for the first generation to produce enough agb ejecta ( to account for the mass of secondary generations ) it should have been at least @xmath9 times more massive than its fraction still bound to the cluster .
this is indeed a lower limit , as it assumes 100% efficiency in converting agb ejecta into second - generation stars .
thus , in this scenario precursors to present - day gcs should have been much more conspicuous objects than their surviving remnants . a mechanism that in principle could alleviate this mass - budget problem is accretion of agb ejecta on stars belonging to the same ( first ) generation , a process first proposed to account for the composition anomalies known at that time ( dantona et al .
1983 ; renzini 1983 ) , and now still entertained in the modern context ( bastian et al .
however , accretion will inevitably differ from star to star , hence may produce some spread in chemical composition , but can not generate discrete , _ quantized _ sub - populations ( renzini 2013 ) , such as e.g. , the seven or so seen in @xmath2 cen ( bellini et al . 2010 ) or the three seen in ngc 2808 ( piotto et al .
actually , recent wfc3/uvis data reveal the presence of at least five distinct populations in ngc 2808 , while m 2 exhibits at least seven of them ( milone et al .
this discreteness of multiple populations is an extremely powerful discriminant for the various scenarios proposed for the origin of multiple populations , excluding those which for their very nature are incapable of producing discrete , distinct populations .
for example , in the fast - rotating - massive - star option ( e.g. krause et al .
2013 and references therein ) stars of secondary populations would form in the extruding disk of massive stars , with each massive star delivering secondary stars with a range of abundances . globally , just a spread of compositions would result . moreover , membership into gcs of fast rotating massive stars would be irrelevant for the production of secondary stars , which then should be equally abundant in gcs as in the galactic halo field , contrary to observations ( renzini 2013 ) .
discreteness of secondary populations therefore demands discrete star formation events , such as a series of star - formation bursts out of an interstellar medium ( ism ) whose composition is slowly changing , interleaved by phases during which the mass of the ism grows thanks to the ejecta from the first generation . in any event
, the requirement of a 20100 times more massive precursor to present day gcs remains valid in the agb scenario , thus making gc formation a major event in the early stages of galaxy evolution .
such precursors may have been dwarf nucleated galaxies , @xmath10@xmath11 m@xmath7 by mass in which , past the supernova era , the agb ejecta stream to the bottom of the potential well in a sort of _ cooling flow _ to form the secondary populations , such as e.g. in the models of dercole et al .
( 2010 , 2011 ) . in these models
most of the stars of the parent dwarf are later stripped by tidal interactions , thus leaving the bare nucleus with its multiple populations .
one pressing question is whether this process of gc formation requires special environmental circumstances , which may be present in one ( proto)galaxy but not in others . in our own milky way ( mw ) , the multiple - population syndrome is virtually ubiquitous among gcs ( piotto et al .
2015 ) , but what about other galaxies ? and especially , what about very massive elliptical galaxies with their thousands of gcs ?
it is indeed quite important for our understanding of gc formation to establish whether the presence of multiple populations is a generic property of gcs , irrespective of the type and mass of the parent galaxy , or whether a special environment ( parent galaxy ) is required . to answer this question one will have to rely on integrated light observations , as
resolving stars in gcs is not currently feasible for distances beyond the local group .
the existence of helium - enriched sub - populations can be inferred from the presence of extremely hot hb stars ( ehb ) , which dominate the uv luminosity of clusters containing them in sufficient numbers .
typical in this respect are the metal - rich clusters ngc 6388 and ngc 6441 that belong to the galactic bulge , where the existence of ehb stars was first inferred from their blue uv@xmath1optical colors and then confirmed by _ hst _
photometry ( rich et al .
1997 ; busso et al .
2007 ) . in this way
, the integrated uv@xmath1optical colors of gcs were thought to offer a unique opportunity to infer the existence of helium - rich sub - populations in distant , unresolved gcs belonging to galaxies of various morphological types , which can not be resolved into individual stars . in this mood ,
the bluer the uv@xmath1optical color at a given @xmath12 color , the stronger the ehb component , and especially so for the reddest @xmath12 clusters .
a first such experiment was attempted by sohn et al .
( 2006 , hereafter s06 ) and kaviraj et al .
( 2007 , hereafter k07 ) for the giant elliptical m 87 , the cd galaxy of the virgo cluster .
these authors used stis fuv ( covering the 11501700 range ) and nuv ( 16003100 range ) data in combination with optical photometry to infer the likely existence of helium - rich sub - populations in a small group of gcs in m 87 .
the presence of a very strong ehb component in these clusters was inferred from their extremely blue uv@xmath1optical colors @xmath13 and @xmath14 for a given @xmath12 . in these studies m 87 gcs appear to be even more uv - strong than the bluest galactic gcs with helium - rich populations , thus suggesting that the multiple - population syndrome is spread also among gcs in this giant elliptical . the s06 and k07 studies have opened the tantalizing opportunity to diagnose the presence of ( likely ) helium - enriched extreme ehb stars in very distant globular clusters , and in particular in a giant elliptical galaxy . yet
, their sample includes only 66 m 87 clusters , due to the very small area covered by the stis observations ( just @xmath15 arcsec@xmath16 ) .
moreover , for the reference population , they had to rely on early integrated - light uv observations of galactic gcs , namely with the _ astronomical netherlands satellite _ ( van albada et al .
1981 ) , with different uv passbands and calibrations compared to the stis data used for the m 87 gcs . for these reasons
, we submitted a _
hst _ proposal ( go-12989 , pi : a. renzini ) to observe the central part of m 87 with the ultraviolet - visible channel of the wide - field camera 3 ( wfc3/uvis ) in the f275w band , exploring in a single shot an area over @xmath17 times larger than that covered by the stis observations and reaching substantially fainter objects .
moreover , for our reference population , we can also use wfc3/uvis ultraviolet data for both m 87 and the mw gc systems , thus avoiding passband differences and calibration issues when comparing m 87 and mw clusters .
the complementary archival optical data consists of very deep ( 50 orbits ) archival f606w and f814w acs / wfc imaging of the central region of m 87 ( go-10543 , pi : e. a. baltz ) .
thus , the plan of the project was to construct uv@xmath1optical color - magnitude diagrams and a @xmath18 and @xmath19 vs. @xmath20 two - color diagrams to characterize the uv properties of a large , representative sample of gcs in m 87 .
our aim was then to address the following questions : * is the presence of helium - enriched gc sub - populations as widespread ( or more so ) in a giant elliptical compared to the milky way ?
* has environment ( present parent galaxy type ) affected the gc formation process by favoring / disfavoring the formation of multiple stellar generations ?
* is the frequency of uv - bright gcs different among the blue / red component of the bimodal color ( metallicity ) distribution of gcs in a giant elliptical such as m 87 ?
* is there a connection between the uv - bright populations in gcs and the uv - upturn in elliptical galaxies ?
however , answering these questions proved to be more difficult than we expected . indeed , the _ hst _ f275w data acquired in the meantime for a large sample of mw gcs ( piotto et al .
2015 ) show that the above uv@xmath1optical two - color plots are not able to provide unambiguous evidence for the presence of an ehb component .
still , the uv data on the m87 and mw gcs are used to highlight possible systematic differences between the two cluster families .
uv@xmath1optical color gradients within the brightest m 87 clusters are also used to tentatively find evidence for the presence in them of helium - enriched ehb stars .
the paper is organized as follows .
sections 2 and 3 describe the _ hst _ data reduction and processing .
the photometric measurement of gcs in m 87 is presented in section 4 while section 5 deals with the completeness of the sample .
section 6 presents the color - magnitude and two - color plots for the m 87 globulars , whose spatial distribution is discussed in section 7 .
section 8 offers an extensive comparison with the milky way gcs and section 9 investigates the possibilities of the detection of internal color gradients within individual gcs .
finally , in section 10 we present and discuss our interpretation of the data and our conclusions .
as mentioned earlier , this work is based on the wfc3/uvis data from go-12989 and on the acs / wfc data from go-10543 providing ` _ flt`-type images of the core of m 87 .
these data sets consist of @xmath21s exposures taken through the wfc3/uvis f275w filter and @xmath22s acs / wfc f606w exposures plus @xmath23s acs / wfc f814w exposures .
all ` _ flt ` images were corrected for charge - transfer efficiency defects ( e.g. , anderson & bedin 2010 ) using either the standard _
hst _ calibration pipeline ( for acs ) , of the official ` fortran ` routine developed for the wfc3 .
the initial catalog of sources in the acs / wfc exposures was obtained using the publicly - available ` fortran ` program ` img2xym_wfc.09x10 ` , which is described in detail in anderson & king ( 2006 ) .
the program employs empirical `` library '' effective point - spread functions ( epsfs ) that vary spatially across the detector .
source positions and fluxes on wfc3/uvis images were measured with the software program ` img2xym_wfc3uv ` , derived from ` img2xym_wfc.09x10 ` , which also employs library , spatially - varying epsfs . both acs and wfc3 software
identify a source if its central pixel is brighter than the surrounding 8 pixel , regardless of the actual shape of the source ( point - like , diffuse , etc . ) .
source positions in each single - exposure catalog were then corrected for geometric distortion using the state - of - the - art solutions provided by anderson & king ( 2006 ) for acs / wfc and by bellini & bedin ( 2009 ) and bellini , anderson & bedin ( 2011 ) for wfc3/uvis . even though m 87 gcs are slightly resolved by _ hst _ and therefore a psf is generally not an optimal representation of their light profiles epsf - derived positions for bright gcs are still found to be accurate to @xmath24 pixel .
our final photometry was obtained from high - quality image stacks .
photometry on the stacks will allow us to obtain a more reliable measure for fainter sources , and to push down the detection limit by @xmath251.6 , 2.2 and 3 mag for f275w , f606w and f814w , respectively .
the preliminary single - exposure catalogs have been used only to setup a common distortion - free reference frame and to define the linear - transformation parameters that map each pixel of each exposure onto the image stacks .
we based our reference frame on the f275w exposures , and we supersampled its pixel size by a factor of 2 .
this means that each pixel on the stacks corresponds to 20 mas . in order to accommodate all meaningful pixels on the stacks
, we shifted the rough center of m 87 ( as defined by the telescope pointing ) to location ( 5000 , 5000 ) on the master frame .
we identified common sources in each single exposure and used their distortion - corrected positions to compute general 6-parameter linear transformations to map each pixel of each exposure on to the master frame .
although the f275w exposures are each over 22 minutes long , only a few dozen sources are clearly detected in each of them . moreover , due to the length of these exposures , each f275w image is deeply contaminated by thousands of cosmic - rays , many of which got included in the single - exposure catalogs as real objects .
( cosmic - ray events are far less abundant in f606w or f814w images , given their shorter exposure times . )
special care was dedicated to identifying common sources between the catalogs to obtain the best linear - transformation coefficients using only real sources . to this aim , we made use of the quality - of - fit parameter ( anderson et al .
2008 ) in order properly select the best - measured objects for the linear transformations . at this point , all the necessary pieces of information required by the stacking algorithms were collected , and we proceeded with the construction of the image stacks themselves .
our wfc3/uvis exposures in f275w were properly dithered with an image - stacking strategy in mind .
unfortunately , dithering in the archival data was minimal .
this means that bad columns , hot / cold pixels , and other detector defects are often present at the same place in the field in multiple exposures and can not simply be `` sigma - clipped '' away .
therefore , before we could create a stack that was representative of the astronomical scene , we had to first identify the warm / hot pixels .
unfortunately , the dq flags did not identify all the warm pixels , so we had to first stack the dithers independently , then compared each exposure against the stacks made from the other dithers in order to identify the inconsistent pixels .
we compared each exposure against the stack made from other dithers and flagged each time a particular pixel was inconsistent at the @xmath26 level with the prediction from the comparison stack .
we then flagged as bad any pixel that was flagged in more than half of the exposures .
this allowed us to identify the bad pixels and construct stacks in f606w and f814w that were free of artifacts .
since the goal of this program was to perform aperture photometry on m 87 s globular clusters , we constructed the stacks in such a way as to carefully preserve flux .
we mapped the four corners of each pixel in each exposure into the reference frame and determined the geometric overlap between each input pixel and each output pixel ( in the stacked frame ) .
these weights were used to construct the output image .
bad pixels were naturally given zero weight .
this procedure results in a small amount of blurring , but this should have a negligible effect on our photometry given the @xmath27 oversampling used and the relatively large apertures we used .
since each exposure can have a slightly different sky level , on account of variations in scattered light , we iterated the stacking procedure .
after each iteration , we resampled the stacked image into the frame of each individual exposure .
this allowed us to determine an average sky offset between each exposure and the average of all of them .
we then adjusted the background level of each exposure by this difference .
this converged after two iterations and the procedure was used to generate a stack for the f275w , f606w and f814w exposures .
the cd galaxy in our field introduces a considerable gradient in our stacks .
since our focus is on the clusters , we decided to iteratively remove this gradient to make background subtraction trivial .
we parametrized the background by its value at an array of @xmath28 elements across the @xmath29-pixel frame .
in other words , we had one array value every @xmath30 pixels .
we iteratively solved for the array that reduced the background to zero everywhere , using a zeno factor of 0.5 and 16 iterations .
the background was smooth enough to use linear interpolation everywhere .
figure [ f : gas ] shows the central region of m 87 on our galaxy - light - subtracted trichromatic stack .
the vast majority of near point - like sources in the figure are m 87 gcs and the color scale has been stretched to emphasize the color difference between blue and red gc candidates .
the granularity of the background in close proximity of the galactic center is mostly caused by background noise , but far from the center the noise is low enough to clearly reveal surface - brightness variations in the light profile of m 87 .
the famous jet is prominent in this figure and various counterjet features are also apparent .
the study of the jet and related features goes beyond the scope of the present paper , but we briefly discuss them in the appendix .
as already mentioned in section [ sec : dataset ] , m 87 gcs are partially resolved , making psf - fitting techniques not appropriate for high - precision photometry .
therefore , we applied standard aperture photometry to measure the flux of the m 87 gcs .
the finding process was performed on the f814w stack only : this stack is made by the largest number of single exposures , and it should therefore be the one with the highest signal - to - noise ratio .
we identified all pixels in each exposure that were local maxima ( larger than their 8 surrounding neighbors ) and were more than 2.5 times the rms of the background noise .
we ignored areas of the field where there was fewer than 3 exposures depth .
a total of 2155 qualifying peaks were found this way .
a visual inspection of their locations on the f814w stack allowed us to remove spurious objects ( mostly obvious galaxies and stars ) from the list , leaving 1913 gc candidates .
aperture photometry was then performed at these locations on the three image stacks , using a circular aperture with a 7-pixel radius ( @xmath31 ) .
the local sky was estimated between 15 and 20 pixels .
the choice of using a 7-pixel aperture is motivated by the fact that it encircles the most - significant pixels of typical m 87 gcs ( being slightly - resolved sources which typical fwhm of @xmath325 pixels ) .
a larger aperture would have made our photometry too sensitive to random background fluctuations , especially for the faint sources in f275w .
for the brightest gcs in m 87 , a larger aperture is more appropriate .
a 15-pixel aperture for these objects is applied and discussed in section [ ss : mass ] .
all of the 1913 objects have positive measured fluxes in f814w and f606w , and 1460 of them also had positive fluxes in the f275w stack , making our uv photometry by far the largest available in the literature for extra - galactic gcs .
finally , the photometry in the three bands was calibrated into the vega - mag flight system following sirianni et al .
( 2005 ) , and using the f275w zero point and aperture correction provided by the stsci webpage . a visual inspection of the f275w image stack revealed the presence of a handful of relatively bright uv objects that were not identified in the f814w stack .
these objects are likely transient sources in m 87 , such as novae , or uv - bright background galaxies .
since our aim is to study m 87 gcs , we will not consider them further in this paper .
in order to infer the completeness level of our source list , as well as to estimate photometric errors , we ran extensive artificial gc tests . for the simulated gcs we used 2d gaussians with a full width at half maximum of 4.5 pixels , which is the average value found for m 87 gcs . while it is true that our stacks are made in such a way to preserve both source positions and fluxes , the actual shape of the gcs in the stacks is the average of several , uncorrelated psfs ( due to different telescope - breathing conditions for each exposure ) .
thus , a 2d gaussian turned out to be a fair representation of these average psfs while accounting also for gcs of m 87 not being point - like sources . instead of the usual practice of simulating gcs with magnitudes and colors following the cluster fiducial sequences on the cmds , we populated the full color - magnitude diagram ( cmd ) space occupied by the real gcs with nearly 17 million sources , and assigned to them a random position on our stacks .
note that m 87 gcs span over 6 magnitudes in the @xmath33 color , hence this approach allows us to better estimate incompleteness and photometric errors across such a wide color range . as for the real gcs
, the finding process was performed only on the f814w stack .
we added one simulated gc at a time , and determined whether its central pixel qualifies as a peak .
if so , we measured its flux using a 7-pixel - radius aperture photometry in all three stacks . before moving to the measurement of the next artificial object , we removed the current simulated gc from the stacks . this way , simulated gcs never interfere with each other , but only with the real sources of the stacks , as it should be .
we consider a simulated gc to be recovered if its measured position in the f814w stack was within 0.75 pixel of the input position , and if its measured photometry was within 0.75 mag of its input photometry in each of the three bands .
there was no need to impose a position constraint on the f275w and the f606w recovered magnitudes since , exactly as we have done for the real gcs , their positions are defined by the measured centroids on the f814w stack .
finally , the completeness level was simply obtained as the ratio between the number of recovered gcs and the number of input gcs . + to estimate the photometric errors , we computed the 68.27@xmath34 percentile ( rms ) of the absolute values of the difference between input and output magnitudes .
these rms values represent our estimate of the photometric uncertainties .
the results of the artificial gc tests are summarized in figures [ f : complecmd ] and [ f : complexy ] .
the cmds of real sources ( black squares ) are shown figure [ f : complecmd ] , one for each color - filter combination at our disposal . in each cmd
we divided the simulated gcs in thousands of small regions ( each containing hundreds of input sources ) , and we determined the completeness level for each of them .
each region is then color - coded according to its measured completeness .
the color ranges from red ( 0% ) to yellow ( 100% ) .
a color bar on top of the first panel illustrates the adopted color - coding , which is kept the same in all panels .
nine contour lines indicate the loci of average 10% to 90% completeness level , with the 50% contour line being heavier than the others , for clarity .
the contour lines in the optical cmd are almost horizontal , meaning that , within the investigated color range , the completeness level of sources measured in the f814w and f606w stacks is similar , with the f606w stack actually going a little deeper than that of f814w one .
this is not the case for the uv cmds , where the completeness level of the sources measured in f275w drops appreciably towards the red part of the cmds .
the cmd in which the incompleteness effects are most evident is the @xmath35 one , where sources have to be found in all the three bands .
the typical size of the photometric errors are also shown , in light blue and for sources of average color , in each cmd .
larger photometric errors are found , as expected , for f275w measurements . we will discuss these cmds in detail in section [ sec : cmd ] .
finally , figure [ f : complexy ] shows the spatial f814w completeness map indicating which parts of m 87 suffer from higher incompleteness levels regardless of the magnitude level .
we adopted the same color - coding as in figure [ f : complecmd ] .
as expected , the completeness drops towards the center due to the higher background noise .
the completeness level is also low around the jet , the location of which can be glimpsed even in this completeness map .
the outer edges of this panel are not mapped by the minimum number of 3 exposures , and no artificial gcs were added there . a red color code ( 0% completeness )
is applied to these marginal regions as well .
figure [ f : redblue ] shows the cmds of m 87 s gcs using the three available color combinations ( panels a , b and c ) , as well as the combined two - color diagram ( panel d ) .
the well - known color bimodality of m 87 gcs ( e.g. , gebhardt & kissler - patig 1999 ) is clearly visible in all the cmds .
while the two gc sequences have a similar spread in @xmath36 , the blue sequence is significantly narrower and more defined than the red one in @xmath37 and @xmath19 .
this effect is related to the increased photometric errors in f275w for the red gcs : the color separation between the two gc sequences in the uv cmds is @xmath38 magnitudes , meaning that , at a given @xmath39 luminosity , red clusters are about 6 to 7 times fainter in f275w on average than their blue counterparts .
six detected objects fall outside the red limit of panel ( a ) , being redder than @xmath40 ( color - coded in red in the other panels of figure [ f : redblue ] ) .
they also occupy a well - defined region in the two - color diagram of panel ( d ) .
a deeper inspection of their images on the stacks suggests they are compact passive galaxies at high redshifts and we will not consider them any further in this paper . a quick look at panels ( b ) and ( c ) of figure [ f : redblue ] also reveals the presence of a group of 25 extremely - blue objects , and we color - coded them in blue in all the panels of the figure .
it is worth noting that these very blue objects seem randomly placed in the optical cmd , and only the uv photometry allows us to single them out .
these very blue objects occupy a well defined , isolated region in the two - color diagram shown in panel ( d ) .
most of these blue objects are faint and appear quite compact in the trichromatic stack : we consider them to be small star - forming galaxies at moderately high redshift ( @xmath41 ) , as they are not f275w _
dropouts_. hereafter , we will remove also these 25 objects from our analysis and therefore we are now left with 1882 likely m 87 gcs , with 1429 of them having also the f275w photometry .
let us now turn our attention on the two - color diagram ; a zoomed - in view of it can be seen in panel ( a ) of figure [ f:4pops ] .
the inset shows a hess diagram in which two main clumps are clearly visible .
there is also a hint for a third , less prominent clump in between the main two ones , located approximately at ( 0.85 , 2.5 ) .
the three clumps are marked with arrows .
panel ( a ) is replicated in panel ( b ) where we decided to arbitrarily split the sample into four groups according to their location in this plot .
the _ blue _
gc component ( 508 objects , hereafter bgc ) occupies a well - defined and compact region in the top - left corner of the panel , while the _ red _ component ( 731 objects , hereafter rgc ) is more spread and stretches across increasing @xmath18 values while keeping a roughly constant optical color .
the _ intermediate _ objects ( 117 by number , hereafter igc ) that seems to stand out in the hess diagram are colored in yellow and appear to be offset in @xmath42 color with respect to the red component ( indeed , they seem to follow the blue - component trend rather than the red - component trend ) , and are separated by a less - populated part of the hess diagram at @xmath43 from the blue component itself . finally , a small number of objects , colored in green , occupy the top - right region of the panel ( 73 objects , hereafter fgc ) .
we stress that the definition of the intermediate gc component is arbitrary , and only marginally supported by the relative clump in the two - color diagram .
still , we will keep this four - group classification through the paper , and see whether the analyses presented in the next sections are able to support ( or disprove ) the presence of igcs as a truly distinct group of clusters in m 87 . turning now our attention to the lower three panels ,
we can notice that while bgcs , igcs and rgcs are always to the left , middle and right side of these cmds , respectively , fgcs are redder on average than rgcs in the @xmath36 cmd , while they become as blue as the bgcs in @xmath18 , and of somewhat intermediate color between bgcs and igcs ( but definitely bluer that rgcs ) in @xmath19 .
hence they _ flip _ in color .
if the fgcs had been to be among the brightest gcs , we would have immediately found the best candidates for the helium - enhanced , multiple - population gcs , as the most prominent ehb extensions are found among the most massive gcs in the milky way . alas , these fgcs are too faint to be the m 87 equivalent of , e.g. , @xmath2 cen or ngc 2808 ( more in section [ sec : mw ] )
. we looked at each of them on the trichromatic stack , and their shapes appear to be essentially indistinguishable from those of the surrounding gcs of the same optical brightness level .
their average fwhm is also the same as that of the other clusters .
moreover , they are also rather isolated , so that light contamination from brighter neighbors is excluded .
because fgcs are faint in all bands , the reader can argue that we are just trying to over - interpret what are just simply photometric - error effects .
while it is true that the larger photometric error of fgcs can in principle explain their larger spread in the cmds , they are never really randomly scattered among the other groups , but are always clumped in specific regions in each cmd . nevertheless , most of these objects are so dim in @xmath44 that , at the faint limit , their total flux amounts to only 10 electrons above the sky background in the f275w stack ( sky noise of @xmath45 electrons ) . the fgcs could be just faint rgcs with large errors in the f275w flux .
an intriguing possibility is that the fgcs are diffuse star clusters , such as those discovered by larsen & brodie ( 2000 ) around two lenticular galaxies , and already found in the virgo cluster ( peng et al .
however , having measured the size of all clusters with @xmath46 we found no difference between fgcs and the other clusters .
another , perhaps more likely possibility , is that they are not gcs at all , but compact emission - line galaxies at moderate redshifts .
nothing really conclusive can be said about fgcs until deeper f275w photometry becomes available .
the top panels of figure [ f : hist ] show the histograms of the color distribution for the three cmds of figure [ f:4pops ] .
we applied a smoothed nave estimator ( silverman 1986 ) for the results to be insensitive to a particular binning starting point .
two peaks are clearly distinguishable in panel ( a1 ) referring to the @xmath47 color distribution , with similar width .
the poisson errors are highlighted by shaded regions , and give an assessment of the quality of the peak detections . in panels ( b1 ) and
( c1 ) for the uv@xmath1optical colors the blue peak is very sharp , while red gcs are more spread in color and apparently show more than one peak .
we divided the @xmath36 histogram in two halves as shown in panel ( a2 ) , color - coded in blue and red .
this represents the classical bimodal subdivision .
we identified the gcs belonging to each component and color - coded accordingly their histograms in panels ( b2 ) and ( c2 ) .
the blue histograms in the middle and right panels both show a secondary peak at redder colors .
the significance of this second peak is marginal , yet it is present in both the @xmath48 and @xmath19 colors , which suggests that it may be a real feature .
the lower panels of the figure adopt the four - group subdivision and the igc group becomes responsible for the intermediate peak . .
see the text for details . ]
it has long been known that the red gcs of m 87 are more centrally concentrated than the blue ones ( see , e.g. , strader et al .
2011 ) , but previous studies have mostly focused on the clusters 1d ( radial ) surface - density profiles . in this section
we investigate the 2d spatial distribution of m 87 gcs , looking for possible ( a)symmetries in the density maps .
we analyzed the spatial distribution of gcs in two ways : first we simply divided gcs into the two classical red and blue groups , as in panel ( a2 ) of figure [ f : hist ] ; then , we adopted the 4-group subdvision as defined in panel ( b ) of figure [ f:4pops ] . in both cases , we focus only on gcs above the 50% completeness limit in the @xmath49 vs. @xmath36 cmd , which corresponds to @xmath50 mag .
by exploring the spatial distribution of gcs using both the classical and the 4-group division , we might learn something more about the nature of igcs and fgcs , e.g. : are igcs and fgcs two new gc groups , or are they just subsamples of the classical blue / red components ?
let us start from the results of the classical subdivision .
since blue / red gcs are defined on the @xmath36 without making use of the uv information , to improve the statistical significance , we included in the analysis all gcs regardless of whether they also have a @xmath44 measurement .
this way , the number of gc above the 50% completeness cut is 1698 .
panel ( a1 ) of figure [ f:2dmaps ] shows the optical cmd for these gcs .
the horizontal line marks the 50% completeness cut .
rejected gcs are plotted in grey .
panel ( b1 ) of the figure shows the blue / red gc subdivision , which is made in the same way as in panel ( a2 ) of figure [ f : hist ] ( the only difference being the 50% completeness cut ) . to derive the gc density maps we proceed as follows . for each gc in the two samples we counted the number of surrounding gcs belonging to the same sample within 1800 pixels ( @xmath51 )
this radius was chosen by trial and error as a compromise between the need of a fine spatial - distribution mapping and an adequate poisson statistics .
larger radii would tend to produce very smoothed maps , washing out most of the intrinsic spatial substructures . on the other hand , smaller
radii tend to produce noisy maps , dominated by small - number statistics .
panels ( c1 ) and ( d1 ) of figure [ f:2dmaps ] show the resulting 2d density maps , for the blue and the red component , respectively .
each point in these panels represents the location of selected gcs with respect to the center of m 87 , and it is color - coded according to the number of surrounding clusters .
colors range from purple ( low - density ) to green ( average ) to red ( high - density ) .
the center of m 87 is marked by a black cross and was identified as the brightest central pixel in the f275w stack ( being the center of m 87 heavily saturated both in f606w and f814w stacks ) .
the direction of the jet is also shown by a black arrow . while the rgc density map appears to have a gaussian - like shape , roughly centered on m 87 , the density map of bgcs appears to be more asymmetric , and
it is flattened in the direction perpendicular to the jet .
so , there is a hint for bgcs being preferentially located about a _ galactic plane _ roughly perpendicular to the jet .
the 1d radial profiles of both samples are also shown , in panel ( e1 ) , for completeness .
for the second test we use additional information from f275w photometry ( in order to divide gcs into four groups ) , thus lowering the total number of available gcs to 1334 ( those above the 50% completeness cut ) .
the optical cmd of selected gcs and their color - distribution are shown in panels ( a2 ) and ( b2 ) of figure [ f:2dmaps ] .
unfortunately , the small number of igcs and fgcs ( 107 and 49 objects , respectively ) does not allow us to derive reliable 2d density maps , and we can only compute 1d radial distributions for them .
as was done for the previous test , for each bgc and rgc we counted the number of surrounding gcs belonging to the same population within 1800 pixels .
the resulting density maps are shown in panels ( c2 ) and ( d2 ) for bgcs and rgcs , respectively .
the available field - of - view ( fov ) is now smaller than that of the first test because of the smaller fov of our wfc3/uvis images with respect to those of the acs / wfc . for both blue and red clusters , in the centermost regions the new contours are more flattened than in the case of the previous test ( and the effect
is more evident for the blue gcs ) , whereas the gc distribution in the outer regions does not show a significant flattening .
this may suggest that igcs ( that were removed in this test ) may have a more symmetric distribution than blue and red gcs .
it is now more evident that in the centermost regions bgcs ( and , to a lesser extend also rgcs ) are flattened in a direction nearly parallel to the jet . in order to better quantify this behavior , we least - squares fitted with ellipses the contours of panels ( c2 ) and ( d2 ) to estimate ellipticity and orientation of each contour . the angle @xmath52 between the jet and the minor axis of the ellipses , and their ellipticities , are reported in table 1 .
the outer contours suffer from edge effects caused by the square fov and by the way we constructed the density maps . as a consequence , ellipticity and orientation for the two outermost contours ( @xmath53 )
are not reliable and in table 1 are reported in parenthesis .
the distribution of both red and blue gcs is more circular at larger radii , and becomes steadily flatter towards the center .
the central flattening of blue gcs is about 50% larger than that of red gcs .
the orientation of the fitted ellipses for the red gcs is approximately parallel ( @xmath54 ) to the jet within @xmath55 . on the other hand , the orientation of the ellipses for the blue gcs changes from being parallel to the jet to being almost perpendicular to it when moving outward .
the agreement among independent contours , i.e. any contours separated by more than the region over which the cluster - density was assessed , provides a sense of the reality of the differences between blue and red gc subsystems . comparing ellipticity and position angle of blue and red clusters with those of m 87 isophotes ( from liu et al .
2005 ) we find that the ellipticities of the gc distributions follow those of the isophotes in the outer regions of our survey ( @xmath56 ) , but then the gc distributions become significantly flattened toward the center whereas the behavior of the isophotes is opposite , becoming even more circular within the central @xmath57 .
the orientation of flattening of red and blue gcs is broadly consistent with that of the outer isophotes of m 87 .
the radial profile of all four groups of gcs is shown in panel ( e2 ) . the igc - supbopulation profile ( in yellow in panel e2 )
is remarkably similar to that of the bgcs , suggesting that the igcs might actually belong to the bgc component , instead of being a separate group .
a special reference needs to be made for the fgc group .
they not only have a unique behavior in the cmds , but they also appear to be the most concentrated group of gcs .
it is hard to distinguish whether this indicates their intrinsically different nature or whether they have larger photometric errors ( especially in the f275w band ) as on average they project over a brighter galaxy background .
this flattening near the center of the gc distribution is quite intriguing indeed , and perhaps even more so that the distribution of blue gcs is more flattened than that of the red ones . at first
, one would have expected the contrary , with the older , metal - poor component to be dynamically hotter because of the past merging history of m 87 .
however , we can not exclude that the difference in flattening , although appreciable , could also arise from a statistical fluctuation given the relatively small number of objects involved in the flattening ( only 115 bgcs and 200 rgcs in the central @xmath58 ) and the necessary smoothing applied to construct the density maps .
given the flattening in the direction of the jet ( hence presumably lying in the equatorial plane of the central supermassive black hole ) this flattening may represent a vestige of an earlier disk morphology of m 87 . however , the galaxy - core dynamics is typical of massive , non - rotating galaxies which have been built up by merging ( see , e.g. , figure a1 in emsellem et al . 2011 ) .
alternatively , rather than a residual characteristic of the formation of these gcs , the flattening may have been acquired in the course of the dynamical evolution of the gc system .
for example , one possibility is that clusters in more radial orbits may have been destroyed by tidal interactions with the supermassive black hole .
we are not in the position to discriminate between these options and we leave this as an interesting open issue .
a direct comparison with mw gcs , where the multi - population phenomenon is now well documented , is mandatory if we are searching for signs of multiple stellar populations in the gcs of m 87 .
there are two _ hst _ treasury programs that made use of the very same detectors and filters used in this work : go-10775 ( pi : a. sarajedini ) , which observed 65 mw gcs through acs / wfc f606w and f814w filters ( sarajedini et al .
2007 ) , and go-13297 ( pi : g. piotto , piotto et al . 2015 ) , nearly finished and targeting the cores of 46 out of the go-10775 65 gcs in three wfc3/uvis filters , including f275w .
additionally , images in f275w for other 12 mw gcs have been acquired with two pilot programs to go-13297 : go-11233 , go-12311 and go-12605 ( both pi : g. piotto ) . moreover , the core of @xmath2 cen was chosen as wfc3/uvis calibration field , and have f275w exposures . at the completion of go-13297 the total number of clusters for which wfc3/uvis f275w and acs / wfc f606w and f814w data exist will be 59 : about a third of the total number of gcs in the milky way . in order to directly compare mw gcs with those in m 87 we need to : ( 1 ) estimate the total light of mw gcs in the three bands ; ( 2 ) apply reddening corrections in order to have the same reddening of m 87 ( @xmath59 , obtained by multiplying by 1.35 the quoted value @xmath60 of harris 2009 ) ; and ( 3 ) bring all mw gcs to the same distance of m 87 ( distance modulus @xmath61 , again from harris 2009 ) . to estimate the integrated total light of the mw gcs we proceeded as follows .
first of all , we note that all go-10775 catalogs are reasonably complete ( 50% ) down to about 5.5 magnitudes below the turnoff , and have comprehensive artificial - star tests that allow proper incompleteness estimates .
a direct sum of the flux of the cluster members present in these catalogs , within a certain radius , is therefore a good estimate of the total cluster flux within that radius .
while we will inevitably lose a sizable fraction of stars fainter than , say the @xmath62 completeness limit ( especially near the cluster center ) .
the dim light of these missed faint sources will only marginally affect the estimated total light of the clusters and will have an even smaller impact on the inferred integrated colors .
in fact , using artificial - star tests , we estimated the amount of light lost from missed stars in f606w and f814w down to the 50% completeness level and found it to be on the order of 4% in both filters and in nearly all go-10775 clusters .
( this value necessarily varies spatially , and the light lost in the core due to crowding effects is higher than in the outer regions . on average , we found a relative 2% increase in lost flux within the centermost @xmath63 with respect to the @xmath64 to @xmath65 radial interval . )
a 4% difference in total flux corresponds to about 0.04 magnitudes . the average @xmath66 color variation due to incompleteness is then found to be of the order of only 0.003 magnitudes .
we also computed the total light of 10 randomly - selected clusters as provided by stars down to 2 , 4 and 6 magnitudes below the turnoff . on average , the total clusters light decreased by 0.029 , 0.005 and 0.0004 magnitudes for the 2 , 4 , and 6 mag cuts in f606w , and by 0.034 , 0.007 and 0.0007 , respectively , in f814w .
the difference in the @xmath67 color is then 0.004 , 0.002 and 0.0003 for the three magnitude cuts , and therefore negligible . given this very small impact of incompleteness on the total clusters light and optical colors , we decided to ignore completeness effects in our calculations . to obtain stellar fluxes in f275w
we used the catalogs of go-10775 as an input , and fitted f275w library psfs to each listed stellar position , so that every source in the go-10775 catalogs will also have a f275w measurement if detected in this band ( soto et al . in preparation ) .
we do not yet have artificial - star tests for the go-13297 data .
nevertheless , in f275w we can easily detect stars down to 34 magnitudes below the turnoff , i.e. 10 magnitudes below the hb ( in f275w ) .
this typically translates into about 2.5 magnitudes below the turnoff in f606w , and we already saw that below this cut we lose only a marginal fraction of light in the f606w passband .
the missed stars are faint , red ms stars that already have little / no contribution to the clusters light in optical , and even less in f275w . unfortunately , this is not the case for the two interesting clusters ngc 6388 and ngc 6441 , the most metal rich clusters with evidence of helium - rich ehb stars ( e.g. , bellini et al .
these heavily reddened bulge clusters would have required a disproportionate number of _ hst _ orbits to reach an f275w depth below the turnoff similar to that of the other clusters .
actually , the data barely reach the turnoff , which is a significant contributor to the uv light especially in metal - rich clusters , see figure 10 .
therefore , these clusters are not included in our present analysis . for the calculation of the total clusters light we selected only bona - fide cluster members , on the basis of their proper motions ( obtained as the difference in stellar position between the go-10775 and go-13297 catalogs ) and on their location on the color - magnitude diagram .
these two selection criteria work very well for all clusters except for ngc 6715 . in this case
, it is impossible to separate cluster members from the core of the sagittarius dwarf spheroidal galaxy using proper motions , but at least the cmd selections allowed us to remove sagittarius rgb contaminants .
needless to say , the fov of go-13297 catalogs ( @xmath68 @xmath69 ) is smaller than the tidal radius of mw gcs . to convert our fov - limited fluxes into the total cluster light ,
we made use of the fraction - of - light radii listed in trager , king & djorgovski ( 1995 ) . in their table 2 , the authors report the value of @xmath70 , @xmath71 , @xmath72 , @xmath73 , @xmath74 , which are the radii within which the 10% to 50% of the total cluster light is enclosed . for each mw gc in our sample
, we summed the flux of all the stars within the largest value @xmath75 ( where @xmath76 is one of the listed fraction - of - light radii ) that was still fully enclosed in the go-13297 fov .
these @xmath75-based fluxes were then accordingly rescaled to synthesize the total clusters light in each of the three passbands .
note that we applied the same scaling factor for uv and optical magnitudes .
if uv light and optical light have different radial gradients , this might introduce a bias .
some of our mw clusters are not listed in trager , king & djorgovski ( 1995 ) ; we could compute integrated clusters lights for only 45 objects .
we adopted the reddening values listed in harris ( 1996 , 2010 edition ) , in order to adjust our integrated magnitudes to the reddening of m 87 , by using the extinction coefficients of sirianni et al .
( 2005 ) for acs / wfc . to date
, no official extinction coefficient is available for the f275w filter on the _ hst _ website .
we adopted the value a@xmath77 , obtained with yes , the york extinction solver , using the reddening laws by cardelli , clayton & mathis ( 1989 ) . finally , we corrected the integrated and ( de)reddened magnitudes for the difference between the mw gcs distance moduli ( from harris 1996 , 2010 edition ) and that of m 87 ( @xmath61 , harris 2009 ) .
the integrated , absolute and de - reddened photometry for mw gcs is listed in table 2 .
figure [ f : mw_piotto ] shows the three cmds ( panels a , b and c ) and the two - color diagram ( panel d ) of m 87 ( in black ) and the 45 mw gcs ( in red ) , for which we already have the data in all three passbands .
outstanding mw gcs in the figure are tagged .
note that none of the 45 mw gcs fall on the bulk of the red sequence of m 87 in @xmath66 .
the metal - rich gc 47 tuc ( [ fe / h][email protected] , from harris 1996 , 2010 edition ) lies just in the middle between blue and red m 87 clusters in the optical color , and just makes it on the red sequence when the color is based on the f275w filter .
moreover , we note that @xmath2 cen itself would be among the brightest gcs in m 87 .
indeed , only a handful of m 87 gcs in our survey field are brighter than @xmath2 cen , but not by much more than 1 magnitude in either @xmath78 or @xmath49 .
it appears that the most luminous gcs in m 87 are likely siblings of g 1 ( mayall ii ) in m 31 , with a mass of @xmath79 m@xmath7 ( ma et al .
the integrated color of ngc 6715 ( m 54 ) is somewhat more uncertain , as this cluster is deeply embedded within the nucleus of the sagittarius dwarf spheroidal galaxy , in such a way that there are a comparable number of cluster and galaxy stars ( that are redder , on average ) even in the centermost few arcmin from the cluster center . from panels ( b ) and ( c ) in figure [ f : mw_piotto ]
we note that the bulk of mw and m 87 blue clusters have approximately the same @xmath80 color , but in the two - color diagram shown in panel ( d ) there is a systematic offset in @xmath67 of @xmath81 mag , also appreciable in panel ( a ) .
several sources of error affect our mw gc photometry : errors in the photometric zero - pointing , in the total flux estimate , in the m 87 distance modulus , in the adopted reddening values , and in our estimate of the total clusters light .
we can exclude faint star incompleteness of our mw gc database as the cause of the shift , as this 0.06 mag difference is much bigger than what would possibly be expected from such an effect , as estimated above .
another possibility is that the @xmath82 color of mw gcs may have been underestimated given that the brightest rgb stars ( above the hb level ) tend to be saturated in the go-10775 images .
the photometry of these rgb stars was corrected for saturation effects , but certainly the result can not be as accurate as in the case of unsaturated stars . whether this is the reason for the 0.06 mag difference could in principle be ascertained , but it requires a non trivial effort and we wish to postpone it to a future paper .
some of these sources of systematic errors affect our three photometric measurements in the same way , but others may indeed generate a systematic color difference between mw and m 87 clusters . at this stage , we can not determine whether this difference in color of @xmath81 mag is an intrinsic property of mw and m 87 gcs or it is due to systematic errors . in analogy to the result of s06 and k07 , we find that m 87 gcs appear systematically bluer than mw gcs in the color - color plot , at fixed optical color .
however , the blue gcs in both galaxies span the same range of colors in @xmath19 , hence we attribute the offset to the @xmath36 color , i.e. , m 87 gcs do not appear to be _ more ultraviolet _ than mw gcs .
our data do not allow us to exclude this offset being due to systematic / calibration errors .
we note that this color shift was more prominent in s06 and k07 fuv@xmath1optical color rather than in the nuv@xmath1optical color .
the stis nuv passband used by s06 and k07 partly overlaps with the f275w passband , but the fuv passband is even bluer than the bluest wfc3/uvis filter : f225w .
such a fuv passband would indeed be more sensitive to the bluest hb stars .
in section 10 we shall return on the comparison with s06 and k07 results .
the fact that the vast majority of the mw gcs in our sample occupy the blue half of the m 87 gcs in all the three cmds should not come as a surprise .
the go-10775 treasury program ( and therefore also the go-13297 one ) did not include heavily - reddened clusters , which are generally harbored in the galactic bulge .
these clusters are metal rich and also intrinsically red . in the milky way bulge
there are some extremely - metal - rich clusters that may lie well within the rgc sequence of m 87 . to verify this , we additionally estimated the integrated light of ngc 6528 ( [ fe / h]=@xmath83 , @xmath84 , from harris 1996 ) , an almost solar - metallicity gc in the bulge , following the very same procedures and selection criteria used for the other clusters .
ngc 6528 was observed through acs / wfc f606w and f814w as part of the go-9453 program ( pi : t. m. brown , see brown et al .
this cluster ( identified and tagged in red in figure [ f : mw_ata ] ) lies on the red side of the red sequence , as one would expect .
so , we can not claim dramatic systematic differences between color and brightness span of the mw and m 87 gc sequences .
the right side of figure [ f : mw_ata ] shows the total - mass scale for the clusters . to estimate the gc masses we assumed the total mass of @xmath2 cen to be @xmath85 @xmath86 ( dsouza & rix 2013 ) , and we linearly scaled this value with the f606w - band luminosities , assuming the same mass - to - light ratio for all clusters .
this scale is just a rough estimate of the mass of mw and m 87 clusters , but it should represent a fair first - order mass ranking .
for example , the total mass of ngc 6397 is about @xmath87 @xmath88 ( heyl et al .
2012 ) , which is about what the scale predicts over an order of magnitude away from the zero point .
this implies that at the faint limit ( roughly at 10% completeness level ) , we can still detect in m 87 some gcs with masses as low as @xmath89 @xmath86 like , e.g. , ngc 6717 ( caloi & dantona 2011 ) , while even less massive clusters are not included in our sample .
peng et al . (
2009 ) showed that the brightest m 87 gcs have half - light radii as large as 9 pc , corresponding to @xmath255.8 pixels on our stacks .
a non - negligible fraction of the light of these clusters necessarily falls outside our adopted 7-pixel - radius aperture photometry . using a 15-pixel - radius aperture photometry ,
the brightest m 87 gcs are found to be @xmath90% brighter than with the 7-pixel - based photometry , meaning that they might be up to @xmath91 more massive than what the mass scale in figure [ f : mw_ata ] would predict .
the mw clusters for which we have extensive optical and uv photometry allow us to undertake a simple experiment to quantify the sensitivity of integrated uv@xmath1optical colors to the distribution of stars on the hb .
figure [ f:47tuc ] shows the @xmath44 vs. @xmath92 cmd of @xmath2 cen ( ngc 5139 , grey ) and 47 tuc ( ngc 104 , black ) with the hb stars of 47 tuc being highlighted in red .
various stellar groups and clumps are recognizable on the hb of @xmath2 cen , corresponding to the many sub - populations of this cluster ( see , e.g. , bellini et al .
2010 ) , including the extreme blue _ hook _ ( see also brown et al . 2010 ) .
notice that the brightest portion of the hb of @xmath2 cen in the f275w band is not the blue hook itself , but instead the next ( redder ) group of stars , with @xmath93 .
thus , a maximum effect on the integrated @xmath19 color would be produced if all hb ( red clump ) stars of 47 tuc were moved to this @xmath94 brightest portion of the hb , as indicated by the red circle in figure [ f:47tuc ] . by doing so , i.e. , by moving all hb stars from their actual location to the red circle , the synthetic colors of 47 tuc become @xmath25 0.01 mag bluer in @xmath36 and @xmath25 0.98 mag bluer in @xmath19 with respect to the untouched cluster .
these new colors would place 47 tuc near the @xmath19 bluest rim of the m 87 distribution in the two - color plot of figure [ f : mw_piotto ] . the green arrow in fig .
8d illustrates where 47 tuc would be placed according to this shift .
the new position on the two - color plot is very close to that of ngc 6717 , a cluster with a blue hb and no red clump ( see , e.g. , fig . 29 of piotto et al .
we emphasize that these synthetic color shifts necessarily overestimate the possible effect of secondary , helium - enriched populations , as these may represent only a fraction , even if a major one , of the entire stellar content of a gc .
thus , a substantially smaller @xmath19 color shift , of order of @xmath95 mag , is to be expected if say only @xmath25half of a cluster population were to populate the brightest portion of the hb in the f275w passband .
this is a shift comparable to ( or smaller than ) the dispersion of the m 87 gc sequences , see panel ( c ) of figure [ f : mw_piotto ] .
a larger effect is to be expected when a bluer uv passband is considered .
we then repeated the same 47 tuc experiment using the f225w passband , with the result that the @xmath96 color shift is @xmath97 mag , hence this color is substantially more sensitive to the presence of helium - enriched , blue hb stars belonging to a cluster s second stellar generation .
one important clue as to how the multiple populations may have originated is given by the spatial distributions of the different populations , which might retain information about their formation sites . among the massive mw gcs ,
the second generation is generally more concentrated towards the center of the cluster compared to the first generation ( see , e.g. , bellini et al . 2009
; milone et al .
2012a ) . as such
, spatial gradients represent a fossil record of the cluster s dynamical history .
this finding has indeed promoted the _ cooling flow _ model of dercole et al .
( 2010 , 2011 ) . over time
, the various stellar generations interact dynamically and may homogenize their radial distribution in those clusters with short relaxation time . in the case of the milky way , only the most massive clusters ( e.g. , @xmath2 cen , 47 tuc , ngc 6441 ) have a half - mass relaxation time long enough ( greater than several gyrs ) to still retain some fossil record of different formation sites of first- and second - generation stars .
obviously , it is impossible to resolve m 87 gcs into stars and study their spatial distribution star by star .
however , helium - rich second - population stars are bluer than he - normal stars .
thus , the detection of a color gradient in the light profile of individual m 87 gcs would provide an indication for the presence in them of he - rich sub - populations . for this experiment we selected clusters brighter than @xmath98 ( corresponding to @xmath99 @xmath88 ) and reasonably bright in uv ( @xmath100 ) , so to have a relatively long relaxation time and enough s / n .
a total of 214 m 87 gcs satisfied these selection criteria , specifically : 150 bgcs , 30 igcs and 34 rgcs .
if we were to apply the same selection criteria to mw gcs , we would have included just seven among the treasury gcs : ngc 2808 , ngc 5024 , @xmath2 cen , ngc 5272 , ngc 5286 , ngc 6715 and ngc 7089 . at the distance of m 87
, 4.6 pc ( the typical half - mass radius for the selected gcs ) corresponds to about 3 pixels on our stacks , and a 3-pixel - radius aperture photometry was measured for all the 214 selected sources .
typically , the flux of 29 pixels is summed within a 3-pixel radius ( depending on the position of the source centroid within the central pixel ) .
we then compared these 3-pixel - radius - based aperture fluxes with those measured in the annulus between 3 and 4 pixels ( @xmath101 pc ) ( enclosing 20 pixels ) , to keep similar poisson statistics .
the flux ratios @xmath102 within the 3-pixel radius were then computed and compared with those between 3-pixel- and 4-pixel - radius apertures @xmath103 .
the top - left panel of figure [ f : color_gr ] shows these values as a function of the @xmath44 magnitude for bgcs , igcs and rgcs ( in blue , yellow and red , respectively ) .
igcs and rgcs have very similar colors within and outside the 3-pixel - radius apertures .
bgcs , on the contrary , have the same color - excess value just as lower limit , and their color is in general appreciably bluer within 3 pixels than outside , with some of them being up to 1% bluer .
we highlighted 2 regions in the plot that define bgcs with a mild color gradient ( 38 gcs , in light grey ) and with a higher color gradient ( 8 gcs , in dark grey ) .
the histogram of the color - excess distribution for bgcs , igcs and rgcs is shown on the top - right panel ( and color - coded accordingly ) .
as a cross check , we repeated the same exercise using the f606w passband instead of the f814w one .
the results are shown in the lower panels of figure [ f : color_gr ] .
the color differences are now smaller than before , but the trend is still clear .
this uv - color excess ( or gradient ) suggests that a good fraction of bgcs may indeed be hosting he - rich second generation of stars that are still more centrally concentrated than first - generation stars .
other uv - bright sources , like blue - straggler stars , are not a valid explanation : while on one hand they are expected to be more concentrated in the core of gcs , nevertheless they are still 23 magnitudes fainter than blue hb stars and substantially less numerous than hb stars among mw gcs ( e.g. , ferraro et al . 2003 ) . finally , figure [ f : color_gr1 ] shows the location in the two - color diagram of the clusters used for the color - gradient analysis .
panel ( a ) shows all m 87 gcs with black points , whereas the 214 selected clusters are color - coded according to their parent population ( bgcs in blue , igcs in yellow , and rgcs in red ) .
the lines used to divide the clusters into subpopulations are also shown in light blue , for clarity . a zoom - in view of the grey box in panel ( a ) is provided in panel ( b ) , where this time we are plotting only the selected clusters ( black dots ) .
clusters for which we found a mild color gradient are highlighted with grey diamonds , while higher - color - gradient clusters are marked by black triangles .
we also checked for the possibility of these color gradients being the result of passband differences in the shape of the psfs .
we found that psf differences contribute less than @xmath104 to the `` high gradients '' seen in figure [ f : color_gr ] . we note that selected bgcs are spread all along the bgc sequence in the two - color diagram of panel ( a ) of figure [ f : color_gr1 ] , but those displaying a mild color gradient are preferentially found among the bluer bgcs in both @xmath18 and @xmath105 colors . on the other hand , clusters with high color gradients
are only found among the bluest m 87 gcs .
we conclude that the gcs with both bluest @xmath18 color for their optical color and with a high color gradient represent the best candidates in our sample for hosting helium - enriched sub - populations producing blue hb stars .
note that this is a preliminary exploration of color gradients in these m 87 gcs ; a more quantitative analysis would require detailed profile fitting , which is beyond the purpose of the present paper .
as illustrated in section [ sec:47 ] and figure [ f:47tuc ] , the hb of helium - rich gc sub - populations can be very bright in the ultraviolet , potentially making the clusters very blue in integrated uv@xmath1optical colors .
this is the case for the wfc3 uv filters we have considered , and even more so for the stis / mama fuv passband ( 11501700 ) , that samples substantially shorter wavelengths than either the f225w and f275w passbands .
fuv photometry should help us to better characterize the possible presence of such extreme helium - enriched sub - populations , as already been attempted by , e.g. , s06 and k07 for m 87 gcs , and by dalessandro et al .
( 2012 ) for mw gcs .
here we want to take advantage of what we have learned so far from our deep f275w , f606w and f814w integrated photometry and extend our analysis to include the fuv . to this end , we cross - identified sources in common between our catalogs and those published in s06 and dalessandro et al .
( 2012 ) . among the 66 m 87 sources with reliable fuv photometry in s06 , only 53 are in common with our catalog .
a visual inspection of the missing sources in our stacks revealed that they are either background galaxies ( having a distorted shape in at least one of our stacks ) , are not found in f814w , or have non - positive flux in f275w .
cmd ( a ) and two - color diagram ( b ) of mw gcs obtained by cross - identifying our mw catalog with the galex fuv photometry of dalessandro et al .
the latter compares to panel ( c ) of figure [ f : tony ] , for m 87 gcs . ]
panel ( a ) of figure [ f : tony ] shows the fov of our m 87 field , with all the reliable sources in our catalog as black dots .
the four fields imaged by stis / mama are tagged and highlighted with circles .
the 53 sources in common with s06 are shown as filled grey circles .
panel ( b ) shows the optical cmd of all our m 87 gcs ( grey dots ) , and of those cross - identified with s06 , color - coded according to their parent population ( bgc , igc or rgc ) .
the two - color diagram employing fuv photometry is presented in panel ( c ) , with photometric error bars .
panels ( d ) , ( e ) and ( f ) show three cmds based on combinations of fuv , wfc3/uvis and acs photometry .
there is marginal evidence that among these 53 common objects the brightest rgcs are also the bluest in the fuv@xmath106 color , with one cluster ( highlighted with an open diamond in panels ( d ) , ( e ) and ( f ) standing out as the bluest of all .
the fact that this cluster is simultaneously red in the optical color and blue in the uv color suggests that it might host two populations : a metal - rich first generation and a he - rich second generation with a very hot hb . as such
, it could represent an extreme analog of ngc 6388 and ngc 6441 in the mw .
note that this object is still red when the uv@xmath1optical color is based on the f275w filter ( panel f ) , indicating that its hb may be mostly occupied by extremely - hot stars ( blue - hook stars in figure 10 ) . in any case
, objects like this one seem to be the exception rather than the rule ( 1 out of 53 ) .
perhaps more interesting is the fact that most rgcs ( and most igcs as well ) are bluer than bgcs in the fuv@xmath107 color , whereas they are redder in all other colors .
this may signal that indeed these clusters harbor many blue - hook stars .
the central field of s06 ( f4 in panel a ) also contains photometry in the nuv passband .
we have 7 gcs in common between our catalog and that of s06 in this field . panel ( g ) of figure [ f : tony ] shows the correlation between the fuv@xmath108 and x@xmath108 colors , where x is either f275w ( in red ) or nuv ( in black ) .
it is worth mentioning the good agreement between our f275w photometry and s06 nuv measurements , despite the different adopted photometric system .
what about our galaxy ? does the mw have similar gcs compared to the anomalous ones we found in m 87 ? to find this out , we cross - identified our mw catalog with the integrated galex fuv photometry published in dalessandro et al .
we did not convert their abmag values into vegamag , for the same reasons stated above in the case of s06 stmag .
there are 23 clusters in common between dalessandro et al .
( 2012 ) and our 45 mw gcs discussed in section [ sec : mw ] . in figure
[ f : dale ] we show the fuv vs. fuv@xmath109 cmd on the left ( panel a ) , and the uv - to - optical two - color diagram on the right ( panel b ) .
all magnitudes are on the same reference system of m 87 gcs ( i.e. , the same distance and reddening of m 87 , but _ not _ the same photometric system ) .
the latter diagram should be compared to panel ( c ) of figure [ f : tony ] for m 87 gcs .
the two brightest objects in fuv are , not surprisingly , ngc 2808 and ngc 7089 ( m2 ) in red in both panels .
the former cluster has at least three sub - populations with he abundances up to @xmath110 ( piotto et al .
the latter hosts seven subpopulations with he content up to 0.33 , with two sub - populations making @xmath111 of the cluster being substantially enriched also in iron as well ( milone et al .
both clusters have a well populated and extended blue hb , with multiple clumps .
yet , they do not stand out for having particularly blue fuv@xmath112 or @xmath113 colors .
the bluest object in panel ( a ) of figure [ f : dale ] is ngc 6254 ( m 10 ) , ( marked in green in both panels ) .
this object is also the bluest in fuv@xmath106 color .
ngc 6254 has a well - populated bhb ( see , e.g. , figure 5 of dalessandro et al .
2013 ) , and the presence of distinct multiple populations within this clusters has just been reported ( piotto et al .
2015 ) . at odds with m 87 gcs
, we find no objects simultaneously red in the optical color and blue in the fuv - based color among these 23 clusters .
moreover , clusters that are red in the optical have about the same uv color ( or are even redder ) in fuv@xmath106 ( panel b ) .
we have imaged through the uv f275w passband the central @xmath0 field of the giant elliptical galaxy m 87 , using the wfc3/uvis camera onboard the _ hst_. in combination with archival acs / wfc f606w and f814w data covering the same field , we have constructed various optical and uv@xmath1optical color - magnitude and two - color plots in an attempt to identify candidate globular clusters hosting a significant helium - enriched sub - population , analogous to those present in several gcs of our own galaxy , that produce very - blue hb stars .
this experiment has attained only partial success .
out of the 1460 candidate gcs identified in all three bands , the best candidates for hosting helium - enriched sub - populations consist of a small group of clusters that are both slightly bluer in @xmath114 for their optical color than mw gcs , and exhibit a uv color gradient . using real stellar data for some mw globulars ,
we have empirically estimated the maximum possible effect on uv@xmath1optical colors of a secondary , helium - enriched population .
assuming that such a population makes 50% of the cluster mass , the maximum blueing of the cluster integrated color is @xmath95 mag in @xmath19 and @xmath115 mag in @xmath116 .
this @xmath117 shift is similar to or smaller than the color width of either the blue or the red gc sequence in m 87 , hence it does not allow us to unambiguously detect clusters with an exceptionally blue hb . clearly , the @xmath116 color would offer a better chance to do this , but wfc3/uvis exposures with the bluer f225w filter would have required considerably longer integration times . for a set of 45 gcs in the mw we have constructed integrated @xmath113 , @xmath118 and @xmath119 colors , directly from deep photometry of individual stars that are members of the clusters .
after reddening corrections , the uv@xmath1optical colors of these mw clusters are compared to those of m 87 .
we extended the comparison to 53 milky - way gcs ( but only to f606w and f814w magnitudes ) , and used them to derive a first - guess estimate of the mass of m 87 gcs .
we included fuv information for 53 m 87 gcs from s06 , and found that the brightest rgcs are also as blue as or even bluer than the brightest bgcs in fuv@xmath106 , with one object being significantly bluer .
a similar comparison is made also for 23 mw gcs using dalessandro et al .
( 2012 ) fuv integrated magnitudes . in this case , the bright mw red clusters in optical are not as blue as the bright blue cluster in neither of the colors employing the fuv , at odd with m 87 clusters .
unfortunately , the lack of a statistically - significant sample of metal - rich mw clusters with fuv photometry does not allow us to make a full comparison with the m 87 gc system .
other results of this investigation include : * both red and blue gcs appear to have a slightly flattened distribution , nearly perpendicular to the famous jet , with this flattening being more pronounced for the blue gcs .
* we find that m 87 blue gcs span the same uv@xmath1optical color range as blue gcs in the mw .
however , they appear to be redder by @xmath81 mag in the optical color @xmath120 , compared to mw gcs , producing a systematic relative shift in the two - color diagram in qualitative agreement with earlier findings by sohn et al .
( 2006 ) and kaviraj et al .
however , we can not exclude that this color shift may result from differences in how colors are measured in the two cluster families .
* only a handful of m 87 gcs in this survey field appear to be more massive than @xmath2 cen , the most massive cluster in the mw , with some as massive as a few @xmath10 m@xmath7 .
* we identify a small sample of m 87 red gcs which are extremely blue in fuv@xmath106 ( fig .
[ f : tony]c ) , which have no known counterparts among the mw gcs ( fig .
[ f : dale ] ) .
such blue colors suggests they may contain a unusual number of extremely blue , helium rich hb stars .
* as a side product , in an appendix we briefly discuss the images of the jet , counterjet and dust lanes near the center of m 87 , which become especially evident in the ultraviolet f275w stack .
we release the photometric catalog of m 87 gcs in the f275w , f606w and f814w bands to the astronomical community , as well as the three high - quality image stacks in the same filters .
* acknowledgments . *
the authors gratefully thank the anonymous referee for his / her useful suggestions that helped improving the manuscript , and thomas puzia for a critical reading of the manuscript .
ab and ja acknowledge support from stsci grant go-12989 .
apm acknowledges the financial support from the australian research council through discovery project grant dp120100475 .
ms acknowledges support from stsci grant go-13297 and becas chile de postdoctorado en el extranjero project 74150088 .
anderson , j. , & king , i. r. 2006 , acs / isr 2006 - 01 ( baltimore : stsci ) ( 2006 ) anderson , j. , sarajedini , a. , bedin , l. r. , et al .
2008 , , 135 , 2055 anderson , j. , & bedin , l. r. 2010 , , 122 , 1035 bastian , n.;,lamers , h. j. g. l. m. , de mink , s. e. , longmore , s. n. , goodwin , s. p. & gieles , m. 2013 , , 436 , 2398 bedin , l. r. , piotto , g. , anderson , j. , et al .
2004 , , 605 , l125 bellini , a. , & bedin , l. r. 2009 , , 121 , 1419 bellini , a. , piotto , g. , bedin , l. r. , et al .
2009 , , 507 , 1393 bellini , a. , bedin , l. r. , piotto , g. , et al .
2010 , , 140 , 631 bellini , a. , anderson , j. , & bedin , l. r. 2011 , , 123 , 622 bellini , a. , piotto , g. , milone , a. p. , et al .
2013 , , 765 , 32 brown , t. m. , ferguson , h. c. , smith , e. , et al .
2005 , , 130 , 1693 brown , t. m. , sweigart , a. v. , lanz , t. , et al .
2010 , , 718 , 1332 busso , g. , cassisi , s. , piotto , g. , et al .
2007 , , 474 , 105 cardelli , j. a. , clayton , g. c. , & mathis , j. s. 1989 , , 345 , 245 caloi , v. , & dantona , f. 2011 , , 417 , 228 ciotti , l. , dercole , a. , pellegrini , s. & renzini , a. 1991 , , 376 , 380 ciotti , l , & ostriker , j.p .
2001 , , 551 , 131 dantona , f. , & caloi , v. 2004 , , 611 , 871 dantona , f. , gratton , r. & chieffi , a. 1983 , mem.s.a.it .
54 , 173 dercole , a. , dantona , f. , ventura , p. , vesperini , e. & mcmillan , s. l. w. 2010 , , 407 , 854 dercole , a. , dantona , f. & vesperini , e. 2011 , , 414 , 130 dsouza , r. , & rix , h .- w . 2013 , , 429 , 1887 dalessandro , e. , schiavon , r. p. , rood , r. t. , et al .
2012 , , 144 , 126 dalessandro , e. , ferraro , f. r. , lanzoni , b. , et al .
2013 , , 770 , 45 de mink , s. e. , pols , o. r. , langer , n. , & izzard , r. g. 2009 , , 507 , l1 emsellem , e. , cappellari , m. , krajnovi , d. , et al .
2011 , , 414 , 888 ferraro , f. r. , sills , a. , rood , r. t. , paltrinieri , b. , & buonanno , r. 2003 , , 588 , 464 gebhardt , k. , & kissler - patig , m. 1999 , , 118 , 1526 harris , w.e .
1996 , aj , 112 , 1487 ( 2010 edition ) harris , w. e. 2009 , , 703 , 939 heyl , j. s. , richer , h. , anderson , j. , et al .
2012 , , 761 , 51 kaviraj , s. , sohn , s. t. , oconnell , r. w. , et al .
2007 , , 377 , 987 king , i. r. , bedin , l. r. , cassisi , s. , et al .
2012 , , 144 , 5 krause , m. , charbonnel , c. , decressin , t. , meynet , g. , & prantzos , n. 2013 , , 552 , aa121 larsen , s. s. , & brodie , j. p. 2000 , , 120
, 2938 liu , y. , zhou , x. , ma , j. , et al .
2005 , , 129 , 2628 ma , j. , de grijs , r. , fan , z. , et al .
2009 , research in astronomy and astrophysics , 9 , 641 milone , a. p. , bedin , l. r. , piotto , g. , et al .
2008 , , 673 , 241 milone , a. p. , piotto , g. , king , i. r. , et al .
2010 , , 709 , 1183 milone , a. p. , piotto , g. , bedin , l. r. , et al .
2012a , , 744 , 58 milone , a. p. , marino , a. f. , piotto , g. , et al .
2012b , , 745 , 27 milone , a. p. , et al . , 2015 , mnras , 447 , 927 norris , j. e. 2004 , , 612 , l25 novak , g. s. , ostriker , j. p. & ciotti , l. 2012 , , 427 , 2734 peng , e. w. , ct , p. , jordn , a. , et al .
2006 , , 639 , 838 peng , e. w. , jordn , a. , blakeslee , j. p. , et al .
2009 , , 703 , 42 piotto , g. , villanova , s. , bedin , l. r. , et al .
2005 , , 621 , 777 piotto , g. , bedin , l. r. , anderson , j. , et al .
2007 , , 661 , l53 piotto , g. 2009 , iau symposium , 258 , 233 piotto , g. , milone , a. p. , anderson , j. , et al .
2012 , , 760 , 39 piotto , g. , milone , a. p. , bedin , l. r. , et al . 2015 ,
aj , 149 , 91 renzini , a. 1983 , mem.s.a.it . , 54 , 335 renzini , a. 2008 , , 391 , 354 renzini , a. 2013 , mem.s.a.it . , 84 , 162 391 , 354 rich , r. m. , sosin , c. , djorgovski , s. g. , et al . 1997 , , 484 , l25 silverman , b. w. 1986 , monographs on statistics and applied probability ( london : chapman & hall ) sirianni , m. , jee , m. j. , bentez , n. , et al .
2005 , , 117 , 1049 sohn , s. t. , oconnell , r. w. , kundu , a. , et al .
2006 , , 131 , 866 strader , j. , romanowsky , a. j. , brodie , j. p. , et al .
2011 , , 197 , 33 trager , s. c. , king , i. r. , & djorgovski , s. 1995 , , 109 , 218 van albada , t. s. , de boer , k. s. & dickens , r. j. 1981 , , 195 , 591 van den bergh , s. 1967 , , 72 , 70
we briefly comment here on some of the features related to the prominent jet , so evident in the stack images .
we may return to some of these issues in the future , but for this paper we are primarily focused on the gc population . besides the famous jet of m 87 , one obvious counterjet feature is clearly visible in figure 1 , with three other faint filamentary structures being possibly related to the counterjet ( such as counterjet - generated shock fronts ) .
the red - green - blue color - scale of this panel is also enhancing the visibility of these dim features .
some of them are bright in f606w ( marked as `` green features '' in the figure ) , while others are almost invisible in this band , yet are still bright in the other two bands ( `` purple features '' ) .
these color differences may be due to different emission - line ratios resulting from different excitations among the features .
the dark areas near the center of m 87 visible in figure 1 are not an artifact of the galaxy - subtraction routines but are real features in the scene .
figure [ f : app ] shows these features in more detail , having been optimized for this purpose .
various dust filaments around the galactic center and the jet are now clearly visible : they are also present in the non sky - subtracted stacks , although they stand out less clearly , due to the scattered light of the galactic core .
they are visible in all three bands , but are definitely more prominent in the f275w image , as expected from obscuring dust clouds .
notice also the sharp edges of some of them , with one being clearly in front of the jet and attenuating the light of its obscured section , suggesting they could be the swan song of recent tidally - disrupted dwarf galaxies .
the phenomenology of the m 87 core is typical of what one expects to happen in a very massive elliptical galaxy .
supernova heating at the known rate of type ia supernov is indeed insufficient to prevent gas outflows to eventually turn to inflows following a central - cooling catastrophe : as stellar mass loss accumulates , the hot gas density increases until cooling exceeds the supernova heating ( ciotti et al .
1991 ) . under such circumstances
, the inflowing material can feed the central supermassive black hole , starting a series of intermittent nuclear activity with duty cycles of several @xmath11 yrs ( ciotti & ostriker 2001 ; novak , ostriker & ciotti 2012 ) .
this kind of recurrent activity is indeed instrumental in keeping the galaxy quenched , once star formation has been quenched for the first time ( by whatever mechanism ) .
coming back to the dark spots , one interesting question concerns the origin of the obscuring dust . forming and growing dust in the interstellar medium itself , at the time of the cooling catastrophe near the center , is one option .
however , the density may not be high enough for rapid nucleation except perhaps near the black hole itself . alternatively
, dust may form in the wind of mass - losing red giants and agb stars and survive sputtering just in the cooler gas resulting from the cooling catastrophe .
we have no decisive argument either way , other than that these dust features are most probably vestiges of the events that lead to the re - fueling of the central black hole
. one can also notice that bright regions appear in between dark patches .
their intrinsic nature may actually be similar to the dark patches , possibly being dust - rich clouds that shine of dust - scattered light from the agn itself .
|
we have imaged with _ hst _
s wfc3/uvis the central @xmath0 region of the giant elliptical galaxy m 87 , using the ultraviolet filter f275w . in combination with archival acs / wfc data taken through the f606w and f814w filters , covering the same field , we have constructed integrated - light uv@xmath1optical colors and magnitudes for 1460 objects ,
most of which are believed to be globular clusters belonging to m 87 .
the purpose was to ascertain whether the multiple - populations syndrome , ubiquitous among galactic globular clusters ( gcs ) , exists also among the m 87 family of clusters . to achieve this goal , we sought those gcs with exceptionally blue uv - to - optical colors , because helium - enriched sub - populations produce a horizontal - branch morphology that is well populated at high effective temperature . for comparison , integrated , synthetic uv@xmath1optical and purely optical colors and magnitudes
have been constructed for 45 galactic gcs , starting from individual - star photometry obtained with the same instruments and the same filters .
we identify a small group of m 87 clusters exhibiting a radial uv@xmath1optical color gradient , representing our best candidate gcs hosting multiple populations with extreme helium content .
we also find that the central spatial distribution of the bluer gcs is flattened in a direction parallel to the jet , while the distribution of redder gcs is more spherical .
we release to the astronomical community our photometric catalog in f275w , f606w and f814w bands and the high - quality image stacks in the same bands .
|
improvements in the zodiacal light emission and scattering to be given by space observations at 3 au , compared with observations near the earth , were discussed by _ mather and beichman ( 1996)_. unfortunately , a rather accurate multi - parametric model of the zodiacal brightness derived by _ kelsall et al .
( 1998 ) _ from the cobe data can not be reliably extrapolated to heliocentric distances as large as 3 au .
we have developed a physical model of the zodiacal cloud incorporating the real dust sources of asteroidal and cometary origin , which makes it possible to evaluate quantitatively the zodiacal light emission and scattering throughout the solar system ( _ gorkavyi et al .
1997a _ ) .
this model considerably improves our previous ` reference model ' based on the use of the continuity equation for distribution function of dust particles ( _ gorkavyi et al .
1997b , c , 1998 _ ) and enables us to obtain more reliable results than a phenomenological modelling of the zodiacal light ( e.g. _ ebbets 1998 _ ) .
below , we describe an improved model that represents a 3d - grid containing @xmath1 cells with a step in ( heliocentric latitude @xmath2 , longitude @xmath3 , and radius @xmath4 ) to be @xmath5 [ au ] ) . using a new numerical approach to the dynamics of minor bodies and dust particles , we increase the number of particle positions employed in each model to @xmath6 without using a supercomputer ( _ ozernoy et al .
2000 , gorkavyi et al . 2000 _ ) .
we compute here the distribution of the zodiacal dust emission and scattering in the solar system .
the processes influencing the dust particle dynamics include gravitational scattering on all planets , except pluto ( inclinations and precession of the planets are neglected ) , mean motion resonances , and the poynting - robertson / solar wind drags .
we employ here 931 sources of dust particles ( with eccentricities @xmath7 ) , which include 284 asteroids , 451 short - period comets , and 196 kuiper belt objects .
we adopt a two - component approximation for dust particle size distribution , which is characterized by the parameter @xmath8 and 0.057 for ( 1 - 2)@xmath0 m and ( 5 - 10)@xmath0 m particles , respectively .
the inferred dust density distribution , @xmath9 , of the ipd at heliocentric distances @xmath10 au enables us to compute the scattered light and the thermal emission of the zodiacal cloud as a function of the observer s latitude and longitude .
the thermal emission and scattering of the zodiacal cloud are given by : @xmath11ds\ ] ] where @xmath12 is the planck function at frequency @xmath13 , @xmath14 is albedo , @xmath15 is the dust temperature , @xmath16 with @xmath17k and @xmath18 ( _ kelsall et al . 1998 _ ) , @xmath19 is the solar flux , @xmath20 is the phase function at scattering angle @xmath21 ( _ kelsall et al .
1998 _ ) .
note that although the phase function only weakly depends on @xmath13 , the value of @xmath22 strongly depends on @xmath13 through the solar flux .
besides , the planck function strongly depends on heliocentric distance , especially in the near ir ( 1 - 5 @xmath0 m ) .
using the inferred distribution of the zodiacal dust , we have computed a variety of zodiacal light maps , both for thermal emission and scattered components , at different locations @xmath23 of the ngst assumed to be at 1 au or 3 au ( see fig . 1 for the @xmath24m - emission of asteroidal dust and fig . 2 for that of kuiperoidal dust ) .
the brightness as a function of latitude @xmath2 and longitude @xmath3 ( in the telescope s frame ) is given in logarithmic scale ( the neighboring contour intensities differ by @xmath25 ) .
the sun s position is ( 0,0 ) for @xmath26 and is shifted to a negative @xmath2 for @xmath27 au . at each location , there is a minimum in the zodiacal light which can be seen as a ` dark spot ' ( or several spots ) in the computed figures .
the positions of those minima are explained by an interplay between dependencies of emissivity upon density and temperature .
our numerical results concerning the computed zodiacal light at different locations of ngst are summarized in tables 1 and 2 .
the data presented in table 2 for @xmath28 au , 2nd column can be compared with the cobe data @xmath29 ( 1.25@xmath0 m band ) , @xmath30 ( 5@xmath0 m band ) , and @xmath31 ( 20@xmath0 m band ) .
interestingly , _ it is the kuiperoidal component ( k1 ) whose shape resembles most closely the profile of the zodiacal cloud _ , which might imply that the role of kuiperoidal particles has been highly underestimated so far .
.components of the model zodiacal cloud .
[ cols="^ , > , > , > , > , > " , ]
the structure of the zodiacal dust cloud computed in the present work is substantially non - uniform : \(iii ) the partial contribution of particles of different origin and size changes with heliocentric distance : as the distance increasess , the contribution of asteroidal particles ( especially of large size ) into the zodiacal light emission sharply decreases , whereas the contribution of the kuiperoidal component increases ; \(iv ) the latitudinal dependence of the zodiacal emission is different for different components of the dust cloud . in the anti - sun direction ( @xmath32 ) , the maximum of the asteroidal dust emission at @xmath33 m is reached at the ecliptic plane , whereas the brightness of the cometary component has its minimum here .
\3 . at @xmath34
au , the zodiacal light brightness in the direction of the pole decreases to @xmath35% at @xmath36 m ; to @xmath37% at @xmath33 m , and to @xmath38% at @xmath39 m compared to that seen near the earth .
\4 . at each wavelength
, there are certain regions on the sky having a minimal zodiacal light brightness . at short wavelengths ( @xmath40 m ) , these regions are around the poles , and at long wavelengths ( such as 5 to 20 @xmath0 m ) these regions are shifted toward the ecliptic .
this result needs to be taken into consideration while planning observations with ngst .
further improvements in our physical modelling of the zodiacal light are feasible . using cobe and iras data ,
the contribution of dust components of different origin into the zodiacal cloud could be determined .
those improvements would make far going implications possible for extragalactic astronomy and cosmology .
[ ebbets , d. 1998 , zodical light model for ngst studies " .
( unpublished ) ] [ gorkavyi , n.n . ,
ozernoy , l.m . , &
mather , j.c .
1997a , apj 474 , 496 ] [ gorkavyi , n.n . , ozernoy , l.m . ,
mather , j.c .
& taidakova , t.a . 1997b , baas 29 , 1310 ] [ 1997c , apj 488 , 268 ] [ 1998 , earth , planets and space , 50 , 539 ] [ 2000 , planetary and space science ( submitted ) ] [ kelsall , t. et al .
1998 , apj 508 , 44 ] [ mather , j.c .
& beichman , c.a .
1996 , unveiling the cosmic infrared background " , ed .
e. dwek , p. 271 ] [ ozernoy , l.m . ,
gorkavyi , n.n . & taidakova , t.a .
2000 , icarus ( submitted ) ]
|
we develop a physical model of the zodiacal cloud incorporating the real dust sources of asteroidal , cometary , and kuiperoidal origin . using the inferred distribution of the zodiacal dust ,
we compute its thermal emission and scattering at several wavelengths ( 1.25 , 5 , and 20 @xmath0 m ) as a function of ngst location assumed to be at 1 au or 3 au . areas on the sky with a minimum of zodiacal light are determined .
|
in previous studies the question of whether the sun has a normal composition has been discussed in detail ( gustafsson 1998 , 2008 ; gustafsson et al .
2010 ; allende prieto 2008 , 2010 ) .
it has been recognized that earlier studies obtained discrepant results regarding possible chemical abundance anomalies in the sun , probably due to relatively large ( 0.05 - 0.10 dex ) abundance uncertainties .
an illustrative example is the conflicting results found for the [ o / fe ] ratio in solar - metallicity thin disk dwarfs . in the seminal study by edvardsson et al .
( 1993 ) , the [ o / fe ] ratio around [ fe / h ] = 0 in disk stars was found to be somewhat subsolar , meaning that the sun could be either somewhat oxygen - rich or somewhat iron - poor , relative to disk stars .
different results were obtained by bensby et al .
( 2004 ) and ramrez et al .
( 2007 ) , who found that solar - metallicity disk stars have [ o / fe ] @xmath0 0 and [ o / fe ] @xmath0 + 0.1 dex , respectively .
these three different studies show that depending on how the analysis is made , worrisome differences of up to 0.15 - 0.20 dex could be found in the [ o / fe ] ratios of disk stars .
even the consistent analysis of three different oxygen abundance indicators by ecuvillon et al .
( 2006 ) , resulted in discrepant oxygen abundances ( solar , sub - solar and super - solar ) regarding the [ o / fe ] ratios of disk stars . a serious issue when comparing the sun to other solar - type stars is the effective temperature scale , as different photometric @xmath1 scales ( e.g. , alonso et al .
1996 ; ramrez & melndez 2005 ; casagrande et al .
2010 ; boyajian et al .
2013 ) may have different zero - points . on this regard , it is important to check the zero - points of a given temperature scale using solar colors as those derived recently by melndez et al .
( 2010 ) , ramrez et al .
( 2012 ) and casagrande et al .
( 2012 ) , using a large sample of solar twins and solar analogs .
the same concern applies to photometric metallicities , i.e. , their zero - point must be tested before they are used to compare the sun to other stars ( e.g. , rnadttir et al .
2010 ; melndez et al .
2010 ; datson et al . 2012 ) .
another problem that may arise when comparing the sun to other stars is the adopted line list and atomic data , and how that data is used to obtain spectroscopic stellar parameters ( @xmath2 ) .
sometimes gf - values are `` calibrated '' using very high resolution solar atlases , and then applied to stars that have been observed using a spectrograph with a much lower resolution , where the line profiles are not as well defined as in the solar atlas .
even worse would be to use solar gf - values determined by a different group using a different set of model atmospheres .
the use of absolute ( laboratory or theoretical ) gf - values is also problematic , as their large uncertainties could imply relatively large errors ( low precision ) in the obtained stellar parameters .
also , although naively we would expect that absolute gf - vaues would result in accurate spectroscopic stellar parameters , this is not necessarily the case due to problems in model atmospheres ( asplund 2005 ) and the simplifications made for the treatment of line formation ( lte vs. nlte ; e.g. , lind , bergemann & asplund 2012 ) . also , precise absolute gf - values exist only for a minority of the lines , meaning that to obtain spectroscopic equilibrium using fei and feii lines we may have to use some imprecise absolute gf - values .
the same applies to the determination of chemical abundances for elements other than iron .
it is also dangerous to use different sets of lines for the sun and for the stars , because due to the large errors in the absolute gf - values , a different set of lines could lead to a different solution , resulting in inaccurate and imprecise stellar parameters and stellar abundances .
blends are also a serious issue , as the dependence of a given line to both stellar parameters and abundance of the main component could be affected by the minor component of the blend .
this problem could be minimized by selecting clean lines or lines with weak blends , or also by deblending or spectrum synthesis .
another issue is both continuum determination during the data reduction and continuum placement in the measurements .
the sun and the sample stars must have a similar determination of the continuum . the way the measurements are done is also very important .
it is increasingly common in the literature to use automatic tools such as ares ( sousa et al .
2007 ) to measure the equivalent widths , yet the measurements done for a given sample star may not be identical to the way the measurement is done for the sun .
even solar twins that may look almost identical to the sun can show small variations in their spectra relative to the solar spectrum , due to somewhat different stellar parameters ( teff , log g , vt ) , different rotation velocities , different chemical compositions and different levels of telluric contamination . for a given line ,
ares does a continuum normalization in a small spectral window , and due to the small variations mentioned above , the continuum fit done by ares may not be necessarily the same for the stars and the sun .
also , ares may use a different number of components ( lines ) to model a given spectral window , and that number may vary from star to star .
although the automatic measurements are good enough for a typical abundance analysis , they may not be adequate to distinguish tiny variations ( 0.01 dex ) between the chemical compositions of stars and the sun . in summary , systematic differences between the stars and the sun could arise due to the ( i ) analysis techniques ( equivalent widths vs. spectrum synthesis ) , ( ii ) stellar parameters , ( iii ) adopted grid of model atmospheres , ( iv ) treatment of line formation ( lte vs. nlte ) , ( v ) adopted gf - values , ( vi ) adopted line lists , ( vii ) spectral resolution , ( viii ) signal - to - noise ratio , ( ix ) problems with the spectrograph , ( x ) adopted solar spectrum ( sky , moon , moons of other planets , asteroids , solar atlas ) , ( xi ) data reduction , ( xii ) determination of the continuum , ( xiii ) blends , ( xiv ) equivalent width measurements , and ( xv ) adopted solar abundances ( e.g. , asplund et al .
2009 ) , when solar abundances are not determined in the same analysis . many of the above problems could be eliminated or largely minimized by planning the observations of the stars and the sun using the same spectrograph and the same setup .
even better would be to observe the sun in the same observing run as the stars , to guarantee that there were no significant changes in the instrument configuration .
the solar spectrum could be obtained using asteroids , which are almost point sources , so that the data reduction is performed in the same way as for the stars .
continuum normalization should be performed homogeneously in the stars and the sun .
having obtained the stellar and solar spectra in the same way , we can now perform a strictly line - by - line differential analysis , so that we make sure that every line in the spectrum is measured in the same way in the stars and the sun , using the same continuum regions and using exactly the same part of the line profile . then
, the abundances can be determined line - by - line , so that uncertainties in the gf - values will cancel out .
it would be even better to perform this kind of analysis using solar twins , stars with spectra very similar to the sun ( e.g. , melndez et al .
2006 , 2009 ; ramrez et al . 2009
; takeda & tajitsu 2009 ; datson et al .
2012 ) . in this way , many of the modeling uncertainties mentioned above are minimized , resulting in extremely precise abundances at the 0.01 - 0.02 dex level , provided very high resolution ( r @xmath3 60000 ) very high s / n ( @xmath3200 ) spectra are employed .
this is the approach first adopted by melndez et al .
( 2009 ) and ramrez et al .
( 2009 ) , that led to the discovery of abundances anomalies in the chemical composition of the sun .
the sun seems chemically anomalous when compared to nearby solar twins in the thin disk , perhaps due to the formation of terrestrial planets in the sun .
the abundances anomalies that may be caused by planet formation are discussed in more detail below .
the processes of star and planet formation are far from fully understood , but we know that they occur nearly at the same time .
thus , it is likely that the formation of both terrestrial and gas giant planets has imprinted signatures in the chemical composition of their host stars . during the final phases of gas accretion by the star , when the proto - planetary nebula cools , dust grain condensates and later coagulates to form planetesimals and finally terrestrial planets in the inner region ,
while in the outer region rocky - ice embryos will accrete hydrogen and helium to form the gas giant planets ( ( * ? ? ?
* morbidelli et al . 2012 ) ) , although giant planets could also form by gravitational disk instability .
the chemical fingerprints left by these processes are key to constrain different models of planet formation .
the classical example of the star - planet connection is the well - known planet - metallicity correlation in solar type stars , indicating that a higher abundance of metals increases the probability of forming giant planets ( @xcite 1997 ) .
this signature has been confirmed by further works ( e.g. @xcite 2005 ; @xcite 2007 ; @xcite 2010 ) and may favor the theory of core - accretion for giant planet formation .
although the correlation between giant planet frequency and overall stellar metallicity has been important in the context of planet formation , it is probably only the tip of the iceberg regarding the connection between stars and planets . in the subsequent decade after the seminal work by @xcite ( 1997 )
, there have been many attempts to explore other relations between planets and the abundance of specific chemical elements , but no clear result was obtained ( @xcite 2007 ) , probably due to the relatively large uncertainties ( about 0.05 dex ) in the determination of chemical abundances ( @xcite 2005 ) , or in same cases due to a bias in the comparison between stars with and without planets , as shown for example for lithium , for which a fare comparison found no difference in the li abundances of the two samples ( @xcite 2010 ) . as mentioned above ,
errors as low as 0.01 dex could be obtained in a strictly differential analysis of a sample of stars that are very similar to each other ( due to the cancellation of systematic errors ) , opening thus new windows on the planet - star connection . below we will discuss recent works on high precision chemical abundances that are giving further clues on the formation of rocky ( @xcite 2009 ) and giant ( @xcite 2011 ) planets . and
@xmath4 was not detected before due to the large errors in standard chemical composition analysis ( @xmath5dex ) . in @xcite ( 2009 ) , the [ x / fe ] values have errors of about 0.01dex .
the solid line is a linear fit to the data , but broken at @xmath6k . *
b ) * variation of the solar photospheric abundances , relative to fe , if the sun s present - day convective envelope were polluted by 2 earth masses of earth - like material and 4 earth masses of cm chondrite rocks .
note the excellent agreement with the linear fit to the solar twin chemical abundance data ( the solid line in this panel is the same as in panel a ) .
adapted from @xcite ( 2012).,width=518 ]
a new era on precise chemical abundance determinations started with the analysis of 11 solar twins by @xcite ( 2009 ) . as these stars have both stellar parameters and spectra nearly indistinguishable from the sun , many systematic errors that plague stellar abundance analyses are canceled in a strictly line - by - line differential analysis of the solar twins relative to the sun , being possible to achieve an unprecedented precision of 0.01 dex ( @xcite 2009 ; @xcite 2012 ) , about a factor of 5 smaller than standard abundance analyses .
this dramatic improvement is critical to explore the small effects that planet formation may imprint on the chemical composition of stars .
@xcite ( 2009 ) found that the sun is peculiar when compared to a sample of 11 solar twins ( fig .
the abundance of refractory elements ( those that condensed at high temperatures in the inner solar nebula ) is systematically smaller in the sun relative to the solar twins .
the amount of material missing is compatible with the amount needed to form the terrestrial planets ( @xcite 2009 ; @xcite 2010 ; @xcite 2010 ) , meaning that the deficiency of refractory elements in the solar photosphere would disappear if the rocky material formed in the solar system were diluted into the present - day solar convective envelope ( @xcite 2010 ; @xcite 2012 ) , as shown in fig .
this opens the exciting possibility of discovering stars that could potentially host terrestrial planets based on a careful chemical abundance analysis , but certainly further work is needed to consider this as a firm signature of rocky planets .
if the above signature is confirmed , then stars with chemical composition similar to solar could potentially host terrestrial planets .
so far the stars that have their chemical composition most similar to the sun are the solar twins hip 56948 ( @xcite 2007 ; @xcite 2012 ) and hip 102152 ( monroe et al .
interestingly , the precise radial velocity observations obtained for hip 56948 at the mcdonald and keck observatories , and for hip 102152 at la silla , show no signs of giant planets within and in the habitable zone , making these stars excellent candidates for hosting habitable rocky planets .
as discussed extensively in melndez et al .
( 2009 , 2012 ) , the abundances anomalies can not be explained by other causes such as contamination from agb stars , snia , snii or hn ( fig .
2 ) , or by galactic chemical evolution processes or age effects . @xcite ( 2011 ) have shown that the peculiar abundance pattern can not be attributed to line - of - sight inclination effects .
also , the abundance trend does not arise due to the particular reflection properties of asteroids .
although the abundance peculiarities may indicate that the sun was born in a massive open cluster like m67 ( @xcite 2011 ) , this explanation is based on the analysis of only one solar twin .
so far the best explanation for the abundance trend seems to be the formation of terrestrial planets .
the kepler mission should detect the first earth - sized planets in the habitable zones of solar type stars .
we look forward to using 8 - 10 m telescopes to perform careful differential abundance analyses of those stars , in order to verify if our chemical signatures indeed imply rocky planets .
dexk@xmath7 , as derived by @xcite.,width=336 ]
regarding the formation of giant planets , we have recently studied the pair of solar analogs 16 cyg a and b. this pair is very important to understand more about giant planet formation , because the star 16 cyg b hosts a gas giant planet at 1.7 a.u .
( @xcite 1997 ) , while no planets have been detected yet around 16 cyg a. the binary pair is supposed to have the same chemical composition , as it was formed from the same natal cloud , unless the formation of the giant planet around 16 cyg b altered its chemical abundances .
our careful differential abundance analysis between 16 cyg a and b showed that the component b is systematically more metal - poor , by about 10% ( @xmath00.04 dex ) , in the two dozen chemical elements analyzed ( @xcite 2011 ) , as shown in fig .
thus , it seems that the formation of the gas giant around 16 cyg b robbed a fraction of the metals present in its parent nebula ( @xcite 2011 ) .
nevertheless , another abundance analysis ( at somewhat lower precision ) published nearly at the same date , showed no difference between the abundance pattern of components a and b ( @xcite 2011 ) .
although the spectra used by @xcite ( 2011 ) have a higher resolving power and a higher precision was achieved , it would be important to perform a new abundance analysis of the pair to confirm the signature of giant planet formation .
the above signatures of terrestrial and giant planets are telling us that there could be a close connection between stellar chemical composition and planet architecture .
solar twins are ideal to obtain very precise stellar abundances but unfortunately planet information is lacking for most of them , so currently we can not study in detail the relation between chemical abundance anomalies and different type of planets .
the synergy between the high precision ( 0.01 dex ) chemical abundances obtained in solar twins ( fig .
4 ) and high precision ( 1 m s@xmath7 ) radial velocities that can be obtained with harps , can give us new insights into the planet - stellar connection . in order to study with unprecedented detail the connection between chemical abundances and planet architecture , we have been granted 88 nights with harps for a large eso programme that started in october 2011 and should continue until 2015 .
this in an international collaboration involving astronomers from brazil , germany , usa and australia . around 70 solar twins are being observed at the eso la silla observatory and already some of them are showing radial velocity variations compatible with planets .
for all the sample stars we have acquired high resolution high s / n spectra using the mike spectrograph at the 6.5 m magellan telescope in las campanas , in order to obtain a homogeneous set of high precision ( 0.01 dex ) chemical abundances .
also , some stars have spectra of even better quality acquired with uves at the vlt .
our initial chemical abundance analyses show that on a star - by - star basis we can obtain chemical abundances with a precision at the 0.004 - 0.008 dex level , i.e. , even better than initially anticipated ( 0.01 dex ) . for the solar twin hip 56948 ,
the element - to - element scatter is only 0.004 dex for the volatile elements and 0.008 dex for the refractories ( @xcite 2012 ) , and our on - going work on the solar twin 18 sco shows also an element - to - element scatter of only 0.008 dex ( fig 4 ) .
similar results are being obtained for other solar twins ( monroe et al .
2013 ) .
@xmath8 0.09 dex),width=288 ]
the solar twin 18 sco presents a complex abundance pattern ( see fig .
5 ) . on top of the typical trend with condensation temperature seen in other solar twins ( melndez et al .
2013 ; monroe et al .
2013 ) , this star also shows a large enhancement of the neutron - capture elements . using yields of agb stars computed by amanda karakas
, we can verify if the observed enhancement is due to agb pollution .
after subtracting both the trend with condensation temperature and the s - process contribution from agb stars , we are left with an abundance signature identical to the r - process in the solar system ( figure 6 ) .
this specific abundance pattern can only be obtained thanks to the high precision ( 0.01 - 0.02 dex ) we achieved for these heavy elements .
our work has shown that it is possible to obtain chemical abundances with a precision of about 0.01 dex or even better ( @xmath00.005 dex ) in some cases ( @xcite 2012 ) .
this abundance precision can be achieved in a strictly differential line - by - line analysis .
our approach minimizes systematic errors due to deficiencies in the adopted model atmospheres .
this is shown in fig .
7 , where the difference in differential abundances obtained using two dissimilar grids of model atmospheres ( mafags models vs. kurucz overshooting models ; see melndez et al .
2012 ) is shown .
the average difference is only 0.001 dex , with an element - to - element scatter of only 0.00075 dex .
the unprecedented precision that we achieve ( @xmath00.01 dex ) was key to discover that the sun is not a typical star .
the abundance anomalies in the sun may be connected to the formation of terrestrial planets ( @xcite 2009 ; @xcite 2009 , 2010 ) . also , giant planet formation may imprint a signature in the chemical composition of stars ( @xcite 2011 ) .
our on - going planet search around solar twins will allow us to study at unparalleled precision any connection that may exist between planet architecture and chemical abundance peculiarities .
finally , our high precision abundances can give us better insights on the neutron - capture elements , which could be important for chemical tagging in our galaxy .
= 0.00075 dex , i.e. , 0.17% .
, width=288 ]
j.m . thanks support from fapesp ( 2012/24392 - 2 ) .
this work is based on observations obtained at keck , magellan / las campanas , mcdonald , vlt / paranal ( 083.d-0871 ) and la silla ( 188.c-0265 ) .
|
the sun is used as the fundamental standard in chemical abundance studies , thus it is important to know whether the solar abundance pattern is representative of the solar neighborhood .
albeit at low precision ( 0.05 - 0.10 dex ) the sun seems to be a typical solar - metallicity disk star , at high precision ( 0.01 dex ) its abundance pattern seems abnormal when compared to solar twins .
the sun shows a deficiency of refractory elements that could be due to the formation of terrestrial planets . the formation of giant planets may also introduce a signature in the chemical composition of stars .
we discuss both planet signatures and also the enhancement of neutron - capture elements in the solar twin 18 sco .
|
quantum mechanical ( qm ) calculations of molecular reaction dynamics @xcite are generally much heavier than quasi - classical trajectory calculations @xcite , in particular for polyatomic processes @xcite .
hence , for several decades , intense researches aim at building semiclassical methods taking into account the largest quantum effects while keeping with the classical description of nuclear motions @xcite .
moreover , developping such methods naturally leads to shed light on the complex frontier between the quantum and classical descriptions of motion .
last but not least , semiclassical approaches may be used as powerful interpretative tools to rationalize the dynamics of molecular processes .
the semiclassical wigner treatment of bimolecular collisions proposed by lee and scully is among them @xcite . when applied to the collinear inelastic collision between he and h@xmath0 , or he and hbr , which involve strong quantum interferences and/or classically forbidden vibrational transitions , this approach leads to final state populations in very good agreement with exact quantum ones ( at least for the lowest vibrational states of the initial diatom ) , contrary to the quasi - classical trajectory ( qct ) method @xcite .
this is illustrated in figs .
[ fig:1 ] and [ fig:2 ] for he plus h@xmath0 ( note that these results are particularly relevant as comparative semiclassical / quantum studies of collinear processes usually provide more stringent tests of the validity of semiclassical approaches than similar studies for realistic three - dimensional collisions ) .
in addition , the treatment of lee and scully is particularly simple to apply .
hence , it is potentially interesting to study realistic molecular collisions involving certain quantum effects ( other than tunneling though a potential energy barrier ) .
however , the original derivation of this treatment @xcite appears to rest on somewhat arbitrary basis , as shown later below .
the goal of the present work is thus to put this treatment on firmer theoretical grounds by deriving it from first principles .
this work is mainly motivated by the recent extention of the semiclassical wigner treatment of heller @xcite to triatomic ( or triatomic - like polyatomic ) photodissociations @xcite .
for the first time , rotational motions were taken into account in the theory , thus making it applicable to realistic fragmentations , and comparison between its predictions and rigorous quantum results in the case of guo s triatomic - like model of methyl iodide photodissociation @xcite showed nearly quantitative agreement .
since the method of lee and scully bears strong resemblance with the one of brown and heller , we hope to be able in the future to include rotational motions in the former so as to make it applicable to realistic bimolecular processes .
the present work is a preliminary step towards this goal .
the inelastic collision is defined in section [ ii ] .
the lee - scully method is summarized in section [ iii ] .
its derivation from first principles is presented in section [ iv ] .
section [ v ] concludes .
we consider the collinear inelastic collision between atom a and diatom bc(@xmath1 ) .
the process takes place in the electronic ground state .
@xmath2 is the distance between a and the center - of - mass of bc , and @xmath3 is the bc bond length . @xmath4 and @xmath5 are the momenta conjugate to @xmath2 and @xmath3 , respectively .
the classical function of hamilton reads + @xmath6 + @xmath7 is the reduced mass of a with respect to bc and @xmath8 , the one of bc .
@xmath9 is the potential energy of abc associated with the electronic ground state . in the asymptotic channel
, bc is assumed to be a harmonic oscillator of frequency @xmath10 .
its initial vibrational state , of energy + @xmath11 + is denoted @xmath12 .
this state is real and normalized to unity .
the total energy available to the separated fragments is denoted @xmath13 .
the initial collision energy is thus @xmath14 and the initial value of @xmath4 is + @xmath15^{1/2}. \label{1a}\ ] ] + the hamiltonian operator is denoted @xmath16 .
the numerical results presented throughout this work were obtained within the framework of the secrest - johnson model of he+h@xmath0 collision @xcite . in this model ,
the potential energy is approximated by + @xmath17 . \label{1b}\ ] ] + @xmath7 and @xmath8 were kept at 2/3 and 1 , respectively , @xmath18 at 1 , @xmath19 at 1 , @xmath20 at 0.3 and @xmath13 at 10 , otherwise stated .
the value of the equilibrium bond length @xmath21 can be arbitrarily chosen . the central quantity of this work , already shown for he+h@xmath0 in figs .
[ fig:1 ] and [ fig:2 ] , is the probability @xmath22 of transition from bc(@xmath1 ) to bc(@xmath23 ) .
lee and scully consider a hybrid model where the initial and final states of bc are treated quantum mechanically , while the initial and final translational states as well as the whole dynamics within the interaction region are treated classical mechanically .
this choice makes sense , in particular to qct users accustomed to introduce _ ad - hoc _ corrections in their trajectory calculations , but from a rigorous point of view , it is arbitrary .
these assumptions lead to the following developments .
the final vibrational state of bc can be written as + @xmath24 + for the @xmath25 s form a complete basis .
hence , @xmath22 satisfies + @xmath26 + now , any quantity @xmath27 defined by @xmath28 + is also rigorously given by the phase space overlap + @xmath29 + where @xmath30 , @xmath31 or 2 , is the wigner density defined by + @xmath32 + @xcite .
a pedestrian demonstration of the strict equivalence between eq . and eqs . and is given in appendix b of ref .
consequently , @xmath22 can be rewritten as + @xmath33 + where @xmath34 and @xmath35 are the wigner distributions corresponding to @xmath36 and @xmath37 , respectively . from the well known analytical expression of @xmath36 and eq .
, @xmath34 can be shown to satisfy @xcite + @xmath38 + where @xmath39 is the @xmath40th laguerre polynomial , + @xmath41 and @xmath42 + @xmath43 appears to be the bc vibrational energy .
@xmath35 is estimated as follows .
@xmath37 is supposed to result from the `` propagation '' in time of @xmath12 from the reagents onto the products .
this loose statement leads , however , to a practical method when extended to the wigner densities @xmath35 and @xmath44 .
@xmath44 can indeed be rewritten as + @xmath45 + propagating @xmath44 in time amounts to move each delta peak @xmath46 in the @xmath47 plane along the classical path starting from @xmath48 .
this path is also specified by @xmath49 , supposed to be large enough for bc to vibrate freely , and @xmath50 .
once the trajectories are back to @xmath49 , where bc has reached its final vibrational state , each delta peak is located at a given point of coordinates @xmath51 , both depending implicitely on @xmath48 .
consequently , + @xmath52 + replacing @xmath53 in eq . by the right - hand - side ( rhs ) of the above equation and integrating over @xmath3 and @xmath5 finally leads to the lee - scully expression + @xmath54 + ( see the third identity of eq .
( 21 ) in ref .
@xcite ) . the lee - scully results presented in figs .
[ fig:1 ] and [ fig:2 ] have been obtained from a simple monte - carlo method based on 10@xmath55 trajectories , with @xmath56 and @xmath57 both randomly selected within the range [ -6,6 ] , and @xmath58 taken at 20 .
we now derive a formulation which closely parallels the time - dependent quantum description as far as possible .
only the main lines of the derivation of transition amplitudes in terms of a time - evolved wave - packet are given in this section , for it is a known result @xcite . for clarity s sake , however , more details
are given in appendix a. we consider at time 0 the wave - packet defined by : + @xmath59 with @xmath60^{1/2}e^{-\frac{1}{2}\left[(k - k_i)/\epsilon\right]^2}. \label{3}\ ] ] + @xmath61 is normalized to unity . here , @xmath58 is sufficiently large for the whole wave - packet to lie within the asymptotic region where @xmath9 does not depend on @xmath2 and the vibrational and translational motions are uncoupled .
@xmath62 has a given positive value , thus making the wave - packet moving inward for sufficiently small values of @xmath63 ( propagation is performed forward in time , otherwise stated ) .
the spreading of the wave - packet is inversely proportional to @xmath64 .
let @xmath65 be the state of inelastic scattering between a and bc at energy @xmath13 , with unit incoming flux in channel @xmath1 . in the asymptotic region
, @xmath66 can be written as + @xmath67^{1/2}e^{-ik_n r}\chi_n(r ) + \sum_{n"}\;s_{n"n}\;\left[\frac{\mu}{2\pi\hbar^2 k_{n"}}\right]^{1/2}e^{ik_{n"}r}\chi_{n"}(r ) \label{4}\ ] ] with @xmath68^{1/2}. \label{5}\ ] ] + the @xmath69 s are transition amplitudes , or _ s_-matrix elements .
the projection of @xmath65 on @xmath61 is given by + @xmath70 + the time - evolved wave - packet can then be written as + @xmath71 + we now consider at an infinite time the projection + @xmath72^{1/2}e^{-ik_{n'}r}\chi_{n'}(r)\psi_t(r , r ) \label{8}\ ] ] + of the outgoing free state associated with @xmath13 and @xmath23 onto the time - evolved wave - packet . from eqs . and
- , it is shown in appendix a that + @xmath73^{1/2 } q_{n'n}\;e^{iet/\hbar}. \label{9}\ ] ] + this is an important expression of the time - dependent approach to semiclassical dynamics derived by heller in the mid - seventies @xcite .
. is analogous to eq .
( 4.4 ) in heller s work .
the only difference is that in the present work , @xmath74 ( see eq . ) can be non zero for negative values of @xmath75 .
this implies the kronecker symbol @xmath76 in eq . , not present in eq .
( 4.4 ) of ref .
@xcite .
analogously to what we have seen in section [ iii ] , any quantity @xmath27 defined by + @xmath78 can be rewritten as @xmath79 + where @xmath80 , @xmath31 or 2 , is the wigner density defined by + @xmath81 + @xcite .
the strict equivalence between eq . and eqs . and can be proved by following the reasoning presented in appendix b of ref .
@xcite limited , however , to the case of one configuration space coordinate . in the present case of two spatial coordinates ,
the developments are more tedious , but present no difficulty . setting @xmath82^{1/2}e^{ik_{n'}r}\chi_{n'}(r )
\label{bb.4}\ ] ] and @xmath83 + we arrive from eqs . and at + @xmath84 + where @xmath85 is related to @xmath86 by eq .
, the translational wigner distribution @xmath87 reads + @xmath88\ ; e^{-ik_{n'}(r+s)}\ ; e^{ik_{n'}(r - s ) } \label{bb.7}\ ] ] + and @xmath34 is given by eqs .
- + using eq . , eq .
transforms to + @xmath89\ ; \frac{2}{\hbar}\ \delta[2(p-\hbar k_{n'})/\hbar ] \label{bb.10}\ ] ] or equivalently , @xmath90\theta(p ) , \label{bb.11}\ ] ] + @xmath91 being the function of heaviside .
is indeed readily obtained from eq . by means of theorem .
provides a formally exact quantum expression of @xmath92 .
we now introduce in the formulation the following classical ingredient : we consider @xmath93 as a solution of the liouville equation @xcite instead of the wigner one @xcite , i.e. , we classically propagate it from @xmath94 to @xmath95 . in the framework of this assumption
, we have + @xmath96 + in this identity , @xmath97 should be understood as the dynamical state of abc reached at time @xmath63 when starting from the initial state @xmath98 at time 0 .
expresses the fact that the probability to lie within @xmath99 does not depend on @xmath63 , a property due to the deterministic nature of classical mechanics . using eq .
, on may rewrite eq . as + @xmath100
+ where for clarity s sake , @xmath97 in @xmath101 have been denoted @xmath102 in order to emphasize that these coordinates determine the dynamical state of abc at time @xmath63 . from eqs .
, and , one may show after some mathematical steps presenting no particular difficulty that + @xmath103 + with @xmath104 given by eqs . - . in fact
, @xmath105 depends on the vibrational energy at @xmath63 ( see eqs . - ) .
moreover , this energy is a constant of motion beyond the frontier between the interaction region and the free products , defined by a given value @xmath106 of @xmath2 .
therefore , @xmath107 can be replaced by @xmath108 in eq . , where @xmath109 and @xmath110 are , respectively , the values of @xmath111 and @xmath112 at the previous frontier . for exactly the same reason , @xmath113 can be replaced by @xmath114 in eq . .
we thus have + @xmath115 the validity of eq .
is conditioned by the ability of classical mechanics to propagate @xmath116 in a realistic way , i.e. , such as it stays sufficiently close to the exact time - evolved wigner density , solution of the wigner equation @xcite , up to the products .
in fact , one may even be more restrictive than that for the following reasons .
first , @xmath117 forces the kinetic energy to be equal to @xmath118 , as deduced from eqs . and .
second , @xmath119 turns out to be negligible for states @xmath51 corresponding to vibrational energies larger than @xmath120 , as illustrated in fig .
[ fig:3 ] for @xmath121 .
consequently , the product @xmath122 in eq . limits the integration to phase space states belonging to the energy range [ @xmath118,@xmath123 .
defining @xmath124 by 1 for @xmath125 within the previous range , and 0 for @xmath125 outside it , eq .
can then be rewritten as + @xmath126 + the validity of eq .
is thus conditioned by the ability of classical mechanics to propagate @xmath127 in a realistic way .
the projection of @xmath128 onto the @xmath129 plane is represented by the green cloud in fig .
[ fig:4 ] for @xmath121 , @xmath130 and @xmath131 ( see eq . ) .
the whole density lies beyond @xmath132 units , defining the frontier of the interaction region for the secrest - johnson potential energy surface ( pes ) @xcite ( see eq . ) .
the projection of the classically propagated density when it bounces against the repulsive wall of the pes is given by the red cloud in the same graphic . at this instant ,
the compression of the wave - packet is maximum . due to the small value of @xmath64 ,
the green cloud is very narrow along the @xmath4 direction , and very broad along the @xmath2 direction , by virtue of the uncertainty principle .
the red cloud extends from -10 to 20 units mainly along the @xmath2-axis , which turns out to be the whole interaction region available at the energy @xmath133 .
the projections of @xmath134 and @xmath128 are represented in fig .
[ fig:5 ] for @xmath135 , a much larger value than previously , the remaining parameters being unchanged .
the first projection corresponds to the brown cloud while the second corresponds to the two green ones .
these clouds are now very broad along the @xmath4 direction , and very narrow along the @xmath2 direction , at the opposite of the previous case . as a matter of fact
, multiplying @xmath134 by @xmath124 strongly alters it for large values of @xmath64 . on the other hand ,
the same operation has not effect for @xmath130 , i.e. , projecting @xmath134 onto the @xmath129 plane still leads to the green cloud in fig .
[ fig:4 ] .
[ fig:6 ] is analogous to fig .
[ fig:4 ] for @xmath135 .
the left and right red clouds come from the classical propagation of the lower and upper green clouds , respectively . for @xmath23 different from @xmath1
, the right red cloud will not contribute to @xmath92 .
this cloud is indeed obtained by classically propagating @xmath136 .
now , the previous density contains the factorized term @xmath44 ( see eq . ) which remains unaltered by the outward propagation and is orthogonal to @xmath34 .
the overlap between these two densities in eq .
is thus zero ( note that the classical propagation is exact here , the pes being flat along the @xmath2-axis and quadratic along the @xmath3-axis , satisfying thereby the conditions for which the wigner equation reduces to the liouville one @xcite ) .
@xmath92 is thus only due to the left red cloud , which extends from about -9 to -4 units along the @xmath2-axis .
the distributions of @xmath2 obtained from the bouncing distributions are represented in fig .
[ fig:7 ] . as anticipated from the shape of the red cloud in fig .
[ fig:4 ] and the inner red cloud in fig .
[ fig:6 ] , the spreading of the distribution is comparable to the size of the interaction region for @xmath130 , while it is @xmath137 times smaller than the same region for @xmath135 .
now , it is well known @xcite that the smaller the extention of a wave - packet in the configuration space , the more accurate the classical propagation of its corresponding wigner density .
the basic reason is that in this limit , one may reasonably approximate the pes by a second order development around the wave - packet , which reduces the wigner equation to the liouville one @xcite . as a consequence , large values of @xmath64 , which strongly narrow the @xmath2-extention of @xmath127 within the interaction region ,
are expected to make its classical propagation more realistic , and the prediction of @xmath92 more accurate than small values of @xmath64 . on the other hand , there is no control of the @xmath3-extention of @xmath127 in the present approach .
this makes it better suited to collisions where the reagents are prepared in the lowest excited vibrational states , less spreaded along the @xmath3 coordinate . for @xmath23
equal to @xmath1 , eq
. has less chance to be valid in the limit of large @xmath64 , as the classically propagated density @xmath116 can not take into account the possible interference between the outgoing and incoming parts of the initial wave - packet , corresponding to the upper and lower green clouds in fig .
[ fig:6 ] .
however , setting @xmath138 @xmath92 is found from eq . to satisfy + @xmath139 |q_{n'n}|^2 , \label{25new}\ ] ] + where @xmath140 results from the interference between the first and second terms on the left - hand - side of eq . .
this interference is of the same nature as the previous one , but expressed in a time - independent framework . since eq . can not deal with it
, it makes sense to remove @xmath140 from eq . , thus assuming that the population @xmath22 satisfies + @xmath141 |q_{n'n}|^2 . \label{26new}\ ] ] + an analytical example is considered in appendix b which supports the previous assumption .
note that when @xmath23 is different from @xmath1 , eq .
is exact .
the main finding of the present analysis is that @xmath64 in eq .
has to be taken at a large value in order to maximize the accuracy of the semiclassical wigner method . at the boundary of the interaction region ,
i.e. , at @xmath142 , eq .
becomes + @xmath143 + using eqs .
, and , eq . can be rewritten as + @xmath144 . \label{24new}\ ] ] + @xmath114 being necessarily positive , the term @xmath145 , coming from eq .
, is not necessary to specify in the above integral . in refs .
@xcite , the alternative set of coordinates @xmath146 was used in place of @xmath98 . in these new coordinates ,
the origin of time corresponds to the instant where the system is at @xmath106 .
the pair @xmath51 specifies the internal state of bc at time 0 and @xmath125 forces @xmath114 to take the value + @xmath147^{1/2 } \label{25}\ ] ] + ( see eq . ) .
@xmath148 lies along a given trajectory .
now , any point along this trajectory can be reached from the previous point by moving along the trajectory a given period of time @xmath149 either forward ( @xmath150 ) or backward ( @xmath151 ) . in other words , given @xmath106 , @xmath152 imposes the classical path , and @xmath63 the location along it .
consequently , @xmath146 allows to span the whole phase space .
in addition to that , one may show the important property @xcite + @xmath153 + ( see , in particular , appendix c of ref .
@xcite ) . from eqs . and and the straightforward integration with respect to @xmath125
, we finally arrive at @xmath154 + not forgetting that the delta term in eq .
imposes the condition + @xmath155 + finally , @xmath102 in eq .
is the value of @xmath97 at time @xmath63 when starting from @xmath156 at time 0 ( the meaning of @xmath102 is thus different here and in section [ iv.c ] ) . to summarize , the internal state @xmath51 of bc is randomly chosen within appropriate boundaries .
together with eq . , they allow to generate a trajectory from @xmath106 at time 0 .
the trajectory is then propagated backward in time , i.e. , in the direction of the reagent molecule , and @xmath157 is time - integrated until the trajectory gets back to the separated fragments and @xmath157 is found to be negligible ( in principle , one should also propagate forward in time .
this is useless , however , as @xmath157 is zero for positive values of @xmath63 ) .
the result is multiplied by the statistical weight @xmath119 in order to get the integrand of eq . .
eq . is , however , not in its final form .
it can indeed be analytically simplified by integrating over @xmath63 as follows .
@xmath158 is given by eq .
where we assume that @xmath64 has a large value so as to maximize the accuracy of @xmath92 , as seen in the previous section .
@xmath134 is thus narrow along the @xmath2-axis , stretched along the @xmath4-axis , and extends along the latter in the upper and lower half planes , as illustrated in figs .
[ fig:5 ] and [ fig:6 ] . along the trajectory defined by ( @xmath159 )
, we have + @xmath160 + within the phase space region corresponding to the free fragments with @xmath4 positive ( after the collision ) . moreover , + @xmath161 \label{30a}\ ] ] + within the phase space region corresponding to the free fragments with @xmath4 negative ( before the collision ) .
both @xmath162 and @xmath163 depend on the `` initial conditions '' @xmath51 within the backward picture of the dynamics .
note that @xmath162 has a negative value , since it corresponds to a approaching from bc . from the previous considerations ,
the fact that @xmath134 lies in the phase space part associated with the free fragments , and eq .
, we have + @xmath164 with @xmath165\right)^2 } \label{31a}\ ] ] and @xmath166 + note that , _ stricto sensu _ , the upper time limit in @xmath167 and the lower time limit in @xmath168 can not be @xmath169 and @xmath170 , respectively
. they should be intermediates times , the former being lower than the latter .
however , the region where @xmath134 takes significant values is expected to be crossed very quickly in both directions , justifying the passage to infinity . also , since @xmath171 does only depend on the vibrational energy corresponding to @xmath172 , and
this energy is a constant of motion in the asymptotic channel , @xmath111 and @xmath112 have been replaced in @xmath167 by @xmath56 and @xmath57 , their values at time @xmath63 where @xmath173 .
moreover , they have been replaced in @xmath168 by @xmath109 and @xmath110 . using the fact that @xmath174 + in order to analytically integrate @xmath167 and @xmath168 , and using eqs . , and , we arrive at + @xmath175 with @xmath176 .
\label{32a}\ ] ] + in addition , keeping @xmath64 at a large value so as to maximize the accuracy of the method clearly reduces @xmath177 to @xmath178 . we finally obtain the first key expression of this work : + @xmath179 + in the case of the secrest - johnson model of he+h@xmath0 collinear collision @xcite , eq
. leads to the blue curves in fig .
[ fig:8 ] ( backward - scw results ) .
a simple monte - carlo approach based on 10@xmath180 trajectories for each value of @xmath23 has been used . the agreement with the qm results
is very satisfying .
the norm , i.e. , the sum of final state populations , appears to be very close to 1 for both @xmath181 and @xmath182 .
is actually not the expression most suitable for numerical applications . in order to get @xmath22 ,
one has to run a batch a trajectories backward in time from @xmath106 with @xmath114 given by eq . .
if @xmath23 runs from 0 to @xmath183 , @xmath184 batches of trajectories must be run in order to obtain the whole final state distribution
. however , the microreversibility principle tells us that , within the framework of quantum mechanics , @xmath185 .
hence , if the semiclassical wigner treatment is realistic , one may rewrite @xmath22 as + @xmath186 + eq .
is just eq . with @xmath1 and @xmath23
@xmath114 is now given by + @xmath187 + last but not least , one may arbitrarily interchange the indices @xmath188 and @xmath189 in eq .
, i.e. , call the initial internal conditions @xmath56 and @xmath57 , keep the radial momentum at the negative value + @xmath190 + run from @xmath106 the resulting trajectories forward in time , and call the final conditions @xmath109 , @xmath110 and @xmath114 once the trajectory is back to @xmath106 . the ensemble of trajectories run is exactly the same as the one involved in eq . .
the only difference is now that @xmath114 is positive , contrary to @xmath162 in eq . .
we finally get the second key expression of this work : + @xmath191 + to sum up , the whole final state populations are determined from only one batch of trajectories run from @xmath106 , with @xmath48 randomly chosen within appropriate boundaries and @xmath162 given by eq . .
moreover , the indices @xmath188 and @xmath189 refer to initial and final conditions , respectively , and trajectories are integrated forward in time .
we have thus reconnected with the traditional way of simulating a molecular collision .
. leads to the blue curves in fig .
[ fig:9 ] ( forward - scw results ) .
they have been obtained from 10@xmath55 trajectories , with @xmath56 and @xmath57 both randomly selected within the range [ -6,6 ] .
the agreement with the qm results is still very satisfying for @xmath181 , a bit less for @xmath182 as compared with the backward method .
this illustrates the fact that microreversibility is not strictly satisfied by the semiclassical wigner method .
the norm is again very close to 1 for both @xmath181 and @xmath182 .
one notes that the forward - scw results are in excellent agreement with the lee - scully ones ( see figs .
[ fig:1 ] and [ fig:2 ] ) . the distribution of the ratio @xmath192 , weighted by @xmath193 in order to take into account the values which more contribute to @xmath22 ( see eq . ) ,
is represented in fig .
[ fig:10 ] for @xmath181 and @xmath194 , corresponding to the three largest populations ( see the upper panel in fig . [ fig:1 ] ) .
similar distributions are found for @xmath182 .
they appear to be peaked around 1 .
one may thus approximate @xmath192 by 1 in eq .
, hence , recovering eq . .
the reason why @xmath192 is on average close to 1 in the model of he+h@xmath0 collision considered here is that a large part of the available energy @xmath13 is channeled into the recoil motion . in the limit where the vibrational energy @xmath195 is negligible as compared to @xmath13 ,
. leads to @xmath196 moreover , @xmath197 + with @xmath198 given by eq . .
hence , we have from eqs . and + @xmath199^{1/2}. \label{41}\ ] ] + but if @xmath13 is much larger than the vibrational energies involved in the problem , we also have + @xmath200 + since both @xmath201 and @xmath114 have nearly the same value , their ratio is roughly equal to 1 .
this explains the shape of the distributions displayed in fig .
[ fig:10 ] .
however , this reasoning is no longer valid when @xmath195 may be larger than the recoil energy . in such a case
, there is _ a priori _ no reason for replacing @xmath192 by 1 in eq . .
the same reasoning holds for @xmath202 in eq . .
last but not least , the forward - scw and lee - scully norms obtained from 5.10@xmath55 trajectories ( with @xmath56 and @xmath57 still randomly selected within the range [ -6,6 ] ) have been calculated . for @xmath181 , the norms obtained by means of eq .
are 0.9984(0.9968 ) , 0.9993(0.9985 ) , 0.9995(0.9984 ) and 1.0003(0.9995 ) for @xmath203 10 , 8 , 6 and 3 , respectively . for @xmath182 ,
the analogous numbers are 0.9952(0.9922 ) , 0.9907(0.9881 ) , 0.9918(0.9862 ) and 1.0002(0.9980 ) .
moreover , it has been checked that the number a trajectories run is sufficiently large for guarantying the order of the norms between the two approaches . as can be observed
, the norm seems to be systematically more underestimated by the lee - scully expression than by the forward - scw one .
though the differences between the two sets of results are very small , they nevertheless support the idea that the lee - scully expression is an approximation to the forward - scw expression ( though an excellent one ) .
note that for @xmath204 , the norms are on average equal to @xmath205 0.94 and the previous order is lost .
however , the extention of the initial phase space density along the @xmath3 coordinate ( see eq . ) is larger than for @xmath181 and 1 , thus making the semiclassical wigner predictions less reliable .
about three decades ago , lee and scully proposed a semiclassical wigner treatment of bimolecular collisions @xcite leading to final state populations in surprisingly good agreement with exact quantum ones in the case of the secrest - johnson model of collinear inelastic collision between he and h@xmath0 @xcite .
the aim of the present paper was to provide a derivation from first principles of the previous method , following the quantum description as far as possible .
the derivation combines elements of ( i ) the time - dependent approach to semiclassical dynamics @xcite , ( ii ) the semiclassical wigner treatment of photodissociation dynamics @xcite , the backward description of molecular collisions @xcite and ( iv ) microreversibility .
the lee - scully expression of final state populations ( eq . ) turns out to be the limiting form of a more general expression ( eq . ) when a significant part of the product available energy is channeled into the recoil motion .
the present derivation is focused on collinear inelastic collisions , but defines the main lines of a strategy that could be applicable to realistic three - dimensional bimolecular processes .
from eqs . , and and integration over @xmath3 , we arrive at + @xmath208^{1/2}\delta_{nj}\int\;dk\;g(k)e^{ikr_i}\frac{1}{2\pi}\int\;dr\;e^{i(k_j - k ) r } \\
+ \left[\frac{\mu}{\hbar^2 k_n}\right]^{1/2}s_{nj}{}^*\int\;dk\;g(k)e^{ikr_i}\frac{1}{2\pi}\int\;dr\;e^{-i(k_n+k ) r}. \label{a1}\end{gathered}\ ] ] + since @xmath209 + we get , by integration over @xmath75 , @xmath210^{1/2}\left[\delta_{nj}\;g(k_n)\;e^{ik_nr_i } + s_{nj}{}^*g(-k_n)\;e^{-ik_nr_i}\right ] .
\label{a3}\ ] ] for @xmath63 tending to infinity , the time - evolved wave - packet @xmath86 ( see eq . ) entirely lies in the asymptotic channel and is moving outward .
its overlap with the incoming part of @xmath211 ( see eq . ) is thus zero and eqs . and lead thus to + @xmath212^{1/2}e^{ik_{n"}r}\chi_{n"}(r)e^{-ie't/\hbar}. \label{a4}\ ] ] replacing in the projection @xmath77 ( see eq . )
@xmath86 by the rhs of eq . and
integrating over @xmath3 leads to + @xmath213^{1/2}c_{j'}(e')s_{n'j}(e ' ) \frac{1}{2\pi}\int\;dr\;e^{i[k_{n'}(e')-k_{n'}(e)]r}e^{-ie't/\hbar}. \label{a5}\ ] ] + for clarity s sake , the dependence of @xmath214 and @xmath215 on the energy is explicitly specified in the previous expression . using eq . , eq . becomes + @xmath213^{1/2}c_{j'}(e')s_{n'j}(e ' ) \delta[k_{n'}(e')-k_{n'}(e)]e^{-ie't/\hbar}. \label{a6}\ ] ] applying the theorem @xmath216 = \sum_k\;\frac{1}{|f'(x_k)| } \delta(x - x_k ) , \label{a7}\ ] ] + where the @xmath217 s are solutions of @xmath218 @xcite , we find + @xmath219 + from eqs . and , we arrive at + @xmath220^{1/2 } \left[g(k_n)\;e^{ik_nr_i}\;s_{n'n}+g(-k_n)\;e^{-ik_nr_i}\;\sum_j\;s_{nj}{}^*s_{n'j}\right]e^{-iet/\hbar}. \label{a9}\ ] ] + however , @xmath221 , where @xmath222 is the _ s_-matrix , thus implying + @xmath223 + from eqs . and , we finally arrive at eq . .
in this example , @xmath62 is kept at 0 in eq . , so @xmath224 .
moreover , the potential energy is given by + @xmath225 + for @xmath226 0 , and @xmath169 for @xmath227 .
if so , no inelastic transition is possible and we have + @xmath67^{1/2}\chi_n(r)\left[e^{-ik_n r}+s_{nn}e^{ik_{n}r}\right ] .
\label{b2}\ ] ] + since @xmath65 is 0 for @xmath228 , @xmath229 squaring both sides of eq . leads to + @xmath230 = \left[\frac{\hbar^2 k_n}{\mu}\right ] |q_{nn}|^2 .
\label{b4}\ ] ] + this is eq . for @xmath231 . besides
, we may apply eq .
, together with eqs . and .
in a first step , one can replace @xmath109 and @xmath110 in @xmath232 by @xmath233 and @xmath234 , owing to the fact that the vibrational energy @xmath43 is a constant of motion throughout the whole collision , and @xmath232 does only depend on @xmath43 ( see eqs . - ) .
since + @xmath235 @xcite , we arrive at @xmath236 + as the translational motion is unperturbed by the bouncing , @xmath114 in @xmath237 can be replaced by @xmath238 . using eqs . , and , one gets after integrating over @xmath239 and @xmath238 , + @xmath240 + since @xmath241 , eq .
can be rewritten as + @xmath242 |q_{nn}|^2 , \label{b9}\ ] ] + which is just eq . for @xmath231 .
the semiclassical expectation of @xmath92 is thus solution of eq . , or eq . , with the interference term removed .
@xmath243 he+h@xmath244 at a total energy of 10 and for @xmath181 .
lee - scully predictions , given by eq . , and
qct ones are compared with qm results in the upper and lower panels , respectively .
qm and qct data come from refs . @xcite and @xcite , respectively .
[ fig:1],width=755 ] for the harmonic oscillator ( see eqs . - ) in terms of the vibrational energy @xmath43 ( red curve ) .
on the rhs of the green vertical line , defined by @xmath245 ( see eq . with the parameters of section [ ii ] ) , @xmath246 appears to be negligible .
[ fig:3],width=680 ] onto the @xmath129 plane for @xmath121 , @xmath130 and @xmath131 .
red cloud : projection of the classically propagated previous cloud when it bounces against the repulsive wall of the secrest - johnson pes .
[ fig:4],width=680 ]
|
the semiclassical wigner treatment of bimolecular collisions , proposed by lee and scully on a partly intuitive basis [ j. chem .
phys .
73 , 2238 ( 1980 ) ] , is derived here from first principles .
the derivation combines e. j. heller s ideas [ j. chem .
phys .
62 , 1544 ( 1975 ) ; 65 , 1289 ( 1976 ) ; 75 , 186 ( 1981 ) ] , the backward picture of molecular collisions [ l. bonnet , j. chem . phys . 133 , 174108 ( 2010 ) ] and the microreversibility principle .
|
the baryons within the observed universe appear to consist essentially only of matter and not antimatter .
there are no regions of gamma ray emission which would correspond to the collision between a galaxy made of normal matter with one made from antimatter , for instance .
moreover , the theory of big bang nucleosythesis gives predictions for the abundances of primordial helium ( both @xmath0he and @xmath1he ) , deuterium and lithium , if the ratio of baryons to photons in the universe is in the range @xmath2 if we do live in a universe with equal quantities of matter and antimatter , then the two must somehow be separated on distance scales greater than the present hubble length , that is , the size of the presently observable universe .
the simplest assumption is that indeed the whole universe has the same preponderance of matter over antimatter .
it is a deep mystery of cosmology why the abundance of baryons should be this peculiar number , @xmath3 , relative to the number of photons .
the mystery is intensified by the fact that the most natural initial condition of the universe , at the time of the big bang or shortly thereafter , _ is _ to have equal numbers of baryons and antibaryons , @xmath4 implying that the net baryon number of the universe was zero .
there are several reasons for believing this
baryon ( @xmath5 ) and lepton ( @xmath6 ) conservation are _ accidental _ symmetries of the standard model ( sm ) : there is no good reason for them to be exact .
for example , a dimension 9 operator consisting of 6 right - handed quark fields in a color singlet state , @xmath7 is allowed by the gauge symmetries , and would lead to neutron - antineutron oscillations at some rate suppressed by the large mass scale @xmath8 . at sufficiently high temperatures @xmath9 in the early universe , however , the effects of such a baryon number violating operator would be unsuppressed .
sphalerons _ , present within the sm itself , violate @xmath5 and @xmath6 ( but not @xmath10 ) .
these are the lowest energy field configurations with chern - simons number @xmath11 , intermediate between neighboring @xmath12-vacua of the @xmath13 electroweak gauge theory .
( a sphaleron can be thought of as the @xmath14 slice of an instanton , such as occur in qcd . in qcd
it is chirality , rather than @xmath15 , which is anomalously violated . )
local transitions which go from one @xmath12-vacuum to a neighboring one must pass through a sphaleron - like configuration .
because of the triangle anomaly in the baryon and lepton currents , each such transition is accompanied by 9 quarks and 3 leptons . at zero temperature ,
the energy barrier between the @xmath12-vacua , which is the sphaleron energy , is near 10 tev , and the tunneling rate for the anomalous transition is so slow as to be entirely irrelevant . but at temperatures above that of the electroweak phase transition , @xmath16 gev , this energy barrier disappears , and sphaleron transitions are fast compared to the hubble expansion rate in the early universe .
3 . grand unified theories also violate b and l through heavy gauge boson ( @xmath17 ) vertices of the form @xmath18 or @xmath19 .
the presence of both such interactions prevents one from assigning a conserved @xmath5 or @xmath6 number to the @xmath17 boson . because of these sources of @xmath5 violation
, we would expect that any initial @xmath5 or @xmath6 asymmetry that might have been present initially would be quickly wiped out , giving @xmath20 as the effective initial condition at the high temperatures of the very early universe .
therefore something must have happened between now and then to produce the observed baryon asymmetry .
it was realized by sakharov in 1967 that three things are needed to spontaneously generate the baryon asymmetry : baryon number violating interactions , @xmath21 ( particle - antiparticle symmetry ) violation , and loss of thermal equilibrium for the @xmath5-violating interactions .
the first two features are present in the sm by virtue of sphalerons and the phase in the ckm matrix for the quarks .
this source of @xmath21 violation is however too weakly coupled to the mechanism of baryon production to produce a large enough asymmetry .
the third condition is unfulfilled in the sm model : the transition between the symmetric and the broken phase of the electroweak theory is so continuous that it is not a phase transition at all .
sphaleron interactions , although they eventually drop out of thermal equilibrium , do so too gradually to allow the universe to go from a state of @xmath22 to one of nonzero @xmath5 .
we therefore need new physics beyond the sm for baryogenesis .
many particle physicists believe that supersymmetry is the most natural direction in which to look for such new physics , and i will also take that point of view .
moreover , i will confine my remarks to electroweak baryogenesis , even though much interesting work on nonelectroweak baryogenesis scenarios has been done in the last year@xcite .
electroweak baryogenesis within the minimal supersymmetric standard model ( mssm ) has evolved far beyond the status of being just a rough , qualitative theory . rather , it has been the subject of intense , highly quantitative scrutiny , thanks to the fact that it is so much in the realm of currently testable physics .
the way it must work is rather well - defined .
first , the electroweak phase transition must be first order , which requires the higgs field potential to have a barrier between the symmetric phase minimum at @xmath23 and the true vacuum state with @xmath24 , as shown in figure [ fig1 ] . under these conditions ,
the phase transition proceeds by the nucleation of spherical bubbles containing the new @xmath24 phase within .
the bubbles quickly expand and fill the universe with the broken phase . between the initial nucleation and the completion of the phase transition
, particles in the high-@xmath25 plasma are encountering the expanding bubble walls ( figure [ fig2 ] ) .
most particles are massless outside the bubble and massive within , so the bubble wall behaves like a quantum mechanical potential which can scatter the particles . in general
there is partial reflection and transmission of particles at the wall , with left - handed particles reflecting into right - handed ones , since spin is conserved , and vice versa .
it is in the reflection process that @xmath21 violation is important .
if @xmath21 is violated on the wall , then particles and antiparticles can have unequal reflection probabilities .
the same is true for left - handed and right - handed fermions .
thus an excess of left - handed quarks versus antiquarks can build up in front of the wall ( compensated by an equal and opposite excess of right - handed quarks so that net baryon number is still zero at this point ) .
the asymmetry between @xmath26 and @xmath27 increases the free energy of the plasma locally , and sphalerons try to minimize this energy by destroying the @xmath26-@xmath27 asymmetry to some extent , redistributing it among the other species of quarks and leptons .
this biases the sphalerons to preferentially create an excess of baryons or antibaryons , which resides in front of the wall for a time , but eventually falls inside the bubble and becomes the baryon asymmetry of the universe ( bau ) that is observed today .
however the bau can survive inside the bubbles only if sphaleron interactions are essentially shut off .
this is because the @xmath21 asymmetry is not operative inside the bubbles ( since chirality is not a good quantum number in the broken phase ) to insure that sphalerons act preferentially to make only baryons or antibaryons .
if sphalerons are not turned off , they will erase whatever bau is created .
the above idea is highly constrained .
for example , it does not appear feasible to alter the basic mechanism by replacing bubbles with other field configurations such as cosmic strings , despite the fact that this possibility has been widely discussed in the literature .
we have recently shown @xcite that the cosmic string scenario , when scrutinized more carefully , underproduces the baryon asymmetry by 10 orders of magnitude .
the reason is essentially that @xmath21 violation at the string walls is proportional to the string velocity squared , @xmath28 , while the density of strings in the network scales like @xmath29 .
it was previously assumed that the strength of @xmath21 violation and the string density could be independently varied . however , the most important constraint on electroweak baryogenesis is the condition for making the sphaleron interactions slow enough to preserve the bau , which turns out to be @xmath30 inside the bubbles , where @xmath31 is the the field value that minimzes the potential energy .
although impossible to achieve in the sm , it _ is _ possible in the mssm , provided that the right - handed top squark and the lightest higgs boson are sufficiently light .
( a light left - handed stop is disfavored by precision electroweak considerations , namely the rho or @xmath25 parameter . )
the significance of the top squark is its large yukawa coupling @xmath32 to the higgs field , whose potential @xmath33 controls the phase transition .
top squark loops such as those shown in figure [ fig3 ] contribute cubic terms of the form @xmath34 to @xmath33 , which make it possible to fulfill eq .
( [ eq1 ] ) . this can be understood in terms of the standard expression for the free energy of a relativistic gas of bosons , represented by the one - loop diagram in figure [ fig3 ] , which when expanded in powers of the boson mass over temperature contains a term proportional to @xmath35 .
since the mass of the squark depends on the higgs field , this increases the magnitude of the negative @xmath36 term in @xmath33 .
we have studied the ewpt in the mssm using the two - loop effective potential , and varying the parameters of the mssm randomly over many thousands of values to search for those which give a strong enough phase transition @xcite .
our results are in reasonable agreement with several similar studies @xcite-@xmath37@xcite , in predicting the mass ranges for the higgs boson and the stop to be @xmath38 the lower bound on @xmath39 is from the latest @xmath40 ( lep ) experimental limit @xcite ( the mssm version of this bound is weaker than in the sm ) , and the upper bound is a function of the heavy stop mass , because its radiative corrections to @xmath39 grow logarithmically with @xmath41 . one might hope that the chances of observing such a light stop at the tevatron would be good .
but experimental limits on @xmath42 often depend on other unknown mssm parameters that determine which production channels are open to @xmath43 .
for example , the greatest sensitivity to squarks is when they are produced along with gluinos , but if the gluino mass exceeds @xmath44300 gev then light squarks are difficult to identify .
@xcite similarly , the baryon production mechanism at the bubble wall prefers values of the @xmath45 parameter and wino mass @xmath46 in the range @xmath47 gev , which would inspire hopes for the imminent discovery of the chargino @xcite .
this range of values is in fact excluded by chargino searches for @xmath48 , but not for @xmath49 .
for these reasons , the most foolproof experimental test of electroweak baryogenesis is the search for the light higgs boson .
the remaining window ( [ eq2 ] ) is rapdily being closed by the lep2 run at cern .
it is a pity , though , that lep2 will terminate with a limit of only @xmath50 gev , leaving a small unexcluded window of higgs mass values .
however it is possible that a better theoretical determination of the higgs mass , using the renormalization group , could help to close this window .
one of the interesting possibilities that has been suggested in connection with electroweak baryogenesis with a light stop is that the su(3 ) gauge group of qcd was temporarily broken . in order to get a sufficiently light right - handed stop ,
its bare mass - squared parameter must have been _ negative _ , so that the field - dependent stop mass has the form @xmath51 this makes the symmetric vacuum unstable toward condensation of the stop field in some random direction in color space .
if @xmath52 is sufficiently negative , it can be energetically preferable to have a period of color breaking before the normal electroweak vacuum state takes over . in this case one
would have the sequence @xmath53 of phase transitions , which would change our view of cosmological history in a very interesting way .
for example , the second transition tends to be very strong , which is favorable for baryogenesis .
but there is an energy barrier impeding this second stage of the transition , due to a positive term in the potential @xmath54 whose presence is mandated by supersymmetry .
it could happen that the rate of tunneling from the color - broken to the electroweak phase is so small that it will never happen in the history of the universe .
guy moore has made the following conjecture : if @xmath55 is ever small enough for transition 1 to take place , then the universe gets stuck in the color broken phase and never completes transition 2 .
although preliminary studies of this question have been done , it deserves a more careful treatment .
we have undertaken such a study , by constructing the full two - loop effective potential @xmath56 for the higgs and stop fields , and computing the nucleation rate for the most likely bubbles interpolating between the color - broken and electroweak phases @xcite .
this involves finding the path in the @xmath57 field space along which the bubble evolves , which gives the lowest bubble energy , hence the fastest rate of transitions .
the field equations with boundary conditions @xmath58 and @xmath59 at the respective ends must be solved along this path .
one needs a value of bubble energy over temperature smaller than @xmath60 to get a tunneling rate per unit volume , @xmath61 , that is competitive with the hubble rate per hubble volume , @xmath62 .
that is , @xmath63 must be less than @xmath64 . however , we find that even when all mssm parameters are adjusted to the values that are optimal for tunneling , the exponent @xmath63 is too large by an order of magnitude .
it therefore appears that one must go beyond the mssm in order to make color - breaking a real possibility just before the electroweak phase transition .
although the electroweak phase transition in the mssm is now well understood , the details of how baryons are produced at the bubble wall are still controversial .
this is a highly complex phenomenon involving quantum reflection or classical forces acting on the particles near the wall , while they are simultaneously being scattered by other particles in the plasma .
it is likely that the quantum boltzmann equation will be needed to put this part of the theory on a more rigorous footing . c. kolda and j. march - russell , hep - ph/9802358 ( 1998 ) ; b.a .
campbell , m.k .
gaillard , h. murayama , and k.a .
olive , nucl .
b538 ( 1999 ) 351 ; k. enqvist and j. mcdonald , phys .
lett . 81 ( 1998 ) 3071;e.w .
kolb , a. riotto and i.i .
tkachev , phys .
lett . b423 ( 1998 ) 348;k .
benakli and s. davidson , hep - ph/9810280 ( 1998 ) ; r. brustein and d.h .
oaknin , hep - ph/9809365 ( 1998);r .
brandenberger , i. halperin and a. zhitnitsky , hep - ph/9808471 ( 1998 ) .
cline , j.r .
espinosa , g.d .
moore and a. riotto , hep - ph/9810261 ( 1998 ) , to appear in phys .
d j.m . cline and g.d .
moore , phys .
81 ( 1998 ) 3317
espinosa , nucl .
b475 ( 1996 ) 273 d. bdeker , p. john , m. laine and m.g .
schmidt , nucl .
b497 ( 1997 ) 387 .
j.r . espinosa and b. de
carlos , nucl .
b503 ( 1997 ) 24 m. carena , m. quiros and c.e.m .
wagner , nucl .
b524 , ( 1998 ) 3 .
m. losada , nucl .
b537 ( 1999 ) 3 .
m. laine and k. rummukainen , phys .
lett . 80 ( 1998 ) 5259 ; nucl .
b535 ( 1998 ) 423 .
r. clare , l3 report to lepc ,
12 november 1998 ( unpublished ) cdf collaboration , phys .
lett . 76 ( 1996 ) 2007 .
m. carena , m. quiros , a. riotto , i. vilja and c.e.m .
wagner , nucl .
b503 ( 1997 ) 387 .
cline , m. joyce and k. kainulainen , phys .
b417 ( 1998 ) 79 .
cline , g.d .
moore and g. servant , preprint mcgill-99/01 , hep - ph/9902220 ( 1999 )
|
i give a synopsis , specifically aimed at nonexperts , of some of the recent developments in electroweak baryogenesis .
the focus of the talk is on the present status of electroweak baryogenesis in supersymmetric models , since this is a plausible and realistic possibility that is currently being probed by experimental searches for the higgs boson and the top squark .
the question of whether it is viable to have a period of color - breaking during the electroweak phase transition is also discussed .
-4 in
|
the spin @xmath0 xxz heisenberg chain is arguably one of the most important exactly solvable quantum models of interacting spins in one dimension .
one the one hand , its study has inspired the development of many modern techniques of quantum integrability such as the bethe ansatz @xcite , the quantum - inverse scattering method @xcite or the vertex - operator approach @xcite . on the other hand , it is related to a variety of physically interesting problems , most notably the theory of quantum magnetism @xcite . in this article , we study the open integrable xxz heisenberg chain with arbitrary spin @xmath2 @xcite .
we focus on a particular value of the anisotropy parameter where the spin chain exhibits an additional symmetry beyond its integrability : _ supersymmetry _ @xcite . for concreteness ,
let us consider the familiar case @xmath3 . for diagonal boundary fields
the spin - chain hamiltonian of a chain of length @xmath4 is given by @xmath5 where @xmath6 and @xmath7 denote the pauli matrices , @xmath8 the anisotropy parameter , and @xmath9 the boundary magnetic fields .
fendley and yang @xcite showed that for @xmath10[eqn : openxxzintro ] the hamiltonian is supersymmetric : up to a constant shift ( that we specify later ) it can be written as the anticommutator of a _
supercharge _ and its adjoint .
this supercharge is a nilpotent operator that maps states of a chain of length @xmath4 to states of a chain of length @xmath11 : it is dynamic .
dynamic supersymmetry on the lattice has since been observed for many other spin chains @xcite , in particular for the periodic and twisted spin @xmath2 xxz heisenberg chains with a particular value of its anisotropy parameter that depends on @xmath12 @xcite . in the following
, we refer to this value as the supersymmetric point . in this article
, we determine a multi - parameter family of boundary magnetic fields for the open integrable xxz heisenberg chains with spin @xmath2 at the supersymmetric point that are compatible with a dynamic lattice supersymmetry .
these boundary terms generically are non - diagonal and may differ at the first and last site of the chain .
we achieve this through a generalisation of the supercharges found in previous works .
furthermore , we identify all values of the parameters for which the spin - chain hamiltonians possess so - called supersymmetry singlets @xcite .
supersymmetry singlets are special eigenstates of the hamiltonian that are annihilated by both the supercharge and its adjoint .
if they exist then they are automatically ground states and therefore of great physical interest .
their existence is related to the existence of a non - trivial cohomology of the supercharge .
we explicitly compute this cohomology .
we use our cohomology results in order to determine sum rules for special scalar products involving the supersymmetry singlets .
specifically , let us denote by @xmath13 a singlet for a spin chain of length @xmath4 .
we consider the overlaps @xmath14 these scalar products and their scaling limit are of interest in the field of quantum quenches and quantum entanglement . in particular , the case @xmath15 is related to an entanglement measure called the _
bipartite fidelity _ @xcite . for one - dimensional quantum critical systems
, this quantity has a large-@xmath4 asymptotic expansion whose first few terms have been predicted by conformal field theory ( cft ) techniques @xcite .
these cft predictions have been confirmed by lattice derivations at the leading order in a few cases @xcite .
here , we use the supersymmetry to show that the scalar product in can ( in a suitable normalisation ) be computed from the sole knowledge of a single special component of each involved singlet . for the hamiltonian , we provide exact finite - size expressions of these components and therefore an explicit formula for the scalar products as a function of @xmath16 .
this allows us to exactly compute the large-@xmath4 expansions of the scalar products and show that they match the cft predictions both at leading _ and _ subleading orders .
the layout of this article is as follows . in [ sec : susyspinchains ] we review the formalism of dynamic lattice supersymmetry for open quantum spin chains .
in particular , we discuss the supercharge of the open integrable xxz heisenberg chain with spin @xmath2 at its supersymmetric point and particular diagonal boundary interactions .
we present a new multi - parameter deformation of this supercharge in [ sec : deformedsusy ] .
the deformation allows us to identify a family of non - diagonal boundary interactions that is compatible with an exact lattice supersymmetry .
the purpose of [ sec : e0states ] is to determine the values of the deformation parameters for which the supercharges possess supersymmetry singlets .
we achieve this by exactly computing the cohomology of the supercharge . in [ sec : sp ] , we analyse a number of properties of these supersymmetry singlets . in particular , we find sum rules for the scalar products .
we analyse them for the supersymmetry singlets of the hamiltonian in [ sec : spxxz ] . in particular , we compute their scaling limit and compare our findings to the predictions of conformal field theory .
we present our conclusions in [ sec : conclusion ] .
in this section , we recall the concept and formalism of dynamic lattice supersymmetry for spin chains @xcite .
furthermore , we discuss the supercharges for the open integrable xxz spin chains with spin @xmath2 at their supersymmetric point that we analyse and generalise in this article .
[ [ supercharge - and - hamiltonian . ] ] supercharge and hamiltonian .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + throughout this article , we consider open quantum spin chains of finite length .
we denote by @xmath17 the hilbert space of a spin chain of length @xmath4 .
it is given by the @xmath4-fold tensor product of single - spin hilbert spaces @xmath18 : @xmath19 we focus on the case @xmath20 where @xmath12 is an arbitrary fixed positive integer .
we refer to @xcite for a more general discussion of models where @xmath18 is a super vector space . the supercharge @xmath21 of our spin - chain models is a length - increasing operator that maps @xmath17 to @xmath22 for each @xmath23 . to @xmath17
was indicated by a subscript : @xmath24 .
we omit these subscripts for the supercharges ( and all other operators ) to keep the notation as simple as possible .
if needed , we write @xmath25 in order to emphasise that we consider the action of the supercharge on @xmath17 . ] when acting on @xmath17 , it is given by the following alternating sum : @xmath26 here , the length - increasing operators @xmath27 are @xmath28 where @xmath29 denotes the so - called _
local supercharge_. one checks that if @xmath30 obeys the relation @xmath31 then the operator @xmath21 is nilpotent , @xmath32 in the sense that the action of @xmath33 yields zero on any element of @xmath17 , for each @xmath23 .
we refer to as the coassociativity property .
the canonical ( complex ) scalar product of the spin - chain hilbert space allows us to define the adjoint supercharge @xmath34 .
it is a length - decreasing operator that maps @xmath17 to @xmath35 for each @xmath36 .
we have @xmath37 for all @xmath38 and each @xmath36 .
it follows from that the adjoint supercharge is also nilpotent , @xmath39 this means that for each @xmath40 , the application @xmath41 yields zero on every element of @xmath17 . in supersymmetric quantum mechanics ,
the hamiltonian is given by the anticommutator of a supercharge @xmath21 and its adjoint @xmath34 : @xmath42 using the specific supercharge we find that the hamiltonian is the sum of a bulk part , describing nearest - neighbour interactions , and boundary terms : @xmath43 here , @xmath44 denotes the hamiltonian density @xmath45 , acting on the sites @xmath46 and @xmath47 . in terms of the local supercharge
it is given by @xmath48 furthermore , the boundary interaction at the first and last site of the chain is encoded in the operator @xmath49 . in terms of the local supercharge
, we find @xmath50 in the boundary terms are the same at both ends of the spin chain . in [ sec : deformedsusy ] , we show for a specific choice of @xmath30 that the definition supercharge can be generalised in order to incorporate unequal boundary terms at both ends without modifying the bulk part . unlike @xmath21 and
@xmath34 the hamiltonian @xmath51 is not a length - changing operator . from and follow the following relations : @xmath52 hence , both @xmath21 and @xmath34 are formally symmetry operators , and therefore the hamiltonian @xmath51 is said to be supersymmetric .
the supersymmetry is dynamic in the sense that the hamiltonians on the left- and right - hand side of act on the hilbert spaces of spin chains whose lengths differ by one .
this supersymmetry leads to special properties of the spectrum of @xmath51 .
the construction implies that this spectrum is real and non - negative .
furthermore , it follows from the commutation relations that the spectra for chains of different length have common eigenvalues .
we discuss these properties in detail below in [ sec : cohomology ] .
[ [ local - supercharges - for - xxz - chains - with - arbitrary - spin . ] ] local supercharges for xxz chains with arbitrary spin .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + the work of this article is based on a particular local supercharge .
it acts on the canonical basis states @xmath53 of @xmath18 according to [ eqn : trigsupercharge ] @xmath54 where we abbreviated @xmath55 .
the coefficients @xmath56 are strictly positive real numbers defined for @xmath57 .
they are given by @xmath58 where @xmath59 is the root of unity @xmath60 it follows that the adjoint local supercharge @xmath61 acts on a basis vector of @xmath62 according to @xmath63 the local supercharge was introduced in @xcite .
it is closely related to the supercharge of the so - called @xmath64 models of fendley , nienhuis and schoutens @xcite .
these models describe supersymmetric fermions on a one - dimensional lattice with an exclusion constraint that limits the length of connected fermion clusters to @xmath12 .
their supercharge splits connected clusters of @xmath65 fermions into pairs of adjacent clusters of @xmath66 and @xmath67 fermions , for @xmath68 , with an amplitude @xmath56 .
locally , the @xmath64 models are equivalent to the spin chains considered in this article
. however , the spin - chain language easily allows us to generalise the supersymmetry and add new features that are rather difficult to implement in the fermion language .
one such feature is particle - hole symmetry for the fermions .
in fact , the particule - hole transformation is a complicated non - local operation because of the exclusion constraint . in the spin - chain language
, it translates to a simple spin - reversal symmetry .
indeed , the special values for the constants @xmath56 in lead to a hamiltonian density that is invariant under a spin - reversal transformation as we shall see below .
another feature is the introduction of boundary interactions that break the particle - number conservation in the fermion model ( without breaking the supersymmetry ) . in the spin - chain language
, this translates to boundary conditions that break the conservation of the magnetisation .
below , we construct a new family of such boundary interactions for the spin - chain models by deforming the action of the supercharge on the boundary sites of the spin chain .
let us now illustrate the nature of the spin - chain hamiltonians resulting from . for @xmath3
, defines a local supercharge for the well - known spin @xmath0 xxz chain with anisotropy parameter @xmath1 , found by fendley and yang @xcite : @xmath69 the corresponding hamiltonian density and the boundary terms are readily evaluated from and .
the full spin - chain hamiltonian is obtained from .
identifying @xmath70 it can conveniently be written in terms of the pauli matrices @xmath71 indeed , one finds the xxz hamiltonian with anisotropy parameter @xmath1 and diagonal boundary magnetic fields : @xmath72 up to a constant shift , this is the hamiltonian that we discussed in [ sec : introduction ] . besides its lattice supersymmetry and its relation to the ( open ) @xmath73 model of supersymmetric fermions @xcite
, it has a few interesting features .
first , it is isospectral to another xxz chain that describes the so - called temperley - lieb stochastic process @xcite .
the connection to a stochastic process implies in particular that the spectrum of is non - negative and contains the non - degenerate ground - state eigenvalue @xmath74 . in [ sec : e0states ] , we use the supersymmetry in order to provide new insights into the properties of the corresponding zero - energy state .
second , we note that belongs to a family of hamiltonians that have a spectral overlap with the hamiltonians of the famous quantum - group invariant xxz spin chains of pasquier and saleur s @xcite . this connection has recently been used in order to proof the reality of the spectra of the pasquier - saleur spin chains @xcite . for arbitrary @xmath12 , the hamiltonian density that derives from
was explicitly computed in @xcite .
we write it as follows : [ eqn : explicithamdensity ] @xmath75 where we abbreviate @xmath76 and @xmath77 .
the coefficients have the symmetry property @xmath78 . for @xmath79 , they are given by @xmath80 furthermore , we have @xmath81 furtermore , the term that describes the boundary interactions is diagonal for any @xmath12 : @xmath82 for @xmath83 , yields a hamiltonian density which coincides , up to a simple unitary transformation , with the hamiltonian density of the integrable spin - one xxz chain ( of fateev and zamolodchikov @xcite ) at its supersymmetric point
. further investigations in @xcite support the conjecture that for arbitrary @xmath12 is , up to a simple unitary transformation , the hamiltonian density of the quantum integrable spin @xmath2 xxz chain at a particular value of its anisotropy parameter .
this observation is consistent with an analysis of the bethe - ansatz equations @xcite but remains to be proven . a possible proof could be obtained from mangazeev s explicit expressions for the @xmath84-matrices of fused vertex models @xcite .
we leave the details of this proof to future investigations .
in this section , we generalise the supersymmetry of the open xxz spin chain at @xmath1 , described by the hamiltonian , and its higher - spin analogues at their supersymmetric point . in [ sec : oneparamdef ] we find a family of local supercharges @xmath85 , depending non - trivially on a complex parameter @xmath86 , that have the same hamiltonian density as @xmath30 defined in .
the resulting boundary terms however depend non - trivially on @xmath86 and generically are non - diagonal . using the properties of @xmath85
, we generalise in [ sec : boundaryconditions ] the action of the supercharge at the first and last site of the spin chain .
the resulting supersymmetric spin - chain hamiltonians have unequal boundary terms at both ends of the chain . in order to construct @xmath85
, we need to discuss two special local supercharges : the image @xmath87 of @xmath30 defined in under spin reversal and a so - called local gauge supercharge .
we start our discussion with @xmath87 and some of its properties .
the spin - reversal operator @xmath84 on @xmath88 is a linear operator defined by the following action on the canonical basis states @xmath89 : @xmath90 we define @xmath91 .
this operator acts on the basis vectors of @xmath18 according to @xmath92 it follows from @xmath93 that @xmath87 has the coassociativity property .
furthermore , it was shown in @xcite that , for the specific choice of the coefficients @xmath56 , the hamiltonian densities @xmath94 of @xmath30 and @xmath95 of @xmath87 are equal : @xmath96 for our construction of @xmath85 , we note furthermore that @xmath30 and @xmath87 obey a certain anticommutation relation up to boundary terms .
indeed , for any @xmath97 we have the relation @xmath98 where the vector @xmath99 is given by @xmath100 next , let us recall the concept of a local gauge supercharge @xmath101 @xcite .
it depends on a vector @xmath102 and acts on any vector @xmath97 according to @xmath103 the operator @xmath101 does not have the coassociativity property .
however , it obeys a similar relation:@xmath104 we define the local supercharge @xmath85 as a linear combination of @xmath30 , @xmath87 and a local gauge supercharge . the idea is to adjust the gauge term in such a way that @xmath85 has the coassociativity property .
let us write @xmath105 where @xmath106 is the vector characterising the gauge term . using , and one checks that @xmath85 satisfies if @xmath107 solves the quadratic equation @xmath108 this equation is solved by other solutions are given by @xmath109 with @xmath110 . ]
@xmath111 the equations and define the local supercharge @xmath85 .
it has the property that @xmath112 and @xmath113 where the limit is taken along the real axis .
the construction implies that @xmath85 has a well - defined transformation property under spin reversal .
indeed , one checks that @xmath114 for non - zero values of @xmath86 .
we note that this is compatible with the values taken at @xmath115 and for @xmath116 .
furthermore , this equation shows that the local supercharge @xmath85 is invariant under spin reversal up to a sign if and only if @xmath117 . [ [ hamiltonian - density . ] ] hamiltonian density .
+ + + + + + + + + + + + + + + + + + + + we will now show that the hamiltonian density of @xmath85 is independent of @xmath86 . to this end
, we use a result for local supercharges that are invariant under a parity transformation .
the parity operator @xmath118 on @xmath88 is the linear operator defined by the following action on the canonical basis states : @xmath119 a local supercharge @xmath30 is parity - invariant if @xmath120 on @xmath18 .
the following property of parity - invariant local supercharges was shown in @xcite : [ lem : parityproperty ] let @xmath30 be a parity - invariant local supercharge and @xmath101 any gauge local supercharge , then the hamiltonian densities of @xmath30 and @xmath121 are equal .
one checks that the local supercharges @xmath30 and @xmath87 defined in and , respectively , are parity - invariant .
we use this observation to prove the following statement : [ prop : hamdensity ] the hamiltonian density @xmath122 of the local supercharge @xmath85 defined by , is independent of the parameter @xmath86 .
it follows from [ lem : parityproperty ] that @xmath122 is equal to the hamiltonian density of @xmath123 .
this hamiltonian density is obtained from . in order to evaluate it , we note that @xmath30 and @xmath87 obey the following relations [ eqn : qqbarrelations ] @xmath124 these relations follow from a straightforward calculation , using and . combining with , we obtain @xmath125 where @xmath94 and @xmath95 denote the hamiltonian densities of @xmath30 and @xmath87 , respectively .
furthermore , @xmath126 as noted above . using the value of @xmath127 given in , it follows that @xmath128 , which is independent of @xmath86 .
we write @xmath129 for the supercharge constructed from @xmath85 .
the corresponding hamiltonian @xmath130 is of the form . in the following ,
we will often write @xmath51 for @xmath131 . [ [ magnetisation . ] ] magnetisation .
+ + + + + + + + + + + + + + we define the magnetisation operator by the following action on the basis states of @xmath17 : @xmath132 from , it follows that @xmath133 in particular , for @xmath134 we have @xmath135 .
using , we find that the hamiltonian density @xmath94 of @xmath30 conserves the magnetisation , @xmath136=0 $ ] .
since @xmath128 for any @xmath86 according to [ prop : hamdensity ] , we conclude that the bulk part of the hamiltonian @xmath137 conserves the magnetisation . for @xmath115 ,
the boundary interactions are given by , which obviously has the property @xmath138=0 $ ] .
this implies that for @xmath115 , the full hamiltonian conserves the magnetisation : @xmath139=0 $ ] .
conversely , for generic values of @xmath86 this conservation law is broken by the boundary terms , which are given by @xmath140 we have @xmath141 .
it is possible albeit tedious to explicitly compute the matrix elements of this operator with respect to the canonical basis for arbitrary @xmath12 .
their evaluation shows that the dependence on @xmath86 is non - trivial for both diagonal and off - diagonal matrix elements .
[ [ parity - and - spin - reversal . ] ] parity and spin reversal .
+ + + + + + + + + + + + + + + + + + + + + + + + + the hamiltonian density @xmath128 is parity - invariant .
indeed , we have @xmath142 on @xmath62 as a consequence of the parity - invariance of the local supercharge @xmath30 .
since the boundary interactions are the same at both ends of the chain , we conclude that for any value of @xmath86 the hamiltonian of the spin chain is parity - invariant : @xmath143 the transformation property of the local supercharge @xmath85 under spin reversal imply that the hamiltonian density @xmath128 is spin - reversal invariant : we have @xmath144 on @xmath62 . because of the boundary terms , this spin - reversal invariance does however not extend to the hamiltonian of the spin chain for generic values of @xmath86 .
we have @xmath145 and therefore spin - reversal invariance if and only if @xmath146 .
[ [ example . ] ] example .
+ + + + + + + + as an example , we discuss the case @xmath3 .
the action of @xmath85 on the basis vectors @xmath147 is given by . ]
@xmath148 one checks the independence of the hamiltonian density on @xmath86 by an explicit calculation . in order to write the boundary terms @xmath149 in a convenient way , we set @xmath150 and express @xmath149 in terms of the identity matrix and the pauli matrices .
we find [ eqn : boundaryhamxxz ] @xmath151 with @xmath152 we conclude that the open xxz chain at @xmath1 is supersymmetric for the family of boundary interactions , parametrised by @xmath153 and @xmath154 .
for @xmath155 the off - diagonal terms of @xmath156 are non - zero and therefore generalise the diagonal boundary interactions of found by fendley and yang @xcite . in this section ,
we modify the action of @xmath157 defined in on the first and last site of the spin chain .
this allows us to show that the lattice supersymmetry can be present for unequal boundary terms at both ends of the spin chain .
both these boundary terms depend on the parameter @xmath86 .
furthermore , each boundary term is individually characterised by an integer label @xmath158 .
the main ingredient of our construction are the vectors @xmath159 where @xmath127 is defined in and @xmath107 in .
because of @xmath160 , we trivially have @xmath161 .
the action of @xmath85 on these vectors is very simple .
indeed , using it is not difficult to show that @xmath162 for any pair @xmath163 , we consider an operator @xmath164 that acts on any @xmath165 according to @xmath166 we note that @xmath167 . using @xmath168 and , one checks that @xmath169 the corresponding hamiltonian @xmath170
is readily evaluated .
it is given by a sum of nearest - neighbour interactions and boundary terms that depend on @xmath171 and @xmath66 : @xmath172 here @xmath94 is the hamiltonian density of @xmath85 and the boundary terms are given by @xmath173 we conclude that all boundary conditions that result from the deformation of the local supercharge and a modification of the action of the supercharge on the first and last site of the spin chain are parameterised by a complex number @xmath86 and two integers @xmath174 .
however , not all choices of these parameters lead to unequal spectra . to see this
, we note that the spectrum of @xmath175 is the same as the spectrum of @xmath176 ( for any real value of @xmath154 ) and @xmath177 where @xmath178 .
an appropriate choice for @xmath154 allows us to conclude that it is sufficient to restrict the parameters to real values for @xmath86 , @xmath179 and @xmath180 .
for example , if @xmath3 there are two distinct cases @xmath181 and @xmath182 for any real value of @xmath86 .
in this section , we analyse whether the family of hamiltonians @xmath175 possesses so - called supersymmetry singlets or zero - energy states . if they exist then the zero - energy states are the ground states of the hamiltonian .
in [ sec : cohomology ] , we recall a few basic facts about the spectrum of supersymmetric hamiltonians . moreover , we explain the relation between zero - energy states and the so - called cohomology of the supercharge .
as we shall see , the structure of the cohomology depends on whether the parameter @xmath86 is non - zero or zero .
we compute the cohomology for @xmath183 in [ sec : genericy ] and for @xmath115 in [ sec : y0 ] .
let us recall the characteristics of the spectrum and the eigenstates of a supersymmetric hamiltonian @xmath184 for a generic supercharge @xmath21 @xcite . clearly , @xmath51 is a hermitian operator and therefore diagonalisable .
furthermore , its spectrum is non - negative . indeed
, the schrdinger equation @xmath185 implies @xmath186 since the left - hand side is non - negative and the norm of an eigenvector @xmath187 non - vanishing , we must have @xmath188 .
we call the non - zero solutions of the schrdinger equation with @xmath189 and @xmath74 positive - energy states and zero - energy states , respectively .
they differ in their behaviour under the action of the supercharge and its adjoint .
[ [ positive - energy - states . ] ] positive - energy states .
+ + + + + + + + + + + + + + + + + + + + + + + the eigenstates with strictly positive energy @xmath189 organise in _ doublets_. they are given by a pair of non - zero vectors @xmath190 with @xmath191 .
the two states in the doublet are called superpartners . since the hamiltonian commutes with the supercharge , the superpartners have the same eigenvalue @xmath192 .
in our case , the supercharge increases the length of the chain .
hence , the supersymmetry leads to spectral degeneracies for chains with lengths differing by one .
[ fig : spectraxxz ] illustrates this spectral degeneracy for the hamiltonians @xmath175 with @xmath3 and unequal boundary conditions at both ends of the chain . for @xmath3 with boundary conditions labelled by @xmath181 for @xmath193 ( left panel ) and @xmath194 sites ( right panel ) as a function of @xmath195 .
the solid lines correspond to exact common eigenvalues in the two spectra.,title="fig : " ] for @xmath3 with boundary conditions labelled by @xmath181 for @xmath193 ( left panel ) and @xmath194 sites ( right panel ) as a function of @xmath195 .
the solid lines correspond to exact common eigenvalues in the two spectra.,title="fig : " ] [ [ zero - energy - states . ] ] zero - energy states .
+ + + + + + + + + + + + + + + + + + + in the following , we focus on the solutions of the schrdinger equation with @xmath74 .
since the spectrum of the hamiltonian is non - negative , any non - zero solution of @xmath196 is automatically a ground state of the system .
the existence of these zero - energy states is , according to , equivalent to the existence of non - zero solutions of the system of equations @xmath197 these equations imply that the zero - energy states are _ singlets _ ( as stated in the introduction ) in the sense that no other eigenstates of the hamiltonian can be obtained by acting on them with the supercharge or its adjoint .
the first equation of requires that a zero - energy state be in the kernel of the supercharge .
we call the elements of @xmath198 _ cocycles_. since @xmath199 ,
the kernel contains all states that are in the image of @xmath21 .
we call the elements of @xmath200 _ coboundaries_. the second equation of leads to the following property of zero - energy states : [ lem : e0notcoboundary ] a zero - energy state is not a coboundary .
let @xmath187 be a zero - energy state and assume that @xmath201 for some vector @xmath202 .
then its square norm is @xmath203 where we used .
hence , @xmath204 , which is a contradiction . [
[ cohomology . ] ] cohomology .
+ + + + + + + + + + + [ lem : e0notcoboundary ] suggests that we consider the kernel of the supercharge modulo its image in order to analyse the existence and the properties of the zero - energy states .
we define therefore the _ cohomology _ of the supercharge @xmath205 where @xmath206 and @xmath207 the elements of @xmath208 are equivalence classes or cohomology classes .
any cohomology class of @xmath208 can be represented by a cocycle @xmath209 , which is called _ its _ _ representative_. any two cocycles of @xmath17 differing by a coboundary represent the same element of @xmath208 .
we denote the equivalence class of a cocycle @xmath165 by @xmath210 $ ] .
hence @xmath211 = [ |\psi\rangle]$ ] for all
if @xmath213 for each @xmath23 then we call the cohomology trivial .
this is the case if and only if all cocycles are coboundaries .
it can be shown @xcite that @xmath208 is isomorphic to the subspace of @xmath17 that is spanned by the zero - energy states of the hamiltonian @xmath51 for a chain of length @xmath4 . furthermore ,
if @xmath214 is a representative of a non - zero element of @xmath208 , then there is a state @xmath215 such that @xmath216 is a zero - energy state @xcite . in the following , we use this connection with cohomology classes and their representatives to investigate some properties of the zero - energy states of our models . in this and the following subsection
, we explicitly compute @xmath217 .
this computation allows us to characterise the space of zero - energy states of the hamiltonian @xmath175 as a function of the parameter @xmath86 , the integer labels @xmath218 and the system size @xmath4 . here
, we consider the case @xmath183 .
we prove the following theorem : [ thm : cohomgenericy ] for @xmath183 and each @xmath219 , the cohomology @xmath220 is trivial .
this theorem implies that for @xmath183 and any length of the chain @xmath4 the hamiltonian @xmath175 does not possess zero - energy states .
thus its spectrum is strictly positive . the proof is based on two lemmas . the first lemma deals with a mapping @xmath221 that is akin to a so - called _ contracting homotopy _ @xcite .
[ lem : contractinghomotopy ] let @xmath21 be an arbitrary supercharge .
suppose that for each @xmath36 there is a mapping @xmath222 such that @xmath223 then for each @xmath36 , we have @xmath224 we show that any cocycle @xmath165 is a coboundary .
indeed , applying to @xmath187 we obtain @xmath225 hence , @xmath226 .
our aim is to construct such a mapping @xmath221 for the supercharge @xmath164 . to this end
, we use the vectors @xmath227 , defined in .
the second lemma needed for our proof of [ thm : cohomgenericy ] establishes that for non - vanishing @xmath86 these vectors span the hilbert space @xmath18 of a single spin : [ lem : basisvectors ] for @xmath183 the vectors @xmath227 constitute a basis of @xmath18 .
the components of @xmath228 with respect to the canonical basis are given by @xmath229 to prove the lemma , it is sufficient to show that the matrix @xmath230 is invertible .
this is indeed the case .
one checks that the entries of the inverse matrix are given by @xmath231 this ends the proof .
we now prove that if @xmath183 then @xmath232 for each @xmath23 and each @xmath219 . for @xmath233 ,
the proof is trivial : one readily checks that @xmath234 , using [ lem : basisvectors ] .
hence , we focus on @xmath36 . the proof is based on the construction of a mapping @xmath235 that obeys for each @xmath158 .
we separately consider the cases @xmath236 and @xmath237 .
let us first consider @xmath236 .
it follows from [ lem : basisvectors ] and standard properties of the tensor product that for @xmath183 , every vector @xmath209 can be written as @xmath238 with unique vectors @xmath239 .
we define the mapping @xmath235 by @xmath240 using action of the local supercharge on the special basis vectors , it is easy to see that @xmath241 .
we combine this identity with the definition of the supercharge and find that @xmath242 for each @xmath243 and @xmath244 .
second , for @xmath237 , we define @xmath245 using the definition of @xmath235 , it is easy to see that holds for @xmath237 and @xmath244 , too . in both cases
, it follows from [ lem : contractinghomotopy ] @xmath232 for any @xmath36 and each @xmath219 .
this ends the proof of the theorem .
we notice that the proof only relies on the existence of a basis @xmath227 of @xmath18 with the property @xmath246 for each @xmath110 .
this is the case for a variety of other physically - relevant spin chains .
an example is the quantum spin @xmath0 xyz chain along a special line of couplings @xcite .
the proof of [ thm : cohomgenericy ] is not generalisable to the cases @xmath115 and @xmath247 .
they have to be separately treated . in this subsection
, we consider the case @xmath248 .
for this value of @xmath86 , the local supercharge reduces to @xmath249 , defined in .
the supercharge @xmath250 is independent of the indices @xmath218 and we simply denote it by @xmath21 . its cohomology is non - trivial .
the results presented here below can easily be modified in order to cover the case @xmath116 . in this case
, the local supercharge reduces to @xmath251 , which is the image of @xmath30 under spin reversal .
in fact , the representatives of @xmath252 are simply obtained by applying the spin - reversal operator to the representatives of @xmath208 . the main result of this subsection is the following theorem : [ thm : occohom ] @xmath253 is spanned by the cohomology class of the state by @xmath254 in order to simplify the notation .
] [ eqn : repe0states ] @xmath255 and @xmath256 where @xmath99 is the state defined in and @xmath257 a positive integer .
this theorem implies that for @xmath115 the spin - chain hamiltonian possesses a zero - energy state for each length @xmath4 .
the state is unique up to normalisation .
this result is expected from the above - mentioned mapping between the spin - chain models at @xmath115 and the @xmath64 models of supersymmetric fermions on open intervals . for these fermion models ,
the dimension of the space of zero - energy states was indeed shown to be one - dimensional in previous works @xcite . to our knowledge ,
the structure of the corresponding cohomology has however so far remained undetermined and unexploited .
[ thm : occohom ] provides this structure for the spin - chain models .
we use it in the forthcoming sections in order to compute certain non - trivial scalar products that involve the zero - energy states .
the proof of [ thm : occohom ] is based on several lemmas .
they establish the existence of an explicit bijection between @xmath258 and @xmath259 for each @xmath23 .
hence , we may construct @xmath258 from @xmath260 and @xmath261 .
we explicitly compute them in the following lemma : [ lem : lowdim ] @xmath262 and @xmath263 are spanned by the cohomology classes of the states @xmath264 and @xmath265 , respectively . for @xmath233
, recall that @xmath266 . according to ,
the only solution to @xmath267 is @xmath268 , up to a factor . for @xmath269
, we consider a cocycle @xmath270 .
we write @xmath271 for its components with respect to the canonical basis of the hilbert space . from @xmath272
, it follows that @xmath273 and @xmath274 we distinguish three cases . _ ( i ) _ for @xmath275 , we find from the relation @xmath276 from , it follows that @xmath277 . _
( ii ) _ for @xmath278 , the relation still holds .
the explicit form of the coefficients @xmath56 , defined in , leads to @xmath279 where @xmath280 are the components of the vector @xmath265 defined in . _
( iii ) _ for @xmath281 , we obtain from the relation @xmath282 combining _
( i ) _ , _ ( ii ) _ and _ ( iii ) _ , we find after some algebra @xmath283 hence @xmath210=\psi_{0,\ell}[|{\protect\raisebox{0.25ex}{$\chi$}}\rangle]$ ] with an arbitrary coefficient @xmath284 .
the cohomology class of @xmath265 can not be zero since this state is a linear combination of basis states of @xmath62 which are clearly not in the image of @xmath30 .
our next aim is to study @xmath253 for @xmath40 . in the following technical lemma ,
we determine a convenient choice of their representatives .
[ lem : technical ] for each @xmath285 any element in @xmath253 can be represented by a cocycle @xmath165 with @xmath286 such that @xmath287 for some vector @xmath288 .
let us a consider the cocycle @xmath289 representing an element of @xmath208 .
then for any @xmath290 the vector @xmath291 is also a cocycle , representing the same element of @xmath208 .
we write the vector @xmath187 ( and likewise @xmath289 , @xmath202 ) as a superposition @xmath292 this leads to @xmath293 in order to prove the lemma , we consider the case @xmath294 : @xmath295 again , we decompose the vector @xmath296 with respect to the last site , @xmath297 , and substitute this decomposition into .
the choices @xmath298 lead to @xmath299 .
this ends the proof . for the next two lemmas ,
we introduce the operator @xmath300 which acts on any vector @xmath301 according to @xmath302 one checks that it commutes with the supercharge @xmath21 , @xmath303 because @xmath265 is annihilated by the supercharge .
it follows that @xmath300 can be extended to a mapping @xmath304 defined on the cohomology @xcite : its action on the cohomology classes is given by @xmath305 = [ s|\psi\rangle]$ ] .
[ lem : surjectivity ] for each @xmath23 the mapping @xmath306 is surjective .
let @xmath187 be a cocycle representing an element of @xmath259 .
we decompose it as in with respect to the last site .
the equation @xmath272 leads to @xmath307 let us consider the case @xmath294 .
from [ lem : technical ] we know that without loss of generality @xmath187 can be chosen in such a way that @xmath308 for some state @xmath309 .
this choice leads to @xmath310 comparing both sides , we obtain @xmath311 according to and , @xmath312 . hence , we find @xmath313 with a cocycle @xmath314 . for the corresponding cohomology classes we find thus @xmath210=s^\sharp[|\psi_{\ell,0}\rangle]$ ] .
[ lem : injectivity ] for each @xmath23 the mapping @xmath306 is injective .
consider an element of @xmath315 .
it can be represented by a cocycle @xmath165 such that @xmath316 for some vector @xmath317 .
as before , it is useful to decompose the state with respect to the last site : @xmath318 .
we find @xmath319 we select on both sides the terms corresponding to @xmath264 on the last site , and find @xmath320 notice that the sum on the right - hand side does not contain any term proportional to @xmath321 on the last site . @xmath322 may however contain such a term . to see this , we decompose @xmath323 .
we apply the supercharge to this decomposition , insert it into and obtain @xmath324 the states @xmath325 can in principle be computed but we wo nt need them .
this equality implies @xmath326 .
hence , @xmath187 is a coboundary .
we conclude that @xmath327 .
this proves the claim .
we are now ready to prove the main result of this section . from [ lem : surjectivity , lem :
injectivity ] we conclude for each @xmath23 the mapping @xmath328 is both surjective and injective .
hence @xmath329 is isomorphic to @xmath258 . by transitivity
, we obtain @xmath330 for each @xmath331 . @xmath260 and
@xmath261 were obtained in [ lem : lowdim ] .
this allows us to compute representatives of the elements of @xmath208 for odd and even @xmath4 from the repeated action of @xmath300 on @xmath264 and @xmath265 , respectively , which leads to .
in this section , we analyse the zero - energy states of the spin - chain hamiltonians @xmath51 with @xmath115 .
our main goal is to unveil some of their properties with the help of [ thm : occohom ] . in [ sec : representations ] , we discuss two representations of the zero - energy states arising from the representatives of the cohomology and their homology analogues .
we deduce from these representations their magnetisation , parity and relations between certain components . in [ sec : scalarproducts ] , we introduce a family of physically - interesting scalar products that involve an arbitrary number of zero - energy states for systems of different lengths .
we use the supersymmetry to show that the knowledge of a single special component of each involved state is sufficient to evaluate the scalar product .
we denote by @xmath13 a zero - energy state of the hamiltonian @xmath51 for a chain of length @xmath4 . for convenience , we allow @xmath233 even though @xmath51 is not defined for chains consisting of a single site
. furthermore , we write @xmath332 for its components with respect to the canonical basis of @xmath17 . [ [ cohomology - representation . ] ] cohomology representation .
+ + + + + + + + + + + + + + + + + + + + + + + + + + it follows from [ thm : occohom ] that @xmath13 can be written as the sum of a multiple of a special representative of @xmath208 and an element of the image of the supercharge .
specifically , we have [ prop : cohomrepr ] for each @xmath331 , we have @xmath333 with @xmath334 and @xmath335 with @xmath336 . for simplicity
, we focus on the case @xmath337 . according to and [ thm : occohom ]
, there is a scalar @xmath338 and a state @xmath334 such that the zero - energy state can be written as @xmath339 the state @xmath340 can not be obtained from @xmath341 .
it is fixed by the requirement that the zero - energy state be annihilated by the adjoint supercharge .
we notice however that the unknown term @xmath342 does not contribute to the scalar product of the zero - energy state with any vector @xmath343 .
indeed , we have @xmath344 a simple choice for @xmath345 is given by a canonical basis vector @xmath346 that _ ( i ) _ has a non - zero scalar product with @xmath254 and _ ( ii ) _ is annihilated by the adjoint supercharge . the requirement _
( i ) _ leads to the constraint [ eqn : constraintsbasisvector ] @xmath347 because the magnetisation of the state @xmath265 vanishes .
the requirement _
( ii ) _ leads to the constraint @xmath348 which follows from the action of the adjoint local supercharge @xmath61 .
the constraints have many solutions .
one solution is given by @xmath349 using , we find that @xmath350 and thus @xmath351 .
this concludes the proof for @xmath337 .
the proof for @xmath352 is similar and uses the vector @xmath353 .
the representations of the zero - energy states given in and allow us to derive a number of simple properties of the zero energy states .
two immediate consequences are : [ corr : nonzerocomps ] for each @xmath331 the components @xmath354 and @xmath355 are non - zero .
consider @xmath337 and suppose that @xmath356 .
then it follows from that @xmath357 which contradicts [ lem : e0notcoboundary ]
. the case @xmath352 is completely analogue .
[ corr : magnetisation ] we have @xmath358 and @xmath359 .
the @xmath74 eigenspace of @xmath51 is one - dimensional .
furthermore the hamiltonian conserves the magnetisation @xmath139=0 $ ] .
hence , we must have @xmath360 for any @xmath23 . to find @xmath361 , it is sufficient to project this equality onto simple basis vectors . for @xmath337 , we find @xmath362 according to [ corr : nonzerocomps ] , @xmath363 is non - zero and therefore @xmath364 .
for @xmath352 the proof is similar . for @xmath337 ,
the constraints have many solutions .
this allows us to relate certain components of the zero - energy state to the special component @xmath354 .
[ prop : specialcompl2n ] for any weakly increasing sequence @xmath365 of integers , we have the component @xmath366 it is trivial to check that holds for @xmath367 and @xmath368 where @xmath369 . hence , using and we find @xmath370 which concludes the proof .
the [ prop : cohomrepr , prop : specialcompl2n ] imply that the zero - energy state is even under the action of the parity operator .
for any @xmath23 we have @xmath371 .
the proof follows the lines of [ corr : magnetisation ] , using [ prop : specialcompl2n ] in the case of even @xmath4 .
[ [ homology - representation . ] ] homology representation .
+ + + + + + + + + + + + + + + + + + + + + + + + up to now , we have focused on the cohomology of the supercharge @xmath21 and the resulting representations of the zero - energy states given in [ prop : cohomrepr ] .
the definition @xmath372 of a supersymmetric hamiltonian suggests that we could as well have considered the adjoint supercharge @xmath34 .
its _ homology _ is defined by @xmath373 where @xmath374 and @xmath375 the existence of a non - degenerate inner product on @xmath17 implies that the cohomology and homology are isomorphic @xcite : @xmath376 in particular , we have @xmath377 we now determine an alternative representation of the zero - energy states using this property .
[ prop : homrepr ] for each @xmath331 , we have @xmath378 with @xmath379 and @xmath380 with @xmath381 .
we focus on the case @xmath337 .
since @xmath382 is one - dimensional , each of its non - zero elements can be represented by a non - zero multiple of a fixed vector in @xmath383 .
this vector is in the kernel ( but not in the image ) of @xmath34 .
we claim that such a vector is given by @xmath384 using , one readily checks that @xmath345 is annihilated by @xmath34 .
furthermore , it can not be in the image of @xmath34 . indeed , otherwise if @xmath385 for some @xmath386 then it follows that @xmath387 .
this contradicts [ corr : nonzerocomps ] .
it follows that the zero - energy states for @xmath337 have the representation @xmath388 for some non - zero scalar @xmath389 and @xmath390 .
the vector @xmath391 can not be determined from homological arguments .
the factor @xmath389 , however , can be found by computing the scalar product of the zero - energy state with suitable states @xmath392 that are annihilated by the supercharge @xmath21 and have a non - zero scalar product with the representative . in the present case , we choose @xmath393 .
a short calculation , similar to , leads to @xmath394 the argument for @xmath352 is similar , with the choice @xmath395 . in this subsection , we use the [ prop : cohomrepr , prop : homrepr ] in order to derive sum rules for the square norm of the zero - energy states and certain scalar products .
an immediate consequence of the ( co)homology representations of the zero - energy states given in [ prop : cohomrepr , prop : homrepr ] is : for each @xmath331 , we have [ eqn : squarenorm ] @xmath396 and @xmath397 for @xmath337 , we find @xmath398 for @xmath352 the proof is similar . the expressions for the square norm motivate the analysis of a family of scalar products involving the zero - energy states of arbitrary length . to be specific ,
let us consider @xmath337 .
we find that @xmath399 on the left - hand side of this equality , we have the projection of the normalised zero - energy state onto an @xmath257-fold tensor product of zero - energy states of the two - site chain . on the right - hand site
, we find the inverse of a special component of the normalised zero - energy state on @xmath400 sites .
it is natural to investigate if the result remains equally simple when we replace the @xmath265 s by the zero - energy energy states of spins chains of generic lengths . to this end
, we cut an open chain of length @xmath4 into @xmath65 subchains of lengths @xmath401 as is illustrated in [ fig : subdivision ] .
( 4,0 ) ; ( 4,0 )
( 5,0 ) ; ( 5,0 )
( 8,0 ) ; in 0,2,3.5,5.5,8 ( 0,-.1 ) ( 0,.1 ) ; ( 0,-.4 )
( 2,-.4 ) ; ( 2,-.4 ) ( 3.5,-.4 ) ; ( 5.5,-.4 )
( 8,-.4 ) ; ( 0,.4 )
( 8,.4 ) ; in 0,0.25, ... ,3.75 ( + .125,0 ) circle ( 1.25pt ) ; in 5,5.25, ... ,7.75 ( + .125,0 ) circle ( 1.25pt ) ; ( 4.55,-0.4 ) node @xmath402 ; ( 4,.4 ) node[above ] @xmath4 ; ( 1.1,-.4 ) node[below ] @xmath403 ; ( 2.8,-.4 ) node[below ] @xmath404 ; ( 6.8,-.4 ) node[below ] @xmath405 ; our aim is to compute projection of the zero - energy state of the complete chain onto the tensor product of the zero - energy states of these @xmath65 subchains .
we consider thus the scalar product @xmath406 the division by the norms of the states makes this quantity normalisation - independent . for @xmath337 and @xmath407 for @xmath408
, we recover ( up to factor ) the left - hand side of .
[ thm : resultforz ] if @xmath16 are even , then @xmath409 if @xmath410 is odd ( for some @xmath411 ) and @xmath412 are even , then @xmath413 in all other cases , the scalar product vanishes .
we note that the results and are remarkably simple .
the zero - energy states typically are very complicated states with many non - zero components .
nonetheless , the scalar product can be inferred from the sole knowledge of a single component of each involved ( normalised ) state .
furthermore , the exchange of @xmath414 and @xmath415 for @xmath416 leaves the result invariant even though it can completely change the subdivision as shown in [ fig : subdivision ] . for simplicity
, we focus on the proof of . using , we find @xmath417 the second term on the right - hand side of this equation vanishes .
indeed , we obtain @xmath418 which is zero because @xmath419 for @xmath420 .
hence , we find @xmath421 using , we write @xmath422 in terms of the square norm and the alternating component . upon division by the norms of the zero - energy states , we obtain .
the proof of is similar . eventually , the scalar product vanishes in all other cases because of the definite magnetisation of the zero - energy states .
there are many generalisations of that have equally simple expressions in terms of special components . as an example
, we consider the product @xmath423 for @xmath424 [ thm : resultforztilde ] if @xmath16 are odd then @xmath425 if @xmath410 is even ( for some @xmath411 ) and @xmath412 are odd , then @xmath426 in all other cases , the scalar product vanishes .
let us prove .
the reasoning is similar to the proof of [ thm : resultforz ] .
we use in order to write @xmath427 the term involving @xmath34 vanishes .
indeed , one checks that if both @xmath187 and @xmath289 are in the kernel of @xmath34 then @xmath428 as a consequence of . applying this property repeatedly ,
we reduce to @xmath429 using , we find .
the proof of is similar , using in addition the parity invariance of the zero - energy state . in all other cases ,
the scalar products vanish because of the definite magnetisation of the zero - energy states .
it is possible to extend the definition of @xmath430 in order to account for the case where @xmath431 for some @xmath411 and the @xmath412 are odd . in this case
, we have @xmath432
in this section , we discuss the scalar products found in [ thm : resultforz , thm : resultforztilde ] for @xmath3 . in this case , the hamiltonian is given by and describes the open xxz spin chain with anisotropy parameter @xmath1 and special diagonal boundary terms . in [ sec : finitesize ] , we provide exact finite - size expressions for certain special components of the normalised zero - energy state .
we use them to exactly evaluate the scalar products @xmath433 and @xmath434 .
furthermore , we evaluate the scaling limits of the scalar products . in [ sec : lbf ] , we compare these scaling limits for @xmath15 to the predictions of conformal field theory .
in contrast to the previous sections , the results presented here below are non - rigorous .
many statements are based on conjectures that we inferred from the exact diagonalisation of the spin - chain hamiltonian .
[ [ a - finite - size - conjecture . ] ] a finite - size conjecture .
+ + + + + + + + + + + + + + + + + + + + + + + + + we computed the zero - energy state @xmath13 of the xxz hamiltonian up to @xmath435 sites using mathematica .
we find that the ratio of the components @xmath436 , @xmath437 and the norm of the corresponding states can be expressed in terms of two integer sequences @xmath438 and @xmath439 .
these two sequences enumerate @xmath440 vertically - symmetric alternating sign matrices and cyclically - symmetric self - complementary plane partitions in a @xmath441 cube , respectively @xcite .
explicitly , they are given by @xmath442 our numerical investigation is consistent with the following conjecture : [ conj : normalisedcomponents ] for each @xmath331 we have @xmath443 the occurrence of sequences that enumerate alternating sign matrices and plane partitions in the ground state of an xxz spin chain at @xmath1 comes by no means as a surprise ( see for example @xcite ) .
we provide a proof of [ conj : normalisedcomponents ] along with many other combinatorial properties of the ground state @xmath13 in a separate publication @xcite . [
[ scaling - behaviour . ] ] scaling behaviour .
+ + + + + + + + + + + + + + + + + + the sequences @xmath444 and @xmath445 are given by ratios of products of factorials .
hence , they can be expressed in terms of barnes @xmath446-function @xcite .
we use the well - known asymptotic expansion of this function to evaluate the components for large system sizes : @xmath447 here @xmath448 and @xmath449 are the constants @xmath450 [ eqn : asymptotics ] we use these expressions to extract the scaling behaviour of @xmath433 , for @xmath3 .
it is obtained when the lengths of the subintervals @xmath16 become large in such a way that the ratios @xmath451 approach certain scaling variables @xmath452 for each @xmath420 .
@xmath433 is then given by an asymptotic series with respect to the system size @xmath4 .
the series coefficients are functions of @xmath453 .
notice that together , [ thm : resultforz ] and imply that this asymptotic series is only well defined if the parity of the integers @xmath16 is fixed .
there are two interesting cases , corresponding to and : 1 .
@xmath414 is even for each @xmath420 . in this case
the length of the chain @xmath4 is even .
using , we obtain @xmath454 2 .
@xmath415 is odd for a certain @xmath171 and @xmath414 is even for each @xmath455 . in this case
, @xmath4 is odd . from
, we obtain @xmath456 we obtain the scaling behaviour of the scalar product @xmath434 from [ thm : resultforztilde ] and .
the resulting leading - order terms of and are up to a factor equal to and , respectively . hence , without loss of generality we focus in the following on @xmath433 .
[ eqn : scalinglimitsz ] the power - law decay of the scalar products in as well as their algebraic dependence on the scaling coordinates @xmath457 suggest that they could be related to correlation functions of conformal field theory ( cft ) . in this section
, we discuss this relation for the special case of @xmath15 .
specifically , we consider the so - called _ logarithmic bipartite fidelity _
( lbf ) : @xmath458 it was introduced by dubail and stphan as an entanglement measure for the ground state of interacting quantum many - body systems in one dimension@xcite . in particular , they predicted the leading - order terms of the asymptotic expansion of the lbf with respect to the system size @xmath4 for one - dimensional quantum critical systems from cft arguments . from and we exactly compute these leading - order terms .
we express them in terms of @xmath459 . in case _
( i ) _ , we obtain @xmath460 in the case _
( ii ) _ , we take the scaling limit with @xmath403 even and @xmath404 odd .
this leads to @xmath461 [ eqn : lbfscaling ] the aim of this section is to show that perfectly matches the cft predictions . to this end
, we briefly discuss the relation between the scaling limit of the open xxz chain at @xmath1 and superconformal cft in [ sec : cft ] .
in [ sec : lbf ] we compare our findings for the scaling limit of the spin chain s lbf to the cft results .
it is well known that in suitable scaling limits , many properties of the xxz spin with anisotropy parameter @xmath462 are accurately described by cft . for open chains and real diagonal boundary terms
, the field theory is expected to be given by a free boson theory with central charge @xmath463 and a compactification radius that depends on the value of @xmath8 @xcite . if @xmath1 then this compactification radius is fixed to a value where the field theory coincides with the first model of the so - called @xmath464 superconformal minimal series @xcite ( see also @xcite for a compact introduction accessible to non - experts ) .
this implies in particular that the cft space of states divides into the so - called ramond and neveu - schwarz sectors .
each sector organises in a finite number of irreducible highest - weight representations of the so - called @xmath464 superconformal algebra . from the representation theory of this algebra
, we know that the corresponding highest - weight states are labeled by a pair @xmath465 . here @xmath94 denotes the conformal weight and @xmath466 the so - called @xmath467 charge of the state . for conformal weights
this is not to be confused with the hamiltonian density discussed in previous sections .
furthermore , for the @xmath467-charges @xmath466 , we follow the conventions of @xcite that differ from the standard choice in the literature on superconformal field theory .
for the first @xmath464 superconformal minimal model that choice is @xmath468 . ] in the following , we focus on the ramond sector . for the first @xmath464 superconformal minimal model , it contains three highest - weight representations .
two of them are singlet representations with the corresponding highest - weight states labeled by @xmath469 furthermore , there is a doublet representation with a highest - weight state labeled by @xmath470 [ eqn : cftdata ] it has a superpartner with the same conformal weight but the opposite @xmath467-charge : @xmath471 .
notice that since we deal with a free boson theory we have @xmath472 for all these states @xcite .
[ [ lattice - and - cft - quantities . ] ] lattice and cft quantities . + + + + + + + + + + + + + + + + + + + + + + + + + + + the xxz hamiltonian possesses an explicit lattice realisation of the supersymmetry .
this suggests that for finite @xmath4 its low - energy eigenstates constitute an approximation to the ramond sector of the cft hilbert space . in order to substantiate this claim
, we identify the representation data with certain properties of the ground state and first excited state of the lattice hamiltonian ( see also @xcite and furthermore @xcite for a related analysis of the @xmath73 model of supersymmetric fermions ) .
the conformal weight @xmath94 of a low - energy eigenstate of the hamiltonian is related to the finite - size scaling of its eigenvalue .
let us denote by @xmath473 the eigenvalues of the hamiltonian at size @xmath4 .
then we have for large @xmath4 and small @xmath46 the following expansion @xcite : @xmath474 where the central charge takes the value @xmath463 .
the factor @xmath475 is the so - called fermi velocity that can be computed by bethe - ansatz techniques @xcite .
for @xmath1 we have @xmath476 .
furthermore , @xmath477 and @xmath478 are non - universal constants that depend on the definition of the hamiltonian . in our case
, these constants have to be zero , @xmath479 , because the spectrum contains the ground - state eigenvalue @xmath480 for any @xmath36 .
furthermore , fixes the conformal weight of the zero - energy states to @xmath481 . to probe the consistency of these assignments ,
we have numerically examined the dependence of the eigenvalue @xmath482 on @xmath4 .
its scaling behaviour matches well with @xmath483 .
furthermore , it is natural to assume that the @xmath467-charge of an eigenstate of the hamiltonian is related to a linear function of its magnetisation .
we fix this function by comparing the @xmath467-charges of the two ramond ground states in and the magnetisation of the lattice ground states that we obtained in [ corr : magnetisation ] .
this suggests that the operator @xmath484 measures the cft @xmath467-charge : we have @xmath485 for each @xmath331 . to see that this choice is consistent , we inspect the first excited state of the hamiltonian .
we denote this state by @xmath486 for a system of length @xmath4 .
our exact diagonalisation data supports the conjecture that for each @xmath331 , we have @xmath487 and @xmath488 . furthermore , the magnetisation of the states that we computed for small systems is compatible with the conjecture @xmath489 for each @xmath331 .
this suggests that the first excited state for spin chains of odd length is a lattice approximation to the cft highest - weight state @xmath490 .
[ [ conformal - weight - of - the - cut . ] ] conformal weight of the cut .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + in order to compare our results and to the predictions from conformal field theory , we need a last ingredient : the @xmath467-charge of the cut . to motivate the introduction of this quantity , recall that the scalar product @xmath491 is non - zero only if the product - state @xmath492 and @xmath13 have the same magnetisation .
let us now rephrase this statement in terms of the @xmath467-charges of the three zero - energy states @xmath493 , @xmath494 and @xmath13 , which we denote by @xmath495 and @xmath496 , respectively .
we find that @xmath491 is non - vanishing only if @xmath497 the quantity @xmath498 is the @xmath467-charge of the cut .
it allows us to formally define the conformal weight of the cut @xmath499 . within cft ,
it is possible to identify @xmath500 with the conformal weight of a so - called boundary condition changing operator .
we refer to @xcite for more details .
[ [ finite - size - scaling - of - the - lbf . ] ] finite - size scaling of the lbf .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + for quantum critical systems , the leading - order terms of the asymptotic expansion of the lbf with respect to the system size @xmath4 is given by @xcite : @xmath501 here , @xmath502 is the central charge of the theory and @xmath503 the conformal weight of the cut .
furthermore , @xmath459 denotes the scaling variable defined above .
the functions @xmath504 and @xmath505 depend on the conformal weights and @xmath467-charges associated to the involved states and the cut .
their explicit general form can be found in @xcite . here
, we only consider the case of a @xmath463 free field theory , where they can be written in terms of the @xmath467 charges alone : @xmath506 and @xmath507[eqn : cftprediction ] here @xmath508 and @xmath509 are non - universal constants .
in particular , @xmath510 is called the _ extrapolation length _ @xcite . apart from these two constants , the expressions are universal . [ [ comparison - to - scaling - limit . ] ] comparison to scaling limit .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + the identification of the @xmath467-charges of the zero - energy states and the cut allow us to finally compare our findings with the cft prediction .
first , we notice that since @xmath463 and @xmath511 , the leading order term in is @xmath512 .
this is consistent with our findings in .
second , for the subleading terms we find the following results . in case _ ( i ) _ we have @xmath513 in case _ ( ii ) _ we have @xmath514 in both cases , the expressions for @xmath504 and @xmath505 match perfectly the scaling limit of the lattice results , provided that we set @xmath515 .
the absence of the @xmath516-term is worth mentioning .
similar instances of vanishing finite - size corrections for the quantum spin @xmath0 xxz chain at @xmath1 were reported in the literature @xcite . in the present case ,
the vanishing is a direct consequence of the supersymmetry as it fixes the @xmath467-charges of the zero - energy states to @xmath517 .
we end our analysis with a comment on the higher - spin cases @xmath518 .
for these cases we expect the @xmath516 term to vanish , too .
indeed , for arbitrary @xmath12 the scaling limit of the spin chains should be described by the ramond sector of the @xmath12-th @xmath464 superconformal minimal model .
indeed , this conjecture was made for the related @xmath64 models of strongly - interacting fermions with supersymmetry @xcite .
it is well - known that the ramond ground states have the conformal weight @xmath519 .
furthermore , for arbitrary @xmath502 the scaling function @xmath505 is given by @xcite @xmath520if all of the involved states are ramond ground states ( or at least lattice approximations thereof ) then this expression identically vanishes .
in this article , we have studied the dynamic lattice supersymmetry for the open xxz heisenberg chains at their supersymmetric point .
we have determined the family @xmath164 of supercharges and identified a family of non - diagonal boundary interactions that are compatible with the supersymmetry .
furthermore , we have computed the cohomology of the supercharges and shown that it is non - trivial if and only if @xmath115 . from this cohomology computation
we have deduced the existence of a zero - energy state , unique up to normalisation , of the spin - chain hamiltonians with @xmath115 . a central result of this article is the sum rules presented in [ thm : resultforz , thm : resultforztilde ] : they reveal that a large family of scalar products involving an arbitrary number of normalised zero - energy states can be computed in terms of certain distinguished components .
we emphasise that these results are solely based on the supersymmetric structure and do not make use of any integrability techniques .
eventually , we have computed the scaling behaviour of logarithmic bipartite fidelity for the open spin @xmath0 quantum xxz chain at @xmath1 with special boundary magnetic fields .
its scaling behaviour matches the predictions from conformal field theory at both leading and sub - leading orders .
the results of this work have a few interesting generalisations .
first , it is natural to consider closed spin chains with periodic or twisted boundary conditions . in this case
, the dynamic lattice supersymmetry only exists in certain ( anti-)cyclic subspaces of the spin - chain hilbert space @xcite .
therefore , the computation of the cohomology of the supercharges is more challenging .
nonetheless , we expect that there is a connection between the cohomology of supercharges of the open and the closed chains , based on known examples in the mathematical literature @xcite
. it should allow to establish some generalisations of [ thm : resultforz , thm : resultforztilde ] to closed spin chains for the class of models discussed in this article .
second , the computation of similar scalar products in off - critical models with lattice supersymmetry , in particular their scaling limit in the vicinity of a critical point , is of potential interest .
an example is the open staggered @xmath73 model @xcite .
we hope to address these problems in forthcoming publications .
this work is supported by the belgian interuniversity attraction poles program p7/18 through the network dygest ( dynamics , geometry and statistical physics ) .
we thank alexi morin - duchesne for many interesting remarks and stimulating suggestions .
furthermore , we thank luigi cantini , jrme dubail , axel marcillaud de goursac and anita ponsaing for discussions .
|
the quantum spin @xmath0 xxz chain with anisotropy parameter @xmath1 possesses a dynamic supersymmetry on the lattice .
this supersymmetry and a generalisation to higher spin are investigated in the case of open spin chains .
a family of non - diagonal boundary interactions that are compatible with the lattice supersymmetry and depend on several parameters is constructed .
the cohomology of the corresponding supercharges is explicitly computed as a function of the parameters and the length of the chain . for certain specific values of the parameters ,
this cohomology is shown to be non - trivial .
this implies that the spin - chain ground states are supersymmetry singlets .
special scalar products involving an arbitrary number of these supersymmetry singlets for chains of different lengths are exactly computed . as a physical application ,
the logarithmic bipartite fidelity of the open quantum spin @xmath0 xxz chain with @xmath1 and special diagonal boundary interactions is determined .
= 1
|
in 2015 , the lhcb collaboration studied the @xmath7 decays with the data sample corresponds to an integrated luminosity of @xmath8 acquired with the lhcb detector from 7 and 8 tev @xmath9 collisions , and observed two pentaquark candidates @xmath6 and @xmath10 in the @xmath11 invariant mass distributions with the significances of more than 9 standard deviations @xcite .
they performed the amplitude analysis on all relevant masses and decay angles of the six - dimensional data using the helicity formalism and breit - wigner amplitudes to describe all resonances .
the breit - wigner masses and widths are @xmath5 , @xmath12 , @xmath13 , and @xmath14 , respectively .
the preferred quantum numbers of the @xmath6 and @xmath10 are @xmath15 and @xmath16 , respectively .
recently , the lhcb collaboration inspected the @xmath7 decays for the presence of @xmath11 or @xmath17 contributions with minimal assumptions about @xmath18 contributions ( such as their number , their resonant or nonresonant nature , or their lineshapes ) , and observed that at more than 9 standard deviations the @xmath7 decays can not be described with the @xmath18 contributions alone , and the @xmath11 contributions play a dominant role in this incompatibility , and obtained model - independent support for the evidences of the @xmath19 decays @xcite .
furthermore , the lhcb collaboration performed a full amplitude analysis of the @xmath20 decays allowing for previously observed conventional ( @xmath21 ) and exotic ( @xmath11 and @xmath22 ) resonances , and observed that a significantly better description of the data can be achieved by either including the two @xmath23 states or the @xmath24 state @xcite
. there have been several possible assignments since the observations of the @xmath6 and @xmath10 , such as the molecule - like pentaquark states @xcite ( or not the molecular pentaquark states @xcite ) , the diquark - triquark type pentaquark states @xcite , the diquark - diquark - antiquark type pentaquark states @xcite , re - scattering effects @xcite , etc .
we can diagnose their resonant or non - resonant nature using photoproduction ( or pionproduction ) off a proton target @xcite .
the photoproduction process does not satisfy the so - called anomalous triangle singularity condition , their presence in the @xmath25 photoproduction is crucial to conforming the resonant nature of such states as opposed to their being kinematical effects . the qcd sum rules is a powerful theoretical tool in studying the ground state heavy baryon states @xcite .
we usually take the diquarks as the basic constituents in constructing the baryon currents and pentaquark currents . the attractive interactions induced by one - gluon exchange favor formation of the diquark states in color antitriplet , flavor antitriplet and spin singlet @xmath26 or in color antitriplet , flavor sextet and spin triplet @xmath27 , where the @xmath28 , @xmath29 and @xmath30 are color indexes @xcite
, the qcd sum rules also favor formation of the diquark states @xmath26 and @xmath27 @xcite , so we can take the scalar or axialvector diquark states as the basic constituents in constructing currents to interpolate the lowest pentaquark states . in our previous works , we observed that the heavy - light scalar and axialvector diquark states have almost degenerate masses @xcite , the masses of the light axialvector diquark states lie @xmath31 above that of the corresponding light scalar diquark states @xcite from the qcd sum rules .
we expect that the ground state diquark - diquark - antiquark type hidden - charm pentaquark states consist of a light scalar ( or axialvector ) diquark , a charm scalar ( or axialvector ) diquark and an anti - charm - quark . in refs.@xcite , we take the light scalar diquark states , heavy axialvector diquark states and heavy scalar diquark states as the basic constituents , construct the scalar - diquark - axialvector - diquark - antiquark type and scalar - diquark - scalar - diquark - antiquark type currents to study the @xmath15 , @xmath16 and @xmath32 hidden - charm pentaquark states using the qcd sum rules .
the predicted masses @xmath33 and @xmath34 support assigning the @xmath6 and @xmath10 to be the @xmath35 and @xmath16 hidden - charm pentaquark states , respectively @xcite . in ref.@xcite , we observe that the empirical energy scale formula , @xmath36 with the effective heavy quark mass @xmath37 , in determining the ideal energy scales of the qcd spectral densities in the qcd sum rules for the tetraquark states @xmath38 , @xmath39 and @xmath40 @xcite , can be successfully applied to study the hidden - charm pentaquark states with a slight modification @xmath41 . in ref.@xcite , we take the light axialvector diquark states , heavy axialvector diquark states and heavy scalar diquark states as the basic constituents , construct the axialvector - diquark - axialvector - diquark - antiquark type and axialvector - diquark - scalar - diquark - antiquark type interpolating currents to study the @xmath42 hidden - charm pentaquark states with the qcd sum rules in a systematic way . in this article , we resort to the same routine as that in refs.@xcite to study the masses and pole residues of the @xmath0 hidden - charm pentaquark states with the qcd sum rules by constructing the axialvector - diquark - scalar - diquark - antiquark type and axialvector - diquark - axialvector - diquark - antiquark type interpolating currents , and revisit the assignment of the @xmath6 . the article is arranged as follows : we derive the qcd sum rules for the masses and pole residues of the @xmath43 hidden - charm pentaquark states in sect.2 ; in sect.3 , we present the numerical results and discussions ; and sect.4 is reserved for our conclusion .
firstly , let us write down the two - point correlation functions @xmath44 in the qcd sum rules , @xmath45 where the @xmath46 are currents to interpolate the @xmath47 hidden - charm tetraquark states , the superscripts @xmath48 and @xmath49 are the spins of the light diquark state and the charm ( or heavy ) diquark state , respectively , @xmath50 , the @xmath51 is the spin of the charm ( or heavy ) antiquark , the subscripts @xmath52 , @xmath53 , @xmath54 are the light quark constituents @xmath55 , @xmath56 or @xmath57 .
now we write down the interpolating currents @xmath46 explicitly , @xmath58
c\bar{c}^{t}_{a}(x ) \ , , \nonumber\\ j^{10\frac{1}{2}}_{udd,\mu}(x)&=&\frac{\varepsilon^{ila } \varepsilon^{ijk}\varepsilon^{lmn}}{\sqrt{3 } } \left [ d^t_j(x ) c\gamma_\mu d_k(x ) u^t_m(x ) c\gamma_5 c_n(x)+2d^t_j(x ) c\gamma_\mu u_k(x ) d^t_m(x ) c\gamma_5 c_n(x)\right ] c\bar{c}^{t}_{a}(x ) \ , , \nonumber \\
j_{ddd,\mu}^{10\frac{1}{2}}(x)&=&\frac{\varepsilon^{ila } \varepsilon^{ijk}\varepsilon^{lmn}}{\sqrt{3 } } d^t_j(x ) c\gamma_\mu d_k(x ) d^t_m(x ) c\gamma_5 c_n(x ) c\bar{c}^{t}_{a}(x ) \ , , \end{aligned}\ ] ] @xmath59 c\bar{c}^{t}_{a}(x ) \ , , \nonumber\\ j^{10\frac{1}{2}}_{uds,\mu}(x)&=&\frac{\varepsilon^{ila } \varepsilon^{ijk}\varepsilon^{lmn}}{\sqrt{3 } } \left [ u^t_j(x ) c\gamma_\mu d_k(x ) s^t_m(x ) c\gamma_5 c_n(x)+u^t_j(x ) c\gamma_\mu s_k(x ) d^t_m(x ) c\gamma_5 c_n(x)\right .
\nonumber\\ & & \left.+d^t_j(x ) c\gamma_\mu s_k(x ) u^t_m(x ) c\gamma_5 c_n(x)\right ] c\bar{c}^{t}_{a}(x ) \ , , \nonumber\\ j^{10\frac{1}{2}}_{dds,\mu}(x)&=&\frac{\varepsilon^{ila } \varepsilon^{ijk}\varepsilon^{lmn}}{\sqrt{3 } } \left [ d^t_j(x ) c\gamma_\mu d_k(x ) s^t_m(x ) c\gamma_5 c_n(x)+2d^t_j(x ) c\gamma_\mu s_k(x ) d^t_m(x ) c\gamma_5 c_n(x)\right ] c\bar{c}^{t}_{a}(x ) \
, , \nonumber\\\end{aligned}\ ] ] @xmath60 c\bar{c}^{t}_{a}(x ) \ , , \nonumber \\ j^{10\frac{1}{2}}_{dss,\mu}(x)&=&\frac{\varepsilon^{ila } \varepsilon^{ijk}\varepsilon^{lmn}}{\sqrt{3 } } \left [ s^t_j(x ) c\gamma_\mu s_k(x ) d^t_m(x ) c\gamma_5 c_n(x)+2s^t_j(x ) c\gamma_\mu d_k(x ) s^t_m(x ) c\gamma_5 c_n(x)\right ] c\bar{c}^{t}_{a}(x ) \ , , \nonumber \\\end{aligned}\ ] ] @xmath61 @xmath62 \gamma_5\gamma^\alpha c\bar{c}^{t}_{a}(x ) \ , , \nonumber\\ j^{11\frac{1}{2}}_{udd,\mu}(x)&=&\frac{\varepsilon^{ila } \varepsilon^{ijk}\varepsilon^{lmn}}{\sqrt{3 } } \left [ d^t_j(x ) c\gamma_\mu d_k(x)u^t_m(x ) c\gamma_\alpha c_n(x)+2d^t_j(x ) c\gamma_\mu u_k(x)d^t_m(x ) c\gamma_\alpha c_n(x ) \right ] \gamma_5\gamma^\alpha c\bar{c}^{t}_{a}(x ) \ , , \nonumber\\ j^{11\frac{1}{2}}_{ddd,\mu}(x)&=&\varepsilon^{ila } \varepsilon^{ijk}\varepsilon^{lmn } d^t_j(x ) c\gamma_\mu d_k(x)d^t_m(x ) c\gamma_\alpha c_n(x ) \gamma_5\gamma^\alpha c\bar{c}^{t}_{a}(x ) \ , , \end{aligned}\ ] ] @xmath63 \gamma_5\gamma^\alpha c\bar{c}^{t}_{a}(x ) \ , , \nonumber\\ j^{11\frac{1}{2}}_{uds,\mu}(x)&=&\frac{\varepsilon^{ila } \varepsilon^{ijk}\varepsilon^{lmn}}{\sqrt{3 } } \left [ u^t_j(x ) c\gamma_\mu d_k(x)s^t_m(x ) c\gamma_\alpha c_n(x)+u^t_j(x ) c\gamma_\mu s_k(x)d^t_m(x ) c\gamma_\alpha c_n(x ) \right.\nonumber\\ & & \left.+d^t_j(x ) c\gamma_\mu s_k(x)u^t_m(x ) c\gamma_\alpha c_n(x ) \right ] \gamma_5\gamma^\alpha c\bar{c}^{t}_{a}(x ) \ , , \nonumber\\ j^{11\frac{1}{2}}_{dds,\mu}(x)&=&\frac{\varepsilon^{ila } \varepsilon^{ijk}\varepsilon^{lmn}}{\sqrt{3 } } \left [ d^t_j(x ) c\gamma_\mu d_k(x)s^t_m(x ) c\gamma_\alpha c_n(x)+2d^t_j(x ) c\gamma_\mu s_k(x)d^t_m(x ) c\gamma_\alpha c_n(x ) \right ] \gamma_5\gamma^\alpha c\bar{c}^{t}_{a}(x ) \ , , \nonumber\\\end{aligned}\ ] ] @xmath64 \gamma_5\gamma^\alpha c\bar{c}^{t}_{a}(x ) \ , , \nonumber\\ j^{11\frac{1}{2}}_{dss,\mu}(x)&=&\frac{\varepsilon^{ila } \varepsilon^{ijk}\varepsilon^{lmn}}{\sqrt{3 } } \left [ s^t_j(x ) c\gamma_\mu s_k(x)d^t_m(x ) c\gamma_\alpha c_n(x)+2s^t_j(x ) c\gamma_\mu d_k(x)s^t_m(x ) c\gamma_\alpha c_n(x ) \right ] \gamma_5\gamma^\alpha c\bar{c}^{t}_{a}(x ) \ , , \nonumber\\ \end{aligned}\ ] ] @xmath65 @xmath66 \gamma_5\gamma^\alpha c\bar{c}^{t}_{a}(x ) \ , , \nonumber\\ j^{11\frac{3}{2}}_{udd,\mu}(x)&=&\frac{\varepsilon^{ila } \varepsilon^{ijk}\varepsilon^{lmn}}{\sqrt{3 } } \left [ d^t_j(x ) c\gamma_\alpha d_k(x)u^t_m(x ) c\gamma_\mu c_n(x)+2d^t_j(x ) c\gamma_\alpha u_k(x)d^t_m(x ) c\gamma_\mu c_n(x ) \right ] \gamma_5\gamma^\alpha c\bar{c}^{t}_{a}(x ) \ , , \nonumber\\ j^{11\frac{3}{2}}_{ddd,\mu}(x)&=&\varepsilon^{ila } \varepsilon^{ijk}\varepsilon^{lmn } d^t_j(x ) c\gamma_\alpha d_k(x)d^t_m(x ) c\gamma_\mu c_n(x ) \gamma_5\gamma^\alpha c\bar{c}^{t}_{a}(x ) \ , , \end{aligned}\ ] ] @xmath67 \gamma_5\gamma^\alpha c\bar{c}^{t}_{a}(x ) \ , , \nonumber\\ j^{11\frac{3}{2}}_{uds,\mu}(x)&=&\frac{\varepsilon^{ila } \varepsilon^{ijk}\varepsilon^{lmn}}{\sqrt{3 } } \left [ u^t_j(x ) c\gamma_\alpha d_k(x)s^t_m(x ) c\gamma_\mu c_n(x)+u^t_j(x ) c\gamma_\alpha s_k(x)d^t_m(x ) c\gamma_\mu c_n(x ) \right .
\nonumber\\ & & \left.+d^t_j(x ) c\gamma_\alpha s_k(x)u^t_m(x ) c\gamma_\mu c_n(x ) \right ] \gamma_5\gamma^\alpha c\bar{c}^{t}_{a}(x ) \ , , \nonumber\\ j^{11\frac{3}{2}}_{dds,\mu}(x)&=&\frac{\varepsilon^{ila } \varepsilon^{ijk}\varepsilon^{lmn}}{\sqrt{3 } } \left [ d^t_j(x ) c\gamma_\alpha d_k(x)s^t_m(x ) c\gamma_\mu c_n(x)+2d^t_j(x ) c\gamma_\alpha s_k(x)d^t_m(x ) c\gamma_\mu c_n(x ) \right ] \gamma_5\gamma^\alpha c\bar{c}^{t}_{a}(x ) \ , , \nonumber\\ \end{aligned}\ ] ] @xmath68 \gamma_5\gamma^\alpha c\bar{c}^{t}_{a}(x ) \ , , \nonumber\\ j^{11\frac{3}{2}}_{dss,\mu}(x)&=&\frac{\varepsilon^{ila } \varepsilon^{ijk}\varepsilon^{lmn}}{\sqrt{3 } } \left [ s^t_j(x ) c\gamma_\alpha s_k(x)d^t_m(x ) c\gamma_\mu c_n(x)+2s^t_j(x ) c\gamma_\alpha d_k(x)s^t_m(x ) c\gamma_\mu c_n(x ) \right ] \gamma_5\gamma^\alpha c\bar{c}^{t}_{a}(x ) \ , , \nonumber\\ \end{aligned}\ ] ] @xmath69 where the @xmath28 , @xmath29 , @xmath30 , @xmath70 , @xmath71 , @xmath72 and @xmath73 are color indices , the @xmath74 is the charge conjugation matrix .
we take the isospin limit by assuming the @xmath55 and @xmath56 quarks have the degenerate masses , and classify the above thirty currents couple to the hidden - charm pentaquark states with degenerate masses into the following twelve types , @xmath75 in calculations , we choose the first current in each type for simplicity . the currents @xmath76 couple potentially to the @xmath0 and @xmath42 hidden - charm pentaquark states @xmath77 and @xmath78 , respectively , @xmath79 where the @xmath80 and @xmath81 are the pole residues or current - hadron coupling constants , the @xmath82 and @xmath83 are the dirac spinors @xcite .
the dirac spinors @xmath82 obey the rarita - schwinger equation @xmath84 , and the relations @xmath85 , @xmath86 ; while the dirac spinors @xmath83 obey the dirac equation @xmath87 . at the phenomenological side
, we can insert a complete set of intermediate hidden - charm pentaquark states with the same quantum numbers as the current operators @xmath46 and @xmath88 into the correlation functions @xmath44 to obtain the hadronic representation @xcite . after isolating the pole terms of the lowest states of the hidden - charm pentaquark states with @xmath0 and @xmath42 respectively ,
we obtain the following results : @xmath89 in this article , we choose the component @xmath90 associated with the tensor structure @xmath91 to study the hidden - charm pentaquark states with @xmath0 , there are no contaminations come from the hidden - charm pentaquark states with @xmath42 in the component @xmath92 associated with the tensor structure @xmath93 .
it is easy to project out the component @xmath90 , @xmath94 we can obtain the hadronic spectral densities through dispersion relation , @xmath95 \nonumber\\ & & + \left[m_{p,-}{\lambda^{-}_{p}}^2 \delta\left(s - m_{p,-}^2\right)-m_{p,+}{\lambda^{+}_{p}}^2 \delta\left(s - m_{p,+}^2\right)\right]\ , , \nonumber\\ & = & \!\not\!{p } \,\rho_{q_1q_2q_3,h}^{j_lj_h j,1}(s)+\rho_{q_1q_2q_3,h}^{j_lj_h j,0}(s ) \
, , \end{aligned}\ ] ] where the subscript index @xmath96 denotes the hadron side ( or phenomenological side ) .
then we introduce the weight function @xmath97 to obtain the phenomenological side of the qcd sum rules , @xmath98\exp\left ( -\frac{s}{t^2}\right ) & = & 2m_{p,-}\lambda^{-}_{p}{}^2\exp\left ( -\frac{m_{p,-}^2}{t^2}\right ) \ , , \\
\int_{4m_c^2}^{s_0}ds \left[\sqrt{s}\rho_{q_1q_2q_3,h}^{j_lj_h j,1}(s)-\rho_{q_1q_2q_3,h}^{j_lj_h j,0}(s)\right]\exp\left ( -\frac{s}{t^2}\right ) & = & 2m_{p,+}\lambda^{+}_{p}{}^2\exp\left ( -\frac{m_{p,+}^2}{t^2}\right ) \ , , \end{aligned}\ ] ] where the @xmath99 are the borel parameters and the @xmath100 are the continuum threshold parameters . in eqs.(19 - 20 ) , we separate the contributions of the negative - parity pentaquark states from that of the positive - parity pentaquark states explicitly .
they do not contaminate each other , and warrant robust predictions .
now we briefly outline the operator product expansion for the correlation functions @xmath101 and use the example @xmath102 to illustrate the routine .
firstly , we contract the @xmath57 and @xmath103 quark fields in the correlation function @xmath102 with wick theorem , @xmath104 \,tr\left[\gamma_5 c_{nn^\prime}(x ) \gamma_5 c s^{t}_{mm^\prime}(x)c\right ] \right . \nonumber\\ & & \left.-4 tr \left[\gamma_\mu s_{kk^\prime}(x ) \gamma_\nu c s^{t}_{mj^\prime}(x)c \gamma_5c_{nn^\prime}(x ) \gamma_5 c s^{t}_{jm^\prime}(x)c\right ] \right\ } \
, , \end{aligned}\ ] ] where the @xmath105 and @xmath106 are the full @xmath57 and @xmath103 quark propagators , respectively , @xmath107 @xmath108 and @xmath109 , the @xmath110 is the gell - mann matrix @xcite , then we compute all the integrals in the coordinate space and in the momentum space to obtain the correlation function @xmath102 at the quark level .
the calculations are straightforward but tedious .
all the correlation functions @xmath44 at the qcd side can be written as @xmath111 where we add the notation @xmath112 to denote the qcd side .
once the analytical expressions of the correlation functions @xmath113 are gotten , we can obtain the qcd spectral densities @xmath114 and @xmath115 through dispersion relation , @xmath116 the lengthy expressions of the qcd spectral densities @xmath114 and @xmath115 are given explicitly in the appendix .
we carry out the operator product expansion to the vacuum condensates up to dimension 10 , and assume vacuum saturation for the higher dimension vacuum condensates .
the gluon condensates are accompanied with large numerical denominators , their contributions to the total qcd spectral densities are less ( or much less ) than the contributions of the dimension 10 vacuum condensates @xmath117 @xcite , see eq.(26 ) , so we neglect the vacuum condensates involving the gluon condensate @xmath118 , @xmath119 , @xmath120 , @xmath121 , @xmath122 , @xmath123 , the predictive ability will not be impaired . in calculations ,
we take into account the contributions of the terms @xmath124 , @xmath125 , @xmath126 , @xmath127 , @xmath128 , @xmath129 and @xmath117 , where @xmath130 now we can take the quark - hadron duality below the continuum thresholds @xmath100 and obtain the following qcd sum rules : @xmath131\exp\left ( -\frac{s}{t^2}\right)\ , , \\ 2m_{p,+}\lambda^{+}_{p}{}^2\exp\left ( -\frac{m_{p,+}^2}{t^2}\right ) & = & \int_{4m_c^2}^{s_0}ds \left[\sqrt{s}\rho_{q_1q_2q_3}^{j_lj_h j,1}(s)-m_c\widetilde{\rho}_{q_1q_2q_3}^{j_lj_h j,0}(s)\right]\exp\left ( -\frac{s}{t^2}\right)\ , .\end{aligned}\ ] ] we differentiate eqs.(27 - 28 ) with respect to @xmath132 , then eliminate the pole residues @xmath133 and obtain the qcd sum rules for the masses of the hidden - charm pentaquark states @xmath134 , @xmath135\exp\left ( -\frac{s}{t^2}\right)}{\int_{4m_c^2}^{s_0}ds \left[\sqrt{s}\rho_{q_1q_2q_3}^{j_lj_h j,1}(s)+m_c\widetilde{\rho}_{q_1q_2q_3}^{j_lj_h j,0}(s)\right]\exp\left ( -\frac{s}{t^2}\right)}\ , , \\ m^2_{p,+ } & = & \frac{\int_{4m_c^2}^{s_0}ds \frac{d}{d(-1/t^2)}\left[\sqrt{s}\rho_{q_1q_2q_3}^{j_lj_h j,1}(s)-m_c\widetilde{\rho}_{q_1q_2q_3}^{j_lj_h j,0}(s)\right]\exp\left ( -\frac{s}{t^2}\right)}{\int_{4m_c^2}^{s_0}ds \left[\sqrt{s}\rho_{q_1q_2q_3}^{j_lj_h j,1}(s)-m_c\widetilde{\rho}_{q_1q_2q_3}^{j_lj_h j,0}(s)\right]\exp\left ( -\frac{s}{t^2}\right)}\ , .\end{aligned}\ ] ] we obtain the masses through fractions , see eqs.(29 - 30 ) , the contributions involving the gluon condensate in the numerator and denominator are canceled out with each other , the main effects of the gluon condensates can be safely absorbed into the pole residues @xmath133 and the remaining tiny effects on the masses can be safely neglected .
the input parameters are shown explicitly in table 1 .
the quark condensates , mixed quark condensates and @xmath136 masses evolve with the renormalization group equation , we take into account the energy - scale dependence according to the following equations , @xmath137^{\frac{4}{9}}\ , , \nonumber\\ \langle\bar{s}s \rangle(\mu)&=&\langle\bar{s}s \rangle(q)\left[\frac{\alpha_{s}(q)}{\alpha_{s}(\mu)}\right]^{\frac{4}{9}}\ , , \nonumber\\ \langle\bar{q}g_s \sigma gq \rangle(\mu)&=&\langle\bar{q}g_s \sigma gq \rangle(q)\left[\frac{\alpha_{s}(q)}{\alpha_{s}(\mu)}\right]^{\frac{2}{27}}\ , , \nonumber\\ \langle\bar{s}g_s \sigma gs \rangle(\mu)&=&\langle\bar{s}g_s \sigma gs \rangle(q)\left[\frac{\alpha_{s}(q)}{\alpha_{s}(\mu)}\right]^{\frac{2}{27}}\ , , \nonumber\\ m_c(\mu)&=&m_c(m_c)\left[\frac{\alpha_{s}(\mu)}{\alpha_{s}(m_c)}\right]^{\frac{12}{25 } } \ , , \nonumber\\ m_s(\mu)&=&m_s({\rm 2gev } ) \left[\frac{\alpha_{s}(\mu)}{\alpha_{s}({\rm 2gev})}\right]^{\frac{4}{9 } } \ , , \nonumber\\ \alpha_s(\mu)&=&\frac{1}{b_0t}\left[1-\frac{b_1}{b_0 ^ 2}\frac{\log t}{t } + \frac{b_1 ^ 2(\log^2{t}-\log{t}-1)+b_0b_2}{b_0 ^ 4t^2}\right]\ , , \end{aligned}\ ] ] where @xmath138 , @xmath139 , @xmath140 , @xmath141 , @xmath142 , @xmath143 and @xmath144 for the flavors @xmath145 , @xmath146 and @xmath147 , respectively @xcite .
furthermore , we set the @xmath148 . .
the input parameters in the qcd sum rules , the values in the bracket denote the energy scales @xmath149 , @xmath150 and @xmath151 , respectively . [ cols="^,^,^,^",options="header " , ] the @xmath152 can be well interpolated by the current @xmath153 @xcite , the @xmath55 and @xmath56 quarks in the @xmath152 form a scalar diquark @xmath154 $ ] ( or @xmath155 ) in color antitriplet , the decays @xmath156 take place through the following mechanism , @xmath157 b ) & \to & [ ud ] c\bar{c}s\to [ ud ] c\bar{c}u\bar{u}s\to p_c^+([ud ] [ uc]\bar{c})k^-(\bar{u}s)\to j/\psi p k^- \ , , \end{aligned}\ ] ] at the quark level . in the decays @xmath158 [ uc]\bar{c})\to j/\psi p$ ] , the scalar diquark @xmath154 $ ] survives in the decay chains , the decays are greatly facilitated and can take place easily . on the other hand , if there exists a light axialvector diquark @xmath159 ( or @xmath160 ) , which has to dissolve to form a scalar diquark @xmath154 $ ] , the decays are not facilitated and can not take place easily , but the decays are not forbidden .
it is also possible to assign the @xmath6 to be the axialvector - diquark - scalar - diquark - antiquark type pentaquark state @xmath2 or the axialvector - diquark - axialvector - diquark - antiquark type pentaquark states @xmath3 and @xmath4 . in fig.1
, we plot the masses of the pentaquark states @xmath2 , @xmath161 , @xmath3 , @xmath162 , @xmath4 and @xmath163 with variations of the borel parameters @xmath99 as an example . in this article , we take the isospin limit , the masses of the pentaquark states @xmath164 and @xmath165 degenerate . from the figure , we can see that the platforms in the borel windows shown in table 2 appear at the minimum values for the axialvector - diquark - axialvector - diquark - antiquark type pentaquark states with positive parity , and the borel platforms are very flat ; other platforms in the borel windows shown table 2 are not as flat , the uncertainties @xmath166 induced by the borel parameters in the borel windows are about @xmath167 . in fig.1 , we plot the masses with variations of the borel parameters at a large interval @xmath168 , which is much larger than the interval of the borel windows shown in table 2 . compared to the values shown in table 3 ,
the masses shown in fig.1 have larger uncertainties due to the larger interval of the borel parameters .
furthermore , from the values in table 3 , we can see that the mass gaps between the pentaquark states with opposite parity @xmath169 and @xmath170 are about @xmath171 and @xmath172 respectively , which is much larger than the mass gap @xmath173 between the pentaquark states with opposite parity @xmath174 .
an additional p - wave costs about @xmath171 for the conventional baryon states , the mass spectra of the pentaquark states have new feature . , where the @xmath175 , @xmath176 , @xmath74 , @xmath177 , @xmath178 and @xmath179 denote the pentaquark states @xmath2 , @xmath161 , @xmath3 , @xmath162 , @xmath4 and @xmath163 respectively , the horizontal lines denote the experimental value of the mass of the @xmath6 . , title="fig:",width=264 ] , where the @xmath175 , @xmath176 , @xmath74 , @xmath177 , @xmath178 and @xmath179 denote the pentaquark states @xmath2 , @xmath161 , @xmath3 , @xmath162 , @xmath4 and @xmath163 respectively , the horizontal lines denote the experimental value of the mass of the @xmath6 . ,
title="fig:",width=264 ] , where the @xmath175 , @xmath176 , @xmath74 , @xmath177 , @xmath178 and @xmath179 denote the pentaquark states @xmath2 , @xmath161 , @xmath3 , @xmath162 , @xmath4 and @xmath163 respectively , the horizontal lines denote the experimental value of the mass of the @xmath6 . ,
title="fig:",width=264 ] , where the @xmath175 , @xmath176 , @xmath74 , @xmath177 , @xmath178 and @xmath179 denote the pentaquark states @xmath2 , @xmath161 , @xmath3 , @xmath162 , @xmath4 and @xmath163 respectively , the horizontal lines denote the experimental value of the mass of the @xmath6 . ,
title="fig:",width=264 ] , where the @xmath175 , @xmath176 , @xmath74 , @xmath177 , @xmath178 and @xmath179 denote the pentaquark states @xmath2 , @xmath161 , @xmath3 , @xmath162 , @xmath4 and @xmath163 respectively , the horizontal lines denote the experimental value of the mass of the @xmath6 . ,
title="fig:",width=264 ] , where the @xmath175 , @xmath176 , @xmath74 , @xmath177 , @xmath178 and @xmath179 denote the pentaquark states @xmath2 ,
@xmath161 , @xmath3 , @xmath162 , @xmath4 and @xmath163 respectively , the horizontal lines denote the experimental value of the mass of the @xmath6 . , title="fig:",width=264 ]
in this article , we construct the axialvector - diquark - scalar - diquark - antiquark type and the axialvector - diquark - axialvector - diquark - antiquark type currents to interpolate the @xmath0 hidden - charm pentaquark states , study the masses and pole residues with the qcd sum rules systematically by calculating the contributions of the vacuum condensates up to dimension 10 in the operator product expansion .
we obtain the masses of the hidden - charm pentaquark states with the strangeness @xmath1 , respectively , which can be confronted to the experimental data in the future . as a byproduct
, the predicted pole residues can be taken as basic input parameters in studying the strong decays of the hidden - charm pentaquark states with the three - point qcd sum rules or the light - cone qcd sum rules . in calculations ,
we use the empirical energy scale formula @xmath180 with the effective heavy quark mass @xmath181 to determine the ideal energy scales of the qcd spectral densities , which works well in studying the hidden - charm tetraquark states and hidden - charm pentaquark states .
the predicted masses of the pentaquark states @xmath2 , @xmath3 and @xmath4 are compatible with the experimental value @xmath5 from the lhcb collaboration
. we can draw the conclusion tentatively that the @xmath6 can be assigned to be the pentaquark state @xmath2 , @xmath3 or @xmath4 , or the @xmath6 has some pentaquark components @xmath2 , @xmath3 and @xmath4 .
more experimental data on its productions and decays are still needed to identify the @xmath6 unambiguously .
the qcd spectral densities @xmath182 , @xmath183 , @xmath184 , @xmath185 , @xmath186 , @xmath187 , @xmath188 , @xmath189 , @xmath190 , @xmath191 , @xmath192 , @xmath193 , @xmath194 , @xmath195 , @xmath196 , @xmath197 , @xmath198 , @xmath199 , @xmath200 , @xmath201 , @xmath202 , @xmath203 , @xmath204 and @xmath205 of the hidden - charm pentaquark states , @xmath206 \nonumber\\ & & + \frac{11\langle\bar{s}g_s\sigma gs\rangle^2}{1536\pi^4}\int dy \ , y(1-y)\left [ 1+\frac{s}{2}\delta\left(s-\widetilde{m}_c^2 \right)\right ] \nonumber\\ & & + \frac{m_s m_c}{12288\pi^8}\int dydz \ , ( y+z)(1-y - z)^3\left(s-\overline{m}_c^2\right)^4 \nonumber\end{aligned}\ ] ] @xmath207 \nonumber\\ & & + \frac{m_s m_c\langle\bar{s}s\rangle\langle\bar{s}g_s\sigma gs\rangle}{768\pi^4}\int dydz \left(\frac{y}{z}+\frac{z}{y}\right)\nonumber\\ & & -\frac{199m_s m_c \langle\bar{s}s\rangle\langle\bar{s}g_s\sigma gs\rangle}{9216\pi^4}\int dy \nonumber\\ & & + \frac{m_s\langle\bar{s}s\rangle^3}{18\pi^2}\int dy \,y(1-y ) \left [ 1+\frac{s}{2}\delta\left(s-\widetilde{m}_c^2 \right)\right ] \nonumber\\ & & -\frac{m_s m_c\langle\bar{s}g_s\sigma gs\rangle^2}{3072\pi^4}\int dy \left(\frac{1-y}{y}+\frac{y}{1-y}\right)\delta \left(s-\widetilde{m}_c^2 \right)\nonumber\\ & & + \frac{17m_s m_c\langle\bar{s}g_s\sigma gs\rangle^2}{4608\pi^4}\int dy \left(1+\frac{s}{2t^2}\right)\delta\left(s-\widetilde{m}_c^2 \right ) \ , , \end{aligned}\ ] ]
@xmath208 @xmath209 \nonumber\\ & & + \frac{m_s m_c}{3072\pi^8}\int dydz \ , ( 1-y - z)^3\left(s-\overline{m}_c^2\right)^4 \nonumber\\ & & -\frac{m_s m_c}{12288\pi^8}\int dydz \ , ( 1-y - z)^4\left(s-\overline{m}_c^2\right)^3\left(3s-\overline{m}_c^2\right ) \nonumber\\ & & - \frac{m_s\langle \bar{s}s\rangle}{768\pi^6}\int dydz \ , ( y+z)(1-y - z)^2\left(s-\overline{m}_c^2\right)^2 \left(4s-\overline{m}_c^2 \right ) \nonumber \\ & & - \frac{m_s\langle \bar{s}s\rangle}{512\pi^6}\int dydz \ , ( y+z)(1-y - z)^3\left(s-\overline{m}_c^2\right ) \left(8s^2 - 7s\overline{m}_c^2+\overline{m}_c^4 \right ) \nonumber \\ & & + \frac{11m_s\langle \bar{s}g_s\sigma gs\rangle}{2048\pi^6}\int dydz \ , ( y+z)(1-y - z ) \left(s-\overline{m}_c^2 \right)\left(3s-\overline{m}_c^2
\right ) \nonumber\\ & & + \frac{m_s\langle \bar{s}g_s\sigma gs\rangle}{1024\pi^6}\int dydz \ , ( y+z)(1-y - z)^2 \left(15s^2 - 16s\overline{m}_c^2 + 3\overline{m}_c^4\right ) \nonumber\\ & & + \frac { m_s m_c\langle\bar{s}s\rangle^2}{6\pi^4}\int dydz \
, \left(s-\overline{m}_c^2 \right ) \nonumber\\ & & + \frac { m_s m_c\langle\bar{s}s\rangle^2}{24\pi^4}\int dydz \ , ( 1-y - z)\left(3s-2\overline{m}_c^2 \right ) \nonumber\\ & & -\frac{89m_s m_c\langle\bar{s}s\rangle\langle\bar{s}g_s\sigma gs\rangle}{2304\pi^4}\int dydz \ ,
\left[1+\frac{s}{2}\delta \left(s-\overline{m}_c^2 \right ) \right]\nonumber\\ & & -\frac{199m_s m_c \langle\bar{s}s\rangle\langle\bar{s}g_s\sigma gs\rangle}{2304\pi^4}\int dy \nonumber\\ & & + \frac{m_s\langle\bar{s}s\rangle^3}{36\pi^2}\int dy \ , \left [ 1+s\,\delta\left(s-\widetilde{m}_c^2 \right)\right ] \nonumber\\ & & + \frac{125m_s m_c\langle\bar{s}g_s\sigma gs\rangle^2}{9216\pi^4}\int dy \left(1+\frac{s}{t^2}\right)\delta\left(s-\widetilde{m}_c^2 \right ) \ , , \end{aligned}\ ] ] @xmath210}{2304\pi^6}\int dydz \ , ( y+z)(1-y - z)^2\left(s-\overline{m}_c^2\right)^3 \nonumber \\ & & - \frac{m_c\left[\langle \bar{q}q\rangle+2\langle \bar{s}s\rangle\right]}{2304\pi^6}\int dydz \ , ( y+z)(1-y - z)^3\left(s-\overline{m}_c^2\right)^2\left(2s-\overline{m}_c^2\right ) \nonumber \\ & & + \frac{11m_c\left[\langle \bar{q}g_s\sigma gq\rangle+2\langle \bar{s}g_s\sigma gs\rangle\right]}{12288\pi^6}\int dydz \ , ( y+z)(1-y - z ) \left(s-\overline{m}_c^2
\right)^2 \nonumber\\ & & + \frac{11m_c\left[\langle \bar{q}g_s\sigma gq\rangle+2\langle \bar{s}g_s\sigma gs\rangle\right]}{24576\pi^6}\int dydz \ , ( y+z)(1-y - z)^2 \left(s-\overline{m}_c^2 \right)\left(5s-3\overline{m}_c^2 \right ) \nonumber\\ & & + \frac{\langle\bar{s}s\rangle\left[2\langle\bar{q}q\rangle+\langle\bar{s}s\rangle \right]}{72\pi^4}\int dydz \ , yz(1-y - z)\left(s-\overline{m}_c^2 \right)\left(2s-\overline{m}_c^2 \right ) \nonumber\end{aligned}\ ] ] @xmath211}{2304\pi^4}\int dydz \,yz \left(3s-2\overline{m}_c^2\right)\nonumber\\ & & + \frac{\langle\bar{s}s\rangle\langle\bar{s}g_s\sigma gs\rangle+\langle\bar{q}q\rangle\langle\bar{s}g_s\sigma gs\rangle+\langle\bar{s}s\rangle\langle\bar{q}g_s\sigma gq\rangle}{4608\pi^4}\int dydz \,(y+z)(1-y - z ) \left(5s-4\overline{m}_c^2\right)\nonumber\\ & & -\frac{m_c\langle\bar{q}q\rangle\langle\bar{s}s\rangle^2}{36\pi^2}\int dy \nonumber\\ & & -\frac{13\langle\bar{s}g_s\sigma gs\rangle\left[2 \langle\bar{q}g_s\sigma gq\rangle+\langle\bar{s}g_s\sigma gs\rangle\right]}{41472\pi^4}\int dydz \ , ( y+z)\left [ 1+\frac{s}{4}\delta\left(s-\overline{m}_c^2 \right)\right ] \nonumber\\ & & + \frac{11\langle\bar{s}g_s\sigma gs\rangle \left [ 2\langle\bar{q}g_s\sigma gq\rangle+\langle\bar{s}g_s\sigma gs\rangle\right]}{4608\pi^4}\int dy \ , y(1-y)\left [ 1+\frac{s}{2}\delta\left(s-\widetilde{m}_c^2 \right)\right ] \nonumber\\ & & + \frac{m_s m_c}{18432\pi^8}\int dydz \ , ( y+z)(1-y - z)^3\left(s-\overline{m}_c^2\right)^4 \nonumber\\ & & + \frac{m_s m_c}{73728\pi^8}\int dydz \ , ( y+z)(1-y - z)^4\left(s-\overline{m}_c^2\right)^3\left(7s-3\overline{m}_c^2\right ) \nonumber\\ & & - \frac{m_s\left[2\langle \bar{q}q\rangle-\langle \bar{s}s\rangle\right]}{1152\pi^6}\int dydz \ , yz(1-y - z)^2\left(s-\overline{m}_c^2\right)^2 \left(5s-2\overline{m}_c^2 \right ) \nonumber \\ & & + \frac{m_s\langle \bar{s}s\rangle}{1152\pi^6}\int dydz \ , yz(1-y - z)^3\left(s-\overline{m}_c^2\right ) \left(35s^2 - 37s\overline{m}_c^2 + 8\overline{m}_c^4 \right ) \nonumber\\ & & - \frac{m_s\langle \bar{s}g_s\sigma gs\rangle}{384\pi^6}\int dydz \ , yz(1-y - z)^2 \left(12s^2 - 15s\overline{m}_c^2 + 4\overline{m}_c^4\right )
\nonumber\\ & & - \frac{m_s\left[\langle \bar{q}g_s\sigma gq\rangle+\langle \bar{s}g_s\sigma gs\rangle\right]}{12288\pi^6}\int dydz \ , ( y+z)(1-y - z)^2 \left(s-\overline{m}_c^2 \right)\left(3s-2\overline{m}_c^2 \right ) \nonumber\\ & & + \frac{m_s\left[19\langle \bar{q}g_s\sigma gq\rangle+3\langle \bar{s}g_s\sigma gs\rangle\right]}{3072\pi^6}\int dydz \ , yz(1-y - z ) \left(s-\overline{m}_c^2 \right)\left(2s-\overline{m}_c^2
\right ) \nonumber\\ & & + \frac { m_s m_c\langle\bar{s}s\rangle\left [ 5\langle\bar{q}q\rangle-\langle\bar{s}s\rangle\right]}{144\pi^4}\int dydz \ ,
( y+z)\left(s-\overline{m}_c^2 \right ) \nonumber\\ & & -\frac { m_s m_c\langle\bar{s}s\rangle\left[\langle\bar{q}q\rangle+\langle\bar{s}s\rangle\right ] } { 144\pi^4}\int dydz \ , ( y+z)(1-y - z)\left(4s-3\overline{m}_c^2 \right ) \nonumber\\ & & + \frac{m_s m_c\left[89\langle\bar{s}s\rangle\langle\bar{s}g_s\sigma gs\rangle+57\langle\bar{s}s\rangle\langle\bar{q}g_s\sigma gq\rangle+32\langle\bar{q}q\rangle\langle\bar{s}g_s\sigma gs\rangle\right]}{9216\pi^4 } \nonumber\\ & & \int dydz \ ,
\left(y+z\right)\left[1+\frac{s}{3}\delta \left(s-\overline{m}_c^2 \right ) \right]\nonumber\\ & & -\frac { m_s m_c \left[231\langle\bar{s}s\rangle\langle\bar{q}g_s\sigma gq\rangle-89\langle\bar{s}s\rangle\langle\bar{s}g_s\sigma gs\rangle+256\langle\bar{q}q\rangle\langle\bar{s}g_s\sigma gs\rangle\right]}{27648\pi^4}\int dy \nonumber\\ & & + \frac{m_s m_c\left[\langle\bar{q}q\rangle\langle\bar{s}g_s\sigma gs\rangle+\langle\bar{s}s\rangle\langle\bar{q}g_s\sigma gq\rangle\right]}{2304\pi^4}\int dydz
\left(\frac{y}{z}+\frac{z}{y}\right)\nonumber\\ & & + \frac{m_s\langle\bar{q}q\rangle\langle\bar{s}s\rangle^2}{27\pi^2}\int dy \,y(1-y ) \left [ 1+\frac{s}{2}\delta\left(s-\widetilde{m}_c^2 \right)\right ] \nonumber\\ & & -\frac{m_s m_c\langle\bar{q}g_s\sigma gq\rangle\langle\bar{s}g_s\sigma gs\rangle } { 4608\pi^4}\int dy \left(\frac{1-y}{y}+\frac{y}{1-y}\right)\delta \left(s-\widetilde{m}_c^2 \right)\nonumber\\ & & -\frac{m_s m_c\langle\bar{s}g_s\sigma gs\rangle\left[19\langle \bar{s}g_s\sigma gs\rangle-53\langle \bar{q}g_s\sigma gq\rangle\right]}{13824\pi^4}\int dy \left ( 1+\frac{s}{2t^2}\right ) \delta\left(s-\widetilde{m}_c^2 \right)\ , , \end{aligned}\ ] ] @xmath212}{576\pi^6}\int dydz \ , ( 1-y - z)^2\left(s-\overline{m}_c^2\right)^3 \nonumber \\ & & + \frac{m_c\left[\langle \bar{q}q\rangle+2\langle \bar{s}s\rangle\right]}{3456\pi^6}\int dydz \ , ( 1-y - z)^3\left(s-\overline{m}_c^2\right)^2\left(5s-2\overline{m}_c^2\right ) \nonumber \\ & & + \frac{11m_c\left[\langle \bar{q}g_s\sigma gq\rangle+2\langle \bar{s}g_s\sigma gs\rangle\right]}{3072\pi^6}\int dydz \ , ( 1-y - z ) \left(s-\overline{m}_c^2
\right)^2 \nonumber\\ & & - \frac{11m_c\left[\langle \bar{q}g_s\sigma gq\rangle+2\langle \bar{s}g_s\sigma gs\rangle\right]}{6144\pi^6}\int dydz \
, ( 1-y - z)^2 \left(s-\overline{m}_c^2 \right)\left(2s-\overline{m}_c^2 \right ) \nonumber\\ & & + \frac{\langle\bar{s}s\rangle\left[2\langle\bar{q}q\rangle+\langle\bar{s}s\rangle \right]}{144\pi^4}\int dydz \ , ( y+z)(1-y - z)\left(s-\overline{m}_c^2 \right)\left(3s-\overline{m}_c^2 \right ) \nonumber\\ & & -\frac{19\left[\langle\bar{s}s\rangle\langle\bar{s}g_s\sigma gs\rangle+\langle\bar{q}q\rangle\langle\bar{s}g_s\sigma gs\rangle+\langle\bar{s}s\rangle\langle\bar{q}g_s\sigma gq\rangle\right]}{2304\pi^4}\int dydz \,(y+z ) \left(2s-\overline{m}_c^2\right)\nonumber\\ & & -\frac{m_c\langle\bar{q}q\rangle\langle\bar{s}s\rangle^2}{9\pi^2}\int dy \nonumber\\ & & + \frac{11\langle\bar{s}g_s\sigma gs\rangle\left[2\langle\bar{q}g_s\sigma gq\rangle+\langle\bar{s}g_s\sigma gs\rangle \right]}{9216\pi^4}\int dy \left [ 1+s\,\delta\left(s-\widetilde{m}_c^2 \right)\right ] \nonumber\\ & & + \frac{m_s m_c}{4608\pi^8}\int dydz \ , ( 1-y - z)^3\left(s-\overline{m}_c^2\right)^4 \nonumber\\ & & -\frac{m_s m_c}{18432\pi^8}\int dydz \ , ( 1-y - z)^4\left(s-\overline{m}_c^2\right)^3\left(3s-\overline{m}_c^2\right ) \nonumber\\ & & - \frac{m_s\left[2\langle \bar{q}q\rangle-\langle \bar{s}s\rangle\right]}{1152\pi^6}\int dydz \ , ( y+z)(1-y - z)^2\left(s-\overline{m}_c^2\right)^2 \left(4s-\overline{m}_c^2 \right ) \nonumber \\ & & - \frac{m_s\langle \bar{s}s\rangle}{768\pi^6}\int dydz \ , ( y+z)(1-y - z)^3\left(s-\overline{m}_c^2\right ) \left(8s^2 - 7s\overline{m}_c^2+\overline{m}_c^4 \right ) \nonumber \\ & & + \frac{m_s\langle \bar{s}g_s\sigma gs\rangle}{1536\pi^6}\int dydz \ , ( y+z)(1-y - z)^2 \left(15s^2 - 16s\overline{m}_c^2 + 3\overline{m}_c^4\right ) \nonumber\\ & & + \frac { m_s\left[19\langle \bar{q}g_s\sigma gq\rangle+3\langle \bar{s}g_s\sigma gs\rangle\right]}{6144\pi^6}\int dydz \ , ( y+z)(1-y - z ) \left(s-\overline{m}_c^2 \right)\left(3s-\overline{m}_c^2
\right ) \nonumber\\ & & + \frac { m_s m_c\langle\bar{s}s\rangle\left [ 5\langle\bar{q}q\rangle-\langle\bar{s}s\rangle\right]}{36\pi^4}\int dydz \ , \left(s-\overline{m}_c^2 \right ) \nonumber\\ & & + \frac { m_s m_c\langle\bar{s}s\rangle\left[\langle\bar{q}q\rangle+\langle\bar{s}s\rangle \right]}{72\pi^4}\int dydz \ , ( 1-y - z)\left(3s-2\overline{m}_c^2 \right ) \nonumber\\ & & -\frac{m_s m_c\left[89\langle\bar{s}s\rangle\langle\bar{s}g_s\sigma gs\rangle+57\langle\bar{s}s\rangle\langle\bar{q}g_s\sigma gq\rangle+32\langle\bar{q}q\rangle\langle\bar{s}g_s\sigma gs\rangle\right]}{6912\pi^4}\int dydz \ ,
\left[1+\frac{s}{2}\delta \left(s-\overline{m}_c^2 \right ) \right]\nonumber\\ & & -\frac{m_s m_c\left [ 231\langle\bar{s}s\rangle\langle\bar{q}g_s\sigma gq\rangle-89\langle\bar{s}s\rangle\langle\bar{s}g_s\sigma gs\rangle+256\langle\bar{q}q\rangle\langle\bar{s}g_s\sigma gs\rangle\right]}{6912\pi^4}\int dy
\nonumber\\ & & + \frac{m_s\langle\bar{q}q\rangle\langle\bar{s}s\rangle^2}{54\pi^2}\int dy \ , \left [ 1+s\,\delta\left(s-\widetilde{m}_c^2 \right)\right ] \nonumber\\ & & -\frac{m_s m_c\langle\bar{s}g_s\sigma gs\rangle\left[19\langle \bar{s}g_s\sigma gs\rangle-269\langle \bar{q}g_s\sigma gq\rangle\right]}{27648\pi^4 } \int dy \left ( 1+\frac{s}{t^2}\right ) \delta\left(s-\widetilde{m}_c^2 \right ) \ , , \end{aligned}\ ] ] @xmath213}{2304\pi^6}\int dydz \ , ( y+z)(1-y - z)^2\left(s-\overline{m}_c^2\right)^3 \nonumber \\ & & - \frac{m_c\left[2\langle \bar{q}q\rangle+\langle \bar{s}s\rangle\right]}{2304\pi^6}\int dydz \ , ( y+z)(1-y - z)^3\left(s-\overline{m}_c^2\right)^2\left(2s-\overline{m}_c^2\right ) \nonumber \\ & & + \frac{11m_c\left[2\langle \bar{q}g_s\sigma gq\rangle+\langle \bar{s}g_s\sigma gs\rangle\right]}{12288\pi^6}\int dydz \ , ( y+z)(1-y - z ) \left(s-\overline{m}_c^2 \right)^2 \nonumber\\ & & + \frac{11m_c\left[2\langle \bar{q}g_s\sigma gq\rangle+\langle \bar{s}g_s\sigma gs\rangle\right]}{24576\pi^6}\int dydz \ , ( y+z)(1-y - z)^2 \left(s-\overline{m}_c^2 \right)\left(5s-3\overline{m}_c^2 \right ) \nonumber\\ & & + \frac{\langle\bar{q}q\rangle \left[\langle\bar{q}q\rangle+2\langle\bar{s}s\rangle \right]}{72\pi^4}\int dydz \ , yz(1-y - z)\left(s-\overline{m}_c^2 \right)\left(2s-\overline{m}_c^2 \right ) \nonumber\\ & & -\frac{19\left[\langle\bar{q}q\rangle\langle\bar{q}g_s\sigma gq\rangle+\langle\bar{q}q\rangle\langle\bar{s}g_s\sigma gs\rangle+\langle\bar{s}s\rangle\langle\bar{q}g_s\sigma gq\rangle\right]}{2304\pi^4}\int dydz \,yz \left(3s-2\overline{m}_c^2\right)\nonumber\\ & & + \frac{\langle\bar{q}q\rangle\langle\bar{q}g_s\sigma gq\rangle+\langle\bar{q}q\rangle\langle\bar{s}g_s\sigma gs\rangle+\langle\bar{s}s\rangle\langle\bar{q}g_s\sigma gq\rangle}{4608\pi^4}\int dydz \,(y+z)(1-y - z ) \left(5s-4\overline{m}_c^2\right)\nonumber\\ & & -\frac{13\langle\bar{q}g_s\sigma gq\rangle \left[\langle\bar{q}g_s\sigma
gq\rangle+2\langle\bar{s}g_s\sigma gs\rangle \right]}{41472\pi^4}\int dydz \ , ( y+z)\left [ 1+\frac{s}{4}\delta\left(s-\overline{m}_c^2 \right)\right ] \nonumber\\ & & + \frac{11\langle\bar{q}g_s\sigma gq\rangle\left[\langle\bar{q}g_s\sigma gq\rangle+2\langle\bar{s}g_s\sigma gs\rangle \right]}{4608\pi^4}\int dy \ , y(1-y)\left [ 1+\frac{s}{2}\delta\left(s-\widetilde{m}_c^2 \right)\right ] \nonumber\\ & & -\frac{m_c\langle\bar{q}q\rangle^2\langle\bar{s}s\rangle}{36\pi^2}\int dy \nonumber\\ & & + \frac{m_s m_c}{36864\pi^8}\int dydz \ , ( y+z)(1-y - z)^3\left(s-\overline{m}_c^2\right)^4 \nonumber\\ & & + \frac{m_s m_c}{147456\pi^8}\int dydz \ , ( y+z)(1-y - z)^4\left(s-\overline{m}_c^2\right)^3\left(7s-3\overline{m}_c^2\right ) \nonumber\\ & & - \frac{m_s\left[4\langle \bar{q}q\rangle-3\langle \bar{s}s\rangle\right]}{2304\pi^6}\int dydz \ , yz(1-y - z)^2\left(s-\overline{m}_c^2\right)^2 \left(5s-2\overline{m}_c^2 \right ) \nonumber \\ & & + \frac{m_s\langle \bar{s}s\rangle}{2304\pi^6}\int dydz \ , yz(1-y - z)^3\left(s-\overline{m}_c^2\right ) \left(35s^2 - 37s\overline{m}_c^2 + 8\overline{m}_c^4 \right ) \nonumber \\ & & + \frac{m_s\left[19\langle \bar{q}g_s\sigma gq\rangle-8\langle \bar{s}g_s\sigma gs\rangle\right]}{3072\pi^6}\int dydz \ , yz(1-y - z ) \left(s-\overline{m}_c^2
\right)\left(2s-\overline{m}_c^2 \right ) \nonumber\\ & & - \frac{m_s\langle \bar{s}g_s\sigma gs\rangle}{768\pi^6}\int dydz \ , yz(1-y - z)^2 \left(12s^2 - 15s\overline{m}_c^2 + 4\overline{m}_c^4\right ) \nonumber\\ & & - \frac{m_s\langle \bar{q}g_s\sigma gq\rangle}{12288\pi^6}\int dydz \ , ( y+z)(1-y - z)^2 \left(s-\overline{m}_c^2 \right)\left(3s-2\overline{m}_c^2 \right ) \nonumber\\ & & + \frac { m_s m_c\langle\bar{q}q\rangle\left [ 3\langle\bar{q}q\rangle-\langle\bar{s}s\rangle\right]}{144\pi^4}\int dydz \ , ( y+z)\left(s-\overline{m}_c^2 \right ) \nonumber\\ & & -\frac { m_s m_c\langle\bar{q}q\rangle\langle\bar{s}s\rangle } { 144\pi^4}\int dydz \ , ( y+z)(1-y - z)\left(4s-3\overline{m}_c^2 \right ) \nonumber\end{aligned}\ ] ] @xmath214}{9216\pi^4}\int dydz \ ,
\left(y+z\right)\left[1+\frac{s}{3}\delta \left(s-\overline{m}_c^2 \right ) \right]\nonumber\\ & & -\frac{m_s m_c \left[288\langle\bar{q}q\rangle\langle\bar{q}g_s\sigma gq\rangle-57\langle\bar{s}s\rangle\langle\bar{q}g_s\sigma gq\rangle-32\langle\bar{q}q\rangle\langle\bar{s}g_s\sigma gs\rangle\right]}{27648\pi^4}\int dy \nonumber\\ & & + \frac{m_s m_c\langle\bar{q}q\rangle\langle\bar{q}g_s\sigma gq\rangle}{2304\pi^4}\int dydz \left(\frac{y}{z}+\frac{z}{y}\right)\nonumber\\ & & + \frac{m_s\langle\bar{q}q\rangle^2\langle\bar{s}s\rangle}{54\pi^2}\int dy \,y(1-y ) \left [ 1+\frac{s}{2}\delta\left(s-\widetilde{m}_c^2 \right)\right ] \nonumber\\ & & -\frac{m_s m_c\langle\bar{q}g_s\sigma gq\rangle^2}{9216\pi^4}\int dy \left(\frac{1-y}{y}+\frac{y}{1-y}\right)\delta \left(s-\widetilde{m}_c^2 \right)\nonumber\\ & & + \frac{m_s m_c\langle\bar{q}g_s\sigma gq\rangle\left[36\langle\bar{q}g_s\sigma gq\rangle-19\langle\bar{s}g_s\sigma gs\rangle\right]}{13824\pi^4}\int dy \left ( 1+\frac{s}{2t^2}\right ) \delta\left(s-\widetilde{m}_c^2 \right ) \ , , \end{aligned}\ ] ] @xmath215}{576\pi^6}\int dydz \ , ( 1-y - z)^2\left(s-\overline{m}_c^2\right)^3 \nonumber \\ & & + \frac{m_c\left[2\langle \bar{q}q\rangle+\langle \bar{s}s\rangle\right]}{3456\pi^6}\int dydz \ , ( 1-y - z)^3\left(s-\overline{m}_c^2\right)^2\left(5s-2\overline{m}_c^2\right ) \nonumber \\ & & + \frac{11m_c\left[2\langle \bar{q}g_s\sigma gq\rangle+\langle \bar{s}g_s\sigma gs\rangle\right]}{3072\pi^6}\int dydz \ , ( 1-y - z ) \left(s-\overline{m}_c^2
\right)^2 \nonumber\\ & & - \frac{11m_c\left[2\langle \bar{q}g_s\sigma gq\rangle+\langle \bar{s}g_s\sigma gs\rangle\right]}{6144\pi^6}\int dydz \
, ( 1-y - z)^2 \left(s-\overline{m}_c^2 \right)\left(2s-\overline{m}_c^2 \right ) \nonumber \\ & & + \frac{\langle\bar{q}q\rangle \left [ \langle\bar{q}q\rangle+2\langle\bar{s}s\rangle\right]}{144\pi^4}\int dydz \
, ( y+z)(1-y - z)\left(s-\overline{m}_c^2 \right)\left(3s-\overline{m}_c^2 \right ) \nonumber\\ & & -\frac{19\left[\langle\bar{q}q\rangle\langle\bar{q}g_s\sigma gq\rangle+\langle\bar{q}q\rangle\langle\bar{s}g_s\sigma gs\rangle+\langle\bar{s}s\rangle\langle\bar{q}g_s\sigma gq\rangle\right]}{2304\pi^4}\int dydz \,(y+z ) \left(2s-\overline{m}_c^2\right)\nonumber\\ & & -\frac{m_c\langle\bar{q}q\rangle^2\langle\bar{s}s\rangle}{9\pi^2}\int dy \nonumber\\ & & + \frac{11\langle\bar{q}g_s\sigma gq\rangle \left [ \langle\bar{q}g_s\sigma gq\rangle+2\langle\bar{s}g_s\sigma gs\rangle\right]}{9216\pi^4}\int dy \left [ 1+s\,\delta\left(s-\widetilde{m}_c^2 \right)\right ] \nonumber\\ & & + \frac{m_s m_c}{9216\pi^8}\int dydz \ , ( 1-y - z)^3\left(s-\overline{m}_c^2\right)^4 \nonumber\\ & & -\frac{m_s m_c}{36864\pi^8}\int dydz \ ,
( 1-y - z)^4\left(s-\overline{m}_c^2\right)^3\left(3s-\overline{m}_c^2\right ) \nonumber\\ & & - \frac{m_s\left[4\langle \bar{q}q\rangle-3\langle \bar{s}s\rangle\right]}{2304\pi^6}\int dydz \ , ( y+z)(1-y - z)^2\left(s-\overline{m}_c^2\right)^2 \left(4s-\overline{m}_c^2 \right )
\nonumber \\ & & - \frac{m_s\langle \bar{s}s\rangle}{1536\pi^6}\int dydz \ , ( y+z)(1-y - z)^3\left(s-\overline{m}_c^2\right ) \left(8s^2 - 7s\overline{m}_c^2+\overline{m}_c^4 \right ) \nonumber \\ & & + \frac{m_s\left[19\langle \bar{q}g_s\sigma gq\rangle-8\langle \bar{s}g_s\sigma gs\rangle\right]}{6144\pi^6}\int dydz \ , ( y+z)(1-y - z ) \left(s-\overline{m}_c^2 \right)\left(3s-\overline{m}_c^2
\right ) \nonumber\end{aligned}\ ] ] @xmath216}{36\pi^4}\int dydz \
, \left(s-\overline{m}_c^2 \right ) \nonumber\\ & & + \frac { m_s m_c\langle\bar{q}q\rangle\langle\bar{s}s\rangle } { 72\pi^4}\int dydz \ ,
( 1-y - z)\left(3s-2\overline{m}_c^2 \right ) \nonumber\\ & & -\frac{m_s m_c\left[32\langle\bar{q}q\rangle\langle\bar{s}g_s\sigma gs\rangle+57\langle\bar{s}s\rangle\langle\bar{q}g_s\sigma gq\rangle\right]}{6912\pi^4}\int dydz \ ,
\left[1+\frac{s}{2}\delta \left(s-\overline{m}_c^2 \right ) \right]\nonumber\\ & & -\frac{m_s m_c \left[288\langle\bar{q}q\rangle\langle\bar{q}g_s\sigma gq\rangle-57\langle\bar{s}s\rangle\langle\bar{q}g_s\sigma gq\rangle-32\langle\bar{q}q\rangle\langle\bar{s}g_s\sigma gs\rangle\right]}{6912\pi^4}\int dy \nonumber\\ & & + \frac{m_s\langle\bar{q}q\rangle^2 \langle\bar{s}s\rangle}{108\pi^2}\int dy \ , \left [ 1+s\,\delta\left(s-\widetilde{m}_c^2 \right)\right ] \nonumber\\ & & + \frac{m_s m_c\langle\bar{q}g_s\sigma gq\rangle \left [ 144\langle\bar{q}g_s\sigma gq\rangle-19\langle\bar{s}g_s\sigma gs\rangle\right]}{27648\pi^4 } \int dy \left ( 1+\frac{s}{t^2}\right ) \delta\left(s-\widetilde{m}_c^2 \right)\ , , \end{aligned}\ ] ] @xmath217 @xmath218 \nonumber\\ & & + \frac{m_s m_c}{12288\pi^8}\int dydz \ , ( y+z)(1-y - z)^3\left(s-\overline{m}_c^2\right)^4 \nonumber\\ & & -\frac{m_s m_c}{49152\pi^8}\int dydz \ , ( y+z)(1-y - z)^4\left(s-\overline{m}_c^2\right)^4 \nonumber\\ & & - \frac{m_s\langle \bar{s}s\rangle}{128\pi^6}\int dydz \
, yz(1-y - z)^2\left(s-\overline{m}_c^2\right)^2 \left(2s-\overline{m}_c^2 \right ) \nonumber\end{aligned}\ ] ] @xmath219 \nonumber\\ & & + \frac{m_s m_c\langle\bar{s}g_s\sigma gs\rangle^2}{2304\pi^4}\int dy \left(16+\frac{7s}{t^2}\right ) \delta\left(s-\widetilde{m}_c^2 \right ) \nonumber\\ & & -\frac{m_s m_c\langle\bar{s}g_s\sigma gs\rangle^2}{3072\pi^4}\int dy \left(\frac{1-y}{y}+\frac{y}{1-y}\right)\delta \left(s-\widetilde{m}_c^2 \right)\nonumber\\ & & -\frac{m_s m_c\langle\bar{s}g_s\sigma gs\rangle^2}{6144\pi^4}\int dy \left ( 1+\frac{s}{t^2}\right ) \delta\left(s-\widetilde{m}_c^2 \right ) \ , , \end{aligned}\ ] ] @xmath220 @xmath221 \nonumber\\ & & -\frac{13\langle\bar{s}g_s\sigma gs\rangle^2}{27648\pi^4}\int dydz \,s\,\delta\left(s-\overline{m}_c^2 \right ) \nonumber\\ & & + \frac{m_s m_c}{3072\pi^8}\int dydz \ , ( 1-y - z)^3\left(s-\overline{m}_c^2\right)^4 \nonumber\\ & & -\frac{m_s m_c}{12288\pi^8}\int dydz \ , ( 1-y - z)^4\left(s-\overline{m}_c^2\right)^4 \nonumber\\ & & - \frac{m_s\langle \bar{s}s\rangle}{1536\pi^6}\int dydz \ , ( y+z)(1-y - z)^2\left(s-\overline{m}_c^2\right)^2 \left(5s-2\overline{m}_c^2 \right ) \nonumber \\ & & - \frac{m_s\langle \bar{s}s\rangle}{512\pi^6}\int dydz \ , ( y+z)(1-y - z)^3\left(s-\overline{m}_c^2\right)^2 \left(2s-\overline{m}_c^2 \right ) \nonumber\\ & & + \frac{11m_s\langle \bar{s}g_s\sigma gs\rangle}{2048\pi^6}\int dydz \ , ( y+z)(1-y - z ) \left(s-\overline{m}_c^2 \right)\left(2s-\overline{m}_c^2
\right ) \nonumber\\ & & + \frac{m_s\langle \bar{s}g_s\sigma gs\rangle}{1024\pi^6}\int dydz \ , ( y+z)(1-y - z)^2 \left(s-\overline{m}_c^2 \right)\left(5s-3\overline{m}_c^2 \right ) \nonumber\\ & & - \frac{m_s\langle \bar{s}g_s\sigma gs\rangle}{2048\pi^6}\int dydz \ , ( 1-y - z)^2 \,s\left(s-\overline{m}_c^2 \right ) \nonumber\\ & & + \frac { m_s m_c\langle\bar{s}s\rangle^2}{6\pi^4}\int dydz \ , \left(s-\overline{m}_c^2 \right ) \nonumber\\ & & + \frac { m_s m_c\langle\bar{s}s\rangle^2}{12\pi^4}\int dydz \ , ( 1-y - z)\left(s-\overline{m}_c^2 \right ) \nonumber\\ & & -\frac{89m_s m_c\langle\bar{s}s\rangle\langle\bar{s}g_s\sigma gs\rangle}{2304\pi^4}\int dydz \nonumber\\ & & -\frac{199m_s m_c \langle\bar{s}s\rangle\langle\bar{s}g_s\sigma gs\rangle}{2304\pi^4}\int dy \nonumber\\ & & + \frac{m_s\langle\bar{s}s\rangle^3}{36\pi^2}\int dy \ , \left [ 1+\frac{s}{2}\delta\left(s-\widetilde{m}_c^2 \right)\right ] \nonumber\\ & & + \frac{m_s m_c\langle\bar{s}g_s\sigma gs\rangle^2}{64\pi^4}\int dy \left(1+\frac{7s}{9t^2}\right ) \delta\left(s-\widetilde{m}_c^2 \right ) \nonumber\\ & & -\frac{m_s m_c\langle\bar{s}g_s\sigma gs\rangle^2}{1536\pi^4 t^2}\int dy \,s\ , \delta\left(s-\widetilde{m}_c^2
\right ) \ , , \end{aligned}\ ] ] @xmath222}{2304\pi^6}\int dydz \ , ( y+z)(1-y - z)^2\left(s-\overline{m}_c^2\right)^3 \nonumber \\ & & + \frac{m_c\left[\langle \bar{q}q\rangle+2\langle \bar{s}s\rangle\right]}{6912\pi^6}\int dydz \ , ( y+z)(1-y - z)^3\left(s-\overline{m}_c^2\right)^3 \nonumber \\ & & + \frac{11m_c\left[\langle \bar{q}g_s\sigma gq\rangle+2\langle \bar{s}g_s\sigma gs\rangle\right]}{12288\pi^6}\int dydz \ , ( y+z)(1-y - z ) \left(s-\overline{m}_c^2
\right)^2 \nonumber\\ & & - \frac{11m_c\left[\langle \bar{q}g_s\sigma gq\rangle+2\langle \bar{s}g_s\sigma gs\rangle\right]}{24576\pi^6}\int dydz \ , ( y+z)(1-y - z)^2 \left(s-\overline{m}_c^2 \right)^2 \nonumber\end{aligned}\ ] ] @xmath223}{72\pi^4}\int dydz \
, yz(1-y - z)\left(s-\overline{m}_c^2 \right)\left(5s-3\overline{m}_c^2 \right ) \nonumber\\ & & -\frac{19\left[\langle\bar{s}s\rangle\langle\bar{s}g_s\sigma gs\rangle+\langle\bar{q}q\rangle\langle\bar{s}g_s\sigma gs\rangle+\langle\bar{s}s\rangle\langle\bar{q}g_s\sigma gq\rangle\right]}{1152\pi^4}\int dydz \,yz \left(4s-3\overline{m}_c^2\right)\nonumber\\ & & -\frac{m_c\langle\bar{q}q\rangle\langle\bar{s}s\rangle^2}{36\pi^2}\int dy \nonumber\\ & & + \frac{11\langle\bar{s}g_s\sigma gs\rangle \left[2\langle\bar{q}g_s\sigma gq\rangle+\langle\bar{s}g_s\sigma gs\rangle \right]}{1536\pi^4}\int dy \ , y(1-y)\left [ 1+\frac{s}{3}\delta\left(s-\widetilde{m}_c^2 \right)\right ] \nonumber\\ & & + \frac{m_s m_c}{18432\pi^8}\int dydz \ , ( y+z)(1-y - z)^3\left(s-\overline{m}_c^2\right)^4 \nonumber\\ & & -\frac{m_s m_c}{73728\pi^8}\int dydz \ , ( y+z)(1-y - z)^4\left(s-\overline{m}_c^2\right)^4 \nonumber\\ & & - \frac{m_s\left[2\langle \bar{q}q\rangle-\langle \bar{s}s\rangle\right]}{192\pi^6}\int dydz \ , yz(1-y - z)^2\left(s-\overline{m}_c^2\right)^2 \left(2s-\overline{m}_c^2 \right ) \nonumber \\ & & - \frac{m_s\langle \bar{s}s\rangle}{576\pi^6}\int dydz \ , yz(1-y - z)^3\left(s-\overline{m}_c^2\right)^2 \left(7s-4\overline{m}_c^2 \right ) \nonumber \\ & & + \frac{m_s\langle \bar{s}g_s\sigma gs\rangle}{192\pi^6}\int dydz \
, yz(1-y - z)^2 \left(s-\overline{m}_c^2 \right)\left(3s-2\overline{m}_c^2 \right ) \nonumber\\ & & + \frac{m_s\left[19\langle \bar{q}g_s\sigma gq\rangle+3\langle \bar{s}g_s\sigma gs\rangle\right]}{3072\pi^6}\int dydz \ , yz(1-y - z ) \left(s-\overline{m}_c^2 \right)\left(5s-3\overline{m}_c^2
\right ) \nonumber\\ & & + \frac { m_s m_c\langle\bar{s}s\rangle\left[5\langle\bar{q}q\rangle -\langle\bar{s}s\rangle\right]}{144\pi^4}\int dydz \ , ( y+z)\left(s-\overline{m}_c^2 \right ) \nonumber\\ & & + \frac { m_s m_c\langle\bar{s}s\rangle\left[\langle\bar{q}q\rangle+\langle\bar{s}s\rangle \right]}{144\pi^4}\int dydz \ , ( y+z)(1-y - z)\left(s-\overline{m}_c^2 \right ) \nonumber\\ & & -\frac{m_s m_c\left[89\langle\bar{s}s\rangle\langle\bar{s}g_s\sigma gs\rangle+57\langle\bar{s}s\rangle\langle\bar{q}g_s\sigma gq\rangle+32\langle\bar{q}q\rangle\langle\bar{s}g_s\sigma gs\rangle\right]}{27648\pi^4}\int dydz \ ,
\left(y+z\right ) \nonumber\\ & & -\frac{m_s m_c \left[231\langle\bar{s}s\rangle\langle\bar{q}g_s\sigma gq\rangle-89\langle\bar{s}s\rangle\langle\bar{s}g_s\sigma gs\rangle+256\langle\bar{q}q\rangle\langle\bar{s}g_s\sigma gs\rangle\right]}{27648\pi^4}\int dy \nonumber\\ & & + \frac{m_s m_c\left[\langle\bar{q}q\rangle\langle\bar{s}g_s\sigma gs\rangle+\langle\bar{s}s\rangle\langle\bar{q}g_s\sigma gq\rangle\right]}{2304\pi^4}\int dydz
\left(\frac{y}{z}+\frac{z}{y}\right)\nonumber\\ & & + \frac{m_s\langle\bar{q}q\rangle\langle\bar{s}s\rangle^2}{9\pi^2}\int dy \,y(1-y ) \left [ 1+\frac{s}{3}\delta\left(s-\widetilde{m}_c^2 \right)\right ] \nonumber\\ & & -\frac{m_s m_c\langle\bar{q}g_s\sigma gq\rangle \langle\bar{s}g_s\sigma gs\rangle}{4608\pi^4}\int dy \left(\frac{1-y}{y}+\frac{y}{1-y}\right)\delta \left(s-\widetilde{m}_c^2 \right)\nonumber\\ & & + \frac{17m_s m_c\langle\bar{q}g_s\sigma gq\rangle\langle\bar{s}g_s\sigma gs\rangle } { 3456\pi^4}\int dy \left(1+\frac{8s}{17t^2}\right ) \delta\left(s-\widetilde{m}_c^2 \right ) \nonumber\\ & & -\frac{m_s m_c\langle\bar{s}g_s\sigma gs\rangle\left[3\langle \bar{q}g_s\sigma gq\rangle+19\langle \bar{s}g_s\sigma gs\rangle\right]}{18432\pi^4}\int dy \left ( 1+\frac{s}{t^2}\right ) \delta\left(s-\widetilde{m}_c^2 \right ) \ , , \end{aligned}\ ] ] @xmath224}{576\pi^6}\int dydz \ , ( 1-y - z)^2\left(s-\overline{m}_c^2\right)^3 \nonumber\end{aligned}\ ] ] @xmath225}{1728\pi^6}\int dydz \ , ( 1-y - z)^3\left(s-\overline{m}_c^2\right)^3 \nonumber \\ & & + \frac{11m_c\left[\langle \bar{q}g_s\sigma gq\rangle+2\langle \bar{s}g_s\sigma gs\rangle\right]}{3072\pi^6}\int dydz \ , ( 1-y - z ) \left(s-\overline{m}_c^2 \right)^2 \nonumber\\ & & - \frac{11m_c\left[\langle \bar{q}g_s\sigma gq\rangle+2\langle \bar{s}g_s\sigma gs\rangle\right]}{6144\pi^6}\int dydz \ , ( 1-y - z)^2 \left(s-\overline{m}_c^2 \right)^2 \nonumber\\ & & + \frac{\langle\bar{s}s\rangle\left [ 2\langle\bar{q}q\rangle+\langle\bar{s}s\rangle\right]}{144\pi^4}\int dydz \ , ( y+z)(1-y - z)\left(s-\overline{m}_c^2 \right)\left(2s-\overline{m}_c^2 \right ) \nonumber\\ & & -\frac{19\left[\langle\bar{s}s\rangle\langle\bar{s}g_s\sigma gs\rangle+\langle\bar{q}q\rangle\langle\bar{s}g_s\sigma gs\rangle+\langle\bar{s}s\rangle\langle\bar{q}g_s\sigma gq\rangle\right]}{4608\pi^4}\int dydz \,(y+z ) \left(3s-2\overline{m}_c^2\right)\nonumber\\ & & + \frac{\langle\bar{s}s\rangle\langle\bar{s}g_s\sigma gs\rangle+\langle\bar{q}q\rangle\langle\bar{s}g_s\sigma gs\rangle+\langle\bar{s}s\rangle\langle\bar{q}g_s\sigma gq\rangle}{2304\pi^4}\int dydz \,(1-y - z)\ , s\nonumber\\ & & -\frac{m_c\langle\bar{q}q\rangle\langle\bar{s}s\rangle^2}{9\pi^2}\int dy \nonumber\\ & & + \frac{11\langle\bar{s}g_s\sigma gs\rangle \left[2\langle\bar{q}g_s\sigma gq\rangle+\langle\bar{s}g_s\sigma gs\rangle \right]}{9216\pi^4}\int dy \ , \left [ 1+\frac{s}{2}\delta\left(s-\widetilde{m}_c^2 \right)\right ] \nonumber\\ & & -\frac{13\langle\bar{s}g_s\sigma gs\rangle\left[2\langle\bar{q}g_s\sigma gq\rangle+\langle\bar{s}g_s\sigma gs\rangle \right]}{82944\pi^4}\int dydz
\,s\,\delta\left(s-\overline{m}_c^2 \right ) \nonumber\\ & & + \frac{m_s m_c}{4608\pi^8}\int dydz \ , ( 1-y - z)^3\left(s-\overline{m}_c^2\right)^4 \nonumber\\ & & -\frac{m_s m_c}{18432\pi^8}\int dydz \ , ( 1-y - z)^4\left(s-\overline{m}_c^2\right)^4 \nonumber\\ & & - \frac{m_s\left[2\langle \bar{q}q\rangle-\langle \bar{s}s\rangle\right]}{2304\pi^6}\int dydz \ , ( y+z)(1-y - z)^2\left(s-\overline{m}_c^2\right)^2 \left(5s-2\overline{m}_c^2 \right ) \nonumber \\ & & - \frac{m_s\langle \bar{s}s\rangle}{768\pi^6}\int dydz \ , ( y+z)(1-y - z)^3\left(s-\overline{m}_c^2\right)^2 \left(2s-\overline{m}_c^2 \right ) \nonumber \\ & & + \frac{m_s\langle \bar{s}g_s\sigma gs\rangle}{1536\pi^6}\int dydz \ , ( y+z)(1-y - z)^2 \left(s-\overline{m}_c^2 \right)\left(5s-3\overline{m}_c^2 \right ) \nonumber\\ & & - \frac{m_s\left[\langle \bar{q}g_s\sigma gq\rangle+\langle \bar{s}g_s\sigma gs\rangle\right]}{6144\pi^6}\int dydz \ , ( 1-y - z)^2 \,s\left(s-\overline{m}_c^2 \right ) \nonumber\\ & & + \frac{m_s\left[19\langle \bar{q}g_s\sigma gq\rangle+3\langle \bar{s}g_s\sigma gs\rangle\right]}{6144\pi^6}\int dydz \ , ( y+z)(1-y - z ) \left(s-\overline{m}_c^2
\right)\left(2s-\overline{m}_c^2 \right ) \nonumber\\ & & + \frac { m_s m_c\langle\bar{s}s\rangle\left[5 \langle\bar{q}q\rangle-\langle\bar{s}s\rangle\right]}{36\pi^4}\int dydz \ , \left(s-\overline{m}_c^2 \right ) \nonumber\\ & & + \frac { m_s m_c\langle\bar{s}s\rangle\left [ \langle\bar{q}q\rangle+\langle\bar{s}s\rangle\right]}{36\pi^4}\int dydz \ ,
( 1-y - z)\left(s-\overline{m}_c^2 \right ) \nonumber\\ & & -\frac{m_s m_c\left[89\langle\bar{s}s\rangle\langle\bar{s}g_s\sigma gs\rangle+57\langle\bar{s}s\rangle\langle\bar{q}g_s\sigma gq\rangle+32\langle\bar{q}q\rangle\langle\bar{s}g_s\sigma gs\rangle\right]}{6912\pi^4}\int dydz \nonumber\\ & & -\frac{m_s m_c \left[231\langle\bar{s}s\rangle\langle\bar{q}g_s\sigma gq\rangle-89\langle\bar{s}s\rangle\langle\bar{s}g_s\sigma gs\rangle+256\langle\bar{q}q\rangle\langle\bar{s}g_s\sigma gs\rangle\right]}{6912\pi^4}\int dy \nonumber\\ & & + \frac{m_s\langle\bar{q}q\rangle\langle\bar{s}s\rangle^2}{54\pi^2}\int dy \ , \left [ 1+\frac{s}{2}\delta\left(s-\widetilde{m}_c^2 \right)\right ] \nonumber\\ & & + \frac{9m_s m_c\langle\bar{q}g_s\sigma gq\rangle \langle\bar{s}g_s\sigma gs\rangle}{864\pi^4}\int dy \left(1+\frac{8s}{9t^2}\right ) \delta\left(s-\widetilde{m}_c^2 \right ) \nonumber\\ & & -\frac{m_s m_c\langle\bar{s}g_s\sigma gs\rangle \left[3\langle \bar{q}g_s\sigma gq\rangle+19\langle \bar{s}g_s\sigma gs\rangle\right]}{13824\pi^4 t^2}\int dy \,s\ , \delta\left(s-\widetilde{m}_c^2
\right ) \ , , \end{aligned}\ ] ] @xmath226}{2304\pi^6}\int dydz \ , ( y+z)(1-y - z)^2\left(s-\overline{m}_c^2\right)^3 \nonumber \\ & & + \frac{m_c\left[2\langle \bar{q}q\rangle+\langle \bar{s}s\rangle\right]}{6912\pi^6}\int dydz \ , ( y+z)(1-y - z)^3\left(s-\overline{m}_c^2\right)^3 \nonumber \\ & & + \frac{11m_c\left[2\langle \bar{q}g_s\sigma gq\rangle+\langle \bar{s}g_s\sigma gs\rangle\right]}{12288\pi^6}\int dydz \ , ( y+z)(1-y - z ) \left(s-\overline{m}_c^2
\right)^2 \nonumber\\ & & - \frac{11m_c\left[2\langle \bar{q}g_s\sigma gq\rangle+\langle \bar{s}g_s\sigma gs\rangle\right]}{24576\pi^6}\int dydz \ , ( y+z)(1-y - z)^2 \left(s-\overline{m}_c^2 \right)^2 \nonumber\\ & & + \frac{\langle\bar{q}q\rangle\left[\langle\bar{q}q\rangle+2\langle\bar{s}s\rangle \right]}{72\pi^4}\int dydz \ , yz(1-y - z)\left(s-\overline{m}_c^2 \right)\left(5s-3\overline{m}_c^2 \right ) \nonumber\\ & & -\frac{19\left[\langle\bar{q}q\rangle\langle\bar{q}g_s\sigma gq\rangle+\langle\bar{q}q\rangle\langle\bar{s}g_s\sigma gs\rangle+\langle\bar{s}s\rangle\langle\bar{q}g_s\sigma gq\rangle\right]}{1152\pi^4}\int dydz \,yz \left(4s-3\overline{m}_c^2\right)\nonumber\\ & & -\frac{m_c\langle\bar{q}q\rangle^2\langle\bar{s}s\rangle}{36\pi^2}\int dy \nonumber\\ & & + \frac{11\langle\bar{q}g_s\sigma gq\rangle \left [ \langle\bar{q}g_s\sigma gq\rangle+2\langle\bar{s}g_s\sigma gs\rangle\right]}{1536\pi^4}\int dy \ , y(1-y)\left [ 1+\frac{s}{3}\delta\left(s-\widetilde{m}_c^2 \right)\right ] \nonumber\\ & & + \frac{m_s m_c}{36864\pi^8}\int dydz \ , ( y+z)(1-y - z)^3\left(s-\overline{m}_c^2\right)^4 \nonumber\\ & & -\frac{m_s m_c}{147456\pi^8}\int dydz \ , ( y+z)(1-y - z)^4\left(s-\overline{m}_c^2\right)^4 \nonumber\\ & & - \frac{m_s\left[4\langle \bar{q}q\rangle-3\langle \bar{s}s\rangle\right]}{384\pi^6}\int dydz \ , yz(1-y - z)^2\left(s-\overline{m}_c^2\right)^2 \left(2s-\overline{m}_c^2 \right ) \nonumber \\ & & - \frac{m_s\langle \bar{s}s\rangle}{1152\pi^6}\int dydz \ , yz(1-y - z)^3\left(s-\overline{m}_c^2\right)^2 \left(7s-4\overline{m}_c^2 \right ) \nonumber \\ & & + \frac{m_s\langle \bar{s}g_s\sigma gs\rangle}{384\pi^6}\int dydz \
, yz(1-y - z)^2 \left(s-\overline{m}_c^2 \right)\left(3s-2\overline{m}_c^2 \right ) \nonumber\\ & & + \frac{m_s\left[19\langle \bar{q}g_s\sigma gq\rangle-8\langle \bar{s}g_s\sigma gs\rangle\right]}{3072\pi^6}\int dydz \ , yz(1-y - z ) \left(s-\overline{m}_c^2 \right)\left(5s-3\overline{m}_c^2
\right ) \nonumber\\ & & + \frac { m_s m_c\langle\bar{q}q\rangle \left [ 3\langle\bar{q}q\rangle-\langle\bar{s}s\rangle\right]}{144\pi^4}\int dydz \ ,
( y+z)\left(s-\overline{m}_c^2 \right ) \nonumber\\ & & + \frac { m_s m_c\langle\bar{q}q\rangle\langle\bar{s}s\rangle } { 144\pi^4}\int dydz \ , ( y+z)(1-y - z)\left(s-\overline{m}_c^2 \right ) \nonumber\\ & & -\frac{m_s m_c\left[32\langle\bar{q}q\rangle\langle\bar{s}g_s\sigma gs\rangle+57\langle\bar{s}s\rangle\langle\bar{q}g_s\sigma gq\rangle\right]}{27648\pi^4}\int dydz \ ,
\left(y+z\right ) \nonumber\\ & & -\frac{m_s m_c \left[288\langle\bar{q}q\rangle\langle\bar{q}g_s\sigma gq\rangle-57\langle\bar{s}s\rangle\langle\bar{q}g_s\sigma gq\rangle-32\langle\bar{q}q\rangle\langle\bar{s}g_s\sigma gs\rangle\right]}{27648\pi^4}\int dy \nonumber\\ & & + \frac{m_s m_c\langle\bar{q}q\rangle\langle\bar{q}g_s\sigma gq\rangle}{2304\pi^4}\int dydz \left(\frac{y}{z}+\frac{z}{y}\right)\nonumber\end{aligned}\ ] ] @xmath227 \nonumber\\ & & -\frac{m_s m_c\langle\bar{q}g_s\sigma gq\rangle^2}{9216\pi^4}\int dy
\left(\frac{1-y}{y}+\frac{y}{1-y}\right)\delta \left(s-\widetilde{m}_c^2 \right)\nonumber\\ & & + \frac{m_s m_c\langle\bar{q}g_s\sigma gq\rangle \left [ 144\langle\bar{q}g_s\sigma gq\rangle-19\langle\bar{s}g_s\sigma gs\rangle\right]}{55296\pi^4}\int dy \left ( 1+\frac{s}{t^2}\right ) \delta\left(s-\widetilde{m}_c^2 \right ) \ , , \end{aligned}\ ] ] @xmath228}{576\pi^6}\int dydz \ , ( 1-y - z)^2\left(s-\overline{m}_c^2\right)^3 \nonumber \\ & & + \frac{m_c\left[2\langle \bar{q}q\rangle+\langle \bar{s}s\rangle\right]}{1728\pi^6}\int dydz \ , ( 1-y - z)^3\left(s-\overline{m}_c^2\right)^3 \nonumber \\ & & + \frac{11m_c\left[2\langle \bar{q}g_s\sigma gq\rangle+\langle \bar{s}g_s\sigma gs\rangle\right]}{3072\pi^6}\int dydz \ , ( 1-y - z ) \left(s-\overline{m}_c^2
\right)^2 \nonumber\\ & & - \frac{11m_c\left[2\langle \bar{q}g_s\sigma gq\rangle+\langle \bar{s}g_s\sigma gs\rangle\right]}{6144\pi^6}\int dydz \ , ( 1-y - z)^2 \left(s-\overline{m}_c^2 \right)^2 \nonumber\\ & & + \frac{\langle\bar{q}q\rangle \left [ 2\langle\bar{q}q\rangle+\langle\bar{s}s\rangle\right]}{144\pi^4}\int dydz \
, ( y+z)(1-y - z)\left(s-\overline{m}_c^2 \right)\left(2s-\overline{m}_c^2 \right ) \nonumber\\ & & -\frac{19\left[\langle\bar{q}q\rangle\langle\bar{q}g_s\sigma gq\rangle+\langle\bar{q}q\rangle\langle\bar{s}g_s\sigma gs\rangle+\langle\bar{s}s\rangle\langle\bar{q}g_s\sigma gq\rangle\right]}{4608\pi^4}\int dydz \,(y+z ) \left(3s-2\overline{m}_c^2\right)\nonumber\\ & & + \frac{\langle\bar{q}q\rangle\langle\bar{q}g_s\sigma gq\rangle+\langle\bar{q}q\rangle\langle\bar{s}g_s\sigma gs\rangle+\langle\bar{s}s\rangle\langle\bar{q}g_s\sigma gq\rangle}{2304\pi^4}\int dydz
\,(1-y - z)\ , s\nonumber\\ & & -\frac{m_c\langle\bar{q}q\rangle^2\langle\bar{s}s\rangle}{9\pi^2}\int dy \nonumber\\ & & + \frac{11\langle\bar{q}g_s\sigma gq\rangle \left[\langle\bar{q}g_s\sigma gq\rangle+2\langle\bar{s}g_s\sigma gs\rangle \right]}{9216\pi^4}\int dy \ , \left [ 1+\frac{s}{2}\delta\left(s-\widetilde{m}_c^2 \right)\right ] \nonumber\\ & & -\frac{13\langle\bar{q}g_s\sigma gq\rangle \left[\langle\bar{q}g_s\sigma
gq\rangle+2\langle\bar{s}g_s\sigma gs\rangle \right]}{82944\pi^4}\int dydz \,s\,\delta\left(s-\overline{m}_c^2 \right ) \nonumber\\ & & + \frac{m_s m_c}{9216\pi^8}\int dydz \ , ( 1-y - z)^3\left(s-\overline{m}_c^2\right)^4 \nonumber\\ & & -\frac{m_s m_c}{36864\pi^8}\int dydz \ , ( 1-y - z)^4\left(s-\overline{m}_c^2\right)^4 \nonumber\\ & & - \frac{m_s\left[4\langle \bar{q}q\rangle-3\langle \bar{s}s\rangle\right]}{4608\pi^6}\int dydz \ , ( y+z)(1-y - z)^2\left(s-\overline{m}_c^2\right)^2 \left(5s-2\overline{m}_c^2 \right )
\nonumber \\ & & - \frac{m_s\langle \bar{s}s\rangle}{1536\pi^6}\int dydz \ , ( y+z)(1-y - z)^3\left(s-\overline{m}_c^2\right)^2 \left(2s-\overline{m}_c^2 \right ) \nonumber \\ & & + \frac{m_s\langle \bar{s}g_s\sigma gs\rangle}{3072\pi^6}\int dydz \ , ( y+z)(1-y - z)^2 \left(s-\overline{m}_c^2 \right)\left(5s-3\overline{m}_c^2 \right ) \nonumber\\ & & - \frac{m_s\langle \bar{q}g_s\sigma gq\rangle}{6144\pi^6}\int dydz \ , ( 1-y - z)^2 \,s\left(s-\overline{m}_c^2 \right ) \nonumber\\ & & + \frac{m_s\left[19\langle \bar{q}g_s\sigma gq\rangle-8\langle \bar{s}g_s\sigma gs\rangle\right]}{6144\pi^6}\int dydz \ , ( y+z)(1-y - z ) \left(s-\overline{m}_c^2 \right)\left(2s-\overline{m}_c^2
\right ) \nonumber\end{aligned}\ ] ] @xmath229}{36\pi^4}\int dydz \ , \left(s-\overline{m}_c^2 \right ) \nonumber\\ & & + \frac { m_s m_c\langle\bar{q}q\rangle\langle\bar{s}s\rangle } { 36\pi^4}\int dydz \ , ( 1-y - z)\left(s-\overline{m}_c^2 \right ) \nonumber\\ & & -\frac{m_s m_c\left[32\langle\bar{q}q\rangle\langle\bar{s}g_s\sigma gs\rangle+57\langle\bar{s}s\rangle\langle\bar{q}g_s\sigma gq\rangle\right]}{6912\pi^4}\int dydz \nonumber\\ & & -\frac{m_s m_c \left[288\langle\bar{q}q\rangle\langle\bar{q}g_s\sigma gq\rangle-57\langle\bar{s}s\rangle\langle\bar{q}g_s\sigma gq\rangle-32\langle\bar{q}q\rangle\langle\bar{s}g_s\sigma gs\rangle\right]}{6912\pi^4}\int dy \nonumber\\ & & + \frac{m_s\langle\bar{q}q\rangle^2\langle\bar{s}s\rangle}{108\pi^2}\int dy \ , \left [ 1+\frac{s}{2}\delta\left(s-\widetilde{m}_c^2 \right)\right ] \nonumber\\ & & + \frac{m_s m_c\langle\bar{q}g_s\sigma gq\rangle^2}{192\pi^4}\int dy \left(1+\frac{s}{t^2}\right ) \delta\left(s-\widetilde{m}_c^2 \right ) \nonumber\\ & & -\frac{19m_s m_c\langle\bar{q}g_s\sigma gq\rangle\langle\bar{s}g_s\sigma gs\rangle } { 13824\pi^4 t^2}\int dy \,s\ , \delta\left(s-\widetilde{m}_c^2 \right ) \ , , \end{aligned}\ ] ] @xmath230 @xmath231\nonumber\end{aligned}\ ] ] @xmath232\nonumber\\ & & + \frac{m_s m_c}{24576\pi^8}\int dydz \ , ( y+z)(1-y - z)^3\left(s-\overline{m}_c^2\right)^4 \nonumber\\ & & -\frac{m_s m_c}{98304\pi^8}\int dydz \ , ( y+z)(1-y - z)^4\left(s-\overline{m}_c^2\right)^4 \nonumber\\ & & - \frac{m_s\langle \bar{s}s\rangle}{512\pi^6}\int dydz \ , yz(1-y - z)^2\left(s-\overline{m}_c^2\right)^2 \left(2s-\overline{m}_c^2 \right ) \nonumber \\ & & - \frac{m_s\langle \bar{s}s\rangle}{1536\pi^6}\int dydz \ , yz(1-y - z)^3\left(s-\overline{m}_c^2\right)^2 \left(7s-4\overline{m}_c^2 \right ) \nonumber \\ & & + \frac{11m_s\langle \bar{s}g_s\sigma gs\rangle}{4096\pi^6}\int dydz \ , yz(1-y - z ) \left(s-\overline{m}_c^2 \right)\left(5s-3\overline{m}_c^2
\right ) \nonumber\\ & & + \frac{m_s\langle \bar{s}g_s\sigma gs\rangle}{512\pi^6}\int dydz \
, yz(1-y - z)^2 \left(s-\overline{m}_c^2 \right)\left(3s-2\overline{m}_c^2 \right ) \nonumber\\ & & + \frac{m_s\langle \bar{s}g_s\sigma gs\rangle}{16384\pi^6}\int dydz \ , ( y+z)(1-y - z)^2 \left(s-\overline{m}_c^2 \right)\left(3s-\overline{m}_c^2 \right ) \nonumber\\ & & + \frac { m_s m_c\langle\bar{s}s\rangle^2}{48\pi^4}\int dydz \ , ( y+z)\left(s-\overline{m}_c^2 \right ) \nonumber\\ & & + \frac { m_s m_c\langle\bar{s}s\rangle^2}{96\pi^4}\int dydz \ , ( y+z)(1-y - z)\left(s-\overline{m}_c^2 \right ) \nonumber\\ & & -\frac{89m_s m_c\langle\bar{s}s\rangle\langle\bar{s}g_s\sigma gs\rangle}{18432\pi^4}\int dydz \ , \left(y+z\right ) \nonumber\\ & & -\frac{199m_s m_c \langle\bar{s}s\rangle\langle\bar{s}g_s\sigma gs\rangle}{18432\pi^4}\int dy \nonumber\\ & & + \frac{3m_s m_c \langle \bar{s}s\rangle\langle \bar{s}g_s\sigma gs\rangle}{1024\pi^4}\int dydz \ , \left(\frac{y}{z}+\frac{z}{y}\right ) \nonumber\\ & & + \frac{m_s m_c\langle\bar{s}s\rangle\langle\bar{s}g_s\sigma gs\rangle}{512\pi^4}\int dydz \left(\frac{y}{z}+\frac{z}{y}\right)(1-y - z)\nonumber\\ & & + \frac{m_s\langle\bar{s}s\rangle^3}{24\pi^2}\int dy \,y(1-y ) \left [ 1+\frac{s}{3}\delta\left(s-\widetilde{m}_c^2 \right)\right ] \nonumber\\ & & + \frac{m_s m_c\langle\bar{s}g_s\sigma gs\rangle^2}{4608\pi^4}\int dy \left(16+\frac{7s}{t^2}\right ) \delta\left(s-\widetilde{m}_c^2 \right ) \nonumber\\ & & -\frac{m_s m_c\langle\bar{s}g_s\sigma gs\rangle^2}{3072\pi^4}\int dydz \left(\frac{z}{y}+\frac{y}{z}\right)\delta \left(s-\overline{m}_c^2 \right)\nonumber\\ & & -\frac{17m_s m_c\langle\bar{s}g_s\sigma gs\rangle^2}{18432\pi^4}\int dy \left ( \frac{1-y}{y}+\frac{y}{1-y}\right ) \delta\left(s-\widetilde{m}_c^2 \right)\nonumber\\ & & - \frac{m_s m_c \langle \bar{s}g_s\sigma gs\rangle^2}{12288\pi^4}\int dy \,\left(1+\frac{s}{t^2}\right)\delta\left(s-\widetilde{m}_c^2 \right ) \ , , \end{aligned}\ ] ] @xmath233 @xmath234 \nonumber\\ & & + \frac{13\langle\bar{s}g_s\sigma gs\rangle^2}{55296\pi^4}\int dydz \ , s\,\delta \left(s-\overline{m}_c^2 \right ) \nonumber\\ & & + \frac{m_s m_c}{12288\pi^8}\int dydz \ , ( 1-y - z)^3\left(s-\overline{m}_c^2\right)^4 \nonumber\\ & & -\frac{m_s m_c}{49152\pi^8}\int dydz \ , ( 1-y - z)^4\left(s-\overline{m}_c^2\right)^4 \nonumber\\ & & - \frac{m_s\langle \bar{s}s\rangle}{3072\pi^6}\int dydz \ , ( y+z)(1-y - z)^2\left(s-\overline{m}_c^2\right)^2 \left(5s-2\overline{m}_c^2 \right )
\nonumber \\ & & - \frac{m_s\langle \bar{s}s\rangle}{1024\pi^6}\int dydz \ , ( y+z)(1-y - z)^3\left(s-\overline{m}_c^2\right)^2 \left(2s-\overline{m}_c^2 \right ) \nonumber\\ & & + \frac{11m_s\langle \bar{s}g_s\sigma gs\rangle}{4096\pi^6}\int dydz \ , ( y+z)(1-y - z ) \left(s-\overline{m}_c^2 \right)\left(2s-\overline{m}_c^2
\right ) \nonumber\\ & & + \frac{m_s\langle \bar{s}g_s\sigma gs\rangle}{2048\pi^6}\int dydz \ , ( y+z)(1-y - z)^2 \left(s-\overline{m}_c^2 \right)\left(5s-3\overline{m}_c^2 \right ) \nonumber\\ & & + \frac{m_s\langle \bar{s}g_s\sigma gs\rangle}{4096\pi^6}\int dydz \ , ( 1-y - z)^2 \,s\left(s-\overline{m}_c^2 \right ) \nonumber\\ & & + \frac { m_s m_c\langle\bar{s}s\rangle^2}{24\pi^4}\int dydz \ , \left(s-\overline{m}_c^2 \right ) \nonumber\\ & & + \frac { m_s m_c\langle\bar{s}s\rangle^2}{48\pi^4}\int dydz \ , ( 1-y - z)\left(s-\overline{m}_c^2 \right ) \nonumber\\ & & -\frac{89m_s m_c\langle\bar{s}s\rangle\langle\bar{s}g_s\sigma gs\rangle}{9216\pi^4}\int dydz \nonumber\\ & & -\frac{199m_s m_c \langle\bar{s}s\rangle\langle\bar{s}g_s\sigma gs\rangle}{9216\pi^4}\int dy \nonumber\\ & & + \frac{3m_s m_c\langle\bar{s}s\rangle\langle\bar{s}g_s\sigma gs\rangle}{1024\pi^4}\int dydz \left(\frac{1}{y}+\frac{1}{z}\right ) \nonumber\end{aligned}\ ] ] @xmath235 \nonumber\\ & & + \frac{m_s m_c\langle\bar{s}g_s\sigma gs\rangle^2}{256\pi^4}\int dy \left(1+\frac{7s}{9t^2}\right ) \delta\left(s-\widetilde{m}_c^2 \right ) \nonumber\\ & & -\frac{m_s m_c\langle\bar{s}g_s\sigma gs\rangle^2}{3072\pi^4}\int dydz \left(\frac{1}{y}+\frac{1}{z}\right)\delta \left(s-\overline{m}_c^2 \right)\nonumber\\ & & -\frac{17m_s m_c\langle\bar{s}g_s\sigma gs\rangle^2}{18432\pi^4}\int dy \left ( \frac{1}{y}+\frac{1}{1-y}\right ) \delta\left(s-\widetilde{m}_c^2 \right)\nonumber\\ & & - \frac{m_s m_c \langle \bar{s}g_s\sigma gs\rangle^2}{6144\pi^4t^2}\int dy \,s\,\delta\left(s-\widetilde{m}_c^2 \right ) \ , , \end{aligned}\ ] ] @xmath236}{4608\pi^6}\int dydz \ , ( y+z)(1-y - z)^2\left(s-\overline{m}_c^2\right)^3 \nonumber \\ & & + \frac{m_c\left[\langle \bar{q}q\rangle+2\langle \bar{s}s\rangle\right]}{13824\pi^6}\int dydz \ , ( y+z)(1-y - z)^3\left(s-\overline{m}_c^2\right)^3 \nonumber \\ & & + \frac{11m_c\left[\langle \bar{q}g_s\sigma gq\rangle+2\langle \bar{s}g_s\sigma gs\rangle\right]}{24576\pi^6}\int dydz \ , ( y+z)(1-y - z ) \left(s-\overline{m}_c^2 \right)^2 \nonumber\\ & & + \frac{11m_c\left[\langle \bar{q}g_s\sigma gq\rangle+2\langle \bar{s}g_s\sigma gs\rangle\right]}{49152\pi^6}\int dydz \ , ( y+z)(1-y - z)^2 \left(s-\overline{m}_c^2 \right)^2 \nonumber\\ & & - \frac{7m_c\left[\langle \bar{q}g_s\sigma gq\rangle+2\langle \bar{s}g_s\sigma gs\rangle\right]}{49152\pi^6}\int dydz \ , \left(\frac{y}{z}+\frac{z}{y } \right)(1-y - z)^2 \left(s-\overline{m}_c^2 \right)^2 \nonumber\\ & & + \frac{m_c\left[\langle \bar{q}g_s\sigma gq\rangle+2\langle \bar{s}g_s\sigma gs\rangle\right]}{24576\pi^6}\int dydz \ , \left(\frac{y}{z}+\frac{z}{y}\right)(1-y - z)^3 \left(s-\overline{m}_c^2 \right)^2 \nonumber\\ & & + \frac{\langle\bar{s}s\rangle \left[2\langle\bar{q}q\rangle+\langle\bar{s}s\rangle \right]}{288\pi^4}\int dydz \ , yz(1-y - z)\left(s-\overline{m}_c^2 \right)\left(5s-3\overline{m}_c^2 \right ) \nonumber\\ & & -\frac{19\left[\langle\bar{s}s\rangle\langle\bar{s}g_s\sigma gs\rangle+\langle\bar{q}q\rangle\langle\bar{s}g_s\sigma gs\rangle+\langle\bar{s}s\rangle\langle\bar{q}g_s\sigma gq\rangle\right]}{4608\pi^4}\int dydz \,yz \left(4s-3\overline{m}_c^2\right)\nonumber\\ & & -\frac{\langle\bar{s}s\rangle\langle\bar{s}g_s\sigma gs\rangle+\langle\bar{q}q\rangle\langle\bar{s}g_s\sigma gs\rangle+\langle\bar{s}s\rangle\langle\bar{q}g_s\sigma gq\rangle}{9216\pi^4}\int dydz \,(y+z)(1-y - z ) \left(2s-\overline{m}_c^2\right)\nonumber\\ & & -\frac{m_c\langle\bar{q}q\rangle\langle\bar{s}s\rangle^2}{72\pi^2}\int dy \nonumber\\ & & + \frac{11\langle\bar{s}g_s\sigma gs\rangle\left[2\langle\bar{q}g_s\sigma gq\rangle+\langle\bar{s}g_s\sigma gs\rangle \right]}{6144\pi^4}\int dy \,y(1-y ) \left[1+\frac{s}{3}\delta \left(s-\widetilde{m}_c^2 \right)\right]\nonumber\\ & & + \frac{13\langle\bar{s}g_s\sigma gs\rangle \left[2\langle\bar{q}g_s\sigma gq\rangle+\langle\bar{s}g_s\sigma gs\rangle \right]}{331776\pi^4}\int dydz \,(y+z ) \left[1+s\,\delta \left(s-\overline{m}_c^2 \right)\right]\nonumber\\ & & + \frac{m_s m_c}{36864\pi^8}\int dydz \ , ( y+z)(1-y - z)^3\left(s-\overline{m}_c^2\right)^4 \nonumber\\ & & -\frac{m_s m_c}{147456\pi^8}\int dydz \ , ( y+z)(1-y - z)^4\left(s-\overline{m}_c^2\right)^4 \nonumber\end{aligned}\ ] ] @xmath237}{768\pi^6}\int dydz \ , yz(1-y - z)^2\left(s-\overline{m}_c^2\right)^2 \left(2s-\overline{m}_c^2 \right ) \nonumber \\ & & - \frac{m_s\langle \bar{s}s\rangle}{2304\pi^6}\int dydz \ , yz(1-y - z)^3\left(s-\overline{m}_c^2\right)^2 \left(7s-4\overline{m}_c^2 \right ) \nonumber \\ & & + \frac{m_s\langle \bar{s}g_s\sigma gs\rangle}{768\pi^6}\int dydz \ , yz(1-y - z)^2 \left(s-\overline{m}_c^2 \right)\left(3s-2\overline{m}_c^2 \right ) \nonumber\\ & & + \frac{m_s\left[19\langle \bar{q}g_s\sigma gq\rangle+3\langle \bar{s}g_s\sigma gs\rangle\right]}{12288\pi^6}\int dydz \ , yz(1-y - z ) \left(s-\overline{m}_c^2 \right)\left(5s-3\overline{m}_c^2
\right ) \nonumber\\ & & + \frac{m_s\left[\langle \bar{q}g_s\sigma gq\rangle+\langle \bar{s}g_s\sigma gs\rangle\right]}{49152\pi^6}\int dydz \ , ( y+z)(1-y - z)^2 \left(s-\overline{m}_c^2 \right)\left(3s-\overline{m}_c^2 \right ) \nonumber\\ & & + \frac { m_s m_c\langle\bar{s}s\rangle \left[5\langle\bar{q}q\rangle-\langle\bar{s}s\rangle \right]}{288\pi^4}\int dydz \ , ( y+z)\left(s-\overline{m}_c^2 \right ) \nonumber\\ & & + \frac { m_s m_c\langle\bar{s}s\rangle \left[\langle\bar{q}q\rangle+\langle\bar{s}s\rangle \right]}{288\pi^4}\int dydz \ , ( y+z)(1-y - z)\left(s-\overline{m}_c^2 \right ) \nonumber\\ & & -\frac{m_s m_c\left[89\langle\bar{s}s\rangle\langle\bar{s}g_s\sigma gs\rangle+32\langle\bar{q}q\rangle\langle\bar{s}g_s\sigma gs\rangle+57\langle\bar{s}s\rangle\langle\bar{q}g_s\sigma gq\rangle\right]}{55296\pi^4}\int dydz \ ,
\left(y+z\right ) \nonumber\\ & & -\frac{m_s m_c \left[231\langle\bar{s}s\rangle\langle\bar{q}g_s\sigma gq\rangle+256\langle\bar{q}q\rangle\langle\bar{s}g_s\sigma gs\rangle-89\langle\bar{s}s\rangle\langle\bar{s}g_s\sigma gs\rangle\right]}{55296\pi^4}\int dy \nonumber\\ & & + \frac{m_s m_c\langle\bar{s}s\rangle\left[\langle\bar{q}g_s\sigma gq\rangle+\langle\bar{s}g_s\sigma gs\rangle\right]}{1536\pi^4}\int dydz
\left(\frac{y}{z}+\frac{z}{y}\right)(1-y - z)\nonumber\\ & & + \frac{m_s m_c\left[9\langle\bar{s}s\rangle\langle\bar{q}g_s\sigma gq\rangle-7\langle\bar{s}s\rangle\langle\bar{s}g_s\sigma gs\rangle+16\langle\bar{q}q\rangle\langle\bar{s}g_s\sigma gs\rangle\right]}{9216\pi^4}\int dydz \left(\frac{y}{z}+\frac{z}{y}\right ) \nonumber\\ & & + \frac{m_s\langle\bar{q}q\rangle\langle\bar{s}s\rangle^2}{36\pi^2}\int dy \,y(1-y ) \left [ 1+\frac{s}{3}\delta\left(s-\widetilde{m}_c^2 \right)\right ] \nonumber\\ & & -\frac{m_s m_c\langle\bar{s}g_s\sigma gs\rangle \left[\langle\bar{q}g_s\sigma gq\rangle+\langle\bar{s}g_s\sigma gs\rangle \right]}{9216\pi^4}\int dydz \left(\frac{z}{y}+\frac{y}{z}\right)\delta \left(s-\overline{m}_c^2 \right)\nonumber\\ & & -\frac{m_s m_c\langle\bar{s}g_s\sigma gs\rangle\left[41\langle \bar{q}g_s\sigma gq\rangle-7\langle \bar{s}g_s\sigma gs\rangle\right]}{55296\pi^4}\int dy \left ( \frac{1-y}{y}+\frac{y}{1-y}\right ) \delta\left(s-\widetilde{m}_c^2 \right)\nonumber\\ & & -\frac{m_s m_c\langle\bar{s}g_s\sigma gs\rangle \left[19\langle\bar{s}g_s\sigma gs\rangle-125\langle\bar{q}g_s\sigma gq\rangle \right]}{55296\pi^4}\int dy \left(1+\frac{s}{2t^2}\right ) \delta\left(s-\widetilde{m}_c^2 \right ) \nonumber\\ & & + \frac{19m_s m_c \langle \bar{s}g_s\sigma gs\rangle\left[\langle \bar{q}g_s\sigma gq\rangle+\langle \bar{s}g_s\sigma gs\rangle\right]}{110592\pi^4}\int dy \,\delta\left(s-\widetilde{m}_c^2 \right ) \ , , \end{aligned}\ ] ] @xmath238}{2304\pi^6}\int dydz \ , ( 1-y - z)^2\left(s-\overline{m}_c^2\right)^3 \nonumber \\ & & + \frac{m_c\left[\langle \bar{q}q\rangle+2\langle \bar{s}s\rangle\right]}{6912\pi^6}\int dydz \ , ( 1-y - z)^3\left(s-\overline{m}_c^2\right)^3 \nonumber \\ & & + \frac{11m_c\left[\langle \bar{q}g_s\sigma gq\rangle+2\langle \bar{s}g_s\sigma gs\rangle\right]}{12288\pi^6}\int dydz \ , ( 1-y - z ) \left(s-\overline{m}_c^2
\right)^2 \nonumber\\ & & - \frac{11m_c\left[\langle \bar{q}g_s\sigma gq\rangle+2\langle \bar{s}g_s\sigma gs\rangle\right]}{24576\pi^6}\int dydz \
, ( 1-y - z)^2 \left(s-\overline{m}_c^2 \right)^2 \nonumber\end{aligned}\ ] ] @xmath239}{49152\pi^6}\int dydz \ , \left(\frac{1}{y}+\frac{1}{z } \right)(1-y - z)^2 \left(s-\overline{m}_c^2 \right)^2 \nonumber\\ & & + \frac{m_c\left[\langle \bar{q}g_s\sigma gq\rangle+2\langle \bar{s}g_s\sigma gs\rangle\right]}{24576\pi^6}\int dydz \ , \left(\frac{1}{y}+\frac{1}{z}\right)(1-y - z)^3 \left(s-\overline{m}_c^2 \right)^2 \nonumber\\ & & + \frac{\langle\bar{s}s\rangle\left[2\langle\bar{q}q\rangle+\langle\bar{s}s\rangle\right]}{288\pi^4}\int dydz \ , ( y+z)(1-y - z)\left(s-\overline{m}_c^2 \right)\left(2s-\overline{m}_c^2 \right ) \nonumber\\ & & -\frac{19\left[\langle\bar{s}s\rangle\langle\bar{s}g_s\sigma gs\rangle+\langle\bar{q}q\rangle\langle\bar{s}g_s\sigma gs\rangle+\langle\bar{s}s\rangle\langle\bar{q}g_s\sigma gq\rangle\right]}{9216\pi^4}\int dydz \,(y+z ) \left(3s-2\overline{m}_c^2\right)\nonumber\\ & & -\frac{\langle\bar{s}s\rangle\langle\bar{s}g_s\sigma gs\rangle+\langle\bar{q}q\rangle\langle\bar{s}g_s\sigma gs\rangle+\langle\bar{s}s\rangle\langle\bar{q}g_s\sigma gq\rangle}{4608\pi^4}\int dydz \ , ( 1-y - z ) s\nonumber\\ & & -\frac{m_c\langle\bar{q}q\rangle\langle\bar{s}s\rangle^2}{36\pi^2}\int dy \nonumber\\ & & + \frac{13\langle\bar{s}g_s\sigma gs\rangle \left[2\langle\bar{q}g_s\sigma gq\rangle+\langle\bar{s}g_s\sigma gs\rangle \right]}{165888\pi^4}\int dydz \ , s\,\delta \left(s-\overline{m}_c^2 \right ) \nonumber\\ & & + \frac{11\langle\bar{s}g_s\sigma gs\rangle \left [ 2\langle\bar{q}g_s\sigma gq\rangle+\langle\bar{s}g_s\sigma gs\rangle\right]}{18432\pi^4}\int dy \
, \left[1+\frac{s}{2}\delta \left(s-\widetilde{m}_c^2 \right)\right]\nonumber\\ & & + \frac{m_s m_c}{18432\pi^8}\int dydz \ , ( 1-y - z)^3\left(s-\overline{m}_c^2\right)^4 \nonumber\\ & & -\frac{m_s m_c}{73728\pi^8}\int dydz \ , ( 1-y - z)^4\left(s-\overline{m}_c^2\right)^4 \nonumber\\ & & - \frac{m_s\left[2\langle \bar{q}q\rangle-\langle \bar{s}s\rangle\right]}{4608\pi^6}\int dydz \ , ( y+z)(1-y - z)^2\left(s-\overline{m}_c^2\right)^2 \left(5s-2\overline{m}_c^2 \right ) \nonumber \\ & & - \frac{m_s\langle \bar{s}s\rangle}{1536\pi^6}\int dydz \ , ( y+z)(1-y - z)^3\left(s-\overline{m}_c^2\right)^2 \left(2s-\overline{m}_c^2 \right )
\nonumber \\ & & + \frac{m_s\langle \bar{s}g_s\sigma gs\rangle}{3072\pi^6}\int dydz \ , ( y+z)(1-y - z)^2 \left(s-\overline{m}_c^2 \right)\left(5s-3\overline{m}_c^2 \right ) \nonumber\\ & & + \frac{m_s\left[\langle \bar{q}g_s\sigma gq\rangle+\langle \bar{s}g_s\sigma gs\rangle\right]}{12288\pi^6}\int dydz \ , ( 1-y - z)^2 \,s\left(s-\overline{m}_c^2 \right ) \nonumber\\ & & + \frac{m_s\left[19\langle \bar{q}g_s\sigma gq\rangle+3\langle \bar{s}g_s\sigma gs\rangle\right]}{12288\pi^6}\int dydz \ , ( y+z)(1-y - z ) \left(s-\overline{m}_c^2
\right)\left(2s-\overline{m}_c^2 \right ) \nonumber\\ & & + \frac { m_s m_c\langle\bar{s}s\rangle \left[5\langle\bar{q}q\rangle -\langle\bar{s}s\rangle \right]}{144\pi^4}\int dydz \ , \left(s-\overline{m}_c^2 \right ) \nonumber\\ & & + \frac { m_s m_c\langle\bar{s}s\rangle \left [ \langle\bar{q}q\rangle+\langle\bar{s}s\rangle\right]}{144\pi^4}\int dydz \ , ( 1-y - z)\left(s-\overline{m}_c^2 \right ) \nonumber\\ & & -\frac{m_s m_c\left[89\langle\bar{s}s\rangle\langle\bar{s}g_s\sigma gs\rangle+32\langle\bar{q}q\rangle\langle\bar{s}g_s\sigma gs\rangle+57\langle\bar{s}s\rangle\langle\bar{q}g_s\sigma gq\rangle\right]}{27648\pi^4}\int dydz \nonumber\\ & & -\frac{m_s m_c\left [ 231\langle\bar{s}s\rangle\langle\bar{q}g_s\sigma gq\rangle+256\langle\bar{q}q\rangle\langle\bar{s}g_s\sigma gs\rangle-89\langle\bar{s}s\rangle\langle\bar{s}g_s\sigma gs\rangle\right]}{27648\pi^4}\int dy
\nonumber\\ & & + \frac{m_s m_c\langle\bar{s}s\rangle \left[\langle\bar{q}g_s\sigma gq\rangle+\langle\bar{s}g_s\sigma gs\rangle\right]}{1536\pi^4}\int dydz \left(\frac{1}{y}+\frac{1}{z}\right)(1-y - z)\nonumber\\ & & + \frac{m_s m_c\left[9\langle\bar{s}s\rangle\langle\bar{q}g_s\sigma gq\rangle-7\langle\bar{s}s\rangle\langle\bar{s}g_s\sigma gs\rangle+16\langle\bar{q}q\rangle\langle\bar{s}g_s\sigma gs\rangle\right]}{9216\pi^4}\int dydz \left(\frac{1}{y}+\frac{1}{z}\right ) \nonumber\\ & & + \frac{m_s\langle\bar{q}q\rangle\langle\bar{s}s\rangle^2}{108\pi^2}\int dy \ , \left [ 1+\frac{s}{2}\delta\left(s-\widetilde{m}_c^2 \right)\right ] \nonumber\end{aligned}\ ] ] @xmath240}{9216\pi^4}\int dydz \left(\frac{1}{y}+\frac{1}{z}\right)\delta \left(s-\overline{m}_c^2 \right)\nonumber\\ & & -\frac{m_s m_c\langle\bar{s}g_s\sigma gs\rangle \left[41\langle\bar{q}g_s\sigma gq\rangle-7\langle\bar{s}g_s\sigma gs\rangle \right]}{55296\pi^4}\int dy \left(\frac{1}{y}+\frac{1}{1-y}\right)\delta
\left(s-\widetilde{m}_c^2 \right)\nonumber\\ & & -\frac{m_s m_c\langle\bar{s}g_s\sigma gs\rangle\left[19\langle\bar{s}g_s\sigma gs\rangle-125\langle\bar{q}g_s\sigma gq\rangle \right ] } { 55296\pi^4}\int dy \left(1+\frac{s}{t^2}\right ) \delta\left(s-\widetilde{m}_c^2 \right ) \nonumber\\ & & + \frac{19 m_s m_c \langle \bar{s}g_s\sigma gs\rangle \left[\langle \bar{q}g_s\sigma gq\rangle+\langle \bar{s}g_s\sigma gs\rangle\right]}{55296\pi^4}\int dy \,\delta\left(s-\widetilde{m}_c^2 \right ) \ , , \end{aligned}\ ] ] @xmath241}{4608\pi^6}\int dydz \ , ( y+z)(1-y - z)^2\left(s-\overline{m}_c^2\right)^3 \nonumber \\ & & + \frac{m_c\left[2\langle \bar{q}q\rangle+\langle \bar{s}s\rangle\right]}{13824\pi^6}\int dydz \ , ( y+z)(1-y - z)^3\left(s-\overline{m}_c^2\right)^3 \nonumber \\ & & + \frac{11m_c\left[2\langle \bar{q}g_s\sigma gq\rangle+\langle \bar{s}g_s\sigma gs\rangle\right]}{24576\pi^6}\int dydz \ , ( y+z)(1-y - z ) \left(s-\overline{m}_c^2 \right)^2 \nonumber\\ & & - \frac{11m_c\left[2\langle \bar{q}g_s\sigma gq\rangle+\langle \bar{s}g_s\sigma gs\rangle\right]}{49152\pi^6}\int dydz \ , ( y+z)(1-y - z)^2 \left(s-\overline{m}_c^2 \right)^2 \nonumber\\ & & - \frac{7m_c\left[2\langle \bar{q}g_s\sigma gq\rangle+\langle \bar{s}g_s\sigma gs\rangle\right]}{49152\pi^6}\int dydz \ , \left(\frac{y}{z}+\frac{z}{y}\right)(1-y - z)^2 \left(s-\overline{m}_c^2 \right)^2 \nonumber\\ & & + \frac{m_c\left[2\langle \bar{q}g_s\sigma gq\rangle+\langle \bar{s}g_s\sigma gs\rangle\right]}{24576\pi^6}\int dydz \ , \left(\frac{y}{z}+\frac{z}{y}\right)(1-y - z)^3 \left(s-\overline{m}_c^2 \right)^2 \nonumber\\ & & + \frac{\langle\bar{q}q\rangle\left[\langle\bar{q}q\rangle+2\langle\bar{s}s\rangle \right]}{288\pi^4}\int dydz \ , yz(1-y - z)\left(s-\overline{m}_c^2 \right)\left(5s-3\overline{m}_c^2 \right ) \nonumber\\ & & -\frac{\langle\bar{q}q\rangle\langle\bar{q}g_s\sigma gq\rangle+\langle\bar{q}q\rangle\langle\bar{s}g_s\sigma gs\rangle+\langle\bar{s}s\rangle\langle\bar{q}g_s\sigma gq\rangle}{9216\pi^4}\int dydz \,(y+z)(1-y - z ) \left(2s-\overline{m}_c^2\right)\nonumber\\ & & -\frac{19\left[\langle\bar{q}q\rangle\langle\bar{q}g_s\sigma gq\rangle+\langle\bar{q}q\rangle\langle\bar{s}g_s\sigma gs\rangle+\langle\bar{s}s\rangle\langle\bar{q}g_s\sigma gq\rangle\right]}{4608\pi^4}\int dydz \,yz \left(4s-3\overline{m}_c^2\right)\nonumber\\ & & -\frac{m_c\langle\bar{q}q\rangle^2\langle\bar{s}s\rangle}{72\pi^2}\int dy \nonumber\\ & & + \frac{11\langle\bar{q}g_s\sigma gq\rangle\left[\langle\bar{q}g_s\sigma gq\rangle+2\langle\bar{s}g_s\sigma gs\rangle \right]}{6144\pi^4}\int dy \ , y(1-y)\left [ 1+\frac{s}{3}\delta\left(s-\widetilde{m}_c^2 \right)\right ] \nonumber\\ & & + \frac{13\langle\bar{q}g_s\sigma gq\rangle \left [ \langle\bar{q}g_s\sigma gq\rangle+2\langle\bar{s}g_s\sigma gs\rangle\right]}{331776\pi^4}\int dydz \,(y+z ) \left[1+s\,\delta \left(s-\overline{m}_c^2 \right)\right]\nonumber\\ & & + \frac{m_s m_c}{73728\pi^8}\int dydz \ , ( y+z)(1-y - z)^3\left(s-\overline{m}_c^2\right)^4 \nonumber\\ & & -\frac{m_s m_c}{294912\pi^8}\int dydz \ , ( y+z)(1-y - z)^4\left(s-\overline{m}_c^2\right)^4 \nonumber\\ & & - \frac{m_s\left[4\langle \bar{q}q\rangle-3\langle \bar{s}s\rangle\right]}{1536\pi^6}\int dydz \ , yz(1-y - z)^2\left(s-\overline{m}_c^2\right)^2 \left(2s-\overline{m}_c^2 \right ) \nonumber \\ & & - \frac{m_s\langle \bar{s}s\rangle}{4608\pi^6}\int dydz \ , yz(1-y - z)^3\left(s-\overline{m}_c^2\right)^2 \left(7s-4\overline{m}_c^2 \right ) \nonumber\end{aligned}\ ] ] @xmath242}{12288\pi^6}\int dydz \ , yz(1-y - z ) \left(s-\overline{m}_c^2 \right)\left(5s-3\overline{m}_c^2
\right ) \nonumber\\ & & + \frac{m_s\langle \bar{s}g_s\sigma gs\rangle}{1536\pi^6}\int dydz \ , yz(1-y - z)^2 \left(s-\overline{m}_c^2 \right)\left(3s-2\overline{m}_c^2 \right ) \nonumber\\ & & + \frac{m_s\langle \bar{q}g_s\sigma gq\rangle}{49152\pi^6}\int dydz \ , ( y+z)(1-y - z)^2 \left(s-\overline{m}_c^2 \right)\left(3s-\overline{m}_c^2 \right ) \nonumber\\ & & + \frac { m_s m_c\langle\bar{q}q\rangle \left[3\langle\bar{q}q\rangle-\langle\bar{s}s\rangle \right]}{288\pi^4}\int dydz \ , ( y+z)\left(s-\overline{m}_c^2 \right ) \nonumber\\ & & + \frac { m_s m_c\langle\bar{q}q\rangle\langle\bar{s}s\rangle } { 288\pi^4}\int dydz \ , ( y+z)(1-y - z)\left(s-\overline{m}_c^2 \right ) \nonumber\\ & & -\frac{m_s m_c\left[32\langle\bar{q}q\rangle\langle\bar{s}g_s\sigma gs\rangle+57\langle\bar{s}s\rangle\langle\bar{q}g_s\sigma gq\rangle\right]}{55296\pi^4}\int dydz \ ,
\left(y+z\right ) \nonumber\\ & & -\frac{m_s m_c \left[288\langle\bar{q}q\rangle\langle\bar{q}g_s\sigma gq\rangle-32\langle\bar{q}q\rangle\langle\bar{s}g_s\sigma gs\rangle-57\langle\bar{s}s\rangle\langle\bar{q}g_s\sigma gq\rangle\right]}{55296\pi^4}\int dy \nonumber\\ & & + \frac{m_s m_c\langle\bar{s}s\rangle\langle\bar{q}g_s\sigma gq\rangle}{1536\pi^4}\int dydz \left(\frac{y}{z}+\frac{z}{y}\right)(1-y - z)\nonumber\\ & & + \frac{m_s m_c\left[16\langle\bar{q}q\rangle-7\langle\bar{s}s\rangle\right]\langle\bar{q}g_s\sigma gq\rangle}{9216\pi^4}\int dydz \ , \left(\frac{y}{z}+\frac{z}{y}\right ) \nonumber\\ & & + \frac{m_s\langle\bar{q}q\rangle^2\langle\bar{s}s\rangle}{72\pi^2}\int dy \,y(1-y ) \left [ 1+\frac{s}{3}\delta\left(s-\widetilde{m}_c^2 \right)\right ] \nonumber\\ & & -\frac{m_s m_c\langle\bar{q}g_s\sigma gq\rangle \langle\bar{s}g_s\sigma gs\rangle}{6912\pi^4}\int dy \left(1+\frac{s}{t^2}\right ) \delta\left(s-\widetilde{m}_c^2 \right ) \nonumber\\ & & -\frac{m_s m_c\langle\bar{q}g_s\sigma gq\rangle \langle\bar{s}g_s\sigma gs\rangle}{9216\pi^4}\int dydz
\left(\frac{z}{y}+\frac{y}{z}\right)\delta \left(s-\overline{m}_c^2 \right)\nonumber\\ & & -\frac{m_s m_c\langle\bar{q}g_s\sigma gq\rangle\left[24\langle\bar{q}g_s\sigma gq\rangle-7\langle\bar{s}g_s\sigma gs\rangle \right ] } { 55296\pi^4}\int dy \left ( \frac{1-y}{y}+\frac{y}{1-y}\right ) \delta\left(s-\widetilde{m}_c^2 \right)\nonumber\\ & & + \frac{m_s m_c\langle\bar{q}g_s\sigma gq\rangle \left [ 24\langle\bar{q}g_s\sigma gq\rangle-\langle\bar{s}g_s\sigma gs\rangle\right]}{18432\pi^4}\int dy \,\left(1+\frac{s}{2t^2}\right)\delta\left(s-\widetilde{m}_c^2 \right ) \nonumber\\ & & + \frac{m_s m_c \langle \bar{q}g_s\sigma gq\rangle\langle \bar{s}g_s\sigma gs\rangle}{36864\pi^4}\int dy \,\delta\left(s-\widetilde{m}_c^2 \right ) \ , , \end{aligned}\ ] ] @xmath243}{2304\pi^6}\int dydz \ , ( 1-y - z)^2\left(s-\overline{m}_c^2\right)^3 \nonumber \\ & & + \frac{m_c\left[2\langle \bar{q}q\rangle+\langle \bar{s}s\rangle\right]}{6912\pi^6}\int dydz \ ,
( 1-y - z)^3\left(s-\overline{m}_c^2\right)^3 \nonumber \\ & & + \frac{11m_c\left[2\langle \bar{q}g_s\sigma gq\rangle+\langle \bar{s}g_s\sigma gs\rangle\right]}{12288\pi^6}\int dydz \ , ( 1-y - z ) \left(s-\overline{m}_c^2
\right)^2 \nonumber\\ & & - \frac{11m_c\left[2\langle \bar{q}g_s\sigma gq\rangle+\langle \bar{s}g_s\sigma gs\rangle\right]}{24576\pi^6}\int dydz \
, ( 1-y - z)^2 \left(s-\overline{m}_c^2 \right)^2 \nonumber\\ & & - \frac{7 m_c\left[2\langle \bar{q}g_s\sigma gq\rangle+\langle \bar{s}g_s\sigma gs\rangle\right]}{49152\pi^6}\int dydz \ , \left(\frac{1}{y}+\frac{1}{z}\right)(1-y - z)^2 \left(s-\overline{m}_c^2 \right)^2 \nonumber\end{aligned}\ ] ] @xmath244}{24576\pi^6}\int dydz \
, \left(\frac{1}{y}+\frac{1}{z}\right)(1-y - z)^3 \left(s-\overline{m}_c^2 \right)^2 \nonumber\\ & & + \frac{\langle\bar{q}q\rangle \left[\langle\bar{q}q\rangle+2\langle\bar{s}s\rangle \right]}{288\pi^4}\int dydz \ , ( y+z)(1-y - z)\left(s-\overline{m}_c^2 \right)\left(2s-\overline{m}_c^2 \right ) \nonumber\\ & & -\frac{19\left[\langle\bar{q}q\rangle\langle\bar{q}g_s\sigma gq\rangle+\langle\bar{q}q\rangle\langle\bar{s}g_s\sigma gs\rangle+\langle\bar{s}s\rangle\langle\bar{q}g_s\sigma gq\rangle\right]}{9216\pi^4}\int dydz \,(y+z ) \left(3s-2\overline{m}_c^2\right)\nonumber\\ & & -\frac{\langle\bar{q}q\rangle\langle\bar{q}g_s\sigma gq\rangle+\langle\bar{q}q\rangle\langle\bar{s}g_s\sigma gs\rangle+\langle\bar{s}s\rangle\langle\bar{q}g_s\sigma gq\rangle}{4608\pi^4}\int dydz \ , ( 1-y - z ) s\nonumber\\ & & -\frac{m_c\langle\bar{q}q\rangle^2\langle\bar{s}s\rangle}{36\pi^2}\int dy \nonumber\\ & & + \frac{11\langle\bar{q}g_s\sigma gq\rangle \left[\langle\bar{q}g_s\sigma gq\rangle+2\langle\bar{s}g_s\sigma gs\rangle \right]}{18432\pi^4}\int dy \ , \left [ 1+\frac{s}{2}\delta\left(s-\widetilde{m}_c^2 \right)\right ] \nonumber\\ & & + \frac{13\langle\bar{q}g_s\sigma gq\rangle \left[\langle\bar{q}g_s\sigma gq\rangle+2\langle\bar{s}g_s\sigma gs\rangle \right]}{165888\pi^4}\int dydz \ , s\,\delta \left(s-\overline{m}_c^2 \right ) \nonumber\\ & & + \frac{m_s m_c}{36864\pi^8}\int dydz \ , ( 1-y - z)^3\left(s-\overline{m}_c^2\right)^4 \nonumber\\ & & -\frac{m_s m_c}{147456\pi^8}\int dydz \ , ( 1-y - z)^4\left(s-\overline{m}_c^2\right)^4 \nonumber\\ & & - \frac{m_s\left[4\langle \bar{q}q\rangle-3\langle \bar{s}s\rangle\right]}{9216\pi^6}\int dydz \ , ( y+z)(1-y - z)^2\left(s-\overline{m}_c^2\right)^2 \left(5s-2\overline{m}_c^2 \right )
\nonumber \\ & & - \frac{m_s\langle \bar{s}s\rangle}{3072\pi^6}\int dydz \ , ( y+z)(1-y - z)^3\left(s-\overline{m}_c^2\right)^2 \left(2s-\overline{m}_c^2 \right ) \nonumber \\ & & + \frac{m_s\left[19\langle \bar{q}g_s\sigma gq\rangle-8\langle \bar{s}g_s\sigma gs\rangle\right]}{12288\pi^6}\int dydz \ , ( y+z)(1-y - z ) \left(s-\overline{m}_c^2 \right)\left(2s-\overline{m}_c^2
\right ) \nonumber\\ & & + \frac{m_s\langle \bar{s}g_s\sigma gs\rangle}{6144\pi^6}\int dydz \ , ( y+z)(1-y - z)^2 \left(s-\overline{m}_c^2 \right)\left(5s-3\overline{m}_c^2 \right ) \nonumber\\ & & + \frac{m_s\langle \bar{q}g_s\sigma gq\rangle}{12288\pi^6}\int dydz \ , ( 1-y - z)^2 \,s\left(s-\overline{m}_c^2 \right ) \nonumber\\ & & + \frac { m_s m_c\langle\bar{q}q\rangle\left[3\langle\bar{q}q\rangle-\langle\bar{s}s\rangle \right]}{144\pi^4}\int dydz \ , \left(s-\overline{m}_c^2 \right ) \nonumber\\ & & + \frac { m_s m_c\langle\bar{q}q\rangle\langle\bar{s}s\rangle } { 144\pi^4}\int dydz \ , ( 1-y - z)\left(s-\overline{m}_c^2 \right ) \nonumber\\ & & -\frac{m_s m_c\left[32\langle\bar{q}q\rangle\langle\bar{s}g_s\sigma gs\rangle+57\langle\bar{s}s\rangle\langle\bar{q}g_s\sigma gq\rangle\right]}{27648\pi^4}\int dydz \nonumber\\ & & -\frac{m_s m_c\left[288 \langle\bar{q}q\rangle\langle\bar{q}g_s\sigma gq\rangle-32\langle\bar{q}q\rangle\langle\bar{s}g_s\sigma gs\rangle-57\langle\bar{s}s\rangle\langle\bar{q}g_s\sigma gq\rangle\right]}{27648\pi^4}\int dy \nonumber\\ & & + \frac{m_s m_c\langle\bar{s}s\rangle\langle\bar{q}g_s\sigma gq\rangle}{1536\pi^4}\int dydz \left(\frac{1}{y}+\frac{1}{z}\right)(1-y - z)\nonumber\\ & & + \frac{m_s m_c\left[16\langle\bar{q}q\rangle-7\langle\bar{s}s\rangle\right]\langle\bar{q}g_s\sigma gq\rangle}{9216\pi^4}\int dydz \ ,
\left(\frac{1}{y}+\frac{1}{z}\right ) \nonumber\\ & & + \frac{m_s\langle\bar{q}q\rangle^2\langle\bar{s}s\rangle}{216\pi^2}\int dy \ ,
\left [ 1+\frac{s}{2}\delta\left(s-\widetilde{m}_c^2 \right)\right ] \nonumber\\ & & -\frac{m_s m_c\langle\bar{q}g_s\sigma gq\rangle \langle\bar{s}g_s\sigma gs\rangle}{3456\pi^4 t^2}\int dy \ , s\ , \delta\left(s-\widetilde{m}_c^2 \right ) \nonumber\end{aligned}\ ] ] @xmath245}{55296\pi^4}\int dy \left ( \frac{1}{y}+\frac{1}{1-y}\right ) \delta\left(s-\widetilde{m}_c^2 \right)\nonumber\\ & & + \frac{m_s m_c\langle\bar{q}g_s\sigma gq\rangle\left[24\langle\bar{q}g_s\sigma gq\rangle-\langle\bar{s}g_s\sigma gs\rangle\right]}{18432\pi^4}\int dy \,\left(1+\frac{s}{t^2}\right)\delta\left(s-\widetilde{m}_c^2 \right ) \nonumber\\ & & + \frac{m_s m_c \langle \bar{q}g_s\sigma gq\rangle \langle \bar{s}g_s\sigma gs\rangle}{18432\pi^4}\int dy \,\delta\left(s-\widetilde{m}_c^2 \right ) \ , , \end{aligned}\ ] ] @xmath246 where @xmath247 , @xmath248 , @xmath249 , @xmath250 , @xmath251 , @xmath252 , @xmath253 , @xmath254 , @xmath255 , when the @xmath256 functions @xmath257 and @xmath258 appear .
this work is supported by national natural science foundation , grant number 11375063 , and natural science foundation of hebei province , grant number a2014502017 .
r. chen , x. liu , x. q. li and s. l. zhu , phys .
* 115 * ( 2015 ) 132002 ; h. x. chen , w. chen , x. liu , t. g. steele and s. l. zhu , phys .
* 115 * ( 2015 ) 172001 ; u. g. meissner and j. a. oller , phys . lett .
* b751 * ( 2015 ) 59 ; c. w. xiao and u. g. meissner , phys . rev .
* d92 * ( 2015 ) 114002 ; n. n. scoccola , d. o. riska and m. rho , phys . rev .
* d92 * ( 2015 ) 051501 ; l. roca , j. nieves and e. oset , phys . rev .
* d92 * ( 2015 ) 094003 ; t. j. burns , eur .
j. * a51 * ( 2015 ) 152 ; h. x. chen , l. s. geng , w. h. liang , e. oset , e. wang and j. j. xie , arxiv:1510.01803 ; g. yang and j. ping , arxiv:1511.09053 ; j. he , phys . lett .
* b753 * ( 2016 ) 547 ; e. wang , h. x. chen , l. s. geng , d. m. li and e. oset , phys . rev .
* d93 * ( 2016 ) 094001 ; g. j. wang , l. ma , x. liu and s. l. zhu , phys . rev .
* d93 * ( 2016 ) 034031 . v. v. anisovich , m. a. matveev , j. nyiri , a. v. sarantsev and a. n. semenova , arxiv:1507.07652 ; r. ghosh , a. bhattacharya and b. chakrabarti , arxiv:1508.00356 ; v. v. anisovich , m. a. matveev , j. nyiri , a. v. sarantsev and a. n. semenova , int . j. mod
* a30 * ( 2015 ) 1550190 ; l. maiani , a. d. polosa and v. riquer , phys . lett . * b750 * ( 2015 ) 37 ; h. y. cheng and c. k. chua , phys .
* d92 * ( 2015 ) 096009 ; v. v. anisovich , m. a. matveev , a. v. sarantsev and a. n. semenova , mod .
a30 * ( 2015 ) 1550212 ; g. n. li , m. he and x. g. he , jhep * 1512 * ( 2015 ) 128 .
q. wang , x. h. liu and q. zhao , phys . rev .
* d92 * ( 2015 ) 034022 ; v. kubarovsky and m. b. voloshin , phys . rev .
* d92 * ( 2015 ) 031502 ; q. f. lu , x. y. wang , j. j. xie , x. r. chen and y. b. dong , phys . rev .
* d93 * ( 2016 ) 034009 ; m. karliner and j. l. rosner , phys
* b752 * ( 2016 ) 329 .
z. g. wang , phys .
b685 * ( 2010 ) 59 ; z. g. wang , eur . phys .
j. * a45 * ( 2010 ) 267 ; z. g. wang , eur .
j. * c68 * ( 2010 ) 459 ; z. g. wang , eur .
j. * a47 * ( 2011 ) 81 ; z. g. wang , commun . theor . phys . * 58 * ( 2012 ) 723 .
|
in this article , we construct the axialvector - diquark - scalar - diquark - antiquark type and the axialvector - diquark - axialvector - diquark - antiquark type interpolating currents , and study the @xmath0 hidden - charm pentaquark states with the strangeness @xmath1 systematically using the qcd sum rules .
the predicted masses of the pentaquark states @xmath2 , @xmath3 and @xmath4 are compatible with the experimental value @xmath5 from the lhcb collaboration , more experimental data are still needed to identify the @xmath6 unambiguously .
+ zhi - gang wang + department of physics , north china electric power university , baoding 071003 , p. r. china pacs number : 12.39.mk , 14.20.lq , 12.38.lg key words : pentaquark states , qcd sum rules
|
the layered structure of the transition metal dichalcogenide tas@xmath1 permits the intercalation of other metal ions , nominally changing the formula to @xmath6tas@xmath1 . if @xmath7 is a 3@xmath8-transition metal , under certain circumstances long range magnetic order may occur .
the details of the order strongly depend on the type , @xmath7 , and amount , @xmath9 , of intercalated ion .
if @xmath10 fe , long range magnetic order is observed for @xmath11 .
further , for @xmath121/4 and 1/3 ordered superlattice form with basal plane axes a=2a@xmath13 and a=@xmath14a@xmath13 , respectively @xcite .
the magnetic ordering temperature is non - monotonic with doping concentration , reaching a maximum at approximately 155 k for @xmath15 @xcite .
band structure calculations for @xmath16 suggest a large imbalance between the two spin directions for the conduction electrons contributed by the iron @xcite , which may hint at a band component of the ferromagnetism . for the @xmath15 concentration ,
however , the measured ratio of the paramagnetic moment ( @xmath174.9 @xmath18/fe ) to saturation ( @xmath174 @xmath18/fe ) moments is only slightly larger than 1 @xcite .
this suggests that for @xmath15 the system is more localized than the case where @xmath16 @xcite .
recent work on single crystals of fe@xmath0tas@xmath1 has shown extremely sharp switching fields in the @xmath19 loops @xcite and anomalous hall effect and magnetoresistance associated with this sharp switching @xcite .
the time dependence of the switching was reported in ref . .
there it was found that the switching speed increased almost linearly with temperature in the range @xmath20 k. the time for magnetization reversal fell in the range of 0.1 - 0.8 ms .
several effects govern the rate of magnetization reversal in materials ; the dynamics of domain evolution and eddy current dissipation in the sample due to the changing magnetic induction being of primary concern .
in this paper we study the magnetic domain structure and formation in single crystals of fe@xmath0tas@xmath1 by utilizing direct magneto - optical ( mo ) visualization based on the faraday effect in transparent ferrimagnetic iron - garnet indicators .
the magneto - optic study reveals unusual slow domain formation and sensitivity of the domain structure on the magnetic and thermal history .
in addition , the radio frequency ( rf ) magnetic susceptibility was measured at 27 mhz via a sensitive tunnel diode resonator ( tdr ) technique .
this technique has been shown to be a sensitive tool to distinguish between local moment and itinerant ferromagnetism @xcite .
the results from fe@xmath0tas@xmath1 strongly suggest local moment ferromagnetic phase transition at 155 k. in addition , rf susceptibility shows thermal hysteresis at approximately the same temperatures that domains are seen to form in mo data .
single crystals of fe@xmath0tas@xmath1 were grown via iodine vapor transport and characterized as described in ref . .
samples were platelets with the crystallographic @xmath21axis perpendicular to the plane of the plate .
this axis is also the magnetic easy axis in these materials .
two different samples have been measured and they produced almost identical results . magneto - optical (
mo ) imaging was performed in a flow - type optical @xmath22he cryostat using faraday rotation of polarized light in an indicator based on bi - doped iron - garnet films with in - plane magnetization @xcite .
the indicator is placed directly on top of a sample with two parallel surfaces in the @xmath9-@xmath23 - plane .
the light propagates through the indicator along the @xmath24 axis and reflects back from a mirror sputtered at the bottom of the indicator ; the sample itself is not seen .
the @xmath24 component of local magnetization in the indicator is proportional to the @xmath24 component of the magnetic induction on the sample surface .
optical contrast shows the distribution of local magnetization in the sample , with brighter areas corresponding to larger magnetic induction values .
image color provides a distinction between positive and negative field ( yellow vs. green , see electronic version of the paper with color figures ) .
radio - frequency magnetic susceptibility , @xmath25 , has been measured using a tunnel - diode resonator technique ( tdr)@xcite .
this technique detects changes in magnetic susceptibility through induced frequency shifts in an @xmath26 circuit . for this work ,
the nominal resonant frequency was 27 mhz .
the tank circuit is self - resonating with losses compensated by a properly biased tunnel diode .
the sample is mounted on a sapphire rod and inserted in the coil .
the coil and sample are thermally isolated from one another .
changes in the sample s magnetic susceptibility induce changes in the resonant frequency and it is this frequency shift that is measured .
it is straightforward to show that @xmath27 where @xmath28 is the empty coil resonant frequency , and @xmath29 is the change in the resonance induced by placing the sample in the coil @xcite .
@xmath30 is a constant factor that depends on sample shape and the ratio of the sample to coil volumes .
determining the value of @xmath30 is difficult leading to complications when making precise statements regarding the absolute value of @xmath25 .
_ changes _ in the susceptibility may be probed with extreme sensitivity .
precise control of the temperature of the circuit components and careful circuit design together allow for a resonance stability on the order of 0.05 - 0.5 hz over several hours , a resolution of 50 parts per billion . with a typical sample size of 1 mm@xmath31
this translates to a sensitivity to changes in @xmath25 on the order of @xmath32 in absolute cgs units .
the excitation magnetic field produced by the tank coil is very small , @xmath33 moe , allowing for a weakly perturbative test of the sample .
the tdr technique has been shown to be well suited to the study of penetration depth in superconductors @xcite and various ferromagnetic properties , particularly distinguishing between local - moment and itinerant ferromagnets @xcite . in the present work
, a triangular sample of area 0.67 mm@xmath34 by 0.05 mm thick was mounted on a sapphire rod and inserted into the tank coil of a tdr mounted on a @xmath22he cryostat operating at a frequency of 27 mhz .
temperature dependent @xmath25 was measured on warming and cooling in dc bias fields up to 1 koe .
the dc field was parallel to the ac excitation field , both of which were aligned with the magnetic easy axis of the sample .
in addition to constant field temperature sweeps , @xmath35 data in the range @xmath36 koe@xmath37 koe were taken at several temperatures to explore the possibility of field dependent effects . in metallic samples the dynamic susceptibility , @xmath25 , is composed of both a spin susceptibility and a diamagnetic skin
effect screening .
the effect of the latter is determined by the chosen measurement frequency and the sample s resistivity . in the vicinity of a ferromagnetic phase transition the spin susceptibility dominates .
additionally , within the measurement frequency range employed here , domain wall resonance effects may become important ( see e. g. ref . ) .
this suggests tdr measurements may detect effects associated with domains .
the resistivity of fe@xmath0tas@xmath1 does not show any special features at and below @xmath38 .
it merely decreases due to a gradual loss of spin - disorder scattering @xcite .
therefore , the skin - effect contribution to the measured susceptibility manifests as a monotonic background .
figure [ remag ] shows the appearance of ferromagnetic domains after the sample was cooled to 5 k in @xmath39 oe magnetic field and the field was slowly increased . in these and all other mo images presented herein ,
the c - axis of the crystal is pointing out of the page . field cooling resulted in a monodomain state for the crystal . in order to nucleate a domain wall , a positive field of @xmath40xx oe was necessary .
the persistence of a monodomain state at 5 k in an oppositely oriented field implies either a small magnetic moment per iron or a large magnetocrystalline anisotropy or both .
the saturated magnetic moment per iron is approximately 4 @xmath18 which is a moderate magnetic moment . in conjunction with
the previously reported @xmath7-@xmath41 data @xcite , these images suggest a very high anisotropy for this compound . as the field
is further increased , unusual , dendritic domains form .
the domains nucleate at a weak spot , probably due to a local imperfection .
the overall domain structure observed under different conditions is different ( compare figs .
[ remag ] [ turnoff ] and [ warmup ] ) , so there is no one - to - one correlation between the distribution of the domains and the structural features of the crystal . however , the gross features of the domain evolution do reflect the general structure of the underlying hexagonal lattice .
this indicates that , in addition to the large c - axis magnetic anisotropy , there exists significant in - plane anisotropy with respect to domain wall motion .
the propagation of a domain dendrite shows that the walls are much more weakly pinned along directions parallel to the crystallographic hexagonal lattice as compared to perpendicular wall motion .
field cooling to a higher temperature ( @xmath42 k ) shows that the monodomain state is broken by lowering the applied field without changing the field direction ( fig .
[ turnoff ] ) .
again , the unusual , dendritic domains form in much the same way as at the lower temperature .
[ warmup ] shows how temperature variation affects domain formation .
the sample was cooled to @xmath43 k in an applied field of 1500 oe .
the field was lowered to zero and the temperature was slowly warmed .
no domains were observed until @xmath44 k , at which temperature the domains grew in a dendritic fashion similar to that observed for the constant temperature field sweeps . for real time videos ,
see ref . .
it should be emphasized that the observed contrast in the mo images is due to the different domains and not the domain walls themselves .
it is easy to consider such narrow domains as walls .
however , the domain walls separate the regions of different contrast ( or color , if read online ) .
this distinction is most evident in the zero field images of fig .
[ turnoff ] bottom row , second from left and all of fig .
[ warmup ] where the domain walls show up as fine , dark lines surrounding the dendritic structure .
we now turn to a discussion of the dynamic magnetic susceptibility .
figure [ zerowinset ] shows @xmath45 in arbitrary units for a single crystal of fe@xmath0tas@xmath1 measured at @xmath46 in the temperature range 2250 k upon warming and cooling .
the ferromagnetic transition is observed as a peak in @xmath25 at @xmath47 k. all rf@xmath2 data are normalized to 1 at this peak temperature . below @xmath48
the measured susceptibility drops in accordance with the resistivity decrease @xcite . in the vicinity of 55 k
there is an observed hysteresis characterized by breaks in the slope of @xmath25 vs. @xmath49 .
the inset shows the effect of applied static fields on this hysteresis .
runs performed in field were field cooled in the next lowest field and then @xmath41 was increased to the value for the run .
to illustrate , the @xmath45 data collected in an applied field of 500 oe was field cooled in 400 oe .
at base temperature ( @xmath502 k ) the field was then increased by 100 oe at 10 oe / s . a plot of @xmath51 vs. @xmath49 , as shown in fig . [ 3way](b ) , allows for a precise determination of the maximum change in slope for both the warming and cooling curves which are taken to delimit the hysteresis region . in fig .
[ 3way](a ) hysteresis region is marked as a function of field .
horizontal dashed lines denote temperatures at which field scans were performed .
figure [ 3way](c ) shows the width of the hysteresis defined as the difference between the upper and lower temperatures from fig . [
3way](a ) .
first , for low fields the hysteresis shifts to lower temperatures with a gradually decreasing width . at a critical field of approximately 700 oe there is a dramatic increase in the low temperature boundary and a corresponding decrease in the width of the hysteresis .
thereafter , the width of the hysteresis grows slightly .
figure [ chih ] presents the field dependent susceptibility data at three temperatures in and above the hysteresis region . in all curves
the sample was zero field cooled , warmed to the measurement temperature and the field was swept up to + 2.6 koe , down to -2.6 koe and back to zero . as a check on the polarity of the field
, experiments were also run sweeping to negative field first .
a slight zero field difference in @xmath25 exists , most notably for the 54 k trace .
however , this is likely due to a small temperature drift on the order of 0.4 k as determined by a comparison of the zero field temperature data .
@xmath52 k,@xmath53 data fall below the zero field thermal hysteresis , but at a field of about 100 oe it crosses the low temperature boundary . at @xmath54
k the data begins in this hysteresis region , while at @xmath55 k the sweep is done above it . from fig
[ 3way](a ) it is obvious that the lowest temperature field scan crossed the low temperature hysteresis boundary at 100 oe . at about 650 and 1000
oe it crossed the interpolated low temperature boundary twice more .
the mid temperature scan ( @xmath54 k ) did not cross any hysteresis boundary .
however , the qualitative features in both of the lower temperature scans are similar . to better compare the three field scans fig . [ deltachi ] shows the change in @xmath25 due to field taking the zero field susceptibility as a reference point .
the high temperature field sweep may be understood in the context of polarizing spins .
as the applied field increases there is an increasing preference for the spins to align with the field . the dynamic response of the moments in the sample decreases as a consequence and the rf susceptibility reflects this . at @xmath46 measured on warming and cooling in fe@xmath0tas@xmath1 crystal .
_ inset : _ detail of the hysteresis region in applied bias fields . both the rf - probe field and the static bias field
were aligned with the crystallographic c - axis.,width=340 ] within the domain formation temperature window , the effect of a magnetic field is to increase the relative fraction of domains parallel to the field at the expense of all others .
the effect of the rf probe field is to change the alignment of the spins by a small amount . in a dc bias field
there is a preferred direction for the spins to point . keeping in mind that the magnetic easy axis is aligned with the dc and rf fields , a spin residing in a domain parallel to the applied field will contribute a greater moment change in response to the rf field than one in an antiparallel domain
this can be seen by considering the thermal effects . in general a spin
will not point exactly along ( parallel or antiparallel ) the line of applied field .
rather it will be tilted away at some angle due to thermal randomization . during the half cycle when the rf field is parallel with the applied field
there will be a slightly greater tendency for all spins to point in that direction .
those spins in domains whose magnetization points along the applied field will rotate more readily toward the field direction as this will minimize both the zeeman and anisotropy energies .
spins in domains with antiparallel magnetizations will also tilt toward the applied field thereby reducing zeeman energy .
however , in this case there is an increase in anisotropy energy .
therefore , the change in the bulk moment from these antiparallel domains is reduced compared with the parallel domains . on the other half of the rf cycle
both parallel and antiparallel spins are working against the zeeman energy from the bias field and there will not be as great a response .
so as the fraction of sample in domains parallel to the applied field grows the measured @xmath2 grows as well .
at some point , the field has essentially saturated the sample and driven it into a single domain .
further increasing of the field increases the zeeman energy effectively freezing out the response of the spins .
the low field region of the 50 and 54 k scans show a characteristic field of 100 oe observed both upon increasing and decreasing the dc bias field . below this field
the susceptibility grows rapidly whereas immediately above 100 oe the change in @xmath25 is much more gradual . in a zero field cooled state , the system is not monodomain . in terms of the preceding argument ,
the rapidly increasing @xmath25 may be explained by the response of any free or nearly free domain walls to the applied field . as mentioned previously ,
this frequency band is suitable for measuring domain wall resonance effects .
it is possible that for this material 27 mhz is close to a natural frequency for the domains , in which case at least some of the observed effects would be frequency dependent .
the only way to be certain of this is to perform a careful study of the susceptibility over a frequency window of @xmath561020 mhz .
if fe@xmath57tas@xmath1 is , in fact , itinerant as band structure calculations suggest @xcite , a comparison of the present results with similar data from the other ferromagnetic members of this family may offer clues regarding to how the crossover between local moment and itinerant ferromagnetism proceeds . in conclusion
, unusual ferromagnetic domains were observed in single crystal fe@xmath0tas@xmath1 .
analysis of the rf - magnetic susceptibility and real - time dynamics of the domain formation imply that this material is a local moment ferromagnet with very large magneto - crystalline anisotropy that increases upon cooling .
further , the evolution of the domain pattern indicates that there is significant anisotropy with respect to domain wall motion within the hexagonal plane .
|
single crystals of fe@xmath0tas@xmath1 have been studied by using magneto - optical ( mo ) imaging and radio - frequency ( rf ) magnetic susceptibility , @xmath2 .
real time mo images reveal unusual , slow dynamics of dendritic domain formation , the details of which are strongly dependent upon magnetic and thermal history .
measurements of @xmath3 show well - defined , local moment ferromagnetic transition at @xmath4 k as well as thermal hysteresis for 50 k@xmath560 k. this temperature range corresponds to the domain formation temperature as determined by mo .
together these observations provide strong evidence for local moment ferromagnetism in fe@xmath0tas@xmath1 crystals with large , temperature dependent magnetic anisotropy .
|
it is well - known that a @xmath0-type three - level atomic medium can become transparent to a weak probe field by applying a strong coupling field , which is the result of the destructive interference between two different excitation pathways to the upper level .
this is the phenomenon of electromagnetically induced transparency ( eit ) @xcite .
the eit has been shown to be important for various applications such as slow light @xcite , light storage @xcite , and so on . besides
, the studies of eit have been extended to multi - level atomic systems .
the double eit windows separated by a narrow absorption peak in the probe absorption spectrum have been observed in the four - level atomic systems @xcite .
recently , the eit effect has been reported in the macroscopic optomechanical systems .
the analogy of eit in optomechanical systems has been shown theoretically @xcite , and observed in a number of experiments in optical cavities @xcite and microwave cavities @xcite .
moreover eit in optomechanical systems in the nonlinear regime was analyzed @xcite .
additionally , the electromagnetically induced absorption , the opposite effect to eit , was discussed in a two - cavity optomechanical system @xcite . in addition
, it has been proven that a strong dispersive coupling between the optical mode and the mechanical mode when the effective optomechanical coupling rate exceeds the optical and mechanical decay rates leads to normal mode splitting @xcite .
the effective optomechanical coupling rate can be enhanced by increasing the power of the driving laser . in this paper
, we investigate the nonlinear response of a ring cavity with two moving mirrors having two close frequencies to a weak probe field in the absence and the presence of a strong coupling field .
we find that there are two transparency windows and an absorption peak in the transmitted probe field in the presence of the coupling field .
the pump - induced broadening of the two eit dips leads to narrowing the central absorption peak . and
the narrow absorption peak associated with double eit windows may have potential application in high - resolution laser spectroscopy @xcite .
we also observe the normal mode splitting in the output probe field at high driving power .
in addition , we show that the normal mode splitting occurs in the stokes field generated by means of the four - wave mixing process . and the four - wave mixing process is completely suppressed on resonance . however , if two movable mirrors in a ring cavity are oscillating at identical frequencies , the eit - like dip can be observed in the output probe field , and the four - wave mixing process is not suppressed on resonance . the paper is organized as follows . in sec .
ii , we introduce the system , give the time evolutions of the expectation values of the system operators , and solve them . in sec .
iii , the expressions for the components of the output field at the probe frequency and the stokes frequency are given . in sec .
iv , we present the numerical results for the output probe field without or with the coupling field , and compare it with that from a ring cavity with two movable mirrors having equal frequencies . in sec .
v , we show the numerical results for the output stokes field from a ring cavity with two moving mirrors having different or equal frequencies , and compare them . finally in sec .
vi , we conclude the paper .
we consider a ring cavity with round - trip length @xmath1 formed by three mirrors , as shown in fig .
[ fig1 ] @xcite .
one of them is not movable and partially transmitting , while the other two are allowed to vibrate and assumed to have 100% reflectivity .
the cavity field at the resonance frequency @xmath2 is driven by a strong coupling field with amplitude @xmath3 at frequency @xmath4 .
meanwhile a weak probe field with amplitude @xmath5 at frequency @xmath6 is sent into the cavity .
the coupling field and the probe field are treated classically here .
the mechanical motions of both movable mirrors are coupled to the cavity field through the radiation pressure force exerted by the photons in the cavity .
the movable mirrors dynamics can be approximated as those of a single harmonic oscillator , with resonance frequency @xmath7 , effective mass @xmath8 , damping rate @xmath9 ( @xmath10 ) .
we assume that the system is in the adiabatic limit where the mechanical frequencies @xmath7 ( @xmath10 ) are much smaller than the cavity free spectral range @xmath11 ( @xmath12 is the speed of light in vacuum ) .
the adiabatic limit @xmath13 ( @xmath10 ) implies that the mechanical frequencies @xmath7 ( @xmath10 ) are very small compared to the cavity resonance frequency @xmath2 , so the moving mirrors are moving so slow that the retardation effect , the casimir effect , and the doppler effect become completely negligible @xcite .
hence the radiation pressure force does not depend on the velocity of the movable mirrors .
assuming that @xmath14 is the wave vector of the cavity field with @xmath15 , the radiation pressure force exerted by the light field on the movable mirror can be calculated from the momentum exchange between the cavity field and the movable mirror , which is @xmath16 , where @xmath17 is the angle between the incident light and the reflected light at the surfaces of the movable mirrors , @xmath18 is the photon number operator of the cavity field , @xmath19 and @xmath12 are the creation and annihilation operators of the cavity field , obey the standard commutation relation @xmath20=1 $ ] . note that the radiation pressure forces acting on the movable mirrors vary linearly with the instantaneous photon number in the cavity . under the action of the radiation pressure force ,
the movable mirrors make small oscillations .
then the small motion of the mirrors changes the length of the optical cavity , and alters the intensity of the cavity field , which in turn modifies the radiation pressure force acting on the mirrors .
therefore the interaction between the cavity field and the mechanical motion in the ring cavity is nonlinear . in a frame rotating at the driving frequency @xmath4
, the hamiltonian of the whole system takes the form @xmath21 here the first three terms are the free energies of the cavity field and two mechanical oscillators , respectively , @xmath22 denote the dimensionless position and momentum quadratures of the two mirrors , @xmath23 , @xmath24 ( @xmath25 ) , and @xmath26=i\delta_{j , k}$ ] ( @xmath27 ) .
the fourth term describes the nonlinear optomechanical interactions between the cavity field and the two movable mirrors , @xmath28 ( @xmath25 ) is the optomechanical coupling strength .
the last two terms give the interactions of the cavity field with the coupling field and the probe field , respectively , @xmath3 is related to the power @xmath29 of the coupling field by @xmath30 , where @xmath31 is the decay rate of the cavity due to the transmission losses through the fixed mirror , @xmath5 is related to the power @xmath32 of the probe field by @xmath33 , @xmath34 is the detuning of the probe field from the coupling field .
starting from the heisenberg equations of motion , taking into account the dissipations of the cavity field and the mechanical oscillators , and neglecting quantum noise and thermal noise , we obtain the time evolutions of the expectation values of the system operators @xmath35\langle c\rangle+\varepsilon\nonumber\\ & & + \varepsilon_{p}e^{-i\delta t}-\kappa \langle c\rangle,\end{aligned}\ ] ] where we have used the mean field assumption @xmath36 and @xmath37 ( @xmath10 ) . since the probe field is much weaker than the coupling field @xmath38 , the steady - state solution to eq .
( [ 2 ] ) can be approximated to the first order in the probe field @xmath5 . in the long time
limit , the solution to eq .
( [ 2 ] ) can be written as @xmath39 where @xmath40 , @xmath41 , @xmath42 , @xmath43 , or @xmath12 .
the solution contains three components , which in the original frame oscillate at @xmath4 , @xmath6 , @xmath44 , respectively . substituting eq .
( [ 3 ] ) into eq .
( [ 2 ] ) , equating coefficients of @xmath45 and @xmath46 , we find the following analytical expressions @xmath47(\omega_{1}^2-\delta^2-i\gamma_{1}\delta)\nonumber\\ & & \times(\omega_{2}^2-\delta^2-i\gamma_{2}\delta)+i[g_{1}^2\omega_{1}(\omega_{2}^2-\delta^2-i\gamma_{2}\delta)\nonumber\\ & & + g_{2}^2\omega_{2}(\omega_{1}^2-\delta^2-i\gamma_{1}\delta)]\big\},\nonumber\\ c_{-}&=&\frac{ic_{0}^2}{|c_{0}|^2d(\delta)^{*}}[g_{1}^2\omega_{1}(\omega_{2}^2-\delta^2+i\gamma_{2}\delta)\nonumber\\ & & + g_{2}^2\omega_{2}(\omega_{1}^2-\delta^2+i\gamma_{1}\delta)],\end{aligned}\ ] ] where @xmath48 and @xmath49 are the effective optomechanical coupling rates , @xmath50 is the effective detuning of the coupling field from the cavity resonance frequency , including the frequency shift induced by the radiation pressure , and @xmath51[\kappa - i(\delta'+\delta)](\omega_{1}^2-\delta^2-i\gamma_{1}\delta)\nonumber\\ & & \times(\omega_{2}^2-\delta^2-i\gamma_{2}\delta)-2\delta'[g_{1}^2\omega_{1}(\omega_{2}^2-\delta^2-i\gamma_{2}\delta)\nonumber\\ & & + g_{2}^2\omega_{2}(\omega_{1}^2-\delta^2-i\gamma_{1}\delta)].\end{aligned}\ ] ]
the output field can be obtained by using the input - output relation @xmath52 . in analogy with eq .
( [ 3 ] ) , we expand the output field to the first order in the probe field @xmath5 , @xmath53 where @xmath54 , @xmath55 , and @xmath56 are the components of the output field oscillating at frequencies @xmath4 , @xmath6 , @xmath44 . here
@xmath56 is called a stokes field , and it is generated via the nonlinear four - wave mixing process , in which two photons at frequency @xmath4 interact with a single photon at frequency @xmath6 to create a new photon at frequency @xmath44 .
thus we find that the components of the output field at the probe frequency and the stokes frequency are @xmath57 respectively . in the absence of the coupling field @xmath58 ,
the components of the output field at the probe frequency and the stokes frequency are given by @xmath59 these are not unexpected results .
let us write the real part of @xmath55 as @xmath60 , which exhibits the absorption characteristic of the output field at the probe frequency .
it can be measured by the homodyne technique @xcite .
in this section , we numerically evaluate how the coupling field modifies the absorption of the ring cavity to the probe field .
the values of the parameters chosen are similar to those in @xcite : the wavelength of the coupling field @xmath61 nm , the coupling constants @xmath62 ghz / nm @xmath63 , @xmath64 ghz / nm @xmath65 , the masses of the movable mirrors @xmath66 ng , the frequencies of the movable mirrors @xmath67 , @xmath68 , where @xmath69 mhz , the cavity decay rate @xmath70 mhz , @xmath71 ( the system is placed in resolved sideband regime ) , the mechanical damping rates @xmath72 khz , the mechanical quality factors @xmath73 , @xmath74 , the angle @xmath75 . and the coupling field is tuned close to the red sideband of the cavity resonance @xmath76 .
the parameters chosen ensure the system operating in the stable regime .
we note that the structure of the quadrature of the output probe field @xmath60 is determined by @xmath77 .
the roots @xmath78 of @xmath77 are complex values .
the real parts @xmath79 of the roots determine the positions of the normal modes of the optomechanical system ; the imaginary parts @xmath80 of the roots describe their widths .
figure [ fig2 ] shows the real parts of the roots of @xmath77 in the domain @xmath81 versus the coupling beam power .
it is seen that the real parts of the roots of @xmath77 are @xmath82 at low driving power . at high driving power ,
one of the real parts is still @xmath83 ( dotdashed curve ) , not changing with increasing the power of the coupling field , the difference between the other two ( dotted curve and dashed curve ) increases with increasing the power of the coupling field , which implies the normal mode spitting @xcite in the output probe field .
figure [ fig3 ] shows the imaginary parts of the roots of @xmath77 versus the coupling beam power .
we see that the imaginary parts of the roots of @xmath77 have three different values at low driving power . at high driving power ,
the imaginary parts of the roots of @xmath77 have three values , two of them are identical , the other one is small .
moreover , in figs .
[ fig2 ] and [ fig3 ] , the dotdashed curves represent the real part and the imaginary part of one of the roots of @xmath77 , respectively , similarly for dotted curves and dashed curves .
hence , at high driving power , the central peak in the quadrature @xmath60 is narrow , two side peaks in the quadrature @xmath60 have the same broad linewidths . in fig .
[ fig4 ] , we plot the quadrature of the output probe field @xmath60 as a function of the normalized probe detuning @xmath84 for three different powers of the coupling field . in the absence of the coupling field , @xmath60 ( solid curve )
has a standard lorentzian absorption peak with a full width at half maximum ( fwhm ) of @xmath85 at the line center @xmath86 .
however , in the presence of the coupling field with power 2 mw , it is seen that the dotdashed curve exhibits two symmetric narrow eit dips centered at @xmath87 and a broad absorption peak centered at @xmath86 .
the fwhm of the two eit dips are about @xmath88 and @xmath89 , respectively .
the fwhm of the central absorption peak is about @xmath90 .
the two transparency dips display that the input probe field could be simultaneously transparent at two symmetric frequencies , which is the result of the destructive interferences between the probe field and the anti - stokes fields at frequencies @xmath91 and @xmath92 generated by the interactions of the coupling field with the movable mirrors .
the absorption peak at the line center implies that the incident probe field is almost fully absorbed by the optomechanical system .
moreover , increasing the power @xmath29 of the coupling field , the two eit dips become broader , which results in narrowing of the central absorption peak .
further , from the dotted curve in fig . [ fig4 ] , one can see that increasing the power of the coupling field to @xmath93 mw results in a larger splitting between the left and right side peaks with the same linewidths , it also results in a narrow central peak .
our calculations show that in the strong coupling limit @xmath94 , the three dressed modes of the system corresponding to the three absorption peaks are @xmath95 , and @xmath96 .
note that the fwhm for the narrow central peak at @xmath97 is about @xmath98 .
in addition , the fwhm for the right and left peaks at @xmath99 are the same @xmath100 , the splitting between the side peaks is about @xmath101 , which is proportional to the power of the coupling field .
these results are consistent with those in figs . 2 and 3 . for comparison ,
we consider the same previous system , but the two mechanical oscillators have the same frequencies @xmath102 , thus their effective optomechanical coupling rates are equal , we assume @xmath103 .
we plot the quadrature of the output probe field @xmath60 as a function of the normalized probe detuning @xmath84 for three different powers of the coupling field , as shown in fig .
when the coupling field is not present , the quadrature @xmath60 ( solid curve ) has a lorentzian absorption lineshape .
however , the presence of the control field with power 2 mw leads to a narrow eit - like dip at the line center ( dotdashed curve ) .
thus the probe field can almost completely propagate through the ring cavity on resonance with almost no absorption .
the fwhm of the eit - like dip is about @xmath104 .
the eit - like dip is attributed to the destructive interference between the input weak probe field and the scattering quantum fields at the probe frequency @xmath6 generated by the interactions of the coupling field with two mirrors having identical frequencies . moreover , in the strong coupling limit @xmath105 , the normal mode splitting exhibits in the quadrature @xmath60 ( dotted curve ) , the two peaks have the same fwhm of @xmath106 , their positions are @xmath107 , the separation between them is about @xmath108 . in order to understand the narrow central peak in fig .
[ fig4 ] and the eit - like dip in fig .
[ fig5 ] , let us introduce the relative coordinates ( @xmath109 , @xmath110 ) and center of mass coordinates ( @xmath111 , @xmath112 ) of the two movable mirrors as @xmath113 with these new coordinates we can write the hamiltonian eq .
( [ 1 ] ) as @xmath114 where @xmath115 .
the fourth term in eq .
( [ 10 ] ) describes the interaction between the relative and center of mass coordinates . the fifth term in eq .
( [ 10 ] ) shows that the interaction between the relative coordinate and the cavity field .
when the two mechanical frequencies are equal @xmath116 , the fourth term vanishes , the center of mass coordinate is decoupled from the relative coordinate , and it is also decoupled from the cavity field , so only the relative coordinate is coupled to the cavity field via radiation pressure , which generates a transparency dip in the output probe field , as shown in fig .
when the two mechanical frequencies are not equal @xmath117 , it is seen that not only the relative coordinate and the cavity field interact with each other , but also the relative and center of mass coordinates interact with each other , which leads to a narrow central peak in the output probe field , as shown in fig .
in this section , we numerically examine the effect of the coupling field on the output stokes field generated via the four - wave mixing process .
figure [ fig6 ] shows the intensity of the generated stokes field @xmath118 as a function of the normalized probe detuning @xmath84 for several values of the driving power when the frequencies of the two mirrors are not equal ( @xmath119 and @xmath120 ) .
it is seen that @xmath121 ( solid curve ) in the absence of the coupling field
. however , in the presence of the coupling field , the @xmath118 ( dotdashed curve and dotted curve ) exhibits the normal mode splitting , the peak separation increases with the power of the coupling field .
moreover , the maximum value of @xmath118 is increased with the power of the coupling field , the maximum value of @xmath118 is about 0.19 when @xmath122 mw .
note that the intensity of the stokes field goes to zero when the detuning @xmath86 .
hence , the four - wave mixing effect is completely suppressed on resonance so that there is only the probe field stored inside the optomechanical system .
the suppression of the four - wave mixing effect on resonance is due to the destructive interference between the stokes fields produced by the coupling field interacting with the two movable mirrors oscillating at frequencies @xmath119 and @xmath120 . for comparison
, we also consider the case of a ring cavity in which the frequencies of the two mirrors are equal ( @xmath102 ) , the intensity @xmath118 of the output stokes field versus the normalized probe detuning @xmath84 for several values of the driving power is shown in fig .
[ fig7 ] . the significant difference between fig .
[ fig7 ] and fig .
[ fig6 ] is that the intensity of the stokes field in fig .
[ fig7 ] is non zero at @xmath86 in the presence of the coupling field .
thus the four - wave mixing process is not suppressed at @xmath86 , which arises from the constructive interference between the stokes fields produced by the coupling field interacting with the two movable mirrors oscillating at the same frequencies @xmath102 .
so there is no apparent normal mode splitting in this case .
to summarize , we have demonstrated how a strong coupling field affects the propagation of a weak probe field in a ring cavity with two movable mirrors whose frequencies are close to each other .
we find double eit dips and a central absorption peak in the output probe field in the presence of the coupling field .
thus this system can become transparent at two different frequencies of a weak probe field . increasing the pump power
gives rise to broadening two eit dips and narrowing the central absorption peak .
in addition , we show that the coupling field induces the normal mode splitting in the output probe field and the output stokes field
. moreover , the four - wave mixing suppression takes place in this optomechanical system when the probe detuning is one - half the sum of the two mechanical frequencies .
however , when two movable mirrors in a ring cavity have identical frequencies , the response of the ring cavity to a weak probe field in the presence of a strong coupling field would be different . we observe a narrow transparency dip in the output probe field , and no four - wave mixing suppression when the probe detuning is equal to the mechanical frequency
harris s e 1997 _ phys . today _ * 50 * 36 harris s e , field j e and imamoglu a 1990 _ phys . rev .
* 64 * 1107 boller k j , imamoglu a and harris s e 1991 _ phys .
lett . _ * 66 * 2593 fleischhauer m , imamoglu a and marangos j p 2005 _ rev .
* 77 * 633 hau l v , harris s e , dutton z and behroozi c h ( 1999 ) _ nature _ ( london ) * 397 * 594 liu c , dutton z , behroozi c h and hau l v ( 2001 ) _ nature _ ( london ) * 409 * 490 lukin m d , yelin s f , fleischhauer m and scully m o 1999 _ phys .
a _ * 60 * 3225 gavra n , rosenbluh m , zigdon t , wilson - gordon a d and friedmann h 2007 _ opt . commun . _
* 280 * 374 goren c , wilson - gordon a d , rosenbluh m and friedmann h 2004 _ phys .
a _ * 69 * 063802 ye c y , zibrov a s , rostovtsev y v and scully m o 2002 _ phys . rev .
a _ * 65 * 043805 paspalakis e and knight p l 2002 _ j. mod .
opt . _ * 49 * 87
|
we study the effect of a strong coupling field on the absorptive property of a ring cavity with two mirrors oscillating at slightly different frequencies to a weak probe field .
we observe double electromagnetically induced transparency windows separated by an absorption peak at line center in the output probe field under the action of a strong coupling field .
we find that increasing driving power can broaden the two transparency windows , which results in narrowing of the central absorption peak . at high driving power ,
the linewidth of the sharp central absorption peak is approximately equal to the mechanical linewidth .
we show the normal mode splitting in both the output probe field and the generated stokes field .
we also find that the suppression of the four - wave mixing process can be achieved on resonance .
|
kellermann & pauliny - toth ( 1969 ) first suggested that the observed upper limit on the maximum radio brightness temperatures of compact self - absorbed radio sources is an inverse compton limit .
they argued that at brightness temperature @xmath5 k energy losses of radiating electrons due to inverse compton effects become so large that these result in a rapid cooling of the system , thereby bringing the synchrotron brightness temperature quickly below this limit .
singal ( 1986 ) on the other hand derived a somewhat tighter upper limit @xmath6 k , without taking recourse to any inverse compton effects .
he used the argument that due to the diamagnetic effects the energy in the magnetic fields can not be less than a certain fraction of that in the relativistic particles and then an upper limit on brightness temperature close to the equipartition value follows naturally .
however bodo , ghisellini & trussoni ( 1992 ) pointed out that this limit on the magnetic field energy changes when the drift currents due to magnetic field gradients at the boundary are considered .
equipartition brightness temperature values have also been used to show that much higher doppler factors are needed to successfully explain the variability events ( singal & gopal - krishna 1985 ; readhead 1994 ) . here
we first examine the dependence of the brightness temperature on the magnetic field and relativistic particle energies and calculate the equipartition value @xmath2 .
we then show that how in a self - absorbed synchrotron case @xmath2 corresponds to a minimum energy configuration for the source .
we further show why brightness temperatures of a source can not rise much above this limit because of the diamagnetic effects , which keep the brightness temperatures close to the equipartition value , well below the limit where inverse compton arguments become important . unless otherwise specified we use cgs system of units throughout .
we want to examine an upper limit on the intrinsic brightness temperature achievable in a synchrotron source .
hence we will not consider here any effects of the cosmological redshift or of the relativistic beaming due to a bulk motion of the radio source , assuming that all quantities have been transformed to the rest frame of the source . in radio sources , specific intensity @xmath7 , defined as
the flux density per unit solid angle at frequency @xmath8 , usually follows a power law in the optically thin part of the spectrum , i.e. , @xmath9 , arising from a power law energy distribution of radiating electrons @xmath10 , with @xmath11 .
the source may become self - absorbed below a turnover frequency @xmath12 ( see pacholczyk 1970 ) , where the peak intensity @xmath13 is related to the magnetic field @xmath14 as , @xmath15 values of various parameters of spectral index @xmath16 , calculated from the tabulated functions in pacholczyk ( 1977 ) are given in table 1 , where we have made the plausible assumption that the direction of the magnetic field vector , with respect to the line of sight , changes randomly over regions small compared to a unit optical depth .
using rayleigh - jeans law , @xmath17 magnetic field @xmath14 can be expressed in terms of the peak brightness temperature @xmath18 , @xmath19 the magnetic field energy density @xmath20 can then be written as , @xmath21 energy density of the relativistic electrons in a synchrotron radio source component is given by ( ginzburg & syrovatskii 1965 ; ginzburg 1979 ) @xmath22 , \end{aligned}\ ] ] where @xmath23 is the characteristic depth of the component along the line of sight .
using equations ( 2 ) and ( 3 ) we get , @xmath24 . \end{aligned}\ ] ] writing @xmath25 we get , @xmath26 for typical values of @xmath27 ( table 1 ) , the denominator in the last term amounts to @xmath28 .
any variations in @xmath12 that we may consider would at most be about an order of magnitude around say , 1 ghz , ( after all it is the brightness temperature limits seen in the radio - band that we are trying to explain ) , which will hardly affect @xmath18 ( equation 7 ) . for example , for typical @xmath16 values of 0.75 to 1.0 , a factor of 10 change in @xmath12 will change @xmath18 by a factor @xmath29 .
further , with the reasonable assumption that a self - absorbed radio source size may not be much larger than @xmath30 a pc , we see that at equipartition ( @xmath31 ) the brightness temperature value is @xmath32 .
actually the @xmath18 values that have been considered hitherto ( here as well as in the past literature ) are for the turnover point in the synchrotron spectrum where the flux density peaks . however , a maxima of flux density is not necessarily a maxima for the brightness temperature as well since the definition of the brightness temperature also involves @xmath33 ( equation 2 ) .
in fact a zero slope for the flux density with respect to @xmath8 would imply for the brightness temperature a slope of @xmath34 .
therefore the peak of the brightness temperature will be at a point where flux density @xmath35 , so that @xmath36 has a zero slope with respect to @xmath8 .
the peak of @xmath37 , can be determined in the following manner ( singal 2009 ) . the specific intensity in a synchrotron self - absorbed source
is given by , @xmath38 , \end{aligned}\ ] ] where @xmath39 are tabulated in pacholczyk ( 1970 ) .
the optical depth varies with frequency as @xmath40 , @xmath41 being the frequency at which @xmath42 is unity . the equivalent brightness temperature ( in rayleigh - jeans limit )
is then given by @xmath43 .\end{aligned}\ ] ] we can maximize @xmath44 by differentiating it with @xmath8 and equating the result to zero .
this way we get an equation for the optical depth @xmath45 , corresponding to the _ peak brightness temperature _ @xmath46 , which is different from the one that is available in the literature for the optical depth @xmath47 at the _ peak of the spectrum_. the equation that we get for @xmath45 is @xmath48 solutions of this transcendental equation for various @xmath16 values are given in table 1 .
it is interesting to note that while the peak of the spectrum for the typical @xmath16 values usually lies in the optically thin part of the spectrum ( @xmath49 ; table 1 ) , peak of the brightness temperature lies deep within the optically thick region ( @xmath50 ) .
both the frequency and the intensity have to be calculated for @xmath45 to get the maximum brightness temperature values .
the correction factors are then given by , @xmath51 @xmath52\ ] ] in table 1 we have listed @xmath53 for various @xmath16 values .
the corresponding correction factors ( @xmath54 ) to the peak brightness temperature values , which need to be applied in both inverse compton and equipartition cases before making a comparison with the observed values , are given in table 1 .
also listed are the accordingly corrected @xmath2 values . from observational data , the deduced values ( readhead 1994 ; homan et al .
2006 ) of the intrinsic brightness temperature are @xmath55 @xmath56 k , quite consistent with the @xmath2 values from table 1 .
we notice that for any increase in the brightness temperature at any given turnover frequency , while the energy density of radiating particles has to go up by a factor @xmath57 ( equation 6 ) , that in the magnetic fields will have to go down by a similar factor ( equation 4 ) .
we can then derive a minimum energy density of the system .
the total energy density @xmath58 can be minimized with respect to @xmath18 to get @xmath59 as the condition for the minimum total energy density .
it should be noted that the relation between @xmath60 and @xmath61 here is somewhat different from the case of extended sources , where we get minimum energy density for an approximate equipartition condition @xmath62 ( see pacholczyk 1970 ) .
the reason being that in extended sources the intensity was treated as an independent quantity , to be determined from observations , while in a self - absorbed case the maximum intensity and therefore @xmath18 is tied to the magnetic field value through equations ( 1 and 3 ) yielding an exact equipartition condition . the equipartition brightness temperature @xmath63 thus corresponds to a minimum energy configuration of the system .
it follows from equation ( 7 ) that an order of magnitude higher @xmath18 values would require @xmath64 to increase by about a factor @xmath65 , i.e. , departure from equipartition will go up by about eight orders of magnitude . actually for a given @xmath12 , the magnetic field energy density will go down by a factor @xmath66 ( equation 4 ) , while that in the relativistic particles will go up by a similar factor ( equation 6 ) .
this implies that the total energy budget of the source will also need be higher by about @xmath66 than from the minimum energy value
. constraints of such extreme energy variations with brightness temperature imply that the latter may not depart much from @xmath2 .
this is supported by the fact that brightness temperatures much lower than @xmath4 k are also not seen at the spectral peak ( see e.g. , kellermann & pauliny - toth 1969 , fig .
4 ) , since brightness temperatures much lower than @xmath2 as well require much larger total energies .
( it is to be emphasized that the brightness temperatures being considered here are the peak values near the turnover in the synchrotron self - absorbed sources , and not the lower values which in any case occur in the optically thin regions ) .
how high could the brightness temperature rise above the equipartition value in a synchrotron source ? following homan et al .
( 2006 ) we can envisage a scenario in which the particle energy density is increased by injecting a large number of additional relativistic particles into the system , e.g. , by increasing @xmath67 , as @xmath68 for any given energy index @xmath69 .
@xmath18 remains below @xmath2 as long as @xmath70 . to increase @xmath18 further , the required change in @xmath67
can be more conveniently calculated from a proportionality expression that can be derived directly from synchrotron self - absorption ( pacholczyk 1970 ) , @xmath71 now an order of magnitude change in @xmath18 for a given magnetic field @xmath14 will require the particle energy density to increase @xmath72 times .
however any change in @xmath60 also brings a change in @xmath14 due to diamagnetic effects . gyrating charged particles
create their own magnetic field , in a direction opposite to the original field , thus giving rise to diamagnetic effects ( ginzburg & syrovatskii 1969 ; singal 1986 ) . if @xmath73 is the original magnetic field ( that is , the magnetic field value in the absence of diamagnetic effects ) then the resultant field is given by ( bodo et al .
1992 ) , @xmath74 this equation actually states that the pressure due to particles and magnetic field in the inner regions is balanced by the surrounding field pressure ( schmidt 1979 ) . at equipartition ,
energy density of radiating particles and magnetic fields is equal , @xmath75 ( say ) , which implies @xmath76
. we can then rewrite equation ( 14 ) as @xmath77 due to diamagnetic effects an increase in @xmath60 not only lowers @xmath61 , but it may also shift the turnover frequency somewhat since from equations ( 4 and 6 ) @xmath78 .
apart from equipartition , another solution of equation ( 15 ) exists for @xmath79 and @xmath80 which does not change @xmath12 .
however the resultant increase in @xmath2 from equation ( 7 ) is only a factor of @xmath81 .
the maximum @xmath60 that can be achieved is only @xmath82 .
larger @xmath60 may lead to a total screening of the field , instabilities or other circumstances like total disruption of the source as the external field pressure @xmath83 may not be able to contain the inner pressure , or the source might expand adiabatically to find a situation which is not too far from equipartition ( ginzburg & syrovatskii 1969 ; bodo et al .
1992 ) . except in such non - equilibrium conditions , where large amount of particle energy might have been injected ( e.g. , near the base of the radio jets ) and
the system has not yet relaxed to equilibrium , the results derived here should hold good .
kellermann & pauliny - toth ( 1969 ) explained the observed upper limit on the radio brightness temperatures of compact self - absorbed radio sources in terms of inverse compton losses .
they derived the ratio of inverse compton to synchrotron radiation losses as , @xmath84 , \end{aligned}\ ] ] here @xmath85 is the synchrotron photon energy density and the second term represents the effect of the second - order scattering . from this
we gather that the two rates are comparable at @xmath86 k. at higher brightness temperatures , say at @xmath5 k , energy losses of radiating electrons due to inverse compton effects would become extremely large , resulting in a rapid cooling of the system and thereby bringing the synchrotron brightness temperature quickly below this value .
however , these inverse compton losses become important _ only if _
@xmath87 k. it should be noted that even though inverse compton scattering increases the photon energy density , yet it does not increase the radio brightness as the scattered photons get boosted to much higher frequencies ( in range of infrared to x - rays ) .
if anything , some photons get removed from the radio band , but the change in radio brightness due to that itself may not be very large .
what could be important is the large energy losses by electrons which may cool the system rapidly . but can brightness temperatures ever rise to such high values for inverse compton losses to come into play at a significant level ? the equipartition conditions may keep the temperatures well below this limit , as departures from minimum energy configuration may not grow very large .
this is not to say that inverse compton effects can not occur ; it is only that conditions in synchrotron radio sources may not arise for inverse compton losses to become very effective . here
it may appear curious that two apparently different theoretical approaches lead to brightness temperature limits which are rather similar ; inverse compton value being only about three times higher than the equipartition one .
the genesis of this similarity lies in the fact that for any given turnover frequency , @xmath18 is exclusively determined by the magnetic field value ( equation 3 ) , and that in both cases , a common factor @xmath88 arises due to @xmath61 in the denominator .
but can magnetic fields get low enough for @xmath18 to rise significantly where inverse compton effects become appreciable ? in the case @xmath60 approaches @xmath89 , from equation ( 15 ) @xmath61 could become very small , thereby making @xmath64 extremely large .
however in this case the turnover frequency @xmath12 will also shift to a very low value .
an order of magnitude increase in @xmath18 from equation ( 13 ) will imply magnetic field @xmath14 falling by a factor @xmath90 ( i.e. , @xmath61 falling by @xmath91 ) and @xmath12 falling by @xmath92 .
even if such an unrealistic drop in magnetic field value to nano - gauss and turnover frequency to khz range were achievable , for one thing this will take @xmath12 outside the radio - band of our interest where the brightness temperature limits have been seen and which we are trying to explain here . but even more important the consequential increase in brightness temperature will still be not contained by inverse compton effects as even the inverse compton limit will go up about an order of magnitude for such a large fall in the turnover frequency ( equation 16 ) .
hence , it is imperative that even in such a scenario inverse compton effects do not get to play any significant role in maintaining the maximum brightness temperature limit in an incoherent synchrotron radio source .
what about the effects of inhomogeneities in the source on the brightness temperature limits ? as mentioned in section 2 , we have assumed that the direction of the magnetic field vector , with respect to the line of sight , changes randomly over the source .
this makes @xmath93 in equation ( 6 ) change by a factor @xmath94 ( ginzburg 1979 ) , while @xmath95 in equation ( 3 ) changes by a factor @xmath96 ( @xmath97 from tables of pacholczyk 1970 , 1977 ) . from equations ( 4 , 6 and 7 ) we see that @xmath98^{1/8}$ ]
thus our derived @xmath18 values in the case of a random magnetic field orientation are higher than those in the homogeneous case by a factor of @xmath99 , a negligible quantity .
another type of inhomogeneity one could consider is when the source may consist of a number of discrete components , each with its own cutoff frequency ; such a scenario is suggested by the flat shape of the observed spectra as well as the vlbi observations ( kellermann & pauliny - toth 1981 ) .
however in such a case the observed brightness temperature value , a sort of average over the various components , can not exceed the highest value among the individual components , which will have an equipartition brightness temperature limit as derived above .
we can thus conclude that under a variety of conditions the diamagnetic effects will limit the brightness temperatures close to the equipartition value , well below the limit where inverse compton effects become important .
lclccclcllccc @xmath100 & @xmath69 & @xmath101 & @xmath102 & @xmath103 & @xmath104 & @xmath105 & @xmath106 & @xmath107 & @xmath108 & @xmath53 & @xmath54 & log @xmath2 + 0.25 & 1.5 & 0.149 & 0.61 & 1.10 & 1.3 & 0.011 & 0.53 & 0.19 & 2.80 & 0.37 & 3.5 & 11.3 + 0.5 & 2.0 & 0.103 & 0.85 & 1.19 & 1.8 & 0.032 & 0.67 & 0.35 & 2.92 & 0.50 & 2.2 & 11.2 + 0.75 & 2.5 & 0.0831 & 0.84 & 1.27 & 2.2 & 0.10 & 0.63 & 0.50 & 3.03 & 0.58 & 1.8 & 11.1 + 1.0 & 3.0 & 0.0741 & 0.74 & 1.35 & 2.7 & 0.18 & 0.53 & 0.64 & 3.13 & 0.64 & 1.6 & 10.9 + 1.5 & 4.0 & 0.0726 & 0.52 & 1.50 & 3.4 & 0.38 & 0.34 & 0.88 & 3.32 & 0.72 & 1.4 & 10.7 + \99 .
bodo , g. , ghisellini , g. , & trussoni , e. 1992 , mnras , 255 , 694 ginzburg , v. l. 1979 , theoretical physics and astrophysics ( oxford : pergamon ) ginzburg , v. l. , & syrovatskii , s. i. 1965 , ara&a , 3 , 297 ginzburg , v. l. , & syrovatskii , s. i. 1969 , ara&a , 7 , 375 homan , d. c. et al 2006 , apj , 642 , l115 kellermann , k. i. , & pauliny - toth , i. i. k. 1969 , apj , 155 , l71 kellermann , k. i. , & pauliny - toth , i. i. k. 1981 , ara&a , 19 , 373 pacholczyk , a. g. 1970 , radio astrophysics ( san francisco : freeman ) pacholczyk , a. g. 1977 , radio galaxies ( oxford : pergamon ) readhead , a. c. s. 1994 , apj , 426 , 51 schmidt , g. 1979 , physics of high temperature plasma ( new york : academic ) singal , a. k. 1986 , a&a , 155 , 242 singal , a. k. 2009 , in ap&ss proc . , turbulence , dynamos , accretion disks , pulsars and collective plasma processes , ed .
s. s. hasan , r. t. gangadhara , & v. krishan ( berlin : springer ) , 273 singal , a. k. , & gopal - krishna 1985 , mnras , 215 , 383
|
we show that an upper limit of @xmath0 k on the peak brightness temperature for an incoherent synchrotron radio source , commonly referred to in the literature as an inverse compton limit , may not really be due to inverse compton effects .
we show that a somewhat tighter limit @xmath1 is actually obtained for the condition of equipartition of energy between radiating particles and magnetic fields which happens to be a configuration of minimum energy for a self - absorbed synchrotron radio source .
an order of magnitude change in brightness temperature from @xmath2 in either direction would require departures from equipartition of about eight orders of magnitude , implying a change in total energy of the system up to @xmath3 times the equipartition value
. constraints of such extreme energy variations imply that brightness temperatures may not depart much from @xmath2 .
this is supported by the fact that at the spectral turnover , brightness temperatures much lower than @xmath4 k are also not seen in vlbi observations .
higher brightness temperatures in particular , would require in the source not only many orders of magnitude higher additional energy for the relativistic particles but also many order of magnitude weaker magnetic fields .
diamagnetic effects do not allow such extreme conditions , keeping the brightness temperatures close to the equipartition value , which is well below the limit where inverse compton effects become important .
|
recent evidence for the existence of intermediate mass black holes ( imbhs ) raises questions about how such objects might form and evolve .
imbh candidates exist in globular clusters @xcite , nearby bulgeless galaxies @xcite , and active galactic nuclei @xcite .
additionally , off - nuclear ultraluminous x - ray sources ( ulxs ) have become increasingly promising imbh candidates @xcite . a source of imbhs
may be the seeds of supermassive black holes ( smbhs ) formed at high redshift ; any seed that does not grow into a smbh would today be observed as an imbh . while the precise mechanism for seed black hole ( bh ) formation is unknown ,
there are several postulated theories .
one possible mechanism for smbh seed formation is the direct collapse of pristine , low angular momentum gas , which forms bhs with mass on the order of @xmath0 m@xmath1 @xcite .
another possibility is that the seeds are the remnants of population iii stars , with masses around @xmath2 m@xmath1 @xcite .
alternatively , the first nuclear star clusters may collapse to form bhs of mass @xmath3 m@xmath1 @xcite .
we perform sph+@xmath4-body simulations of massive disk galaxies to explore the evolution of seed bhs in a cosmological context .
while previous works have addressed the growth of central smbhs @xcite , here we specifically focus on those which do not end up in galaxy centers .
we include bh growth by gas accretion and merging , as well as radiative feedback in order to form a fully self - consistent picture of bh growth and evolution .
we find that the tidal stripping of galaxies containing smbh seeds leads to a population of `` wandering '' bhs within the larger galaxy halo .
such a phenomenon has been predicted in the case of idealized simulations of merging galaxies @xcite , though fully cosmological simulations in the current @xmath5cdm paradigm are needed to test a suite of non - idealized scenarios . wandering bhs
may also be created by a gravitational slingshot from 3-body black hole interactions @xcite , gravitational recoil , or a distribution of population iii star remnants which have not spiralled into the central bh @xcite . in each of these cases ,
the timescale for dynamical friction is longer than a hubble time @xcite , leaving a potentially substantial population of `` wandering '' imbhs in galaxy halos .
the recent discovery of an off - nuclear imbh candidate @xcite leads us to explore whether such an object can be explained by a `` wandering '' bh .
the object hlx-1 is offset from the nucleus of its host spiral galaxy , and exhibits an x - ray luminosity of @xmath6 ergs s@xmath7 .
this luminosity implies a lower limit to a black hole ( bh ) mass of 500 m@xmath1 , but the true mass may be much larger .
we discuss under what circumstances hlx-1 would be observable .
preliminary results suggest that hlx-1 can not be described by an isolated wandering bh , but a bh traveling with the remnant core of its parent galaxy may explain its observed properties .
we use the n body code gasoline @xcite , an sph tree code which incorporates star formation , gas cooling and hydrodynamics , and supernova feedback , and successfully models realistic galaxies .
studies of cosmological simulations with gasoline have produced galaxies that follow the mass - metallicity relation @xcite and the hi tully - fisher relation @xcite , exhibit cold flow gas accretion @xcite , and reproduce the distribution of damped lyman alpha systems at high redshift @xcite .
additionally , @xcite have shown how high resolution and physically motivated supernova feedback allow for the formation of a bulgeless dwarf galaxy with a dark matter core .
the ability to form realistic disk galaxies is critical to our analysis of bh physics , because of the need to accurately trace the angular momentum of inspiraling gas which comprises the disk and eventually fuels the bh , as well as the star formation histories responsible for dispensing metals into the ism via supernova feedback . for this letter
, we have simulated four different milky way - mass halos , which were selected from a uniform , 50 mpc volume and resimulated at a high resolution using the volume renormalization technique @xcite .
all simulations are run with gravitational softening @xmath8 kpc , gas particle masses of @xmath9 m@xmath1 , dark matter particle masses of @xmath10 m@xmath1 , a chabrier initial mass function @xcite , and a _ wmap _ year 3 cosmology @xcite .
star formation and supernova recipes are described in detail in @xcite and @xcite ; we adopt parameter values @xmath11 and @xmath12 .
stars are allowed to form when gas reaches a threshold density of 1.0 amu @xmath13 , which allows us the most physical representation of star formation given our resolution . while our dwarf galaxy satellites are not resolved to the level of @xcite , for our purposes here
this resolution is sufficient .
simulation properties are described in table 1 .
circular velocity @xmath14 is measured using the width at 20% of the peak of the simulated hi line profile ( see * ? ? ?
the disk scale length @xmath15 is determined by simultaneously fitting an exponential + sersic profile to the projected edge - on stellar surface density of the galactic disk .
lcccll h239 & 7408639 & @xmath16 & 247.7 & 242.9 & 2.82 + h258 & 7347383 & @xmath17 & 225.4 & 238.8 & 3.81 + h277 & 6624914 & @xmath18 & 262.7 & 230.3 & 2.74 + h285 & 7373714 & @xmath19 & 238.2 & 241.5 & 2.24 + there is much uncertainty regarding the formation of the initial `` seed '' black holes which grow to become smbhs ; however , they must form early on and grow quickly in order to form @xmath20 quasars @xcite .
we maintain a physically motivated scenario by assuming seed black holes form via the direct collapse of extremely low metallicity gas as in @xcite . however , our choice of seed formation scenario is not crucial .
we focus here on the dynamical evolution of seed bhs , which will be the same regardless of where we set the initial mass ( i.e. 100 m@xmath1 for the first stars or @xmath21 m@xmath1 for direct collapse ) .
the dynamical friction timescale is longer than the hubble time even for these relatively heavy seeds , given the local densities involved here .
seeds are allowed to form if the parent gas particle meets the criteria for star formation ( see * ? ? ?
* ) and additionally has zero metallicity .
we designate a probability that a newly formed star will instead become a seed black hole with mass @xmath22 .
this probablility is set to 0.10 in order to reproduce the predicted black hole seed halo occupation probability at @xmath23 @xcite .
allowing seed bhs to form out of only pristine gas results in the truncation of seed bh formation at a redshift of @xmath24 due to the efficient diffusion of metals produced by the first supernovae into the ism .
this criterion causes seed bhs to form in the regions of early bursts of star formation , which are primarily the centers of the first massive halos to form in the simulation . in the event that more than one bh forms at the same time within a softening length , the bhs are merged into a single bh particle , with @xmath25 as the total of the individual masses
however , we make no assumptions regarding the large - scale properties of the host halo , allowing the bhs to form based only on the properties of the local environment .
bhs are not fixed at their halo centers , but are allowed to move as the forces upon them dictate .
we minimize the effects of two - body interactions by adopting dark matter particle masses that are similar to gas particle masses in the high resolution region , which greatly helps the central bhs stay at their halo centers .
black holes are presumed to accrete gas isotropically , following the formula for bondi hoyle accretion .
the accretion rate is limited by the eddington rate , assuming a 10% radiative efficiency .
in addition , a fraction of the rest mass energy of the accreted gas is converted to radiative energy , which is then isotropically imparted onto the surrounding gas @xcite . to dissipate the feedback energy in a realistic way
, we use a blast - wave feedback approach similar to the supernova feedback recipe described in @xcite .
we set the fraction of radiated energy given to the surrounding gas to be 0.1% .
black holes are allowed to merge if they _ ( a ) _ are within one another s softening length and _ ( b ) _ fulfill the criterion @xmath26 , where @xmath27 and @xmath28 are the differences in velocity and acceleration of the two bhs , and @xmath29 is the distance between them .
we identify galaxies and their halos with ahf @xcite , which identifies a virial radius based on the overdensity criterion for a flat universe @xcite . for each output
, we identify every halo and all of the particles it contains ; thus , we can track a bh s halo residency at each timestep .
we use this process to define whether a black hole is stripped from its parent halo to reside in another .
a bh is denoted as `` stripped '' if its parent halo overdensity can no longer be detected as a separate halo .
the bh is then identified as a resident of the larger galaxy , most likely in the outer regions .
we provide a dynamical mechanism to place massive bhs in galaxy halos , which is simply a consequence of hierarchical merging in the @xmath5cdm paradigm . detecting these objects may help constrain the mechanism of seed bh formation , and we investigate a few scenarios in which a wandering bh may be observed as a ulx . for our four simulated halos ,
we show the radial distribution of bhs at @xmath30 , defined as the distance between the halo center and the bh , in figure [ fig : radial ] .
each of the four galaxies has a central bh and between 5 and 15 `` wandering '' bhs whose distances range from 10 - 100 kpc .
the vast majority of wandering bhs have grown in mass by less than 2% since their formation . a small number ( 5 out of 36 total )
have undergone bh
bh mergers early in their lifetimes and thus have larger masses , but only one of these has experienced any substantial accretion events . in all cases accretion
is effectively quenched when the host galaxies are torn apart and the bhs are left to wander throughout the halo , and thus any bhs present here are at or within a few factors of their original seed mass .
local group dwarf galaxies are promising targets for imbh searches @xcite . in figure
[ fig : maxmass ] we show the masses of bh - hosting satellite galaxies before the stripping process has begun .
this distribution peaks between @xmath31 m@xmath1 ; at halo masses of @xmath32 m@xmath1 the satellites become increasingly dark .
since very few stars form in these halos , it is unlikely that we are underestimating the number hosting massive bh seeds , given our scenario .
we estimate the absolute magnitudes of the satellites in these mass ranges using the starburst99 @xcite code , using stellar ages and metallicities from our simulations .
this mass range corresponds to galaxies with magnitudes fainter than -15 in the @xmath33 band , which includes several local group dwarfs , including the large and small magellanic clouds . examining these local dwarfs for
imbhs may help constrain the locations and masses of smbh seeds .
a confirmed detection of an imbh would provide an upper limit to the initial mass of smbh seeds and possibly allow us to differentiate between the various proposed formation mechanisms of such seeds .
[ htb ] while wandering bhs may be present in the milky way halo , observing such objects would require either a fortuitous gravitational lensing detection , or a triggered accretion event , resulting in x - ray emission .
we investigate situations where a wandering bh could undergo substantial accretion in a galactic halo , which could explain the existence of objects such as hlx-1 .
while the majority of ulxs are attributed to star formation regions @xcite , a smaller number of off - planar ulxs are not cospatial with star formation , and we focus on the possible origins of these objects here .
we first examine whether a wandering bh passing through a dense portion of the host galaxy disk could reproduce the properties of off - nuclear ulxs .
we trace the orbits of each bh and examine the frequency with which they cross the disk region , which we define as three disk scale lengths ( @xmath15 ) .
the last few gyr of orbital history for the bhs in our simulated galaxy h239 are shown in figure [ fig : orbit ] .
several of the bhs pass through the @xmath34 limit , which we denote as a disk passage .
the average bh disk passage rate for all of the simulations combined is 10.6 per gyr ; however , this rate increases with time ( 13.3 per gyr for the last few billion years of galaxy evolution ) due to the increased population of wandering bhs in the halo ( since more mergers have occurred ) .
the likelihood of observing such a disk passage depends on the column density of gas in the disk , the number of passages made , and the velocity of the bhs .
we use results from hi measurements of the surface densities of local disk galaxies @xcite to estimate an average disk column density of 10 m@xmath1 pc@xmath35 .
the ability of the passing bh to attract a sufficient amount of gas to be observed as a ulx depends most strongly on its velocity , which ranges between 154 - 662 km s@xmath7 , and on average is @xmath36 470 km s@xmath7 for bhs at pericenter . to estimate a bh s luminosity
, we developed a simple scenario where the passing bh is able to collect the gas within a radius of influence determined by its mass and pericenter velocity : @xmath37 .
we perform a monte carlo simulation taking into account the observed ranges of bh velocities , and estimate that the amount of mass collected by a bh during its disk passage ranges from @xmath38 m@xmath1 . assuming an accretion luminosity @xmath39 with @xmath40 and the canonical luminosity of @xmath41 ergs s@xmath7 , the bh will accrete and radiate in a high state for a mean duration of 2000 years , though a slower bh may radiate for up to @xmath21 years
. however , if the ulx transitions from a high state to a low state at any point , the accretion rate would drastically decrease and the observed emission could continue for much longer . this simple scenario , however , is not sufficient to reproduce the properties of hlx-1 , which is at least 1 kpc away from the plane of its host galaxy .
a bh passing through the disk moving at a few hundred km s@xmath7 will only travel a distance of @xmath36 tens of parsecs during the duration of the predicted accretion event .
such an event would appear to be a disk ulx source , possibly indistinguishable from those cospatial with star formation regions .
hlx-1 is also 1000 times more luminous and 100 times less massive ( wiersema et al .
, in prep . )
than the assumed values in our calculation above . taking all of these factors into account ,
it is unlikely that an isolated wandering bh passing through a galaxy disk can reproduce the properties of off - nuclear ulxs .
however , our wandering bh scenario would be feasible if hlx-1 is not an isolated bh , but is traveling with a bound clump of gas and stars .
the magnitude and colors of the detected optical counterpart of hlx-1 are consistent with a globular cluster @xcite .
this cluster may in fact be the stellar remnants of a stripped dwarf galaxy core that is still bound to the bh .
hlx-1 could be similar to the object g1 in m31 , a globular cluster that may actually be the nucleus of a stripped dwarf galaxy @xcite and harbors the most promising local group candidate for an imbh @xcite .
if we presume that the wandering bhs in our simulations retain a nuclear star cluster and gas reservoir from their parent halos , then the instance of a wandering bh passing near the center of the primary could cause instabilities in its accretion disk , triggering an accretion event of sufficient magnitude to power a ulx with a high luminosity as seen in hlx-1 .
our simulations do not have sufficient resolution to follow the tidally stripped cores of galaxies in detail , though other simulations have shown that a tidally stripped dwarf galaxy can retain its core after a close passage with the primary @xcite .
thus , in the likely instance that a wandering bh retains the core of its host galaxy , its passage near the galaxy disk can explain the origin and properties of hlx-1 .
previous studies have estimated luminosities for massive bhs wandering through the ism , but prior to this letter none have explored the issue in a cosmological context .
@xcite showed that imbhs with masses ranging from @xmath42 m@xmath1 can produce luminosities of ulxs if they pass through or near molecular clouds .
smbh gravitational recoil events passing through the disk may exhibit x - ray emission of @xmath43 ergs s@xmath7 @xcite .
@xcite performed an n - body+sph simulation of imbhs in a galaxy merger , and found that a few halo imbhs reside in orbits that pass through the disk , which in our scenario may be observable as ulxs .
we provide compelling dynamical scenario for the presence of `` wandering '' massive bhs in the halos of galaxies . a natural consequence of the hierarchical build up of galaxies in a @xmath5cdm scenario , the tidal stripping of galaxies containing seed bhs can populate the halo of a massive disk galaxy with wandering bhs .
these objects often retain their original seed mass , are found throughout the galaxy halo , and may pass through the galactic disk at an average rate of 10.6 gyr@xmath7 .
we predict that local group dwarf galaxies such as the magellanic clouds are likely to host imbhs .
detections of these wandering bhs may give an upper limit to the initial mass of bh seeds , and may allow us to differentiate between the various proposed formation mechanisms of such seeds .
our scenario provides a physically motivated explanation for off - nuclear ulxs as imbhs which have been stripped from their host galaxies , if they retain a gas reservoir / accretion disk that , when dynamically destabilized , might be funneled toward the black hole and form an accretion disk .
jmb gratefully acknowledges the national science foundation graduate research fellowship program .
fg acknowledges support from hst go-1125 , nasa itp nnx08ag84 g , and nsf itr grant phy-0205413 ( also supporting trq ) .
mv acknowledges support from sao tm9 - 0006x and nasa atp nnx10ac84 g awards .
simulations were run using computer resources and technical support from nas .
the authors thank charlotte christensen , adrienne stilp , amy kimball and brant robertson for assistance and helpful comments , as well as the referee whose comments greatly improved the paper .
|
we perform sph+n - body cosmological simulations of massive disk galaxies , including a formalism for black hole seed formation and growth , and find that satellite galaxies containing supermassive black hole seeds are often stripped as they merge with the primary galaxy .
these events naturally create a population of `` wandering '' black holes that are the remnants of stripped satellite cores ; galaxies like the milky way may host 5 15 of these objects within their halos .
the satellites that harbor black hole seeds are comparable to local group dwarf galaxies such as the small and large magellanic clouds ; these galaxies are promising candidates to host nearby intermediate mass black holes . provided that these wandering black holes retain a gaseous accretion disk from their host dwarf galaxy , they give a physical explanation for the origin and observed properties of some recently discovered off - nuclear ultraluminous x - ray sources such as hlx-1 .
|
the existence of light hypernuclei , such as @xmath12he , shows the low energy @xmath13-@xmath14 interaction to be strongly attractive , though not sufficient to bind the two - baryon hyperdeuteron .
the @xmath13-@xmath14 interaction is of especial interest since it is influenced by the strange quark content of the @xmath13-hyperon .
however , in contrast to the nucleon - nucleon case , due to the short lifetime of the @xmath13 , direct measurements of low - energy @xmath13-@xmath14 scattering are sparse and the resulting parameters rather poorly known .
bubble chamber measurements @xcite , based on samples of a few hundred secondary events , have allowed determinations of the elastic cross section down to @xmath13 laboratory momenta @xmath15 mev / c . in the low energy region , where only @xmath16-waves are important , the spin - averaged total cross section is of the form @xmath17 here @xmath18 is the @xmath0 centre - of - mass momentum and @xmath19 and @xmath20 are , respectively ,
the @xmath16-wave scattering lengths and effective ranges in the @xmath0 spin - singlet and triplet states . separate values of these parameters have been claimed for the two spin states @xcite and these are shown in fig .
[ ar_other ] . however , the error bars are large , strongly and systematically correlated and hard to quantify , since such data should really only support the determination of an average scattering length @xmath21 and effective range @xmath22 @xcite . already for laboratory momenta @xmath23 mev / c ,
the differential cross section is significantly non - isotropic , indicating the presence of @xmath24 or higher waves @xcite , and so it is not surprising that the @xmath16-wave parameters deduced from such experiments depend upon the upper momentum cut assumed .
values of the scattering length and effective range have also been deduced through the study of the @xmath0 final state interaction ( fsi ) in the @xmath25 reaction with stopped @xmath26-mesons @xcite .
here it is the of the @xmath0 effective mass spectrum near the kinematic limit which is sensitive to the parameters . in impulse approximation
the amplitude for this process is proportional to that for @xmath27 and , if the fermi motion in the target deuteron is neglected , the reaction is purely @xmath28-wave with no spin - flip .
the final @xmath0 system is therefore in the same spin - triplet state as the @xmath29 pair in the deuteron and the values @xmath30 and @xmath31 so determined , which are also shown in fig .
[ ar_other ] , are consistent with those obtained from the scattering experiments @xcite .
phenomenological investigations of the @xmath0 interaction by the jlich @xcite and nijmegen @xcite groups yield low energy scattering parameters in agreement with the results of fig .
[ ar_other ] , though it is impossible to quantify the systematic uncertainties inherent in such models .
it should , however , be noted that their spin - singlet potential is more attractive than the triplet , which is necessary to ensure the correct spin assignments of the ground states of @xmath12h and @xmath32he @xcite .
the final state interaction in the @xmath33 reaction at low @xmath0 effective masses is also sensitive to the @xmath0 scattering parameters @xcite .
the recent data on this reaction very close to threshold @xcite , taken at the cosy-11 magnetic spectrometer @xcite installed at the cooler synchrotron cosy - jlich @xcite , allow us to extract information complementary to that obtained from elastic scattering because it is possible to reach lower centre - of - mass momenta . in the present paper we aim to constrain the @xmath0 scattering parameters from the of the double - differential @xmath33 cross section . in section 2 a simplified model is outlined to describe the principal interactions between the three outgoing particles .
experimental details , including event selection , are discussed in section 3 .
the fsi model was already used in ref .
@xcite to determine the precise beam energy by fitting the integrated total cross sections as a function of beam energy with assumed values of the @xmath0 input parameters . to avoid biassing the analysis , in the present work we determine the @xmath0 parameters using only the structure of the dalitz plot at each energy and not its normalisation .
for this purpose we apply the maximum likelihood method to obtain a map of the confidence levels for the @xmath0 scattering parameters .
this formalism and the definition of the event weights are shown in section 4 . even taking the @xmath0 spin - triplet and singlet to be identical ,
the resulting values of the average scattering length and effective range presented in section 5 are strongly and systematically correlated in the fit , such that it is hard to quote error bars .
however , the position of the nearby pole in the @xmath0 scattering amplitude , corresponding to a virtual bound state of the system , is much more stable , being unbound by @xmath34 mev .
our conclusions are presented in section 6 .
as discussed in ref .
@xcite , if the basic production mechanism is of short range then the energy dependence of the @xmath1 cross section close to threshold is dominated by the available three - body phase - space @xmath35 , modified by final - state interactions . in principle
one should consider fsi s in all the three two - body subsystems , , , and . since the strong interaction in the first case appears to be more than an order of magnitude larger than for the other two @xcite , we concentrate on the dominant factor @xmath36 in the @xmath13-@xmath14 system . in addition , however , the coulomb repulsion in the proton - kaon subsystem @xmath37 , where @xmath38 denotes the c.m .
momentum in @xmath39 subsystem , is also important at the low energies pertaining in our experiment .
keeping only these two interactions our _ ansatz _ for the production cross section is @xmath40 the @xmath16-wave assumption implicit here is justified for our experiment since , for excess energies @xmath41 mev , the maximum momentum in any two - body subsystem is below 100 mev / c .
it should be noted that the bubble chamber data @xcite cover a higher range of @xmath13-@xmath14 momenta .
choosing as independent variables @xmath42 and @xmath43 , the squares of the effective masses in the @xmath14-@xmath26 and @xmath13-@xmath26 subsystems , integration of eq .
( [ eq_iffr ] ) over the angular variables leads to a number of events per pixel in the dalitz plot distribution of the form @xmath44 structure in the dalitz plot must be associated with the functions @xmath45 and @xmath46 . now
the coulomb distortion factor defined in ref .
@xcite contains no free parameters @xcite .
in contrast the @xmath45 factor depends on the scattering lengths and effective ranges in the triplet and singlet states . since the spin dependence is expected to be small @xcite , as shown by some of the extracted numbers in fig .
[ ar_other ] , and our experiment is not sensitive to singlet / triplet differences , we used mean values of the scattering length and effective range in the parametrization of @xmath45 . in analogy with eq .
( 1 ) , we take the popular watson form for the final state interaction @xcite @xmath47 a typical dalitz plot distribution calculated from eqs .
( [ eq_dffr ] ) and ( [ eq_fsi ] ) with @xmath48 fm and @xmath49 fm is shown in fig .
[ dal_mod_q5_new ] .
the measurement of the @xmath1 reaction was performed at the cosy - jlich synchrotron , using the internal target facility cosy-11 @xcite .
outgoing protons and positively charged kaons were identified by means of particle momentum reconstruction in the magnetic field combined with time - of - flight measurement .
the four - momentum , and hence the missing mass ( mm ) , corresponding to the unobserved @xmath13-hyperon was calculated from energy - momentum conservation .
details of the experimental technique are given elsewhere @xcite .
an example of the missing mass spectrum is shown in fig .
[ mm_q5 ] .
the function @xmath50 used to fit the peak corresponding to good @xmath1 events , is combined with the smooth background @xmath51 .
only events which deviate by less than @xmath52 from the central value were accepted and a weight @xmath53}$ ] was assigned to each of them to describe the probability that the particular event resulted from a @xmath54 final state rather than being a background signal .
our data were well described by a gaussian form for @xmath50 with @xmath55 mev .
the closer @xmath56 is to the known @xmath13 mass @xmath57 , the larger the weight @xmath58 for the event .
thus @xmath58 can be interpreted as a penalty factor which is imposed on events where the observables are significantly modified by the application of the kinematic fit procedure . the experimental dalitz plot for @xmath59 mev , shown in fig .
[ dal_exp_q5]a , demonstrates that the whole kinematically allowed region is occupied by data , so that there are no forbidden zones in our acceptance .
this experimental acceptance , calculated _ via _ a monte - carlo simulation with a pure phase - space generator , is however non - uniform with a _
colosseum_-like pattern , as illustrated in fig .
[ dal_exp_q5]b .
comparison of the experimental dalitz plot ( fig .
[ dal_exp_q5]a ) with the model calculation ( fig . [ dal_mod_q5_new ] ) , folded with the detector acceptance ( fig
. [ dal_exp_q5]b ) , allows a determination of the model parameters @xmath21 and @xmath22 at each excess energy @xmath60 .
although they influence the statistical confidence , the relative counting rates at different @xmath60 are not used , which is important since the variation of cross section with energy has already been employed in fixing the absolute beam energy @xcite .
the maximum likelihood method was applied to determine best values of the @xmath21 and @xmath22 parameters , though it must be stressed that these will be strongly correlated due to the form of the @xmath45 factor of eq .
a set of model dalitz plots @xmath61 , as defined in eq .
( [ eq_dffr ] ) and ( [ eq_fsi ] ) , was generated at fixed excess energy @xmath60 over a grid in @xmath62 with @xmath63 $ ] fm and @xmath64 $ ] fm and a step - size of @xmath65 fm in each variable . defining for brevity @xmath66 and @xmath67 , the probability @xmath68 that an event will be detected with some @xmath69 value is @xmath70 where @xmath71 is the acceptance function .
applying the experimental event weight @xmath72 , the likelihood of a single event is @xmath73 ^{w_i } ~\equiv~\mbox{$\cal{p}$ } _ i^{w_i}\end{aligned}\ ] ] and the corresponding global likelihood function @xmath74 since @xmath72 is the probability that the @xmath75-th event is a @xmath54 and not a background reaction , if the whole experiment were repeated @xmath76 times with perfect background subtraction , then each @xmath75-th event would appear @xmath77 times .
the likelihood corresponding to each @xmath78 point is given by @xmath79 and the global likelihood function is : @xmath80 .
both likelihood functions @xmath81 and @xmath82 have their extrema at the same @xmath83 points and thus it is sufficient to perform the experiment once , searching for the maximum of the @xmath81 function . in order to amalgamate the likelihood functions from measurements at different values of @xmath60 ,
the @xmath76 was chosen in such a way that @xmath84 was equal to the number of measured events @xmath85 .
as shown by eadie @xcite , the quantity @xmath86 $ ] has an asymptotic @xmath87 distribution corresponding to the two degrees of freedom , _
i.e. _ two parameters @xmath62 .
the equation @xmath88 defines two - dimensional contours in the @xmath62 plane for the desired confidence level cl @xmath89 .
contours in the @xmath62 plane of the global likelihood function calculated from a sample of about 2400 events measured at six excess energies from 2.7 to 6.7 mev are shown in fig .
[ cl_all_q ] . those solutions with a positive value of @xmath21 are to be excluded since they would imply the existence of a bound @xmath0 system and the hyperdeuteron has never been found . the branch with negative @xmath21 yields a long narrow ridge with a very strong correlation between the scattering length and effective range , such that it is only a combination of the two parameters which is well determined by the experiment .
the averaged values from the literature @xcite@xcite lie close to the ridge but generally slightly outside the 99%-confidence contour .
the parameter which is in fact well determined by this experiment is the energy of the nearby pole in the @xmath0 scattering amplitude . to see this , rewrite the final - state - interaction factor of eq .
( 4 ) in the form @xmath90 + the @xmath0 scattering amplitude has poles at @xmath91 , where @xmath92\ \
\mbox{\rm and}\ \ \gamma_2=\frac{1}{\bar{r}}\left[1 + \sqrt{1-\frac{2\bar{r}}{\bar{a}}}~\right ] \:\cdot \label{eq_xx2}\end{aligned}\ ] ] + it is straightforward from eq .
( [ eq_xx2 ] ) to transform the likelihood function into the new variables and the contours in the resulting @xmath93 are shown in fig .
[ cl_gg ] together with the literature values @xcite@xcite .
note that the functional form of eq .
( [ eq_xx1 ] ) clearly shows that the data are sensitive only to the magnitudes of the @xmath94 and that the results must be symmetric under the interchange @xmath95 .
we therefore establish the convention that @xmath96 .
if @xmath97 is positive then that would correspond to a bound state of the @xmath0 system , whereas if it is negative then it is an antibound or virtual state of the kind with which one is familiar from the low energy proton - proton singlet @xmath16-wave .
the two branches seen in the maximum likelihood contours of fig . [ cl_all_q ] correspond to the mere reversal of the sign of @xmath97 .
since it is highly unlikely that there would be another singularity of the @xmath13-@xmath14 amplitude very close to zero energy , we can assume that @xmath98 , in which case we deduce from our fit that @xmath99 @xmath100 .
it has been argued @xcite that the single pole limit of letting @xmath101 in fact provides a better representation of @xmath45 than the scattering length effective range form of eq .
( [ eq_fsi ] ) .
our value of @xmath97 corresponds to the average @xmath13-@xmath14 system being unbound by an amount @xmath34 mev . in order to obtain values of @xmath21 and @xmath22 separately
, we must use extra experimental information such as for example @xmath13-@xmath14 elastic scattering cross section data .
it is seen from eq .
( 1 ) that the normalisation of this at zero energy is proportional to the square of the scattering length . the low energy data of ref .
@xcite are shown in fig .
[ elas_lp ] together with a fit to the data on the basis of eq .
( 1 ) , where the parameters @xmath21 and @xmath22 are constrained to lie on the maximum likelihood ridge of fig .
[ cl_gg ] .
this is achieved with @xmath102 fm and @xmath103 fm , corresponding to @xmath104 @xmath100 and @xmath105 @xmath100 .
the fit therefore confirms that @xmath106 and that the @xmath0 scattering data do not realistically allow for a separation between singlet and triplet parameters .
it is worth noting that our production data are sensitive to much lower values of @xmath60 , as indicated by the arrow , than the scattering data , and it is this region which determines best the value of @xmath97 and hence the position of the pole of the virtual bound state .
through an analysis of the two - dimensional structure of the dalitz plot for the @xmath107 reaction at fixed energies within a simple final - state - interaction model , we have established a strong constraint between the spin - averaged @xmath0 scattering length and effective range .
the data allow us to fit accurately the position of the spin - average virtual bound state @xmath97 . since the data were taken at excess energies which are inaccessible to low energy elastic scattering experiments , the results are complementary and it is appropriate to make a combined fit of the whole data set , leading to new values of @xmath62 . the total cross section data of ref .
@xcite have recently been analysed to determine values of @xmath62 @xcite .
their argument is , however , somewhat cyclic since a final - state interaction with fixed scattering length and effective range was already used in the experimental analysis to fix the beam energy @xcite . in principle
therefore these values should then be found by the fitting procedure , though there are still of course the ambiguities discussed in this paper .
we avoid falling into this trap by not using the relative normalisation of the event rate as a function of the beam energy within our fitting procedure .
it is rather the structure of the two - dimensional dalitz plots which fixes our parameter values .
though we have implicitly assumed that the @xmath0 system produced from the near - threshold @xmath108 reaction is the same 3:1 spin - average seen in bubble chamber scattering experiments , this is not guaranteed .
just as in the @xmath109-capture experiment @xcite , the basic reaction mechanism could favour the production of a particular spin combination in the final state .
if the present experiment were extended through the use of a polarised beam and target , then it would be possible to repeat the current analysis separately in the singlet and triplet final states to separate these important quantities .
+ we should like to thank dr a. szczurek for very helpful discussions .
the research project was supported by the bmbf , the polish committee for scientific research , the bilateral cooperation between germany and poland , represented by the internationales bro dlr for the bmbf , and the ffe program of the forschungszentrum jlich .
one of the authors ( cw ) is grateful to the fz - jlich for the consultancy which supported some of this work .
|
constraints on the spin - averaged @xmath0 scattering length and effective range have been obtained from measurements of the @xmath1 reaction close to the production threshold by comparing model phase - space dalitz plot occupations with experimental ones .
the data fix well the position of the virtual bound state in the @xmath0 system . combining this with information from elastic @xmath0 scattering measurements at slightly higher energies , together with the fact that the hyperdeuteron is not bound , leads to a new determination of the low energy @xmath0 scattering parameters .
+ 0.86 cm 0.86 cm 1.8 cm 0.4 cm 0.42 cm 0.88 cm 0.42 cm 22.5 cm 15.05 cm 0.35 cm 4ex * low - energy * + * the * reaction * * + j.t .
balewski@xmath2 , a. budzanowski@xmath3 , c. goodman@xmath4 , d. grzonka@xmath5 , m. hofmann@xmath5 , l. jarczyk@xmath6 , a. khoukaz@xmath7 , k. kilian@xmath5 , t. lister@xmath7 , p. moskal@xmath8 , w. oelert@xmath5 , i.a .
pellmann@xmath9 , c. quentmeier@xmath7 , r. santo@xmath7 , g. schepers@xmath7 , t. sefzick@xmath5 , s. sewerin@xmath5 , j. smyrski@xmath6 , a. strzakowski@xmath6 , c. thomas@xmath7 , c. wilkin@xmath10 , m. wolke@xmath5 , p. wstner@xmath11 , d. wyrwa@xmath8 + _ @xmath3 institute of nuclear physics , cracow , poland _ + _ @xmath5 ikp , forschungszentrum jlich , germany _ + _ @xmath4 iucf , bloomington , indiana , usa _ + _ @xmath6 institute of physics , jagellonian university , cracow , poland _
+ _ @xmath7 ikp , westflische wilhelms universitt , mnster , germany _ + _ @xmath10 university college london , london wc1e 6bt , united kingdom _
+ _ @xmath11 zel , forschungszentrum jlich , germany _ * pacs : * 12.38.qk , 13.85.hd , 25.40.ve + * keywords : * scattering length , effective range , dalitz plot 4ex
|
feature selection is a central theme in analyzing many variants of magnetic resonance imaging ( mri ) data . supervised approaches that are highly capable of performing regression or classification , but do not rely on features , are at best specialized maps between input data and the output labels .
they lack the crucial component of `` inference '' to produce generalizations _ about _ the input data .
meanwhile , the inference process and ability to find generalizeable features or structure in the data is at the core of scientific discovery ; in mri research , such structures are necessary for the general goal of understanding the brain .
inferring the latent structure is generally the goal of unsupervised learning , which has had a wide success in analyzing mri data .
when combined with supervised learners , these structures have a diagnostic value . independent component analysis ( ica )
@xcite is a representative approach , which has found success as a means for inferring the latent structure in brain imaging data represented via a linear mixture of maximally - independent sources @xcite .
while linear mixture models have been very successful in neuroimaging applications , their success relative to nonlinear models @xcite is due to simple and tractable inference , not due to a strong belief that linearity is a correct depiction of the latent structure . for linear mixtures ,
non - gaussian sources are necessary to ensure uniqueness , as for gaussian sources one can not guarantee any independence beyond the correlation @xcite .
luckily , the converse is true : non - gaussian sources with linear mixtures assure maximum independence under a generative learning framework , such as maximum likelihood estimation ( mle ) @xcite .
requiring the prior distribution be non - gaussian , while enabling inference and learning with linear ica methods , breaks down when the relationship between data and sources is nonlinear , necessitating more advanced methods .
although nonlinear versions of ica @xcite , as well as some alternative nonlinear methods @xcite exist , each comes with its shortcomings , and none have been successful enough to supplant linear ica .
alternatively , nonlinear independent component estimation ( nice ) @xcite is a method for drawing from a family of nonlinear transformations , @xmath0 , parameterized by feed forward networks ( ffns ) such that computing the jacobian and the inverse are tractable .
while nice can estimate sources from nonlinear mixings better than ica in simulations , it is also limited to square transformations and requires principle component analysis ( pca ) to be practical in a medical imaging setting @xcite .
in addition to being constrained to square transformations , ica and many nonlinear variants can not incorporate multimodal data in a natural way .
the linear mixing assumption is harder to justify when modes are drawn from fundamentally different distributions , such as mri , electroencephalography ( eeg ) , and other variables such as age , gender , and clinical diagnoses .
the above issues are ongoing challenges for realizing the full potential of a deep independence network ( din ) or a general infomax approach @xcite for feature extraction in medical imaging .
it is possible that lack of progress has been due to the strong requirement in ica that the data be the output of a deterministic map of sources . as an alternative
, we propose learning features in a directed graphical model setting using recent advances in variational inference and demonstrate the effectiveness of this approach with mri data .
linear mixture models such as ica fall under a more general category of _ volume - preserving bijective maps _
@xcite , such that we learn a deterministic parameterized transformation , @xmath1 , along with a prior distribution of the sources : @xmath2 where @xmath3 is the density of the data , @xmath4 is the density of the sources , and @xmath5 is the jacobian . for ica
, we have two constraints : first , @xmath6 , is a linear transformation with square unmixing matrix , @xmath7 , and second , the prior distribution of the sources , @xmath8 , is non - gaussian .
probabilistic ica ( pica ) @xcite relaxes the square requirement , but learning is still reliant on a linear transformation as well as a noise operator with known covariance .
being a linear transformation , computing the jacobian , and hence learning , is tractable , but this can not be said about general nonlinear transformations .
nonlinear independent component estimation ( nice ) gets around this problem by parameterizing @xmath0 as a feed - forward network ( ffn ) , such that the affine transformation at each layer is lower or upper triangular , but it is still limited to square transformations .
a directed graphical model or _ bayesian network _ is a generative model that represents the density of the data as the marginal of the joint : @xmath9 , which is composed of a set of prior and conditional distributions that make up an acyclic graph .
directed graphical models have been used in various medical imaging settings @xcite , but have been limited to relatively simple , often linear configurations .
a special case of the bayesian network is the _ directed belief network _ : a hierarchical model that represents the joint via layers of latent variables that within a layer are conditionally independent : @xmath10 where @xmath11 is the prior distribution of the top or @xmath12th layer
. directed graphical models are most commonly trained using the maximum - likelihood estimation ( mle ) method , which maximizes the log - likelihood of the data by adjusting parameters of the conditional and prior distributions . when present , latent variables need to be marginalized out at each stage during the process ; but training can become difficult as marginalizing the joint distribution over the latent variables is often computationally infeasible .
potentially , learning can be aided by the use of a posterior , @xmath13 , such that @xmath14 ; however , the exact posterior can be equally intractable , particularly when the conditional distributions are complex ( e.g. , parameterized by highly nonlinear functions ) .
some recent advances allow us to more easily train directed graphical models . _ variational inference _ makes use of an approximate posterior to compute the variational lower bound of the log - likelihood , @xmath15 : @xmath16 where we have used a monte carlo estimate for the generative term of the bound
, @xmath17 are @xmath18 independent samples drawn from the approximate posterior , and @xmath19 is the entropy of the approximate posterior . the most notable advances in variational inference were made in `` helmholtz machines '' @xcite that model the approximate posterior and conditional distributions by deep neural networks @xcite . in this model , the difficulty is offset from inference to training the approximate posterior modeled by the `` recognition network '' .
for example , suppose the conditional distribution is modeled by an ffn , such that the output makes up the parameters of a multivariate gaussian distribution with mean , @xmath20 , and diagonal covariance , @xmath21 .
let us assume as well that the approximate posterior has a logistic distribution with mean , @xmath22 , and scale , @xmath23 : @xmath24 where @xmath0 and @xmath25 are multilayer ffns with parameters @xmath26 and @xmath27 , @xmath28 are visible variables corresponding to data , and @xmath8 are the latent variables ( or sources ) .
finally , assume @xmath29 , the prior distribution of the latent variables , is a spherical multivariate logistic distribution .
the lower bound in equation [ eq : approx_logp ] becomes : @xmath30\end{aligned}\ ] ] the gradient of the first term above w.r.t the variational parameters , @xmath27 , is not normally possible due to the stochastic variables @xmath31 .
however , in the case of logistic latent variables , the following re - parameterization makes learning possible via back - propagation : @xmath32 commonly known as a _
variational autoencoder _ ( vae )
@xcite , this type of re - parameterization is available for a number of continuous distributions , such as gaussian , poisson , and gumbel , but is not available for helmholtz machines with discrete latent variables , though other good methods exist @xcite .
as the prior is factorized , the lower bound corresponds to learning a generative model with maximally - independent latent variables , a feature desirable in many research settings .
this approach should , in principle , work for any directed graph with continuous latent variables , given the appropriate approximate posterior and prior .
visualizing a latent variable , @xmath33 , of a directed belief network involves calculating the marginal over all other latent variables : @xmath34 this is computationally infeasible with most configurations .
alternatively , we can draw @xmath18 samples from the approximate posterior , @xmath35 to approximate the marginal : @xmath36 however , this approximation typically requires a large number of samples to be accurate ( e.g. @xmath37 with the mnist dataset ) .
in addition , this only provides a single point in the marginal , which is a continuous function of @xmath33 . in reality
, we are interested in how changes in @xmath33 effect generation of the image .
therefore , we use the following fast approximation to determine the projection " of the @xmath38th latent variable : @xmath39 where @xmath40 and @xmath41 is reserved for parameters of the prior distribution that encode first and second order statistics respectively . for instance
, for a logistic distribution , @xmath40 would be the center of unit @xmath38 and @xmath41 would be the scale factor .
this approximation does not capture the full generative effect of the latent variables , but it is sufficient for this demonstration .
for our medical imaging study , we used the mri dataset from the combined schizophrenia studies in plis et .
al . @xcite .
whole brain mris were obtained on a 1.5 t signa ge scanner using identical parameters and software , and the resulting dataset was segmented into grey matter regions with @xmath42 voxels in each sample . for quality control ,
the correlation coefficient of each mri volume was calculated and any volumes with mean coefficient of @xmath43 standard deviations below the mean across all volumes were categorized as noisy and removed . the resulting dataset had @xmath44 subjects and @xmath45 healthy controls . for our generative model , we used a logistic prior , @xmath46 , with @xmath47 units and a 2-layer generation " feed - forward network ( ffn ) with a deterministic intermediate layer with @xmath48 softplus ( @xmath49 ) units to parameterize a gaussian conditional distribution , @xmath50 .
our approximate posterior , @xmath51 , was a multivariate factorized logistic which was parameterized by a 2-layer recognition " ffn with @xmath48 hyperbolic tangent ( @xmath52 ) deterministic units .
we learn @xmath27 , @xmath26 , and @xmath53 by maximizing the variational lower bound , and trained our model with a batch size of @xmath54 using the rmsprop algorithm @xcite for @xmath55 epochs .
as the latent variable projections from section [ sec : viz ] were both positive and negative and the prior distribution is symmetric with respect to our choice of positive scale factors , we reversed the sign of our projections if the mean of voxels above @xmath43 standard deviations was negative .
for each latent variable or component " , we calculate the approximate posterior for each subject , @xmath56 , and then used logistic regression to schizophrenia using the approximate posterior means , @xmath57 .
each component was tested for significance by using the resulting @xmath58 values from the logistic regression in a one - sample @xmath59-test .
visual inspection of the latent variables revealed a diverse set of features that were mostly identifiable as regions of interest , with very little noisy features .
there was significantly more overlap between features than is typical with ica with pca preprocessing or rbm with mri data @xcite , which may or may not be beneficial depending on the research setting .
latent variables that showed high significance to schizophrenia ( @xmath60 ) are shown in figure [ fig : features ] with the complete set in the supplementary material .
[ fig : s_corr ] [ fig : comp_corr ] the means of the approximate posterior , @xmath57 , were used as input to a classification task , using simple logistic regression and @xmath61-fold class - balanced cross validation .
the resulting classification rate , @xmath62 , is significantly above chance .
the conclusion is that , despite lacking information about the labels in the mle objective and much lower dimensionality , the latent variables have a similar amount of information necessary to perform diagnosis .
this is also apparent in the correlation matrix in figure [ fig : s_corr ] using the components that showed high significance to schizophrenia .
finally , the components were grouped by calculating the correlation of approximate posterior centers across subjects .
figure [ fig : comp_corr ] shows several groupings , as well as some inter - group relationships .
we have demonstrated variational autoencoders as a means of training nonlinear directed graphical models for extracting maximally indepdent features from mri data .
our results show both relevant structure and preservation of information relevant to schizophrenia diagnosis .
this work opens the door for further studies using helmholtz machines for medical imaging research , including multimodal and multilayer analysis .
|
independent component analysis ( ica ) , as an approach to the blind source - separation ( bss ) problem , has become the de - facto standard in many medical imaging settings . despite successes and a large ongoing research effort , the limitation of ica to square linear transformations
have not been overcome , so that general infomax is still far from being realized . as an alternative
, we present feature analysis in medical imaging as a problem solved by helmholtz machines , which include dimensionality reduction and reconstruction of the raw data under the same objective , and which recently have overcome major difficulties in inference and learning with deep and nonlinear configurations .
we demonstrate one approach to training helmholtz machines , variational auto - encoders ( vae ) , as a viable approach toward feature extraction with magnetic resonance imaging ( mri ) data .
|
the analytic description of the vacuum spacetime surrounding a rapidly rotating neutron star is still an open problem .
the analytic structure of the spacetime outside a slowly rotating star , and its relation to the kerr metric , has been well understood since the seminal works of hartle ( 1968 ) and hartle & thorne ( 1969 ) . on the other hand , numerical solutions of the einstein equations for stars rotating up to the mass - shedding limit
are now routinely obtained with a number of different methods , such as the komatsu , eriguchi and hachisu ( 1989 ) method ( see stergioulas 2003 , for an extensive comparison of the different existing numerical methods ) .
these numerical solutions are indeed useful for modelling astrophysical systems , for studying linear perturbations of rapidly rotating relativistic stars and as initial data for dynamical evolutions of spacetimes in numerical relativity ( see e.g. stergioulas & friedman 1998 , stergioulas , kluzniak & bulik 1999 , stergioulas & font 2001 ) . despite the availability of numerical solutions , a consistent analytic representation of the vacuum
metric outside a rapidly rotating neutron star is desirable for several reasons . in the first place ,
having an analytic form for the metric simplifies the computation of the _ stationary _ properties of the spacetime .
for example , if an accurate analytic solution were available , geodesics in the neutron star exterior could be studied analytically , and one could find closed - form expressions for the radii and frequencies of the innermost stable circular orbits ( iscos ) . in turn , this would simplify the calculation of properties of accretion disks , of epicyclic frequencies , of accretion luminosities , and so on . furthermore , having an analytic solution could prove useful to the study of _ dynamical _ properties of the spacetime , such as gravitational wave emission .
one of the unsolved problems in gravitational - wave theory is the study of the quasinormal modes of rapidly rotating neutron stars .
these can be computed either in the frequency domain , as an eigenvalue problem , or in the time domain , evolving numerically the ( linearized or full ) einstein equations and then computing the outgoing radiation .
the major technical issue in this problem is related to the difficulty of imposing outgoing - wave boundary conditions at infinity , since a rapidly rotating neutron star spacetime is expected to deviate significantly from petrov type d. having in hand an accurate analytic metric for the exterior spacetime one could envisage the possibility of computing the weyl scalars in closed form , looking for neutron star models which are , in some suitably defined sense , `` close to petrov type d '' ( baker & campanelli 2000 ) .
if the spacetime is `` close enough to type d '' one could then apply approximation schemes to impose the outgoing - wave boundary conditions .
the idea here is to improve the presently available methods , which are generally based on the use of the zerilli functions ( see e. g. abrahams _ et al .
_ 1992 , allen _ et al .
_ 1998 , rupright _ et al .
_ 1998 ) - i.e. , on perturbations of _ spherically symmetric _ vacuum spacetimes .
only recently , the teukolsky formalism for perturbations of kerr black holes has been used for the purpose of wave extraction in the final phase of binary black holes mergers ( baker _ et al .
_ 2002 ) . until the development of a powerful integral equation method , devised by sibgatullin in 1984 ( see sibgatullin 1991 and manko & sibgatullin 1993 for details ) , finding analytic solutions to the einstein equations for stationary axisymmetric spacetimes was largely a matter of guesswork .
one typically had to choose some ansatz to simplify the mathematical problem of obtaining the solution ; then one verified _ a posteriori _ that the obtained solution had physically acceptable properties . in sibgatullin s
method one knows the physical characteristics of the solution to be constructed from the very beginning , through the choice of the axis expressions of the ernst potentials .
a complete analytic representation of axisymmetric spacetimes can be obtained in terms of a series expansion whose coefficients are the physical multipole moments ( fodor , hoenselaers & perjes 1989 , ryan 1995 ) . in principle , this gives an approximation to a numerical spacetime that can be made arbitrarily accurate : one would need to include a sufficiently large number of multipole moments and match them to some given numerical solution .
however , such a procedure involves a very large number of expansion coefficients , which makes it difficult to use for practical purposes .
some applications of this idea have already appeared : for example , shibata & sasaki ( 1998 ) derived formulae for the location of the isco around rapidly rotating neutron stars . quite recently , manko _ et al . _
( 2000b ) were able to find a new asymptotically - flat solution to the ernst equations for the einstein - maxwell system .
this solution is very interesting because it is given _ in closed form_. furthermore , when two of its parameters ( i.e. , the charge and magnetic moment ) are set to zero , the solution depends only on _ three parameters _ :
mass , angular momentum and a third parameter @xmath0 , which is related to the spacetime s physical quadrupole moment . with this simplification ,
the solution reduces to a particular three - parameter specialization of the kinnersley - chitre ( 1978 ) solution ( a generalization of the tomimatsu - sato @xmath1 spacetime ) .
notice however that kinnersley and chitre only constructed the relevant ernst potential ( they did not provide explicit expressions for the corresponding metric functions ) .
furthermore , the kinnersley - chitre solution is restricted to the subextreme case ( @xmath2 ) . on the other hand , in the solution by manko _
et al . _ , when electric and magnetic fields are set to zero @xmath3 and @xmath4 are allowed to assume arbitrary real values , because the parameter set in their solution is analytically extended .
therefore the kinnersley - chitre solution is obtained as a particular case of the analytic solution in manko _
( 2000b ) when certain restrictions are imposed on the parameters of that solution .
there have been attempts in the literature to fix the free parameters in analytic exterior solutions by matching them to numerical solutions .
however , different matching conditions were used .
for example , sibgatullin & sunyaev ( 1998 , 2000 ) fixed the free parameters appearing in a different analytic solution using the radii of marginally stable circular orbits , or a suitably defined redshift parameter at the stellar equator . for their metric , which is different from the one we consider here , they found that corrections due to the quadrupole moment can accurately reproduce the properties of the `` exact '' exterior spacetime only for several equations of state ( eoss ) , with the exception of eoss with large phase transitions .
a simple , closed form expression for the analytic metric used in sibgatullin & sunyaev ( 1998 , 2000 ) was given explicitly by sibgatullin ( 2002 ) .
a matching procedure based on the redshift parameter was again used by stute & camenzind ( 2002 ) .
our own preference here is to avoid matching using _
local _ properties and , instead , match the solution s mass - quadrupole moment , which is a _ global _ property of the spacetime .
furthermore , it is well known that deviations from the slow - rotation behavior in rapidly rotating stars , due to the stellar oblateness , are determined mainly by the mass - quadrupole moment .
the quadrupole moment was also used in matching the analytic and numerical solution in manko _
( 2000a ) .
the plan of the paper is as follows . in section [ numgravfield ]
we describe the procedure to numerically compute the spacetime describing a rapidly rotating compact star using the komatsu - eriguchi - hachisu ( 1989 ) self - consistent field method , as modified by cook , shapiro and teukolsky ( 1994 , henceforth cst ) .
in particular , we discuss how to implement this method for a numerical evaluation of the spacetime s multipole moments . in section [ analgravfield ]
we present the analytic solution recently obtained by manko _
( which is only valid in the vacuum prevailing outside the rotating neutron star ) and describe its multipolar structure . in section [ match ]
we describe our procedure to match manko s analytic solution to the numerically obtained spacetime , and derive the coordinate transformation relating the two metrics .
section [ checksol ] is devoted to a discussion of the tests we used in order to understand `` how close '' the analytic and numerical spacetimes are . as we will discuss in the following , there are two possible families of analytic solution for which the mass - quadrupole moment of the analytic solution matches to the mass - quadrupole moment of the numerical spacetime .
the current - octupole moment of the first family of solutions is very close to the current - octupole moment of the numerical spacetime , while the second solution is close to the kerr spacetime .
an examination of the metric functions on the equatorial plane and on the rotation axis confirms that the first solution is also the one which better approximates the numerically obtained metric functions . as an independent check ,
we compute the location of iscos in the spacetime surrounding the rotating star using different approaches .
in particular we locate iscos using the analytic solution , and compare the results thus obtained : 1 ) to the iscos found by numerical integration of the einstein equations , and 2 ) to the analytic formulae for the isco s obtained by shibata & sasaki ( 1998 ) , truncated at different orders of approximation . in most cases where a corotating isco exists , the analytic solution has an accuracy consistently better than the shibata - sasaki expansion .
only in some cases the higher - order multipoles that are missing in the analytic solution significantly increase the error in computing the location of the isco .
finally , we compare our matching procedure to previous work by manko _
( 2000a ) and by stute & camenzind .
the conclusions follow .
to begin with , in this section we briefly discuss the procedure for obtaining highly - accurate numerical solutions for the spacetime of rapidly rotating neutron stars and for computing their multipole moments . for more details the reader
is referred to the review article by stergioulas ( 2003 ) .
the interior and exterior spacetime of a stationary , axisymmetric star is described by a metric in the following form : [ cstmetr ] ds^2=-e^2dt^2+b^2 e^-2 r^2 ^ 2(d - dt)^2 + e^2(dr^2+r^2d^2 ) , where @xmath5 and @xmath6 are four metric functions to be determined by solving four field equations . in the numerical method of komatsu _ et al . _
( 1989 , henceforth keh ) one defines two auxiliary functions @xmath7 , @xmath8 through the relations @xmath9 and @xmath10 .
then , three out of the four field equations are written in the following integral forms & & |(r,)=-_n=0^e^-|/2 + & & \ { _ 0^dr_0 ^ 1 d r^2 f_2n^2(r , r ) p_2n()s_|(r, ) } + & & p_2n ( ) , [ rho ] & & |(r,)=-2r_n=1^e^-|/2 + & & \ { _ 0^dr_0 ^ 1 d r^2 f_2n-1 ^ 1(r , r ) 2n-1 s_|(r, ) } + & & , & & ( r,)=-1r_n=1^e^(2|-|)/2 + & & \ { _ 0^dr_0 ^ 1 d r^3 f_2n-1 ^ 2(r , r)2n(2n-1 ) p_2n-1 ^ 1()s_(r, ) } + & & p_2n-1 ^ 1 ( ) , [ omega ] where f_n^1(r , r)&=&(rr)^nrr , + f_n^1(r , r)&=&(rr)^nr>r , + f_n^2(r , r)&=&1r(rr)^nrr , + f_n^2(r , r)&=&1r(rr)^nr>r , and @xmath11 , @xmath12 and @xmath13 are lengthy source terms , whose expressions can be found in keh . in the equations above ,
@xmath14 , while @xmath15 denotes the legendre polynomials and @xmath16 the associated legendre functions . the metric function @xmath17 is determined by an ordinary differential equation .
we compute numerical equilibrium models using the code by stergioulas and friedman ( 1995 ) ( see nozawa _ et al .
_ 1998 and stergioulas 2003 for extensive accuracy tests ) .
the numerical code uses the cst formulation , in which the keh equations are written in terms of a compactified coordinate @xmath18 defined through the relation r = r_e(s1-s ) , where @xmath19 is the ( coordinate ) radius of the stellar equator .
this allows the computation of the whole exterior spacetime out to infinity , which is important in detailed comparisons of the numerical metric to the analytic metric . for a configuration that is stationary , axisymmetric , symmetric with respect to reflections in the equatorial plane and asymptotically flat
, the spacetime can be characterized by two sets of scalar multipole moments : the even - valued mass moments ( @xmath20 ) and the odd - valued current moments ( @xmath21 ) .
ryan ( 1997 ) presented a method for extracting the multipole moments from the asymptotic form of the metric functions .
the lowest - order appearance of each moment in terms of a power series in @xmath22 is determined by the expansions |=_n=0^
-2m_2nr^2n+1p_2n ( ) , [ rhotilde ] and = _ n=1^. [ omegatilde ] by comparison with equations ( [ rho ] ) and ( [ omega ] ) one finds that [ mn ] m_2n=12_0^dr_0 ^ 1 d r^2n+2 p_2n()s_|(r, ) , and [ sn ] s_2n-1=14n_0^dr_0 ^ 1 d r^2n+2
p_2n-1 ^ 1()s_(r, )
. thus , in general , any model of a rapidly rotating neutron star has an infinite number of mass- and current - multipole moments . in order to match an analytic exterior metric to a numerically - computed interior metric and to check the accuracy of the matching procedure , we computed the mass - quadrupole moment @xmath23 and the current - octupole moment @xmath24 .
an alternative , asymptotic method for evaluating the multipole moments was introduced by laarakkers & poisson ( 1997 ) .
we also used their method in order to cross - check the results obtained from the integral relations ( [ mn ] ) and ( [ sn ] ) . the idea , in this case , is to evaluate numerically the coefficient of @xmath25 in the general expression for ( [ rho ] ) - or analogously , the coefficient of @xmath26 in ( [ omega ] ) - at the outermost grid points ( i.e. , as @xmath27 ) , and multiply the result by the appropriate factor ( containing powers of @xmath28 ) that can be obtained from equations ( [ rhotilde ] ) and ( [ omegatilde ] ) .
we have checked that the two methods typically agree to better than one part in @xmath29 .
the equilibrium solutions for a given eos form a two - parameter family .
in particular , _ stable _ equilibrium solutions are bounded by four limit sequences .
these limits are shown in figure [ statickepler ] , which displays the gravitational mass @xmath3 vs. the central energy density @xmath30 for one the eoss derived by akmal , pandharipande and ravenhall ( 1998 , henceforth apr ) . as an illustrative example we consider the apr eos which does not include boost interactions ( we refer to the original paper for details ) .
the qualitative picture does not change when we consider other eoss ( see eg .
cst , where plots are presented for a representative sample of eoss ) .
the solid line is the _
static _ limit - that is , the sequence of nonrotating solutions to the standard tolman - oppenheimer - volkoff equations .
the long - dashed line is the _ mass - shedding _ ( kepler ) limit , which is determined by the condition that the centrifugal force exactly balances the gravitational attraction at the stellar equator , in which case a fluid element on the equator has the same angular velocity as a free particle in a keplerian orbit at the same location .
both sequences terminate at high central density at the _ stability _ limit , where equilibrium solutions are marginally stable to axisymmetric perturbations ; and they terminate at low central densities at the _ low - mass _ limit , below which a neutron star can not form ( not shown in figure [ statickepler ] ) . within the class of stable equilibrium solutions , cst pointed out the significance of constant rest - mass sequences , called _ evolutionary sequences _ , since an isolated neutron star , slowly losing energy and angular momentum via some dissipative process ( e.g. electromagnetic emission or gravitational - wave radiation ) , must evolve conserving the total baryon number , and hence its rest mass @xmath31 .
an accreting neutron star in a binary system will not evolve along a constant rest - mass sequence : the actual sequence depends on several parameters , such as the magnetic field , accretion rate etc .
nevertheless , the constant rest - mass sequences in cst have been used in the past in evaluating the accuracy of analytic exterior solutions and we will also use them here solely for the same reason .
we compute three constant rest - mass sequences for each eos : * the sequence corresponding to a canonical neutron star having gravitational mass @xmath32 in the non rotating limit , * the sequence terminating at the maximum - mass model in the non rotating limit ( maximum - mass normal sequence ) . * a supramassive sequence , i.e. , a sequence which does not terminate at a nonrotating model .
we include the following set of eoss . for comparison with cst
we include eoss a , au , fps and l. we refer to their paper for an extensive discussion of each eos .
we supplement the set of eoss considered by cst with a relatively new model : the model derived by akmal , pandharipande and ravenhall ( 1998 ) from hamiltonian many - body theories of nuclear matter , including boost corrections in the hamiltonian ( henceforth , we will refer to this model as apr - b , where `` b '' stands for `` boosted '' ) . in tables ( [ eosa][eosaprb ] ) we give numerical results for the structure properties of the models we have computed . all models have been computed using a resolution of ( @xmath33 angular points ) @xmath34 ( @xmath35 radial points ) , corresponding to a typical accuracy of at least one part in @xmath29 in all quantities .
each table corresponds to a constant rest mass sequence , and lists : the total central energy density @xmath30 in units of @xmath36 g @xmath37 ; the angular velocity @xmath38 in units of @xmath29 s@xmath39 ; the moment of inertia @xmath40 in units of @xmath41 g @xmath42 ( for rotating models only ) ; the gravitational mass @xmath3 in solar masses ; the ratio of rotational kinetic energy to gravitational binding energy @xmath43 ; the equatorial circumferential radius of the star @xmath44 in km and the height ( in km ) of corotating ( @xmath45 ) and counterrotating ( @xmath46 ) iscos from the surface of the star ( if an isco does not exist , the corresponding entry is omitted ) .
the height of an isco is defined as the difference between the circumferential radius at the isco and the circumferential equatorial radius of the star .
the next three columns give the first few physical multipoles in geometrized units of @xmath47 : namely , we list the mass quadrupole moment @xmath48 in km@xmath49 , the angular momentum @xmath50 in km@xmath51 , and the current octupole moment @xmath24 in km@xmath52 .
we have checked our code by reproducing the quadrupole moments computed by laarakkers & poisson and found excellent agreement .
the accuracy in computing @xmath24 was checked by comparing the integral form to the asymptotic form mentioned in section [ numsol ] , finding good agreement .
this shows that using the compactified coordinate introduced in cst allows a very accurate numerical determination of relatively high - order multipoles .
the last column gives the value of the `` quadrupole '' parameter @xmath0 ( in km ) for which the analytic solution provides a good approximation of the numerical spacetime .
when matching the quadrupole moment of the numerical and analytic spacetimes is not possible , the corresponding entry is omitted .
more details on the procedure we followed to obtain the values listed in this column will be given in section [ match ] .
in this section we summarize the properties of the vacuum analytic solution obtained by manko _
. we will concentrate in particular on the multipolar structure of the solution , since our ultimate purpose will be to reproduce accurately the first few multipoles of the numerical spacetimes we discussed in the previous section . in the vacuum region surrounding a stationary and axisymmetric star ,
the spacetime only depends on three metric functions ( while four metric functions are needed for the interior ) .
the most general form of the metric ( papapetrou 1953 ) is [ manko ] ds^2=-f(dt - wd)^2+f^-1\ { e^2 ( d^2+dz^2)+^2d^2 } .
here @xmath53 , @xmath54 and @xmath55 are functions of the quasi - cylindrical weyl - lewis - papapetrou coordinates @xmath56 . starting from this metric one can write down the vacuum einstein - maxwell equations as two equations for two complex potentials @xmath57 and @xmath58 , following a procedure due to ernst ( 1968 ) . the equations are : ( re \{e}+||^2)^2e&= & ( + 2^ * ) + ( re \{e}+||^2)^2&= & ( + 2^*)once the potentials are known , the metric can be reconstructed .
sibgatullin ( 1991 ) devised a powerful procedure for reducing the solution of the ernst equations to simple integral equations .
basically , one starts with a choice for the values of the ernst potentials on the symmetry axis e(z)(=0,z ) , f(z)(=0,z ) , solves two complex - valued integral equations , and checks that the obtained solution satisfies the expression of the ernst potentials in terms of physical multipoles : & & e=1 - 1 + , = q1 + + & & ( = 0)=_n=0^m_n z^-(n+1 ) , + & & q(=0)=_n=0^q_n z^-(n+1 ) the real parts of @xmath59 are the mass multipoles , the imaginary parts of @xmath59 are the current multipoles , the real parts of @xmath60 are the electric multipoles and the imaginary parts of @xmath60 are the magnetic multipoles .
after more than ten years of work in the field , manko _ et al . _
( 2000b ) were finally able to find a vacuum solution involving five parameters ( mass , angular momentum , charge , magnetic dipole moment and mass quadrupole moment ) which can be expressed in terms of relatively simple rational functions .
we are particularly interested in solutions having no charge or magnetic dipole moment .
if we denote by @xmath3 the gravitational mass of the star , by @xmath4 the specific angular momentum ( @xmath61 ) , and introduce a parameter @xmath0 which can be related to the mass quadrupole moment , their choice for the axis values of the ernst potentials is : @xmath62 with & = & -m^2b^2m^2-(a - b)^2 , + d&=&14 . to be able to write the metric in rational form , one must introduce generalized spheroidal coordinates @xmath63 where @xmath64 and k=. the inverse transformation between the two sets of coordinates is = k(1-y^2)^1/2(x^2 - 1)^1/2,z = kxy . the metric is then written as @xmath65,\end{aligned}\ ] ] with f = ed , e^2 = e16k^8(x^2-y^2)^4 , w=-(1-y^2)fe , and d&= & \{4 ( k^2x^2-y^2)^2 + 2kmx [ 2k^2(x^2 - 1 ) + & + & ( 2+ab - b^2)(1-y^2 ) ] + & + & ( a - b ) [ ( a - b)(d-)-m^2b ] ( y^4 - 1)-4d^2 } ^2 + & + & 4y^2\{2k^2(x^2 - 1)[kx(a - b)-mb]-2mb(1-y^2 ) + & + & [ ( a - b)(d-)-m^2b ] ( 2kx+m)(1-y^2 ) } ^2 + & & + e&= & \{4 [ k^2(x^2 - 1)+(1-y^2)]^2 + & + & ( a - b ) [ ( a - b)(d-)-m^2b ] ( 1-y^2)^2}^2 + & - & 16k^2(x^2 - 1)(1-y^2)\ { ( a - b)[k^2(x^2-y^2)+2y^2 ] + & + & m^2by^2 } ^2
+ & & + f&= & 8k^2(x^2 - 1)\ { ( a - b)[k^2(x^2-y^2)+2y^2 ] + y^2m^2b } + & & \ { kmx [ ( 2kx+m)^2 - 2y^2(2+ab - b^2)-a^2+b^2 ] + & -&2y^2(4d - m^2b^2 ) } + \ { 4[k^2(x^2 - 1)+(1-y^2)]^2 + & + & ( a - b ) [ ( a - b)(d-)-m^2b ] ( 1-y^2)^2 } + & & ( 4(2kmbx+2m^2b)[k^2(x^2 - 1)+(1-y^2 ) ] + & + & ( 1-y^2)\ { ( a - b)(m^2b^2 - 4d ) + & -&(4kmx+2m^2 ) [ ( a - b)(d-)-m^2b ] } ) . in order for the solution to satisfy the requirements of axisymmetry , stationarity and reflection - symmetry in the equatorial plane , all three parameters @xmath3 , @xmath4 and @xmath0 must be real .
here we examine the multipolar structure of the analytic solution by manko _
et al . _ for rotating and nonrotating solutions .
the only nonvanishing multipole moments of the solution are the gravitational mass @xmath66 , the quadrupole moment @xmath67 , the angular momentum @xmath68 and the current octupole @xmath69 .
the quadrupole moment and the current octupole moment are given in terms of the three parameters @xmath3 , @xmath4 and @xmath0 as [ quadanal ] q =- m(d -- ab+a^2 ) , and @xmath70-b(d-\delta ) \right\}.\ ] ] however , since @xmath4 and @xmath0 are independent parameters , setting @xmath4 equal to zero does not automatically imply a vanishing @xmath71 and @xmath24 , as would be the case for a realistic solution of a nonrotating perfect fluid star . instead , the nonrotating solution ( @xmath72 ) has a quadrupole moment equal to @xmath73 and a current octupole moment equal to @xmath74 it is obvious that there exists no real value of the parameter @xmath0 for which the quadrupole moment vanishes for a nonrotating star .
for @xmath75 , the solution is oblate ( @xmath76 ) with a minimum quadrupole deformation obtained for @xmath77 @xmath78 at @xmath79 , the nonrotating multipole moments @xmath71 and @xmath24 diverge , while for @xmath80 , the nonrotating solution is prolate ( @xmath81 ) with a minimum quadrupole deformation of @xmath82 at @xmath83 . obviously , the analytic solution by manko _ et al . _ does not reduce continuously to the schwarzschild solution as the rotation vanishes .
it can only reduce to other forms of nonrotating vacuum solutions ( e.g. the well - known weyl solutions ) that could be matched to other interior solutions , such as nonrotating stars with non - isotropic stresses , inducing nonvanishing quadrupole deformations .
nevertheless , as we will show next , the analytic solution can approximately describe a rapidly rotating fluid star , when the rotation rate is large enough , so that the quadrupole deformation induced by the rotation roughly exceeds the minimum nonvanishing oblate quadrupole deformation of the solution in the absence of rotation , i.e. roughly when @xmath84 since the quadrupole moment is roughly proportional to @xmath85 , one expects that the analytic solution could be relevant for rotation rates of roughly @xmath86 , where @xmath87 is a dimensionless measure of the angular momentum of the star .
we will confirm this expectation by direct comparisons with numerical solutions in the next section .
it is interesting that the kerr solution can still be obtained from the analytic solution , if one accepts the following imaginary form for the parameter @xmath0 [ kerrlimit2 ] b = i , with @xmath88 . in this case , one recovers the correct expressions @xmath89 and @xmath90 .
the three parameters of the analytic solution ( @xmath91 and @xmath0 ) can be set at will .
however , only certain combinations of values can correspond to specific models of rapidly rotating neutron stars . since the solution has four nonvanishing multipole moments , but only three free parameters , one can at most match three multipole moments of any given numerical solution .
the fourth multipole moment will then be determined by the analytic solution , and its relative difference with the ( known ) numerical value will be a measure of the accuracy of the analytic solution .
the four multipole moments are not equally important for specifying a solution , @xmath24 being the least important , even for the most rapidly rotating models .
therefore we choose to match the analytic exterior solution to known numerical solutions by matching the gravitational mass @xmath3 , the specific angular momentum @xmath4 and the mass - quadrupole moment @xmath71 .
one then hopes that the resulting analytic solution will yield a value for the current - octupole moment @xmath24 that is close to the corresponding value in the numerical model . as we will show
, there exists a branch of solutions for which this is indeed the case .
( 2000a ) also used the quadrupole moment to match numerical and analytic solutions , but their examples correspond to the other branch of solutions , for which the analytic value of @xmath24 does not agree well with the numerical value . for a given model of a rapidly rotating neutron star ,
we first construct a highly accurate numerical solution , as described in section [ numgravfield ] . in the analytic solution ( [ solution ] ) ,
we set @xmath3 and @xmath4 to be equal to the obtained numerical values .
the remaining parameter @xmath0 is then determined by solving the equation @xmath92 where @xmath93 is the value of the quadrupole moment obtained by the numerical code . a plot of @xmath94 as a function of the parameter @xmath0 , for the most rapidly rotating model of the maximum - mass sequence for eos fps ,
is shown in fig .
[ analytic_bf ] .
two possible real solutions for @xmath0 exist : a solution that is usually negative , @xmath95 , and a solution that is always positive , @xmath96 .
thus , for each set of physical parameters @xmath91 and @xmath71 , there exists two different branches of solutions , with parameters ( @xmath3 , @xmath4 , @xmath95 ) and ( @xmath3 , @xmath4 , @xmath96 ) , respectively . in the remainder of the paper
, we will refer to these two different branches as the negative solution ( - ) and the positive solution ( + ) . as we will show next
, these two branches correspond to very different spacetimes . in the previous section
, we estimated that the analytic solution should be relevant for rapidly rotating neutron stars only for values of @xmath97 roughly larger than 0.5 .
[ analytic_bs ] shows a more slowly rotating model than the model shown in fig .
[ analytic_bf ] , along the same evolutionary sequence .
it is obvious that no solution to equation ( [ qmatch ] ) exists for any real value of the parameter @xmath0 . along each sequence
there is a critical rotation rate above which one can match the numerical interior solution to the analytic exterior solution . in tables ( [ eosa][eosaprb ] ) we list all computed physical properties for the selected sequences .
the last column lists the parameter @xmath98 of the _ negative _ branch of the analytic solution ( when it exists ) .
this is the relevant branch for rapidly rotating neutron stars , as we will show in section [ checksol ] . in tables ( [ eosa][eosaprb ] ) some models appear having @xmath99 .
these models do _ not _ belong to the positive branch @xmath96 .
they are instead models which are very close to the critical value of the rotation parameter , @xmath100 : in these particular cases , both solutions to equation ( [ qmatch ] ) can happen to be positive . however ,
in general ( as long as a model is rotating somewhat above the critical rate ) the negative branch has @xmath101 .
typical values of @xmath102 above which the analytic solution exists are listed in table [ jcrit ] for a subset of the considered eoss . for smaller masses , @xmath102 is usually smaller : therefore , for `` canonical '' neutron stars
( having mass @xmath103 in the non rotating limit ) the analytic solution is valid over a wider range of @xmath97 . in terms of the angular velocity at the mass - shedding limit for uniformly rotating stars ,
the critical rotation rates are given in the right column of table [ jcrit ] .
the critical rotation rate @xmath104 ranges from @xmath105 to @xmath106 for the @xmath107 sequence , with the lower ratio corresponding to the stiffest eos . for the maximum - mass sequence
the ratio is @xmath108 , nearly independent of the eos . in conclusion
, the analytic exterior solution can be useful for studying rapidly rotating neutron stars .
the exterior gravitational field of massive neutron stars created in binary neutron star mergers , supported temporarily by differential rotation against collapse , could also be described , to some accuracy , by the analytic solution ( the accuracy will depend on how significant the higher multipole moments are in the case of strong differential rotation ) . if the eos is very stiff , such as eos l , then the analytic solution is also valid for for description of accreting neutron stars in low - mass - x - ray binaries ( lmxb ) , with rotational periods of a few milliseconds . before presenting specific tests of the accuracy of the analytic solution
, we need to describe the coordinate transformation that relates the interior metric ( [ cstmetr ] ) to the exterior _ vacuum _
metric ( [ manko ] ) ( see islam 1985 ) . for the interior metric ( [ cstmetr ] )
, we define the cylindrical coordinates [ cylind ] r , z r. in vacuum , einstein s field equations imply that @xmath109 one can therefore define a new coordinate [ trho ] b , satisfying the two - dimensional laplace equation ( [ laplace ] ) , and a second coordinate @xmath110 satisfying the cauchy - riemann conditions @xmath111 the coordinate @xmath112 is obtained by integration of the above cauchy - riemann conditions , requiring that @xmath113 in the equatorial plane ( at @xmath114 ) .
it is easy to show that @xmath115^{-1 } \left ( d \tilde \rho^2 + d \tilde z^2 \right),\ ] ] and setting [ mpa2 ]
f & = & e^2 -^2 ^2 e^-2 , + [ mpa ] w & = & - , + e^2 & = & f ^-1 e^2 , [ mpa3 ] the metric in the exterior takes the desired form ( [ manko ] ) .
since the transformation ( [ z1],[z2 ] ) for the coordinate @xmath112 can not , in general , be solved analytically , one can relate a solution for the interior metric ( [ cstmetr ] ) to the exterior metric ( [ manko ] ) only through numerical integration .
the current - octupole moment @xmath24 , like the quadrupole moment @xmath116 , is a function of @xmath4 , @xmath0 and @xmath3 .
once we have fixed @xmath0 by matching the quadrupole moment to the numerical spacetime through equation ( [ qmatch ] ) , there are no more free parameters to be specified ; the current - octupole @xmath24 can be computed using equation ( [ s3 ] ) , and then compared to the value of @xmath24 computed for the numerical metric .
therefore , @xmath24 serves as a good error indicator for the accuracy of the solution .
in fact , the value of @xmath24 obtained analytically for the two branches of solutions , @xmath96 and @xmath95 , can be used to distinguish which of the solutions is more relevant for rapidly rotating neutron stars .
[ s3fig ] displays @xmath24 , for the two branches of the analytic solution , along with the value of @xmath24 for the numerical solution and for the kerr solution , for the evolutionary sequence corresponding to eos fps and having maximum mass in the nonrotating limit ( see table [ eosfps ] ) .
the error for the ( - ) solution is very small , at most of the order of @xmath117 . on the other hand
the error for the ( + ) solution is quite large ( up to 56 @xmath118 for the fastest rotating model ) . in this case
the solution is closer to the kerr value than to the value corresponding to numerical models of rapidly rotating neutron stars .
therefore , in the remainder of this paper we will only use the ( - ) branch of solutions to the matching condition ( [ qmatch ] ) . in table
[ s3error ] we display the relative error @xmath119 in the analytic value of @xmath24 ( when compared to the numerical solution ) for the critically rotating and maximally rotating models of all evolutionary sequences considered in this paper . for most sequences ,
the relative error in @xmath24 can be as large as @xmath120 for the critically rotating models , reducing to a few percent only for the models at the mass - shedding limit .
typically , the largest errors appear for the @xmath121 sequences , while the sequences that terminate at the maximum mass static model have the smallest errors . in most cases
the error is larger for slower rotating models .
this shows that for those models @xmath24 is still influenced by its nonzero value in the non rotating case ( for the analytic solution , the octupole moment @xmath24 , like the quadrupole moment @xmath71 , does not vanish for @xmath72 and @xmath122 ) . for more rapidly rotating models
this influence diminishes and @xmath24 becomes almost entirely of rotational origin , agreeing better with the numerical solution . comparing the various eoss , one sees that the error in @xmath24 for soft eoss , such as eos a , is smaller than the corresponding error for very stiff eoss , such as eos l. the critical @xmath121 model for eos l shows an unusually large relative error of @xmath123 in @xmath24 .
this , again , is related to the compactness of the various models and to the value of the multipole moments for @xmath72 . as a second test of the accuracy of the analytic solution for rapidly rotating neutron stars
, we performed a direct comparison of specific metric components for several representative models , using all eoss in our sample . here , we focus on the most rapidly rotating model of the maximum mass sequence with eos fps , since the other cases we examined showed similar behavior .
for this model we computed the metric components @xmath124 , @xmath125 and @xmath126 on the equatorial plane and along the symmetry axis using the analytic metric and the kerr metric
. then we compared the relative error of both metrics with respect to the corresponding components of the numerical metric .
[ nu_equat ] shows the relative error of the @xmath124-component of the analytic metric and of the kerr metric in the equatorial plane . for the analytic metric
the error is only @xmath127 at the surface of the star ( located at @xmath128 ) , and decreases monotonically with increasing distance , becoming of order @xmath129 near infinity . in comparison ,
the relative difference between the kerr metric and the numerical metric is @xmath130 at the equator ( i.e. , four times larger than the error in the analytic metric ) .
the relative difference between the kerr metric and the numerical metric also decreases with increasing distance , as expected , and for distances larger than about 200 equatorial radii , the difference between the analytic solution and the kerr solution is negligible . in other words , at such a distance the error in the analytic solution is dominated by the kerr contribution at first order in the rotational parameter , while the effects of the higher - order multipole moments @xmath71 and @xmath24 have become unimportant .
the corresponding figure for the relative error in @xmath131 on the equatorial plane is nearly identical to fig .
[ nu_equat ] for @xmath132 .
when we consider @xmath133 in the equatorial plane , the relative error at the surface is @xmath134 for the analytic metric and @xmath135 for the kerr metric .
this larger error for @xmath133 should be expected : this metric component vanishes in the nonrotating limit , so it is more sensitive to contributions by the higher - order multipole moments @xmath71 and @xmath24 than the metric components @xmath124 and @xmath131 . in order to compare the metric components on the symmetry axis , we first need to integrate the cauchy - riemann conditions ( [ caurie ] ) and obtain the coordinate @xmath112 in terms of the coordinate @xmath136 .
this can be done easily , once the numerical solution for the metric function @xmath137 is obtained .
[ nu_offaxis ] shows the relative error in @xmath124 for the analytic solution and the kerr solution along the symmetry axis .
the location of the surface ( as determined from the numerical solution ) is at @xmath138 . at the surface ,
the relative error for the analytic solution is @xmath139 , while it is @xmath140 for the kerr metric .
thus , we see that the effect of a large quadrupole moment @xmath71 shows up predominantly in the metric components along the symmetry axis , while in the equatorial plane this effect is very small .
the reason for this difference is that a rapidly rotating star becomes very oblate , so that the stellar surface on the symmetry axis is located deeper in the gravitational potential well than the surface in the equatorial plane .
the specific example shown in the above figures has polar to equatorial axes ratio of @xmath141 , thus the equatorial radius is roughly twice as large as the polar radius .
the analytic value of @xmath124 on the surface in the equatorial plane is @xmath142 , while it is @xmath143 on the surface on the symmetry axis ( the asymptotic value at large distances is @xmath144 ) .
gravity is stronger on the polar surface , and this justifies a larger relative error in @xmath124 there . at about 3 polar radii ,
the relative error in @xmath124 along the symmetry axis decreases to the @xmath145 level for both the analytic and kerr solutions .
the above direct comparison of metric components shows that the analytic metric is a good approximation to the numerical one ( or , at least , a much better approximation than the kerr metric ) in the equatorial plane , where one expects particle orbits to be astrophysically more relevant . for gravitational - wave extraction in numerical relativity
, the larger inaccuracies near the polar surface could influence the waveforms . in order to minimize this effect ,
the extraction should be done as far as possible from the surface of the star . in any case , the analytic metric is everywhere more accurate than the kerr metric .
thus a perturbative wave - extraction scheme , built with the analytic metric as a background , should yield more accurate waveforms than those obtained with techniques available at present ( which are based on a perturbative extraction of waveforms around a schwarzschild or kerr background ) .
it is well known that not all orbits around relativistic stars are stable . for nonrotating stars ,
the isco is located at a circumferential radius of @xmath146 .
depending on the eos and the mass of the star , the isco can be located outside the stellar surface .
rotation introduces a preferred direction in the @xmath147 coordinate , so iscos around a rotating star belong to two distinct families : a corotating and a counterrotating one . for moderate rotation rates ,
the effect of rotation is to shorten the distance between the surface and the corotating isco . for rapid rotation
, the large quadrupole moment of the star reverses this trend ( notice that even in some newtonian stellar models , large higher - order multipole moments can introduce an isco ( see , zdunik & gourgoulhon , 2001 ; amsterdamski _ et al .
the counterrotating isco radius normally increases with rotation .
detailed computations of iscos for a large number of models and eoss are presented in cst ; we also refer the reader to that paper for the equations defining the iscos that were used in our numerical computations .
testing the accuracy of the analytic solution in computing the properties of iscos is important , as iscos are related to several astrophysical properties of rapidly rotating neutron stars in lmxbs .
an accretion disk can not extend to radii located within the isco , and this sets an upper limit to the keplerian frequency of particles orbiting a star .
this idea could be used , e.g. , in determining whether compact stars in lmxbs are composed of strange matter ( stergioulas , kluzniak & bulik 1999 , gondek _ et al .
in addition , the location of the isco could play a role in the mechanism producing the khz quasi - periodic oscillations observed in many lmxbs ( van der klis 2000 ) : see , e.g. , kluzniak _ et al . _
( 2003 ) .
a circular orbit in the equatorial plane is one for which @xmath148 , and hence @xmath149 the equation for geodesic motion along the radial coordinate @xmath150 reads @xmath151 where @xmath152 and @xmath153 are the conserved energy and angular momentum per unit mass , determined by the conditions @xmath154 .
geodesics become unstable when @xmath155 , or @xmath156 & & \nn \\ + w ' f^2 \sqrt{w'^2f^4+f ' \tilde\rho(2f - f ' \tilde\rho ) } & & \nn \\ { [ } 2f^2 + 2f'^2\tilde\rho^2-f\tilde\rho(4f'+f '' \tilde\rho ) { ] } + \tilde\rho(2f - f ' \tilde\rho ) & & \nn \\ \bigl\{3f ' f^2 - 4f'^2 f\tilde\rho+f'^3\tilde\rho^2+f^2[f '' \tilde\rho & & \nn \\
-w '' f\sqrt{w'^2f^4+f ' \tilde\rho(2f - f ' \tilde\rho ) } ] \bigr\ } \bigr ) / & & \nn \\ \bigl(f^2\tilde\rho^2 \bigl\{w'^2f^4 + 3f ' f\tilde\rho - f'^2\tilde\rho^2 & & \nn \\ -f^2[2+w ' \sqrt{w'^2f^4+f ' \tilde\rho(2f - f ' \tilde\rho ) } ] \bigr\ } \bigr)=0 & & \end{aligned}\ ] ] for corotating orbits ( cf . stute & camenzind , 2002 ) , where @xmath157 indicates a partial derivative with respect to @xmath150 . for counter - rotating orbits
, one can simply use the above equation and change the sign of the star s angular momentum .
shibata & sasaki have used a more general representation of axisymmetric vacuum solutions ( in the form of a series expansion that is completely determined by the physical multipole moments of the spacetime : see fodor , hoenselaers and perjes 1989 and ryan 1995 ) and derived an approximate analytic formula for the location of the isco .
their formula depends on the stellar mass , angular momentum , mass quadrupole , current octupole and mass @xmath158-pole moments . including all terms up to order @xmath159 in the rotation parameter
, they find the following equation for the circumferential radius of the corotating isco : [ ssisco ] r_isco&=&6m(1 - 0.54433j-0.22619j^2 + 0.17989q_2 + & -&0.23002j^3 + 0.26296jq_2 - 0.05317q_3 + & -&0.29693j^4 + 0.44546j^2q_2 - 0.06249q_2 ^ 2 + & + & 0.01544q_4 - 0.11310jq_3 ) . in the previous expression we have introduced dimensionless parameters @xmath160 , @xmath161 and @xmath162 . in the case of the kerr metric , the approximate expression for the location of the corotating isco up to order @xmath163 is ( see , e.g. , shapiro & teukolsky 1983 ) [ kisco ] r^kerr_isco&=&6m(1 - 0.54433j-0.04630j^2 + & -&0.02016j^3 - 0.01110j^4).the location of the counter - rotating isco is obtained from the above formulae by reversing the sign of @xmath97 and @xmath24 . for illustrative purposes we again focus on the three sequences of eos fps ( table [ eosfps ] ) .
we have carried out the calculation for other eoss as well .
although there are quantitative differences between the various models , the qualitative behavior and the relative accuracy between the numerical and analytic solutions remain similar to those shown here . for each sequence
we find the relative error in computing corotating and counter - rotating isco radii with respect to the numerical solution .
we perform this comparison for the analytic solution obtained through our matching procedure ( when a solution to the matching condition exists ) , for the shibata - sasaki formula ( [ ssisco ] ) and for the kerr formula ( [ kisco ] ) . in the case of the shibata - sasaki formula
we do not explicitly compute the moment @xmath164 from the numerical solution , but we follow the same approximation adopted by shibata & sasaki .
namely , we set @xmath165 , where @xmath166 is expected to take values ranging between 0 and 2 . again , following shibata & sasaki , we normally set @xmath167 ( unless otherwise noted ) . fig .
[ 14e ] shows the relative errors in computing the iscos for the sequence having @xmath107 in the non rotating limit . in this and in the following figures , negative values of @xmath97
correspond to counterrotating orbits , while positive values of @xmath97 correspond to corotating orbits . in the corotating case
the isco disappears at slow rotation rates , even before the analytic solution becomes valid ; therefore , for this sequence , we can only compare the accuracy in finding counter - rotating iscos .
for the kerr solution the error increases monotonically with @xmath168 , reaching @xmath169 for the fastest rotating model . on the other hand ,
the error for the shibata - sasaki formula with @xmath167 is only @xmath170 for the fastest rotating model . for this sequence
the analytic solution is initially close to the shibata - sasaki formula , but then shows a rather large error , that becomes @xmath171 for the fastest rotating model .
the explanation for this behavior is that at very large rotation rates the inclusion of the multipole moment @xmath164 is important , but this multipole moment is absent in the analytic solution .
when we omit @xmath164 in the shibata - sasaki formula ( while still keeping all other mixed terms up to order @xmath159 ) by setting @xmath172 , we obtain an error which is much closer to the error made using the analytic solution . on the other hand , including only terms up to @xmath173 in the shibata - sasaki formula gives a much smaller error , comparable to ( or better than ) the error of the formula when all orders up to @xmath159 with @xmath167 are included .
what this comparison underlines is the importance of being consistent up to a certain order in the rotation parameter .
the shibata - sasaki formula has small error when used consistently up to @xmath173 or up to @xmath159 , but a large error when only a few mixed terms up to @xmath159 are included .
the error of the shibata - sasaki formula to order @xmath159 should improve , if one would include the precise values for @xmath164 , instead of the crude estimate of @xmath167 .
the analytic solution suffers from the inconsistency that while the @xmath164 moment vanishes , it is still an exact analytic solution .
this means that mixed terms containing @xmath97 , @xmath71 and @xmath24 up to order @xmath159 are present .
it follows that , for the counterrotating iscos in fig .
[ 14e ] , the analytic solution is not as accurate as a consistent application of the shibata - sasaki formula . notice that the non - monotonic increase in the error for the counter - rotating isco with the shibata - sasaki formula at large rotation rates is a consequence of the moment @xmath164 becoming important near the mass - shedding limit : for sequences of larger mass , as the ones we examine next , @xmath164 appears to be much less important .
the comparison of the error in computing the iscos is much more favorable for the analytic solution in the case of the other two sequences we examined .
[ mme ] shows the errors for the evolutionary sequence that terminates at the maximum - mass nonrotating model . in this case
, a corotating isco exists for some models for which the analytic solution is valid .
the error made with the analytic solution is @xmath174 for the fastest rotating model , and it is consistently better than the error of the shibata - sasaki formula .
for counter - rotating orbits the error for the analytic solution is somewhat smaller than for corotating orbits ( which is expected , as the isco for corotating orbits is normally at larger radii ) .
however , the error is consistently larger than the error of the shibata - sasaki formula .
the corresponding errors for the supramassive sequence , shown in fig .
[ sme ] , are very similar to the errors for the sequence in fig .
[ mme ] .
manko _ et al . _
( 2000a ) also use the quadrupole moment @xmath71 in order to match the analytic solution to a numerical one .
however , they redefine the parameter @xmath0 as @xmath175 where @xmath176 is a new parameter , with the motivation that now @xmath176 measures the departure of the analytic solution from the kerr metric .
we find that the above redefinition is not necessary , as it does not change the solution : in other words , a solution with a given @xmath0 has a corresponding value of @xmath176 . in manko _
( 2000a ) it is not mentioned that @xmath71 can be set to the numerical value only for a limited range of the parameter @xmath0 ( or , equivalently , of @xmath176 ) .
moreover , we find that the illustrative solutions they give do _ not _ correspond to the negative branch of solutions for @xmath0 ( that , as we have seen , are those relevant for rapidly rotating stars ) but rather to the positive branch , which is closer in behavior to the kerr metric .
stute & camenzind ( 2002 ) fix the the third parameter in the analytic solution , @xmath0 , by matching the value of the metric function @xmath124 at the stellar equator .
however , this is a local quantity , so that the analytic and numerical metrics are matched only at a single point in the @xmath177 plane .
experimenting with this choice , we found that one can obtain numerical solutions that are also limited to rapidly rotating stars .
however , since the parameter @xmath0 is not fixed directly , but only indirectly , one has to solve a nonlinear equation in order to obtain @xmath0 for a given value of @xmath124 at the equator .
this procedure could lead to multiple solutions , and one has to choose the one closest to a rapidly rotating neutron star by examining other properties of the spacetime ( e.g. the higher multipole moments ) .
as we have seen , fixing @xmath71 also leads to multiple solutions , however , we find that the procedure of fixing directly three leading multipole moments ( @xmath3 , @xmath97 and @xmath71 ) and selecting the desired solution according to the value of a fourth multipole moment ( @xmath24 ) is more intuitive that fixing @xmath3 , @xmath97 and the value of a metric function a single point in the spacetime . in the matching procedure used by stute & camenzind the parameter @xmath0 is not chosen to be real .
they rather impose that @xmath0 continuously reduces to the schwarzschild value , @xmath178 , in the nonrotating limit ( @xmath179 ) .
however , the schwarzschild and kerr solutions can only ( formally ) be obtained as limiting cases of the manko solution by analytic continuation in the complex-@xmath0 plane . in a sense ,
the black hole solutions are `` isolated points '' on the pure - imaginary axis of the complex-@xmath0 plane , while solutions representing neutron star exteriors lie on the pure - real axis .
therefore , the requirement imposed by stute & camenzind violates one of the original requirements of the analytic solution ( namely , that all three parameters of the solution must be real ) .
if one follows this procedure , the resulting metric components are , in general , complex .
for complex values of @xmath0 one could , in principle , use the real parts of some quantities in order to compute an estimate for the location of the isco , as was done by stute & camenzind .
however , in such cases , even the coordinates in which the metric is expressed become complex numbers . furthermore
, additional multipole moments appear that are not present in the numerical solution , rendering the analytic solution inappropriate for describing the physical properties of a rotating neutron star . finally , an important point in the matching procedure is to use the correct correspondence between the coordinates in the analytic exterior spacetime ( [ manko ] ) and the numerical spacetime ( [ cstmetr ] ) .
stute & camenzind transformed the analytic metric to boyer - lindquist like coordinates , but these are _ not _ the coordinates used in ( [ cstmetr ] ) .
this can easily be seen when one considers that the metric ( [ cstmetr ] ) reduces to the schwarzschild metric in _ isotropic _ coordinates ( not in the usual schwarzschild coordinates ) in the nonrotating limit .
we have investigated the properties of a closed - form analytic solution for the exterior spacetime of rapidly rotating neutron stars .
we matched it to highly - accurate numerical solutions , imposing that the quadrupole moment of the numerical and analytic spacetimes be the same .
for the analytic solution we considered , such a matching condition can be satisfied only for very rapidly rotating stars .
we found that solutions belong to two branches , only one of which is a good approximation to the exterior of rapidly rotating neutron star spacetimes . in order to evaluate the accuracy of the analytic solution in describing rapidly rotating neutron stars
, we presented a comparison of the radii of iscos obtained with a ) the analytic solution , b ) the kerr metric , c ) an analytic series - expansion derived by shibata & sasaki and d ) a highly - accurate numerical code . in most cases we found that the analytic solution has an accuracy consistently better than the shibata - sasaki expansion up to @xmath163 , for corotating orbits . only for counterrotating orbits
does the higher - order shibata - sasaki expansion perform better than the analytic solution .
we have only shown direct comparisons for three constant rest - mass sequences and one representative eos ( fps ) ; however our qualitative conclusions also hold for other eoss .
the analytic solution we studied in this paper could become useful in constructing outgoing - wave boundary conditions for simulations of pulsating relativistic stars , and for the computation of quasinormal modes of oscillation as an eigenvalue problem ( a long - standing problem in relativistic astrophysics ) .
another potential application is the study of high - frequency variability in accretion disks around rapidly rotating relativistic stars .
we emphasize , however , that this analytic solution is only valid for rapidly rotating stars , contrary to previous claims in the literature . for stars of intermediate rotation rates one can use the exterior analytic solution by hartle & thorne ( 1968 ) , valid to second order in the rotation rate .
this approximate solution is determined by the three multipole moments @xmath3 , @xmath97 and @xmath71 , but higher - order multipole moments are ignored
. it would be interesting to determine whether the region in which the second - order hartle - thorne metric is valid to some accuracy , overlaps with the region in which the analytic solution considered here is valid .
such a study , along with a characterization of the spacetimes using invariant quantities ( constructed in the newman - penrose formalism ) will be reported elsewhere ( berti _ et al .
_ , in preparation ) .
we wish to thank marco bruni , john l. friedman , kostas kokkotas , mina maniopoulou , vladimir manko , masaru shibata , nail sibgatullin , frances white and leszek zdunik for useful discussions and correspondence .
we are grateful to the anonymous referee for a very careful reading of the paper and many suggestions .
this work has been supported by the eu programme improving the human research potential and the socio - economic knowledge base ( research training network contract hprn - ct-2000 - 00137 ) . &
@xmath40 & @xmath3 & @xmath43 & @xmath44 & @xmath45 & @xmath46 & @xmath23 & @xmath180 & @xmath24 & @xmath0 + & @xmath36 g @xmath37 & @xmath29 s@xmath39 & @xmath41 g @xmath42 & @xmath181 & @xmath182 & km & km & km & km@xmath49 & km@xmath51 & km@xmath52 & km + + & @xmath183 & @xmath184 & @xmath182 & @xmath185 & @xmath186 & @xmath187 & @xmath188 & @xmath188 & @xmath184 & @xmath186 & @xmath184 & @xmath182 + & @xmath189 & @xmath190 & @xmath191 & @xmath192 & @xmath193 & @xmath194 & @xmath195 & @xmath196 & @xmath197 & @xmath198 & @xmath199 & @xmath182 + & @xmath200 & @xmath201 & @xmath202 & @xmath203 & @xmath204 & @xmath205 & @xmath206 & @xmath207 & @xmath208 & @xmath209 & @xmath210 & @xmath182 + & @xmath211 & @xmath212 & @xmath213 & @xmath214 & @xmath215 & @xmath216 & @xmath217 & @xmath218 & @xmath219 & @xmath220 & @xmath221 & @xmath222 + & @xmath223 & @xmath224 & @xmath225 & @xmath226 & @xmath227 & @xmath228 & @xmath229 & @xmath230 & @xmath231 & @xmath232 & @xmath233 & @xmath234 + & @xmath235 & @xmath236 & @xmath237 & @xmath238 & @xmath239 & @xmath240 & @xmath182 & @xmath241 & @xmath242 & @xmath243 & @xmath244 & @xmath245 + & @xmath246 & @xmath247 & @xmath248 & @xmath249 & @xmath250 & @xmath251 & @xmath182 & @xmath252 & @xmath253 & @xmath254 & @xmath255 & @xmath256 + & @xmath257 & @xmath258 & @xmath259 & @xmath260 & @xmath261 & @xmath262 & @xmath182 & @xmath263 & @xmath264 & @xmath265 & @xmath266 & @xmath267 + & @xmath268 & @xmath269 & @xmath270 & @xmath271 & @xmath272 & @xmath273 & @xmath182 & @xmath274 & @xmath275 & @xmath276 & @xmath277 & @xmath278 + & @xmath279 & @xmath280 & @xmath281 & @xmath271 & @xmath282 & @xmath283 & @xmath182 & @xmath284 & @xmath285 & @xmath286 & @xmath287 & @xmath288 + + & @xmath289 & @xmath184 & @xmath182 & @xmath290 & @xmath186 & @xmath291 & @xmath292 & @xmath292 & @xmath184 & @xmath186 & @xmath184 & @xmath182 + & @xmath293 & @xmath294 & @xmath295 & @xmath296 & @xmath297 & @xmath298 & @xmath299 & @xmath300 & @xmath301 & @xmath302 & @xmath303 & @xmath182 + & @xmath304 & @xmath305 & @xmath225 & @xmath306 & @xmath307 & @xmath308 & @xmath309 & @xmath310 & @xmath311 & @xmath312 & @xmath313 & @xmath182 + & @xmath314 & @xmath315 & @xmath316 & @xmath317 & @xmath318 & @xmath319 & @xmath320 & @xmath321 & @xmath322 & @xmath323 & @xmath324 & @xmath182 + & @xmath325 & @xmath326 & @xmath327 & @xmath328 & @xmath329 & @xmath330 & @xmath331 & @xmath332 & @xmath333 & @xmath334 & @xmath335 & @xmath182 + & @xmath336 & @xmath337 & @xmath338 & @xmath339 & @xmath340 & @xmath341 & @xmath342 & @xmath343 & @xmath344 & @xmath345 & @xmath346 & @xmath347 + & @xmath348 & @xmath349 & @xmath350 & @xmath351 & @xmath352 & @xmath353 & @xmath354 & @xmath355 & @xmath356 & @xmath357 & @xmath358 & @xmath359 + & @xmath360 & @xmath361 & @xmath362 & @xmath363 & @xmath364 & @xmath365 & @xmath366 & @xmath367 & @xmath368 & @xmath369 & @xmath370 & @xmath371 + & @xmath372 & @xmath373 & @xmath374 & @xmath375 & @xmath376 & @xmath377 & @xmath378 & @xmath379 & @xmath380 & @xmath381 & @xmath382 & @xmath383 + & @xmath384 & @xmath385 & @xmath386 & @xmath387 & @xmath388 & @xmath389 & @xmath182 & @xmath390 & @xmath391 & @xmath392 & @xmath393 & @xmath394 + + & @xmath395 & @xmath396 & @xmath397 & @xmath398 & @xmath399 & @xmath400 & @xmath401 & @xmath402 & @xmath403 & @xmath404 & @xmath405 & @xmath182 + & @xmath406 & @xmath407 & @xmath408 & @xmath398 & @xmath409 & @xmath410 & @xmath411 & @xmath412 & @xmath413 & @xmath414 & @xmath415 & @xmath182 + & @xmath416 & @xmath417 & @xmath418 & @xmath419 & @xmath420 & @xmath421 & @xmath422 & @xmath423 & @xmath424 & @xmath425 & @xmath426 & @xmath182 + & @xmath427 & @xmath428 & @xmath429 & @xmath430 & @xmath431 & @xmath432 & @xmath433 & @xmath434 & @xmath435 & @xmath436 & @xmath437 & @xmath182 + & @xmath438 & @xmath439 & @xmath440 & @xmath441 & @xmath442 & @xmath443 & @xmath444 & @xmath445 & @xmath446 & @xmath447 & @xmath448 & @xmath182 + & @xmath449 & @xmath450 & @xmath451 & @xmath452 & @xmath453 & @xmath454 & @xmath455 & @xmath456 & @xmath457 & @xmath458 & @xmath459 & @xmath182 + & @xmath460 & @xmath461 & @xmath462 & @xmath463 & @xmath464 & @xmath465 & @xmath466 & @xmath467 & @xmath468 & @xmath469 & @xmath470 & @xmath471 + & @xmath472 & @xmath473 & @xmath474 & @xmath475 & @xmath476 & @xmath477 & @xmath478 & @xmath479 & @xmath480 & @xmath481 & @xmath482 & @xmath483 + & @xmath484 & @xmath485 & @xmath486 & @xmath487 & @xmath488 & @xmath489 & @xmath490 & @xmath491 & @xmath492 & @xmath493 & @xmath494 & @xmath495 + & @xmath496 & @xmath497 & @xmath498 & @xmath499 & @xmath500 & @xmath501 & @xmath502 & @xmath503 & @xmath504 & @xmath505 & @xmath506 & @xmath507 + & @xmath40 & @xmath3 & @xmath43 & @xmath44 & @xmath45 & @xmath46 & @xmath23 & @xmath180 & @xmath24 & @xmath0 + & @xmath36 g @xmath37 & @xmath29 s@xmath39 & @xmath41 g @xmath42 & @xmath181 & @xmath182 & km & km & km & km@xmath49 & km@xmath51 & km@xmath52 & km + + & @xmath508 & @xmath184 & @xmath182 & @xmath185 & @xmath186 & @xmath509 & @xmath510 & @xmath510 & @xmath184 & @xmath186 & @xmath184 & @xmath182 + & @xmath511 & @xmath512 & @xmath513 & @xmath192 & @xmath514 & @xmath515 & @xmath516 & @xmath517 & @xmath518 & @xmath519 & @xmath520 & @xmath182 + & @xmath521 & @xmath522 & @xmath523 & @xmath524 & @xmath525 & @xmath526 & @xmath527 & @xmath528 & @xmath529 & @xmath530 & @xmath531 & @xmath532 + & @xmath533 & @xmath534 & @xmath535 & @xmath214 & @xmath536 & @xmath365 & @xmath182 & @xmath537 & @xmath538 & @xmath539 & @xmath540 & @xmath541 + & @xmath542 & @xmath543 & @xmath544 & @xmath545 & @xmath546 & @xmath489 & @xmath182 & @xmath547 & @xmath548 & @xmath549 & @xmath550 & @xmath551 + & @xmath552 & @xmath553 & @xmath554 & @xmath555 & @xmath239 & @xmath556 & @xmath182 & @xmath557 & @xmath558 & @xmath559 & @xmath560 & @xmath561 + & @xmath562 & @xmath563 & @xmath271 & @xmath564 & @xmath565 & @xmath566 & @xmath182 & @xmath567 & @xmath568 & @xmath569 & @xmath570 & @xmath571 + & @xmath572 & @xmath573 & @xmath574 & @xmath575 & @xmath576 & @xmath577 & @xmath182 & @xmath578 & @xmath579 & @xmath580 & @xmath581 & @xmath582 + & @xmath583 & @xmath584 & @xmath585 & @xmath586 & @xmath587 & @xmath588 & @xmath182 & @xmath589 & @xmath590 & @xmath591 & @xmath592 & @xmath593 + & @xmath594 & @xmath595 & @xmath596 & @xmath260 & @xmath597 & @xmath598 & @xmath182 & @xmath599 & @xmath600 & @xmath601 & @xmath602 & @xmath603 + + & @xmath604 & @xmath184 & @xmath182 & @xmath605 & @xmath186 & @xmath606 & @xmath607 & @xmath607 & @xmath184 & @xmath184 & @xmath184 & @xmath182 + & @xmath608 & @xmath609 & @xmath610 & @xmath611 & @xmath612 & @xmath613 & @xmath614 & @xmath615 & @xmath616 & @xmath617 & @xmath618 & @xmath182 + & @xmath619 & @xmath620 & @xmath621 & @xmath622 & @xmath623 & @xmath205 & @xmath624 & @xmath625 & @xmath626 & @xmath627 & @xmath628 & @xmath182 + & @xmath629 & @xmath630 & @xmath631 & @xmath632 & @xmath633 & @xmath509 & @xmath634 & @xmath635 & @xmath636 & @xmath637 & @xmath638 & @xmath182 + & @xmath639 & @xmath640 & @xmath641 & @xmath642 & @xmath643 & @xmath644 & @xmath645 & @xmath646 & @xmath647 & @xmath648 & @xmath649 & @xmath182 + & @xmath650 & @xmath651 & @xmath652 & @xmath653 & @xmath565 & @xmath654 & @xmath655 & @xmath656 & @xmath657 & @xmath658 & @xmath659 & @xmath182 + & @xmath660 & @xmath661 & @xmath662 & @xmath663 & @xmath664 & @xmath665 & @xmath666 & @xmath667 & @xmath668 & @xmath669 & @xmath670 & @xmath182 + & @xmath671 & @xmath672 & @xmath673 & @xmath674 & @xmath675 & @xmath676 & @xmath677 & @xmath678 & @xmath679 & @xmath680 & @xmath681 & @xmath682 + & @xmath683 & @xmath684 & @xmath685 & @xmath686 & @xmath687 & @xmath688 & @xmath689 & @xmath690 & @xmath691 & @xmath692 & @xmath693 & @xmath694 + & @xmath695 & @xmath515 & @xmath696 & @xmath697 & @xmath698 & @xmath598 & @xmath699 & @xmath700 & @xmath701 & @xmath702 & @xmath703 & @xmath704 + + & @xmath705 & @xmath706 & @xmath707 & @xmath708 & @xmath709 & @xmath503 & @xmath710 & @xmath711 & @xmath712 & @xmath713 & @xmath714 & @xmath182 + & @xmath715 & @xmath716 & @xmath717 & @xmath708 & @xmath718 & @xmath672 & @xmath719 & @xmath720 & @xmath721 & @xmath722 & @xmath723 & @xmath182 + & @xmath724 & @xmath725 & @xmath726 & @xmath727 & @xmath728 & @xmath729 & @xmath730 & @xmath731 & @xmath732 & @xmath733 & @xmath734 & @xmath182 + & @xmath735 & @xmath736 & @xmath737 & @xmath738 & @xmath739 & @xmath740 & @xmath741 & @xmath742 & @xmath743 & @xmath744 & @xmath745 & @xmath182 + & @xmath746 & @xmath747 & @xmath748 & @xmath749 & @xmath750 & @xmath751 & @xmath752 & @xmath753 & @xmath754 & @xmath755 & @xmath756 & @xmath182 + & @xmath757 & @xmath758 & @xmath759 & @xmath760 & @xmath761 & @xmath762 & @xmath763 & @xmath764 & @xmath765 & @xmath766 & @xmath767 & @xmath182 + & @xmath768 & @xmath769 & @xmath770 & @xmath771 & @xmath772 & @xmath773 & @xmath774 & @xmath775 & @xmath776 & @xmath777 & @xmath778 & @xmath182 + & @xmath779 & @xmath780 & @xmath781 & @xmath782 & @xmath783 & @xmath784 & @xmath785 & @xmath786 & @xmath787 & @xmath788 & @xmath789 & @xmath182 + & @xmath790 & @xmath791 & @xmath792 & @xmath793 & @xmath794 & @xmath795 & @xmath796 & @xmath786 & @xmath797 & @xmath798 & @xmath799 & @xmath800 + & @xmath801 & @xmath802 & @xmath803 & @xmath804 & @xmath805 & @xmath806 & @xmath807 & @xmath808 & @xmath809 & @xmath810 & @xmath811 & @xmath812 + & @xmath40 & @xmath3 & @xmath43 & @xmath44 & @xmath45 & @xmath46 & @xmath23 & @xmath180 & @xmath24 & @xmath0 + & @xmath36 g @xmath37 & @xmath29 s@xmath39 & @xmath41 g @xmath42 & @xmath181 & @xmath182 & km & km & km & km@xmath49 & km@xmath51 & km@xmath52 & km + + & @xmath813 & @xmath184 & @xmath182 & @xmath185 & @xmath186 & @xmath814 & @xmath815 & @xmath815 & @xmath184 & @xmath184 & @xmath184 & @xmath182 + & @xmath816 & @xmath817 & @xmath818 & @xmath819 & @xmath514 & @xmath762 & @xmath820 & @xmath696 & @xmath821 & @xmath822 & @xmath823 & @xmath182 + & @xmath824 & @xmath825 & @xmath248 & @xmath826 & @xmath827 & @xmath828 & @xmath182 & @xmath829 & @xmath830 & @xmath831 & @xmath832 & @xmath833 + & @xmath834 & @xmath835 & @xmath836 & @xmath837 & @xmath838 & @xmath839 & @xmath182 & @xmath840 & @xmath841 & @xmath842 & @xmath843 & @xmath844 + & @xmath845 & @xmath846 & @xmath847 & @xmath214 & @xmath848 & @xmath849 & @xmath182 & @xmath850 & @xmath851 & @xmath852 & @xmath853 & @xmath854 + & @xmath855 & @xmath856 & @xmath857 & @xmath226 & @xmath858 & @xmath859 & @xmath182 & @xmath860 & @xmath861 & @xmath862 & @xmath863 & @xmath864 + & @xmath865 & @xmath866 & @xmath867 & @xmath555 & @xmath868 & @xmath869 & @xmath182 & @xmath870 & @xmath871 & @xmath872 & @xmath873 & @xmath874 + & @xmath875 & @xmath876 & @xmath877 & @xmath878 & @xmath879 & @xmath880 & @xmath182 & @xmath881 & @xmath882 & @xmath883 & @xmath884 & @xmath885 + & @xmath886 & @xmath887 & @xmath888 & @xmath249 & @xmath476 & @xmath889 & @xmath182 & @xmath890 & @xmath891 & @xmath892 & @xmath893 & @xmath894 + & @xmath895 & @xmath896 & @xmath897 & @xmath898 & @xmath899 & @xmath900 & @xmath182 & @xmath901 & @xmath902 & @xmath903 & @xmath904 & @xmath905 + + & @xmath906 & @xmath184 & @xmath182 & @xmath907 & @xmath186 & @xmath908 & @xmath909 & @xmath909 & @xmath184 & @xmath184 & @xmath184 & @xmath182 + & @xmath910 & @xmath911 & @xmath912 & @xmath913 & @xmath914 & @xmath915 & @xmath916 & @xmath917 & @xmath918 & @xmath919 & @xmath920 & @xmath182 + & @xmath921 & @xmath922 & @xmath923 & @xmath924 & @xmath925 & @xmath926 & @xmath927 & @xmath928 & @xmath929 & @xmath930 & @xmath931 & @xmath182 + & @xmath932 & @xmath933 & @xmath934 & @xmath935 & @xmath936 & @xmath937 & @xmath903 & @xmath938 & @xmath939 & @xmath940 & @xmath941 & @xmath182 + & @xmath942 & @xmath943 & @xmath944 & @xmath945 & @xmath946 & @xmath947 & @xmath948 & @xmath949 & @xmath950 & @xmath951 & @xmath952 & @xmath182 + & @xmath953 & @xmath954 & @xmath955 & @xmath956 & @xmath957 & @xmath251 & @xmath958 & @xmath959 & @xmath960 & @xmath961 & @xmath962 & @xmath963 + & @xmath964 & @xmath965 & @xmath966 & @xmath967 & @xmath968 & @xmath969 & @xmath970 & @xmath971 & @xmath972 & @xmath973 & @xmath974 & @xmath975 + & @xmath976 & @xmath977 & @xmath978 & @xmath979 & @xmath980 & @xmath981 & @xmath982 & @xmath412 & @xmath983 & @xmath984 & @xmath985 & @xmath986 + & @xmath987 & @xmath988 & @xmath989 & @xmath990 & @xmath991 & @xmath992 & @xmath993 & @xmath994 & @xmath995 & @xmath996 & @xmath997 & @xmath998 + & @xmath999 & @xmath1000 & @xmath1001 & @xmath1002 & @xmath1003 & @xmath1004 & @xmath182 & @xmath1005 & @xmath1006 & @xmath1007 & @xmath1008 & @xmath1009 + + & @xmath1010 & @xmath1011 & @xmath1012 & @xmath1013 & @xmath1014 & @xmath1015 & @xmath1016 & @xmath1017 & @xmath1018 & @xmath1019 & @xmath1020 & @xmath182 + & @xmath1021 & @xmath928 & @xmath1022 & @xmath1013 & @xmath1023 & @xmath1024 & @xmath1025 & @xmath1026 & @xmath1027 & @xmath1028 & @xmath1029 & @xmath182 + & @xmath1030 & @xmath1031 & @xmath1032 & @xmath1033 & @xmath1034 & @xmath1035 & @xmath1036 & @xmath1037 & @xmath1038 & @xmath517 & @xmath1039 & @xmath182 + & @xmath1040 & @xmath1041 & @xmath1042 & @xmath1043 & @xmath1044 & @xmath1045 & @xmath1046 & @xmath780 & @xmath1047 & @xmath1048 & @xmath871 & @xmath182 + & @xmath1049 & @xmath1050 & @xmath398 & @xmath1051 & @xmath1052 & @xmath1053 & @xmath1054 & @xmath780 & @xmath1055 & @xmath1056 & @xmath1057 & @xmath182 + & @xmath1058 & @xmath1059 & @xmath1002 & @xmath1060 & @xmath1061 & @xmath477 & @xmath1062 & @xmath1026 & @xmath1063 & @xmath1064 & @xmath1065 & @xmath182 + & @xmath1066 & @xmath1067 & @xmath1068 & @xmath1069 & @xmath1070 & @xmath665 & @xmath1071 & @xmath1072 & @xmath1073 & @xmath1074 & @xmath1075 & @xmath1076 + & @xmath1077 & @xmath1078 & @xmath1079 & @xmath1080 & @xmath1081 & @xmath1082 & @xmath1083 & @xmath762 & @xmath1084 & @xmath1085 & @xmath1086 & @xmath1087 + & @xmath1088 & @xmath1089 & @xmath1090 & @xmath1091 & @xmath1092 & @xmath1093 & @xmath1094 & @xmath1095 & @xmath1096 & @xmath1097 & @xmath1098 & @xmath1099 + & @xmath801 & @xmath1100 & @xmath1101 & @xmath1102 & @xmath1103 & @xmath1104 & @xmath182 & @xmath1105 & @xmath1106 & @xmath1107 & @xmath1008 & @xmath1108 + & @xmath40 & @xmath3 & @xmath43 & @xmath44 & @xmath45 & @xmath46 & @xmath23 & @xmath180 & @xmath24 & @xmath0 + & @xmath36 g @xmath37 & @xmath29 s@xmath39 & @xmath41 g @xmath42 & @xmath181 & @xmath182 & km & km & km & km@xmath49 & km@xmath51 & km@xmath52 & km + + & @xmath1109 & @xmath184 & @xmath182 & @xmath185 & @xmath186 & @xmath1110 & @xmath182 & @xmath182 & @xmath184 & @xmath186 & @xmath184 & @xmath182 + & @xmath1111 & @xmath1112 & @xmath1113 & @xmath819 & @xmath1114 & @xmath1115 & @xmath182 & @xmath182 & @xmath1116 & @xmath1117 & @xmath1118 & @xmath1119 + & @xmath1120 & @xmath1121 & @xmath1122 & @xmath1123 & @xmath1124 & @xmath1125 & @xmath182 & @xmath1126 & @xmath1127 & @xmath1128 & @xmath1129 & @xmath1130 + & @xmath1131 & @xmath1025 & @xmath1101 & @xmath203 & @xmath1132 & @xmath1133 & @xmath182 & @xmath1134 & @xmath1135 & @xmath1136 & @xmath1137 & @xmath1138 + & @xmath1139 & @xmath1140 & @xmath1141 & @xmath837 & @xmath1142 & @xmath1143 & @xmath182 & @xmath1144 & @xmath1145 & @xmath1146 & @xmath1147 & @xmath1148 + & @xmath1149 & @xmath1150 & @xmath940 & @xmath214 & @xmath1151 & @xmath1152 & @xmath182 & @xmath1153 & @xmath1154 & @xmath1155 & @xmath1156 & @xmath1157 + & @xmath1158 & @xmath1159 & @xmath1160 & @xmath1161 & @xmath1162 & @xmath1163 & @xmath182 & @xmath1164 & @xmath1165 & @xmath1166 & @xmath1167 & @xmath1168 + & @xmath1169 & @xmath1170 & @xmath1171 & @xmath1172 & @xmath1173 & @xmath1174 & @xmath182 & @xmath1175 & @xmath1176 & @xmath1177 & @xmath1178 & @xmath1179 + & @xmath1180 & @xmath1181 & @xmath1182 & @xmath1183 & @xmath1184 & @xmath1185 & @xmath182 & @xmath1186 & @xmath1187 & @xmath1188 & @xmath1189 & @xmath1190 + & @xmath1191 & @xmath1192 & @xmath1193 & @xmath555 & @xmath1194 & @xmath1195 & @xmath182 & @xmath898 & @xmath1196 & @xmath1197 & @xmath1198 & @xmath1199 + + & @xmath1200 & @xmath184 & @xmath182 & @xmath1201 & @xmath186 & @xmath1202 & @xmath1203 & @xmath1203 & @xmath184 & @xmath186 & @xmath184 & @xmath182 + & @xmath1204 & @xmath1205 & @xmath1206 & @xmath1207 & @xmath1208 & @xmath1209 & @xmath1210 & @xmath1211 & @xmath208 & @xmath1212 & @xmath1213 & @xmath182 + & @xmath1214 & @xmath1215 & @xmath1216 & @xmath1217 & @xmath1218 & @xmath1219 & @xmath1220 & @xmath1221 & @xmath1222 & @xmath493 & @xmath1223 & @xmath182 + & @xmath1224 & @xmath1225 & @xmath1226 & @xmath1227 & @xmath1228 & @xmath1229 & @xmath1230 & @xmath1231 & @xmath1232 & @xmath1233 & @xmath1234 & @xmath182 + & @xmath1235 & @xmath1236 & @xmath1237 & @xmath1238 & @xmath1239 & @xmath711 & @xmath1240 & @xmath1115 & @xmath1241 & @xmath1242 & @xmath1243 & @xmath182 + & @xmath1244 & @xmath1245 & @xmath1246 & @xmath1247 & @xmath1248 & @xmath1249 & @xmath1250 & @xmath711 & @xmath1251 & @xmath1252 & @xmath1253 & @xmath1254 + & @xmath1255 & @xmath1256 & @xmath1257 & @xmath1182 & @xmath1258 & @xmath1259 & @xmath1260 & @xmath1261 & @xmath1262 & @xmath1263 & @xmath1264 & @xmath1265 + & @xmath1266 & @xmath1267 & @xmath1268 & @xmath1269 & @xmath1270 & @xmath1271 & @xmath1272 & @xmath1273 & @xmath1274 & @xmath353 & @xmath1275 & @xmath1276 + & @xmath1277 & @xmath1278 & @xmath1279 & @xmath1280 & @xmath1281 & @xmath1282 & @xmath1283 & @xmath1284 & @xmath1285 & @xmath1286 & @xmath1287 & @xmath1288 + & @xmath1289 & @xmath1290 & @xmath1291 & @xmath1292 & @xmath1293 & @xmath1294 & @xmath1295 & @xmath1296 & @xmath1297 & @xmath676 & @xmath1298 & @xmath1299 + + & @xmath1300 & @xmath1301 & @xmath1302 & @xmath1303 & @xmath1304 & @xmath1305 & @xmath1306 & @xmath1307 & @xmath1308 & @xmath1309 & @xmath1310 & @xmath182 + & @xmath1311 & @xmath1312 & @xmath1313 & @xmath1303 & @xmath1314 & @xmath1315 & @xmath1316 & @xmath1317 & @xmath1318 & @xmath1319 & @xmath1320 & @xmath182 + & @xmath1321 & @xmath1322 & @xmath1323 & @xmath1324 & @xmath1325 & @xmath1326 & @xmath1327 & @xmath1328 & @xmath1329 & @xmath1330 & @xmath1331 & @xmath182 + & @xmath1332 & @xmath1333 & @xmath1334 & @xmath1335 & @xmath1336 & @xmath1337 & @xmath1338 & @xmath1328 & @xmath1339 & @xmath1340 & @xmath1341 & @xmath182 + & @xmath1342 & @xmath1343 & @xmath1344 & @xmath1345 & @xmath957 & @xmath1346 & @xmath1347 & @xmath1348 & @xmath1349 & @xmath1350 & @xmath1351 & @xmath182 + & @xmath1352 & @xmath1353 & @xmath1354 & @xmath1355 & @xmath1356 & @xmath1357 & @xmath1358 & @xmath1359 & @xmath1360 & @xmath1053 & @xmath1361 & @xmath1362 + & @xmath1363 & @xmath1364 & @xmath1365 & @xmath444 & @xmath1366 & @xmath1367 & @xmath1368 & @xmath1369 & @xmath1370 & @xmath1371 & @xmath1372 & @xmath1373 + & @xmath1374 & @xmath1375 & @xmath1376 & @xmath1377 & @xmath1378 & @xmath1379 & @xmath989 & @xmath1380 & @xmath1381 & @xmath1382 & @xmath1383 & @xmath1384 + & @xmath1385 & @xmath1386 & @xmath1387 & @xmath1388 & @xmath1389 & @xmath1174 & @xmath1390 & @xmath1391 & @xmath1392 & @xmath1393 & @xmath1394 & @xmath1395 + & @xmath1396 & @xmath1397 & @xmath1398 & @xmath1399 & @xmath1400 & @xmath1401 & @xmath1402 & @xmath1403 & @xmath1404 & @xmath1405 & @xmath1406 & @xmath1407 + & @xmath40 & @xmath3 & @xmath43 & @xmath44 & @xmath45 & @xmath46 & @xmath23 & @xmath180 & @xmath24 & @xmath0 + & @xmath36 g @xmath37 & @xmath29 s@xmath39 & @xmath41 g @xmath42 & @xmath181 & @xmath182 & km & km & km & km@xmath49 & km@xmath51 & km@xmath52 & km + + & @xmath1408 & @xmath184 & @xmath182 & @xmath1409 & @xmath186 & @xmath1410 & @xmath1411 & @xmath1412 & @xmath184 & @xmath186 & @xmath184 & @xmath182 + & @xmath1413 & @xmath1414 & @xmath1415 & @xmath1123 & @xmath1114 & @xmath1416 & @xmath182 & @xmath1417 & @xmath1418 & @xmath1419 & @xmath1420 & @xmath182 + & @xmath1421 & @xmath1422 & @xmath1423 & @xmath524 & @xmath1424 & @xmath1425 & @xmath182 & @xmath1426 & @xmath1427 & @xmath1428 & @xmath1429 & @xmath1430 + & @xmath1431 & @xmath1432 & @xmath878 & @xmath1433 & @xmath1434 & @xmath1435 & @xmath182 & @xmath1436 & @xmath1437 & @xmath249 & @xmath1438 & @xmath1439 + & @xmath1440 & @xmath1441 & @xmath1442 & @xmath837 & @xmath1443 & @xmath1444 & @xmath182 & @xmath1445 & @xmath1446 & @xmath888 & @xmath1447 & @xmath1448 + & @xmath1449 & @xmath1450 & @xmath1451 & @xmath1452 & @xmath1453 & @xmath1454 & @xmath182 & @xmath1455 & @xmath1456 & @xmath907 & @xmath1457 & @xmath1458 + & @xmath1459 & @xmath1460 & @xmath1461 & @xmath545 & @xmath1462 & @xmath1463 & @xmath182 & @xmath1464 & @xmath1465 & @xmath1466 & @xmath1467 & @xmath1468 + & @xmath1469 & @xmath1470 & @xmath1471 & @xmath555 & @xmath1472 & @xmath1473 & @xmath182 & @xmath1474 & @xmath1475 & @xmath1476 & @xmath1477 & @xmath1478 + & @xmath1479 & @xmath1480 & @xmath1481 & @xmath564 & @xmath1482 & @xmath1483 & @xmath182 & @xmath840 & @xmath1484 & @xmath1485 & @xmath1486 & @xmath1487 + & @xmath1488 & @xmath1489 & @xmath1490 & @xmath249 & @xmath1491 & @xmath1492 & @xmath182 & @xmath1493 & @xmath1494 & @xmath770 & @xmath1495 & @xmath1496 + + & @xmath1497 & @xmath184 & @xmath182 & @xmath653 & @xmath186 & @xmath1498 & @xmath1499 & @xmath1499 & @xmath184 & @xmath184 & @xmath184 & @xmath182 + & @xmath1500 & @xmath1280 & @xmath727 & @xmath1501 & @xmath1502 & @xmath729 & @xmath1503 & @xmath1037 & @xmath1504 & @xmath1505 & @xmath1506 & @xmath182 + & @xmath1507 & @xmath1508 & @xmath1509 & @xmath1510 & @xmath1511 & @xmath1512 & @xmath1513 & @xmath1514 & @xmath1515 & @xmath1516 & @xmath1517 & @xmath182 + & @xmath1518 & @xmath1519 & @xmath872 & @xmath1520 & @xmath1521 & @xmath1522 & @xmath1523 & @xmath566 & @xmath1524 & @xmath1525 & @xmath1526 & @xmath182 + & @xmath1527 & @xmath1528 & @xmath1529 & @xmath1530 & @xmath1453 & @xmath1531 & @xmath1532 & @xmath1533 & @xmath1534 & @xmath1535 & @xmath1536 & @xmath182 + & @xmath1537 & @xmath1538 & @xmath1539 & @xmath1122 & @xmath1540 & @xmath1541 & @xmath1542 & @xmath880 & @xmath1543 & @xmath1544 & @xmath1545 & @xmath182 + & @xmath1546 & @xmath1547 & @xmath1548 & @xmath1549 & @xmath1550 & @xmath1551 & @xmath1552 & @xmath1553 & @xmath1554 & @xmath1555 & @xmath1556 & @xmath182 + & @xmath1557 & @xmath1558 & @xmath1559 & @xmath1560 & @xmath1561 & @xmath1562 & @xmath1563 & @xmath1564 & @xmath1565 & @xmath1566 & @xmath1567 & @xmath182 + & @xmath999 & @xmath1568 & @xmath1569 & @xmath1570 & @xmath1571 & @xmath1572 & @xmath1573 & @xmath1574 & @xmath1575 & @xmath1576 & @xmath1577 & @xmath1578 + & @xmath1579 & @xmath1580 & @xmath1581 & @xmath243 & @xmath1582 & @xmath1583 & @xmath1584 & @xmath1585 & @xmath1586 & @xmath1587 & @xmath1588 & @xmath1589 + + & @xmath1590 & @xmath1591 & @xmath1592 & @xmath1593 & @xmath1594 & @xmath1595 & @xmath1596 & @xmath1597 & @xmath1598 & @xmath1599 & @xmath1600 & @xmath182 + & @xmath1601 & @xmath1591 & @xmath1602 & @xmath1603 & @xmath1604 & @xmath1605 & @xmath1606 & @xmath1607 & @xmath1608 & @xmath1609 & @xmath1610 & @xmath182 + & @xmath1507 & @xmath1611 & @xmath1612 & @xmath1613 & @xmath1614 & @xmath1615 & @xmath1616 & @xmath1617 & @xmath1618 & @xmath1619 & @xmath1620 & @xmath182 + & @xmath1621 & @xmath1622 & @xmath1623 & @xmath1624 & @xmath1625 & @xmath1626 & @xmath1627 & @xmath1628 & @xmath1629 & @xmath1630 & @xmath1631 & @xmath182 + & @xmath1518 & @xmath1632 & @xmath1368 & @xmath1101 & @xmath1633 & @xmath1634 & @xmath1635 & @xmath1636 & @xmath1637 & @xmath1638 & @xmath1639 & @xmath182 + & @xmath1640 & @xmath1641 & @xmath1642 & @xmath785 & @xmath1643 & @xmath1644 & @xmath1645 & @xmath1646 & @xmath1647 & @xmath1648 & @xmath1649 & @xmath182 + & @xmath1527 & @xmath1650 & @xmath1651 & @xmath1652 & @xmath750 & @xmath1653 & @xmath1654 & @xmath1655 & @xmath1656 & @xmath1657 & @xmath1658 & @xmath182 + & @xmath1537 & @xmath1659 & @xmath1660 & @xmath1661 & @xmath1662 & @xmath1663 & @xmath1664 & @xmath1665 & @xmath1666 & @xmath1667 & @xmath1668 & @xmath182 + & @xmath1546 & @xmath1669 & @xmath1670 & @xmath1671 & @xmath1672 & @xmath1673 & @xmath1674 & @xmath1675 & @xmath1676 & @xmath1677 & @xmath1678 & @xmath182 + & @xmath1557 & @xmath1679 & @xmath1680 & @xmath1681 & @xmath1682 & @xmath1683 & @xmath1684 & @xmath1685 & @xmath1686 & @xmath1687 & @xmath1688 & @xmath1689 + @xmath104 + & a(1.44 ) & @xmath1690 & @xmath1691 + & a(mm ) & @xmath1692 & @xmath1693 + & fps(1.44 ) & @xmath1694 & @xmath1695 + & fps(mm ) & @xmath1692 & @xmath1693 + & l(1.44 ) & @xmath1696 & @xmath1697 + & l(mm ) & @xmath1698 & @xmath1699 + @xmath1700 + & a(1.44 ) & @xmath1701 & @xmath1702 + & a(mm ) & @xmath1703 & @xmath1704 + & a(sm ) & @xmath1704 & @xmath1702 + & au(1.44 ) & @xmath1701 & @xmath1703 + & au(mm ) & @xmath1705 & @xmath1706 + & au(sm ) & @xmath1701 & @xmath1707 + & fps(1.44 ) & @xmath1701 & @xmath1702 + & fps(mm ) & @xmath1708 & @xmath1704 + & fps(sm ) & @xmath1706 & @xmath1702 + & l(1.44 ) & @xmath1709 & @xmath1702 + & l(mm ) & @xmath1708 & @xmath1704 + & l(sm ) & @xmath1710 & @xmath1702 + & aprb(1.4 ) & @xmath1711 & @xmath1712 + & aprb(mm ) & @xmath1713 & @xmath1712 + & aprb(sm ) & @xmath1701 & @xmath1701 +
|
we investigate the properties of a closed - form analytic solution recently found by manko _
et al . _
( 2000b ) for the exterior spacetime of rapidly rotating neutron stars .
for selected equations of state we numerically solve the full einstein equations to determine the neutron star spacetime along constant rest mass sequences .
the analytic solution is then matched to the numerical solutions by imposing the condition that the quadrupole moment of the numerical and analytic spacetimes be the same . for the analytic solution we consider , such a matching condition can be satisfied only for very rapidly rotating stars . when solutions to the matching condition exist , they belong to one of two branches . for one branch the current octupole moment of the analytic solution
is very close to the current octupole moment of the numerical spacetime ; the other branch is more similar to the kerr solution .
we present an extensive comparison of the radii of innermost stable circular orbits ( iscos ) obtained with a ) the analytic solution , b ) the kerr metric , c ) an analytic series expansion derived by shibata and sasaki ( 1998 ) and d ) a highly accurate numerical code . in most cases
where a corotating isco exists , the analytic solution has an accuracy consistently better than the shibata - sasaki expansion .
the numerical code is used for tabulating the mass - quadrupole and current - octupole moments for several sequences of constant rest mass .
gravitation relativity stars : rotation stars : neutron
|
classical cepheids , also known as @xmath1 cep - type stars or just cepheids , are population i instability strip pulsators crucial for understanding stellar structure and evolution .
most galactic cepheids pulsate in one mode with remarkably stable amplitudes and periods , although space - based photometry reveals that they are not entirely regular clocks .
observations with _
corot , kepler _ and _ most _ detect short - term , low - amplitude irregular variability or `` jitter '' in the pulsation cycles of some classical cepheids @xcite .
one exception is v473 lyrae ( @xmath2 = 19@xmath315@xmath45949 , @xmath5 = + 2755346 , @xmath6 = 6.18 mag ) , the only known classical cepheid in the galaxy that undergoes strong amplitude and phase modulations .
the changes in the pulsation amplitude were first noted by @xcite .
@xcite estimated the modulation period to be 3.33.4 years , which was later measured to be 1210 d by @xcite , and 1258 d by @xcite , respectively .
the origin of this modulation was extensively discussed in the literature , with theories ranging from mode beating @xcite to magneto - convective cycles @xcite .
a recent analysis of extensive ground - based observations collected during 19662011 @xcite confirmed that v473 lyrae pulsates in the second overtone , modulated with a primary period of @xmath7 d. @xcite also discovered a second , longer modulation cycle with a period of @xmath8 d , and concluded that the modulation closely resembles the blazhko effect seen in rr lyrae stars .
v473 lyrae is unique among galactic cepheids : the closest counterparts are v1154 cyg , in which the _ kepler _ space telescope detected modulation but with a very low amplitude @xcite and sv vul , with a suspected , much longer modulation period of order 30 yr @xcite . beyond our galaxy ,
strong amplitude modulation was recently found in three classical cepheids in the magellanic clouds @xcite .
although modulation appears to be very rare among cepheids , the phenomenological properties of the light variations of v473 lyr closely resemble the blazhko effect , a common feature among rr lyrae stars .
thanks to space - based photometry , we now know that many modulated rr lyrae stars show low - amplitude additional modes and/or alternating high- and low - amplitude cycles known as period doubling @xcite .
numerical models suggest that mode resonances that generate the period doubling can also be responsible for the modulation itself @xcite .
it is not unreasonable , therefore , to postulate that similar mechanisms could operate in modulated cepheids too .
unfortunately , the pulsation period of v473 lyr ,
@xmath9 = 1.4909d , means that one can not observe consecutive pulsation maxima from a single ground - based site .
this makes it virtually impossible to detect period doubling , a challenge also faced in ground - based rr lyrae observations .
therefore , we proposed to gather continuous observations of the star with the _ most _ space telescope for a timespan of one month in order to search for cycle - to - cycle variations .
_ most _ serendipitously also observed another cepheid , u tra , just before the run that targeted v473 lyr . a campaign aimed at the wolf - rayet star wr 71 included this beat cepheid within the same field of view .
u tra ( @xmath2 = 16@xmath307@xmath41901 , @xmath5 = 6254380 , @xmath6 = 7.89 mag ) , along with tu cas were the first cepheids in which beating ( the characteristic feature caused by double - mode pulsation ) was observed @xcite .
u tra pulsates in the fundamental mode and the first overtone , with periods @xmath10d and @xmath11d @xcite . after an investigation of the relative amplitudes of the pulsation
modes over time , @xcite proposed that pulsation energy may have been redistributed from the fundamental mode to the first overtone , an intriguing concept that can be tested with new measurements . together with earlier observations of rt aur and sz tau @xcite ,
_ most _ s cepheid sample now totals 4 stars .
this paper is structured as follows : section [ sec2 ] describes the various observations we carried out ; sections [ secv473 ] and [ secutra ] include the analysis and obtained results for v473 lyr and u tra , respectively ; section [ secsum ] closes with discussion and summary of the findings presented in the paper .
the stars were observed by the _ most _ ( microvariability & oscillations of stars ) space telescope : a canadian microsatellite designed to observe bright stellar targets continuously for up to 2 months at a time @xcite .
we also obtained spectroscopy for both stars with various ground - based instruments . in the case of u tra , in the field of the _
most _ primary science target wr 71 , each orbit was shared with another primary science field .
u tra was observed for 28 days during 29 may 25 june 2014 .
as _ most _ has to use the science ccd for guiding purposes too , it takes short individual exposures that are stacked on - board to boost the stellar signals .
the sampling cadence is 123.8 sec . in the case of v473 lyrae ,
the star is located in the sky where _ most _ must point at too large an angle relative to the sun , heating the satellite beyond the thermal limits for its battery .
but the v473 lyrae run coincided with the eclipse season " for _ most _ , when the sun skims briefly below the earth horizon once per orbit for a few minutes .
it was possible to point _ most _ at v473 lyrae for a few minutes per orbit , leading to an effective sampling cadence of 101 minutes .
each orbit contains 210 measurements , with a cadence of 62.2 sec .
the star was observed for 27 days during 28 june
23 july 2014 , but communication problems with the satellite and deliberate interruptions to observe other targets introduced some longer gaps .
we have data from 57 per cent of the _ most _ orbits in the overall run .
most photometric reduction processes are well established ( see , e.g. , * ? ? ?
* ; * ? ? ?
* ) and have recently been successfully applied to other cepheids @xcite .
_ most _ data experience modulation of scattered earthshine with the 101-min orbital period .
also , as the instrument regains thermal equilibrium after switching from two widely - separated points in the sky , there are short - lived trends after switching from one field to another .
stray light modulation is filtered in the reduction process ; extreme outliers were removed by sigma - clipping the flux values . in instances of substantially poor pointing , data were filtered by sigma - clipping outliers in the pointing positions and the measured flux values . to preserve the intrinsic cepheid variations ,
slow trends were filtered from the data and clipping applied to the residuals .
the filtered data were phased with the orbital period and averaged in bins to remove modulations caused by stray light .
we note that v473 lyr is much less affected by stray light than u tra as it was only observed when _ most _ was in the shadow of the earth .
finally , the large - amplitude variation was restored to the data .
any remaining slow trends were then subtracted by a simple linear fit .
the final light curves ( times , magnitudes and errors ) are shown in fig .
[ most - lc ] .
@xmath12konkoly observatory , mta csfk , konkoly thege mikls t 15 - 17 , h-1121 budapest , hungary + @xmath13elte gothard astrophysical observatory , h-9704 szombathely , szent imre herceg t 112 , hungary + @xmath14department of physics and astronomy , university of british columbia , 6224 agricultural road , vancouver , bc v6 t 1z1 , canada + @xmath15department of mathematics , physics & geology , cape breton university , 1250 grand lake road , sydney , nova scotia , canada , b1p
6l2 + @xmath16canadian coast guard college , dept . of arts , sciences , and languages ,
sydney , nova scotia , b1r 2j6 , canada + @xmath17dpartement de physique and centre de recherche en astrophysique du qubec ( craq ) , universit de montral , c.p .
6128 , succ .
centre - ville , montral , qubec , h3c 3j7 , canada + @xmath18ritter observatory , department of physics and astronomy , university of toledo , toledo , oh 43606 - 3390 , usa + @xmath19department of physical sciences , kutztown university , kutztown , pa 19530 ,
usa + @xmath20barfold observatory , glenhope , victoria 3444 , australia + @xmath21saser the southern astro spectroscopy email ring + @xmath22international centre for radio astronomy research , university of western australia , 35 stirling hwy , crawley , wa 6009 , australia + @xmath23harvard - smithsonian center for astrophysics , 60 , garden street , cambridge ma 02138 , usa + @xmath24instituut voor sterrenkunde , celestijnenlaan 200d , b-3001 leuven , belgium + @xmath25department of physics , university of antwerp , groenenborgerlaan 171 , 2020 antwerp , belgium + @xmath26department of physics and astronomy , university of louisville , louisville , ky 40292 , usa + @xmath27department of physics and astronomy , vanderbilt university , nashville , tn 37235 , usa + @xmath28department of physics , lehigh university , bethlehem , pa 18015 , usa + @xmath29department of physics , fisk university , 1000 17th avenue north , nashville , tn 37208 , usa + @xmath30las cumbres obs
. global telescope network , 6740 cortona dr .
, suite 102 , santa barbara , ca 93117 , usa + @xmath31american association of variable star observers , cambridge , ma 02138 , usa + @xmath32center for theoretical physics of the polish academy of sciences , al .
lotnikow 32/46 , 02 - 668 warsaw , poland + @xmath33faculty of physics , university of warsaw , pasteura 5 , 02 - 093 warszawa , poland + @xmath34national centre for nuclear research , hoa 69 , 00 - 681 warsaw , poland + @xmath35international centre for radio astronomy research ( icrar ) , curtin university , gpo box u1987 , perth , wa 6845 , australia + @xmath36institute for computational astrophysics , department of astronomy and physics , saint mary s university , halifax , ns b3h 3c3 , canada + @xmath37institut fr astronomie , universitt wien , trkenschanzstrasse 17 , a-1180 wien , austria + @xmath38department of astronomy and astrophysics , university of toronto , toronto , on m5s 3h4 , canada + [ lastpage ]
|
space - based photometric measurements first revealed low - amplitude irregularities in the pulsations of cepheid stars , but their origins and how commonly they occur remain uncertain . to investigate this phenomenon , we present _ most _ space telescope photometry of two cepheids .
v473 lyrae is a second - overtone , strongly modulated cepheid , while u trianguli australis is a cepheid pulsating simultaneously in the fundamental mode and first overtone .
the nearly continuous , high - precision photometry reveals alternations in the amplitudes of cycles in v473 lyr , the first case of period doubling detected in a classical cepheid . in u tra , we tentatively identify one peak as the @xmath0 or 0.61type mode often seen in conjunction with the first radial overtone in cepheids , but given the short length of the data , we can not rule out that it is a combination peak instead .
ground - based photometry and spectroscopy were obtained to follow two modulation cycles in v473 lyr and to better specify its physical parameters .
the simultaneous data yield the phase lag parameter ( the phase difference between maxima in luminosity and radial velocity ) of a second - overtone cepheid for the first time .
we find no evidence for a period change in u tra or an energy exchange between the fundamental mode and the first overtone during the last 50 years , contrary to earlier indications .
period doubling in v473 lyr provides a strong argument that mode interactions do occur in some cepheids and we may hypothesise that it could be behind the amplitude modulation , as recently proposed for blazhko rr lyrae stars .
[ firstpage ] stars : variables : cepheids stars : individual : v473 lyrae stars : individual : u tra
|
the study of dirac points , i.e. the contact points between different energy bands with an approximate linear dispersion relation , has become a major issue since the experimental breakthrough in graphene - based electronics @xcite .
indeed , the low - energy electronic properties of graphene are governed by a pseudo - relativistic 2d dirac equation for massless fermions situated at the @xmath0 and @xmath1 corners of the brillouin zone @xcite .
the dirac points are topologically protected and a gap is opened only when the inversion symmetry of the lattice or the time - reversal symmetry are broken . the possibility to generate topological phase transitions in graphene - like systems has recently attracted a great deal of attention . within a tight - binding description , an anisotropy in the nearest - neighbor hopping parameters
makes the dirac points move away from the high - symmetry @xmath0 and @xmath1 points and , under appropriate conditions , merge at time - reversal invariant points in the first brillouin zone @xcite .
most saliently , this merging of dirac points is associated with a topological phase transition between a semimetallic phase and a gapped band - insulating phase .
an experimental investigation of the merging transition in graphene turns out to be problematic , since in order to appropriately modify the hopping parameters , an unphysically large strain needs to be applied to the graphene sheet @xcite .
an alternative system for the study of such topological transitions is that of ultracold atoms trapped in a honeycomb optical lattice . since the seminal realization of the superfluid - mott - insulator transition in the bose - hubbard model , ultracold atoms in optical lattices
have become promising systems to emulate condensed - matter physics .
indeed , the lattice geometry , the dimensionality , the atomic species , as well as the interactions can be engineered with a high degree of precision .
the more involved triangular and honeycomb geometries were recently realized experimentally and exotic correlated states of matter have been observed experimentally @xcite or predicted theoretically @xcite .
the application of a time - periodic perturbation on the optical lattice introduces yet another parameter scale into the system .
a periodic shaking of the optical lattice , up to the khz frequency range , has been implemented by placing one of the mirrors used to create the optical lattice on a piezoelectric material , such that the mirror can be moved back and forth in the direction of the beam @xcite .
the floquet formalism shows that the hopping energy of the atoms in the shaken lattice is renormalized by a bessel function , as a function of the shaking frequency and amplitude , thus allowing both the magnitude and the sign of the hopping parameter to change .
this rather counter - intuitive phenomenon , as compared to the standard tight - binding physics , has been experimentally observed in a one - dimensional cold - atomic system @xcite . in this paper
, we consider ultracold fermions trapped in a _ shaken _ honeycomb optical lattice . within the floquet formalism , we derive an effective hamiltonian that generalizes that of a graphene - like material under strain . in particular , we find that the alignment and merging of dirac points in momentum space are now accessible with ultracold fermions in the shaken optical lattice and the phase diagram consists of various phases of the corresponding solid - state system that are otherwise difficult to realize . furthermore , by taking into account a hubbard - like interaction for spinful fermions , we study the density profiles for the homogenous and the trapped gas within a hartree - fock theory .
the outline of this paper is the following : in sec.[sec01a ] we introduce the time - dependent hamiltonian and in sec.[sec01b ] we derive the time - independent one , by applying the floquet formalism . in sec.[sec02 ]
we investigate the merging and alignment of dirac points , when the optical lattice is shaken along specific directions .
the description is extended to include interactions in sec.[sec03 ] , where we derive the dependence of the density on the chemical potential .
implications of our results for experiments are discussed in sec.[sec04 ] .
finally , our conclusions are presented in sec.[conc ] .
in this section , we derive a time - independent effective description for ultracold atoms trapped in a periodically shaken honeycomb optical lattice by utilizing a floquet theory . for simplicity , we focus on a system of single - component fermionic atoms and consider only single - particle terms in this section . the results from the floquet theory are valid for fermionic atoms with internal degrees of freedom as well as for bosonic atoms . in particular
, the hyperfine state of fermionic atoms , playing the role of an effective spin-@xmath2 degree of freedom for electrons , will be considered when interaction effects are taken into account in sec .
iv . in the tight - binding limit ,
the system of ultracold fermionic atoms trapped in a 2d _ shaken _ honeycomb optical lattice can be described by the hamiltonian @xmath3 which consists of two distinct parts .
the static part @xmath4 is simply the tight - binding hamiltonian in the honeycomb lattice , where @xmath5 ( @xmath6 ) and @xmath7 ( @xmath8 ) are , respectively , fermionic creation and annihilation operators on the lattice site @xmath9 in the @xmath10 ( @xmath11 ) sublattice .
the three vectors @xmath12 connect an @xmath10-lattice site with its three nearest - neighbor ( nn ) @xmath11-lattice sites and are given in terms of the distance @xmath13 between nn sites , where @xmath14 is the laser wave number ( see fig.[fig : lattice ] ) . here
, @xmath15 characterize the energy gained in hopping to the nn and next - nearest - neighbor ( nnn ) sites , respectively , and @xmath16 is the on - site energy .
we remark that the nnn hopping is taken into account because the nn hopping may be rendered vanishingly small in the effective time - independent description . in this regime , the nnn may become the dominant kinetic term . in a square lattice , where the potential is separable in independent @xmath17 and @xmath18 components ,
the nnn hopping is identically zero @xcite .
however , the nnn hopping can be nonzero in the honeycomb lattice , since its potential is not separable in @xmath17 and @xmath18 components .
nevertheless , it may be expressed as the sum of two triangular lattices . , red dots , and @xmath11 , blue dots ) .
the vectors @xmath19 , @xmath20 , and @xmath21 connect a site on the @xmath10 sublattice to its nearest neighbors on the @xmath11 sublattice . ]
the time - dependent part of the hamiltonian ( [ basich ] ) , @xmath22 describes the harmonic shaking of the lattice in the direction @xmath23 with a driving frequency @xmath24 in the co - moving frame of ref .
@xcite . as a consequence of the transformation to the co - moving frame
, @xmath25 describes atoms of mass @xmath26 experiencing a position - dependent sinusoidal force .
the unavoidable complexity that arises when dealing with a quantum many - body system out of equilibrium has recently motivated the development of new theoretical tools , for example time - dependent density matrix renormalization group @xcite , time - dependent dynamical mean field theory @xcite , and exact diagonalization @xcite
. however , for a periodically driven quantum system , the floquet theory offers a simplified description of the system , in the form of a time - independent effective hamiltonian , if the period @xmath27 is the shortest time scale in the problem @xcite . in this limit , the atoms can not follow the shaking motion adiabatically and remain thus at their average lattice position , albeit with renormalized hopping parameters .
the system is thus considered to be in a stationary state and the knowledge of equilibrium physics can be employed .
let us consider the floquet hamiltonian defined by @xmath28 , where @xmath29 is periodic in time @xcite .
the eigenvalue equation is then given by @xmath30 where @xmath31 is the quasienergy defined uniquely up to a multiple of @xmath32 .
any solution @xmath33 is part of a set of solutions @xmath34 with integer @xmath35 , which all correspond to the same physical solution .
hence , the spectrum of the floquet hamiltonian possesses a brillouin - zone - like structure @xcite .
the interest therefore lies with the states in the first brillouin zone , i.e. states with quasi - energies @xmath36 .
the space in which the states @xmath33 are defined is the composite of the hilbert space spanned by square integrable functions on configuration space , @xmath37 , and the space of @xmath38-periodic functions .
the state @xmath33 may be written down in an orthonormal basis in the composite space according to @xmath39 |\alpha ( q ) \rangle,\ ] ] where @xmath40 are coefficients to normalize @xmath33 and the operator @xmath41 can be any @xmath38-periodic hermitian operator .
therefore , we can conveniently choose @xmath41 to be @xmath42 , such that @xmath43 . if the condition @xmath44 \ , \exp [ i ( n - n ' ) \omega t ] \ , h_0 \exp[-i \hat{f}(t ) ] \rangle_t | \ , \alpha ( q ) \rangle \ll \hbar \omega \label{eq : condition}\ ] ] is satisfied for any two states @xmath45 and @xmath46 , then the eigenvalues @xmath47 are approximately @xmath48 \ , h_0 \ , \exp[-i \hat{f}(t ) ] \rangle_t \,| \,\alpha ( q ) \rangle .
\label{eq : floquet - approximation}\ ] ] here , @xmath49 denotes the time average of the operator @xmath50 over the period @xmath38
. the condition ( @xmath51 ) will hold for @xmath52 , if @xmath53 is nearly constant during the period @xmath38 , which is small if @xmath24 is large . in this case , states with different @xmath35 do not mix . if @xmath24 is large enough , such that the condition ( @xmath51 ) also holds for @xmath54 , then the energy spectrum will split up into energy bands labelled by an index @xmath35 , where the details within the energy band are determined by @xmath53 . because the states with a different index @xmath35 are separated by an energy which is a multiple of @xmath32 and because the spectrum possesses a brillouin - zone - like structure ,
only the terms with @xmath55 need to be taken into account .
the effective hamiltonian @xmath56 , which gives rise to the same spectrum as the floquet hamiltonian , is then defined by @xcite @xmath57 \ , h_{0 } \
, \exp[-i \hat{f}(t ) ] \bigg>_t \nonumber \\ & = \bigg < \sum^\infty_{n=0 } \frac{i^n}{n ! }
[ \hat{f}(t ) , h_{0}]_n \bigg>_t .
\label{eq : effective - ham}\end{aligned}\ ] ] here , @xmath58_n$ ] denotes the multiple commutator , which is defined by @xmath58_{n+1 } = [ \hat{f},[\hat{f},\hat{g}]_n]$ ] and @xmath58_0 = \hat{g}$ ] .
effective hamiltonians corresponding to eq.([eq : effective - ham ] ) have been derived for linear shaking of a one - dimensional lattice @xcite and for elliptical shaking of a triangular lattice @xcite . for the shaken honeycomb lattice studied here ,
the condition ( @xmath51 ) is satisfied if @xmath59 , and the effective hamiltonian becomes @xmath60 where the renormalized nn hopping parameters @xmath61 are given by @xmath62 and the renormalized nnn hopping parameters are given by @xmath63 ( see appendix [ app1 ] for detailed calculations ) . in these expressions
, @xmath64 denotes the zeroth order bessel function of the first kind , which shows a damped oscillation around zero . in terms of the renormalized nn and nnn hopping parameters , the diaginalization of the effective hamiltonian ( [ hamiltonian - eff ] ) yields the dispersion relation @xmath65 where @xmath66 is the band index , and we have defined the functions @xmath67 and @xmath68.\ ] ]
in this section , the honeycomb lattice with anisotropic hopping is studied . in the first two subsections , only nn hopping is considered for illustration reasons .
indeed , this allows for a simple understanding of the main consequences of shaking on dirac - point motion and dimensional crossover . in subsec .
[ sec02c ] , we discuss how the picture evolves when nnn hopping is included , and the sign of the hopping parameters is investigated in subsec .
[ sec02d ] .
since the system with two nn equal hopping parameters and a single independent one captures the essential features of the systems with three independent nn hopping parameters , we will focus on this system .
the numbering of the @xmath61s is chosen such that @xmath69 , which can be achieved by shaking in a direction parallel or perpendicular to @xmath19 .
although the atoms in the optical lattice are charge - neutral objects , we shall adopt the language from condensed - matter physics and call a zero - gap phase with a pair of dirac cones and a vanishing density of states at the band - contact points a _ semimetal _ , whereas a gapped phase is called _
band insulator_. furthermore , nnn hopping induces a _ metallic _
phase for small values of @xmath61 because of an overlap between the two bands that yields a non - vanishing density of states at the energy level of the band - contact points .
if the latice is shaken in the direction perpendicular to @xmath19 ( direction 1 in fig.[fig : lattice ] ) , @xmath70 remains equal to @xmath71 , whereas @xmath72 and @xmath73 are renormalised to a smaller value .
an increase in the shaking amplitude results in a decrease in @xmath74 , which is depicted by the arrow @xmath75 in fig.[fig : phase - diagram - abs](a ) .
when the hopping parameters change according to this arrow @xmath75 , the energy spectrum evolves from fig.[fig : merging - energy - dispersion](a ) to fig.[fig : merging - energy - dispersion](b ) .
the dirac points , originally situated at the corners @xmath0 and @xmath1 of the first brillouin zone , start to move in the @xmath76-direction along the vertical edges of the latter .
this motion is depicted by the arrows in fig.[fig : phase - diagram - abs](b ) . even if the two dirac points are no longer located at the high - symmetry points @xmath0 and @xmath1 , they remain related by time - reversal symmetry , such that their berry phases @xmath77 and @xmath78 are opposite .
this non - zero berry phase topologically protects each of the dirac points and thus the semimetallic phase remains robust until @xmath79 , where the two points merge at a time - reversal invariant momentum , i.e. half of a reciprocal lattice vector @xcite . in the present example , this point is situated at the center of the vertical edges of the first brillouin zone , and the band dispersion becomes parabolic in the @xmath80-direction while remaining linear in the @xmath81-direction [ see fig.[fig : merging - energy - dispersion](b ) ] .
the merged dirac points are no longer topologically protected due to the annihilation of the opposite berry phases .
consequently , a further increase of the shaking amplitude , which results in a further decrease of @xmath82 , leads to the opening of a gap between the two bands .
thus , the system undergoes a topological phase transition from a semimetal to a band insulator .
this merging transition was also studied in a static setup in ref .
@xcite , where the hopping amplitudes @xmath71 were proposed to be modified by a change in the intensity of one of the lasers used to create the optical lattice .
in contrast to this static setup , shaking the honeycomb lattice allows one to completely annihilate some of the nn hopping parameters and to even change their sign .
this sign change occurs at the zeros of the bessel function [ see eq.([eq : renormalized - gamma ] ) ] . for an example system of @xmath83k atoms in a lattice created by lasers with a wavelength of 830 nm , which is shaken in the direction perpendicular to @xmath19 , the situation @xmath84 is encountered for @xmath85 which corresponds to the first zero of the bessel function . at this particular point , and if @xmath86 in addition , the system consists of a set of effectively decoupled horizontal bonds along which the atoms are solely allowed to hop .
this yields two flat bands at @xmath87 [ see fig.[fig : merging - energy - dispersion](c ) ] that may be viewed as the extreme limit of the band - insulating phase .
alternatively , one may view this situation upon decreasing the value of @xmath82 as a dimensional crossover from a 2d band insulator to a zero - dimensional ( 0d ) system .
a small non - zero value of @xmath82 simply provides a weak dispersion of these decoupled bands ( not shown ) . , @xmath88 , and @xmath86 . the labels @xmath89 and @xmath76 represent the @xmath81 and @xmath80 components of the momentum , respectively
the @xmath81 and @xmath80 axes have been chosen such that the nn vectors are given by eq.([eq : nn - vec ] ) .
( a ) the isotropic case , where @xmath90 .
( b ) the merged dirac points , where @xmath91 .
( c ) the zero dimensional case , where @xmath84 .
( d ) the aligned dirac points , where @xmath92 . ] , @xmath88 , and @xmath86 . the labels @xmath89 and @xmath76 represent the @xmath81 and @xmath80 components of the momentum , respectively .
the @xmath81 and @xmath80 axes have been chosen such that the nn vectors are given by eq.([eq : nn - vec ] ) .
( a ) the isotropic case , where @xmath90 .
( b ) the merged dirac points , where @xmath91 .
( c ) the zero dimensional case , where @xmath84 .
( d ) the aligned dirac points , where @xmath92 . ] , @xmath88 , and @xmath86 . the labels @xmath89 and @xmath76 represent the @xmath81 and @xmath80 components of the momentum , respectively .
the @xmath81 and @xmath80 axes have been chosen such that the nn vectors are given by eq.([eq : nn - vec ] ) .
( a ) the isotropic case , where @xmath90 .
( b ) the merged dirac points , where @xmath91 .
( c ) the zero dimensional case , where @xmath84 .
( d ) the aligned dirac points , where @xmath92 . ] , @xmath88 , and @xmath86 . the labels @xmath89 and @xmath76 represent the @xmath81 and @xmath80 components of the momentum , respectively .
the @xmath81 and @xmath80 axes have been chosen such that the nn vectors are given by eq.([eq : nn - vec ] ) .
( a ) the isotropic case , where @xmath90 . ( b ) the merged dirac points , where @xmath91 .
( c ) the zero dimensional case , where @xmath84 .
( d ) the aligned dirac points , where @xmath92 . ] . here , we have chosen @xmath86 . ( b )
dirac - point motion in the first brillouin zone for a shaking direction perpendicular to @xmath19 [ direction @xmath75 in the phase diagram ( a ) ] .
( c ) dirac - point motion in the first brillouin zone for a shaking direction parallel to @xmath19 [ direction @xmath10 in the phase diagram ( a ) ] .
the contour plots depict the dispersion of the isotropic system with an arbitrary color scale .
the area with higher contrast is the first brillouin zone . ]
another dimensional crossover , from 2d to 1d , may be obtained if the lattice is shaken in the direction parallel to one of the nn vectors ( direction 2 in fig.[fig : lattice ] ) . here , we choose @xmath93 to maintain the symmetry @xmath94 . in this case ,
both @xmath70 and @xmath82 are renormalized by bessel functions , albeit with different arguments .
since all hopping parameters are renormalized , the trajectory of the system in the phase space upon increasing the shaking amplitude is not a straight line , as was the case for shaking perpendicular to a hopping direction , and has a new feature : the alignment of dirac points , which occurs for @xmath95 .
the first zero of @xmath70 is found at @xmath96 for the same system of @xmath83k atoms mentioned above . here , for illustrative purposes , a simplified trajectory of the system
is depicted in fig.[fig : merging - energy - dispersion ] by arrow @xmath10 , which corresponds to the motion of the dirac points in reciprocal space as shown in fig.[fig : phase - diagram - abs](c ) . as @xmath70 approaches zero ,
the dirac points align in lines parallel to the @xmath81-axis at @xmath97 and the energy barriers between the aligning points are lowered .
consequently , when @xmath95 , the energy spectrum contains lines where the two energy band meet and the dispersion is linear , as is shown in fig.[fig : merging - energy - dispersion](d ) . the dispersion relation ( [ dispersionrel ] ) then reads simply @xmath98 and one clearly sees the 1d character . indeed , there is no dispersion in the @xmath89-direction , as is also evident from fig.[fig : merging - energy - dispersion](d ) , and the system may be viewed as completely decoupled 1d chains in which the zig - zag arrangement is of no importance . in this particular limit , the sites a and b are therefore no longer inequivalent such that the unit cell is effectively divided by two , and the size of the first brillouin zone is consequently doubled .
the aligned dirac points may thus , alternatively , be viewed as due to an artificial folding of the second ( outer ) half of the first brillouin zone into its inner half .
however , this aspect is very particular in that the brillouin zone immediately retrieves its original size when @xmath70 is small , but non - zero , or if nnn hoppings are taken into account . in both cases , one needs to distinguish the two different sublattices and one obtains a dispersion in the @xmath89-direction .
the actual behavior of the system for an increasing shaking amplitude is discussed in sec.[sec04 ] .
this behavior is more complicated because all three nn hopping parameters are renormalized , which , beyond the alignment , leads to the merging of dirac points and the opening of a gap also in the case of shaking parallel to one of the nn vectors . in the absence of nnn
hopping , the 0d limit can be reached in addition .
the major consequence of nnn hopping is to break particle - hole symmetry , as may be seen from eq .
( [ dispersionrel ] ) , where non - zero values of @xmath99 yield @xmath100 .
its relevance depends sensitively on the shaking direction , because of the different renormalization of the nn hopping parameters .
the band structure with nnn hopping included is depicted in fig.[fig : nnn - energy - dispersion ] for different shaking directions . in the case of a shaking perpendicular to @xmath19 , only @xmath82
are decreased , whereas @xmath101 remains the leading energy scale in the band structure . as a reference direction , but it may naturally be replaced by any other direction @xmath102 , in which case @xmath103 remains constant . ]
the band structure for the unshaken lattice is depicted in fig.[fig : nnn - energy - dispersion](a ) for @xmath104 , and one notices that the main features of the band structure , namely the dirac points , are unaltered with respect to the case @xmath86 in fig.[fig : merging - energy - dispersion](a ) , apart from the flattening of the upper band as compared to the lower one .
when approaching the merging transition @xmath79 , the value of which is determined by the zeros of @xmath105 in eq .
( [ fq ] ) and that therefore does not depend on the nnn hopping parameters , the band width remains dominated by the largest hopping parameter @xmath70 , such that the band structure [ fig.[fig : nnn - energy - dispersion](b ) ] at the transition is essentially the same as in fig.[fig : merging - energy - dispersion](b ) for @xmath86 . in the 0d limit , with @xmath106 the originally flat bands [ fig.[fig : merging - energy - dispersion](c ) ] acquire the weak dispersion of a triangular lattice as a consequence of the non - zero nnn hopping parameters .
however , as expected from the above arguments , the dispersion is on the order of @xmath107 and thus small as compared to the energy separation @xmath108 between the two bands .
in contrast to a shaking direction perpendicular to @xmath19 , nnn hopping has more drastic consequences if the lattice is shaken in the direction parallel to @xmath19 . in this case , all nn hopping parameters are decreased , and the relative importance of nnn hopping is enhanced .
notice further that the nnn lattice vectors @xmath109 are now perpendicular to the shaking direction such that @xmath110 remains unrenormalized .
also in this case , the system is approaching the 1d limit , with @xmath111 [ see fig.[fig : nnn - energy - dispersion](d ) ] .
however , in contrast to fig.[fig : merging - energy - dispersion](d ) , the chains remain coupled by nnn hopping that yields a dispersion in the @xmath89-direction . furthermore ,
as mentioned above , the a and b sites are now not equivalent from a crystallographic point of view , such that the outer parts of the first brillouin zone can not be folded back into the inner one , as may be seen from fig.[fig : nnn - energy - dispersion](d ) .
finally , for particular values of the shaking amplitude in the direction parallel to @xmath19 , the nn hopping parameters can be decreased in such a manner as to render @xmath82 more relevant . in this case , the two bands can overlap in energy , as depicted in figs.[fig : nnn - energy - dispersion](e ) and [ fig : nnn - energy - dispersion](f ) for @xmath112 ( in which case @xmath113 ) and @xmath114 ( with @xmath115 ) , respectively . in the latter example
there are no band contact points , in spite of the overlap between the two bands , and the system would be in an insulating phase if nnn hopping terms were not taken into account .
this overlap in energy between the two bands yields a non - zero density of states at any energy , such that the semi - metallic ( or insulating ) phase vanishes and yields , at half - filling , a metallic phase with particle and anti - particle pockets in the first brillouin zone . , @xmath88 , and @xmath116 .
the labels @xmath89 and @xmath76 represent the @xmath81 and @xmath80 components of the momentum , respectively .
the @xmath81 and @xmath80 axes have been chosen such that the nn vectors are given by eq.([eq : nn - vec ] ) .
( a ) the homogeneous case , where @xmath90 .
( b ) the merged dirac points , where @xmath91 .
( c ) the zero dimensional case , where @xmath84 .
( d ) the aligned dirac points , where @xmath92 . ( e )
an example of the metallic phase with @xmath112 .
( f ) another example of the metallic phase with @xmath114 . ] , @xmath88 , and @xmath116 . the labels @xmath89 and @xmath76 represent the @xmath81 and @xmath80 components of the momentum , respectively .
the @xmath81 and @xmath80 axes have been chosen such that the nn vectors are given by eq.([eq : nn - vec ] ) .
( a ) the homogeneous case , where @xmath90 .
( b ) the merged dirac points , where @xmath91 .
( c ) the zero dimensional case , where @xmath84 .
( d ) the aligned dirac points , where @xmath92 . ( e )
an example of the metallic phase with @xmath112 .
( f ) another example of the metallic phase with @xmath114 . ] , @xmath88 , and @xmath116 . the labels @xmath89 and @xmath76 represent the @xmath81 and @xmath80 components of the momentum , respectively .
the @xmath81 and @xmath80 axes have been chosen such that the nn vectors are given by eq.([eq : nn - vec ] ) .
( a ) the homogeneous case , where @xmath90 .
( b ) the merged dirac points , where @xmath91 .
( c ) the zero dimensional case , where @xmath84 .
( d ) the aligned dirac points , where @xmath92 .
( e ) an example of the metallic phase with @xmath112 .
( f ) another example of the metallic phase with @xmath114 . ] , @xmath88 , and @xmath116 . the labels @xmath89 and @xmath76 represent the @xmath81 and @xmath80 components of the momentum , respectively .
the @xmath81 and @xmath80 axes have been chosen such that the nn vectors are given by eq.([eq : nn - vec ] ) .
( a ) the homogeneous case , where @xmath90 .
( b ) the merged dirac points , where @xmath91 .
( c ) the zero dimensional case , where @xmath84 .
( d ) the aligned dirac points , where @xmath92 . ( e )
an example of the metallic phase with @xmath112 .
( f ) another example of the metallic phase with @xmath114 . ] , @xmath88 , and @xmath116 . the labels @xmath89 and @xmath76 represent the @xmath81 and @xmath80 components of the momentum , respectively .
the @xmath81 and @xmath80 axes have been chosen such that the nn vectors are given by eq.([eq : nn - vec ] ) .
( a ) the homogeneous case , where @xmath90 .
( b ) the merged dirac points , where @xmath91 .
( c ) the zero dimensional case , where @xmath84 .
( d ) the aligned dirac points , where @xmath92 . ( e )
an example of the metallic phase with @xmath112 .
( f ) another example of the metallic phase with @xmath114 . ] , @xmath88 , and @xmath116 . the labels @xmath89 and @xmath76 represent the @xmath81 and @xmath80 components of the momentum , respectively .
the @xmath81 and @xmath80 axes have been chosen such that the nn vectors are given by eq.([eq : nn - vec ] ) .
( a ) the homogeneous case , where @xmath90 .
( b ) the merged dirac points , where @xmath91 .
( c ) the zero dimensional case , where @xmath84 .
( d ) the aligned dirac points , where @xmath92 . ( e )
an example of the metallic phase with @xmath112 .
( f ) another example of the metallic phase with @xmath114 . ] as already alluded to in the previous sections , the shaking of a honeycomb lattice can lead to a sign change of the hopping parameters .
quite generally , fig.[fig : phase - diagram - sign ] shows that changing the relative signs of the nn hopping parameters results in a translation of the energy spectrum in momentum space .
this effect was also mentioned in ref .
@xcite . indeed
, the relative signs determine at which of the four time - reversal invariant momenta in the first brillouin zone the merging of dirac points and the semimetal - insulator transition take place when @xmath117 .
however , the sign change of the nn hopping parameters can be transformed away by a gauge transformation @xcite . nevertheless , the sign of the nnn hopping parameter is important , since it determines whether the upper or the lower band is flattened . .
( a ) to ( h ) contour plots of the energy bands , where the color scaling is arbitrary and the first brillouin zone is the area with higher contrast . the value of the nn hopping parameters for each contour plot are given by the position of the corresponding letter in the phase diagram and @xmath86 .
the dark regions indicate energies close to zero , whereas brighter regions are further away in energy from the fermi level at half filling . ]
until now , we have considered single - component fermionic atoms and , due to the pauli principle , the absence of @xmath118-wave interaction naturally results in an ideal fermi lattice gas , albeit with an unusual band structure . by trapping two hyperfine states of the fermionic atoms ,
hubbard - like interaction terms arise , @xmath119 where the fermionic operators now acquire an additional spin index @xmath120 , which is summed over , and @xmath121 is the interaction energy . naturally , in order to be able to apply the floquet theory in the presence of interactions , the hamiltonian @xmath53 in eq .
( [ eq : condition ] ) must now be replaced by @xmath122 .
this is the case in our study because we investigate the weak - coupling limit with @xmath123 .
since the interaction term commutes with the shaking , it is not renormalized , similarly to the on - site energy term proportional to @xmath16 in eq.([eq : h0 ] ) .
however , it has been shown that complications may arise when a multiple of the energy @xmath121 is in resonance with a harmonic of @xmath124 , @xmath125 , for integer @xmath26 and @xmath35 . whereas the limit @xcite @xmath126 is not considered here because it is in contradiction with the small-@xmath121 large - frequency limit , critical resonances may occur for @xmath127 @xcite .
nevertheless , it has been shown in ref .
@xcite that these resonances , which occur in higher - order perturbation theory , are strongly suppressed in the large-@xmath26 limit . in the weakly - interacting regime considered here , the ground state is adiabatically connected to that of the non - interacting system , with no broken symmetry .
first , we use the fourier transform of the creation and annihilation operators , @xmath128 , to find the hamiltonian in momentum space . within a hartree - fock theory
, we introduce a mean - field decoupling of the interaction terms , @xmath129 such that the expectation values of both sides are equal . for the mean value
we take @xmath130 where @xmath131 is the number of sites per sublattice , @xmath132 is the density of atoms with momentum @xmath133 and spin index @xmath134 , and @xmath135 is the kronecker delta .
we then obtain the mean - field hamiltonian @xmath136 where the total density is defined by @xmath137 .
the hamiltonian ( [ eq : ham - mf1 ] ) may be rewritten in a matrix form : @xmath138 where we have introduced the functions @xmath139,\ ] ] and @xmath105 is defined in eq .
( [ fq ] ) the hamiltonian ( @xmath140 ) can then be diagonalized by the unitary operator @xmath141 which yields @xmath142 because the @xmath143 and @xmath144 quasiparticles are free , the partition function corresponding to the hamiltonian ( @xmath145 ) reads @xmath146 \bigg\ } + \log\bigg\ { 1 + \exp\big[-\beta \big(h(\mu , \textbf{q } ) + |f(\textbf{q})| \big ) \big ] \bigg\ } \bigg ) \bigg],\ ] ] where @xmath147 with @xmath148 denoting the boltzmann s constant and @xmath38 the temperature .
the total number of particles @xmath149 is given by @xmath150 , and one obtains @xmath151 } + \frac{1}{1 + \exp\big[\beta \big(h(\mu , \textbf{q } ) + |f(\textbf{q})| \big ) \big ] } \bigg ) .
\label{eq : nrparticles}\ ] ] since the expression inside the sum does not depend on spin , summing over @xmath134 yields a factor 2 .
one recognizes in eq.([eq : nrparticles ] ) the fermi - dirac distribution function @xmath152^{-1}$ ] .
the number of particles @xmath149 is related to the density @xmath35 , which is defined here as the number of particles per lattice site , i.e. @xmath153 . converting the sum over @xmath133 into an integral , the following self - consistent equation for the density is derived @xmath154 + n_{fd}\big[\beta \big(h(\mu , \textbf{q } ) + |f(\textbf{q})| \big ) \big ] \bigg\ } , \label{eq : density - interaction - os}\ ] ] where the integral is restricted to the first brillouin zone , the surface of which is @xmath155 . + in fig.[fig : density - mu](a ) , the density @xmath156 is plotted for several values of @xmath157 . for the isotropic case , @xmath158 , the result of zhu _
et al_. is reproduced @xcite . for the 0d limit ,
the flat line due to the gap in the spectrum is clearly visible at the chosen temperature .
fig.[fig : density - mu](b ) confirms that repulsive interactions lead to a lower density than in a system without interactions for the same chemical potential .
fig.[fig : density - mu](c ) agrees with the observation that the nnn hopping breaks the particle - hole symmetry .
this effect is also visible in fig.[fig : density - mu](d ) , where the dependence of the density on the chemical potential is calculated for a shaking vector where the system is in the zero - gapped semimetallic phase for @xmath86 and in the metallic phase for @xmath116 . as a function of the chemical potential @xmath16 . unless specified otherwise in the figure , the nn hopping parameters @xmath159 , the nnn hopping parameter @xmath86 , the interaction strength @xmath160 , and the inverse temperature @xmath161 .
( a ) effect of the renormalization of the nn hopping parameters .
( b ) effect of the interaction strength @xmath121 for the isotropic case .
( c ) effect of the nnn hopping parameter @xmath116 in the shaken lattice .
( d ) the metallic phase . for the isotropic cases , @xmath112 , whereas for the 0d cases @xmath114 .
these systems are in the metallic phase for @xmath116 , whereas for @xmath86 they are in the semi - metallic and the insulating phase , respectively . ] as a function of the chemical potential @xmath16 . unless specified otherwise in the figure , the nn hopping parameters @xmath159 , the nnn hopping parameter @xmath86 , the interaction strength @xmath160 , and the inverse temperature @xmath161 .
( a ) effect of the renormalization of the nn hopping parameters .
( b ) effect of the interaction strength @xmath121 for the isotropic case .
( c ) effect of the nnn hopping parameter @xmath116 in the shaken lattice .
( d ) the metallic phase . for the isotropic cases , @xmath112 , whereas for the 0d cases @xmath114 .
these systems are in the metallic phase for @xmath116 , whereas for @xmath86 they are in the semi - metallic and the insulating phase , respectively . ] as a function of the chemical potential @xmath16 . unless specified otherwise in the figure , the nn hopping parameters @xmath159 , the nnn hopping parameter @xmath86 , the interaction strength @xmath160 , and the inverse temperature @xmath161 .
( a ) effect of the renormalization of the nn hopping parameters .
( b ) effect of the interaction strength @xmath121 for the isotropic case .
( c ) effect of the nnn hopping parameter @xmath116 in the shaken lattice .
( d ) the metallic phase . for the isotropic cases , @xmath112 , whereas for the 0d cases @xmath114 .
these systems are in the metallic phase for @xmath116 , whereas for @xmath86 they are in the semi - metallic and the insulating phase , respectively . ] as a function of the chemical potential @xmath16 . unless specified otherwise in the figure , the nn hopping parameters @xmath159 , the nnn hopping parameter @xmath86 , the interaction strength @xmath160 , and the inverse temperature @xmath161 .
( a ) effect of the renormalization of the nn hopping parameters .
( b ) effect of the interaction strength @xmath121 for the isotropic case .
( c ) effect of the nnn hopping parameter @xmath116 in the shaken lattice .
( d ) the metallic phase . for the isotropic cases , @xmath112 , whereas for the 0d cases @xmath114 .
these systems are in the metallic phase for @xmath116 , whereas for @xmath86 they are in the semi - metallic and the insulating phase , respectively . ]
honeycomb optical lattices have recently been realized experimentally , although the existing setups have only been used to investigate bosonic atoms @xcite .
shaking of a lattice has been experimentally implemented in a one - dimensional one by a periodic modulation on the position of the reflecting mirrors @xcite . for a honeycomb lattice
, the shaking could be realized by means of an acousto - optical device , as proposed for a triangular lattice in ref .
@xcite .
the magnitude of the nn hopping parameter @xmath71 in a honeycomb optical lattice has been evaluated in ref .
@xcite , @xmath162 in terms of the recoil energy @xmath163 and the magnitude of the potential barrier between nearest - neighbor lattice sites @xmath164 .
the magnitude of the nnn hopping parameter @xmath107 is not yet known , but could be determined from numerical band structure calculations . in a typical experimental situation , we expect the ratio @xmath165 to be in the @xmath166 range , in agreement with the parameter chosen in the discussion of sec.[sec02c ] . in a typical experiment ,
the shaking amplitude would be increased from zero to a finite value .
fig.[fig : phases - linear ] shows in which order the system goes through the different phases and dimensions upon increasing the shaking amplitude . here , also
the values of the shaking amplitude required for the dimensional crossovers are given for the same system as discussed in sec.[sec02c ] and @xmath167 .
if the shaking direction is perpendicular to one of the nn vectors , the system will be in the gapped insulating phase beyond a certain value of the shaking amplitude , since the bessel function crosses the value 0.5 only once and never obtains the value -0.5 .
if the shaking is parallel to one of the nn vectors , the system will be in the metallic phase beyond a certain value of the shaking amplitude .
nevertheless , it is still possible to induce a merging of dirac points and to open up a gap in the spectrum , since the nn hopping parameters are renormalized such that the value of one of them will in general differ from that of the other two .
however , whether the system is actually driven into an insulating phase or remains metallic depends on the precise value of the ratio @xmath165 . in nm correspond to the optical lattice discussed in sec.[sec02c ] with @xmath168 , @xmath169khz , @xmath170 nm , and containing @xmath83k atoms .
( a ) shaking perpendicular to one of the nn hopping directions .
( b ) shaking parallel to one of the nn hopping directions . ] in experiments , an overall harmonic trapping potential is imposed to confine the atoms .
it is described by @xmath171 where @xmath172 is the trapping frequency , and @xmath9 is the position measured from the center of the trap . by applying the local density approximation ( lda ) , one finds that the chemical potential evolves radially according to @xmath173 .
fig.[fig : density - r](a ) shows the density profile for several ratios of @xmath157 , without nnn hopping or interactions .
the case with @xmath174 , when the system is in the extreme limit of the band insulating phase , can be well distinguished from the other cases .
fig.[fig : density - r](b ) shows that stronger interactions lead to a higher density away from the center of the trap .
this effect becomes visible when the density starts to deviate from one particle per lattice site .
next - nearest - neighbor hopping leads to a higher density at the edge of the cloud compared to the case without nnn hopping , which can be seen from comparing figs.[fig : density - r](a ) and ( c ) and from fig.[fig : density - r](d ) .
the latter shows the effect of nnn hopping on the density profile for the case where the nnn hopping gives rise to the metallic phase for two different shaking vectors . in the first case , @xmath112 , which gives @xmath113 , such that without nnn hopping , the system is in the zero - gapped semi - metallic phase and the dirac points are located very close to the corners of the first brillouin zone . in the second case , @xmath114 , which results in @xmath115 , such that without nnn hopping the system is in the insulating phase and the two energy bands are almost flat . as a function of the distance from the trap s centre @xmath175 , which is expressed in units of the nearest - neighbor distance @xmath144 .
the trapping frequency has been chosen such that the trapping potential is given by @xmath176 .
the chemical potential @xmath16 for each case has been chosen such that the density at the trap s centre @xmath35 is one particle per site .
this corresponds to half - filling , since we consider 2 species of fermions . unless specified otherwise in the figure , the nn hopping parameters @xmath159 , the nnn hopping parameter @xmath86 , the interaction strength @xmath160 , and the inverse temperature @xmath161 .
( a ) effect of the renormalization of the nn hopping parameters .
( b ) effect of the interaction strength @xmath121 for the isotropic case .
( c ) effect of the nnn hopping parameter @xmath116 in the shaken lattice .
( d ) the metallic phase . for the isotropic cases , @xmath112 , whereas for the 0d cases @xmath114 .
these systems are in the metallic phase for @xmath116 , whereas for @xmath86 they are in the semi - metallic and the insulating phase , respectively . ] as a function of the distance from the trap s centre @xmath175 , which is expressed in units of the nearest - neighbor distance @xmath144 .
the trapping frequency has been chosen such that the trapping potential is given by @xmath176 .
the chemical potential @xmath16 for each case has been chosen such that the density at the trap s centre @xmath35 is one particle per site .
this corresponds to half - filling , since we consider 2 species of fermions . unless specified otherwise in the figure , the nn hopping parameters @xmath159 , the nnn hopping parameter @xmath86 , the interaction strength @xmath160 , and the inverse temperature @xmath161 .
( a ) effect of the renormalization of the nn hopping parameters .
( b ) effect of the interaction strength @xmath121 for the isotropic case .
( c ) effect of the nnn hopping parameter @xmath116 in the shaken lattice .
( d ) the metallic phase . for the isotropic cases , @xmath112 , whereas for the 0d cases @xmath114 .
these systems are in the metallic phase for @xmath116 , whereas for @xmath86 they are in the semi - metallic and the insulating phase , respectively . ] as a function of the distance from the trap s centre @xmath175 , which is expressed in units of the nearest - neighbor distance @xmath144 .
the trapping frequency has been chosen such that the trapping potential is given by @xmath176 .
the chemical potential @xmath16 for each case has been chosen such that the density at the trap s centre @xmath35 is one particle per site .
this corresponds to half - filling , since we consider 2 species of fermions . unless specified otherwise in the figure , the nn hopping parameters @xmath159 , the nnn hopping parameter @xmath86 , the interaction strength @xmath160 , and the inverse temperature @xmath161 .
( a ) effect of the renormalization of the nn hopping parameters .
( b ) effect of the interaction strength @xmath121 for the isotropic case .
( c ) effect of the nnn hopping parameter @xmath116 in the shaken lattice .
( d ) the metallic phase . for the isotropic cases , @xmath112 , whereas for the 0d cases @xmath114 .
these systems are in the metallic phase for @xmath116 , whereas for @xmath86 they are in the semi - metallic and the insulating phase , respectively . ] as a function of the distance from the trap s centre @xmath175 , which is expressed in units of the nearest - neighbor distance @xmath144 .
the trapping frequency has been chosen such that the trapping potential is given by @xmath176 .
the chemical potential @xmath16 for each case has been chosen such that the density at the trap s centre @xmath35 is one particle per site .
this corresponds to half - filling , since we consider 2 species of fermions . unless specified otherwise in the figure , the nn hopping parameters @xmath159 , the nnn hopping parameter @xmath86 , the interaction strength @xmath160 , and the inverse temperature @xmath161 .
( a ) effect of the renormalization of the nn hopping parameters .
( b ) effect of the interaction strength @xmath121 for the isotropic case .
( c ) effect of the nnn hopping parameter @xmath116 in the shaken lattice .
( d ) the metallic phase . for the isotropic cases , @xmath112 ,
whereas for the 0d cases @xmath114 .
these systems are in the metallic phase for @xmath116 , whereas for @xmath86 they are in the semi - metallic and the insulating phase , respectively . ]
we emphasize that , in the present paper , we only discuss weak correlations that adiabatically affect the density .
however , when increasing further the onsite interaction , one may expect correlated phases with inhomogeneous density , even at half - filling .
a detailed study of these correlated phases is a vast research issue that is yet ongoing and that is beyond the scope of the present paper . here
, we only provide a glimpse on how the density , which we discussed above in the weak - coupling limit , may evolve in view of some phases studied in the literature .
mean - field calculations indicate a transition to an antiferromagnetic state above a value of @xmath177 @xcite , whereas more sophisticated quantum monte - carlo calculations indicate an intermediate spin - liquid phase between the semimetal and the anti - ferromagnetic phase @xcite . the spin - liquid phase may be viewed as a mott insulator with a charge localization on the lattice sites , and recent slave - rotor calculations indicate that such spin - liquid phases dominate the phase diagram for @xmath178 @xcite , which is the parameter range where the dirac points would merge in the absence of interactions .
the precise transition between the weakly - interacting liquid phases and these strongly - correlated mott insulators could in principle be determined with the help of the above - mentioned density measurements .
a more promising technique for detecting dirac - point motion is momentum - resolved raman spectroscopy .
this technique has been proposed as an equivalent of angle - resolved photoemission spectroscopy for cold atom systems @xcite .
it has not yet been realized experimentally , while momentum - resolved radio - frequency spectroscopy , which is a very similar technique , has already been implemented @xcite .
notice further that another very similar technique , momentum - resolved bragg spectroscopy , has been applied to ultracold bosonic atoms in a static optical lattice by ernst _
et al_. @xcite .
momentum - resolved spectroscopy can allow us to indirectly visualize the band structure . in momentum - resolved raman spectroscopy ,
two laser pulses with frequencies @xmath179 and @xmath180 are irradiated upon the system .
if the frequency difference is in resonance with a transition @xmath181 between atomic hyperfine states , @xmath182 , some atoms are excited in a second - order process to the higher hyperfine state .
then , with state - selective time - of - flight measurements , the dispersion of the atoms in the new state are measured , from which the dispersion of the original atoms can be derived . when the atoms are confined in a trapping potential and the laser pulses are focussed on the center of the trap ,
the quality of the results obtained by raman spectroscopy is comparable to those of a homogeneous system @xcite .
furthermore , raman spectroscopy yields better results for a system with strong interactions compared to standard time - of - flight measurements @xcite .
notice that momentum resolved raman spectroscopy was originally proposed to be applied to a gas of ultracold fermionic atoms at equilibrium and not for a shaken lattice .
we therefore discuss , in this final paragraph , why we think that this technique may also be applied to the present case .
naturally , as long as the frequencies of the additional lasers in the raman - spectroscopy setup are small with respect to the shaking frequency , @xmath183 , even the full system satisfies the condition ( [ eq : condition ] ) for the validity of floquet theory . as in the case of interactions ,
one needs , however , to avoid resonances between the different laser frequencies that could become critical @xcite . the opposite limit , in which the laser frequencies and that of the hyperfine transition are larger than the shaking frequency , is more delicate .
however , even then , the shaken system remains at quasi - equilibrium as long as the intensities of the lasers used in raman spectroscopy are weak , such that they only constitute a small perturbation .
the atomic dynamics probed even at high frequencies is therefore still that of the atoms at quasi - equilibrium , with the band strucure obtained from floquet theory .
furthermore , in the experimental studies by zenesini _
et al_. @xcite , time - of - flight measurements were used to determine the momentum distribution of bosonic atoms in a shaken lattice .
apart from the time - scale considerations , there are also some length scales that need to be taken into account .
there are indeed two requirements for the correct size of the focus of the laser beams . on the one hand
, it needs to be larger than the lattice spacing , such that sufficiently many atoms can be excited , while on the other hand the focus of the beams should be small enough , in order to have an approximately flat trapping potential inside the focus area .
in addition , choosing the length of the pulses could possibly be a problem , since for shorter pulses , the excited atoms will be less affected by the lattice potential , whereas for longer pulses more atoms can be excited , leading to a stronger signal .
in conclusion , we have investigated the band engineering of fermionic atoms in an optical honeycomb lattice with the help of a periodic shaking of the lattice .
if the shaking frequency @xmath24 is large enough , i.e. if @xmath124 constitutes the largest energy scale in the system , the floquet theory may be applied and the system is at quasi - equilibrium in the sense that the atoms can not follow the rapid motion associated with the shaking .
depending on the direction of the shaking , one may render the hopping amplitudes in the quasi - static lattice anisotropic , due to a renormalization of the nn and nnn hopping parameters by bessel functions that go through zero and change sign . as a consequence ,
dimensional crossovers can be induced in absence of the nnn hopping . for a shaking direction parallel to one of the nn vectors ( such as e.g. @xmath19 )
, one can make one of the nn hopping parameters vanish , @xmath184 .
the system then undergoes a transition from 2d to 1d , while the dirac points align simultaneously .
shaking in the perpendicular direction ( @xmath185 ) allows one to decrease two nn hopping amplitudes simultaneously while maintaining @xmath70 unrenormalized . in this case , a dimensional crossover from 2d to 0d is induced for @xmath186 , leading to two flat energy bands , beyond the merging of dirac points @xcite , which occurs at @xmath187 .
a nonzero value of @xmath107 breaks the particle - hole symmetry and leads to a coupling among the 1d chains and the 0d dimers , for the @xmath95 and @xmath106 cases , respectively , and thus to a weak 2d dispersion .
the merging and the alignment of dirac points , however , are not affected . moreover , for a shaking direction parallel to @xmath19 , one pair of nnn hopping amplitudes [ @xmath109 ] remains unrenormalized , and its relative importance is thus enhanced when compared to the decreasing nn hopping amplitudes . in this limit , beyond the semi - metallic and the band - insulating phases , a novel metallic phase can appear that consists of particle and hole pockets with a non - vanishing density of states even at half - filling .
furthermore , we have investigated the role of weak repulsive on - site interactions .
the resulting ground state is then adiabatically connected to that of the non - interacting system , and we have self - consistently calculated the dependence of the atomic density on the ( local ) chemical potential .
the density profiles of the different phases , e.g. the gapless semimetal or the gapped band insulator , and the different dimensionality may be measured experimentally by in - situ density measurements .
moreover , momentum - resolved raman spectroscopy may be a promising technique to measure the band structure associated with these different phases .
we thank gilles montambaux , guangquan wang , andreas hemmerich , and marco di liberto for fruitful discussions .
we also thank christoph lschlger for informing us about an error in the calculations .
this work was financially supported by the anr project nanosim graphene under grant no .
anr-09-nano-016 and by the netherlands organization for scientific research ( nwo ) .
for the studied case , @xmath189 , where @xmath53 was given in eq.([eq : h0 ] ) . since the nn hopping is usually larger than the nnn hopping , @xmath190 , and since we let the chemical potential be in the range @xmath191 , the dominant energy scale in the hamiltonian @xmath53 is @xmath71 .
therefore , if @xmath59 , the condition ( [ eq : condition ] ) is satisfied and the floquet theory may be applied . in the general case ,
the effective hamiltonian is given by @xcite @xmath192_n \bigg \rangle_t , \label{eq : app - h - eff}\ ] ] where for the shaken honeycomb lattice , we choose @xmath193 using the ( nonvanishing ) commutation relations @xmath194 = & a^\dagger_{\textbf{r } ' } \ , b_{\textbf{r}+\textbf{d}_j } \
, \delta_{\textbf{r } ' , \textbf{r } } \ , , \nonumber \\ [ a^\dagger_{\textbf{r}'}\,a_{\textbf{r } ' } , b^\dagger_{\textbf{r}+\textbf{d}_j } \ , a_{\textbf{r } } ] = & - b^\dagger_{\textbf{r}+\textbf{d}_j } \ , a_{\textbf{r } ' } \ , \delta_{\textbf{r } ' , \textbf{r } } , \nonumber \\ [ a^\dagger_{\textbf{r}'}\,a_{\textbf{r } ' } , a^\dagger_{\textbf{r } } \ , a_{\textbf{r}+\textbf{d}_i -\textbf{d}_j } ] = & a^\dagger_{\textbf{r } ' } \ , a_{\textbf{r}+\textbf{d}_i -\textbf{d}_j } \ , \delta_{\textbf{r}',\textbf{r } } \nonumber \\ & - a^\dagger_\textbf{r } \ , a_{\textbf{r } ' } \ , \delta_{\textbf{r}',\textbf{r}+\textbf{d}_i -\textbf{d}_j } , \label{eq : commutators}\end{aligned}\ ] ] which are valid for both fermionic and bosonic creation and annihilation operators , and equivalent ones for the creation and annihilation operators on the @xmath11 sublattice , the multiple commutator in eq.([eq : app - h - eff ] ) becomes @xmath195_n & = \left[\frac{m \omega^2}{\hbar \omega } \sin(\omega t ) \right]^n \nonumber \\
\bigg\ { & - \gamma \sum_{j=1}^3 \sum_{\textbf{r } \in a } ( \textbf{d}_j \cdot \boldsymbol{\rho})^n \left [ ( -1)^n a^\dagger_\textbf{r } b_{\textbf{r } + \textbf{d}_j } + b^\dagger_{\textbf{r } + \textbf{d}_j } a_\textbf{r } \right ] \nonumber \\ & - \gamma ' \sum^3_{i=1 } \sum^3_{j=1,j\neq i } \bigg ( \sum_{\textbf{r } \in a } [ ( \textbf{d}_j - \textbf{d}_i ) \cdot \boldsymbol{\rho}]^n \ , a^\dagger_\textbf{r } \ , a_{\textbf{r } + \textbf{d}_i - \textbf{d}_j }
\nonumber \\ & + \sum_{\textbf{r } \in b } [ ( \textbf{d}_j - \textbf{d}_i ) \cdot \boldsymbol{\rho}]^n \ , b^\dagger_\textbf{r } b_{\textbf{r } + \textbf{d}_i - \textbf{d}_j } \bigg ) \bigg\ } \nonumber \\ & - \mu \bigg ( \sum_{\textbf{r } \in a } \ , a^\dagger_\textbf{r } a_\textbf{r } + \sum_{\textbf{r } \in b } \ , b^\dagger_\textbf{r } b_\textbf{r } \bigg ) .\end{aligned}\ ] ] finally , after performing the time average and evaluating the sum over @xmath35 , the effective hamiltonian ( [ hamiltonian - eff ] ) is obtained . c. becker , p. soltan - panahi , j. kronjger , s. drscher , and k. bongs , and k. sengstock , new j. phys .
* 12 * , 065025 ( 2010 ) ; a. klinger , s. degenkolb , n. gemelke , k .- a .
brickman soderberg , c. chin , rev .
instrum . * 81 * , 013109 ( 2010 ) ; j. struck , c. lschlger , r. le targat , p. soltan - panahi , a. eckardt , m. lewenstein , p. windpassinger , and k. sengstock , science * 333 * , 996 ( 2011 ) ; p.
soltan - panahi , d .- s .
lhmann , j. struck , p. windpassinger , and k. sengstock , nature physics * 8 * , 71 ( 2012 ) .
white and a.e .
feiguin , phys .
lett . * 93 * , 076401 ( 2004 ) ; a.j .
daley , c. kollath , u. schollwck , and g. vidal , j. stat .
: theor . exp .
* p04005 * ( 2004 ) ; d. gobert , c. kollath , u. schollwck , and g. schutz , phys .
e * 71 * , 036102 ( 2005 ) ; u. schollwck , rev .
phys . * 77 * , 259 ( 2005 ) ; d. gobert , c. kollath , u. schollw ck , and g. schtz , phys .
e * 71 * , 036102 ( 2005 ) ; c. kollath , a. m. luchli , and e. altman , phys .
. lett . * 98 * , 180601 ( 2007 ) .
|
inspired by the recent creation of the honeycomb optical lattice and the realization of the mott insulating state in a square lattice by shaking , we study here the shaken honeycomb optical lattice . for a periodic shaking of the lattice , a floquet theory may be applied to derive a time - independent hamiltonian . in this effective description
, the hopping parameters are renormalized by a bessel function , which depends on the shaking direction , amplitude and frequency . consequently
, the hopping parameters can vanish and even change sign , in an anisotropic manner , thus yielding different band structures . here , we study the merging and the alignment of dirac points and dimensional crossovers from the two dimensional system to one dimensional chains and zero dimensional dimers .
we also consider next - nearest - neighbor hopping , which breaks the particle - hole symmetry and leads to a metallic phase when it becomes dominant over the nearest - neighbor hopping .
furthermore , we include weak repulsive on - site interactions and find the density profiles for different values of the hopping parameters and interactions , both in a homogeneous system and in the presence of a trapping potential .
our results may be experimentally observed by using momentum - resolved raman spectroscopy .
|
the quasar q2237 + 0305 at redshift @xmath3 is gravitationally lensed by a nearby galaxy at @xmath4 ( huchra _ et al . _ 1985 ) .
the galaxy core lies nearly perfectly along the line of sight .
such a configuration results in a symmetric , cross - like arrangement of the four quasar component images , with relative separations in this case between 1.2 and 1.8 arcsec .
several facts make this lens system useful .
first , the closeness of the lensing spiral galaxy allows us to study it in great detail .
second , the large leverage between lens plane and source plane that results from this proximity reduces the time scale for microlensing .
third , the symmetric arrangement of the four components leads to a small relative time delay ( of order a day or shorter , _ cf .
_ wambsganss & paczyski 1994 ) .
this last fact helps to distinguish intrinsic fluctuations of the quasar from microlensing - induced changes : the former have to show up in all four components almost simultaneously , whereas the latter are completely independent of each other , and occur on a timescale of months .
finally , the fact that the four quasar components are bright , well separated , and of comparable optical brightness makes q2237 + 0305 an easy target for various photometric , spectroscopic and astrometric studies .
recently , q2237 + 0305 was detected at radio wavelengths with the vla at 3.6 cm and 20 cm ( falco _ et al . _
the relative positions of the components measured in radio agree well with the optical positions .
furthermore , the relative flux ratios of the components in each wavelength regime are similar , with the exception that component d is brighter in the radio , compared with the optical light .
thus , the relative brightnesses in radio agree much better with the ratios predicted by various models ( rix , schneider & bahcall 1992 ) , than do those in the optical ; the slight discrepancy in the optical is possibly caused by a demagnification due to microlensing ( _ cf . _ wambsganss , paczyski & schneider 1990 ; witt & mao 1994 ) , or by dust in the lensing galaxy .
the system q2237 + 0305 was the first in which microlensing was detected ( irwin _ et al . _ 1989 ; corrigan _ et al . _
there are various photometric monitoring programs underway ; these continue to show fluctuations in the component intensities which can be attributed to microlensing .
recent results of some of these campaigns were published by lewis , irwin & hewett ( 1995 ) , and ostensen _
et al . _ ( 1996 ) . here
we present new data obtained with the wfpc2 camera aboard the hubble space telescope .
the motivation and the technical details of the observations are explained in section 2 . in section 3
we present the results of these observations , including the photometry and the astrometry of the four components .
we discuss our results and their implications in the final section 4 .
the primary goal of the proposed observations was to use the q2237 + 0305 lens system as a `` zwicky telescope , '' _
i.e. _ to take advantage of the lensing magnification which nature provides in this case to image the background qso s structure with unprecedented and otherwise unavailable resolution .
we chose q2237 + 0305 in part because yee & de robertis ( 1991 ) had claimed to see ly-@xmath0 emission distinct from the quasar components , possibly resulting from an intense star - forming region associated with the quasar .
in addition , the quasar s distance redshifts the ly-@xmath0 line to 3270 , which is available to hst but is not so red that the light of the lensing galaxy swamps the quasar image .
high surface brightness , spatially extended structure in the quasar image will most likely appear in ly-@xmath0 .
such structure could result from the broad line region ( blr ) , from a larger and more quiescent quasar hii region , or possibly even from intense star formation regions in the host galaxy .
also , if we were so fortunate as to resolve the continuum emitting region in some way , it is also expected to be relatively bright in the uv .
meanwhile , the light of the relatively red bulge component of the lensing galaxy , through which we see the quasar components , provides little uv background to swamp low surface brightness parts of the source image .
we proposed obtaining deep wfpc2 images in the f336w and f300w bands .
both bands are near the ly-@xmath0 line , but ly-@xmath0 would appear brighter in f336w than f300w .
this difference allows us to distinguish between continuum and ly-@xmath0 emission .
we hoped , in the most optimistic scenario , that this project could yield the first well - resolved uv image of a classical high redshift qso and thus provide a fundamentally important datum in our attempts to understand quasars . a ly-@xmath0 ring image , similar to the radio einstein rings " , was an extreme but real possibility .
finally , even if no resolved images were detected , the data would provide useful upper bounds on sizes and surface brightnesses of structures near the core of this qso and so are still of considerable interest . on 3 november 1995 , we obtained a total of twenty - five exposures using the wfpc2 camera aboard the hubble space telescope .
ten of these images were exposed on one of the wide field chips ; of these images , four were taken using the f300w filter and six were taken using the f336w filter .
the wide field images were exposed for approximately 1200 sec each .
the remaining fifteen images were exposed on the planetary camera chip ; of these images , six were taken using the f300w filter and nine were taken using the f336w filter .
the planetary camera images were exposed for approximately 800 sec each .
thus , we had four combinations of chip and filter .
in addition , within each of these combinations , the images were dithered by approximately half - integer numbers of pixels ( _ e.g .
_ @xmath5 or @xmath6 ) .
this dithering procedure allowed us to recover resolution higher than that of the chips pixels .
see table 1 for details on all of these exposures .
the pc chip scale is @xmath7 mas / pixel , and the scale of the wf3 chip is @xmath8 mas / pixel ( holtzmann _ et al . _ 1995 ) .
thus , the pc achieves finer resolution while the wf3 yields greater signal - to - noise .
we would expect the wf3 chip to detect large and diffuse signals more readily than the pc chip , and the pc chip to better perceive features close to the quasar core .
additionally , the pc chip is more suited to astrometry than the wf3 chip . because of contamination difficulties , the pc chip is also preferable for photometry . for the purpose of some of the work described here , it was convenient to combine the various images .
this combination required using the _ iraf _ routine crrej to remove cosmic rays , and the _ iraf _ routine drizzle to account for the dithering of the images ( fruchter & hook 1997 ) .
thus , we obtained four final images : two filters ( f300w and f336w ) times two chips ( pc and wf3 ) .
these four images appear in figure 1 .
finally , we note that the field around q2237 + 0305 is extremely empty in our uv images .
only two field stars of significant brightness were detected in our exposures , both of them on exposures 1125 of table 1 , when the quasar appeared on the pc chip .
one was @xmath9 17 mag and appeared on wf2 ; the other was @xmath9 19 mag and appeared on wf3 .
this dearth of field stars , and their absence on the pc chip , posed difficulties in the photometric and astrometric calibration .
as well as addressing the issues related to the physics of quasars outlined in section 2 , these exposures provide excellent astrometric and good photometric information .
thus , before discussing the possibility of resolving a ly-@xmath0 emitting region in the lensed quasar , we describe our photometric and astrometric results .
as we outline below , we obtained most of these results from the uncombined images . in order to obtain the position of the galaxy core , however , it was necessary to use the combined images .
finally , we examine the possibility of a resolvable blr or continuum - emitting region , using the combined images to obtain maximum signal - to - noise .
we measure the relative brightness of the four quasar components a , b , c , d ( identified in figure 1 ) in each of the twenty - five exposures , using the _ iraf _ procedure qphot .
the zeropoint magnitudes for the stmag system used here ( the constant flux system used by space telescope ) were derived from the photflam values associated with each filter and chip .
the errors in the zeropoints are about 0.02 mag ( holtzmann _ et al . _ 1995 ) .
the other major source of error is contamination , which accumulates on the wfpc2 after each monthly decontamination and degrades the throughput of the instrument .
the effect of contamination is more severe for filters of shorter wavelength , and varies between the chips .
for f300w and f336w , it is about 0.01 0.05 mag . following whitmore , heyer & baggett ( 1996 )
, we used the synphot package and a synthetic quasar spectrum with @xmath10 to estimate the effect of contamination in our case .
our results yield about 0.010.02 mag contamination in the pc chip , and 0.040.06 mag contamination in the wf3 chip . because the contamination corrections are rather uncertain , we use only the pc chip to obtain the overall photometry .
we average our results for each filter in the pc chip and list them in table 2 .
the errors listed there are simply the @xmath11 variance found in the determination of the mean , added in quadrature with the errors expected in the zeropoint and in the contamination correction ( @xmath9 0.01 mag in f336w , @xmath9 0.02 mag in f300w ) . in figures 2a and 2b ,
we examine the time dependence of the component brightnesses by plotting the photometry for each image of table 1 .
most of the exposures of the a component on wf3 have saturated pixels and are therefore not included in the plot .
other exposures are not included because cosmic rays interfere with the determination of their photometry .
it is useful to examine two time scales .
first , because of the contamination and zeropoint uncertainties , we examine each chip / filter combination separately . this examination probes a timescale of only 34 hours , but allows us to assume that zeropoint and contamination problems are uniform for a single chip / filter combination .
accordingly , the error bars on the data points in figures 2a and 2b do not include the zeropoint and contamination errors , but only the statistical photometric errors convolved with 0.01 mag pixel - centering errors . to test for variation , we compare the measured @xmath11 ( @xmath12 ) dispersion about the mean ( dotted lines figures 2a and 2b ) with the estimated @xmath11 ( @xmath13 ) errors on the data points .
if the ratio @xmath14 of these two quantities were large , we could claim possible variation . for comparison , for our 17 mag field star on wf2 @xmath15 . for the quasar , in the pc chip
, @xmath14 exceeds 1 only for the b component in the f336w filter ; this exception is due to a single anomalously bright image of b which we do not believe is intrinsic variation . in the wf3 chip ,
the situation is more difficult , especially in the f336w filter , where for the b , c , and d components @xmath16 . excepting these final images , we state that over the first 10 hours component
a remains constant to within 0.02 mag and the other components remain constant to within 0.05 mag .
to examine the situation in the last 5 hours more thoroughly , we plot d@xmath17b and c@xmath17b magnitude differences in f336w versus time in figure 3 . the error bars and dotted lines
have the same meanings as in figure 2 .
we notice that while c and b remain at constant relative brightness throughout the run , and between the pc chip to the wf3 chip , d brightens relative to b and c in the final three images .
namely d@xmath17b changes by about 0.1 mag over the course of about 1.5 hrs .
it is possible that this change is due to either intrinsic variability of the quasar or microlensing .
we note that it is more likely that d brightens than that b and c dim .
first , an explanation due to microlensing would clearly require two unlikely coincidences : that images b and c would start to vary at the same time and by the same amount ; second , since the estimates of the time delay between b and c from lens models are typically @xmath9 10 hours ( rix _ et al . _ 1992 ) , intrinsic variation of the quasar source would not cause them to vary together either .
it is worth noting that the variations of b , c , and d , taken individually , are not excessive in these last three images ( @xmath16 compared to 1.5 for a field star ) .
therefore , we can not exclude that the large change in d@xmath17b is simply a coincidence between the errors in components b , c , and d ; that is , that b and c just happen to appear somewhat dim in these images and that d just happens to appear to brighten . to investigate the second interesting time scale ,
we look at the difference between the photometry on the pc chip and on the wf3 chip in each filter ; this examination extends the timescale to @xmath18 hours for f300w and @xmath19 hours for f336w , but increases our errors because we have to account for the zeropoint and contamination difficulties .
in fact , a glance at figures 2a and 2b show that there must indeed be large contamination correction or zeropoint errors , as there are shifts in the apparent magnitudes of up to 0.1 mag halfway through the run , when we changed to the wf3 chip .
thus , although the pc and wf3 magnitudes are separated by @xmath20 even when we account for contamination correction and zeropoint errors , we do not regard these differences as intrinsic .
( note that any zeropoint and contamination errors do not affect our above conclusions regarding b@xmath17d ) .
photometry in the f336w filter was performed by rix _
et al . _ ( 1992 ) .
their results agreed with ours for components a and d , but they found components b and c to be significantly brighter than we did ( about 0.4 magnitudes each ) .
note that the color differences between components ( presumably due to reddening by the galaxy core of the c and d components ) preclude a meaningful comparison of our relative brightnesses to those observed in bands other than f336w or f300w .
since the pc chip is better suited to astrometry than the wf3 chip , we measure the quasar and galaxy positions using only the pc chip . for each of the fifteen exposures on this chip , we determine the position of each quasar component ( @xmath21 , @xmath22 , @xmath23 , and @xmath24 ) using the _ iraf _ procedures imcentroid and metric .
because the foreground galaxy has such a weak signal in our bands , we neglect the effect of any gradient in the galaxy light distribution on our calculated positions .
because there are small rotations of the camera from exposure to exposure , we want to define the quasar geometry in a way which does not depend on the ( slightly uncertain ) orientation of the camera .
thus , we pick five quasar component pairs ( @xmath25 , @xmath26 , @xmath27 , @xmath28 , and @xmath29 ) and for each exposure measure the distances between the members of each pair . thus , after measuring the positions of @xmath21 , @xmath22 , and @xmath23 in an exposure , we can calculate the distances @xmath25 , @xmath30 , and @xmath28 .
even if the camera orientation changes from image to image , these distances should remain constant , as they are invariant under rotation .
indeed , they are quite constant , and we are able to average their values over all the exposures . these distances are listed in table 3 .
note that the standard deviations are all approximately 1.5 mas .
when we perform the same procedure on the wf3 images , they yield results which are consistent with those of the pc chip , but with standard deviations of order 10 mas . since the galaxy is quite dim in the f300w and f336w filters , it is necessary to measure the position of its core using coadded images .
we coadd the f300w exposures and f336w exposures separately , measure the distance of the galaxy core from components @xmath22 and @xmath24 in both coadded images , and then average the results to obtain the distances @xmath31 and @xmath32 listed in table 3 .
such a measurement is not as accurate as the distance determinations between the ( point - like ) quasar components , since the images undergo at least one shifting operation in the process of being coadded , in addition to being subjected to cosmic - ray rejection algorithms .
furthermore , the galaxy core is not point - like , but rather extended . thus there could be systematic errors in the @xmath31 and @xmath32 distances which are larger than the statistical errors listed .
we use these distances to compare our results to those of other investigators .
essentially , the distances provide a more robust comparison of results than coordinates do because of the rotation problem described above ; using distances instead of coordinates eliminates the need to worry about whether somebody else s coordinate system is exactly lined up with ours . in rows two through six of table 3 ,
we have listed the results of a number of other investigators .
immediately it is clear that our results differ systematically from most of the previous results . in particular ,
our distances are between 0.1 2.0 % larger than those of all other investigators . only for the @xmath31 distance
is our result smaller .
this observation raises the possibility that we have determined our absolute scale inaccurately . nevertheless ,
because the wfpc2 is not limited by atmospheric seeing , because it is a high quality post - repair _ hst _ observation , and because it is in a band in which the galaxy is faint , we are still confident of our results .
the deflection angles and hence image separations in a simple gravitational lens system are proportional to the lens mass enclosed by the images ; thus , our result implies an increase of 0.2
4.0% for this quantity , assuming a singular isothermal sphere model .
this factor would be negligibly small compared to uncertainties in almost any other system , but detailed modeling of q2237 + 0305 yields a mass for the inner region of the lensing galaxy which is thought to be accurate at such a level ( rix _ et al . _
1992 ) . here
we discuss possible reasons for the systematic discrepancies between our astrometric results and those of others .
yee ( 1988 ) , whose values are consistently about @xmath33 below ours , measured his astrometry in wavelengths ranging from 495 nm upwards .
it is possible that the shape of the quasar region that dominates at these higher wavelengths is different than the shape of the region that dominates around 300 nm ; such a shape difference could cause a slightly different image configuration .
probably a more important effect is that in the higher wavelengths , the foreground galaxy is bright , and the gradient of its light distribution can skew the determined positions of the quasar components . of course
, he accounted for this effect using an iterative method which fit for the galaxy light distribution and the quasar component positions together .
however , this procedure demands an accurate knowledge of the psfs , and can be unreliable .
the same comments apply to the results of irwin ( 1989 ) and racine ( 1991 ) ; furthermore , racine ( 1991 ) determined his plate scale by scaling his results to those of yee ( 1988 ) , and therefore did not produce an independent result . finally ,
et al . _ ( 1992 ) and crane _ et al . _ ( 1991 ) both suffer from the fact that the astrometry was performed with a pre - repair _ hst _ instrument .
it is worth noting , however , that the rix _
_ ( 1992 ) results do agree with ours within their @xmath11 error bars , though their individual values are consistently smaller than ours .
ideally , one would prefer to approach this question of scale by measuring the angular distance to a field star from the quasar components . unfortunately , there are no stars on the pc chip in our exposures bright enough to yield reliable astrometry . there
do exist two field stars that appear in the wf2 and wf3 chips during our pc exposures .
however , even though the _ iraf _ metric routine will yield astrometry for these relative to the quasar in the pc chip , such measurements would not yield unambiguous information about the scale of the pc chip , since the wf3 scale would also be included in the measurement .
in addition , most of the other investigators either had to mask the relevant comparison stars because of their brightness ( racine 1991 ) , did not have it in their field of view ( crane _ et al .
_ , 1991 ) , or had our problem , that the stellar image was on a different chip than the quasar ( rix _ et al . _ 1992
so measuring such an angular distance would not help to compare our results with other measurements .
although for the purpose of comparing datasets it is preferable to measure distances , for modeling the system it is more useful to have the positions given in a particular coordinate system .
therefore , we also have listed in table 2 the right ascension and declination displacements relative to component a , in units of arcsec .
note that the errors are the same as the errors in the distances of table 3 ; that is , we do not include the errors incurred because of the uncertainty in determining the coordinate system .
the main purpose of these observations was to place limits on or , with luck , detect the size and brightness of the broad - line region ( blr ) possibly associated with the quasar .
the idea was that even if the unlensed blr of the quasar were slightly smaller than the pixel size and its brightness below the detection limit , the blr would be stretched and magnified due to the lensing magnification , so that one could possibly see arc - like structures .
for this purpose it was best to use the combined images ; however , one must temper any enthusiasm for the resulting signal - to - noise and resolution gains by the awareness of possible spurious signals created by the combination process .
we employed two methods to examine possible tangential extension of the broad - line region .
the first consisted of a mostly qualitative examination of quasar components and their colors ; the second consisted of a quantitative examination of what sort of signal we could possibly detect around the quasars .
upon examining the images visually , it seems clear that there is no measurable tangential extension .
this absence is evident in figure 4 , which depicts a sum of all the components for the f336w exposure in the pc chip . in an attempt to detect a tangentially extended component in the image
, we rotated our image of each component into a coordinate system such that the center of the galaxy was along the horizontal to the left , and the clockwise direction tangent to the galaxy was vertically down .
then we combined all of the images , scaling each to the same magnitude before combination .
presumably , such a depiction would enhance any tangential extension , which would appear vertically in figure 4 .
the image here is nearly circularly symmetric , but the contours are @xmath34 longer in the vertical direction than in the horizontal direction .
similar images created for the other filter / chip combinations showed approximately the same result .
we know that there is asymmetry in the psf ; however , since we rotate the image of each component before summing , we expect that this asymmetry is suppressed to some degree .
to evaluate the completeness of this suppression , we perform the same rotation and summation , using as the base image a field star of about 17 mag on the wf2 chip ( thus , instead of starting from four separate images we start each rotation from the same image ) . in the f336w filter
, the resulting asymmetry amounted to @xmath9 5% excess in the vertical direction ; in the f300w filter , the asymmetry was 23% .
we also performed the rotation and summation on synthetic psfs produced by the program tinytim ( krist 1992 ) .
the contours determined for these combined images are also long in the vertical direction , by @xmath9 24% .
therefore , we conclude that the slight ellipticity observed in figure 4 can not be considered a reliable signature of structure .
in addition to searching in general for tangential extension , we can search specifically for features due to ly-@xmath0 emission . since f336w would transmit the red - shifted ly-@xmath0 emission approximately three to four times more efficiently than would the f300w filter
, we can difference the two filters in an attempt to pick out such features .
because the images are brighter in general in the f336w filter , we found it best to scale the images accordingly before differencing ; this scaling was designed to make the cores of the components cancel , so one could pick out features surrounding the core .
as the components differ somewhat in color ( c and d being redder than a and b ) , we performed this scaling and subtraction separately for each component .
figure 5 depicts this subtraction for the images in both chips .
note that in the pc exposures , the a and b components reveal a slight positive feature downwards and leftwards of each , at an angle of forty - five degrees from the negative - x axis and at a distance from the core of approximately 0.1 arcsec .
there are at least three explanations of this feature to explore .
first , it could be an image of a real blr associated with the quasar .
second , it could be a result of the dithering process .
finally , it could result from a difference in the diffraction patterns for the two filters .
we favor this third explanation . if the features were images of a ly-@xmath0 region associated with the quasar , the region would have about @xmath35 of the integrated brightness of the quasar .
this explanation of the features could be confirmed if we could determine whether the features existed around the c and d components , as well ; since these components are expected to have the opposite parity of a and b , a feature due to lensing would appear on the opposite side in c and d than in a and b. that is , a lensed feature appearing on the lower left in the a component would appear on the upper right of the c component .
unfortunately , the noise around the c and d components is too severe to allow such a test .
however , there is another property we would expect if the feature were due to a ly-@xmath0 region near the quasar .
we would guess that the features would be oriented differently with respect to their associated components .
in particular , we would guess that each feature would lie in a direction nearly perpendicular to the line connecting its associated component and the galaxy .
the direction towards the galaxy at a differs from that at b by thirty degrees , so we would expect the line connecting the component and the feature at a to differ from that at b by a similar amount
. however , both of the observed features are forty - five degrees from the x - axis , which argues against their interpretation as a blr .
since the dithering occurs along the same direction ( forty - five degrees from the x - axis ) , we suspected that perhaps a problem with accounting for the dither had caused the feature .
thus , we experimented with putting artificial errors in the dithering shift ; that is , we purposely misaligned the images before recombining them .
if the feature had been caused by an error in the dithering , we would expect the misalignments to affect the feature .
although we tried misalignments as drastic as half a pixel , they had no effect on the feature .
therefore , we conclude that the dithering process most likely did not cause the feature .
the remaining explanation is that the diffraction patterns for the two filters differ enough that their difference causes a visible residual . as there are no bright stars in our exposures on the pc chip with which to test this proposition , we resort to the synthetic psfs provided by the program tinytim .
we subtract synthetic psfs in the f336w and f330w filters ( subsampled by a factor of two , as the image in figure 5 is ) . because the diffraction ring patterns are of slightly different sizes in the two filters , one obtains a pattern of rings in the differenced image . examining the first positive ring from the center
, we discovered that it was about the same distance from the center and about the same brightness as the features in the real images ; interestingly , it had an asymmetry that caused it to be brighter on the lower left than elsewhere along the ring .
we avoid taking this result too literally , but we do conclude that the differing diffraction patterns of the two filters could cause the feature we see .
even so , we can use this feature to place an upper limit on the brightness of compact regions that could exist without our detecting them . for the a component
, the accompanying feature has an integrated brightness about 6.9 magnitudes dimmer than the component ; the corresponding figure for the b component is about 7.5 magnitudes .
thus we can state with confidence that there is no ly-@xmath0 feature further than 100 mas from the quasar core and brighter than about 23.5 magnitudes in the f336w band . using the lensing model of rix _
et al . _ ( 1992 ) , this would preclude any features further than 20 mas ( @xmath1 pc , assuming @xmath2 ) from the quasar core in the source plane and brighter than 25 mag before magnification .
we can place another quantitative limit on the size and brightness of a tangentially extended image surrounding the quasar components .
to do so , consider one of the quasar components . draw a circle centered on it .
evaluate the f336w surface brightness at each point around that circle . this evaluation yields brightness as a function of angle .
if the coordinates are such that the galaxy is in the direction @xmath36 , and the quasar component were tangentially extended , the brightness function would have bumps at @xmath37 and @xmath38 degrees .
upon examination of such plots , we see no such bumps . to place a limit on the size of bump we could measure ,
we calculate the standard deviation @xmath39 of the surface brightness values around the circle about their mean .
then we require some signal - to - noise ratio ( s / n ) for a meaningful result , and our limit becomes ( s / n)@xmath39 .
figure 6 reveals the results of this procedure for the images in the pc chip and the f336w filter , which is most likely to reveal ly-@xmath0 structure .
the horizontal axis is the radius in mas , and the vertical axis is the standard deviation in flux units .
depending on the assumed lensing model , we may divide the radius by the appropriate magnification for each component to obtain the limits appropriate for the source plane ; thus , we can use the lensing magnification to `` squeeze '' the plot radially .
for example , a limit of @xmath40 erg @xmath41 s@xmath42 arcsec@xmath43 at 100 mas , given a magnification of 5 , would become a limit of @xmath40 erg @xmath41 s@xmath42 arcsec@xmath43 at 20 mas .
however , we are obliged to multiply the vertical axis by our desired signal - to - noise ratio ; thus we must slide the plot up ( on a logarithmic axis ) . in the example
above , our limit of @xmath40 erg @xmath41 s@xmath42 arcsec@xmath43 at 20 mas , given a required ( s / n ) of 5 , would yield , a limit of @xmath44 erg @xmath41 s@xmath42 arcsec@xmath43 at 20 mas .
the results of such a squeezing and sliding appear in figure 7 for the model 2a of rix _
et al . _ ( 1992 ) , with a signal - to - noise ratio of 5 .
we compare our two methods by noting that the brightest pixels in the features associated with the a and b components have a surface brightness of about @xmath40 erg @xmath41 s@xmath42 arcsec@xmath43 and are located about 100 mas from each component .
this position is indicated as a cross in figure 6 .
thus , the feature is just at our calculated threshold . with the conservative requirement of a 5@xmath39 result , as in figure 7
, this feature would be well below the noise . in figure 7
, we have included the source plane position of the feature on every plot ; for this purpose we used its position as determined by its appearance next to the b component . that this feature is somewhat submerged in our predicted noise
encourages our belief that figure 6 provides a reasonable upper limit for the brightness of other possible features .
finally , we can compare our results with other observations of regions surrounding high - redshift quasars , in particular those of bremer _
et al . _ ( 1992 ) .
these investigators found extended ly-@xmath0 emission around two @xmath45 quasars , at fluxes of approximately @xmath46 erg @xmath41 s@xmath42 and extensions of about 4 arcsec . from our results , any emission at this level over such a large scale can be excluded for quasar q2237 + 0305 .
a final consideration is the possibility of detecting a faint fifth component of the quasar .
predictions of the brightness of this component relative to a are uncertain by orders of magnitude ; for example , the results of schneider _ et al . _ ( 1988 ) indicate @xmath47 , kent & falco ( 1988 ) find @xmath48 , and rix , schneider & bahcall ( 1992 ) find @xmath49 for all their models . in this regard ,
it is worth noting that the galaxy core is 6.5 mag dimmer than component a , and is close to the threshold of detectability .
since there is not the slightest indication of a fifth component in the image , it must be at least 6.5 mag dimmer than component a , in our filters f336w and f300w .
our results do not appear to support the findings of racine ( 1991 ) , who detected a faint fifth component 4.5 mag dimmer than a in the r and i bands .
however , a central fifth component could be brighter with respect to a in longer wavelength bands as a result of reddening due to dust
. the fifth component would shine through the very center of the lensing galaxy , making any extinction estimates wildly uncertain .
in this paper , we have presented the results of hubble space telescope exposures on the wfpc2 camera of quasar q2237 + 0305 ; specifications of the images are in table 1 and we display the image configuration in figure 1 .
the f336w and f300w filters were chosen because the relatively red galaxy bulge through which we see the quasar components is faint in the uv , and because one of the filters is very close to the redshifted ly-@xmath0 line of the quasar ( at about 3270 ) , so that it is plausible that any spatially extended structures in the quasar environment would emit in this wavelength regime .
we have determined : 1 . the photometry for the four components in the f336w and f300w bands , which is listed in table 2 .
the relative brightnesses of the components are known to vary with time due to microlensing . at the time of our observations ,
component a was the brightest , followed by component b which was fainter by 0.3 mag .
components c and d were fainter than component a by about 1.3 and 1.4 mag , respectively . on the timescale of 34 hours , we can state that we observe no variation in component a of greater than 0.02 mag . for the other components , over a period of 10 hours they remain constant to within about 0.05 mag . in the final five hours
there is some evidence for variation of component d of about 0.1 mag .
2 . the astrometry of the four components and the galaxy core , which is listed in tables 3 and 4
. the ( 1@xmath39- ) uncertainty of these astrometric measurements are about 1.5 mas .
we found the system to be consistently larger ( by between 0.1% and 2% ) than previous studies have found .
we discuss above why we think our values are reliable .
3 . the existence of a feature near the a and b components which is bright in f336w , as shown in figure 5 with a subtraction of f300w from f336w .
we conclude that this signal is probably an artifact of the differing diffraction patterns of the two filters .
however , we use the brightness of the feature to obtain an upper limit on the brightness of any real ly-@xmath0 regions , plotted as the crosses on figures 6 and 7 .
an estimate of the upper limits these images yield for the brightness of any extended image near the quasar , as a function of angular distance from the quasar in the source plane .
these upper limits depend on the lensing model and the desired signal - to - noise .
they are plotted in figures 6 and 7 .
an upper limit on the brightness of central fifth component in our band which is 6.5 mag fainter than component a.
we wish to thank ray lucas , jean surdej , and brad whitmore of the space telescope science institute for useful help and advice .
in addition , we would like to thank the authors of the drizzle algorithm , andy fruchter of stsci and richard hook of st - ecf .
we acknowledge financial support from stsci grant go-05937.01 - 94a.a01 .
bremer , m. n. , a. c. fabian , w. l. w. sargent , c. c. steidel , a. boksenberg , r. m. johnstone : 1992 , _ mon . not .
r. astro .
soc . _ * 258 * , 23p .
corrigan , r. t. , m. j. irwin , j. arnaud , g. g. fahlman , j. m. fletcher , _ et al . _ : 1991 , _ astron .
j. _ * 102 * , 34 .
crane , p. , r. albrecht , c. barbieri , j. c. blades , a. boksenberg _ et al . _ : 1991 , _ astrophys .
j. _ * 369 * , l59 .
falco , e. e. , j. lehr , r. a. perley , j. wambsganss , m. v. gorenstein : 1996 , _ astron .
j. _ , * 112 * , 897 .
fruchter , a. and hook , r. : 1997 , _ pasp _ , submitted .
huchra , j. , m. gorenstein , s. kent , i. shapiro , g. smith , _ et al .
_ : 1985 , _ astron
. j. _ * 90 * , 691 .
holtzmann , j. , _
et al . _ : 1995 , _ publ .
pac . _ * 107 * , 156 .
irwin , m. j. , p. c. hewett , r. t. corrigan , r. i. jedrzejewski , r. l. webster : 1989 , _ astron .
j. _ * 98 * , 1989 .
kent , s. m. and e. e. falco : 1988 , _ astron .
j. _ * 96 * , 1570 .
krist , j. 1992 , the tiny tim user s manual ( baltimore : stsci ) lewis , g. f. , m. j. irwin , p. c. hewett : 1995 , _ astrophys .
j. _ , submitted august 1995 .
ostensen , r. , s. refsdal , r. stabell , j. teuber , p. i. emanuelsen , _
et al . _ : 1996 , _ astron .
astrophys . _ * 309 * , 59 .
racine , r. : 1991 , _ astron .
j. _ * 102 * , 454 .
rix , h .- w .
, d. p. schneider , j. n. bahcall : 1992 , _ astron .
j. _ * 104 * , 959 .
schneider , d. d. , e. l. turner , j. e. gunn , j. n. hewitt , m. schmidt , and c. r. lawrence : 1988 , _ astron .
j. _ * 95 * , 1619 .
wambsganss , j. , b. paczyski : 1994 , _ astron .
j. _ * 108 * , 1156 .
wambsganss , j. , b. paczyski , p. schneider : 1990 , _ astrophys .
j. _ * 358 * , l33 .
whitmore , b. , i. heyer , s. baggett : 1996 , stsci instrument science report wfpc2 96 - 04 .
witt , h.j .
, and mao , s. : 1994 , _ astrophys . j. _ * 429 * , 66 .
yee , h. k. c. : 1988 , _ astron .
j. _ * 95 * , 1331 .
yee , h. k. c. , m. m. de robertis : 1991 , _ astrophys .
j. _ * 381 * , 386 .
ccccc 1 & wf3 & f300w & 1100 & 02 & wf3 & f300w & 1100 & 03 & wf3 & f300w & 1300 & 2504 & wf3 & f300w & 1300 & 2505 & wf3 & f336w & 1100 & 06 & wf3 & f336w & 1200 & 07 & wf3 & f336w & 1200 & 08 & wf3 & f336w & 1100 & 2509 & wf3 & f336w & 1200 & 25010 & wf3 & f336w & 1200 & 25011 & pc & f336w & 700 & 012 & pc & f336w & 700 & 013 & pc & f336w & 700 & 014 & pc & f336w & 800 & 015 & pc & f336w & 700 & 016 & pc & f336w & 700 & 25017 & pc & f336w & 800 & 25018 & pc & f336w & 800 & 25019 & pc & f336w & 800 & 25020 & pc & f300w & 700 & 021 & pc & f300w & 800 & 022 & pc & f300w & 700 & 023 & pc & f300w & 700 & 25024 & pc & f300w & 800 & 25025 & pc & f300w & 800 & 250 ccccc a & 16.65 @xmath50 0.02 & 16.85 @xmath50 0.03 & 0.000 @xmath50 0.0015 & 0.000 @xmath50 0.0015 b & 16.93 @xmath50 0.04 & 17.10 @xmath50 0.03 & -0.671 @xmath50 0.0015 & 1.697 @xmath50 0.0015 c & 17.95 @xmath50 0.03 & 18.24 @xmath50 0.04 & 0.634 @xmath50 0.0015 & 1.210 @xmath50 0.0015 d & 18.10 @xmath50 0.03 & 18.41 @xmath50 0.03 & -0.866 @xmath50 0.0015 & 0.528 @xmath50 0.0015 galaxy core & n / a & n / a & -0.081 @xmath50 0.010 & 0.937 @xmath50 0.010 +
|
we present and analyze observations of the quadruple lensed quasar q2237 + 0305 , obtained with the _ hst _ wfpc2 camera in the f336w and f300w filters .
twenty - five exposures were performed within 15 hours real time on 3 november 1995 . on a timescale of 34 hours , we observe no variation in component a of greater than 0.02 mag .
the other components remain constant over a period of 10 hours to within about 0.05 mag . in the final 5 hours
there is some evidence ( not conclusive ) for variation of component d by about 0.1 mag .
the exposures indicate that component a is brighter than component b by about 0.3 mag .
components c and d are fainter than component a by about 1.3 and 1.4 mag , respectively .
our results place an upper limit on any fifth ( central ) component of 6.5 mag fainter than component a. we determine the astrometric properties of the lens system , using only the exposures of the higher resolution planetary camera chip .
we measure the relative distances of the four components with high accuracy .
our values are systematically larger than those of other investigators ( by 0.1% to 2.0% ) .
we discuss the reasons why we believe our results are reliable . the f336w filter had been chosen for the observations because it corresponds to the redshifted ly-@xmath0 line of the quasar .
this filter might have allowed us to see extended ly-@xmath0 emission from the broad - line region ( blr ) of the quasar as ly-@xmath0 arcs , and hence to determine the physical size of the blr . however , the quasar components in this filter are consistent with a point source .
we conclude that there can not be any ly-@xmath0 feature in the image plane brighter than about 23.5 mag in f336w and further from the quasar core than 100 mas .
according to a lensing model by rix , scheider & bahcall ( 1992 ) , this would preclude any such features in the source plane further than 20 mas ( @xmath1 pc , assuming @xmath2 ) from the quasar core and brighter than 25 mag before magnification .
|
formation of localized coherent structures during the interaction of intense laser pulses with plasmas is an important topic of the laser plasmas research , which is vital for diagnostic purposes in the experiments with laser ion acceleration , the fast ignition of controlled thermonuclear fusion , the investigation of warm dense matter , high energy density phenomena , and laboratory astrophysics ( see article @xcite and references therein ) .
when the laser pulse interacts with a homogeneous plasma region , we expect the pulse to penetrate inside the plasma and propogate with minor energy losses in case of underdense plasma .
quantitavely speaking , we expect laser pulse to penetrate if @xmath0 , where @xmath1 is the laser carrier frequency and @xmath2 is the electron plasma frequency .
the larger the @xmath3 ratio is , the lower is the laser pulse depletion rate .
however , eventually a finite duration laser pulse completely depletes due to stimulated raman scattering , various pulse filamentation instabilities and transformation to various localized coherent structures @xcite .
these processes lead to the formation of langmuir waves , electromagnetic solitons , and electron vortices retaining electromagnetic energy . besides self - focusing channels and langmuir waves , we consider two classes of coherent structures that happen to form in laser plasmas simulations . in two - dimentional plasmas , we can distinguish solitons @xcite and electron vortices @xcite .
electron vortices are coherent localized structures , which keep quasistatic magnetic flux locked inside the electron cavity .
solitons enclose the electromagnetic wave inside the density gap .
though having slightly different conditions of generation , in 3d case these structures are usually combined in one , posessing some features from both of them @xcite .
on an electron timescale @xmath4 , these coherent structures are quasistationary . in the case of immobile ions ,
an analytical stationary solution exists , see section ii .
however , on a proton timescale @xmath5 , we expect these vortices to evolve , as the uncompensated proton charge results in coloumb explosion of the whole vortex @xcite
. hereafter , we will call it `` vortex explosion '' . here
we investigate the structure and evolution of the relativistic electron vortices .
we present a simple analytical vortex model which illustrates the main features of realistic vortex structure .
we carry out two - dimensional ( 2d ) particle - in - cell ( pic ) simulations using the remp code based on the density decomposition scheme @xcite .
we discuss in detail a number of effects occurring during the evolution of electron vortices : the coulomb explosion of the uncompensated proton core of the vortex , multistream processes that lead to shell - like structure formation , instability at the vortex boundary leading to the edge field intensification and transformation into smaller vortices .
the paper is organized as follows . in the next section ,
the analytical solution for stationary vortex structure is presented . in section iii
, we describe the simulation parameters . in section iv
, we discuss simulation results of the evolution of the electron vortex , addressing the observed effects . in the concluding section
, we summarize the results obtained .
before discussion of the simulations of the vortex structure evolution , we present theoretical estimates for the analytical structure of the electron vortex in the magnetized collisionless plasma region . we start from maxwell equations @xmath6 and relativistic plasma motion equations @xmath7 here , current @xmath8 and charge density @xmath9 can be written as @xmath10 and relativistic gamma - factors for electrons and ions are @xmath11 all equations are in dimensionless form .
hereafter , we will measure spatial parameters in @xmath12 , temporal in @xmath13 , densities in critical densities @xmath14 , electromagnetic fields in @xmath15 , where @xmath16 is electron mass , @xmath17 is the absolute value of electron charge , @xmath18 is the speed of light in vacuum .
in addition , we note that in dimentionless units @xmath19 becomes @xmath20 and velocities are measured in fractions of light speed in vacuum . changing to cylindrical coordiantes
@xmath21 , we further assume asymmetry and homogeneity along the @xmath22-axis , i.e. @xmath23 .
we denote vector components by subscripts corresponding to respective coordinates , e.g. , for electric field e components are @xmath24 . besides , we assume that @xmath25 then eq . ( [ eq : divb ] ) becomes identity and from eqs .
( [ eq : et])-([eq : vit ] ) together with definitions eqs .
( [ eq : j])-([eq : n ] ) we obtain @xmath26 we also assume that plasma is neutral at infinity and the total charge is zero : @xmath27 now we assume @xmath28 , @xmath29 , @xmath30 , @xmath31 , @xmath32 and neglect eqs ( [ eq : virt - cyl])-([eq : vitt - cyl ] ) .
then eqs ( [ eq : ert - cyl ] ) , ( [ eq : bzt - cyl ] ) and ( [ eq : vett - cyl ] ) become identities and from eqs ( [ eq : ett - cyl ] ) , ( [ eq : err - cyl ] ) and ( [ eq : vert - cyl ] ) we obtain @xmath33 where prime denotes derivative with respect to @xmath34 , @xmath35 . neglecting the last term in eq ( [ eq : ebv ] ) in the limit of non - relativistic electron motion , @xmath36 , and resolving this system wit respect to @xmath37 , @xmath38 and @xmath39
we obtain @xmath40 where we assume @xmath41 and @xmath42 .
the system eqs ( [ sol : vet])-([sol : bz ] ) gives the solution provided that a reasonable profile is set for @xmath43 .
there could be infinite number of vortex shapes , however not all of them are stable . here
we do not consider the problem of stability .
we assume the dependence of @xmath44 on @xmath34 as @xmath45 it corresponds to a situation when the inner region of the vortex is represented by the homogeneous proton region , and to some extent in cyllindrical geometry @xmath44 grows lineary with radius , until the electric field is being shielded by outside plasmas . then from eq ( [ sol : bz ] ) we obtain @xmath46 with the following asymptotic at @xmath47 @xmath48 we assume that @xmath39 vanishes at infinity , which leads to @xmath49 in addition , we impose the condition on @xmath39 to have a local maximum at @xmath50 .
then @xmath51 which ensures positive number under the root sign in eq ( [ sol : b0 ] ) . for @xmath52 , see the profiles in fig .
[ fig:1 ] .
@xmath44 , @xmath39 , @xmath38 and @xmath37 in the vortex for @xmath52 .
, width=302 ] this simple vortex model highlights the main features of electron vortex - electron cavitation , localization of the magnetic flux inside the electron cavity and electron density increase in the vicinity of vortex boundary @xcite .
though we are going to consider mildly relativistic electron vortices , the above consideration of nonrelativistic case fairly reproduces the main features we are going to observe in the next section .
in order to investigate the processes that occur during the electron vortex evolution , we conduct a series of 2d pic simulations . in our simulations , we set the homogeneous square of cold electron - proton plasma with the density about @xmath53 and size of @xmath54 with zero temperatures of protons and electrons . for the sake of simplicity , we form a symmetric electron vortex , artificially generating the localized magnetic field distribution during a number of initial timesteps . as we are not interested in the physical reasons of vortex formation , but would like to discuss the evolution of these coherent structures , this method is suitable for our purposes .
thus , we add the following profile of the magnetic field in our numerical scheme @xmath55 where @xmath56 is the maximum amplitude of the added magnetic field profile , @xmath57 is the typical vortex width , @xmath58 and @xmath59 are initial coordinates of the vortex center .
we use this addition to the magnetic field calculations each timestep during the first 5 time units . for @xmath60 and @xmath61 , at time equals to 10 , we observe a quasistatic vortex structure with the maximum magnetic field amplitude of @xmath62 and @xmath63 in diameter .
these parameters are chosen in the way to satisfy the conditions for `` magnetized electrons and unmagnetized ions '' approach that is used in the theoretical description of the electron vortex evolution @xcite .
quantitavely speaking , we set the value of the maximum magnetic field inside the cavity satisfying the relation @xmath64 which means that the field energy is enough to expel the electrons out from the axis region of the vortex ( left inequality ) , but not strong enough to influence proton dynamics significantly @xcite .
the computational grid is @xmath65 with 32 nodes per each cell .
the initial particle - in - cell number is equal to 16 .
the total number of particles is about @xmath66 .
the integration timestep equals to 0.0125 time units .
the total time of simulations is 300 time units .
as we expected , the simulation with immobile ions justifies that there is a quasistationary configuration of electron fluid rotating in the localized magnetic field , which is similar to one described in section ii .
however , on a proton timescale @xmath67 , we find out that these vortices expand due to noncompensated positive charge enclosed in the electron cavity . here
we consider how the vortex evolves during the simulation time . in case of @xmath68 ( see figure [ fig : firstdens ] ) we observe an axisymmetric electron cavity with a substantinal increase of the electron density around the shock - like vortex boundary , up to @xmath69 .
a ) , e ) @xmath39 component of the magnetic field and f ) @xmath70 component of the electric field for t=50.,title="fig : " ] b ) , e ) @xmath39 component of the magnetic field and f ) @xmath70 component of the electric field for t=50.,title="fig : " ] c ) , e ) @xmath39 component of the magnetic field and f ) @xmath70 component of the electric field for t=50.,title="fig : " ] d ) , e ) @xmath39 component of the magnetic field and f ) @xmath70 component of the electric field for t=50.,title="fig : " ] e ) , e ) @xmath39 component of the magnetic field and f ) @xmath70 component of the electric field for t=50.,title="fig : " ] f ) , e ) @xmath39 component of the magnetic field and f ) @xmath70 component of the electric field for t=50.,title="fig : " ] the magnetic flux is concentrated inside the cavity .
these facts give us a description of the evolution of the vortex using the snowplow model @xcite .
let us assume that the magnetic flux inside the cavity is constant in time ( i.e. , we assume an adiabatic approximation ) .
then , we can suppose that the whole dynamics is determined by the magnetic field pressure on a walls of axisymmetric vortex .
we write @xmath71 where @xmath72 is the mass of plasma pushed to the radius @xmath73 by the magnetic flux and @xmath74 , where index @xmath75 denotes to the initial values of the magnetic field and vortex radius .
we can rewrite this equation in dimentionless variables @xmath76 and @xmath77 : @xmath78 which is the same as in @xcite .
the solution could be written as follows : @xmath79 for which the asymptotics in @xmath80 limit are : @xmath81 , @xmath82 , and @xmath83 ( note the misprint in @xcite ) . comparing the obtained asymptotics with simulations , we plot the figure [ fig : asympt ] .
we observe quite a good agreement of the radius evolution between theory and simulations , despite the simplification of theoretical model in comparison with the simulation setup .
, in comparison to the power law from snowplow model @xmath84 .
] it is also interesting to discuss the proton phase space distributions .
figure [ fig : firstphase ] shows the @xmath85 and @xmath86 phase space distributions of protons . at @xmath87 phase plot we may see two antisymmetric regions , corresponding to coloumb explosion in the external plasma the region where @xmath88 correspond to the coloumb explosion of the cyllindrical noncompensated positive charge of protons .
the subsequent decrease of the proton momentum is due to the finite velocity of interacting electromagnetic field propagation .
while exploding due to the coloumb repulsion , more and more protons are involved in the process , starting to move radially .
the @xmath89 figure shows that there are numerous proton shells that move radially in an axisymmetric way .
the shell - nature of the dynamics leads to a number of multistream instabilities that come into play during the simulation .
a ) and b ) @xmath86 for t=50.,title="fig : " ] b ) and b ) @xmath86 for t=50.,title="fig : " ] at figure [ fig : instability ] , we observe that the @xmath90 derivative tends to infinity in @xmath87 phase plots .
physically , it means that there are multiple proton shells located at the same spatial coordinate but having the broad range of energies .
this reveals a rise of a multistream instability , leading to the enhanced proton energy gain rate .
we plot the same figures for @xmath91 . here
, we see that the dense proton ring has been divided into the three most energetic parts , which are seen at both density profile and energy spectrum of protons .
it is also noticeable that there are some phase space folds revealing on a @xmath87 plot , that also correspond to these independently moving proton circles .
a ) ) , e ) @xmath85 phase plot for protons - correspond to the moment of @xmath92 at @xmath93 , b ) , d ) , and f ) - the same for the moment of manifestation of the multistream instability , @xmath91 , g ) proton energy spectrum at @xmath91 , three peaks corresponding to the most energetic ion associations are seen.,title="fig : " ] b ) ) , e ) @xmath85 phase plot for protons - correspond to the moment of @xmath92 at @xmath93 , b ) , d ) , and f ) - the same for the moment of manifestation of the multistream instability , @xmath91 , g ) proton energy spectrum at @xmath91 , three peaks corresponding to the most energetic ion associations are seen.,title="fig : " ] c ) ) , e ) @xmath85 phase plot for protons - correspond to the moment of @xmath92 at @xmath93 , b ) , d ) , and f ) - the same for the moment of manifestation of the multistream instability , @xmath91 , g ) proton energy spectrum at @xmath91 , three peaks corresponding to the most energetic ion associations are seen.,title="fig : " ] d ) ) , e ) @xmath85 phase plot for protons - correspond to the moment of @xmath92 at @xmath93 , b ) , d ) , and f ) - the same for the moment of manifestation of the multistream instability , @xmath91 , g ) proton energy spectrum at @xmath91 , three peaks corresponding to the most energetic ion associations are seen.,title="fig : " ] e ) ) , e ) @xmath85 phase plot for protons - correspond to the moment of @xmath92 at @xmath93 , b ) , d ) , and f ) - the same for the moment of manifestation of the multistream instability , @xmath91 , g ) proton energy spectrum at @xmath91 , three peaks corresponding to the most energetic ion associations are seen.,title="fig : " ] f ) ) , e ) @xmath85 phase plot for protons - correspond to the moment of @xmath92 at @xmath93 , b ) , d ) , and f ) - the same for the moment of manifestation of the multistream instability , @xmath91 , g ) proton energy spectrum at @xmath91 , three peaks corresponding to the most energetic ion associations are seen.,title="fig : " ] g ) ) , e ) @xmath85 phase plot for protons - correspond to the moment of @xmath92 at @xmath93 , b ) , d ) , and f ) - the same for the moment of manifestation of the multistream instability , @xmath91 , g ) proton energy spectrum at @xmath91 , three peaks corresponding to the most energetic ion associations are seen.,title="fig : " ] after some time of vortex evolution , we observe the effect of bending of the electric current associated with the electron vortex . besides , the vortex boundary starts to break into a group of small electron vortices .
figure [ fig : vortices ] shows electron and proton densities , @xmath22-component of the magnetic field and @xmath94-component of electric current .
this effect may correspond to the manifestation of various instabilities , having some similarity with the hydrodynamical instability of vortex boundary @xcite and the electromagnetic kelvin - helmholtz instability @xcite .
it is also important to note that not only electron dynamics is affected by this instability , as we observe the anisotropy in proton acceleration as well , see figure [ fig : vortices]e )
. a ) , d ) x - component of electric current @xmath95 , and e ) proton phase plot in @xmath86 coordinates for @xmath96 .
a group of small vortices along the vortex boundary are seen.,title="fig : " ] b ) , d ) x - component of electric current @xmath95 , and e ) proton phase plot in @xmath86 coordinates for @xmath96 .
a group of small vortices along the vortex boundary are seen.,title="fig : " ] c ) , d ) x - component of electric current @xmath95 , and e ) proton phase plot in @xmath86 coordinates for @xmath96 .
a group of small vortices along the vortex boundary are seen.,title="fig : " ] d ) , d ) x - component of electric current @xmath95 , and e ) proton phase plot in @xmath86 coordinates for @xmath96 .
a group of small vortices along the vortex boundary are seen.,title="fig : " ] e ) , d ) x - component of electric current @xmath95 , and e ) proton phase plot in @xmath86 coordinates for @xmath96 . a group of small vortices along the vortex boundary are seen.,title="fig : " ]
in this article , we presented analytical description and computer simulation results on the stationary state and evolution of one type of coherent structures that are observed in laser plasmas - electron vortices . these structures are often seen in 2d pic simulations of various laser - plasma configurations .
while being quasistatic in the immobile ion approach , they do evolve on the ion timescale for the two - species simulations .
we observed the manifestation of various stream instabilities , which enhance maximum energies of protons and lead to shell - like structure of proton distribution .
we have obtained @xmath97 protons accelerated by the combination of coloumb explosion of the mildly relativistic vortex with @xmath98 t @xmath99 and multistream instability . at final stages of the vortex evolution ,
ion rotational motion becomes significant in addition to their expansion .
this leads to disintegration of the vortex boundary into the association of a small electron vortices , making an impact on a ion acceleration as well .
the obtained results will be useful for developing the theory describing the electromagnetic turbulence in relativistic plasmas @xcite and diagnostics of the results of the experiments with petawatt - class laser systems , that are being build nowadays @xcite .
this work was supported by the eli project no . cz.02.1.01/0.0/0.0/15_008/0000162 .
we also would like to acknowledge the support from russian foundation for basic research ( grant no . 15 - 02 - 03063 ) .
l. romagnani , a. bigongiari , s. kar , s. v. bulanov , c. a. cecchetti , m. galimberti , t. zh .
esirkepov , r. jung , t. v. liseykina , a. macchi , j. osterholz , f. pegoraro , o. willi , and m. borghesi , phys .
lett . * 105 * , 175002 ( 2010 ) s. v. bulanov , f. califano , g. i. dudnikova , t. zh .
esirkepov , i. n. inovenkov , f. f. kamenets , t. v. liseikina , m. lontano , k. mima , n. m. naumova , k. nishihara , f , pegoraro , h. ruhl , a. s. sakharov , y. sentoku , v. a. vshivkov , v. v. zhakhovskii , `` relativistic interaction of laser pulses with plasmas '' , reviews of plasma physics , vol .
22 , publisher : kluwer academic / consultants bureau , new york , editors : v. d. shafranov , pp.227 - 335 s. v. bulanov , m. lontano , t. zh .
esirkepov , f. pegoraro , and a. m. pukhov , phys .
lett . * 76 * , 3562 ( 1996 ) ; n. m. naumova , j. koga , k. nakajima , t. tajima , t. zh .
esirkepov , s. v. bulanov , and f. pegoraro , phys .
plasmas * 8 * , 4149 ( 2001 )
esirkepov , k. nishihara , s. v. bulanov , and f. pegoraro , phys .
* 89 * , 275002 ( 2002 )
|
the interaction of high intensity laser radiation with underdense plasma may lead to the formation of electron vortices . though being quasistationary on an electron timescales , these structures tend to expand on a proton timescale due to coloumb repulsion of ions . using a simple analytical model of a stationary vortex as initial condition ,
2d pic simulations are performed .
a number of effects are observed such as vortex boundary field intensification , multistream instabilities at the vortex boundary , and bending of the vortex boundary with the subsequent transformation into smaller electron vortices .
keywords : relativistic laser plasmas , electron vortices , quasi - static magnetic field , particle - in - cell simulation
|
a point spread function ( psf ) is a continuous two - dimensional probability - distribution function which describes the scattering pattern of photons from a point source ( star ) . encoding a psf as a continuous mathematical function works well for many ground - based astronomical observations due to the significant blurring caused by turbulence in the earth s atmosphere and dome / telescope seeing .
ground - based psfs are typically characterized by having a lot of the power in their spatial - frequency distributions at low spatial frequencies .
space - based psfs frequently have significant amounts of power at higher spatial frequencies due to the lack of blurring caused by atmospheric turbulence .
adaptive optics can produce psfs with characteristics found in both uncorrected ground - based psfs and space - based psfs : low - spatial - frequency features ( e.g. , broad halos ) are frequently combined with high - spatial - frequency features ( e.g. , due to segmented mirrors ) .
some psf - fitting stellar photometric reduction programs describe the psf as a combination of continuous mathematical functions and a residual matrix which contains the difference between the mathematical model of the psf and an observed ( `` true '' ) psf .
this artificial breaking of the psf into analytical and discrete components is not without mathematical risk .
such residuals can have small features which are described with higher spatial frequencies than are present in the actual observational data a problem that can usually be mitigated by sampling residuals at higher spatial resolutions than the observational data .
what if we dispose of the use of continuous mathematical functions to model _ any _ part of the psf and just use a matrix to describe _ all _ of the psf ? is precise and accurate stellar photometry and astrometry possible using matrix psfs with oversampled stellar image data ? if that is possible , then what extra information , if any , is required in order to do precision photometric reductions with matrix psfs on undersampled data ?
this article describes how precise and accurate stellar photometry may be obtained using psfs encoded as a matrix .
the following section derives the theoretical performance limits of psf - fitting stellar photometry and astrometry .
some of the key features of the matphot algorithm are presented in 3 .
a demonstration computer program , called @xmath0 , based on the current implementation of the matphot algorithm , is described in 4 .
simulated ccd ( charge - coupled device ) stellar observations are analyzed with @xmath0 in 5 and the performance of the matphot algorithm is compared with theoretical expectations .
concluding remarks are given in 6 .
an appendix explains box - and - whisker plots which are used extensively in this article .
a point response function , @xmath1 , is the convolution of a point spread function , @xmath2 , and a detector response function , @xmath3 : @xmath4 the psf describes the two - dimensional distribution of _ photons _ from a star _ just above the detector . _
although stellar photons are distributed as a point source above the earth s atmosphere , a stellar image becomes a two - dimensional distribution as the stellar photons are scattered by atmospheric turbulence .
the blurred stellar image is then further degraded by passage of the stellar photons through the combined telescope and camera optical elements ( such as mirrors , lenses , apertures , etc . ) .
the psf is the convolution of all these blurring effects on the original point - source stellar image .
the two - dimensional discrete ( sampled ) detector response function ( drf ) describes how the detector electronics convert stellar photons ( @xmath5 ) to electrons ( @xmath6 ) including such effects as the diffusion of electrons within the detector substrate or the reflection ( absorption ) of photons on ( in ) the gate structures of the detector electronics .
the psf is a two - dimensional probability - distribution function describing the scattering pattern of a photon .
the volume integral of the psf is one : @xmath7 ; photons , after all , have to be scattered _
somewhere_. it is important to note that since the angular extent of a psf can be quite large , the volume integral the psf _ over any given observation _ is frequently less than one due to the limited spatial coverage of the observation .
the volume integral of a prf is , by definition , one or less : @xmath8 where a value that is less than one indicates a loss of stellar photons during the detection / conversion process within the detector . while the quantum efficiency ( qe ) variations within a single detector are generally not a major problem with state - of - the - art charge - coupled devices , intrapixel qe variations can be significant with some near - infrared detector technologies currently being used in astronomical cameras ( e.g. , @xcite , @xcite ) .
a perfect drf gives a prf that is a _ sampled version _ of the psf : @xmath9 where @xmath10 pixel of the prf located at ( @xmath11 ) is the volume integral of the psf over the area of the @xmath10 pixel .
the actual limits of the above volume integral reflect the appropriate mapping transformation of the @xmath12 and @xmath13 coordinates onto the ccd pixel coordinate system .
the @xmath14 of a prf is defined as the volume integral of the _ square _ of the _ normalized _ prf : @xmath15 physically , @xmath14 is a shape parameter which describes the `` pointiness '' of a prf ; @xmath14 values range from a maximum of one ( all of the stellar flux is found within a single pixel ) to a minimum of zero ( a flat stellar image ) .
for example , cameras that are out of focus have broad psfs with @xmath14 values near zero
. a normalized gaussian psf with a standard deviation of @xmath16 pixels , @xmath17 , \label{eq : gaussian}\ ] ] that has been _ oversampled _ with a perfect drf will have a @xmath14 value of @xmath18 a critically - sampled normalized gaussian prf has a @xmath14 of @xmath19 and any prf with a @xmath14 value greater than that value ( @xmath20@xmath21 ) can be described as being undersampled .
diffraction limited optics , theoretically , give @xmath14 values that decrease ( i.e. , psfs become flatter ) with increasing photon wavelength for a fixed pixel ( detector ) size . with real astronomical cameras ,
the value of @xmath14 frequently depends on _ where the center of a star is located within the central pixel _ of the stellar image .
for example , the _ hubble space telescope _ _ ( hst ) _
wfpc2 planetary camera prf at a wavelength of 200 nm has an observed @xmath14 value of 0.084 if the prf is centered in the middle of a pc pixel or 0.063 if the prf is centered on a pixel corner ( table 6.5 of * ? ? ?
* ) ; at 600 nm the observed @xmath14 values range from 0.066 ( pixel - centered ) to 0.054 ( corner - centered ) .
the wide - field cameras of the _ hst _ wfpc2 instrument have pixels which are approximately half the angular resolution of the pc camera pixels ; stellar images on the wf cameras are undersampled and the observed range of wf camera @xmath14 values are 0.1020.120 at 200 nm and 0.0980.128 at 600 nm . the _ effective - background area _ , @xmath22 , of a prf is defined as the _ reciprocal _ of the volume integral of the _ square _ of the prf : @xmath23^{-1 } \ , .
\label{eq : beta}\ ] ] alternatively , the effective - background area ( a.k.a . _ equivalent - noise area _ or _ effective solid angle _ ) of a prf is equal to the reciprocal of the product of its @xmath14 and the square of its volume : @xmath24^{-1 } \!\!\!= \,\frac{1}{{\mbox{v}}^{\,2}\,{\mbox{$\mathsf{sharpness}$ } } } \ , .
\label{eq : beta2}\ ] ] the effective - background area of a normalized gaussian prf is @xmath25 px@xmath26 , where @xmath16 is the standard deviation in pixels ; a critically - sampled normalized gaussian prf has an effective - background area of @xmath27 px .
@xcite notes that numerical integration of a realistic ground - based stellar profile gives an effective - background area of @xmath28 instead of the value of @xmath29 for a normalized gaussian profile .
consider a ccd observation of two overlapping stellar images .
assuming that we already know the psf and the drf of the observation , a simple model of the observation will have seven parameters : two stellar intensities is , by definition , is always less than or equal to the _ measured _ stellar intensity @xmath30 . ]
@xmath31 in electrons , four coordinate values , giving the stellar positions @xmath32 in pixels , and @xmath33 which is the _ observed _ background sky level . ] in electrons ( which is assumed to be the same for both stars ) .
these observational parameters are not independent for overlapping stars in the presence of photon and ccd readout noise
. the conservation of electron flux will require that if @xmath34 increases then @xmath35 must decrease and vice versa for a given value of @xmath33 .
the most accurate photometry possible is obtained when these dependent parameters are fitted simultaneously .
any reasonable model of two overlapping stellar images will be a nonlinear function when the positions and intensities are to be determined simultaneously .
the technique of nonlinear least - squares fitting was developed to provide for the simultaneous determination of dependent or independent parameters of nonlinear model functions .
assume that we have a calibrated ccd observation with @xmath36 pixels and that @xmath37 is the number of electrons in the @xmath10 pixel which is located at the position of @xmath38 and has a measurement error of @xmath39 electrons .
let @xmath40 be an observational model of the ccd electron pixel values that has two coordinates @xmath41 and @xmath42 parameters . for notational convenience ,
let the vector @xmath43 represent the coordinates @xmath38 of the @xmath44 pixel and the vector @xmath45 represent all the model parameters [ @xmath46 .
the observational model of the @xmath10 pixel can thus be compactly written as follows : @xmath47 .
the measure of the goodness of fit between the data and the model , called chi square , is defined as @xmath48 the theory of least - squares minimization states that the optimum value of the parameter vector @xmath45 is obtained when @xmath49 is minimized with respect to each parameter simultaneously . if @xmath50 is the optimal parameter vector , then @xmath51 is the absolute minimum of the @xmath42-dimensional manifold @xmath49 . for some small correction parameter vector @xmath52 one can approximate @xmath53 by its taylor series expansion : @xmath54 where @xmath55_{jk } & \equiv & \frac{\partial^2\,{\mbox{$\chi^2({\mbox{$\mathbf{p}$}})$}}}{\partial a_j \partial a_k } \nonumber \\ & \approx & \left[\frac{\partial\,{\mbox{$\chi^2({\mbox{$\mathbf{p}$}})$}}}{\partial a_j}\right ] \left[\frac{\partial\,{\mbox{$\chi^2({\mbox{$\mathbf{p}$}})$}}}{\partial a_k}\right]\end{aligned}\ ] ] is the @xmath56 element of the @xmath42@xmath57@xmath42 hessian matrix @xmath58 of @xmath49 [ e.g. , @xcite ; @xcite ]
the approximation for the calculation of the hessian matrix elements is frequently used whenever the computation of the second partial derivative is numerically unstable . if @xmath53 is a _ local _ minimum of @xmath59 manifold , then it can be shown that @xmath60 by solving this equation for the correction vector @xmath52 , one can determine a better parameter vector as follows : @xmath61 when the parameter vector ( @xmath45 ) is redefined to be the better parameter ( @xmath62 ) , the hessian matrix and the gradient of @xmath49 can then be recalculated to determine a new correction vector ( @xmath52 ) .
this process repeats until the correction vector is sufficiently small generally when the difference between solutions is no longer statistically significant .
if the fitting process has not failed , then the optimal parameter vector ( @xmath50 ) should be very close to the true parameter vector .
once the optimal parameter vector has been determined , the covariance matrix @xmath63 may then be calculated by inverting the hessian matrix @xmath58 computed with the optimal parameter vector .
the standard errors ( one standard deviation ) of the fitted parameters can be estimated as follows : @xmath64_{jj } } = \left [ \sum_{i=1}^{n } \frac{1}{\sigma_i^2 } \left ( \frac{\partial { \mbox{$m_i$}}}{\partial p_j } \right)^2\ , \right]^{-1/2 } \label{eq : errorest}\ ] ] where @xmath65 is the standard error associated with the @xmath66 parameter ( @xmath67 ) .
usage of equation ( [ eq : errorest ] ) for error estimates is based on the critical assumption that fitted model parameters are independent ( indicated by negligibly small off - diagonal elements of the covariance matrix ) .
it is important to note that whenever this critical assumption is violated , the results produced by least - squares fitting may not be statistically reliable , which is to say , they may no longer be physically meaningful . the theoretical photometric performance limits for psf - fitting ccd stellar photometry can be derived using a simple observational model consisting of a prf and a constant sky level .
consider a ccd observation of single isolated star on a flat sky background .
assuming one already knows the prf of the observation at the location of the star , a simple model of the observation would have just two parameters : the stellar intensity ( @xmath68 ) in electrons , and the observed background sky level ( @xmath33 ) in electrons .
the observational model for the @xmath10 pixel would be @xmath69 where @xmath70 is the volume integral of the prf and @xmath71 is the value of the @xmath44 pixel of the _ normalized _ prf @xmath72 . in the case of bright stars ,
most of the electrons found in the @xmath44 pixel of the observation will come from the star and not the sky : @xmath73 the actual number of electrons found in the @xmath44 pixel will be described by a poisson distribution with a mean and variance of @xmath74 .
the measurement error ( one standard deviation ) for the @xmath44 pixel would thus be @xmath75 all other noise sources ( due to , for example , the observed background sky , instrumental readout noise , flat - field calibrations errors , etc . )
are assumed , in this special case , to be negligibly small .
the variance of the stellar intensity measurement error of bright stars can be estimated using equations ( [ eq : errorest ] ) , ( [ eq : photmodel ] ) , and ( [ eq : bphoterr ] ) : @xmath76^{-1 } \nonumber \\ & \approx & \frac{{\mbox{$\mathcal{e}$}}}{{\mbox{v } } } \left [ { \mbox{$\displaystyle \mathop{\int\!\!\!\int}_{\!-\infty}^{\ + \infty } $ } } \tilde{{\mbox{$\psi$ } } } \,dx\,dy \right]^{-1 } \nonumber \\ & = & \frac{{\mbox{$\mathcal{e}$}}}{{\mbox{v } } } \ , \label{eq : myvarebright}\end{aligned}\ ] ] as expected from photon statistics .
a bright isolated star with an intensity of @xmath77 photons imaged with a perfect ccd detector would have a stellar image with @xmath77 @xmath6 @xmath78 and a stellar intensity measurement error of @xmath79 @xmath6 .
the same star imaged with an inefficient ccd detector with a quantum efficiency of 25% ( @xmath80 ) would have a stellar image with @xmath20@xmath81 @xmath6 which would have a poisson noise error of @xmath20@xmath82 @xmath6 .
the _ measured _ stellar intensity is @xmath83 @xmath6 with an rms measurement error of @xmath84 @xmath6 which is two times larger than it would be with a perfect detector and four times larger than the poisson noise error of the _ observed _ number of electrons . solving for _ measured
_ stellar intensity @xmath85 instead of the _ observed _ stellar intensity @xmath86 enables the creation of stellar photometric reduction programs capable of dealing with intrapixel qe variations through the accurate modeling of the image formation process within the detector .
while it is certainly convenient to assume that one s detector has negligible intrapixel qe variation , in the real world even nasa - grade ccd detectors , like those found in the _ hst _ wfpc2 instrument , can have peak - to - peak intrapixel sensitivity variations that are greater than 0.02 mag ( @xmath872% ) ( see figs . 5 and 6 of * ? ? ?
* ) . most of the electrons found in the @xmath44 pixel of an observation of a _ faint _ isolated star on a flat sky background will come from the sky and not from the star . in that case , the measurement error associated with @xmath44 pixel is approximately the effective - background noise level : @xmath88 where @xmath89 @xmath33 is the constant observed background sky level which is assumed to be a poisson distribution with a mean of @xmath33 electrons , and @xmath90 is the rms readout noise . the variance of the stellar intensity measurement error of faint stars can be estimated using equations ( [ eq : errorest ] ) , ( [ eq : photmodel ] ) , ( [ eq : fphoterrirms ] ) , ( [ eq : fphoterr ] ) , ( [ eq : fphoterr_approx ] ) , and ( [ eq : beta2 ] ) : @xmath91^{-1 } \nonumber \\ & \approx & \frac{{\mbox{$\sigma^2_{\rm{rms}}$}}}{{\mbox{v}}^{\,2 } } \left [ { \mbox{$\displaystyle \mathop{\int\!\!\!\int}_{\!-\infty}^{\ + \infty } $ } } \tilde{{\mbox{$\psi$}}}^2 \,dx\,dy \right]^{-1 } \nonumber \\ & = & { \mbox{\boldmath$\beta$}}\,{\mbox{$\sigma^2_{\rm{rms}}$}}\\ & \approx & { \mbox{\boldmath$\beta$}}\ , \left [ { \mbox{$\mathcal{b}$}}+ { \mbox{$\sigma^2_{\rm{ron}}$}}\right ] \ , \label{eq : myvarefaint}\end{aligned}\ ] ] where @xmath22 is the effective - background area of the prf . equation ( [ eq : myvarefaint ] ) agrees with equation ( 9 ) of @xcite for a perfect @xmath92 noiseless @xmath93 detector .
an important additional noise source for the photometry of faint stars is the systematic error due to the uncertainty of the measurement of the background .
if the sky background is assumed to be flat , then the rms measurement error of the constant sky background can be estimated using equations ( [ eq : errorest ] ) , ( [ eq : photmodel ] ) , ( [ eq : fphoterrirms ] ) , ( [ eq : fphoterr ] ) , and ( [ eq : fphoterr_approx ] ) : @xmath94^{-1/2 } \nonumber \\ & = & \frac{{\mbox{$\sigma_{\rm{rms}}$}}}{\sqrt{n } } \\ & \approx & \sqrt { \frac{{\mbox{$\mathcal{b}$}}+ { \mbox{$\sigma^2_{\rm{ron}}$}}}{n } } \ .
\label{eq : sigmab}\end{aligned}\ ] ] given a ccd observation with no readout noise , equation ( [ eq : sigmab ] ) reduces to the value of @xmath95 expected from simple sampling statistics .
the portion of the rms stellar intensity measurement error that is caused by the error in the determination of the local sky level is @xmath96 @xcite . while this error is frequently negligible for bright stars , it is generally significant for faint stars .
including the uncertainty in the determination of the constant observed background sky level thus gives a more realistic estimate for the rms stellar intensity measurement error for _ faint _ stars : @xmath97 precise and accurate stellar photometry of faint stars requires an excellent determination of the observed background sky which in turn requires accurate background sky models .
given a valid background sky model , small apertures will be more sensitive to background sky measurement errors than large apertures . a realistic photometric performance model for ccd psf - fitting photometry can be created by combining the bright and faint star limits developed above .
the theoretical _ upper limit _ for the photometric signal - to - noise ratio ( @xmath98 ) of ccd psf - fitting photometric algorithms is as follows : @xmath99 } } \ . \label{eq : practical - snr}\end{aligned}\ ] ] these approximations assume , for the sake of simplicity , that any noise contribution due to dark current and quantization noise is negligible .
while these additional noise sources can be added to create an even more realistic performance model for stellar photometry , the assumption of low dark current and minimal quantization noise is realistic for state - of - the - art astronomical - grade ccd imagers .
the resulting photometric error is approximately @xmath100 where the constant @xmath101 is an approximation for pogson s ratio @xmath102 @xcite .
the cramr - rao lower bound ( crlb ) is the lower bound on the variance of _ any _ unbiased estimator .
since it is physically impossible to find an unbiased estimator that beats the crlb , the crlb provides a performance benchmark against which any unbiased estimator can be compared .
the cramr - rao lower bound for stellar photometry of a single isolated star imaged by a two - dimensional photon - counting detector has been derived several times in the astrophysical literature ( see , e.g. , appendix a of @xcite , @xcite , and @xcite ) .
the generalization for a crowded field with overlapping stellar images is given in @xcite .
the cramr - rao lower bound for the bright star limit of stellar photometry of a single isolated star is @xmath103 which is equation ( [ eq : myvarebright ] ) with a perfect detector .
the crlb for the faint star limit of stellar photometry of a single isolated star is @xmath104 which is equation ( [ eq : myrmsefaint ] ) with a noiseless detector and a negligible background measurement error ( @xmath105 ) .
the photometric performance model has bright and faint star limits which are the same , respectively , as the bright and faint star cramr - rao lower bounds for stellar photometry of a single isolated star on a flat sky background imaged with a perfect noiseless detector .
the theoretical astrometric limits for psf - fitting ccd stellar photometry can be derived using a simple observational model consisting of a gaussian prf and a constant sky level .
consider a ccd observation of single isolated star on a flat sky background .
a gaussian is a good model for the psf of a ground - based ccd observation since the central core of a ground - based stellar profile is approximately gaussian @xcite . in this case
the psf would have three parameters : two coordinate values giving the location ( @xmath106 ) of the star on the ccd and the standard deviation of the gaussian ( @xmath16 ) in pixels [ see equation ( [ eq : gaussian ] ) ] .
an _ imperfect but uniformly flat _
drf ( @xmath107 ) gives a value for the @xmath44 pixel of the prf located at @xmath108 which is equal to the product of the volume of the prf and the value of the volume integral of the psf over the area of the @xmath44 pixel : @xmath109 the actual limits of the above volume integral reflect the appropriate mapping transformation of the @xmath12 and @xmath13 coordinates onto the ccd pixel coordinate system .
if the prf has been _ oversampled _ , the value of the @xmath44 pixel of the prf is approximately equal to the product of the volume of the prf and the value of the psf at the _ center of the @xmath44 pixel _ : @xmath110 where @xmath111 a simple model of the observation will require two additional parameters : the stellar intensity ( @xmath68 ) in electrons and the observed background sky level ( @xmath33 ) in electrons .
the @xmath44 pixel of the observational model would be @xmath112 where @xmath70 is the volume integral of the prf and @xmath113 is the value of the @xmath44 pixel of the _ normalized _ prf @xmath114 . in the case of bright stars ,
most of the electrons found in the @xmath44 pixel of the observation will come from the star and not the sky : @xmath115 the actual number of electrons found in the @xmath44 pixel will be described by a poisson distribution with a mean and variance of @xmath74 .
the measurement error ( one standard deviation ) for the @xmath44 pixel would thus be @xmath116 all other noise sources ( e.g. , due to the observed background sky , instrumental readout noise , flat - field calibrations errors , etc . )
are assumed to be negligibly small .
the variance of the stellar @xmath117 position measurement error of a bright isolated _ oversampled _ star can be estimated using equations ( [ eq : errorest ] ) , ( [ eq : astmodel ] ) , ( [ eq : abphoterr ] ) , and ( [ eq : gaussian ] ) : @xmath118^{-1 } \nonumber \\ & \approx & \frac{1}{{\mbox{$\mathcal{e}$}}{\mbox{v } } } \left [ \sum_{i=1}^{n } \frac{1}{{\mbox{$\mathsf{g}$}}_i } \left ( \frac{\partial}{\partial { \mbox{$\mathcal{x}$ } } } { \mbox{$\mathsf{g}$}}_i \right)^2\ , \right]^{-1 } \nonumber \\ & \approx & \frac{{\mbox{$\mathcal{s}$}}^4}{{\mbox{$\mathcal{e}$}}{\mbox{v } } } \left [ { \mbox{$\displaystyle \mathop{\int\!\!\!\int}_{\!-\infty}^{\ + \infty } $ } } { \mbox{$\mathsf{g}$ } } ( x , y ; { \mbox{$\mathcal{x}$ } } , { \mbox{$\mathcal{y}$ } } , { \mbox{$\mathcal{s}$ } } ) \,(x-{\mbox{$\mathcal{x}$}})^2 \,dx\,dy \right]^{-1 } \label{eq : sxbt1 } \nonumber \\ & = & \frac{{\mbox{$\mathcal{s}$}}^2}{{\mbox{$\mathcal{e}$}}{\mbox{v } } } \nonumber \\ & \approx & \frac{{\mbox{$\mathcal{l}$}}^2}{{\mbox{$\mathcal{e}$}}{\mbox{v } } } \ , , \label{eq : sxb_cssl}\end{aligned}\ ] ] where @xmath119 is the _ critical - sampling scale length _ of the prf , @xmath120 , one sees that critical - sampling scale length has been designed to be a proxy for @xmath16 for _ any _ prf . ] in pixel units ( @xmath121 ) , which , unlike @xmath16 , is defined for all prfs . by definition , the critical - sampling scale length of a critically - sampled prf imaged with a perfect detector is one pixel ; @xmath122 indicates that the prf is _ oversampled _ , while @xmath123 indicates that the prf is _
undersampled_. in the special case of a critically - sampled bright star imaged with a perfect detector , one finds that the astrometric performance limit ( in pixel units ) is equal to the reciprocal of photometric error performance limit : @xmath124 let us again assume that the noise contribution from the star is negligibly small and that the variance of the measurement error of the @xmath44 pixel can be replaced with an average constant rms value .
the variance of the stellar @xmath117 position measurement error of a faint isolated _ oversampled _ star can be estimated using equations ( [ eq : errorest ] ) , ( [ eq : astmodel ] ) , ( [ eq : fphoterrirms ] ) , ( [ eq : fphoterr ] ) , ( [ eq : fphoterr_approx ] ) , and ( [ eq : gaussian ] ) : @xmath125^{-1 } \nonumber \\ & \approx & \frac{{\mbox{$\sigma^2_{\rm{rms}}$}}}{{\mbox{$\mathcal{e}$}}^2{\mbox{v}}^{\,2 } } \left [ \sum_{i=1}^{n } \left ( \frac{\partial}{\partial { \mbox{$\mathcal{x}$ } } } { \mbox{$\mathsf{g}$}}_i \right)^2\ , \right]^{-1 } \nonumber \\ & \approx & \frac{{\mbox{$\sigma^2_{\rm{rms}}$}}\,{\mbox{$\mathcal{s}$}}^4}{{\mbox{$\mathcal{e}$}}^2{\mbox{v}}^{\,2 } } \left [ { \mbox{$\displaystyle \mathop{\int\!\!\!\int}_{\!-\infty}^{\ + \infty } $ } } { \mbox{$\mathsf{g}$}}^2 ( x , y ; { \mbox{$\mathcal{x}$}},{\mbox{$\mathcal{y}$ } } , { \mbox{$\mathcal{s}$ } } ) \,\,(x-{\mbox{$\mathcal{x}$}})^2 \,dx\,dy \right]^{-1 } \nonumber \\ & = & 8\pi \,{\mbox{$\sigma^2_{\rm{rms}}$}}\frac{{\mbox{$\mathcal{s}$}}^4}{{\mbox{$\mathcal{e}$}}^2{\mbox{v}}^{\,2 } } \nonumber \\ & \approx & 8\pi \,{\mbox{$\sigma^2_{\rm{rms}}$}}\left ( \frac{\mathcal{l}^2}{{\mbox{$\mathcal{e}$}}{\mbox{v } } } \right)^{\!2 } \nonumber \\ & \approx & 8\pi \,{\mbox{$\sigma^2_{\rm{rms}}$}}\left ( { \mbox{$\sigma_{\mbox{{\mbox{$\mathcal{x}$}}:\,bright}}^2$}}\right)^2 \\ & \approx & 8\pi \left ( { \mbox{$\mathcal{b}$}}+ { \mbox{$\sigma^2_{\rm{ron}}$}}\right ) \left ( { \mbox{$\sigma_{\mbox{{\mbox{$\mathcal{x}$}}:\,bright}}^2$}}\right)^2 \label{eq : sxf_cssl_approx } \,.\end{aligned}\ ] ] a realistic performance model for ccd psf - fitting astrometry can be created by combining the bright and faint star limits developed above .
the expected _ lower limit _ of the rms measurement error for the stellar @xmath117 position for a single isolated star on a flat sky can be estimated as follows : @xmath126 } \\ & \approx & \sqrt { \frac{\mathcal{l}^2}{{\mbox{$\mathcal{e}$}}{\mbox{v } } } \ , \left [ 1 + 8\pi \left ( { \mbox{$\mathcal{b}$}}+ { \mbox{$\sigma^2_{\rm{ron}}$}}\right ) \frac{\mathcal{l}^2}{{\mbox{$\mathcal{e}$}}{\mbox{v } } } \ , \right ] } \ , . \label{eq : practical - astrometric - model - x}\end{aligned}\ ] ] the rms stellar @xmath127 position measurement error is , by symmetry , the same as for @xmath128 : @xmath129 the cramr - rao lower bound for stellar astrometry depends not only on the signal - to - noise ratio but also on the _ size _ and _ shape _ of the detector . for well - sampled data , the size and shape of the detector can be ignored and a crlb can be found for a perfect noiseless detector with infinitely small pixels .
this is called the _
photonic limit_. the determination of the crlb for astrometry becomes much more complicated with _ undersampled _ observations .
astrometric precision degrades when the size of the detector is comparable to the size of the stellar image the quality of the position estimation is then dependent on the fraction of photons falling _ outside _ of the central pixel .
the worst - case scenario for stellar astrometry occurs when _ all _ of the light from a star falls within a single pixel : all one knows for sure , in that unfortunate case , is that the star is located _ somewhere _ within the central ( and only ) pixel . the photonic limit ( pl ) for stellar astrometry of a bright well - sampled single isolated normalized gaussian star is @xmath130 @xcite .
using @xmath131 as a proxy for @xmath16 , one has the generalized form for any psf : @xmath132 which is equation ( [ eq : sxb_cssl ] ) with a perfect detector .
the photonic limit for stellar astrometry of a faint well - sampled single isolated normalized gaussian star is @xmath133 @xcite . using @xmath131 as a proxy for @xmath16
, one has the generalized form for any psf : @xmath134 which is equation ( [ eq : sxf_cssl_approx ] ) with a perfect noiseless detector .
the astrometric performance model has bright and faint star limits which are the same , respectively , as the bright and faint star photonic astrometric limits , which are the cramr - rao lower bounds for stellar astrometry of a single isolated gaussian star on a flat sky background imaged with a perfect noiseless detector with infinitely small pixels .
the cramr - rao lower bound for stellar astrometry of a single isolated gaussian star on a flat sky background imaged with a perfect noiseless ccd _ with square pixels _
@xcite quickly approaches the photonic limits with _
well - sampled _ observations ; undersampled observations will have larger astrometric errors than predicted by the photonic limits .
following @xcite and @xcite , i now compare the astrometric error of bright isolated stars with their photometric error .
the ratio of the astrometric error of a bright isolated star and the critical - sampling scale length of the prf is equal to the ratio of the stellar intensity measurement error and the stellar intensity : @xmath135 for example , a bright isolated critically - sampled star with one million electrons imaged on a perfect detector ( @xmath136 , @xmath137 , and @xmath138 ) would , theoretically , have a signal - to - noise ratio of a thousand ( @xmath139 ) , a stellar intensity measurement error of @xmath140 , and an rms position error in @xmath12 of one - thousandth of a pixel @xmath141 .
such astrometric accuracy may be difficult to achieve in practice under normal ground - based observing conditions even with state - of - the - art astronomical - grade ccd cameras .
the astrometric error of faint isolated stars is related to their photometric error as follows : @xmath142 for example , a faint isolated critically - sampled star imaged with a perfect detector with a 20.0% intensity measurement error and a negligible background measurement error ( @xmath105 ) would , theoretically , have an astrometric error of @xmath200.283 @xmath143 $ ] px .
these results suggest the following practical lower bound for astrometric errors with respect to photometric errors : for example , a star with one - percent stellar photometry will have no better than one - percent astrometry with respect to the critical - sampling scale length .
if the star is critically sampled , then the astrometric precision will be no better than 0.01 px .
all of the above derivations are based on the assumption that that flat - field calibration errors are negligible .
the relation between photometry and astrometry for bright isolated stars can fail with large flat - field calibration errors .
a _ discrete _ point spread function is a sampled version of a continuous two - dimensional point spread function .
the shape information about the photon scattering pattern of a discrete psf is typically encoded using a numerical table ( matrix ) .
analytical _ psf has the shape information encoded with continuous two - dimensional mathematical functions . in order to do accurate stellar photometry and astrometry with discrete
psfs one needs to able to ( 1 ) accurately shift discrete psfs to new positions within the observational model , and ( 2 ) compute the position partial derivatives of discrete psfs .
the next two subsections describe how these tasks may be accomplished using numerical analysis techniques . building a realistic observation model
requires the placement of a star at the desired location within the model ; this is done by determining the prf at required location and then multiplying it by the stellar intensity .
with psfs encoded by mathematical functions , one just computes the psf at the desired location in the observational model . with discrete psfs ,
one ideally takes a reference psf ( typically derived / computed for the center of a pixel ) and shifts it to the desired location using a perfect two - dimensional interpolation function . but how is this done in practice ?
the sinc function , @xmath144 , is , theoretically , a perfect two - dimensional interpolation function .
unfortunately , the sinc function decays with @xmath145 and never actually reaches zero .
one can use a windowed interpolant in order to improve computational speed but one must be cautious about aliasing effects caused by using a windowed function . in the case of stellar photometry and astrometry , _ aliasing effects will generally only be seen with bright stars _ since a large number of photons are required in order to adequately sample the higher spatial frequencies of the psf .
the following 21-pixel - wide _ damped sinc function _ interpolant does an excellent job interpolating discrete psfs : @xmath146 ^ 2\right ) } \label{eq : dampedsinc}\end{aligned}\ ] ] @xcite .
note that since the two - dimensional sinc function is separable in @xmath12 and @xmath13 , this interpolant can be coded to be computationally fast and efficient .
this interpolant , from the zodiac c library written by marc buie of lowell observatory , was specifically designed for use with 32-bit floating numbers .
aliasing problems due to critically - sampled or undersampled data may be overcome by using discrete psfs that are _ supersampled _ at 2 , 3 , or more times more finely than the observational data . in order to have a realistic observational model , once the supersampled discrete psf has been interpolated to the correct position , a new _ degraded _
( rebinned ) version of the discrete psf must be created which has the same spatial resolution as the observational data .
while the mathematics of determining the position partial derivatives of individual stars within the observational model with respect to the @xmath12 and @xmath13 direction vectors is the same regardless of how the shape information in a psf is encoded , the implementation methodology for the computation of position partial derivatives of discrete psfs is very different than the one used for analytical psfs .
the position partial derivatives of discrete psfs can be determined using _ numerical differentiation techniques _ on the discrete psf .
it is a standard practice in numerical analysis to approximate the first , second , or higher , derivatives of a tabulated function @xmath147 with multi - point formulae .
@xcite give 18 different multi - point formulae which can be used ( with varying degrees of accuracy ) to approximate the first derivative of the tabulated function @xmath147 .
the following five - point differentiation formula , @xmath148\ , , \label{eq : five - point}\end{aligned}\ ] ] ( p. 914 of * ? ? ? * ) works well with discrete psfs @xcite .
this approximation takes just 4 additions and 3 multiplications which generally makes it considerably faster to compute than the traditional determination of the partial derivative of the volume integral of the psf above a ccd pixel .
the concepts presented above outline the unique and fundamental features of the matphot algorithm for accurate and precise stellar photometry using discrete point spread functions .
while the key features of a ccd stellar photometric reduction algorithm can be described in an article , the full implementation of such an algorithm generally exists as a complex computer program consisting of many thousands of lines of computer code .
since good algorithms can be poorly implemented , it can be difficult to differentiate between a poor algorithm and a poorly - coded implementation of a good algorithm .
confidence in a complex algorithm can be established by developing an implementation of the algorithm which meets theoretical performance expectations .
the following subsection describes a real - world implementation of the matphot algorithm that meets the theoretical performance expectations for accurate and precise stellar photometry and astrometry which were derived in 2 .
i have written a c - language computer program , called @xmath0 and support software is freely available at the official matphot website at noao : http://www.noao.edu/staff/mighell/matphot ] , which is based on the current implementation of the matphot algorithm for precise and accurate stellar photometry using discrete point spread functions .
the @xmath0 code demonstrates the precision and accuracy of the matphot algorithm by analyzing simulated ccd observations based on user - provided discrete psfs encoded as fits images @xcite .
discrete psfs are shifted within the observational model using the 21-pixel - wide damped sinc interpolation function given in equation ( [ eq : dampedsinc ] ) .
position partial derivatives of discrete psfs are computed using the five - point differentiation formula given in equation ( [ eq : five - point ] ) .
accurate and precise stellar photometry and astrometry of _ undersampled _ ccd observations can be obtained with the @xmath0 code when it is presented with
_ supersampled _
discrete psfs that are sampled 2 , 3 , or more times more finely than the observational data .
the @xmath0 code is based on a robust implementation of the levenberg - marquardt method of nonlinear least - squares minimization ( @xcite , @xcite , also @xcite ) . when presented with simulated observations based on a gaussian psf with a known @xmath149 value ( full - width - at - half - maximum ) value of a gaussian is equal to @xmath150 times the standard deviation , @xmath16 , of the gaussian : @xmath151 [ see equation ( [ eq : gaussian ] ) ] . ]
, the @xmath0 code can analyze the observation two different ways : ( 1 ) the matphot algorithm can be used with a discrete gaussian psf , or ( 2 ) analytical techniques @xcite can be used with an analytical gaussian psf .
i now demonstrate that the theoretical performance limits of 2 provide practical performance metrics for photometry and astrometry of ccd stellar observations that are analyzed with oversampled gaussian point spread functions .
twenty thousand _ oversampled _ ccd stellar observations were simulated and analyzed using the @xmath0 code .
the ccd detector was assumed to be perfect ( @xmath152 ) with a ccd readout noise value of @xmath153 @xmath154 .
stars were simulated using an _ analytical _ gaussian psf with a @xmath155 @xmath121 located near the center of 60@xmath5760 pixels , the input stellar intensities ranged from @xmath156 to @xmath157 mag mag @xmath158 @xmath159 ( electron ) @xmath158 @xmath160 ( photon ) for a point response function volume of one @xmath161 . ] ( @xmath162 ) , and a flat background was assumed with a value of @xmath163 @xmath6 .
photon and readout noise was simulated , respectively , using poisson and gaussian random noise generators and the resulting observed background sky measurement error was @xmath164 @xmath6 .
the median effective - background area of the prf of these observations was @xmath165 @xmath166 .
all the simulated observations were analyzed with @xmath0 using an _ analytical _ gaussian psf with @xmath167 @xmath121 .
the binned absolute photometric errors are shown as black box - and - whiskers plots ( see appendix [ appendix : a ] ) on the _ top _ panel of fig .
the absolute photometric error of an observation is the absolute value of the difference between the measured ( estimated ) and true ( actual ) stellar magnitude : @xmath168 .
the four _ grey _ limits seen on the _ top _ panel of fig .
[ fig1 ] are theoretical predictions ( derived from 2.3.4 ) for the median ( 50% cumulative fraction : grey solid curve ) , top hinge ( 75% : bottom of the grey band ) , top fence ( @xmath2098.35% : top of band ) , and 5-@xmath169 outlier ( @xmath2099.99997% : grey dashed curve ) values .
if the rms photometric error is called @xmath170 , then the values of these theoretical limits are approximately equal to @xmath171 , @xmath172 , @xmath173 , and @xmath174 , respectively .
if the photometric performance model is correct and @xmath0 has been coded correctly , then ( 1 ) the observed median values ( central bar in each box ) should intersect the theoretical median value , ( 2 ) most of the top whiskers should be found inside the band , and ( 3 ) most of the outliers should be found above the top of the band and all of the outliers should found below the 5-@xmath169 outlier limit .
comparing the absolute photometric errors of the 20,000 simulated ccd observations with the grey theoretical limits , one sees that the photometric performance of the @xmath0 code is very well predicted by the model given in 2.3.4 .
the binned total astrometric errors are shown as black box - and - whiskers plots on the _ bottom _ panel of fig .
[ fig1 ] . the total astrometric error of an observation is the distance between the measured ( estimated ) and true ( actual ) position of a star : @xmath175 .
the four _ grey _ limits seen on the _ bottom _ panel of fig .
[ fig1 ] are theoretical predictions ( derived from 2.4.4 ) for the median ( 50% cumulative fraction : grey solid curve ) , top hinge ( 75% : bottom of the grey band ) , top fence ( @xmath2098.97% : top of band ) , and 5-@xmath169 outlier ( 99.99997% : grey dashed curve ) values .
the values of these theoretical limits are approximately equal to @xmath176 , @xmath177 , @xmath178 , and @xmath179 , where @xmath180 is the rms measurement error for the stellar @xmath117 position . if the astrometric performance model is correct and @xmath0 has been coded correctly , then ( 1 ) the observed median values should intersect the theoretical median value , ( 2 ) most of the top whiskers should be found inside the band , and ( 3 ) most of the outliers should be found above the top of the band and all of the outliers should found below the 5-@xmath169 outlier limit . comparing the total astrometric errors of the 20,000 simulated ccd observations with the grey theoretical limits , one sees that the astrometric performance of the @xmath0 code is very well predicted by the model given in 2.4.4 .
figure [ fig2 ] shows the _ relative _ stellar intensity errors and the _ relative _ @xmath117 position errors of the 20,000 stars analyzed in fig .
the relative stellar intensity error is the difference between the measured ( estimated ) and true ( actual ) stellar intensity values divided by the estimated stellar intensity error : @xmath181 .
the relative @xmath117 position error is the difference between the measured ( estimated ) and true ( actual ) stellar @xmath117 position values divided by the estimated @xmath117 error : @xmath182 . if @xmath0 has been coded correctly , the relative error distributions for the stellar parameters @xmath68 , @xmath117 , and @xmath127 should be _ normally _ distributed .
the five _ grey _ limits seen on each panel are theoretical predictions ( based on the normal distribution ) for , from bottom to top , the bottom fence ( @xmath200.35% cumulative fraction : bottom of the bottom grey band ) , bottom hinge ( 25% : top of bottom band ) , median ( 50% : grey solid line at zero ) , top hinge ( 75% : bottom of top band ) , top fence ( @xmath2099.65% : top of top band ) values .
if the relative errors for @xmath68 and @xmath117 are indeed normally distributed , then ( 1 ) the observed median values should be near zero , ( 2 ) most of the whiskers should be found inside the bands , and ( 3 ) most of the outliers should be beyond the fence values . comparing the relative errors for @xmath68 and @xmath117 of the 20,000 simulated ccd observations with the grey theoretical limits
, one sees that these errors are , as expected , normally distributed . *
the @xmath0 code works well with oversampled analytical gaussian psfs and its performance can be very well predicted with the photometric and astrometric models derived in 2 .
the 20,000 simulated ccd observations analyzed in figs .
[ fig1 ] and [ fig2 ] were _ reanalyzed _ with @xmath0 using an oversampled _ discrete _ gaussian psf with a @xmath149 of 3.0 @xmath121 .
figure [ fig3 ] shows the resultant absolute photometric errors and total astrometric errors .
figure [ fig4 ] shows the resultant relative errors for @xmath68 and @xmath117 .
notice how similar figs .
[ fig1 ] and [ fig3 ] and figs .
[ fig2 ] and [ fig4 ] are to each other . despite the very different way the shape information of the psf was encoded ( i.e. , _ discrete _ versus _ analytical _ representations )
, @xmath0 produced nearly identical photometric and astrometric results .
how similar are the measured stellar positions ?
figure [ fig5 ] shows the _ relative _ @xmath117 and @xmath127 position
_ differences _ between the previous analytical and numerical analyses with the @xmath0 code .
the _ top _
panel shows the difference between the numerical @xmath117 result and the analytical @xmath117 result divided by the estimated error of the analytical result .
similarly , the _ bottom _ panel shows the difference between the numerical @xmath127 result and the analytical @xmath127 result divided by the estimated error of the analytical result .
the relative differences between the numerical and analytical methods are _ not _ normally distributed observe how much smaller the values on the ordinate of fig .
[ fig5 ] are compared to those of figs .
[ fig2 ] and [ fig4 ] .
[ fig2 ] and [ fig4 ] _ are _ normally distributed and the source of the scatter is photon noise .
figure [ fig5 ] indicates that the relative differences between the numerical and analytical methods for astrometry are less than one - fifteenth the difference due to photon noise . in other words , _ the computational noise due to the chosen analysis method ( numerical versus analytical ) is insignificant when compared to the unavoidable photon noise due to the random arrival of photons in any astronomical ccd observation . _ * the @xmath0 code works as well with oversampled discrete gaussian psfs as it does with oversampled analytical gaussian psfs . while the volume , @xmath70 , of the prf was carefully tracked throughout the derivation of the photometric and astrometric performance models in 2 , all previous simulations have assumed a perfect detector @xmath183 .
let us now check to see if the effects of a prf volume integral that is less than one has been correctly accounted for in the performance models of 2 by analyzing simulated observations imaged on a very inefficient detector @xmath184 .
twenty thousand oversampled ccd stellar observations were simulated assuming a very inefficient detector with @xmath185 .
stars were simulated using a discrete gaussian psf with a @xmath155 @xmath121 located near the center of 60@xmath5760 pixels , the input stellar intensities ranged from @xmath186 to @xmath157 mag ( @xmath187 to @xmath77 @xmath5 ) ; and the _ observed _ background sky level was assumed to be a constant value of @xmath188 @xmath6 ( @xmath189 @xmath5 , @xmath190 ) all other simulation parameters were the same as before . all the simulated observations were analyzed with @xmath0 in the same way as described for the numerical experiment shown in fig .
[ fig3 ] except that the volume of the prf was set to @xmath191 in order to simulate the use of an inefficient detector which converts only @xmath2011.1% of photons to electrons .
figure [ fig6 ] shows the absolute photometric errors and total astrometric errors of this numerical experiment .
the median effective - background area of prf of these observations was @xmath192 @xmath166 which is , as expected , 81 @xmath193 times larger than the median value reported in fig .
[ fig3 ] . comparing the simulation results with the grey theoretical limits
, one sees that the photometric and astrometric performance of the @xmath0 code is very well predicted by the theoretical performance models given in 2 .
the black dash - dot curves in each panel of fig .
[ fig6 ] shows the expected median response with a perfect detector ; these curves are the same as the solid grey median curves found in fig .
the _ observed _ stellar intensities and observed background sky level are nine times fainter than was seen in the numerical experiment shown in fig .
[ fig3 ] and the median photometric and astrometric errors in fig .
[ fig6 ] are , as expected , @xmath203 @xmath194 times larger when the inefficient detector is used . * the @xmath0 code and the theoretical performance models work well with prfs that have volumes of less than one .
twenty thousand _ undersampled _ ccd stellar observations were simulated using an analytical gaussian with a @xmath195 @xmath121 , the other simulation parameters were the same as given in 5.1.1 .
the median effective - background area of prf of these observations was @xmath196 @xmath166 @xmath183 .
all the simulated observations were analyzed with @xmath0 using a _
discrete _ gaussian psf with @xmath195 @xmath121 .
figure [ fig7 ] shows the absolute photometric errors and total astrometric errors of this numerical experiment . while the photometric and astrometric results for stars with @xmath197 @xmath6 are fine , the results for stars brighter than that limit are seen to quickly degrade in accuracy with the brightest stars having median errors that are @xmath2040 times worse than expected . what starts going wrong at @xmath198 @xmath6 ?
figure [ fig8 ] shows a one pixel wide slice through a pixel - centered discrete gaussian psf with @xmath199 @xmath121 that was shifted half of a pixel in @xmath200 to the right using damped sinc function given in equation ( [ eq : dampedsinc ] ) .
the dashed black curve looks fine , but when expanded by a factor of 100 , one sees that _ negative side lobes have been created _ due to the fact that the nyquist - shannon sampling theorem has been violated .
doing a sinc interpolation ( damped or otherwise ) on undersampled data is never a good idea the `` ringing '' seen in fig .
[ fig8 ] is a classic signature of an edge that is too sharp to be adequately expressed with the limited spatial information contained in an undersampled observation .
the biggest negative side lobe of the shifted psf has a value of about @xmath2010.0006 .
although that may seem to be a small value compared to the total volume integral of one , it is actually quite disastrous because _ negative psf values have no physical meaning_. it is now clear what has gone wrong for stars with @xmath202 @xmath6 . at stellar intensity values
greater than 17,000 electrons , the intensity - scaled undersampled psf models can have negative side lobes that are larger than the rms observed background sky noise level @[email protected]@xmath204 . at stellar intensity values
greater than 167,000 electrons , the observational models have physically nonsensical _ negative sky values_. * aliasing ( ringing ) effects will generally only be seen with bright stars since a large number of photons are required in order to adequately sample the higher spatial frequencies of a psf .
* fitting undersampled observations of bright stars with undersampled psfs results in poor photometry and astrometry .
a supersampled psf is a psf with pixels that have greater spatial resolution ( higher spatial frequencies ) than the actual pixels in the observational data .
for example , a 2@xmath572 supersampled psf uses 4 pixels to describe every physical pixel of the ccd observation ; each supersampled pixel has twice the spatial resolution of the actual pixels in the observation .
twenty thousand _ undersampled _ ccd stellar observations were simulated using an analytical gaussian with a @xmath195 @xmath121 ; the other simulation parameters were the same as before .
all the simulated observations were analyzed with @xmath0 using a 2@xmath572 _ supersampled _ discrete gaussian psf with @xmath195 @xmath121 ( @xmath205 @xmath166 ; @xmath206 ) .
figure [ fig9 ] shows the absolute photometric errors and total astrometric errors of this numerical experiment . by providing @xmath0 with extra information , in the form of a supersampled psf ,
the nyquist - shannon sampling theorem was no longer violated and excellent photometry and astrometry was done with this undersampled data set .
figure [ fig10 ] shows the relative errors for @xmath68 and @xmath117 .
the relative stellar intensity errors are normally distributed .
however , the relative @xmath117 position errors are almost , but not quite , normally distributed .
the @xmath0 code is accurately measuring the stellar positions ( i.e. , the median difference , @xmath207 , values are zero ) but the rms position error estimates ( @xmath180 ) are slightly _ underestimated _ ( the top and bottom whiskers for @xmath208 @xmath6 are seen to extend beyond the grey bands ) . the same effect is seen with @xmath127 . using a higher - resolution supersampled psf ( 3@xmath573 , 4@xmath574 , @xmath209 ) does not eliminate the small underestimation by @xmath0 of position errors .
_ the position errors estimated by @xmath0 are close to the photonic limit , but the actual errors for undersampled observations are close to the astrometric crlb with square ccd pixels _ @xcite .
* accurate and precise ccd stellar photometry and astrometry may be obtained with undersampled ccd observations if supersampled psfs are used during the psf - fitting process .
let us now investigate what happens when critically - sampled data is fit with a critically - sampled psf .
twenty thousand _ critically - sampled _
ccd stellar observations were simulated using an analytical gaussian with a @xmath210 @xmath121 ; the other simulation parameters were the same as before .
all the simulated observations were analyzed with @xmath0 using a _ critically - sampled _ discrete gaussian psf with @xmath210 @xmath121 ( @xmath211 @xmath166 ; @xmath206 ) .
figure [ fig11 ] shows the absolute photometric errors and total astrometric errors of this numerical experiment .
figure [ fig12 ] shows the relative errors for @xmath68 and @xmath117 .
looking carefully at figs .
[ fig11 ] and [ fig12 ] , one sees that the photometric and astrometric performance is well matched to the theoretical expectations except for the brightest three bins ( @xmath212 @xmath6 ) .
twenty thousand _ critically - sampled _
ccd stellar observations were simulated using an analytical gaussian with a @xmath210 @xmath121 ; the other simulation parameters were the same as before .
all the simulated observations were analyzed with @xmath0 using a 2@xmath572 _ supersampled _ discrete gaussian psf with @xmath210 @xmath121 ( @xmath211 @xmath166 ; @xmath206 ) .
figure [ fig13 ] shows the absolute photometric errors and total astrometric errors of this numerical experiment .
figure [ fig14 ] shows the relative errors for @xmath68 and @xmath117 .
the photometric and astrometric performance is well matched to the theoretical expectations for all stellar intensities .
comparing fig .
[ fig7 ] with fig .
[ fig9 ] and fig .
[ fig11 ] with fig .
[ fig13 ] , one sees that one can obtain excellent stellar photometry and astrometry with the matphot algorithm for all stellar intensities even if the observational data is undersampled as long as the discrete psfs used to do the model fitting are sampled finely enough to have sufficient spatial frequency coverage such that the nyquist - shannon sampling theorem is not violated . comparing fig .
[ fig3 ] with fig .
[ fig11 ] , one sees that the breakpoint for the matphot algorithm between undersampled and oversampled data is @xmath213 @xmath166 or , in terms of a gaussian full - width - at half maximum , @xmath214 @xmath121 . *
if a discrete psf is close to being critically sampled , then one should use a supersampled discrete psf which is oversampled in terms of supersampled pixels @xmath215 .
in other words , _ if the equivalent - background area is less than 21 square pixels ( @xmath216 @xmath166 ; gaussians : @xmath217 @xmath121 ) , then one should use a supersampled discrete psf which has an equivalent - background area of at least 21 square supersampled pixels ( @xmath218 @xmath219 ; gaussians : @xmath220 @xmath221 ) . _ let us now investigate the photometric and astrometric performance of the matphot algorithm with an ugly ( realistic ) space - based psf .
figure [ fig15 ] shows a simulated _ next generation space telescope _ ( _ ngst _ ) @xmath222-band ccd stellar observation .
this simulated observation used a 2@xmath572 supersampled psf which was based on a 8-meter trw - concept @xmath223 diffraction - limited primary mirror with 1/13 wave rms errors at @xmath223 ; the original version of the psf was kindly provided by john krist ( stsci ) . the six - sided `` snowflake '' pattern seen in fig .
[ fig15 ] is mainly due to fact that the primary mirror is composed of segmented hexagonal - shaped mirrors .
observers will note that this psf is very similar to optical psfs seen with the 10-m telescopes at the w. m. keck observatory .
the 6.5-m _ james webb space telescope _
( _ jwst _ ) is likely to have similar looking near - infrared psfs once it achieves first light in @xmath202011 . the _ ngst _ psf is so complicated that it is unlikely that it could be represented adequately with a continuous analytical mathematical function .
space - based observations frequently have high spatial frequencies which make them ideal candidates for photometric and astrometric analysis using discrete psfs .
twenty thousand ccd stellar observations were simulated using the simulated @xmath222-band _ ngst _
2@xmath572 supersampled psf described above ; the other simulation parameters were the same as before .
all the simulated observations were analyzed with @xmath0 with the psf used to create the simulated observations . figure [ fig16 ] shows the absolute photometric errors and total astrometric errors of this numerical experiment .
figure [ fig17 ] shows the relative errors for @xmath68 and @xmath117 .
the photometric and astrometric performance is well matched to the theoretical expectations for all stellar intensities .
although only gaussian psfs were used in previous numerical experiments , the excellent fit seen in the top panel of fig .
[ fig16 ] between the theoretical photometric performance model ( 2.3.4 ) and actual @xmath0 measurements using such an ugly discrete psf is not surprising once one remembers that the theoretical photometric performance model was derived from an abstract point response function .
the development of the theoretical astrometric performance model , however , required differentiation of the point response function which i assumed to be an oversampled analytical gaussian function .
the analytical gaussian bright star astrometric limit was transformed to the general form by assuming that the gaussian - specific @xmath224 term could be replaced with the more general @xmath225 term , which , by definition , can be computed for any prf .
the same assumption was then used to derive the general faint star astrometric limit .
the excellent fit seen in the bottom panel of fig .
[ fig16 ] indicates that this bold assumption is not only useful but practical .
many numerical experiments with very ugly discrete psfs have shown that the theoretical astrometric performance model of 2.4.4 works well with ugly discrete point response functions .
if the matphot algorithm is optimally extracting photometric and astrometric information from a stellar observation , _ and _ @xmath0 has been correctly coded , _ and _ the ccd observation has been properly calibrated , _ and _ the prf used in the observational model is correct , _ and _ accurate estimates of the measurements errors for each pixel have been made , _ then _ one expects that the @xmath59 goodness - of - fit value reported by @xmath0 to be distributed as a @xmath59 distribution with the number of degrees of freedom equal to the difference between the number of pixels in the observation and the number of free parameters .
figure [ fig18 ] shows that this prediction about the precision and accuracy of the matphot algorithm has been verified : the cumulative distribution of the @xmath59 reported by @xmath0 ( _ thin black curve _ ) is seen to lie on top of the cumulative @xmath59 distribution of for 3596 [ @xmath226 @xmath227 pixels @xmath201 4 free parameters @xmath228 , @xmath117 , @xmath127 , and @xmath229 degrees of freedom ( _ thick grey curve _ ) .
* the @xmath0 code works well with ugly discrete psfs and its performance can be well predicted using the general theoretical photometric and astrometric performance models given in 2 . *
the @xmath59 goodness - of - fit value reported by @xmath0 is a statistically reliable measure of the quality of a photometric and astrometric reduction of a stellar observation obtained with the matphot algorithm using ugly discrete psfs .
let us now investigate the photometric and astrometric performance of the matphot algorithm with an ugly psf and an ugly detector .
suppose one has a detector where every pixel has been divided up into 16 square regions called `` subpixels '' .
let us call the first row and first column of subpixels `` gate structures '' which are optically inactive with 0% qe . the remaining 9 subpixels are the optically active part of the pixel with a 100% qe .
by definition , such a pixel would have a very large intrapixel qe variation with only 56.25% of the total pixel area being capable of converting photons to electrons .
a few extra lines of code were added to the @xmath0 program to simulate the image formation process with such an ugly detector .
the new version of @xmath0 is called @xmath230 and was designed specifically for use with 4@xmath574 supersampled psfs . ten thousand ccd stellar observations of @xmath231 mag stars ( @xmath20@xmath232 @xmath5 ) were simulated and analyzed with @xmath230 using a 4@xmath574 supersampled version of the simulated @xmath222-band _ ngst _ psf described above .
the _ observed _ background level was assumed to be a constant value of @xmath233 @xmath6 ( @xmath189 @xmath5 , @xmath234 ) and all other simulation parameters were the same as before .
the measured prf volume of these simulated observations was @xmath235 which is consistent with the expected value of 0.5625 from the physical structure of a single pixel .
the median and semiquartile range of the effective - background area @xmath236 of these observations was , respectively , 28.10 and 4.82 @xmath237 .
the median critical - sampling scale length of these observations was @xmath238 @xmath121 indicating that these observations were _ undersampled _ , as expected .
the optically inactive gate structures of the pixel cause the observed number of electrons in each stellar image to be significantly less than the number of photons which fell on the detector . _
the total amount of loss was dependent on where the center of the star fell within the central pixel of the stellar image .
_ figure [ fig19 ] shows that stars centered in the middle of the active area of a pixel suffered a @xmath2040% loss ( @xmath239 mag ) while those centered on gate structures _ ( grey points ) _ lost up to 47% ( @xmath240 mag ) .
although this numerical experiment may seem to be very artificial , large intrapixel sensitivity variations can be found in cameras currently installed on the _ hubble space telescope_. @xcite reported peak - to - peak variations of 0.39 mag at the @xmath241 band ( f110w ) and 0.22 mag at the @xmath242 band ( f160w ) of the nic3 camera of the _ hst _ nicmos instrument .
the peak - to - peak variation of @xmath200.2 mag at f160w with nic 3 was independently confirmed by @xcite .
the mean _ observed _ stellar magnitude for these @xmath231 mag stars was @xmath243 mag .
the photometric performance model predicts an rms measurement error of @xmath244 mag for these bright stars . with an average loss of 44% and an rms measurement error that is _ ten times larger _ than expected from photon statistics , the observed stellar magnitudes were neither precise or accurate .
figure [ fig20 ] shows that @xmath230 was able to do an excellent job in recovering the true stellar magnitude of the 10,000 @xmath231 mag stars despite being presented with a worst - case scenario of undersampled observations with an ugly psf imaged on an ugly detector with a very large intrapixel qe variation .
the mean _ measured _ stellar magnitude reported by @xmath230 was @xmath245 mag and the mean rms error estimated by @xmath230 was @xmath246 mag .
the photometric performance of @xmath230 is fully consistent with theoretical expectations which were derived for an ideal detector with no intrapixel qe variation .
twenty thousand ccd stellar observations were simulated and analyzed with @xmath230 using the same 4@xmath574 supersampled version of the simulated @xmath222-band _ ngst _ psf .
the input stellar intensities ranged from @xmath156 to @xmath157 mag ( @xmath247 to @xmath77 @xmath5 ) .
the observed background level was assumed to be a constant value of @xmath233 @xmath6 ( @xmath189 @xmath5 , @xmath234 ) and all other simulation parameters were the same as before .
the median and semiquartile range of the effective - background area @xmath236 of these observations was , respectively , 28.04 and 4.77 @xmath237 .
figure [ fig21 ] shows the absolute photometric errors and total astrometric errors of this numerical experiment . comparing the simulation results with the grey theoretical limits
, one sees that the photometric and astrometric performance of the @xmath230 code is well predicted by the theoretical performance models given in 2 . *
excellent stellar photometry and astrometry is possible with ugly psfs imaged onto ugly detectors as long as the image formation process _ within the detector _ is accurately modeled by the photometric reduction code .
after developing theoretical photometric and astrometric performance model for point spread function ( psf)-fitting stellar photometry , i described the unique features of the matphot algorithm for accurate and precise stellar photometry and astrometry using discrete point spread functions .
i conducted numerical experiments with the @xmath0 implementation of the matphot algorithm and demonstrated that the computational noise due to the chosen analysis method ( numerical versus analytical ) is insignificant when compared to the unavoidable photon noise due to the random arrival photons in any astronomical ccd observation .
the matphot algorithm was specifically designed for use with space - based stellar observations where psfs of space - based cameras frequently have significant amounts of power at higher spatial frequencies .
using simulated _ ngst _ ccd observations , i demonstrated millipixel relative astrometry and millimag photometry is possible with very complicated space - based discrete psfs .
the careful reader will observe that i have not discussed _ how _ a discrete psf is derived .
the matphot algorithm will optimally determine the brightness and position of a star in a ccd observation when provided with the correct psf and drf functions which need to be determined _ beforehand _ through calibration procedures .
photometric and astrometric accuracy and precision degrades if either the psf or drf is poorly known .
psf reconstruction ( calibration ) is a complicated topic in its own right and has been the subject of many articles , instrumentation reports , and entire workshops .
the challenges of psf reconstruction are many .
an astronomer may be faced with trying to derive a psf from an observation * with a variable psf within the field of view , * that has too few bright stars , * that might be undersampled , * that might be poorly dithered , * that might be poorly flat - fielded , * that exhibits significant charge transfer efficiency variations , * that has variable charge diffusion within the ccd substrate , * with significant photon loss due to charge leakage , * that might not actually be linear below the 1% level . while many of these problems can be overcome by the proper design of instruments or experiments , their solution is beyond the scope of this article which has sought to determine the practical limits of psf - fitting stellar photometry .
the analysis presented in this article has assumed that psfs are perfectly known a situation that is rarely , if ever , physically possible .
the _ cores _ of observationally based psfs are generally much better determined than the broad _ wings _ due to simple photon statistics
. the effect of large instrumental calibration errors can also be significant .
for example , flat - field limitations can dramatically impact the achievable levels of photometric and astrometric precision .
an investigation based on theory of the effect of psf errors and flat - field calibration error on the limits of psf - fitting stellar photometry would be very difficult .
an investigation based on numerical experiments , however , might be a much more tractable proposition . in any case ,
a through investigation of the effects of calibration errors on the limits of psf - fitting stellar photometry is best left to another article .
i would like to thank the anonymous referee whose careful reading and insightful comments have improved this article .
special thanks are due to ian roederer who checked all of the mathematical proofs in an early draft of the manuscript .
jessica moy has my heartfelt thanks for cheerfully helping me acquire copies of many articles in technical journals that are not available in the noao library collection .
i would also like to thank the following people for providing stimulating conversations , encouragement , and support during the extensive research and development effort behind the creation of the matphot algorithm and and its various software implementations : joseph bredekamp , susan hoban , dot appleman , taft armandroff , nick buchholz , marc buie , harvey butcher , julian christou , chuck claver , lindsey davis , michele de le pea , mike fitzpatrick , ken freeman , dan golombek , richard green , steve howell , george jacoby , buell jannuzi , stuart jefferies , ivan king , john krist , todd lauer , john mackenty , john mather , mike merrill , dave monet , mark morris , jeremy mould , jan noordam , john norris , sudhakar prasad , the late alex rogers , steve ridgway , pat seitzer , james schombert , donald west , and sidney wolff .
this research has been supported in part by the following organizations , institutions , and research grants ( in reverse chronological order ) : national aeronautics and space administration ( nasa ) , interagency order no .
s-13811-g , which was awarded by the applied information systems research program ( nra 01-oss-01 ) of nasa s science mission directorate , interagency order no .
s-67046-f which was awarded the long - term space astrophysics program ( nra 95-oss-16 ) of nasa s office of space science , center for adaptive optics ( cfao ) , institute for pure and applied mathematics ( ipam ) , kitt peak national observatory , national optical astronomy observatory , association of universities for research in astronomy inc .
( aura ) , national science foundation , mount stromlo and siding spring observatories , australian national university , the netherlands foundation for astronomical research ( astron ) , and the netherlands organization for the advancement of pure research ( zwo ) , and the kapteyn astronomical institute of groningen . 99 abramowitz , m. , & stegun , i. 1964 , handbook of mathematical functions with formulas , graphs , and mathematical tables , applied mathematics series , 55 , eds .
m. abromowitz and i. stegun ( washington , d.c .
: nbs ) arfken , g. b. 1970 , mathematical methods for physicists , 2nd ed .
, ( new york : academic press ) biretta , j. , et al .
2001 , wfpc2 instrument handbook , version 6.0 ( baltimore : stsci ) hook , r.n . , & fruchter , a.s .
2000 , in asp conf .
216 , astronomical data analysis software and systems ix , eds .
n. manset , c. veillet , and d. crabtree ( san francisco : asp ) , 521 irwin , m.j .
1985 , mnras , 214 , 575 , p. , greenfield , p. , & jedrzejewski , r. 1992 , a&a , 253 , 329 king , i.r . 1971 ,
pasp , 83 , 199 king , i.r .
1983 , pasp , 95 , 163 lauer , t.r . 1999 ,
pasp , 111 , 1434 levenberg , k. 1944 , q. app .
, 2 , 164 marquardt , d. 1963 , j. siam , 11 , 431 mighell , k. j. 1989 , mnras , 238 , 807 mighell , k. j. 1999 , in asp conf . ser .
172 , astronomical data analysis software and systems viii , eds.d .
m. mehringer , r. l. plante , and d. a. roberts ( san francisco : asp ) , 317 mighell , k. j. 2002 , in proceedings of spie , vol . 4847 , astronomical data analysis ii , eds .
starck and f. d. murtagh ( bellingham , wa : spie ) , 207 pogson , n. 1856 , mnras , 17 , 12 , m. a. c , jakobsen , p. , colina , l. , lelievre , g. , macchetto , f. , nieto , j. l. , & di serego alighieri , s. 1989 , a&a , 215 , 195 press , w. h. , flannery , b. p. , teukolsky , s. a. , & vetterrling w. t. 1986 , numerical recipes .
the art of scientific computing ( cambridge : cambridge university press ) tukey , j. w. 1977 , exploratory data analysis , ( reading , ma : addison - wesley ) wells , d. c. , greisen , e. w. , & harten , r. h. 1981 , a&as , 44 , 363 winick , k. a. 1986 , josa a , 3 , 1809
a box - and - whisker plot ( a.k.a . _ box plot _ ) is a graphical method of showing a data distribution .
a box is drawn showing the inner quartile range of the data which , by definition , includes half of all the data values ( see fig .
[ figa1 ] ) .
median _ of the data is shown with a bar inside the box .
the bottom end of the box is the lower quartile ( 25% ) of the data ; @xcite , the creator of the box - and - whiskers plot , calls this value the _ lower hinge _ , @xmath248 , value .
the top end of the box is the upper quartile ( 75% ) of the data which is called the _ upper hinge _ , @xmath249 , value .
the _ step _
value is 1.5 times the inner quartile range : @xmath250 .
the _ top fence _
value is the sum of the upper hinge and step values : @xmath251 . the _ bottom fence _
value is the difference between the lower hinge and step values : @xmath252 .
the _ top whisker _ is drawn from the upper hinge value to the largest data value that is less than or equal to the top fence value : @xmath253 .
similarly , the _ bottom whisker _ is drawn from the lower hinge value to the smallest data value that is greater than or equal to the bottom fence value : @xmath254 .
data values that are greater than the top fence value or less than the bottom fence value are called _ outliers _ and are plotted at their appropriate value beyond the whiskers . for a normal distribution , which is a gaussian distribution with a mean of zero and a standard deviation of one ,
the bottom fence , bottom hinge , median , top hinge and top fence values are , respectively , @xmath2012.6980 ( 0.35% cumulative fraction ) , @xmath2010.6745 ( 25% ) , zero ( 50% ) , 0.6745 ( 75% ) , 2.6980 ( 99.65% ) .
figure [ figa1 ] shows a data set of 1000 normal deviates .
the histogram of the data with 0.1-wide bins is shown with thin black lines .
the cumulative fraction distribution of the data is shown as a thick gray curve .
the box - and - whisker plot of the data is shown with thick black lines below the histogram ; arrows show the relationship between various box values and the cumulative fraction distribution .
the mean and standard deviation of this data set are , @xmath2010.0341 and 0.9739 , respectively . the bottom fence , bottom whisker , bottom hinge , median , top hinge , top whisker , and top fence values of this data set are , respectively , @xmath2012.5511 ( 0.25% cumulative fraction ) , @xmath2012.4940 ( 0.30% ) , @xmath2010.6522 ( 25.10% ) , @xmath2010.0231 ( 50.00% ) , 0.6137 ( 75.10% ) , 2.4580 ( 99.50% ) , 2.5126 ( 99.57% ) .
the seven outlier values of this data set , @xmath2012.9500 , @xmath2012.6320 , 2.5390 , 2.7150 , 2.7430 , 2.8270 , 2.8530 , are plotted in fig . [ figa1 ] as diamonds beyond the whiskers .
|
the key features of the matphot algorithm for precise and accurate stellar photometry and astrometry using discrete point spread functions are described . a discrete point spread function ( psf )
is a sampled version of a continuous psf which describes the two - dimensional probability distribution of photons from a point source ( star ) just above the detector .
the shape information about the photon scattering pattern of a discrete psf is typically encoded using a numerical table ( matrix ) or a fits image file .
discrete psfs are shifted within an observational model using a 21-pixel - wide damped sinc function and position partial derivatives are computed using a five - point numerical differentiation formula .
precise and accurate stellar photometry and astrometry is achieved with undersampled ccd observations by using supersampled discrete psfs that are sampled 2 , 3 , or more times more finely than the observational data .
the precision and accuracy of the matphot algorithm is demonstrated by using the c - language @xmath0 code to analyze simulated ccd stellar observations ; measured performance is compared with a theoretical performance model .
detailed analysis of simulated _ next generation space telescope _ observations demonstrate that millipixel relative astrometry and millimag photometric precision is achievable with complicated space - based discrete psfs .
[ firstpage ] techniques : image processing , photometric astrometry instrumentation : detectors methods : analytical , data analysis , numerical , statistical
|
following previous work @xcite , the multiplicity distribution ( md ) generating function in the dummy variable @xmath0 , @xmath1 has been recognised to be related to the grand canonical partition function of a system of @xmath2 particles in statistical mechanics when @xmath0 is identified with the fugacity variable ; @xmath3 is here the probability of detecting @xmath2 charged particles in full phase space ( @xmath4 ) and it is linked to the canonical partition function .
it has to be stressed that in a realistic experiment @xmath2 will never become infinity .
let us call @xmath5 the maximum finite value of @xmath2 which can be detected in an experiment .
the generating function @xmath6 in this case is reduced to a polynomial of degree @xmath5 in @xmath0 : @xmath7 let us now consider the algebraic equation @xmath8 in the complex @xmath0-plane . in the first place , notice that none of the roots @xmath9 @xmath10 of equation ( [ eq:3 ] ) can be real and positive , since the @xmath3 in ( [ eq:2 ] ) are positive for any @xmath2 . in the second place ,
recall that the @xmath5-truncated generating function ( as any polynomial ) can be factorised in terms of its roots as follows : @xmath11 by applying these considerations to phase transitions of a lattice gas ( and of the ising model ) , c.n .
yang and t.d .
lee @xcite have found that the roots @xmath9 lie on a circle centered at the origin of the complex @xmath0-plane .
the circle is open in a small sector bisected by the positive real axis ; no zero lies in that sector .
when @xmath5 is odd , at least one root is real : it will of course be on the negative side .
@xmath12 turns out to be an analytic function along the positive real axis for any @xmath5 .
a phase transition in the sense of a non analytic behaviour of the thermodynamical functions ( like pressure for instance ) can only occur at points of the positive real @xmath0 axis which are `` in the thermodynamical limit '' accumulation points of the zeroes of the algebraic equation ( [ eq:3 ] ) . for @xmath13
, one should expect in general that the zeroes become closer and closer to the real positive axis and that the distance between the zeroes vanishes .
the first application of these ideas to particle production goes back to the 70 s @xcite , but the field flourished again ( results were obtained in the context a discrete approximation to qcd cascades @xcite ) in 1995 , when e. de wolf @xcite studied , using the jetset monte carlo generator , the zeroes of the generating function of the @xmath2-charged particle multiplicity distribution in @xmath14 annihilation in the @xmath15-rapidity interval @xmath16 at 1000 gev c.m .
the zeroes were found to lie on a circle of approximately unit radius centered at the origin of the complex @xmath0-plane .
intriguing questions were also raised on the development of this approach in different models and in different classes of events . after recalling and extending the main theorems on the subject
, the present paper explores the properties of the distribution of zeroes in high energy collisions in the framework of the weighted superposition mechanism of different classes of events : each class is described by a negative binomial ( pascal ) md , whose characteristic parameters are @xmath17 , the average charged particle multiplicity , and @xmath18 , linked to the dispersion @xmath19 by @xmath20 .
in 1997 , t.c . brooks , k.l . kowalsky and c.c .
taylor @xcite applied the enestrm - kakeya theorem @xcite to the scaled @xmath5-truncated generating function @xmath21 with coefficients given by @xmath22 and the parameter @xmath23 defined as the minimum of the ratio of two consecutive @xmath24 coefficients : @xmath25=\min[p_{n}/p_{n+1}]=p_{n_{\min}}/p_{n_{\min}+1}. \label{eq:6}\ ] ] the coefficients @xmath24 , for @xmath26 , are real and positive and satisfy the condition @xmath27 required by the enestrm - kakeya ( ek ) theorem in order to ensure that @xmath28 differs from zero in the complex @xmath0-plane for @xmath29 . equation ( [ eq:5 ] ) can be rewritten in terms of the variable @xmath30 as @xmath31 with @xmath32 and @xmath33 by choosing now the maximum of the ratio between two consecutive @xmath34 coefficients : @xmath35=\max[p_{n}/p_{n+1}]=p_{n_{\max}}/p_{n_{\max}+1}\ ] ] one finds that the coefficients @xmath34 satisfy once again the ek theorem : they are indeed real and positive and ordered following the rule @xmath36 .
therefore @xmath37 differs from zero for @xmath38 and the zeroes of @xmath37 lie necessarily in the region @xmath39 . accordingly ,
all the zeroes of the truncated generating function @xmath12 fall in the annular region in the complex @xmath0-plane centered at the origin and limited by @xmath40 $ ] and @xmath41 $ ] .
the positive real axis is of course excluded .
the mentioned theorem turns out to be quite useful when applied to the class of power series distributions , defined as follows : @xmath42 the corresponding @xmath5-truncated generating function is @xmath43 where @xmath44 is a normalisation factor . in the polynomial ( [ eq:11 ] ) ,
the term @xmath45 can be neglected and @xmath46 replaced by the variable @xmath47 , producing a new scaled generating function in @xmath47 : @xmath48 which as we will see in the following is of particular interest for classifying different classes of events described in terms of nb ( pascal ) md with different parameters .
in fact , applying the above considerations to the generating function @xmath49}(z)=\sum_{n=0}^{\infty}p_{n}^{[nb]}\,z^n=\left(\frac{k}{k+\bar{n}(1-z)}\right)^k \label{eq:13}\ ] ] with @xmath50}=\frac{k(k+1 ) \dots ( k+n-1 ) { \bar{n}}^n k^k}{n ! { ( \bar{n}+k)}^{n+k } } , \label{eq:14}\ ] ] leads to the @xmath5-truncated version of @xmath51}(z)$ ] : @xmath52}(z)=a_{m}^{[nb]}\sum_{n=0}^{m}p_{n}^{[nb ] } z^n = a_{m}^{[nb]}\sum_{n=0}^{m}\prod_{i=1}^{n}\left(\frac{k+i-1}{i}\right)p_{0 } b^n z^n \label{eq:15}\ ] ] with @xmath53 .
notice that in ( [ eq:13 ] ) , @xmath51}(z)$ ] has a singularity for @xmath54 . by defining now in ( [ eq:15 ] ) @xmath55}=\prod_{i=1}^{n}\left(\frac{k+i-1}{i}\right )
\label{eq:16}\ ] ] and @xmath56 the new renormalised generating function , neglecting the inessential factor @xmath45 , becomes @xmath57}(u)=a_{m}^{[nb]}\sum_{n=0}^{m}\prod_{i=1}^{n}\left(1+\frac{k-1}{i}\right ) u^n = a_{m}^{[nb]}\sum_{n=0}^{m}c_{n}^{[nb ] } u^n .
\label{eq:17}\ ] ] notice that the not - truncated version @xmath58}(u)$ ] of ( [ eq:17 ] ) has a singularity at @xmath59 .
relation ( [ eq:17 ] ) has been used in its not - truncated form in ref .
@xcite where @xmath60}$ ] has been identified with the @xmath2-particle _ canonical partition function _ ; @xmath61}(u)$ ] is in this framework the _ truncated grand canonical partition function_. it follows that if parameter @xmath18 is less than one ( as in the case of the third class of events in pp collisions at 14 tev c.m .
energy @xcite ) , then the coefficients @xmath60}$ ] are all real and positive , and decrease starting from @xmath62}$ ] .
the conditions for the applicability of the ek theorem are therefore also here satisfied , and the zeroes of the scaled generating function ( [ eq:17 ] ) lie all outside the circle of unit radius @xmath63 .
this situation should be contrasted with the application of the ek theorem when the parameter @xmath18 is greater than one ( as in the case of the soft and semi - hard classes of events in pp collisions @xcite ) . in this case in fact , the coefficients @xmath60}$ ] are in increasing order and the zeroes lie all inside the circle of unit radius .
as indicated in @xcite and shown in @xcite , for the ( not - rescaled ) nb ( pascal ) md eq .
( [ eq:15 ] ) , one finds @xmath64=\frac{p_{n}^{[nb]}}{p_{n+1}^{[nb]}}=\frac{1}{b}\left(\frac{n+1}{n+k}\right ) ; \label{eq:18}\ ] ] thus all zeroes of the algebraic equation of the @xmath5-truncated generating function ( [ eq:3 ] ) lie inside the annular region of the complex @xmath0-plane delimited by @xmath40 $ ] and @xmath41 $ ] , with @xmath65=\frac{1}{b}\left(\frac{m+1}{m+k}\right)\leq | z | \leq
r[n_{\min}]=\frac{1}{kb}\equiv\frac{1}{a}\;\;\;\;(k<1 ) \label{eq:19}\ ] ] and @xmath66=\frac{1}{kb}=\frac{1}{a}\leq | z | \leq r[n_{\max}]=\frac{1}{b}\left(\frac{m+1}{m+k}\right)\;\;\;(k>1 ) .
\label{eq:20}\ ] ] notice that @xmath67 \rightarrow \frac{1}{b}$ ] when @xmath68 , and @xmath67 \rightarrow 1 $ ] when @xmath69 ( i.e. @xmath70 ) . in the rescaled version ( [ eq:17 ] ) , @xmath67=\frac{n+1}{n+k}$ ]
: it tends to 1 when @xmath71 with constant @xmath18 .
it is remarkable that for @xmath72 , @xmath67 $ ] is also equal to one and all zeroes lie on the circle of unit radius , the @xmath59 point being excluded ( see figures 1 and 2 ) .
the role of parameter @xmath73 in the case of the nb ( pascal ) md is underlined by the fact that the sum of the inverse of the zeroes of the algebraic equation ( [ eq:3 ] ) is given by @xmath74 recall in fact that @xmath75 and also that the sum of ( @xmath76 ) by ( @xmath76 ) product of zeroes equals to @xmath77 for example for @xmath78 one gets @xmath79 it follows that @xmath80 notice that for the nb ( pascal ) md , eq . ( [ eq:21 ] ) becomes @xmath81 equation ( [ eq:21 ] ) is indeed a _ general property _ of polynomials and is _ independent _ of the value of @xmath5 . in the @xmath47-rescaled version , one has @xmath82 in addition , it should be pointed out that the nb parameters of the coefficients of the @xmath5-truncated generating function ( [ eq:15 ] ) , namely @xmath83 and @xmath84 , can be calculated directly from the zeroes of equation ( [ eq:3 ] ) .
one finds : @xmath85 and @xmath86 in the limit @xmath13 , the parameters @xmath83 and @xmath84 coincide of course with the standard parameters of the not - truncated distributions .
one question still to be answered concerns the multiplicity of zeroes of the algebraic equation ( [ eq:3 ] ) , both in general and in the specific case of the nb ( pascal )
let us consider two polynomials @xmath87 and @xmath88 of degree @xmath89 and @xmath2 respectively : @xmath90 and @xmath91 it is known that @xmath87 and @xmath88 possess a not constant common factor iff the determinant @xmath92 is equal to zero .
it follows that , called @xmath93 the first derivative of the polynomial @xmath87 with respect to @xmath0 , @xmath87 will have at least one double root iff @xmath94 . in order to be sure that the zeroes of a @xmath5-truncated generating function @xmath12 are all distinct , it is therefore sufficient to show that none of the m roots @xmath95 is such that @xmath96 . by rewriting @xmath97}(\bar n , k$ ] ) for simplicity in terms of parameters @xmath73 and @xmath98 ( see equation ( [ eq:19 ] ) for the definition of @xmath73 )
, one has @xmath50}(a , b)=\frac{1}{n!}a(a+b ) \dots ( a+b(n-1))p_{0 } , \label{eq:26}\ ] ] and by properly normalising @xmath12 one finds , by induction , @xmath99},{g\,'}_{m}^{[nb]}}=a_{m}^{[nb]}\frac{a^m{[(a+b)(a+2b ) \dots ( a+mb)]}^{m-1}{p_{0}}^{m-1}}{{(m!)}^{m-1}}. \label{eq:27}\ ] ] since clearly @xmath100},{g\,'}_{m}^{[nb]}}$ ] is always positive for @xmath73 and @xmath98 greater than zero , it follows that all the zeroes of the algebraic equation @xmath101}(z)=0 $ ] are distinct . since all zeroes of @xmath101}(z)=0
$ ] are distinct , @xmath5 conditions are needed to uniquely determine the set of zeroes . for @xmath72 for instance , one should expect that @xmath102 in fact , from ( [ eq:16 ] ) @xmath103}}{c_{m}^{[nb]}}=-\frac{m}{m+k-1 } \stackrel{k \rightarrow 1}{\longrightarrow } -1.\ ] ] the same limit is obtained for @xmath104}}{c_{m}^{[nb]}}\right)^2 - 2\frac{c_{m-2}^{[nb]}}{c_{m}^{[nb ] } } \stackrel{k \rightarrow 1}{\longrightarrow } -1\ ] ] as well as for the other powers up to @xmath5 .
notice that the same conditions are satisfied for @xmath105 and constant @xmath18 : consequently , also in this case the zeroes stay on the circle of unit radius , as will be discussed in the following .
we propose in this section to apply the previously discussed theorems and remarks to the study , in the rescaled complex @xmath0-plane ( i.e. , in the complex @xmath47-plane ) , of maps of zeroes of the algebraic equations obtained by truncating at integer value @xmath5 the nb ( pascal ) md generating function .
this distribution is used in a statistical model of multiparticle production to describe different classes of events ( or substructures ) in different collisions .
these substructures are for instance 2- and 3-jets events in @xmath14 annihilation or events without mini - jets ( soft events ) and with mini - jets ( semi - hard events ) in pp collisions .
recently @xcite , in pp collisions , the occurrence of a third class of hard events at 14 tev c.m .
energy has been postulated also in the framework of the weighted superposition mechanism of different classes of events , each described by a nb ( pascal ) md with characteristic nb parameters .
the occurrence of this third class of events was motivated by the fact that , by assuming both strong kno scaling violations and qcd behaviour , extrapolations of nb parameters in the lhc energy domain implied , in the semi - hard sample of events , a surprising reduction in the average number of clans @xcite . in order to deepen the analysis of the properties of collective variables in terms of the maps of complex zeroes and in order to study the possible occurrence of a phase transition , we focus our attention on the three mentioned classes of events at 14 tev c.m .
energy in pp collisions . accordingly , the nb parameters used in the following are those described in @xcite : @xmath106 @xmath107 .
we find that for the soft and semi - hard components , the parameter @xmath18 is greater than one and the zeroes stay _ inside _ the circle of unit radius in the complex @xmath47-plane ( see figures 1a and 1b ) . for the hard component ,
the parameter @xmath18 is smaller than one and all the zeroes lie _ outside _ the circle of unit radius @xmath108 ( see figure 1c ) . in order to make our statements more transparent , in figures 2a , 2b , 2c the behaviour of the three classes of events in the region around the point
@xmath59 has been magnified for different cut - offs but fixed value of @xmath18 . in figure 3
the above maps are shown together close to the point @xmath59 , with the same fixed cut - off for three values of @xmath18 corresponding to the three classes of events .
notice that for @xmath72 all the zeroes stay on the circle of unit radius with the @xmath59 point excluded .
in addition , when @xmath13 the distribution of the zeroes of the algebraic equation obtained from the truncated and rescaled generating function converges towards the @xmath59 point for all classes of events with both @xmath109 and @xmath110 .
as mentioned in the introduction , in order to test the occurrence of a phase transition one has to find the behaviour of the zeroes of the grand canonical partition function @xmath111 in the thermodynamical limit , i.e. , with the size of the system going to infinity .
the discussion is based on the approach contained in @xcite , where the generating function @xmath6 has been related to @xmath111 by @xmath112^{-1}. \label{eq:28}\ ] ] relation ( [ eq:28 ] ) implies that the zeroes of the @xmath5-truncated grand canonical partition function @xmath113 can be obtained from the zeroes of the @xmath5-truncated generating function @xmath114 simply by rescaling those zeroes by a factor @xmath98 ( which was identified in @xcite it should be stressed again with the fugacity ) .
the behaviour of the zeroes will then depend on the volume and the number of particles : when the md is of nb type , as we are using now , it is sufficient to know the behaviour of the nb parameters . at present however , a realistic link between @xmath18 and @xmath17 and the system size ( the volume ) is under investigation , hence postponed to a forthcoming paper . here
we limit ourselves to explore what happens when the thermodynamical limit corresponds to @xmath5 ( and @xmath17 ) going to infinity with @xmath18 fixed . in the first place we will check whether the ek theorem can be used to confine the zeroes to a narrowing region of the complex plane .
recalling eq .
( [ eq:18 ] ) , we notice that for @xmath72 one has @xmath67=\frac{1}{b}$ ] .
thus minimum and maximum radii coincide and all the zeroes of @xmath12 lie on a circle of radius @xmath115 , independently of the chosen cut - off @xmath5 .
numerical studies show that the opening around the positive real axis is reduced when @xmath13 ; this suggests that @xmath54 is an accumulation point of zeroes .
it means of course that in this scenario for @xmath72 , the point @xmath59 is an accumulation point of zeroes of the grand canonical partition function : since @xmath47 on the real axis coincides with the fugacity , only for @xmath116 will a phase transition take place .
for @xmath109 the zeroes of the grand canonical partition function obey the following relation @xmath66=\frac{1}{k } \leq | u | \leq r[n_{\max}]=\frac{m+1}{m+k } \label{eq:29}\ ] ] while for @xmath110 the two limits are exchanged .
such a region does not shrink .
we have to rely again on numerical root - finding , which still indicates the closing in of the zeroes to @xmath59 ( see figures 1 and figures 2 ) , and at the same time shows that they are much closer together than the bound given by relation ( [ eq:29 ] ) allows .
this feature again points in the direction of the presence of a possible phase transition at @xmath116 . in conclusion ,
the use of the rescaled complex variable @xmath117 marks visibly the different behaviour of the map of zeroes of the hard class of events with respect to the maps of the soft and semi - hard ones in our statistical model .
in addition , it leads to identify the coefficients @xmath60}$ ] of eq .
( [ eq:16 ] ) with the canonical partition function for a system of @xmath2 particles , and @xmath118 with the truncated grand canonical partition function of the same system ( in this framework @xmath119 , i.e. , the parameter @xmath18 coincides with the canonical partition function for a system with one particle , see also @xcite ) .
it should also be pointed out that in the rescaled version of this approach the nb parameter @xmath18 alone controls the geometry of the distribution of the zeroes .
this is indeed a quite general property which is expected to be valid also in other collisions and especially in those collisions involving events with large numbers of particles , like heavy ion collisions .
finally , it has been shown that the nb parameter @xmath18 is equal to minus the sum of the inverse of the zeroes , independently of the value of the cut - off @xmath5 .
all these facts suggest to investigate further the thermodynamical content of the nb parameter @xmath18 , and especially its dependence on temperature and volume , in view of a possible phase transition in the @xmath116 limit .
being indeed the fugacity @xmath120 ( with @xmath121 the chemical potential ) , one finds that @xmath121 goes to 0 in the limit @xmath116 : this is a situation which is reached in a high temperature and high density real boson gas and which is expected to occur in a quark - gluon plasma .
7 a.giovannini , s.lupia and r.ugoccioni , phys.rev.d 65 , 094028 ( 2002 ) c.n.yang and t.d.lee , phys.rev .
87 , 404 ( 1952 ) ; 87 , 410 ( 1952 ) k.j .
biebl and j. wolf , nucl.phys .
b 44 , ( 1972 ) 301 b. andersson , g. gustafson and j.samuelsson , nucl.phys .b 463 , ( 1996 ) 217 e. de wolf , in proceedings of the xiv int . symp . on multiparticle dynamics , eds .
a.giovannini , s.lupia , r.ugoccioni , world scientific , singapore 1995 , p.15 t.c.brooks , k.l.kowalski and c.c.taylor , phys.rev.d 56:58575861 , 1997 m.marden , _ geometry of polynomials _
( mathematical surveys number 3 , american mathematical society , providence , rhode island , 1966 , p.136137 ) a.giovannini and r.ugoccioni , phys.rev.d 59 , 094020 ( 1999 ) and phys.rev.d 60 , 074077 ( 1999 ) a.giovannini and r.ugoccioni , phys.rev.d 68 , 034009 ( 2003 ) fig .
1a : rescaled map of zeroes , in the complex @xmath47-plane , of the nb ( pascal ) md at 14 tev c.m .
energy with @xmath122 and cut - off @xmath123 ( @xmath5 has been taken equal to @xmath124 ) . zeroes _ lie inside _ the circle of unit radius @xmath108 .
+ fig .
1b : rescaled map of zeroes , in the complex @xmath47-plane , of the nb ( pascal ) md at 14 tev c.m .
energy with @xmath125 and cut - off @xmath126 ( @xmath5 has been taken equal to @xmath127 ) . zeroes _ lie inside _ the circle of unit radius @xmath108 .
+ fig .
1c : rescaled map of zeroes , in the complex @xmath47-plane , of nb ( pascal )
md at 14 tev c.m .
energy with @xmath128 and cut - off @xmath129 . zeroes _ lie outside _ the circle of unit radius @xmath108 .
+ fig .
2a : the rescaled map of zeroes of the nb ( pascal ) md of fig .
1a has been magnified around the point @xmath59 for different cut - offs @xmath5 ( white and black triangles correspond to @xmath123 and @xmath130 respectively , and white and black squares to @xmath131 and @xmath132 ) . as @xmath5 increases
, the point @xmath59 is shown to become an accumulation point of zeroes _ from the inside_. + fig .
2b : the rescaled map of zeroes of the nb ( pascal ) md of fig .
1b has been magnified around the point @xmath59 for different cut - offs @xmath5 ( white triangles correspond to @xmath126 , white squares to @xmath133 and black triangles to @xmath134 ) . as @xmath5 increases
, the point @xmath59 is shown to become an accumulation point of zeroes _ from the inside_. + fig . 2c : the rescaled map of zeroes of the nb ( pascal ) md of fig .
1c has been magnified around the point @xmath59 for different cut - offs @xmath5 ( white triangles and squares correspond to @xmath129 and @xmath135 respectively , and black triangles and squares to @xmath136 and @xmath137 ) .
as @xmath5 increases , the point @xmath59 is shown to become an accumulation point of zeroes _ from the outside_. + fig . 3 : maps of zeroes of the nb ( pascal ) md are shown for different values of the parameter @xmath18 and the same fixed cut - off @xmath136 .
white triangles correspond to @xmath138 of fig . 1a , white squares to @xmath139 of fig .
1b and black triangles to @xmath140 of fig.1c .
the full line describes the circle of unit radius @xmath108 .
the figure illustrates that as discussed in the text the point @xmath59 is an accumulation point of zeroes _ from the inside _
( for @xmath109 ) and _ from the outside _ ( for @xmath110 . )
|
theorems on zeroes of the truncated generating function in the complex plane are reviewed . when examined in the framework of a statistical model of high energy collisions based on the negative binomial ( pascal ) multiplicity distribution ,
these results lead to maps of zeroes of the grand canonical partition function which allow to interpret in a novel way different classes of events in pp collisions at lhc c.m.energies .
|
binary neutron star systems are rare , and even among them , the double pulsar system j0737@xmath03039 is extraordinary , since both the neutron stars are detected as radio pulsars .
the system consists of the recycled 22.7 ms pulsar `` a '' @xcite and the young 2.8 s pulsar `` b '' @xcite , in a 2.454 hr eccentric ( @xmath2 ) binary orbit which happens to be nearly edge - on to us . as well as being a test bed for general relativity and theories of gravity ( e.g. * ? ? ?
* ) , the double pulsar is rich in observational phenomena , including a short eclipse of a by the magnetosphere of b and orbital modulation of the radio flux of b due to the influence of a @xcite .
the individual pulses from b show drifting features due to the impact of the low - frequency electromagnetic wave in the relativistic wind from a @xcite , while the eclipse of a is modulated at half the rotational period of b @xcite .
clearly , the two neutron stars have both gravitational and electromagnetic interactions with each other , and the double pulsar system should provide a unique laboratory to investigate the interactions between the magnetospheres and relativistic winds of the two pulsars . in this context , the detection of x - ray emission from the j0737@xmath03039 system @xcite is particularly exciting .
energetic pulsars generate several forms of x - ray emission : quasi - blackbody emission from the cooling neutron star surface and/or from heated polar caps ; pulsed non - thermal emission from the pulsar magnetosphere ; and at larger distances from the pulsar , synchrotron emission from a pulsar wind nebula ( pwn ) powered by the relativistic particle outflow .
all of these processes may be taking place in the j0737@xmath03039 system ( see , e.g. , * ? ? ?
specifically , the x - rays could be pulsed magnetospheric or thermal emission from pulsar a ( as seen for several other recycled pulsars ; see * ? ? ?
* ) , could originate in the colliding winds of a and b @xcite , or could be produced by the shock generated when one or both of the pulsar winds interacts with the interstellar medium @xcite .
the electrodynamics of pulsar winds have been studied in considerable detail through the extended pwne typically seen around young and/or high - velocity pulsars @xcite . in systems such as the crab nebula , the pwn is an expanding synchrotron bubble centered on the pulsar .
such nebulae act as calorimeters , revealing the geometry and energetics of the pressure - confined outflow and its termination shock .
however , the termination shocks seen in such pwne are typically at distances @xmath3 from their pulsars ( where the light cylinder radius of a pulsar rotating at a frequency @xmath4 is @xmath5 ) . in contrast , the two neutron stars in the double pulsar system are separated by @xmath6 and only @xmath7 ; a termination shock between them can thus probe the properties of a pulsar s relativistic wind at smaller separations from the central engine than ever studied before .
additionally , detection of an orbital phase dependence in the x - ray emission might be expected ( e.g. , * ? ? ?
such variability could constrain the geometry of the emission site , thus providing new insights into the wind physics close to the pulsar . here
we report on _ chandra _ observations of the double pulsar which have high enough time resolution to test for pulsations from either pulsar and for orbital variability , the latter of which might be expected in the bow shock or colliding winds interpretation .
forming histograms of count rates as a function of phase , we test for modulated x - ray emission at the periods of pulsar a and pulsar b , as well as for orbital modulation .
the j0737@xmath03039 system was observed with the _ chandra x - ray observatory _ using the high resolution camera ( hrc - s ) in `` timing mode '' , which provides the highest available time resolution , with events corrected for the instrumental wiring error and time - tagged to 16 @xmath8s accuracy .
the observations spanned 10.5 binary orbits but were split into two segments for spacecraft operational reasons .
the first segment of 55 ks began on 2006 february 28 , while the second segment began @xmath9 ks after the end of the first , and spanned 38 ks .
the pulsar system was unambiguously detected as a point source in both segments , at a position @xmath10 @xmath11 ( j2000 ) , consistent with positions previously determined at x - ray and radio wavelengths @xcite at the @xmath1205 pointing accuracy of chandra .
x - ray photons were extracted from a 1 radius circle at the detected position of the j0737@xmath03039 system , and the times of arrival for the photons were corrected to the solar system barycenter using the jpl planetary ephemeris de405 .
of the 411 photons extracted , we estimate that @xmath13 counts were contributed by the x - ray background .
of course , we can not identify which of the extracted photons came from the background , and nor can we assign photons to the individual pulsars . instead , we use tempo and timing solutions from @xcite to calculate the binary orbital phase and the rotational phases of both pulsars a and b at which each photon was emitted . in keeping with our request that the observation be split ( if necessary ) into integer orbit blocks ,
the orbital phase of the extracted photons ranged from 0.840 to 6.854 in the first segment , and from 0.497 to 4.531 in the second .
thus only @xmath125% of the orbit was sampled 11 times , while the rest was uniformly sampled 10 times . in the analysis reported below
, we have ignored the minor oversampling , but we have verified that discarding 3 detected photons to force uniformity in orbital coverage does not affect our results .
forming a histogram of count rate as a function of the rotational phase of a , we detect x - ray pulses from pulsar a , as illustrated in .
the uncertainties on each bin ( here and elsewhere in this work ) are 1@xmath14 ( 68% ) confidence intervals estimated according to @xcite .
the pulsations are double - peaked and deeply modulated , with a pulsed fraction @xmath15 of @xmath16 . to estimate the significance of the detection
, we calculate the pearson @xmath17 statistic for the pulse profile with 16 bins ( degrees of freedom @xmath18=15 ) , and find @xmath19 , corresponding to a probability of only @xmath20 ( @xmath21 ) that the profile is drawn from a uniform distribution .
a visual comparison of the radio pulse profile of psr j0737@xmath03039a @xcite with the x - ray profile shows a distinct resemblance ( ) .
@xcite model the radio pulse as two cuts through a wide cone of emission centered on a single magnetic pole of a , which has its spin and magnetic axes nearly aligned ( @xmath22 ) .
although a wide range of misalignment is currently permitted by radio observations @xcite , both peaks of the observed pulse appear to come from one magnetic pole , implying a very wide fan beam in some geometries .
the x - ray pulse profile also shows two peaks , whose locations fall within the range in pulse phase delimited by the radio peaks when the pulses are phase - aligned.s level between hrc - s and radio observations of the recycled pulsar psr b1821@xmath024 . since we have to predict and account for both orbital and rotational phase , our timing errors are somewhat larger , but insignificant compared to the bin width of @xmath121.4 ms . ]
this suggests that the x - ray emission is from a narrower cone than the radio beam .
specifically , the peaks in the x - ray profile ( ) are located at pulse phases @xmath23 and @xmath24 , as estimated by binning the observed x - ray photons at various resolutions , and the peak - to - peak separation is @xmath25 ( @xmath26 ) , while the peaks in the 1.4 ghz radio profile @xcite are at @xmath27 and @xmath28 , separated by @xmath29 , and the outer rims of the radio emission profile are at @xmath30 , @xmath31 .
the x - ray emission also shows a significant `` bridge '' between the two peaks , implying that the cone of x - ray emission is ( partially ) center - filled in this model , unlike the broader , hollow radio emission cone .
the detected pulses are quite unlike the typical x - ray emission observed from other recycled pulsars with comparable spin parameters ( e.g. , psr j0437@xmath04715 , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) , which show broad , roughly sinusoidal pulsations with low pulsed fractions and thermal spectra . instead
, the pulsations from pulsar a resemble non - thermal pulses seen only from the most energetic recycled pulsars ( e.g. , psr b1821@xmath024 , * ? ? ?
* ) , even though a is slower rotating and has a lower spindown energy loss rate ( @xmath32 erg s@xmath33 ) . in this context
, we note that the formation scenario for double neutron star binary systems @xcite can result in a shorter accretion episode and thus a higher surface magnetic field strength compared to other recycled pulsars .
pulsar a has an inferred dipole magnetic field @xmath34 g , comparable to psr b1821@xmath024 and the binary pulsar b1534 + 12 , but significantly higher than other recycled pulsars that show predominantly thermal x - ray emission .
both pulsars b1821@xmath024 and j0737@xmath03039a also lie above the death line for curvature radiation estimated by @xcite , suggesting that the processes that power non - thermal magnetospheric emission in psr b1821@xmath024 may also operate for pulsar a , although the two differ substantially in period and @xmath35 . the absence of any useful energy resolution in _
hrc data precludes spectral fits to the data , but previous _ chandra _ acis observations can be well - modeled by a power law with a photon index @xmath36 @xcite , and _ xmm _ data is well - fit by a power law with a photon index @xmath37 @xcite .
joint fits to the _ chandra _ and _ xmm _ data @xcite allow for both power law ( @xmath38 ) and thermal black body ( @xmath39 kev ) interpretations . similar fit parameters ( @xmath40 or @xmath41 kev )
were found by @xcite as well .
additionally , @xcite show that a two - component fit with a fixed power law index @xmath42 and a black body component ( @xmath43 kev ) is consistent with the _ chandra _ acis and _ xmm _ data , although two components are not statistically required .
the x - ray spectrum , in combination with our detection of sharp , double - peaked x - ray pulses , is thus consistent with a purely magnetospheric origin for the x - ray emission , but it is also possible that the observed x - ray pulsations consist of both non - thermal magnetospheric emission and broad sinusoidal thermal pulsations from the hot polar cap .
the pulse profile of a shows a floor of x - ray emission ( ) , corresponding to a count rate of @xmath44 cts ks@xmath33 at every phase .
an image of the off - pulse counts reveals no extended nebular structure , and their distribution is consistent with the on - pulse photons .
other recycled pulsars also show emission at all pulse phases , whether their pulsations are broad and thermal ( e.g. , psr j0437@xmath04715 , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) or narrower and primarily non - thermal ( e.g. , psr j0218 + 4232 , @xcite ; psr b1821@xmath024 , @xcite ) .
such unpulsed emission is usually ascribed to thermal x - rays emitted from the neutron star surface .
assuming that the entire x - ray flux of the double pulsar system arises only from the combined thermal and non - thermal emission from psr j0737@xmath03039a , we find that the maximum amplitude sinusoid @xmath45 that is consistent with the observed profile at @xmath46 could account for as much as @xmath47 of the observed x - ray counts , although the actual fraction is likely to be far lower .
rotational phase - resolved spectroscopy with substantially more x - ray counts will be required to verify or rule out such a two - component model . in order to investigate possible orbital variations in the x - ray pulse profile of a ,
9-bin pulse profiles were constructed for each quadrant of the orbit .
each of the four profiles was then compared to the pulse profile constructed by averaging the other three quadrants .
the resulting @xmath48 values range between 0.8 and 1.3 ( with @xmath49 degrees of freedom ) , consistent with no variations . while we lack the s / n to definitively rule out any differences between the x - ray pulse profiles , no orbital variations are detected in the radio pulse profiles of a either ( e.g. * ? ? ?
we note in passing that our estimate of the pulsed fraction @xmath50 is only marginally consistent with the upper limit of 60% on the pulsed fraction ( assuming sinusoidal pulses ) inferred by @xcite from _ xmm - newton _ observations .
since the detected pulse is non - sinusoidal , a direct comparison is not possible , but @xmath47 of our detected photons are above the estimated minimum count rate baseline , and @xmath51 are @xmath52 above the baseline level .
the _ xmm _
pn observations of @xcite , which were in continuous clocking mode , were totally dominated by the background due to the one - dimensional readout , while the _ xmm _ mos chips lack the time resolution to detect pulses from a , leading to a limit which is less robust compared to the _ chandra _ hrc detection presented here .
we repeated the analysis described in 3 for psr j0737@xmath03039b .
the results are shown in .
no x - ray pulsations are detected , either by folding the full span of data , or by selecting counts which are in the lowest emission bins of the pulse profile of a , @xmath53 .
for the folded profile from the entire data span , we calculate @xmath54 , corresponding to a 17% probability that the data are drawn from a uniform distribution . as described by @xcite , epoch folding is not as sensitive to broad , smooth pulses as the family of rayleigh statistics @xmath55 , which also avoid the need to bin data . therefore , we also calculated the @xmath56 statistic ( @xmath57 , for @xmath58 ; * ? ? ?
* ) , which is well suited to searching for an unknown modulation shape .
we find @xmath59 , at @xmath60 , corresponding to a null hypothesis probability ( i.e. , the probability that we are sampling a uniform distribution ) close to unity .
pulsar b shows significant enhancements in radio emission at some parts of its orbit @xcite , but folding x - ray photons selected from those orbital phase ranges does not show any evidence for pulsations either .
the non - detection is unsurprising , since pulsar b has rotational parameters and a spindown energy loss rate ( @xmath61 erg s@xmath33 ) similar to other `` ordinary '' pulsars ( ages @xmath62 yr ) , which are not known for their x - ray emission .
the spindown luminosity of pulsar b is only @xmath63 , and so pulsar a is expected to dominate any x - ray emission from the system .
systems such as the crab nebula @xcite and the nebula around psr b1509@xmath058 @xcite provide the best current constraints on the behavior of pulsar winds at large distances ( @xmath64 ) from the neutron star , but few constraints exist for the close - in behavior .
the magnetization parameter @xmath14 , the ratio of poynting flux to the kinetic energy flux in the wind , is a key descriptor of such systems .
optical , near - infrared , and x - ray images at sub - arcsecond resolution reveal that the shock has an axisymmetric structure of equatorial arcs ( wisps ) and polar jets ( knots ) that vary on short time - scales @xcite , and that the wind transforms from a poynting - dominated outflow ( @xmath65 ) near the pulsar to a kinetic - energy - dominated outflow ( @xmath66 ) at the termination shock @xcite .
recent work has begun to elucidate the collimation mechanism that produces the axisymmetric structure ( e.g. , * ? ? ?
* ) , while the conversion of poynting flux to mechanical energy remains poorly understood .
the wind interaction of a neutron star with a stellar binary companion allows constraints on the wind behavior at @xmath67 , and such interaction can produce radio and high energy emission signatures . for example , the be star
pulsar binary b1259@xmath063 produces unpulsed radio emission @xcite as well as unpulsed high energy emission ( e.g. , * ? ? ?
* ) , which arise from the shock formed between the stellar outflow and the pulsar wind @xcite .
the interaction of the pulsar b1957 + 20 with its white dwarf binary companion is expected to produce orbital modulation in the x - ray emission ( see , e.g. , * ? ? ?
* ; * ? ? ?
* ) , although the observational evidence for such modulation @xcite is not significant .
as opposed to the interaction between a neutron star relativistic wind and the particle wind of a stellar companion , the double pulsar presents a situation where the relativistic wind interacts with the magnetosphere of another neutron star . additionally , the system separation is @xmath6 and only @xmath68 .
the detection of orbital modulation in the system would thus be of particular interest , since it probes the behavior of the pulsar wind in a high-@xmath14 regime .
we note that @xcite find possible orbital phase dependence for the double neutron star binary b1534 + 12 , but not for the j0737@xmath03039 system .
their result is based on earlier _ xmm _ and _ chandra _ acis data that lacked the time resolution to detect pulsations from pulsar a. given our detection of deeply modulated pulsed emission from psr j0737@xmath03039a , we attempted to detect orbital modulation in the x - ray emission by folding x - ray photons from the off - pulse phase of a , @xmath69 , corresponding to the three bins with lowest photon counts in . the 25 counts thus selected from the observation ( corresponding to a reduced effective exposure of 16.7 ks ) were folded as a function of orbital phase , and the results are shown in the top panel of .
we find an apparent enhancement at a phase @xmath120.69 . at the epoch of observation ,
that phase bin includes an orbital longitude @xmath70 , corresponding to a s crossing of the ascending node of the orbit .
however , there is no obvious physical mechanism that could produce such an enhancement , and the binned distribution has @xmath71 , corresponding to a chance probability of 11.5% . as in 4 , we calculate the @xmath56 statistic with the un - binned orbital phase values .
we find @xmath72 at the fifth harmonic @xmath73 , which allows the null hypothesis that we have sampled uniformly distributed data at @xmath74 , a probability that is small but not insignificant .
we also check for the enhancement by folding all the x - ray photons , and by selecting and folding photons from the bridge of emission between the two peaks of a s pulse profile ( @xmath75 ) , where the contribution of the pulsar itself is reduced .
the results are shown in the middle and bottom panels of , and in each case , we again calculate the @xmath56 statistic .
@xmath76 at @xmath60 when including all the extracted photons , corresponding to a null hypothesis probability of 52% .
for the photons chosen from between the two peaks of a s pulse , @xmath77 at @xmath60 , which allows the null hypothesis at a probability close to unity .
together , these results lend weight to the conclusion that the apparent orbital modulation seen above ( with a chance probability of 5% ) is , in fact , not real .
we have also confirmed the absence of significant modulation by binning as a function of orbital longitude rather than phase , with very similar results . as outlined in
3 , it is more likely that the unpulsed x - rays have their origin in thermal emission from the surface of pulsar a. we thus confirm the negative result found by @xcite . from the drifting sub - pulses detected in b s radio emission @xcite ,
it is apparent that the low - frequency electromagnetic wave in the relativistic wind from a influences the emission of pulsar b , and several models have been proposed where the formation of a shock between the two pulsars should produce orbital modulation in their emission ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) . however , only a small fraction of the wind power emitted by a ( and half of the power emitted by b ) is intercepted by the shock between the two pulsars , reducing proportionately the maximum x - ray flux that the shock emits . for example , if we assume that the wind energy is radiated isotropically from a , and that it is intercepted by a sphere centered on b with radius @xmath78 , then the power intercepted by the shock , @xmath79 .
if , instead , a s wind is intercepted at the surface where pressure balance is achieved between the wind from a and the magnetosphere of b , at @xmath80 lt - s from b @xcite , then we have @xmath81 . finally , if the shock roughly coincides with the region centered on b that eclipses the radio pulses from a , we have @xmath82 ( although the processes that contribute to radio eclipses are likely to be quite different from those that cause x - ray emission ) .
of course , the wind radiated from a is unlikely to be isotropic , especially if the magnetic and rotational axes are nearly - aligned @xcite , and the shock geometry is not described simply by intersecting spheres centered on a and b. nevertheless , the conservative geometric estimates above demonstrate that the x - ray power output from the shock is @xmath83 , possibly modulated at the orbital period .
interestingly , spectral fits to the _ chandra _ and _ xmm _ data imply an x - ray efficiency @xmath84 @xcite , where @xmath85 is the x - ray luminosity in the 0.510 kev range .
thus , if the entire @xmath86 were converted to x - ray emission , at least two of our proposed scenarios above would have resulted in a higher x - ray efficiency for the j0737@xmath03039 system than actually observed .
since we detect x - ray pulses from a which account for a significant proportion ( and arguably @xmath12100% ) of the observed x - ray emission , all of @xmath86 evidently does not appear as x - ray emission .
( we note that for @xmath32 erg s@xmath33 , the relations derived by @xcite for x - ray luminosity in the 210 kev range predict a maximum x - ray efficiency @xmath87 , consistent with observations . )
the wind interaction in the double pulsar system is fundamentally different energetically from wind confinement in a crab - like pulsar wind nebula , since the termination shock of the wind is much closer to pulsar a ( @xmath6 ) than in crab - like nebulae ( @xmath88 ) .
all modern wind models , whether for a steady - state , force - free , magnetohydrodynamic outflow @xcite or a wave - like , striped outflow @xcite , predict values of the magnetization parameter @xmath89 ( probably @xmath90 ) at these distances , unlike termination shocks in pulsar wind nebulae , where @xmath91 . for a high-@xmath14 shock ,
@xcite estimate an upper limit on the power fed into the accelerated electrons ( and hence on the x - ray luminosity of the shock ) of @xmath92 . in summary , as a result of the high expected value of @xmath14 and the small solid angle over which the wind from a is intercepted by b , the shock produced at the interaction region is unlikely to show significant x - ray emission . similar arguments apply to the absence of unpulsed radio emission from the system @xcite as well .
with 89 ks of _ chandra _ hrc observations , we have detected deeply modulated emission from psr j0737@xmath03039a .
the off - pulse emission reveals no extended structure .
no pulsations were detected from psr j0737@xmath03039b , and no orbital modulation was detected either .
although we can not absolutely rule out orbital modulation or emission from other mechanisms such as bow shocks , we have shown that the entire x - ray emission from the j0737@xmath03039 system can be explained as arising from pulsar a alone , either as non - thermal magnetospheric emission , or as a combination of magnetospheric and thermal emission .
pulse phase - resolved spectroscopy will allow discrimination between these two scenarios . like the dog that did not bark in the night
, the absence of significant and detectable orbital modulation in the x - ray emission from the j0737@xmath03039 system is noteworthy .
the wind from pulsar a impinges on and compresses the magnetosphere of b , leading to deep orbital modulation in the detected radio pulsations from b @xcite , and the impact of the low - frequency electromagnetic wave in the relativistic wind from a is also seen in the drifting sub - pulses of emission observed from b @xcite . given the strong influence of a on the radio emission from b , it may seem natural to ascribe the x - ray emission from the psr j0737@xmath03039 system to a particle shock formed at the wind - magnetosphere interaction site . as we show here ,
such an interpretation is neither favored by theory , nor required by the x - ray observations .
our observations reveal no significant orbitally modulated shock emission , consistent with models for relativistic winds that require a poynting - dominated wind close to the pulsar , and with only a small fraction of the energy outflow from a interacting with the termination shock .
we thank michael kramer for making current timing solutions available for psr j0737@xmath03039a and b , dick manchester for providing us radio pulse profiles , scott ransom for creating presto , and for his guidance in using it , and zaven arzoumanian for helpful discussions about tempo .
we also thank george pavlov , slavko bogdanov , nanda rea , and the anonymous referee for their helpful comments on the manuscript .
sc acknowledges support from the university of sydney postdoctoral fellowship program .
support for this work was provided by nasa through chandra award go5 - 6046x to the harvard college observatory . , m. , stairs , i. h. , manchester , r. n. , mclaughlin , m. a. , lyne , a. g. , ferdman , r. d. , burgay , m. , lorimer , d. r. , possenti , a. , damico , n. , sarkissian , j. m. , hobbs , g. b. , reynolds , j. e. , freire , p. c. c. , & camilo , f. 2006 , science , 314 , 97 , r. n. , kramer , m. , possenti , a. , lyne , a. g. , burgay , m. , stairs , i. h. , hotan , a. w. , mclaughlin , m. a. , lorimer , d. r. , hobbs , g. b. , sarkissian , j. m. , damico , n. , camilo , f. , joshi , b. c. , & freire , p. c. c. 2005 , , 621 , l49 , m. a. , lyne , a. g. , lorimer , d. r. , possenti , a. , manchester , r. n. , camilo , f. , stairs , i. h. , kramer , m. , burgay , m. , damico , n. , freire , p. c. c. , joshi , b. c. , & bhat , n. d. r. 2004 , , 616 , l131
|
the double pulsar system j0737@xmath03039 is not only a test bed for general relativity and theories of gravity , but also provides a unique laboratory for probing the relativistic winds of neutron stars .
recent x - ray observations have revealed a point source at the position of the j0737@xmath03039 system , but have failed to detect pulsations or orbital modulation .
here we report on _ chandra x - ray observatory _ high resolution camera observations of the double pulsar .
we detect deeply modulated , double - peaked x - ray pulses at the period of psr j0737@xmath03039a , similar in appearance to the observed radio pulses .
the pulsed fraction is @xmath1 .
purely non - thermal emission from pulsar a plausibly accounts for our observations .
however , the x - ray pulse morphology of a , in combination with previously reported spectral properties of the x - ray emission , allows the existence of both non - thermal magnetospheric emission and a broad sinusoidal thermal emission component from the neutron star surface .
no pulsations are detected from pulsar b , and there is no evidence for orbital modulation or extended nebular structure .
the absence of orbital modulation is consistent with theoretical expectations of a poynting - dominated relativistic wind at the termination shock between the magnetosphere of b and the wind from a , and with the small fraction of the energy outflow from a intercepted by the termination shock .
|
neutron star ( ns ) astronomy has been recently invigorated by the identification of a new class of magnetically dominated emitters .
known as anomalous x - ray pulsars ( axps ) and soft gamma - ray repeaters ( sgrs ) , these objects are apparently young , isolated neutron stars ( nss ) , whose properties differ markedly from those of the crab pulsar , previously considered prototypical of the young nss ( for a review see @xcite ) .
all axps and at least 3 of the 4 sgrs are identified as relatively slow ( @xmath5 s ) pulsars .
many are located at the centers of recognized , young supernova remnants ( snrs ) , directly associating them with their supernova explosions .
these objects emit predominantly at x - ray energies and are distinguished by their characteristic spectral signature .
this radiation can not be accounted for by rotational energy losses alone , as for the radio pulsars , but is most likely powered by the decay of an enormous magnetic field characterized by a dipole with @xmath6 g , at the pole .
collectively , these isolated nss are understood within the context of the magnetar theory ( @xcite ) .
a unique pulsar has been discovered whose remarkable properties offer great promise for deciphering the emission mechanism(s ) of magnetars .
is a 5.54 s x - ray pulsar whose measured and derived physical parameters are fully characteristic of an axp , but expresses behavior not previously associated with any such object ( * ? ? ?
* hereafter paper i ) . is a transient axp ( taxp ) it was discovered during a bright impulsive outburst @xcite that is still fading steadily .
even more surprising is the discovery of highly variable radio emission , providing the first confirmed example of radio flux from an axp @xcite , and the subsequent detection of radio pulsations at the x - ray period @xcite .
the outburst that resulted in the detection of the transient axp occurred sometime between 2002 november and 2003 january @xcite .
since then , over the course of a year , regular scans of the region with rxte recorded an exponential flux decay with a time - constant of @xmath7 days from a maximum of @xmath8 erg @xmath9 s@xmath10 . in comparison ,
the previous average quiescent flux , with its softer spectrum , gives @xmath11 erg @xmath9 s@xmath10 ( paper i ) .
this contrast is unprecedented for an axp .
further rxteobservations yielded sgr - like episodes of bursts @xcite , similar to those seen from 1e 2259 + 586 @xcite .
a search for an optical / ir counterpart detected a fading ir source within the chandra error circle , similar to ones associated with other axps , confirming its identification with @xcite . in this paper
we present the latest results from taxp including new xmm - newton observations extending to 2006 march .
these observations allow us to characterize the spectral evolution of a taxp in outburst as it returns to quiescence .
we show that the spectrum has now begun a marked transition back to the nominal quiescent state , with a distinct temperature and flux evolution .
this long - term evolution provides the unique possibility of decomposing its fading spectral components in a manner unavailable for other magnetars ( ( * ? ? ?
* hereafter paper ii ) ) . in the following , we used a revised distant to the ns , @xmath12 , quoted in units of @xmath13 kpc and based on the radio pulsar dm measurement of @xcite .
herein we present a total of seven xmm - newton observations of of which four have been previously described in papers i , ii , and ( * ? ? ?
* hereafter paper iii ) .
the three new data sets were obtained using similar observing modes as for previous observations and reduced and analyzed in an identical manner .
a log of these observations is recorded in table 1 . a full report on the reduction and analysis of these data sets
will be presented in gotthelf ( in prep . ) .
figure [ fig : flux ] presents an up - to - date light curve of taxp derived by adding the xmm - newton flux measurements to the rxte results of @xcite . for comparison ,
the xmm - newton fluxes were extracted assuming a simple power - law spectral model fitted in the @xmath14 kev energy band .
the combined data points are well fitted by an exponential decay model with overall time - constant of @xmath15 d , somewhat shorter but consistent with the initial rxte trend , given the extended monitoring interval . to allow for a systematic offset in the rxte flux measurements kev conversion factor of 2.27 counts s@xmath10 per cpu @xmath16 erg s@xmath10 @xmath9 . ]
, these data were rescaled by a factor of 1.42 to match the xmm - newton results .
the implied flux range for the initial outburst is @xmath17 erg s@xmath10 @xmath9 , over the interval during which the target was inaccessible to rxte .
clearly the flux in this band has all but returned to its pre - outburst value . kev energy band , for direct comparison with the rxte results .
the _ solid line _ is a best combined fit to an exponential decay model ( see text for parameters ) .
, scaledwidth=45.0% ] [ tab : table ] [ cols="<,^,^,^,^,^,^,^ " , ] + note uncertainties in spectral parameters are 90% confidence for two interesting parameters .
+ @xmath18epic - pn exposure / livetime in units of ks . + @xmath19interstellar hydrogen absorbing column density in units of @xmath9 .
+ @xmath20absorbed 0.510 kev flux in units of erg @xmath9 s@xmath10 .
+ @xmath21unabsorbed bolometric luminosity in units of erg s@xmath10 assuming a distance of @xmath22 kpc .
+ spectra of axps are nominally modeled assuming a two - component power - law plus blackbody model .
the spectral and temporal trends for allow us to strongly reject this model ( see paper ii for a comprehensive discussion ) .
instead , we find that a two - temperature blackbody model gives equally acceptable spectral fits but is better motivated physically ( see 5 ) . a summary of spectral results is presented in table 1 and the spectra , fitted with the double blackbody model , are displayed in figure [ fig : spectra ] ( the 2003 oct observation is excluded here for clarity ) . although the hot blackbody component ( bb2 ; fig .
[ fig : luminosity ] ) initially dominated the emission at low energies , it fades more rapidly than the cooler `` warm '' emission component ( bb1 ; fig .
[ fig : luminosity ] ) .
evidently the spectral fits after 2004 march deviate significantly at the lower end of the xmm - newton energy band , below 0.7 kev .
it is not yet clear if this is an instrumental artifact , however , taken at face value there seems to be an absorption feature at @xmath23 ev with a gaussian width of @xmath24 ev .
the mean equivalent width of this feature is @xmath25 ev , but its strength is somewhat time dependent .
perhaps as the hotter component contributes less , this feature becomes more evident , suggesting an association exclusively with the cooler component .
further research is needed to quantify this feature and consider its reality , or physical implications .
the spectrum of can be thought of as the combined flux from two concentric emitters , `` hot spots '' , whose temperature and size are evolving at different rates , effectively changing the overall shape of the spectrum with time ( paper ii ) . with a set of spectral measurements spanning two and a half years
, a clear trend has emerged .
the bolometric luminosities of the two components are shown in figure [ fig : luminosity ] .
this reveals a cooler component decreasing exponentially in time with @xmath0 d , while the hotter temperature flux declines with @xmath1 d. as might be expected , the shorter time constant is very close to the one described by the long - term flux above @xmath26 kev ( figure [ fig : flux ] ) , where the hotter blackbody component dominates .
however , the latest two data points show that this component has now fallen faster than the original exponential . although the _ bolometric luminosities _ of the two components are quite different at the latest epoch , their measured _ fluxes _ are now nearly equal in the @xmath27 kev range .
we also note that , while it is possible to fit an alternative power - law decay model to the hot blackbody flux , such a model would require a decay index that steepens with time .
based on the decay rates and approximate outburst time we estimate an initial bolometric luminosity for the two spectral components of @xmath28 erg s@xmath10 and @xmath29 erg s@xmath10 , respectively .
the total luminosity at the peak of the outburst is comparable to that of a persistent axp , suggesting that this taxp is anomalously faint .
we can now compute the implied fluences of @xmath30 erg and @xmath31 erg , for the two components , respectively .
the energy of the outburst event is many orders - of - magnitude lower than that for a typical sgr flare .
d and @xmath32 d for the warm and hot temperature components , respectively .
the recent data for the hot component now deviates from this model .
the fitted quantities have been extrapolated to the ( @xmath33 ) quiescent range measured in paper i ( _ cross - hatched area _ ) .
[ fig : luminosity],scaledwidth=45.0% ] the temperatures and inferred blackbody emitting areas have also evolved measurably with time .
figure [ fig : params ] displays the time histories of these parameters for each spectral component , obtained from the model fits . based on the three latest data points , the temperatures now show a definite cooling trend for each spectral component , as suggested by the broken line in figure [ fig : params ] . prior to mid 2004 , the temperatures likely remained nearly constant , after which they both fell at a rate of @xmath34 per year ( @xmath35 kev yr@xmath10 ; @xmath36 kev yr@xmath10 ) .
the size of the effective hot spots ( derived from the blackbody emission areas ) also followed distinct evolutions .
the hot component has been shrinking exponentially since the initial xmm - newton observation , while the warm component has steadily increased in size .
a notable exception is the initial data point which deviates from the trend for both the areas and temperatures .
this may be associated with a glitch or rotation instability , as suggested by the pulse timing results around this epoch ( 3 ) .
the area of the warm component ( @xmath37 ; table 1 ) may be reaching a maximum , corresponding to the whole ns surface .
this would then provide a lower - limit on the ns radius of @xmath38 km , ignoring relativistic effects ( redshift and light bending ) .
in contrast , the area of the hotter component continues to shrink and its contribution to the flux is severely diminished . per year since mid 2004 .
prior to that time the temperature of the hotter component was essentially constant .
this trend for the warm component is somewhat ambiguous .
[ fig : params ] , scaledwidth=45.0% ] 0.15 in per year since mid 2004 .
prior to that time the temperature of the hotter component was essentially constant .
this trend for the warm component is somewhat ambiguous .
[ fig : params ] , scaledwidth=45.0% ] as shown in figure [ fig : luminosity ] , the x - ray luminosity of taxp has nearly returned to its historic quiescent level .
this is also true of the temperature and inferred blackbody area .
the pre - burst rosat spectrum of 1992 march 7 is reasonably well fitted with a single blackbody of @xmath39 kev covering @xmath40 @xmath41 and @xmath42 erg s@xmath10 ( paper i ) .
the latest xmm - newtonspectrum that overlaps the rosat band is again dominated by the cooler blackbody component .
furthermore , the current pulsed fraction is consistent with the quiescent upper - limits .
thus the double blackbody model is sufficient to describe the observed properties , and may not require a third `` quiescent '' component to explain the pre - outburst observations .
xmm - newton data from the next observing window ( 2006 september ) is likely to be consistent with the rosat result in the soft energy band , effectively defining a return to the quiescence state .
the barycentered pulse period of measured at each xmm - newtonobserving epoch is shown in figure [ fig : spindown ] .
these data points were derived following the method outlined in paper i. prior to epoch 2004 , the spin - down rate derived from the rxte and initial xmm - newton measurements were highly erratic , ranging from @xmath43 s s@xmath10 ( @xcite ; paper iii ) .
this temporal behavior is likely associated with the outburst event .
the latest period measurements suggest that the spin - down rate has settled down somewhat .
a linear model fit to the last 5 period measurements yields a period derivative of @xmath44 s s@xmath10 with a @xmath45 for 3 dof , corresponding to a null hypothesis probability of 0.27 .
compared to the radio @xmath46 s s@xmath10 of 17 march - 7 may @xcite , there is evidently still timing noise in the pulsar s spin - down .
the revised x - ray rate implies a nominal characteristic age @xmath47 kyr , surface magnetic field @xmath48 g , and spin - down power @xmath49 erg s@xmath10 , values typical for a magnetar .
, scaledwidth=45.0% ] the energy - dependent modulation of the pulsar has evolved noticeably over time with the declining flux .
figure [ fig : profiles ] displays the pulse profiles at each xmm - newton epoch , folded at the best determined periods .
the data has been broken up into six energy bands , with the corresponding background in each band subtracted .
the pulsed fractions , defined as the pulsed emission divided by the total flux , are indicated on the corner of each panel of the figure .
these profiles are aligned so that phase zero is the same at each epoch , for ease of comparison . in any case , within an observation , the phase zero does not change with energy .
this suggests a geometric interpretation for the modulation , such as from localized emission on the rotating ns star surface .
the pulsed fraction has generally decreased with time , most notably at x - ray energies below @xmath50 kev ; at higher energies the trend is less clear due to the large uncertainties derived for those pulsed fractions .
furthermore , at all epochs , the pulsed fraction clearly increases with energy , an equally interesting result discussed below .
the evolving spectral shape and pulse profiles of taxp provide an important and unique ( so far ) diagnostic of the emission geometry , and ultimately , the emission mechanism(s ) of nss . because the spectrum is well modeled by two blackbody components , and the phase alignment of the pulse profiles are energy independent , it is most natural to consider emission from two concentric regions on the ns surface .
the hot component is associated with a smaller hot - spot , while the blackbody model predicts a larger annulus for the warm component .
the complete collection of observed pulsed fractions is consistent with this model .
the smaller , hotter spot always dominates the spectrum above @xmath26 kev and offers a natural explanation for the higher modulation at these energies . however , at lower energies , this contribution gradually fades relative to cooler emission ( which is less modulated ) , contributing less and less to the pulsed emission over time .
we note in passing that a fitted power - law spectral component would have to dominate the soft ( @xmath51 kev ) x - rays at _ all _ epochs , while varying in its contribution to the hard ( @xmath52 kev ) x - rays .
such a power law would drive an evolution of the pulse shapes that is opposite of what is observed .
thus , we find that the detailed evolution of the x - ray emission from further supports the assumption of a purely thermal spectral model , and leaves no evidence of a steep power law as is commonly fitted to individual observations of axps . since the first observation of it was apparent that the broad - band pulse shape is not a simple sine function ; the pulse peak is relatively sharp with a broader inter - pulse trough ( paper i ) .
this effect is more pronounced at higher energy , where the profile is nearly triangular in shape .
this suggests that the pulse profiles can be decomposed into a triangle function and a sinusoidal function , a model that was explored for the earlier data sets and detailed in paper ii & iii .
this model continues to be appropriate for the new data and is used to extract unbiased pulsed fraction measurements for the profiles shown in figure [ fig : profiles ] .
given the success in modeling the pulse profiles with these functions it is natural to associate the two pulsed components uniquely with the two spectral components , i.e. , the triangle shape for the hotter blackbody component and the sinusoidal shape for the cooler one .
however , we find that we can not model all the profiles in a consistent manner with just a simple superposition of these temporal components , based on the implied flux ratio in each energy band @xcite .
instead we conclude that either the two spectral components contain an admixture of the two shapes or there is a third , unmodeled spectral component present .
this is a direction of active research , to consider the correct characterization of the phase dependent spectral contribution .
considering that the absence of radio emission is a defining characteristic of axps , the serendipitous radio detection from came as a great surprise . a chance search of vla data taken about a year after outburst reveals an unresolved point source with a flux of @xmath53 mjy at 1.43 ghz , located at the precise coordinates of the taxp @xcite .
other archival vla data obtained at various frequencies provide upper - limits before and after outburst .
a follow - up vla observation in 2006 march yielded a flux of @xmath54 mjy at 1.43 ghz .
together , these results indicate highly erratic radio emission .
a second surprise came with the detection around that time of pulsed radio emission at the x - ray period @xcite .
this search , at @xmath55 ghz using the parkes radio observatory , revealed a narrow , bright pulse with a high degree of linearly polarization @xcite .
a series of multi - frequency measurements in 2006 april - may confirms the erratic flux behavior and provides an unusual spectral slope of @xmath56 , where @xmath57 ( cf .
@xmath58 for a typical radio pulsar ) . a revised distance to the pulsar of @xmath13 kpc
is inferred from the pulsar s dispersion .
a search for pulsations in archival data acquired in 1997/1998 produced a null result and argues against significant emission prior to the outburst event .
a summary of these radio observations is shown in figure [ fig : radio ] .
mjy at 1.4 ghz . for clarity ,
multi - frequency measurements coinciding with the 1.4 ghz pulsed emission data points are not shown ( from @xcite ) .
[ fig : radio ] , scaledwidth=45.0% ]
although transient axps such as are relatively rare , their short active duty cycle suggests the existence of a larger population of unrecognized young nss . provides a unique window into this population , with prior measurements in the quiescent state and detailed observations during its active , pulsed phase
ultimately , we hope to use this taxp to give insight into the emission mechanisms of magnetars , in general . with the discovery of pulsed radio flux , this may carry over to interpreting emission mechanism of radio pulsars , as well .
below we now summarize our initial attempts to interpret the results in the context of a physical model .
first we outline the basic arguments for the two - temperature blackbody spectral model as a plausible alternative to the nominal power - law plus blackbody model . while both models give reasonable fits to the spectrum , it s important to note that the power - law component is used to model the _ low energy residuals and not any high - energy tail_. without a low energy cut - off of this component , it is not possible to connect it with the observed ir emission without an energy catastrophe . as detailed in paper
ii , invoking synchrotron self - absorption is excluded by radius / magnetic - field inconsistently .
in contrast , the extrapolated spectrum of the warm blackbody component does not exceed the measured ir flux .
furthermore , currently there is no acceptable physical model to anchor a power - law component .
as discussed earlier , the two components of the double blackbody model can be associated with a pair of hot spots on the surface of the star .
the warm temperature component , covering a large fraction of the ns surface , is consistent with the decreased pulsed fraction at lower energies .
the hotter blackbody component is consistent with a smaller emission area and greater modulation at higher energies .
a comprehensive discussion of this issue can be found in paper ii .
determining the emission geometry on the ns is of great interest .
the natural conclusion of applying the double blackbody model to the time resolved spectra and pulse profiles is that concentric hot regions give rise to the observed modulation .
we consider a model of the phase - resolved spectrum taking into account the viewing geometry and offset of the hot spots from the rotation axis ( perna & gotthelf 2006 , in prep . ) .
this is based on the ns emission model given in @xcite that includes general relativistic effects ( redshift and light bending ) .
our goal is to match spectrum and energy dependent pulse shape in order to determine viewing geometry , distance , and ns size .
preliminary work is able to reproduce the pulsed fraction to a reasonable degree but the pulse shape remains elusive . a framework for a theoretical interpretation of the emission from is suggested by the magnetar coronal model of @xcite .
according to this model , the large outburst was generated by a starquake that resulted in a transition to an active coronal state that caused energy to be stored in the twisted b - field of the coronal loop .
particles ( mostly @xmath59 ) are accelerated within this loop and impact the ns surface with gev energy .
this heats up the loop footprint resulting in the observed hot spot emission .
the decay timescale of the coronal loop , and thus the hotter component of the double blackbody luminosity , is of order of a few years .
the decay rate is determined by ohmic dissipation of current in the excited loop .
a cooler component likely arises from deep crustal heating associated with the initial outburst and possibly earlier ones .
the features of this model are in general accordance with the observational properties and inferred model for .
|
we present the latest results from a multi - epoch timing and spectral study of the transient anomalous x - ray pulsar xte j1810197 .
we have acquired seven observations of this pulsar with the newton x - ray multi - mirror mission ( xmm - newton ) over the course of two and a half years , to follow the spectral evolution as the source fades from outburst .
the spectrum is arguably best characterized by a two - temperature blackbody whose luminosities are decreasing exponentially with @xmath0 d and @xmath1 d , respectively .
the temperatures of these components are currently cooling at a rate of 22% per year from a nearly constant value recorded at earlier epochs of @xmath2 kev and @xmath3 kev , respectively .
the new data show that the temperature @xmath4 and luminosity of that component have nearly returned to their historic quiescent levels and that its pulsed fraction , which has steadily decreased with time , is now consistent with the previous lack of detected pulsations in quiescence .
we also summarize the detections of radio emission from xte j1810197 , the first confirmed for any axp .
we consider possible models for the emission geometry and mechanisms of xte j1810197 . 0.5 cm
|
the question of statistical isotropy in the cosmic microwave background ( cmb ) has received much attention within the cosmological community ever since the release of the first - year wilkinson microwave anisotropy probe ( wmap ; bennett et al .
2003a ) in 2003 .
the reasons for this are two - fold . on the one hand ,
the current cosmological concordance model is based on the concept of inflation @xcite , which predicts a statistically homogeneous and isotropic universe .
since inflation has proved highly successful in describing a host of cosmological probes , most importantly the cmb and large - scale power spectra , this undeniably imposes a strong theoretical prior towards isotropy and homogeneity . on the other hand
, many detailed studies of the wmap sky maps , employing higher - order statistics , have revealed strong hints of both violation of statistical isotropy and non - gaussianity .
some early notable examples include unexpected low-@xmath21 correlations @xcite , a peculiar large cold spot in the southern galactic hemisphere @xcite , and a dipolar distribution of large - scale power @xcite .
today , the literature on non - gaussianity and violation of statistical isotropy in the wmap data has grown very large , indeed ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) , and it would be unwise not to consider these issues very seriously . of particular interest to us
is the question of hemispherical distribution of power in the wmap data , first reported by @xcite and later confirmed by , e.g. , @xcite and @xcite .
the most recent works on this topic include those presented by @xcite , who found that the power asymmetry extends to much smaller scales than previously thought , and by @xcite , who quantified the large - scale power asymmetry in the 3-year wmap data using an optimal bayesian framework .
a separate , but possibly physically related , line of work was recently presented by @xcite , who considered the specific model for violation of lorenz invariance in the early universe , proposed by @xcite .
this model involves cmb correlations with a quadrupolar distribution on the sky , and is thus orthogonal to the current dipolar model .
surprisingly , when analyzing the 5-year wmap data , @xcite found supportive evidence for this model at the @xmath22 significance level , when considering angular scales up to @xmath23 .
thus , assuming that the wmap observations are free of unknown systematics , there appears to be increasing evidence for both dipolar and quadrupolar structure in the cmb power distribution , at all angular scales . in this paper
, we repeat the bayesian analysis of @xcite , but double the angular resolution of the data . nevertheless , we are still limited to relatively low angular resolutions , since the method inherently relies on brute - force evaluation of a pixel - based likelihood , and therefore scales as @xmath24 . yet , simply by spending more computer resources we are able to increase the pixel resolution from @xmath25 to 32 and decrease the degradation smoothing scale from @xmath3 to @xmath6 fwhm .
this provides additional support for multipoles between @xmath26 and 80 .
while not sufficient to provide a full and direct comparison with the results of @xcite , it is a significant improvement over the results presented by @xcite .
the bayesian analysis framework used in this paper are very similar to that employed by @xcite .
we therefore only give a brief overview of its main features here , and refer the reader interested in the full details to the original paper and references therein . the starting point for our analysis
is the phenomenological cmb signal model first proposed by @xcite , @xmath27 \mathbf{s}(\hat{n } ) + \mathbf{n}(\hat{n } ) .
\label{eq : data_model}\ ] ] here @xmath28 denotes the observed data in direction @xmath29 , @xmath30 is an intrinsically isotropic and gaussian random field with power spectrum @xmath31 , @xmath32 is an auxiliary modulating field , and @xmath33 denotes instrumental noise . obviously , if @xmath34 , one recovers the standard isotropic model .
however , we are interested in a possible hemispherical asymmetry , and we therefore parametrize the modulation field in terms of a dipole with a free amplitude @xmath0 and a preferred direction @xmath1 , @xmath35 the modulated signal component is thus an anisotropic , but still gaussian , random field , with covariance matrix @xmath36 \mathbf{s}_{\textrm{iso}}(\hat{n } , \hat{m})[1+a\,(\hat{m}\cdot\hat{p})],\ ] ] where @xmath37 we now introduce one new feature compared to the analysis of @xcite , for two reasons .
first , we are interested in studying the behaviour of the modulation field as a function of @xmath21-range , and therefore want a mechanism to restrict the impact of the modulation parameters in harmonic space .
second , we also want to minimize the impact of the arbitrary regularization noise ( see section [ sec : data ] ) on the modulation parameters at high @xmath21 s .
therefore , we split the signal covariance matrix into two parts , one modulated low-@xmath21 part and one isotropic high-@xmath21 part , @xmath38 where only multipoles between @xmath39 are included in @xmath40 , and only multipoles at @xmath41 are included in @xmath42 .
( note that we are not proposing a physical mechanism for generating the modulation field in this paper , but only attempt to characterize its properties .
this split may or may not be physically well - motivated , but it does serve a useful purpose in the present paper as it allows us to study the scale dependence of the modulation field in a controlled manner . ) including instrumental noise and possible foreground contamination , the full data covariance matrix reads @xmath43 the noise and foreground covariance matrices depend on the data processing , and will be described in greater detail in [ sec : data ] . we also have to parametrize the power spectrum for the underlying isotropic component , @xmath31 .
following @xcite , we choose a simple two - parameter model with a free amplitude @xmath44 and tilt @xmath45 for this purpose , @xmath46 here @xmath47 is a pivot multipole and @xmath48 is a fiducial model , in the following chosen to be the best - fit @xmath49cdm power law spectrum of @xcite .
since both the signal and noise are assumed to be gaussian , the log - likelihood now reads @xmath50 up to an irrelevant constant , with @xmath51 . the posterior distribution for our model is given by bayes theorem , @xmath52 here @xmath53 is a prior , and @xmath54 is a normalization factor often called the `` bayesian evidence '' .
note that we now have included an explicit reference to the hypothesis ( or model ) , @xmath55 , in all factors , as we will in the following compare two different hypotheses , namely `` h1 : the universe is isotropic ( @xmath56 ) '' versus `` h2 : the universe is anisotropic ( @xmath57 ) '' .
we adopt uniform priors for all priors in the following .
specifically , we adopt@xmath58 $ ] and @xmath59 $ ] for the power spectrum , and a uniform prior over the sphere for the preferred axis , @xmath1 .
the modulation amplitude prior is chosen uniform over @xmath60 $ ] , where @xmath61 is sufficiently large to fully encompass the non - zero parts of the likelihood .
if more liberal priors are desired , the interested reader can easily calculate the corresponding evidence by subtracting the logarithm of the volume expansion factor from the results quoted in this paper . with these definitions and priors , the posterior distribution ,
@xmath62 is mapped out with a standard mcmc sampler . the bayesian evidence , @xmath63 is computed with the `` nested sampling '' algorithm @xcite . for further details on both procedures ,
we refer the interested reader to @xcite . for easy reference ,
we recall jeffreys interpretational scale for the bayesian evidence @xcite : a value of @xmath64 indicates a result `` not worth more than a bare mentioning '' ; a value of @xmath65 is considered as `` significant '' evidence ; a value of @xmath66 is considered `` strong to very strong '' ; and @xmath67 ranks as `` decisive '' .
llcccccc ilc & kq85 & 64 & @xmath68 & @xmath69 & 3.3 & 7.3 & @xmath70 + @xmath71-band & kq85 & 64 & @xmath72 & @xmath73 & 3.8 & & + @xmath71-band & kq85 & 40 & @xmath68 & @xmath74 & 3.5 & & + @xmath71-band & kq85 & 80 & @xmath75 & @xmath76 & 3.7 & & + @xmath77-band & kq85 & 64 & @xmath78 & @xmath79 & 3.5 & & + ilc & kq85e & 64 & @xmath80 & @xmath81 & 2.6 & & + @xmath82-band & kq85e & 64 & @xmath83 & @xmath84 & 3.9 & & + @xmath71-band & kq85e & 64 & @xmath85 & @xmath86 & 2.7 & & + @xmath77-band & kq85e & 64 & @xmath87 & @xmath88 & 2.5 & & + ilc & kp2 & @xmath89 & @xmath90 & @xmath91 & 2.8 & 6.1 & @xmath92
in this paper we analyze several downgraded versions of the five - year wmap temperature sky maps , namely the template - corrected @xmath82- , @xmath71- and @xmath77-band maps , as well as the `` foreground cleaned '' internal linear combination ( ilc ) map @xcite .
each map is downgraded to low resolution as follows @xcite : first , each map is downgraded to a healpix resolution of @xmath93 , by smoothing to an effective resolution of @xmath6 fwhm and properly taking into account the respective pixel windows .
we then add uniform gaussian noise of @xmath94 rms to each pixel , in order to regularize the pixel - pixel covariance matrix at small angular scales .
the resulting maps have a signal - to - noise ratio of unity at @xmath95 , and are strongly noise dominated at @xmath96 .
two different sky cuts are used in the analyses , both of which are based on the wmap kq85 mask @xcite . in the first case , we directly downgrade the kq85 cut to the appropriate @xmath97 , by excluding any healpix pixel for which more than half of the corresponding sub - pixels are missing .
this mask is simply denoted kq85 . in the second case ,
we smooth the mask image ( consisting of 0 s and 1 s ) with a beam of @xmath6 fwhm , and reject all pixels with a value less than 0.99 .
we call this expanded mask kq85e . the two mask remove 16.3% and 26.9% of the pixels , respectively
. the instrumental signal - to - noise ratio of the wmap data is very high at large angular scales , at about 150 for the v - band at @xmath98 .
the only important noise contribution in the downgraded sky maps is therefore the uniform regularization noise , which is not subject to the additional beam smoothing .
we therefore approximate the noise covariance matrix by @xmath99 .
note that this approximation was explicitly validated by @xcite for the 3-year wmap data , which have higher instrumental noise than the 5-year data .
we also marginalize over a fixed set of `` foreground templates '' , @xmath100 , by adding an additional term to the data covariance matrix on the form @xmath101 , with @xmath102 , for each template .
in addition to one monopole and three dipole templates , we use the @xmath71ilc difference map as a template for both the @xmath71-band and ilc maps , the @xmath82ilc difference for the @xmath82-band , and the @xmath77ilc difference for the @xmath77-band .
however , these foreground templates do not affect the results noticeably in either case , due to the sky cuts used .
the main results from the analysis outlined above are summarized in table [ tab : results ] .
we consider nine different data combinations ( ie . ,
frequency bands , masks and multipole range ) , and show 1 ) the best - fit modulation axis and amplitude , both with 68% confidence regions ; 2 ) the statistical significance of the corresponding amplitude ( ie . ,
@xmath103 ) ; and 3 ) the raw improvement in @xmath104 and bayesian log - evidence for the modulated model over the isotropic model .
the last items are shown for the ilc with the kq85 sky cut only . for reference
, we also quote the ilc result for the kp2 mask @xcite reported by @xcite when analysing the @xmath25 and @xmath3 fwhm data .
the reason for providing the full evidence for only one data set is solely computational .
the total cpu cost for the full set of computations presented here was @xmath105 cpu hours , and the evidence calculation constitutes a significant fraction of this . on the other hand , the evidence is closely related to the significance level @xmath103 , and one can therefore estimate the evidence level for other cases in table [ tab : results ] given the two explicit evidence values and significances .
we have therefore chosen to spend our available cpu time on more mcmc posterior analyses , rather than on more evidence computations .
we consider first the results for the ilc map with the kq85 mask and @xmath106 . in this case ,
the best - fit amplitude is @xmath107 , non - zero at the @xmath9 confidence level .
the best - fit axis points towards galactic coordinates @xmath108 , with a 68% uncertainty of @xmath109 .
these results are consistent with the results presented by @xcite , who found an amplitude of @xmath110 and a best - fit axis of @xmath111 for @xmath112 .
second , we see that these results are only weakly dependent on frequency , as both the v - band and w - band for the same mask and @xmath21-range have amplitudes within @xmath113 of the ilc map , with @xmath114 and @xmath115 and non - zero at @xmath22 and @xmath116 , respectively .
( we have not included the q - band analysis for the kq85 mask , as there were clearly visible foreground residuals outside the mask for this case . )
the corresponding marginal posteriors are shown in figure [ fig : amplitude ] , clearly demonstrating the consistency between data sets .
figure [ fig : axes ] compares the best - fit axes of the three data sets , and also indicates the axes reported by @xcite and @xcite .
next , we also see that the results are not strongly dependent on the choice of mask , as the amplitudes for the extended kq85e mask are consistent with the kq85 results , even though it removes an additional 10% of the sky .
however , we do see , as expected , that the error bars increase somewhat by removing the additional part of the sky , and this reduces the absolute significances somewhat .
finally , the best - fit modulation amplitudes for the v - band data and kq85 mask are @xmath117 for @xmath118 , @xmath119 for @xmath120 and @xmath121 for @xmath122 at @xmath116 , @xmath22 and @xmath123 , respectively .
this is an interesting observation for theoreticians who are interested in constructing a fundamental model for the effect : taken at face value , these amplitudes could indicate a non - scale invariant behaviour of @xmath0 , as also noted by @xcite . on the other hand ,
the statistical significance of this statement is so far quite low , as a single common value @xmath124 is also consistent with all measurements .
better measurements at higher @xmath21 s are required to unambiguously settle this question .
shortly following the release of the first - year wmap data in 2003 , @xcite presented the early evidence for a dipolar distribution of power in the cmb temperature anisotropy sky , considering only the large angular scales of the wmap data .
next , @xcite presented the evidence for a quadrupolar distribution of cmb power , and found that this feature extended over all @xmath21 s under consideration .
finally , @xcite found that the dipolar cmb power distribution is present also at high @xmath21 s .
the evidence for violation of statistical isotropy in the cmb field is currently increasing rapidly , and the significance of these detections are approaching @xmath125 . in this paper , we revisit the high-@xmath21 claims of @xcite , by applying an optimal bayesian framework based on a parametric modulated cmb model to the wmap data at higher multipoles than previously considered with this method , albeit lower than those considered by @xcite . in doing so
, we find results very consistent with those presented by @xcite : the evidence for a dipolar distribution of power in the wmap data increases with @xmath21 .
for example , when considering the v - band data and kq85 sky cut , the statistical significance of the modulated model increases from @xmath126 at @xmath127 , to @xmath22 at @xmath120 , and @xmath123 at @xmath122 . the bayesian evidence now also ranking within the `` strong to very strong '' category on jeffreys scale
. however , it should be noted that the bayesian evidence is by nature strongly prior dependent , and if we had chosen a prior twice as large as the one actually used , the corresponding log - evidence for the ilc map would have fallen from @xmath18 to 1.7 , ranking only as `` substantial '' evidence .
for this reason , it is in many respects easier to attach a firm statistical interpretation to the posterior distribution than the bayesian evidence .
it is interesting to note that the absolute amplitude @xmath0 may show hints of decreasing with @xmath21 .
it is premature to say whether this is due simply to a statistical fluctuation , or whether it might point toward a non - scale invariant underlying physical effect , in which case the amplitude @xmath0 should be replaced with a function @xmath128 . either case is currently allowed by the data . to answer this question , and
further constrain the overall model , better algorithms are required .
the current approach relies on brute - force inversion of an @xmath129 covariance matrix , and therefore scales as @xmath24 or @xmath130 .
however , already the present analysis , performed at @xmath93 , required @xmath13150000 cpu hours , and increasing @xmath97 by an additional factor of two would require @xmath1313 million cpu hours .
more efficient algorithms are clearly needed . to summarize , there is currently substantial evidence for both dipolar ( hansen et al .
2008 and this work ) and quadrupolar power distribution @xcite in the wmap data , and this is seen at all probed scales .
the magnitude of the dipolar mode is considerably stronger than the quadrupolar mode , as a @xmath132 significance level is reached already at @xmath133 for the dipole , while the same significance was obtained at @xmath134 for the quadrupole .
these observations may prove useful for theorists attempting to construct alternative models for these features , either phenomenological or fundamental .
considerable efforts have gone towards this goal already ( e.g. , * ? ? ?
* ; * ? ? ?
? * ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) , but so far no fully convincing model has been established .
clearly , more work is needed on both the theoretical and observational side of this issue .
fortunately , it is now only a few years until planck will open up a whole new window on these issues by producing high - sensitivity maps of the cmb polarization , as well as measuring the temperature fluctuations to arc - minute scales
. we will then be able to measure the properties of the dipole , quadrupole and , possibly , higher - order modes of the modulation field to unprecedented accuracy .
hke acknowledges financial support from the research council of norway .
the computations presented in this paper were carried out on titan , a cluster owned and maintained by the university of oslo and notur .
some of the results in this paper have been derived using the healpix @xcite software and analysis package .
we acknowledge use of the legacy archive for microwave background data analysis ( lambda ) . support for lambda is provided by the nasa office of space science .
|
motivated by the recent results of hansen et al .
( 2008 ) concerning a noticeable hemispherical power asymmetry in the wmap data on small angular scales , we revisit the dipole modulated signal model introduced by gordon et al .
( 2005 ) .
this model assumes that the true cmb signal consists of a gaussian isotropic random field modulated by a dipole , and is characterized by an overall modulation amplitude , @xmath0 , and a preferred direction , @xmath1 .
previous analyses of this model has been restricted to very low resolution ( ie .
, @xmath2 pixels , a smoothing scale of @xmath3 fwhm and @xmath4 ) due to computational cost . in this paper
, we double the angular resolution ( ie . , @xmath5 pixels and @xmath6 fwhm smoothing scale ) , and compute the full corresponding posterior distribution for the 5-year wmap data . the results from our analysis are the following : the best - fit modulation amplitude for @xmath7 and the ilc data with the wmap kq85 sky cut is @xmath8 , non - zero at @xmath9 , and the preferred direction points toward galactic coordinates @xmath10 .
the corresponding results for @xmath11 from earlier analyses was @xmath12 and @xmath13 .
the statistical significance of a non - zero amplitude thus increases from @xmath14 to @xmath9 when increasing @xmath15 from 40 to 64 , and all results are consistent to within @xmath16 .
similarly , the bayesian log - evidence difference with respect to the isotropic model increases from @xmath17 to @xmath18 , ranking as `` strong evidence '' on the jeffreys scale .
the raw best - fit log - likelihood difference increases from @xmath19 to @xmath20 .
similar , and often slightly stronger , results are found for other data combinations .
thus , we find that the evidence for a dipole power distribution in the wmap data increases with @xmath21 in the 5-year wmap data set , in agreement with the reports of hansen et al .
( 2008 ) .
|
the study of open clusters naturally introduces many advances , because they are physically related groups of stars held together by mutual gravitational attraction that were formed at roughly the same time from one large cosmic gas and dust cloud .
their evolutionary stages range from clouds where star formation still takes place at this moment to very old aggregates with turn - off points as late as solar type stars .
therefore , they represent samples of population i stars of constant age and comparable intrinsic chemical composition , best suited to study processes related to stellar evolution and formation , and to fix lines or loci in several most important astrophysical diagrams .
these procedures are statistical methods , for example fitting isochrones to determine the age , reddening , and distance of an open cluster , independent of individual peculiarities of members . as the open clusters drift along their orbits , they are excellent tracers for the global kinematics and dynamical characteristics of the milky way itself . however , some of their members escape the cluster , due to velocity changes in mutual close encounters or tidal forces in the galactic gravitational field .
the escaped stars continue to orbit on their own as field stars .
= 80 mm as a summary it can be concluded that open clusters allow the study of a significant number of astrophysical parameters and models from the global scale of the milky way down to processes in individual stars from the pre - main - sequence to post - agb evolution .
but what about the accuracy of the presently known open - cluster parameters ?
surprisingly enough , there is , to our knowledge , no work within the last ten years which deals with this question in detail .
although there are several new catalogues ( dias et al . 2002 and kharchenko et al .
2005 ) published and a flood of new parameter estimates , mainly because the common use of ccds permits the observation of fainter and more distant clusters , there is no published statistical analysis describing if and how the accuracy has improved . if we look at recent statistical papers dealing with open - cluster parameters ( e.g. chen , hou & wang 2003 , dias & lpine 2005 ) , none of them have any detailed error treatment included .
although apparent `` mean values '' for the age , distance and metallicity are used , the level of accuracy is completely ignored in the final conclusions . in this work
we try to fill this gap by presenting a detailed analysis of the current available open - cluster parameters and the corresponding errors . for this purpose
we have searched the literature up to november 2005 for individual cluster parameter estimates .
the primary goal is to establish a list of `` standard '' open clusters covering a wide range of ages , reddenings and distances from the sun selected on the basis of the smallest errors from the available parameters in the literature .
these clusters should serve as a standard set for future investigations .
we also compare the derived mean values with those of the most recent catalogue by dias et al .
( 2002 ) resulting in an overall satisfactory agreement .
for the determination of mean values for the age , reddening and distance , we have used the following data sources : janes , tilley & lyng , ( 1988 ) , malysheva ( 1997 ) , dambis ( 1999 ) dutra & bica ( 2000 ) , loktin et al .
( 2001 ) , tadross ( 2001 ) , lata et al .
( 2002 , and references therein ) , kharchenko et al .
( 2005 ) and an updated list of individual publications as listed in webda ( http://www.univie.ac.at/webda/recent_data.html ) which includes more than 300 different papers .
the latter was mainly performed within the framework of ads ( http://cdsads.u-strasbg.fr/ ) .
the individual references were further checked for sources from the literature to guarantee a maximum of available estimates accessible to the community .
all selected publications list @xmath0 , log@xmath1 and the distance from the sun .
the only exception is malysheva ( 1997 ) who lists @xmath2 which was converted by a factor of 1.43 to @xmath0 . for the further analysis we have used @xmath3=3.1@xmath0 which is the best mean value for most regions in the milky way ( winkler 1997 ) .
we have carefully checked whether the listed parameters for each open cluster in the individual references are independent of each other and not used by two authors twice .
kharchenko et al .
( 2005 ) , for example , used data by loktin et al .
such duplicated data were not used in the subsequent analysis .
janes et al .
( 1988 ) studied all photometric measurements available at that time to them and introduced weights to derive the cluster parameters .
they summarize the knowledge and results for the given time and supersede the catalogue by lyng ( 1987 ) . the catalogue of dias et al .
( 2002 ) was not used at this stage because it already lists averaged values from the literature .
furthermore we made a comprehensive comparison with the results of this work ( sect .
[ compd ] ) .
the metallicity is also a free parameter when fitting isochrones to photometric data .
chen et al . ( 2003 ) compiled a catalogue of metallicities and analysed this sample .
they derived a radial iron gradient of @xmath40.063(8)dexkpc@xmath5 for the milky way with an intrinsic spread of the overall open cluster metallicities of @xmath60.2dex .
this is in excellent agreement with the scatter of the metallicity found for f and g type stars in the solar neighbourhood ( karatas , bilir & schuster 2005 ) .
the effect of different metallicities to the isochrone fitting procedure is extensively discussed by pinsonneault et al .
the @xmath7 values range from + 0.1 to @xmath40.3mag for [ fe / h][email protected] and constant effective temperature sensitive color indices . taking a difference of 0.3mag for the distance modulus introduced by an incorrect treatment of the metallicity ,
the error for 1kpc is 150pc or 15% .
however , these discrepancies are only valid for effective temperatures cooler than 8000k . for hotter temperatures , the difference in the metallicities result mainly in a shift of the temperature sensitive color indices ( schaerer et al .
lebreton ( 2000 ) presented a comparison of galactic field stars with known metallicities as well as accurate parallax measurements and appropriate isochrones ( see figure 6 therein ) .
there is an apparent mismatch at any level of metallicity .
however , the most significant differences are at large underabundances ( @[email protected] ) which are normally not found in open clusters of the milky way
. there are only a few independent estimates of the metallicity in the literature .
twarog , ashman & anthony - twarog ( 1997 ) , gratton ( 2000 ) and chen et al .
( 2003 ) give a summary of the current knowledge with an error estimate of the individual values . since then very few new estimates have been published .
we have therefore not made a new and redundant error analysis of the metallicities for open clusters .
all individual data were considered as completely equal and no weights were attributed .
no assessment of the corresponding reference for the used methods , isochrones and so on was done . this should guarantee that no unknown bias is introduced in the data set .
we have rejected ngc 2451 , which recent studies have shown that is formed by the superposition of two clusters at different distances , and basel 11 , which was wrongly assigned in the original survey . therefore the former parameters determined for the joined data are no longer significant .
furthermore , pena & peniche ( 1994 ) reported significant inconsistencies for the analysis of ngc 7209 which we also neglected .
we have also not included the given `` intrinsic '' errors of individual references .
the simple reason is that most authors list only an overall error ( if any ) for all measurements which is highly unrealistic and would introduce an unknown bias . for the final list
, we have averaged the data of open clusters for which at least three independent estimates of the age , reddening and distance are available .
it contains 6437 individual estimates for 395 open clusters . as the last step
, the standard deviations were calculated .
= 82 mm = 82 mm
fitting isochrones to photometric data is not straightforward .
the goodness of the fit depends on the following characteristics of the colour - magnitude diagram of an open cluster : * number of members * percentage of contamination by non - members * age , i.e. visibility of the turn - off point , an important source of errors for young open clusters * presence of spectroscopic binaries which shifts the main sequence by a maximum of 0.754mag * presence of a red giant clump which makes an isochrone fitting much more significant * differential reddening * available photometric indices , @xmath9 is , for example , an excellent indicator for the reddening and membership * quality of the data * metallicity in addition , there is also a `` human factor '' which enters the estimate of the best fit to observational data .
andersen & nordstrm ( 1999 ) show how different research groups interpret the same data set quite differently , using the excellent example of ngc 3680 ( figure 3 therein ) .
this is not because one set of parameters is better than the other but the technique of isochrone fitting itself is not unambiguous .
for this reason , very sophisticated computer algorithms were developed to overcome this problem ( jrgensen & lindegren 2005 ) .
the method is rather simple : the locations of the color - magnitude diagrams are divided into boxes and the incidence of members within each box is counted .
those numbers are compared with the same parameterization of standard isochrones .
loktin et al .
( 2001 ) and kharchenko et al .
( 2005 ) , for example , used such a black box to derive cluster parameters .
however , jrgensen & lindegren ( 2005 ) concluded that the systematic errors of computer algorithms are at least on the same level as those introduced by the human factor .
another error source is the choice of appropriate isochrones .
grocholski & sarajedini ( 2003 ) discussed the apparent differences of the five most commonly used sets of isochrones .
they concluded that none of the theoretical isochrones reproduce the observational data in a consistent manner over the magnitude and colour range of the zero age main sequence .
in particular , there are significant zero point and shape differences between the models and the observations .
there are three reasons for these discrepancies : 1 ) the input physics has to be improved ; 2 ) the transformation between theoretical and observational parameters ; and 3 ) more standard open clusters have to be established for testing new models . for our statistical analysis
we have chosen the following parameters : * @xmath10 : the absolute error of the mean distances * @xmath11 : the standard deviation of the mean reddening values * @xmath12 : the absolute error of the mean ages ( in units of myrs and not in the logarithmic scale ) the parameters themselves are , in general , _ not _ independent from each other .
for example , the distance depends on the total absorption .
figure [ means ] shows the errors of the mean for the distance , reddening and age versus the number of available estimates in the literature for the complete sample of 395 open clusters .
the errors of the means for the distance and reddening decrease with more available observations while the mean errors for the age estimate are more or less constant over the whole range .
note that there are also a few parameters which are estimated more than ten times , but these few data are not shown to avoid any poor number statistics .
the distribution of the different error estimates ( fig .
[ means_d ] ) gives an idea of the current accuracy for the open - cluster parameters .
the reddening is best known mainly because the main sequence can be fitted more easily for almost all open clusters . for about 90% of all clusters , the error is below 0.1mag .
the distances are also rather well determined : for about 80% of all aggregates , the absolute error is less than 20% .
as expected , the estimate of the age of open clusters is the most difficult and introduces large uncertainties .
only 11% of the investigated open clusters have errors less than 20% .
even more than 30% exhibit an absolute error larger than 50% with extreme values of more than 200% .
these characteristics are most certainly not due to wrong estimates or typographical errors in the reference but an intrinsic problem of the method itself
. how should one overcome the problem of a proper age estimate for open clusters ?
we have shown that the knowledge of the reddening and determining distances is satisfactory .
if we look at the metallicity , the situation is rather poor .
most authors use solar abundances ( e.g. dutra & bica 2000 , loktin et al .
2001 and kharchenko et al .
2005 ) throughout the isochrone fitting process , this procedure is in most cases justified , but normally photometric data will not allow us do it any better .
we believe that the only way out of the dilemma is to establish `` standard '' open clusters covering a wide range of distances , ages and reddening values which should allow , for example the testing of isochrones .
[ cols="<,<,>,>,>,>,>,^,^ , > , > , > , > " , ]
the current best approximation , many authors use is the list of parameters published by dias et al .
we have used the open - cluster parameters of the october 2005 version of the catalogue available at http://www.astro.iag.usp.br/@xmath13wilton/ for comparison with our results .
the motivation for this comparison is to have an estimate of the statistical error introduced when using `` mean '' catalogue parameters instead of performing an independent analysis of its own .
we are aware that the catalogue of dias et al .
is a kind of compendium of different sources from the literature .
they do not perform any weighting or averaging in the tradition of lyng .
our analysis , on the other hand , is based on averaged values from widely different methods and authors .
it is therefore very interesting to examine if an independent check of this catalogue can confirm its validity .
for the analysis we have calculated the differences of the open - cluster parameters of the catalogue by dias et al .
( 2002 ) and our mean estimate for the sample of 395 open clusters . for the reddening
we have just used the differences as they were calculated . otherwise , those difference were normalized in units of the standard deviation of the mean for the distance and age .
the interpretation of those parameters is the following : for values between @xmath41 and + 1 , the error of averaging the literature parameter is statistically at the same level as the uncertainty introduced by using those catalogued ones . fitting a gaussian distribution to the complete sample of values should result in a width of less than 2@xmath14 if both `` error sources '' have the same statistical significance and can therefore not be disentangled ( rees 1987 ) .
figure [ dias ] shows the distribution of the differences of the distance , reddening and age as well as the basic parameters of a gaussian fit to these distributions .
the widths of all three distributions are in the expected range taking into account the overall statistical errors on the cluster parameters ( fig .
[ means_d ] ) .
we therefore conclude that if one uses the parameters of the catalogue by dias et al .
( 2002 ) the expected errors are comparable to those which we derived for averaging the independent values from the literature , a result which is in favor of the parameters from the mentioned catalogue from a statistical point of view .
if one aims at a statistical study of a group of clusters , a standard set for comparison is necessary .
mermilliod ( 1981 ) , for example used the following clusters to establish empirical isochronous curves : alpha persei , hyades , ic 4665 , ngc 457 , ngc 884 , ngc 2281 , ngc 2287 , ngc 2362 , ngc 2516 , ngc 3532 , ngc 3766 , ngc 6231 , ngc 6475 and the pleiades .
those clusters still serve as references when studying red giants ( eigenbrod et al .
another interesting work in this respect is the one by piatti , claria & bica ( 1998 ) who presented composite @xmath15 versus @xmath16 diagrams for the following `` standard '' open clusters : ic 4651 , ngc 3680 , ngc 3766 , ngc 6025 , ngc 6067 , ngc 6231 , ngc 6242 , ngc 6259 , ngc 6451 and ngc 6633 .
they selected these clusters because of accurately determined fundamental parameters .
the final diagrams are used as a homogeneous set of empirical isochrones in the age range between 5myr and 4gyr for the study of faint reddened open clusters .
the lists of mermilliod ( 1981 ) and piatti et al .
( 1998 ) only coincide in three cases : ngc 3766 , ngc 6231 and ngc 6475 .
this already indicates that a common set of standard open clusters is very much needed .
the lists of standard clusters given in the above mentioned references have been chosen based on the available photometric , spectroscopic and kinematical data . however , as an example , paunzen & maitzen ( 2002 ) have shown that the photometric data of ngc 6451 are not intrinsically consistent and suffer from an offset of one magnitude in johnson @xmath17 .
our approach is different and not based on the current available data for individual open clusters . on the contrary , our list should trigger new observations to further improve the knowledge of the listed cluster parameters .
we present a list of open clusters with a significant number of independent investigations and parameters with small errors .
we have chosen the rather strict criteria ( especially for the age estimate ) for the errors of the parameters : * @xmath10@xmath1825% * @[email protected] * @xmath12@xmath1830% in total , 72 open clusters , listed in table [ st_ocls ] , fulfill these criteria .
figure [ st_ocls_p ] shows the distribution of the distances , reddenings and ages for these proposed standard clusters .
they cover the complete range of known cluster parameters within the milky way .
these open clusters should serve as observational test cases for further theoretical developments .
a closer inspection of these open clusters in webda reveals that , for example , the data of ngc 6208 and ngc 6811 are still very unsatisfactory which calls for further high precision observational investigations .
using open clusters for studying global and local phenomena is very much dependent on the accuracy of the used cluster parameters such as the age , reddening and distance .
this is also vital for testing theoretical models .
we have made a statistical study of the published cluster parameters summarizing the results for 395 open clusters . while the reddening seems to be quite accurate ,
the ages and distances suffer from severe uncertainties . for about 90% of all studied clusters , the error of the reddening is below 0.1mag . for the distance we find absolute errors of less than 20% for about 80% of the aggregates .
however , only 11% of the investigated open clusters have errors of the ages which are less than 20% ; there are extreme discrepancies of more than 200% .
this calls for a homogeneous set of isochrones together with a solid fitting technique .
we have compared the errors of the mean cluster parameters with the differences of those listed in the open cluster catalogue by dias et al .
( 2002 ) which is widely used .
both error distributions have the same statistical significance .
using the parameters from the mentioned catalogue , therefore , introduces the same error level as using averaged values from the literature , an important result when using this catalogue . as last step , a set of 72 open clusters with the most accurate ( in a statistical sense ) parameters were established .
those clusters cover a wide range of reddenings , ages as well as distances and should serve in the future as a standard set for further observational efforts and for testing theoretical models .
+ + we thank our referee , jean - claude mermilliod , for very useful comments which helped to improve this manuscript significantly .
this research was performed within the projects _
p17580 _ and _ p17920 _ of the austrian fonds zur frderung der wissenschaftlichen forschung ( fwf ) .
m. netopil acknowledges the support by a `` forschungsstipendium '' of the university of vienna .
use was made of the simbad database , operated at the cds , strasbourg , the webda database , operated at the university of vienna and the nasa astrophysics data system .
999 andersen j. , nordstrm b. , 1999 , in gimnez a. , guinan e.f .
, montesinos b. , eds , asp conf .
173 , theory and tests of convection in stellar structure , san francisco , p. 31
chen l. , hou j.l .
, wang j.j . , 2003 ,
aj , 125 , 1397 dambis a.k
. , 1999 , astronomy letters , 25 , 7 dias w.s . ,
lpine j.r.d .
, 2005 , apj , 629 , 825 dias w.s .
, alessi b.s .
, moitinho a. , lpine j.r.d . , 2002 ,
a&a , 389 , 871 dutra c.m . , bica e. , 2000 , a&a , 359 , 347 eigenbrod a. , mermilliod j .- c .
, clara j.j . , andersen j. , mayor m. , 2004 , a&a , 423 , 189 gratton r. , 2000 , in pallavicini r. , micela g. , sciortino s. , eds , asp conf .
198 , stellar clusters and associations : convection , rotation , and dynamos , san francisco , p. 225
grocholski a.j .
, sarajedini a. , 2003 , mnras , 345 , 1015 janes k.a . , tilley c. , lyng , g. , 1988 , aj , 95 , 771 jrgensen b.r . , lindegren l. , 2005 , a&a , 436 , 127 karatas y. , bilir s. , schuster w.j . , 2005 , mnras , 360 , 1345 kharchenko n.v . , piskunov a.e .
, rser s. , schilbach e. , scholz r .- d . , 2005 , a&a , 438 , 1163 lata s. , pandey a.k . , sagar r. , mohan v. , 2002 , a&a , 388 , 158 lebreton y. , 2000 , ara&a , 38 , 35 loktin a.v . , gerasimenko t.p . , malisheva l.k . , 2001 ,
, 20 , 607 lyngg . , 1987 , catalogue of open cluster data , 5@xmath19 edition , cds , strasbourg malysheva l.k . , 1997 , astronomy letters , 23 , 585 mermilliod j .- c . , 1981 , a&a , 97 , 235 paunzen e. , maitzen h.m . , 2002 , a&a , 385 , 867 pena j.h . , peniche r. , 1994 , rmxaa , 28 , 139 piatti a.e . , clara j.j
, bica e. , 1998 , apjs , 116 , 263 pinsonneault m.h . ,
terndrup d.m . ,
hanson r.b .
, stauffer j.r . , 2004 , apj , 600 , 946 rees d.g . , 1987 , foundations of statistics , chapman & hall , london schaerer d. , meynet g. , maeder a. , schaller g. , 1993 , a&as , 98 , 523 tadross a.l . , 2001 , new astronomy , 6 , 293 twarog b.a . , ashman k.m . , anthony - twarog b.j . , 1997 ,
aj , 114 , 2556 winkler h. , 1997 , mnras , 287 , 481
|
we aim to characterize the current status of knowledge on the accuracy of open - cluster parameters such as the age , reddening and distance .
these astrophysical quantities are often used to study the global characteristics of the milky way down to very local stellar phenomena . in general , the errors of these quantities are neglected or set to some kind of heuristic standard value .
we attempt to give some realistic estimates for the accuracy of available cluster parameters by using the independent derived values published in the literature . in total ,
6437 individual estimates for 395 open clusters were used in our statistical analysis .
we discuss the error sources depending on theoretical as well as observational methods and compare our results with those parameters listed in the widely used catalogue by dias et al .
( 2002 ) .
in addition , we establish a list of 72 open clusters with the most accurate known parameters which should serve as a standard table in the future for testing isochrones and stellar models .
open clusters and associations : general
|
how might superpartners be detected if they exist ? they could be found in direct production at colliders as energy or luminosity is increased and a threshold is crossed in either . observing them
of course requires triggering on such events , and separating the signal from backgrounds . or , their effects could be seen from one - loop contributions .
most possible deviations from the , whether particle production or loop effects , could not be interpreted as signals of supersymmetry , since supersymmetric signals are strongly constrained as to which processes they can contribute to .
for example , susy production events should have missing energy ( assuming @xmath7-parity conservation ) , while deviations caused by loop effects should show up most strongly in processes such as @xmath8 and @xmath9 ( but _ not _ in @xmath10 quark asymmetries ) . in the past couple of years
some evidence of such effects has been reported@xcite .
here we do not wish to describe this evidence or argue that it is , in fact , evidence for superparticles , but to observe that it can only be interpreted as such evidence and be consistent with all existing data if certain parameters lie in certain ranges .
therefore , it is appropriate to study the predictions of models with such parameters .
we can summarize these ranges approximately by & & m_1 m_2 m_z < , > > , & & 45 65 .
[ massh ] here @xmath11 and @xmath12 are the ( unsigned ) neutralino and chargino mass eigenvalues , @xmath13 and @xmath14 are the left and right charged slepton masses , @xmath15 is the sneutrino mass , @xmath16 and @xmath17 are the @xmath18 and @xmath19 gaugino masses at the electroweak scale , @xmath20 is the coefficient of the @xmath21 term in the superpotential at the electroweak scale , and @xmath5 is the ratio of the higgs vevs .
we also take the right slepton mass @xmath14 to be of order 125 gev , but this choice only indirectly affects the lep analysis .
we further impose @xmath22 , so that we qualitatively expect events such as @xmath23 at fermilab@xcite .
the models discussed here do not depend on the light stop mass indicated in ( [ inputs ] ) for their interest .
however , the relative size of @xmath24 and @xmath25 affects the signatures considerably , so we examine all cases .
we use the word `` models '' to denote different choices of the above parameters , all constrained to be in the ranges specified by ( [ inputs ] ) and ( [ massh ] ) .
our entire analysis is in terms of an effective general lagrangian at the weak scale , including soft supersymmetry breaking parameters .
the scenario defined by ( [ inputs ] ) is at present a well - motivated one to study for its implications for signals at lep ( and fermilab ) , and that is the purpose of this paper .
further , the detectable signatures for lep are rather non - standard .
for example , the largest susy signals may occur as events with two photons and large missing energy , an excess of single photon events with recoil mass above about 140 gev , an excess of events with two soft charged leptons and large missing energy , etc . even though some superpartner production cross sections are large ( of order 1 picobarn ) , giving hundreds of produced events , most are simply invisible !
the signals that have been most often studied from charginos and neutralinos are relatively small in this scenario . the key features which give these models their unique character are large branching ratios to invisible final states of the sneutrinos and @xmath26 , as well as a significant branching ratio for the radiative decay of @xmath27 . in the following we
first summarize the masses and decays predicted for individual superpartners , and then describe the resulting cross sections and signatures .
table [ decays ] summarizes the properties of the relevant superpartners .
the constraints in eq .
( [ inputs ] ) imply@xcite that @xmath28 is dominantly higgsino ( the approximately symmetric combination of @xmath29 and @xmath30 ) , @xmath27 is mainly photino , and @xmath26 is mostly antisymmetric higgsino , but with a non - negligible zino component .
the charginos are neither pure wino nor pure charged higgsino ( the @xmath31 mixing matrix is approximately off - diagonal , while all of the @xmath32 mixing matrix elements are roughly the same magnitude ) .
all of the results of table [ decays ] follow straightforwardly .
we have included a row for the stop since it is motivated if there are deviations from the in @xmath33 and @xmath34@xcite , by electroweak baryogenesis@xcite , and to some observers by aspects of fermilab data@xcite . however , the crucial features of the class of models we are studying do not change if we take the stop mass to be heavier than indicated in table [ decays ] . as described in detail in ref .
@xcite , an important feature of the susy parameters chosen in ( [ inputs ] ) is a large value of @xmath35 .
consequently , photons will play a significant role in the signals within this class of models .
the remainder of the @xmath27 decay rate is to 3-body final states . of the possible
3-body decays , @xmath36 has the largest branching ratio , because of the relatively low slepton masses .
a key result of the mass hierarchy ( [ massh ] ) concerns the sneutrino decay modes .
the only two - body modes open are @xmath37 and @xmath38 .
in fact , the decay to @xmath39 dominates , rendering the sneutrino almost entirely invisible . likewise , since @xmath40 dominates , @xmath26 is also mainly invisible .
since @xmath41 and @xmath42 are among the largest cross sections , most sparticle production at lep is invisible .
when @xmath43 or @xmath26 do have a visible decay , it contains a single @xmath44 , or possibly @xmath45 .
the heaviest neutralino , @xmath46 , has decays which are very similar to @xmath26 .
the dominant @xmath46 decay mode is @xmath47 .
up to a quarter of the total @xmath46 decay rate is to @xmath48 .
so while most @xmath46 s are invisible , a significant fraction will produce @xmath49 .
if the higgs mass is small enough and the @xmath46-@xmath28 mass splitting large enough , the decay @xmath50 opens up , suggesting the possibility of a significant unconventional source of higgs bosons at lep .
however , the amount of available phase space is small , limiting this branching ratio to at most a percent or two , rendering such prospects dim .
charginos and stops decay very differently depending on their relative mass .
here we focus upon @xmath51 and @xmath52 since @xmath53 and @xmath54 are likely to be too massive to be important at lep in the short term .
there are three interesting possibilities : @xmath55 , @xmath56 , and @xmath57 . in region
the stop decays exclusively to @xmath58 , as no other two - body modes are kinematically allowed . for the chargino ,
both @xmath59 and @xmath60 final states are allowed .
generically we expect both modes to have sizeable branching ratios , although phase space suppression or the size of the stop mixing angle can cause one or the other of the two modes to dominate . in region
the decay @xmath61 accounts for virtually 100% of the rate , while @xmath62 dominates the chargino decays , so that stops mainly end up as a @xmath63 ( @xmath64 invisible ) final state . in some cases , @xmath65 can be significant ( _ i.e. _ a few percent ) , though seldom dominant .
this is because we exclude the small corner of parameter space where @xmath66 , since constraints from lep161 and lep172 suggest that @xmath65 is probably small for @xmath52 .
region combines features from the first two regions : we have @xmath67 ( as in region ) , but the @xmath52 decays mostly to @xmath59 , ( as in region ) . in the cases where the dominant @xmath52 decay is to @xmath59 , we note that most reported limits on @xmath68 do not apply , both because the @xmath52 decay is non - standard and because the @xmath52 cross section is reduced by @xmath43 exchange for our range of @xmath15 .
we discuss those signals which are large enough to be detectable at lep with @xmath69 pb@xmath6 per detector at a center of mass energy @xmath70 .
the cross section estimates presented below were obtained using the spythia monte carlo @xcite .
although @xmath71 and @xmath72 production are among the largest cross sections , they are almost entirely invisible , as described above . however , on occasion we have @xmath73 , implying @xmath74 , where we use @xmath75 to stand for a set of invisible particles .
since @xmath75 includes two @xmath28 s and a neutrino , and one of the @xmath28 s must combine with the neutrino to form an on - shell sneutrino , the _ minimum _ missing invariant mass is @xmath76 .
similarly , on occasion @xmath71 will give rise to @xmath77 , with a missing invariant mass of at least @xmath78 .
the production of @xmath79 followed by the dominant @xmath27 and @xmath26 decays also leads to a single photon plus missing energy : in fact , this mode accounts for the majority of the @xmath80 total rate .
photons can also be radiated from the initial electrons ( or from the @xmath81-channel chargino in @xmath72 production , although this contribution is small ) ; the threshold recoil mass here for @xmath72 is about 150 gev , and for @xmath71 about 120 gev .
other channels that can give a photon include @xmath82 with a radiated @xmath44 , @xmath83 , etc . , but these give a smaller contribution@xcite
. the entire effect can be large , giving an excess over the single photon rate for large missing invariant mass as one signature for supersymmetry .
in the models we are considering , the total susy - related @xmath80 rate is typically between 100 and 300 fb .
once the signal is detected , it constrains the @xmath28 , @xmath27 , @xmath26 , and @xmath43 masses .
of course , there is a background for this @xmath80 channel from @xmath84 and direct @xmath85 production via @xmath86-exchange .
most of the background is not in the region of interest here ( missing mass well above @xmath87 ) , but enough is that some study is required to determine the best cuts on the observed photon .
the authors of ref .
@xcite have studied largely invisible susy signatures and how they might appear in the @xmath80 channel .
the susy cases they examine do not overlap ours , but some of the phenomenology is the same .
their comments on backgrounds are relevant , particularly for @xmath88 events where neither lepton is detected .
another channel that can give a signal is production followed by @xmath89 and @xmath90 provides another ( but much smaller ) source of @xmath91 events . ]
e^+e^- i. [ twophoton ] this channel is particularly interesting because about six events with missing invariant masses greater than 100 gev have been reported@xcite from lep161 and lep172 running , combining all four detectors and both energies .
the events are precisely in the region where we expect a signal in our models ( _ i.e. _ missing invariant mass above 100 gev ) .
it is very important to know the background well , since the signal may only be a few times larger .
the best estimate of the background is by s. ambrosanio @xcite , who has done a careful calculation at tree - level using comphep3.0 @xcite . for photons satisfying the requirements @xmath92 , @xmath93 , and @xmath94
, he finds a cross section of @xmath95 fb , which gives a background of 1.6 events for perfect photon detection efficiency . assuming an average photon efficiency of 0.8 , the final expected background is about 1.3 events .
we have made checks using pythia @xcite ( which is not ideal for such calculations ) , and find numbers consistent with ambrosanio s , and certainly not noticeably larger . if a signal is found , a precise evaluation of the higher - order corrections to this background would be useful .
as far as we know , this has not yet been carried out .
the signal has a missing invariant mass @xmath96 larger than @xmath78 , and photons that are never very soft ( because we require @xmath97 ) .
the energy range for the photons produced in ( [ twophoton ] ) is e_min , max = ( ^2 - ^2 ) ( 1 ) [ eminmax ] where @xmath98 and @xmath99 are the neutralino masses and @xmath100 is the beam energy . although initial state radiation and detector resolution effects will result in some photons having less energy than this minimum , the majority of the signal events in our models will have @xmath101 .
in contrast , the background ( from @xmath102 and @xmath103 ) has a missing invariant mass distribution which is concentrated around the @xmath104 peak ; those background events which do have large @xmath96 tend to contain low - energy photons .
thus , the analysis should impose a minimum energy requirement of about 8 gev for each photon , and define three regions for @xmath96 , say @xmath105 , @xmath106 , and @xmath107 .
the signal we predict is entirely in the region @xmath107 , while little background is in this region ( see above ) . of course
, if softer photons or events from the @xmath104 peak are included in the signal region it will be harder to detect a signal .
even with our tightly constrained parameter space there is a large variation in @xmath108 , but the majority of the parameter space gives cross sections at lep184 from 50400 fb .
( if we consider only those models which imply 310 events of @xmath91 at lep161+lep172 , then this range narrows to 100220 fb . ) note that this channel is independent of the single photon channel as a susy signal .
once the signal is detected , it constrains the @xmath28 and @xmath27 masses .
if the value of @xmath35 is only somewhat larger than 50% , the process @xmath109 can become important .
production also provides a source of @xmath49 events ( with differing kinematics ) . however , at lep184 the cross section times branching ratio for this mode is at most about 10 fb . ]
the signal in this case is large missing energy ( @xmath110 ) and two leptons with a pair mass many widths below the @xmath104 ( @xmath111 ) .
our parameter space contains models with up to 200 fb in this channel . the other channels at lep184 which could have sizeable cross sections are @xmath112 and @xmath113 ( see fig . [ lepfig ] ) .
for charginos , we just combine the single chargino results above . in region
, chargino decays to @xmath60 and @xmath59 can be comparable , so there are three different signatures : @xmath114 ( @xmath115 ) , @xmath116 ( @xmath110 ) , and @xmath117 ( @xmath118 ) . here and below @xmath119 and @xmath120
are any charged leptons , but @xmath120 can be different from @xmath119 . in regions and ,
the decay @xmath62 dominates , so chargino pairs primarily give @xmath114 ( @xmath115 ) . because we have @xmath121 , the decay @xmath122 never dominates .
however , it can lead to @xmath123 ( @xmath124 ) as an additional signature , but at a much reduced rate . in all cases ,
@xmath114 ( @xmath115 ) will be important as the chargino pair signature , with all combinations of @xmath125 and @xmath126 possible . since @xmath127 , @xmath128 , and @xmath129
are not guaranteed to be exactly degenerate , the relative number of each type of lepton pairs can not be precisely predicted .
once such events are seen , they will provide information about the sneutrino mass splittings .
for stops we proceed similarly . in regions and
stop pairs produce @xmath130 ( @xmath110 ) . in region stop pairs give the signature @xmath131 ( @xmath115 ) .
note that the @xmath10 and @xmath132 can be very soft .
the observation of @xmath114 ( @xmath115 ) could signal either charginos or stops .
the presence of a soft @xmath133 pair , which might simply appear as a large hadron multiplicity , would tell us that @xmath134 .
the biggest potential background to the @xmath114 channel is @xmath135 .
however , leptons coming from @xmath86-pair production are fairly stiff ( @xmath136 at @xmath70 ) .
events containing two leptonic tau decays produce leptons which are much softer .
however , this mode is suppressed by the branching ratio factor @xmath137 ^ 2 = 1.5 \times 10^{-3}$ ] . ] on the other hand , the @xmath52-@xmath43 mass difference in our models tends to be smaller than 10 gev , and is frequently only a few gev .
consequently , the maximum lepton energy coming from @xmath52 decay in @xmath112 events at lep184 is only a few gev .
thus , a good event selection strategy would veto energetic leptons , but retain leptons which are as soft as possible .
finally , in some of our models , left - handed slepton pairs are ( barely ) light enough to be produced at lep184 , giving @xmath138 .
unfortunately , the cross section is phase - space limited , and amounts to only 23 fb per slepton flavor .
in the scenario described in this paper , many light superpartners are produced at fermilab .
however , the combination of the near invisibility of @xmath26 , @xmath46 , and @xmath43 with the backgrounds at the tevatron makes for few good signals .
the sparticles which give potentially visible decays are the charginos ( especially @xmath54 ) , @xmath27 , and the lighter stop ( and , possibly , the gluino and heavier squarks : see the end of this section ) . at this point
, some model dependence enters the discussion , as the relative masses of the @xmath54 and @xmath51 are important in determining their decay modes .
however , two signals which are likely to be important are the inclusive @xmath139 and @xmath140 rates .
the first of these two signals is of interest in a wider category of models than considered here @xcite , while the second is special in that it has no significant parton - level background .
we begin our discussion with @xmath139 .
both the collider detector at fermilab ( cdf ) and d - zero ( d0 ) collaborations have reported results on searches for such a signal @xcite . the cdf results are still preliminary , and do not yet include an upper limit on the @xmath139 rate .
d0 , however , reports the 95% c.l .
upper limit @xmath141 fb for photons satisfying the following cuts : transverse energy @xmath142 , pseudorapidity @xmath143 and missing transverse energy @xmath144 within the group of models we are studying here , we obtain this signal from @xmath145 , @xmath146 , @xmath147 , @xmath148 , and @xmath149 production .
since the production cross section for @xmath54 pairs at the tevatron is large ( @xmath150700 fb in our models ) , it potentially provides the largest contribution to the signal .
first , let us assume that the stop is heavy ( @xmath151 ) . then ,
@xmath152 , and , including the remaining branching ratios , we obtain a contribution of around 100 fb . however , when we account for the effects of the cuts employed by d0 , we find that the total @xmath139 rate is never more than 60 fb , even when the other initial states are added .
thus , our models are consistent with current fermilab search limits .
the expected number of @xmath139 events actually decreases if the stop mass is lowered .
first , a smaller stop mass implies a smaller @xmath153 branching ratio@xcite .
second , if @xmath154 , then the decay @xmath155 opens up , allowing for a @xmath140 final state when the two charginos decay differently .
this possibility is especially interesting since there is no significant parton - level source of such events within the standard model .
depending upon the value of @xmath156 , we estimate that the size of the @xmath140 signal from @xmath145 could be up to 100 fb .
if @xmath24 is even somewhat lighter still , the decay @xmath157 becomes allowed . in this case ,
@xmath158 production followed by @xmath159 and @xmath160 provides an additional source of @xmath140 events .
although @xmath161 tends to be only a few percent at most , the @xmath162 production cross section is enormous ( @xmath163 pb ) .
thus , even a modest 1% value for this branching ratio can lead to an additional 100 fb of @xmath140 production . in connection with this possibility
, we note that if @xmath157 is allowed , then it is likely ( but not certain ) that @xmath164 , in which case the signal becomes @xmath165 .
additional consequences for the tevatron which arise in models which allow top - to - stop decays are discussed in refs .
@xcite .
finally , we remark that if charginos and sleptons are in the mass ranges of table [ decays ] , then in many models gluinos and squarks of the first two families fall in the mass range 200300 gev .
these have large cross sections at fermilab and might also be observable@xcite , possibly even in the present data sample .
the signatures of some such events could cause them to be included in the top quark sample .
this paper reports the predictions for lep of a particular , interesting , region of the susy parameter space .
it is worth reporting these predictions because they are quite different from those of most susy analyses .
for example , the largest cross sections ( @xmath71 and @xmath72 ) are almost completely invisible ; they only show up occasionally as a single photon plus large missing energy .
the cleanest susy signature may be @xmath166 plus large missing energy , from @xmath167 .
chargino pairs mainly give @xmath168 plus large missing energy , where @xmath119 and @xmath120 can be different leptons and are soft .
stop pairs and selectron pairs are also possible .
if no signals are observed at lep184 , this susy scenario is almost , but not quite , eliminated .
events from lep161 and lep172 running as @xmath147 production would become untenable in the absence of a signal at lep184 .
] at lep194 , this scenario is completely excluded if no signatures are seen with @xmath69 pb@xmath6 per detector . note that most , but not all , sets of masses consistent with eqs .
( [ inputs ] ) and ( [ massh ] ) are also consistent with all present data .
this is because if ( [ inputs ] ) and ( [ massh ] ) do indeed describe the real world , then the sparticles are on the verge of being detected .
if such signals are seen , it will be easy ( and fun ) to extract from even limited data the remaining parameters of the chargino , neutralino and left - handed slepton sectors , @xmath5 , and ( if present ) light stop to good accuracy , even in the most general framework of a softly broken supersymmetric theory .
if a model such as those examined here was indeed observed , the implications for the structure of the fundamental theory will be unusually interesting , since some features are likely to be different from those in most minimal susy models . for example , @xmath169 seems to be nearer to unity than to @xmath170 , and @xmath171 ( which could happen , for example , from the @xmath172-terms coming from an extra @xmath18 @xcite ) .
if these signals are seen it is very likely that the lightest susy higgs boson is accessible at lep , and it is certainly detectable at fermilab within a few years .
this research is supported in part by the u.s .
department of energy .
we appreciate conversations with j. gunion , d. stickland , d. treille , and g. wilson .
we are very grateful to g. kribs for checking some results and for helpful discussions , and to s. ambrosanio for discussions and the use of some of his computer programs to check some of the results .
finally , we thank s. ambrosanio , g. kribs , and s. martin for reading the manuscript and providing useful feedback .
g.l . kane and s. mrenna ,
* 77 * , 3502 ( 1996 ) .
g. mahlon and g.l .
kane , phys .
rev . * d55 * , 2779 ( 1997 ) .
h. komatsu and j. kubo , phys .
157b * , 90 ( 1985 ) ; h.e .
haber , g.l .
kane , and m. quiros , phys .
160b * , 297 ( 1985 ) ; h.e .
haber and d. wyler , nucl .
b323 * , 267 ( 1989 ) ; s. ambrosanio and b. mele , phys .
* d53 * , 2541 ( 1996 ) ; * d55 * , 1399 ( 1997 ) .
s. ambrosanio , private communication .
additional information and plots are available at http://feynman.physics.lsa.umich.edu/~ambros / phys/2photon+emiss/. further discussion of the @xmath173 background is contained in s. ambrosanio , g.d .
kribs , and s.p .
martin , hep - ph/9703211 , as well as the last reference in @xcite .
boos , m.n .
dubinin , v.a .
ilyin , a.e .
pukhov , and v.i .
savrin , `` comphep : specialized package for automatic calculations of elementary particle decays and collisions , '' hep - ph/9503280 , and references therein .
s. dimopoulos , m. dine , s. raby , and s. thomas , phys . rev
. lett . * 76 * , 3494 ( 1996 ) ; s. ambrosanio , g.l .
kane , g.d .
kribs , s.p .
martin , and s. mrenna , phys .
lett . * 76 * , 3498 ( 1996 ) ; s. dimopoulos , s. thomas , and j.d .
wells , phys .
* d54 * , 3283 ( 1996 ) ; s. ambrosanio , g.l .
kane , g.d .
kribs , s.p .
martin , and s. mrenna , phys . rev .
* d54 * , 5395 ( 1996 ) .
.masses and main decays of sparticles relevant to lep , for models satisfying the mass hierarchy of eq .
( [ massh ] ) . unless specific quark or lepton flavors are indicated , we quote the sum over all flavors .
[ decays ] [ cols="^,^,^,^,^,^,^,^ " , ]
|
there is a class of supersymmetric models which is well - motivated by hints of evidence for susy and consistent with all existing data .
it is important to study the predictions of these models .
they are characterized by @xmath0 ( where @xmath1 and @xmath2 are neutralino and chargino mass eigenstates ) , @xmath3 , @xmath4 , and @xmath5 near 1 .
their lep signatures are mostly unusual .
most produced superpartners are invisible ! a good signature is two photons plus large missing energy .
there are also excess events at large recoil mass in the single photon plus nothing channel .
we list the main signatures for charginos , stops , etc . , which are also likely to be unconventional
. this class of models will be definitively tested at lep194 with 100 pb@xmath6 per detector , and almost definitively tested at lep184 .
# 1_t_#1 # 1_n_#1 # 1_#1 # 1#2_#1^#2 # 1_c_#1 # 1__#1 # 1#2_#1^#2
|
as pointed out by koponen , fractal or inverse power law distributions are of interest in modeling various meaningful situations in solid state physics @xcite .
an example , treated in ref .
@xcite , is the thermalization of a nonequilibrium electron - phonon system . + until recently , however , there has been little guidance on how to generalize the kinetic theory of electrons and phonons obeying non - gibbsian statistics . in a very recent paper @xcite
a generalized kinetic theory for electrons and phonons has been proposed .
modified collision terms were introduced , in order to allow applications not only to electrons , but also to other particles ( obeying a general statistics ) which interact with a crystal lattice .
+ in the present paper our aim is to introduce some models for the generalization of the ktep equations as well as to study equilibrium and stability for the gktep equations .
we introduce a couple of models for the gktep , relevant to non standard quantum statistics recently proposed in the literature , and compare them from the point of view of kinetic theory .
both of them give rise to inverse power decays of the distribution function with respect to energy . in absence of external disturbances ,
it is possible to treat phonons as background at equilibrium .
this assumption implies a drastic simplification of the gktep , since we have just to deal with the equation for electrons .
equilibrium solutions and their stability are investigated by means of lyapunov s theory .
the connection between mathematical results and thermodynamics is shown .
first we recall the gktep equations and their stationary solutions .
let @xmath0 be the distribution function of phonons ( quasi - momentum @xmath1 , energy @xmath2 of branch @xmath3 of the phonon spectrum and @xmath4 the distribution function of electrons [ quasi - momentum @xmath5 , energy @xmath6 .
+ the generalized kinetic equations for phonons and electrons read @xcite : @xmath7 where @xmath8 and @xmath9 .
the right hand sides are given by @xmath10\\ \nonumber & & + \sum_{g_1g_3}w_{pp}({{\mbox{\boldmath${k}$}}},\,{{\mbox{\boldmath${k}$}}}_1\rightarrow{{\mbox{\boldmath${k}$}}}_3)\,\delta(\omega_{g_3}-\omega_g-\omega_{g_1})\\ & & \times[\phi(n_{g_3})\,\psi(n_g)\,\psi(n_{g_1})- \phi(n_g)\,\phi(n_{g_1})\,\psi(n_{g_3})]\big\rbrace\,\frac{d{{\mbox{\boldmath${k}$}}}_1 } { 8\,\pi^3 } \ , \\ \nonumber & & \\ \nonumber \left(\frac{\partial n_g}{\partial\ , t}\right)_{pe}&=&2\int w_{pe}({\bf p}\rightarrow{\bf p}^\prime,\,{{\mbox{\boldmath${k}$}}})\,\delta(\epsilon_{\bf p^\prime}+\omega_g-\epsilon_{\bf p})\\ & & \times[\varphi(n_{\bf p})\,\psi(n_{\bf p^\prime})\,\psi(n_g)-\phi(n_g)\,\varphi(n_{\bf p^\prime})\ , \psi(n_{\bf p})]\,\frac{d{{\mbox{\boldmath${p}$}}}}{8\,\pi^3 } \ , \\ \nonumber & & \\
\nonumber \left(\frac{\partial n_{\bf p}}{\partial t}\right)_{ep } & = & \sum_g\int\big\lbrace w_{ep}({{\mbox{\boldmath${p}$}}}^\prime,\,{{\mbox{\boldmath${k}$}}}\rightarrow{{\mbox{\boldmath${p}$}}})\,\delta(\epsilon_{\bf p}-\epsilon_{\bf p^\prime}-\omega_g)\\ \nonumber & & \times[\phi(n_g)\,\varphi(n_{\bf p^\prime})\,\psi(n_{\bf p } ) -\varphi(n_{\bf p})\,\psi(n_{\bf p^\prime})\,\psi(n_g)]\\ \nonumber & & + w_{ep}({{\mbox{\boldmath${p}$}}}^\prime\rightarrow{{\mbox{\boldmath${p}$}}},\,{{\mbox{\boldmath${k}$}}})\,\delta(\epsilon_{\bf p}+\omega_g-\epsilon_{\bf p^\prime})\\ & & \times[\varphi(n_{\bf p^\prime})\,\psi(n_{\bf p})\,\psi(n_g )
-\phi(n_g)\,\varphi(n_{\bf p})\,\psi(n_{\bf p^\prime})]\big\rbrace\,\frac{d{{\mbox{\boldmath${k}$ } } } } { 8\,\pi^3 } \ , \label{15}\end{aligned}\ ] ] where the @xmath11 s are transition probabilities and electron - electron interactions have been neglected .
observe that in the limit @xmath12 the bloch - boltzmann - peierls ( bbp ) equations @xcite are recovered .
the non negative functions @xmath13 , @xmath14 for phonons [ @xmath15 $ ] , and @xmath16 , @xmath17 for electrons ( @xmath18 $ ] ) represent the arrival and departure state availability , respectively .
moreover we assume : + _ i _ )
@xmath19 and @xmath20 are monotonically increasing functions of @xmath21 and @xmath22 , respectively @xcite ; + _ ii _ ) the following conditions hold @xmath23 conditions ( [ c1 ] ) mean that no transition occurs when the initial state is empty , conditions ( [ c2 ] ) mean that the emptiness of the arrival state neither inhibits nor enhances a transition and conditions ( [ c3 ] ) are characteristic of fermions and bosons , respectively .
+ in @xcite it has been shown that , implicitly , the equilibrium solutions of the generalized system are : @xmath24 ( @xmath25 means `` at equilibrium '' ) , where @xmath26 is the absolute temperature of the mixture and @xmath27 is the chemical potential of the electron gas .
preliminarily , consider certain particles , whose equilibrium distribution function is @xmath28 , which obey a statistics described by the parameter @xmath29 $ ] .
hereinafter hatted quantities are referred to these particles . in particular , for @xmath30 we have fermions , classical particles , and bosons , respectively .
suppose , we invert this distribution function with respect to @xmath31 [ where @xmath32 than we obtain : @xmath33 in particular , for electrons @xmath34 , for phonons @xmath35 , since , in this case , @xmath36 the statistics of the particles we deal with is described also by the couple of function @xmath37 and @xmath38 , essentially positive , and obeying the following conditions : @xmath39 in particular we have @xmath40 from ( [ 1 ] ) and ( [ 2 ] ) we know that at equilibrium @xmath41 now , eqs .
( [ 3 ] ) and ( [ 4 ] ) give @xmath42 let us introduce the following assumptions : + _ i _ ) @xmath43 is given by eq .
( [ 5 ] ) not only at equilibrium ; + _ ii _ ) @xmath37 does not depend on @xmath44 , and we take @xmath45 ; + _ iii _ ) @xmath46 .
+ properties _ i _ ) , _ ii _ ) and _ iii _ ) actually hold for standard statistics ( fermi - dirac , maxwell - boltzmann , and bose - einstein ) and , more in general , we postulate them . from eq .
( [ 5 ] ) , by accounting for _ ii _ ) and _ iii _ ) , we get @xmath47 we introduce now two models which account for an inverse power law decay of the distribution function with respect to energy : + * 1 * ) the first model that we consider has been proposed by bykkili et al . @xcite .
relevant applications to the blackbody problem can be found in @xcite .
preliminarily , we define the @xmath48-deformed exponential and logarithm @xcite : @xmath49^{\frac{1}{q-1 } } } \ , \hspace{2 cm } \ln_q(x)={1-(1/x)^{q-1 } \over q-1 } \ .\end{aligned}\ ] ] the deformed quantum distribution function is given by @xmath50 where the parameter @xmath48 is greater or equal to 1 ( in the last case the standard statistics are recovered ) . since @xmath51\},$ ] from eq .
( [ 6 ] ) we get @xmath52 we have @xmath53 and @xmath54 for @xmath55 , otherwise it either vanishes ( for @xmath56 ) or it diverges ( for @xmath57 ) , @xmath58 and @xmath59 for fermions we observe that @xmath60 is not fulfilled .
moreover , in the case of bosons the condition @xmath61 does not hold .
+ figures 1a and 1b describe the couple of function @xmath17 and @xmath16 versus both @xmath62 and @xmath48 .
figures 2a and 2b describe the couple of functions @xmath14 and @xmath13 versus both @xmath63 and @xmath48 .
figure 1b shows clearly the violation of condition @xmath60 .
both @xmath17 and @xmath14 start with a vanishing derivative , while both @xmath16 and @xmath13 start with an infinite derivative .
+ * 2 * ) recently kaniadakis @xcite has proposed a new classical distribution function , together with its quantum counterparts .
an application to bose - einstein condensation can be found in @xcite .
preliminarily , we define the new @xmath64-deformed exponential and logarithm : @xmath65 \ , \hspace{1.5 cm } \ln_{\{\kappa\}}(x)={1\over\kappa}\,\sinh\left(\kappa\,\ln x\right ) \ .\end{aligned}\ ] ] observe the remarkable property @xmath66 .
the deformed quantum distribution function is given by @xmath67 where the parameter @xmath64 is taken here greater or equal to 0 ( in the last case the standard statistics are recovered ) . since @xmath68 $ ] , from eq .
( [ 6 ] ) we obtain @xmath69 we have @xmath53 and @xmath54 for @xmath70 , otherwise , it either vanishes ( for @xmath56 ) or it diverges ( for @xmath57 ) , and @xmath71 \ .\end{aligned}\ ] ] for electrons , the last result agrees with the condition @xmath60 . in the case of bosons ,
the condition for @xmath14 when @xmath72 is fulfilled as well .
figure 1c and 1d describe the couple of functions @xmath17 and @xmath16 versus both @xmath62 and @xmath64 .
figures 2c and 2d describe the couple of functions @xmath14 and @xmath13 versus both @xmath63 and @xmath64 .
figure 1d shows clearly the fulfillment of conditions @xmath60 .
both @xmath17 and @xmath14 start with a vanishing derivatives , while both @xmath16 and @xmath13 start with an infinite derivative .
a usual approximation to the bbp equations consists in considering the relaxation due to p
p interactions to be much quicker than the one due to e p interactions .
this assumption amounts to fix , in the kinetic equation for electrons , the distribution of phonons as the equilibrium function ( [ 1 ] ) at a given temperature @xmath26 .
+ the kinetic equation for electrons reads now @xmath73 where @xmath74 means that @xmath75 has been substituted for @xmath21 . in order to study equilibrium and its stability ,
it is useful to introduce the following functional : @xmath76\,d\,{{\mbox{\boldmath${p}$ } } } \ , \label{54}\end{aligned}\ ] ] which can be rewritten as follows : @xmath77\,\frac{d{{\mbox{\boldmath${k}$}}}\,d{{\mbox{\boldmath${p}$}}}}{8\,\pi^3 } \
\le\ 0 \ , \label{55}\end{aligned}\ ] ] since @xmath78 .
+ in the space homogeneous and forceless case , equilibrium solutions are given by @xmath79 * proof . * in fact , we observe that eq .
( [ 57 ] ) @xmath81 eq .
( [ 56 ] ) [ see eq .
( [ 15 ] ) ] . on the other hand , from eq .
( [ 54 ] ) it follows that eq .
( [ 56 ] ) @xmath82 .
tanks to eq .
( [ 55 ] ) , we obtain @xmath83 eq . ( [ 57 ] ) @xmath84 eq .
( [ 57 ] ) , by taking into account eq .
( [ 1 ] ) and @xmath85 , shows that , at equilibrium , @xmath86+\epsilon_{\bf p}/t$ ] is a collisional invariant for electrons , that means @xmath87 is given by eq .
( [ 2 ] ) . in order to study the stability of such an equilibrium solution , we introduce the following functional @xmath88 where @xmath89 and @xmath90 ( observe that @xmath91 is a convex function of @xmath22 , since @xmath20 has been assumed to be monotonically increasing ) . + we can prove an h theorem for the present problem : * proof . * by utilizing the definition of @xmath92 together with eq .
( [ 53 ] ) , one obtains @xmath93 , where the dot means the time derivative . moreover , since @xmath94 by taking into account electron conservation , we have @xmath95 so that we can write @xmath96\right\}\,d{{\mbox{\boldmath${p}$}}}\ \ge0 \ , \end{aligned}\ ] ] due to the convexity of @xmath97 let us now interpret this result on a physical ground .
first , we define the concentration n , the energy density @xmath98 , and the entropy density @xmath99 of the electron gas as follows : @xmath100 where the factor 2 inside these integrals accounts for degeneracy .
observe that this definition of @xmath99 is consistent with the definitions of @xmath26 and @xmath27 , which we have already given .
in fact , at equilibrium , the following thermodynamical relationships @xcite are recovered : @xmath101 now , from the definition of @xmath92 , it is easy to realize that the meaning of @xmath102 is nothing but the clausius inequality @xmath103
|
in the present paper our aim is to introduce some models for the generalization of the kinetic theory of electrons and phonons ( ktep ) , as well as to study equilibrium solutions and their stability for the generalized ktep ( gktep ) equations .
we consider a couple of models , relevant to non standard quantum statistics , which give rise to inverse power law decays of the distribution function with respect to energy . in the case of electrons in a phonon background , equilibrium and stability
are investigated by means of lyapounov theory .
connections with thermodynamics are pointed out .
|
the interaction of the soft pions at the comparatively low energies is described by the effective lagrangian @xmath0,\ ] ] including three isovector pion fields @xmath1 , @xmath2 , and an auxiliary scalar field @xmath3 obeying the constraint @xmath4 where @xmath5 mev is the pion decay constant . the pion field represents the chiral phase of the quark condensate , that is why the parameterization through the unitary matrix is quite natural in this approach .
constructing the matrix @xmath6 the lagrangian ( [ lpi ] ) takes the form @xmath7 the lagrangian ( [ leff ] ) involves non - linear terms responsible for the many - pion interaction and allows for the classical solutions .
the important class of plane wave type solutions for the pion field was described by a.a.anselm @xcite , the disoriented chiral condensate ( dcc ) being the particular case of it with the wave vector @xmath8 @xcite . in the collision of high energy particles ( or , better , heavy ions )
the system is `` warmed up '' to the temperatures at which the chiral symmetry is restored .
then , in the course of cooling , the symmetry breaks again and the scalar , @xmath9 , or pseudoscalar , @xmath10 , condensate with the quantum numbers of @xmath3 or @xmath11 mesons settles out . all four orientation ( 3 pions and @xmath3-meson )
are equivalent in the isotopic @xmath12 space , and if the pion condensate , @xmath10 , is produced somewhere it means the formation of dcc domain . possible scenarios for the classical pion field ( dcc ) production at the high energies and the experimental signatures of dcc have been widely discussed in the literature @xcite .
one of the most significant evidence of dcc is the distribution over the neutral and charged pions .
the multiplicity distribution obeys the poisson law for independent production , so the number of @xmath13 mesons is 1/3 of the total number of pions in the large multiplicity event while the distribution of the ratio @xmath14 is close to @xmath15 ( here @xmath16 is the number of @xmath13 , @xmath17 is the total number of pions ) . in case of a classical pion field creation the orientation of the isotopic vector @xmath18 , which determines the ratio of charged and neutral pions ,
is chosen once for the whole classical field domain .
it results into the large probability , @xmath19 of the events with anomalous small number of neutral pions @xcite
. there have not been still a real progress in experimental attempts to observe such events ( in @xmath20 collisions at fnal collider @xcite and in nucleus - nucleus reactions in cern@xcite ) although the hope to create dcc in this conditions was not originally much . the heavy ion collisions at the new colliders ( rhic , lhc ) seem to be more promising . in this paper
we deal with another possibility - with classical pion fields , which can be produced and exist in the presence of an external source i.e. nucleons or quarks of an incident nucleus .
contrary to the skyrmion type fields for which the isotopic direction of the vector is strongly correlated with the coordinate space position ( so called `` hedgehog ansatz '' ) here the direction is described by the single vector @xmath18 at the whole space just as it is for anselm solution or dcc .
therefore the distribution ( [ sqrtf ] ) for the ratio of charged and neutral particles is still valid for the treated fields .
the structure and physical nature of these solutions is rather similar to electromagnetic ( coulomb ) field of a charged particle , however the nonlinear pion interaction ( [ leff ] ) allows for the existence of several solutions with different energies .
the nucleon or nucleus in the cloud of such fields can be regarded as an excited state or resonance . below we shall consider the spherically symmetric solutions at first in the chiral limit , putting the pion mass @xmath21 , and then we shall add the mass term to the lagrangian ( [ leff ] ) .
the mass term looks like the source in the classical equation and gives rise to the piece of the form @xmath22 , where @xmath23 .
the contribution of the external source is opposite in sign to the mass term . for a weak fields @xmath24 the source acts as an attractive potential and allows thereby for a stable solution
a stationary state . even for a strong field
@xmath25 the stationary state is possible in the case of large size source .
once this classical solution is created in the heavy ion collision ( if created ) it can live for the long time slowly decaying through the emission of very soft pions ( in the nucleus reference frame ) .
we describe the classical equation for the fields surrounding the nucleon and possible excitations ( resonances ) of this system in section 2 .
the more interesting case of a large nucleus is considered in section 3 .
the pion interaction with fermions ( quarks or nucleons ) is given by the microscopic lagrangian @xmath26 further we shall treat it as an interaction with external classical source generally described by isoscalar @xmath27 and isovector @xmath28 densities . for the simplicity we shall deal with the situation when the source has zero isospin , so that @xmath29 , and the lagrangian takes the form @xmath30 the last term being zero for @xmath31 . upon varying the lagrangian with respect to the pion matrix @xmath32 , the equation of motion yields @xmath33\,=\,\frac g{2 f_\pi}\ , \rho(x)\,\bigl[u^+- u\bigr].\ ] ] in what follows the matrix @xmath34 is sought among the kind of exact solutions found in refs.@xcite : @xmath35 where @xmath36 is an arbitrary but constant unitary matrix .
the function @xmath37 obeys the equation @xmath38 we are interested in the stationary solutions for which @xmath39 where @xmath40 is the laplace operator . supposing the function @xmath37 to be decreasing at the space infinity this equation can be transformed to the nonlinear integral form @xmath41 the total energy accumulated in the pion field is generally finite : @xmath42.\ ] ] there are two different regimes for the equation ( [ e ] ) - the regime of small and large size of the source . in the first case the density is concentrated in the small space , and since we are looking for a smooth function we can put @xmath43 in the integrand ( [ int ] ) .
then we obtain the coulomb - like potential at the finite distances from the source @xmath44 with the charge - like constant @xmath45 obeying the equation @xmath46 in which the dimensionless parameter @xmath47 , @xmath48 is proportional to the radius of the source .
the equation ( [ fsinf ] ) has a set of solution for @xmath49 @xmath50.\ ] ] the lesser the effective dimensionless radius of the source , @xmath47 , the more energy levels appear .
for @xmath51 they form two quasi - continuous bands for even and odd states approximated for not very large @xmath52 as @xmath53,\ ] ] with energy spacing @xmath54 between the levels in the each band . the lower levels in the odd band in ( [ band ] ) have the negative energy .
one has to emphasize here that @xmath55 are only the pion field energies whereas the total energy comprises , except this , the energy of the source itself , that is the energy of the quarks or nucleons inside .
the smaller the volume they occupy the large energy they have .
therefore there is a competition between these two effects which , could lead under certain conditions to a bound state formation . in the second regime
there are no finite energy solutions except for the @xmath56 .
indeed , after rescaling the variables @xmath57 , @xmath58 , where @xmath59 is a characteristic size , where the density is non zero , the equation ( [ int ] ) reads @xmath60 so only the zero solution is allowed for large enough @xmath59 .
the simple estimation shows that the critical radius @xmath59 needed for the nontrivial solution should be smaller than 1 fm in the order of magnitude .
the energy @xmath61 for @xmath62 fm is of the order of the mass difference between the baryon resonances and it is not excluded that some / part of this states are due to an excitation of the `` coulomb - like '' pion field considered above .
the last statement on the absence of the solution for the large source is valid only for the fixed constant @xmath63 while the value of @xmath64 increases for a heavy ion with the atomic number @xmath65 faster than the characteristic radius @xmath66 .
this is an interesting case which should be discussed in more detail .
let us put the nucleon density @xmath67 inside the nuclear and include the pion mass .
it results into the equation @xmath68 where the density @xmath69 and @xmath70 is the nuclear radius . for the normal nuclear density the effective strength of the source @xmath71 @xmath72 . for the weak fields
the equation can be linearized and takes the simple form @xmath73 the well known spherically symmetric solution reads @xmath74 with @xmath75 and @xmath76 in order to match the solutions at @xmath77 the logarithmic derivative of the first expression should obey the condition @xmath78 the fine tuning of nuclear _ radius _ @xmath70 , which is needed to satisfy ( [ n4 ] ) , looks , of course , rather unlikely .
however it is possible to provide the matching in non - linear case by choosing the value of @xmath79 ( for large enough @xmath80 ) .
we find such solutions numerically for a reasonable ion radius @xmath81 - @xmath82 fm with the field amplitudes @xmath83 - @xmath84 , respectively .
the results are shown in fig.1 by the solid curves . to calculate the energy of the pion field we include the pion mass term @xmath85 in ( [ e ] ) and obtain a very small value @xmath86 mev for @xmath87
( this is still a weak field ) but @xmath88 mev for the case of @xmath89 .
the fields shown by the solid curves in fig.1 look as the result of the pion condensation in a heavy nuclear which was studied many years ago @xcite mainly in terms of the fermy liquid theory . a possibility to observe the pion condensation in heavy ion collisions is discussed in @xcite .
the lagrangian ( [ lsq ] ) is correct being written in terms of the quarks and pions .
we have applied it to the nucleon instead of the quarks . taking the value @xmath91 we assume the whole nucleon mass to be generated by the interaction of the nucleon with the classical @xmath3 field @xmath92 . in the framework of the effective lagrangian ( [ lpi]),([leff ] )
it leads to a very strong interaction of the soft pions with the nucleons . on the other hand
the @xmath93 scattering amplitude is rather small corresponding to a so - called @xmath3-term , which is about 20 - 30 times less than the proton mass .
that is why we take a smaller value of the coupling @xmath63 ( say , @xmath94 @xmath72 ) in the better agreement with the nucleon @xmath3-term .
then a heavy ion acts as a rather weak source . for @xmath95 @xmath72 and a reasonable ion size ( @xmath96 fm )
the resulting attractive potential is even not sufficiently strong to form a bound state .
however it is possible to form for @xmath94 @xmath72 the quasistationary long - living state , which decays only due to the non - linear effects .
indeed , in a weak field approximation one can find the time dependent solution in the form @xmath97 , where the function @xmath37 satisfies eq.([n2 ] ) with @xmath98 replaced by @xmath99 . choosing the eigenenergy @xmath100 mev to provide the matching at @xmath77 we obtain the solution shown in fig.1 by the dotted curve ( for @xmath101 fm ) .
being produced in a heavy ion collision ( if it is ) it can live for the long time slowly decaying through the emission of very soft pions ( in the nucleus reference frame ) due to non - linear terms in the lagrangian ( [ leff ] ) .
the life - time of the solution decreases with the field amplitude @xmath102 , which determines the multiplicity @xmath103 of the pions in the classical field .
the dotted curve presented in fig.1 corresponds to @xmath104 the existence of such a solution may help in production of a `` breather '' like classical pion fields @xcite .
interaction with the source ( nuclear matter ) makes this solution more stable and long - lived .
the presence of the source could also stabilize under certain conditions the solutions of `` pion string '' kind @xcite , for which the direction of the isotopic vector @xmath18 is correlated with the azimuthal angle in the coordinate @xmath105 plane .
nevertheless the production of the classical pion field we discussed first ( solid curves in fig.1 ) can not be completely excluded in the high energy heavy ion collisions . if a kind of `` quark bag '' ( large drop of quark plasma ) would be formed after collision instead of the outgoing `` nuclear ''
, then we , probably , come back to the _ quark_-pion lagrangian ( [ lsq ] ) with a large coupling @xmath63 and the corresponding value @xmath106
the interaction now would be enough to form a stationary classical pion field ( pion condensate ) in the drop of quark matter .
* acknowledgments * + one of us ( a.s . ) is grateful to prof .
k. goeke for hospitality and to prof .
l. mclerran for discussions . +
this work is supported by grants dfg - rffi-436-m3/540 and rfbr 98 - 02 - 17636 .
99 a.a.anselm , phys.lett . *
b217 * ( 1989 ) 169 .
a.a.anselm , m.bander , jetp.lett .
* 59 * ( 1994 ) 503 .
j.d.bjorken , int.j.mod.phys .
* a7 * ( 1992 ) 4189 ; acta phys.pol . *
b23 * ( 1992 ) 561 .
p.blaizot , a.krzywicki , phys.rev . *
d46 * ( 1992 ) 246 .
k.l.kowalski , c.c.taylor , cwruth-92 - 6 , hep - ph/9211289 .
j.d.bjorken , k.l.kowalski , c.c.taylor , slac - pub-6109 ( 1993 ) .
k.rajagopal , f.wilczek , nucl.phys . *
b399 * ( 1992 ) 395 .
j.d.bjorken et al . ,
minimax : a revised proposal for t-864 , april 1993 .
p.v.ruuskanen , z.phys . *
c38 * ( 1988 ) 219 ; acta phys.pol .
* b18 * ( 1987 ) 551 .
s.gavin , a.gocksch , r.d.pisarski , phys.rev.lett .
* 72 * ( 1994 ) 2143 .
s.gavin , b.mller , phys.lett . *
b329 * ( 1994 ) 486 .
ian.i.kogan , jetp lett .
* 59 * ( 1994 ) 307 .
a.a.anselm , m.g.ryskin , a.g.shuvaev .
problems of high energy physics , international winter school on theoretical high energy physics , pnpi , st.petersburg , 1995 ; z.phys * a354 * ( 1996 ) 333 .
i.v.andreev , jetp lett .
* 33 * ( 1981 ) 367 .
minimax collab . : t.c.brooks et al .
, hep - ex/9906026 .
wa98 collab .
: m.m.aggarwal et al . ,
* b420 * ( 1998 ) 169 .
a.b.migdal , rev .
* 50 * ( 1978 ) 107 , + a.b.migdal , e.e.saperstein , m.a.troitskii , d.n.voskresensky , phys
* 192 * ( 1990 ) 179 .
h.j.pirner , d.n.voskresensky , phys . lett .
* b343 * ( 1995 ) 25 .
v.a.gani , a.e.kudryavtsev , t.i .
belova , b.l.druzhinin , phys . at .
* 62 * ( 1999 ) 895 .
x.zhang , t.huang and r.h.brandenberger , phys . rev .
* d58 * ( 1998 ) 027702 ; hep - ph/9711452 .
|
classical pion field similar to disoriented chiral condensate ( dcc ) is considered in the presence of the external source .
this field is similar to dcc in the sense that its isotopic orientation is specified with a single vector at the whole space .
we study the classical field solutions in the nonlinear sigma - model both in the chiral limit with massless pion and for the finite pion mass . in both cases
the field resembles the coulomb field of charged particle however the nonlinear pion interactions lead to the existence of several solutions . in the massless case and for the very small size of the source
there is the lot of classical solutions with finite discrete energies . in the more realistic situation of large nucleus ( heavy ion )
there are no stable solutions of the above type , but there is the possibility for the formation of the quasistationary states
. they can live for a long time slowly decaying through the emission of very soft pions .
the structure and the energies of these solutions is investigated numerically .
= 0.7 cm epsf 1.5 truecm * classical pion fields in the presence of source * + m.g.ryskin and a.g.shuvaev + theory department , st.petersburg nuclear physics institute + 188350 , gatchina , st.petersburg , russia . + e - mail : [email protected] + e - mail : [email protected]
|
in recent years graphics processors ( gpus ) have gained widespread acceptance as an architecture useful for scientific computing . in many application domains their use brings manyfold increase in performance @xcite .
most often gpus are used for computational kernels , the most computationally intensive parts of applications . in the finite element method ( fem )
domain , the first cited case of using gpus is @xcite where finite elements are applied in interactive visualisation .
other examples include implementation for gpu clusters of higher order fem approximations in earthquake modelling and wave propagation problems @xcite or gpu implementations of some variants of discontinuous galerkin approximation @xcite .
the finite element solution process can be divided into two parts : creation of the matrix for the system of linear equations , based on a suitable weak statement of the approximated problem and the solution of the system of equations ( or some update of the vector of global degrees of freedom , based on the created matrix or several matrices like e.g. for certain time stepping schemes and the right hand side vector ) .
the two parts can be implemented as different software components .
the second component , the solver of linear equations , whose implementation can often neglect numerical and computational details associated with finite element approximation , is more frequently investigated in the context of high performance computing and especially gpu acceleration ( see e.g. @xcite ) . sometimes the main stress is put on computational kernels of linear solvers , particularly linear algebra operations , that are properly optimized @xcite . as a justification for such an approach
, one can use the fact that quite often the time required for the solution of the system of linear equations strongly dominates the time for the whole solution process .
this always happens in situations where weak statements employed are simple , degree of approximation is low and , moreover , the system of linear equations is ill - conditioned or non - definite , causing iterative solvers to converge slowly or enforcing the use of direct solvers .
however for complex weak formulations and high orders of approximation the time for creation of system of linear equations can be higher than the time for its solution .
the most time consuming part of the system creation is numerical integration of terms from the weak statement . because of that
, optimization of numerical integration routines has been for many years the subject of investigations in the higher order and spectral approximation communities @xcite . numerical integration in finite element codes
is used to create entries for element stiffness matrices that are further assembled into the global stiffness matrix , the matrix for the associated system of linear equations .
again this two processes can be investigated together or treated separately .
the problem of combined integration and assembly for particular applications ported to gpus has been described e.g. in @xcite for nonlinear elasticity and in @xcite for electromagnetics .
an extensive study presented in @xcite concerns 3d discontinuous galerkin approximations for hyperbolic problems .
higher order tetrahedral elements are considered , with the special focus on particular operators associated with explicit time stepping mechanism and dg discretization .
the approach presented in @xcite for dg problems follows a generally applicable technique that considers numerical integration and global assembly as steps of evaluation of fem operators @xcite .
this technique is important not only in the context of explicit time stepping methods , but also e.g. for the implementation of iterative solvers of linear equations and the application of boundary conditions from fem weak statements .
this approach has formed a basis for the attempts to create tools for automated generation of finite element codes @xcite .
the continuation of these investigations in the context of gpu computing was recently described in @xcite where different assembly techniques for finite element approximations are considered .
the performance of presented techniques is tested for 2d low order approximations .
opencl and cuda programming environments are compared and portability of opencl kernels between gpus and cpus is examined . in the similar context ( automated code generation and low order approximations ) the problem of finite element integration is presented in @xcite .
the implementation is tested for 2d and 3d elliptic problems ( laplace and elasticity operators ) and the performance obtained for gpus is reported .
the three mentioned above papers suggest automatic generation of optimised code for each processor architecture , based on some abstract specification of computations , as a solution to an important implementation problem of the portability of software in face of diversity of multi - core architectures . the authors confirm the observations from other application domains @xcite , that also in the context of finite element assembly , even in the case of the same programming model ( e.g. opencl ) employed for different architectures , in order to obtain the best performance , it is necessary to produce separate versions of procedures for each architecture .
thorough investigations of the finite element assembly process on gpus were presented in @xcite . in the article several methods of assembly , that imply also numerical integration techniques , were shown .
the performance of the proposed techniques was tested for scalar problems and non - adaptive 2d meshes .
the approach presented in @xcite was extended in @xcite to 3d problems , however for first order polynomials only .
nevertheless , the techniques presented in the papers can form the basis for various implementation strategies of the assembly process , also in the case where numerical integration is performed separately and different types of approximation are considered .
there may be several reasons for considering numerical integration separately from the global assembly .
first , the solver employed for the solution of the system of linear equations may perform itself the conversion of element matrices to some specific sparse format of the global matrix @xcite .
second , the assembly may be performed by a different hardware component ( such an approach for large scale problems has been adopted e.g. in @xcite ) . and
finally , considering numerical integration independently of the global assembly gives software developers more flexibility in designing the final solution strategy . as an additional argument one may consider also the relative importance of both operations .
the assembly of an element matrix to the global stiffness matrix involves three memory references and one floating point operation . for higher order approximations that we consider in the paper , computing a single entry in the element matrix may require , as we show in detail later in the paper , hundreds of memory accesses and floating point operations . even considering that for numerical integration the references
can be realized using faster local memories without data races , still the proper implementation of numerical integration can be considered as having its own merits and importance , independent of the global assembly .
when porting computations to graphics processors , one encounters another specific problem the issue of precision .
some gpus do not support double precision arguments , some perform double precision operations much slower than single precision .
it has been proven in many studies that the solution of linear equations , especially for large scale problems , requires the use of double precision unknowns , because of accumulation of round - off errors . when investigating the implementation of that stage of finite element solution process , the solution might be to select the phase of calculations or particular operations that can be performed in single precision and after the single precison phase perform improvement of results @xcite . sometimes ,
when the solution process does not require the solution of the system of linear equations , due to the requirements of a particular application , the whole finite element simulation process can be performed in single precision .
this can happen for example in the case of explicit algorithms for time dependent problems @xcite .
even in the case of performing the stage of linear system solution , numerical integration can be performed in single precision .
it forms a part of the whole approximation process and has to satisfy specific requirements concerning accuracy , in order to guarantee the convergence of finite element solutions @xcite . for many problems , due to the approximate character of numerical integration itself ,
performing calculations in single precision does not introduce errors large enough to be comparable to discretization errors @xcite .
in such cases numerical integration can be performed in single precision and the resulting entries to element stiffness matrices or the global stiffness matrix can be promoted to double precision before the solution of the system of equations .
for some problems , e.g. for problems with terms of different orders that scale in a different way , double precision is required also for numerical integration @xcite . however , this can be considered a special case , while general purpose calculations should be performed in single precision , due to shorter execution times and lower memory requirements .
the advantages of single precision computations can be especially pronounced for higher order calculations .
we consider the problem of numerical integration for higher order 3d finite element approximations .
the main feature of the problem that we address is the large variation of required resources , processing time and memory , for different orders of approximation , i.e. different degrees of polynomials employed as finite element shape functions .
we investigate the characteristics of the problem and try to design a portable kernel for different problems and different processor architectures .
hence , we restrict ourselves to opencl programming model .
it allows us to construct kernels for graphics processors supporting the model ( currently nvidia , amd and intel hardware ) , as well as for cpus .
our previous investigations @xcite as well as results reported in other papers ( e.g. @xcite ) show that opencl implementations achieve performance only several percent lower than native cuda model for nvidia gpus .
we situate our research in the context of creating a general purpose finite element framework modfem @xcite , for continuous and discontinuous galerkin approximations .
the code has modular structure in which different software components are responsible for different phases of finite element calculations @xcite .
each component can exist in several variants and final applications for different problems and solution strategies are created using suitable compilation options .
the code has also a layered structure with separate modules responsible for shared memory computations , that form a core layer and that are further extended to distributed memory environments by suitable overlays @xcite , belonging to a different layer .
the code operates on unstructured adaptive meshes with proper load balancing for parallel execution @xcite .
the code has abilities to run on clusters with message passing and is currently ported to different new computer architectures , such as gpus and heterogeneous processors , like cbe @xcite . in each particular code created by modfem framework
there is a separate approximation module that is responsible for the creation of element stiffness matrices .
the matrices are further transferred to linear equations solvers using a generic interface @xcite .
thanks to the layered structure of the code there is only one numerical integration routine , in the core layer , for both , shared and distributed memory environments .
the interface between numerical integration procedure and the solver of linear equations allows for employing different , direct and iterative , solvers of linear equations , possibly using thin adapters @xcite .
the same integration routine is also used for different types of meshes and problems solved .
this approach proved successful in many application domains @xcite . as a main problem of the current investigations we pose the problem whether it possible to maintain this portability and genericity by creating a single kernel executed on gpus for different problems , meshes and computer architectures .
we consider numerical integration only , leaving the problem of assembly ( and also application of boundary conditions ) to different software and possibly hardware components . in our view
this smaller grain of considered gpu code gives more flexibility in designing solution strategies , where cpu and gpu cores can work together , each group performing the tasks it is most suitable for @xcite . in order to maximize this flexibility of the approach we consider different options for passing the results of calculations .
the algorithms presented in the paper do not prevent the use of different assembly techniques , such as e.g. presented in the discussed above studies devoted to the subject .
we attempt to investigate the possibility of creating a general purpose numerical integration procedure executed on graphics processors that can either replace or complement the cpu integration routine .
we do not attempt to create gpu - only code .
considering relatively small sizes of memory for current accelerators with gpus , in the order of several gbytes , this would limit the sizes of problems that we solve . since we aim at large scale problems we want to use large memories , in the range of hundreds of gbytes , for each shared memory node .
the use of gpus as accelerators for cpus , corresponds also to our distributed memory strategy where we assume two separate levels of decomposition .
gpu calculations fit into shared memory level calculations controlled by cpu code and due to the design philosophy can be directly used in distributed memory calculations , based on cpu code and its message passing overlays . following the arguments presented above
, we assume that our general purpose finite element numerical integration procedure does not require double precision calculations and we restrict our interest to single precision implementations .
we plan to consider extensions to double precision for particular application domains that would pose such requirements .
the investigations presented in the paper form the continuation of an earlier work . in @xcite we presented numerical integration on gpus for 2d linear elasticity , cuda environment and older generations of gpus .
this work was extended to opencl programming model in @xcite and to 3d laplace operators in @xcite .
we investigate separately numerical integration procedures for cbe processors in @xcite and @xcite .
the paper is organised in the following way : * section [ problem_formulation ] defines the problem of numerical integration for finite element approximations * section [ programming_model ] describes the programming model that we use and its mapping to selected gpu architectures * section [ parallel_implementation ] contains details of several versions of numerical integration procedures for gpus that we developed * section [ computational_tests ] presents results of tests performed to assess the performance of the code * section [ conclusions ] contains final remarks .
in the current section we define the problem of numerical integration for which we will further design gpu kernels . we restrict ourselves to defining the notation that we use and computational aspects of the problem . for more information on numerical aspects of integration
we refer to the works presented in the introduction and for the whole context of higher order finite element approximation to any of the excellent textbooks on the subject ( e.g. @xcite , @xcite ) .
3d finite element numerical integration and assembly consist of a loop over elements and faces and possibly some other mesh entities in order to create entries to the global stiffness matrix and the global right hand side vector .
the creation can involve different operations
integration over elements , integration over faces and other forms of application of boundary conditions , application of other constraints , etc .
the most time consuming part , especially for higher orders of approximation , is the calculation of element integrals , and following the discussion presented in the introduction , we restrict our attention to this process . for numerical integration
we use quadratures with the order determined by the order of approximation and the characteristics of the problem solved .
integration is always performed on a reference element , following the suitable change of integration domain from a real element in physical space to the reference element . within a 3d reference element
quadrature points have coordinates @xmath0 : @xmath1[i_s];\ ; \ ; i_q = 1 , 2 , ... , n_q ; \ ; i_s = 1 , 2 , 3;\ ] ] and weights @xmath2 @xmath3 , i_q = 1 , 2 , ... , n_q\ ] ] with @xmath4 denoting the number of quadrature points .
computed integrals correspond to different terms in the weak statement of the problem considered .
the problem dependent contribution consist in defining a set of coefficients for numerical integration .
there is a different coefficient for each combination of indices , that we denote by @xmath5 and @xmath6 , corresponding to different spatial derivatives for test and trial functions respectively , including terms with functions itself , that we associate with @xmath7 and @xmath8 .
moreover , for vector problems , the same approximation may be used for different vector components ( different unknowns in the solved system of pdes ) and hence for each combination of indices @xmath5 and @xmath6 there may be a small matrix of coefficients with the dimension @xmath9 equal to the number of equations in the solved system of pdes .
in such a case the array of coefficients may be defined as : @xmath10[j_{e}][i_d][j_d];\ ; \ ; i_{e},j_{e } = 1 , 2 , ... , n_{e } ; \ ; i_d , j_d = 0,1,2,3 ; \;\ ] ] in the most general , non - linear or quasi - linear case there may be different values of coefficients at each integration point .
hence for the generic numerical integration algorithm we have to consider an array of coefficients of the form : @xmath11[i_{e}][j_{e}][i_d][j_d];\ ; \ ; i_q = 1 , 2 , ... , n_q ; i_{e},j_{e } = 1 , 2 , ... , n_{e } ; i_d , j_d = 0,1,2,3\ ] ] as it was stated above , the integration of terms from the finite element weak statement is performed using the change of variables , from a real element to a reference element , of one of several possible types ( e.g. tetrahedron , prism , cube , to name the most popular ) .
hence the derivatives with respect to physical coordinates , that appear in the weak statement , have to be computed by the chain rule , using derivatives of trial and test functions with respect to reference coordinates .
these calculations involve the jacobian matrix of the transformation from the real element to the reference element .
the matrix is usually obtained by inverting the matrix of its inverse transformation , the transformation from the reference to the real element .
this last transformation can be easily obtained from the description of geometry of real elements .
the jacobian matrix of the transformation from the reference to the real element is also used for computing the volume element in integrals , and later in the paper when the notion of the jacobian matrix is used it denotes this matrix . the transformation from the reference to the real element
is performed using the representation of the geometry of the real element .
different options are possible here , with the simplest of defining linear or multi - linear elements , whose geometry is determined by the position of their vertices .
more complex cases include popular isoparametric elements , where the geometry is specified using polynomials of the same order as the solution @xcite and recently introduced isogeometric analysis , where elements are geometrically represented using spline approximations @xcite .
element stiffness matrices are computed for finite element approximations to trial and test functions , represented as linear combinations of finite element basis functions , that in turn are constructed using element shape functions .
the number of shape functions for an element , @xmath12 ( equal to the number of local degrees of freedom for the element ) is related to the order of approximation and possible particular choices of approximation . in order to present concisely the numerical integration problem we denote shape functions and their derivatives by a single array @xmath13[i_{dof}];\ ; \ ; i_d = 0,1,2,3 ; \ ; i_{dof } = 1 , 2 , ... , n_{\mathrm{sh}}\ ] ] where for the first index , its value 0 refers to shape functions and the values 1,2,3 refer to derivatives with respect to local , reference element coordinates , @xmath14 , and the second index corresponds to different degrees of freedom . in order to compute integrals we use the values of shape functions and their derivatives at all integration points : @xmath15[i_d][i_{dof } ] ; \ ; \ ; i_q = 1 , 2 , ... , n_q ; \ ; i_d = 0,1,2,3 ; \ ; i_{dof } = 1 , 2 , ... , n_{\mathrm{sh}};\ ] ] these values are the same for all real elements of a given approximation order . in calculations of terms from the weak statement
we use the values of shape functions and their derivatives with respect to physical coordinates .
these values will be denoted by : @xmath16[i_d][i_{dof}];\ ; \ ; i_q = 1 , 2 , ... , n_q ; \ ; i_d = 0,1,2,3;\ ; i_{dof } = 1 , 2 , ... , n_{\mathrm{sh}};\ ] ] the derivatives of shape functions with respect to physical coordinates are different for each element , since they are computed using the entries of the jacobian matrix of the transformation from the real element to the reference element , that are different for each element . the jacobian matrix of the transformation from the real element to the reference element is also used for computing derivatives of the solution at previous non - linear iterations and/or previous time steps .
this derivatives ( and values of the solution itself ) are used in non - linear and time dependent problems to compute problem dependent coefficients @xmath17 . the finite element numerical integration problem for a single element can be defined in the following way : given a set of data defining the geometry of an element , a set of degrees of freedom corresponding to solutions at previous non - linear iterations and/or previous time steps and a set of data specifying problem dependent input to the integration procedure create an element stiffness matrix @xmath18[j_{e}][i_{dof}][j_{dof}];\ ; \ ; i_{e},j_{e } = 1 , 2 , ... , n_{e};\ ; i_{dof},j_{dof } = 1 , 2 , ... , n_{\mathrm{sh}};\ ] ] corresponding to the finite element formulation of the problem solved .
read problem dependent coefficients , @xmath17 read geometry data for element possibly read old
element degrees of freedom from previous iterations / time steps initialize element stiffness matrix , @xmath19 prepare quadrature data , @xmath0 and @xmath2 read values of shape functions and their derivatives with respect to local element coordinates , @xmath20 $ ] read or calculate jacobian matrix , its determinant ( * det * ) and its inverse calculate derivatives of shape functions with respect to physical coordinates , @xmath21 $ ] based on @xmath22 $ ] and old degrees of freedom calculate coefficients at quadrature point , @xmath23 $ ] @xmath24[j_{e}][i_{dof}][j_{dof}]+=$ ] @xmath25*det*@xmath26 \times { \mbox{\boldmath $ c$}}^q[i_{e}][j_{e}][i_d][j_d][i_q ] \times$ ] @xmath27[i_{dof}][i_q ] \times { \mbox{\boldmath $ \psi$}}^q[j_d][j_{dof}][i_q]$ ] we conclude the definition of the problem by providing algorithm [ num_int_simple ] , a generic procedure that calculates the entries to the element stiffness matrix .
the procedure corresponds to mathematical expressions involved in numerical integration , indicating necessary summations . from the mathematical point of view
, the order of performing the three outermost loops of algorithm [ num_int_simple ] is irrelevant ( but implies some changes in details of the algorithm ) . in our implementation
we follow the common practice from classical finite element codes , corresponding directly to algorithm [ num_int_simple ] , where the loop over integration points is the outermost loop .
thanks to this we do not have to store in some fast memory the values of all shape functions at all integration points , instead we need only quickly accessible storage for the values of shape functions and their derivatives at a single integration point .
the other option , with precomputed values of all shape functions at all integration points and the loops over degrees of freedom ( shape functions ) as the outermost loops , is also possible , especially for low order 2d approximations @xcite . in practice ,
calculations are often performed in a way that is different that specified by algorithm [ num_int_simple ] .
the four innermost loops are small , with the ranges for @xmath5 and @xmath6 equal 4 and usually less or equal 5 for @xmath28 ( 3 for elasticity , 4 for incompressible navier - stokes , 5 for compressible navier - stokes ) .
this suggests to manually unroll all the loops .
moreover arrays @xmath17 are usually sparse and substantial savings can be obtained , when instead of performing the four innermost loops from algorithm [ num_int_simple ] , suitable , optimized calculations are performed for a block of entries of the matrix @xmath19 associated with a single pair @xmath29 .
this leads to algorithm [ num_int_real ] , that we further implement for graphics processors .
read problem dependent coefficients , @xmath17 read geometry data for element possibly read old
element degrees of freedom from previous iterations / time steps initialize element stiffness matrix , @xmath19 prepare quadrature data , @xmath0 and @xmath2 read or calculate values of shape functions and their derivatives with respect to local element coordinates , @xmath20 $ ] read or calculate jacobian matrix , its determinant ( * det * ) and its inverse calculate derivatives of shape functions with respect to physical coordinates , @xmath21 $ ] based on @xmath22 $ ] and old degrees of freedom calculate coefficients at quadrature point , @xmath23 $ ] update a block of entries of @xmath19 associated with a pair @xmath29 in a manner implied by the structure of non - zero entries of the array of coefficients @xmath17 , multiplying each non - zero entry of @xmath30 associated with a pair @xmath31 by @xmath25*det*@xmath32 \times { \mbox{\boldmath $ \psi$}}^q[i_d][i_{dof}][i_q ] \times { \mbox{\boldmath $ \psi$}}^q[j_d][j_{dof}][i_q]$ ] and performing suitable summations over ranges of @xmath5 and @xmath6 to assess the particular character of the problem of numerical integration for 3d approximations with different orders of approximation @xmath33 , we present in table [ tab_1 ] the sizes of the arrays involved in calculations for an example 3d reference element the prism with shape functions being products of polynomials from a set of complete polynomials of a given order for triangular bases and 1d polynomials associated with vertical direction . as an example quadrature , the most popular in finite element codes , gaussian quadrature is selected .
as can be seen from the table , the number of times the innermost calculations are performed for @xmath34 equals 27,869,184 .
that is one of reasons why for higher orders of approximation special techniques for integration are designed @xcite .
the order from which it becomes advantageous to switch to different techniques that presented in the current paper depends on the characteristics of the problem statement and the computing environment .
the value of 7 that we choose for our study is to certain extent arbitrary ( in some cases it can be profitable to switch to different techniques of integration even from orders in the range 4 - 5 ) .
nevertheless , for the rest of the paper we consider for our higher order finite element approximations the range of polynomial degrees @xmath33 from 2 to 7 .
.parameters determining the computational characteristics of 3d finite element numerical integration the number of shape functions and the number of gaussian integration points for the standard prismatic element and different degrees of approximating polynomials [ cols= " < , > , > , > , > , > , > , > " , ] time for : & p=2 & p=3 & p=4 & p=5 & p=6 & p=7 + & 0.06 & 0.08 & 0.1 & 0.7 & 2.9 & 15 + & 2.00 & 5.81 & 7.4 & 14.5 & 24.7 & 46 + & 11.40 & 60.20 & 179.5 & 481.8 & 1144.0 & 2457 + & 0.07 & 0.11 & 0.2 & 0.7 & 3.6 & 11 + & 0.40 & 1.06 & 1.9 & 4.4 & 11.3 & 22 + & 11.18 & 60.12 & 165.2 & 478.8 & 1137.0 & 2439 + & & & & & & + & 0.44 & 6.27 & 29.22 & 163.0 & 614.8 & 1981 + & 2.10 & 27.16 & 124.0 & 679.3 & 3010 & 9894 + & 208 & 230 & 235 & 240 & 204 & 200 + & 24.61 & 147.6 & 469.1 & 1641 & 5298 & 14817 + & & & & & & + & 0.44 & 6.30 & 29.37 & 163.9 & 617.9 & 1990 + & 1.21 & 17.80 & 80.07 & 439.9 & 2117 & 7203 + & 366 & 354 & 366 & 372 & 291 & 276 + & 25.99 & 149.1 & 442.0 & 1433 & 4458 & 12213 + & & & & & & + & 0.43 & 6.23 & 31.42 & 158.9 & 590.5 & 1868 + & 1.91 & 23.06 & 114.2 & 477.4 & 2318 & 7357 + & 228 & 270 & 274 & 274 & 254 & 253 + & 24.41 & 143.5 & 459.4 & 1439 & 4605 & 12280 + & & & & & & + & 0.44 & 6.25 & 33.47 & 159.3 & 592.0 & 1873 + & 1.30 & 19.05 & 94.31 & 490.9 & 1932 & 6199 + & 338 & 328 & 334 & 324 & 306 & 302 + & 26.07 & 150.4 & 456.2 & 1484 & 4273 & 11208 + time for : & p=2 & p=3 & p=4 & p=5 & p=6 & p=7 + & 0.03 & 0.04 & 0.2 & 1.58 & 4.8 & 22 + & 1.20 & 3.21 & 5.4 & 11.18 & 17.1 & 52 + & 7.50 & 37.12 & 131.5 & 390.4 & 902.9 & 1995 + & 0.10 & 0.41 & 1.7 & 6.2 & 13.6 & 31 + & 0.58 & 1.74 & 4.1 & 16.9 & 27.9 & 60 + & 2.41 & 12.19 & 42.5 & 162.8 & 318.4 & 639 + & & & & & & + & 0.43 & 5.56 & 30.05 & 168.6 & 659.0 & 2206 + & 3.13 & 53.73 & 325.3 & 1779 & 9496 & 34568 + & 139 & 103 & 92 & 94 & 69 & 63 + & 10.77 & 91.32 & 458.9 & 2177 & 10418 & 36618
+ & & & & & & + & 0.44 & 5.62 & 30.36 & 170.2 & 665.1 & 2226 + & 4.06 & 34.61 & 190.8 & 985 & 4684 & 17094 + & 109 & 162 & 159 & 172 & 141 & 130 + & 13.35 & 76.68 & 332.0 & 1404 & 5632 & 19202 + & & & & & & + & 0.43 & 7.02 & 31.57 & 157.8 & 602.5 & 1906 + & 3.90 & 63.74 & 297.5 & 1477 & 6428 & 21359 + & 111 & 110 & 106 & 106 & 93 & 89 + & 11.54 & 101.3 & 431.0 & 1876 & 7350 & 23408 + & & & & & & + & 0.44 & 7.05 & 24.59 & 158.3 & 604.6 & 1912 + & 5.33 & 67.90 & 305.9 & 1519 & 6152 & 20742 + & 82 & 103 & 103 & 104 & 98 & 92 + & 14.62 & 109.9 & 447.1 & 1937 & 7101 & 22850 + * each presented table contains timing results for subsequent stages of the whole procedure of numerical integration .
this reflects the possibility of arranging the stages in different ways .
the initial stage is important when the first group of elements is sent for integration on gpu .
then , for next groups it is possible not to allocate the gpu memory and send only input data for processed elements . also the stage of copying back
the computed stiffness matrices can be eliminated in certain solution strategies , as we have mentioned already . *
the timing results are presented for a single finite element of a given order .
these results were obtained by running the kernels for a large number of elements ( specified for our test runs in tables [ tab_3 ] and [ tab_4 ] ) and dividing the times of performing different stages of execution by the number of elements .
it should be made clear that the results depend on the total number of elements , since the times for card initialization and data transfers to and from gpu memory does not scale linearly with the number of elements .
* the nonlinear dependence mentioned above can appear for low numbers of elements .
for sufficiently high numbers , the linear dependence should emerge and the results presented in tables [ tab_5][tab_7 ] should be applicable to different problem sizes by simple scaling . the linear scaling should be observed especially for higher orders of approximation where gpu calculations , that scales linearly , take more time than card initialization and data transfers .
the fact of linear scaling of the numerical integration algorithm with respect to the number of elements for sufficiently high numbers of elements has been confirmed experimentally in several studies @xcite . *
as it was mentioned already , for cpu cores and gpu compute units we parallelize the loop over elements .
if the processor resources are sufficient , than , thanks to the embarrassingly parallel character of the loop over elements in the numerical integrations algorithm , the performance of calculations should scale linearly ( even in the strong sense ) with the number of cores and compute units .
we observed such relations for different gpus with the nvidia tesla architecture when performing computational experiments reported in @xcite .
* it should be noted , that both types of scaling , with respect to the number of elements and with respect to the number of cores / compute units , are much more complex when assembly process is taken into account . in such a case also the dependence of performance results on a particular problem solved ( due to the different properties of the global stiffness matrix ) can be observed @xcite . in the case of numerical integration alone , the particular problem solved is irrelevant to the performance obtained .
* execution times were always measured on the host side .
we introduced synchronisation procedures for opencl events in the host code and measured elapsed times for each stage of calculations .
the reason for this was to explicitly indicate the overhead associated with the subsequent stages . in recent modifications to gpu programming models
it is possible to overlap different stages of execution for different kernel invocations and reduce the overheads @xcite . *
measuring elapsed time means adopting a user perspective .
we do it for particular stages of calculations and also for the total execution time of kernels .
the times measured on the host side were always longer than the times reported by execution environment profilers .
hence the performance that we report is the performance observed by the user , which is lower than the actual performance of the hardware .
* times for gpu initialization were relatively constant for all orders of approximation since they were mainly associated with creation of opencl memory objects in gpu memory and the total size of memory objects for each order was similar , due to the assumption that by changing the number of elements per kernel the whole available gpu memory is utilized . however , for each order of approximation the times per element are different , which is implied by the changing size of the stiffness matrix for a single element . * for computing performance in gflops , only the time for kernel execution ( measured on the host side )
was taken into account . * to obtain the total execution time , we measured , in an external procedure responsible in our code for global matrix assembly , the time spent in the host code and the integration kernels .
the main results from tables [ tab_5 ] , [ tab_6 ] and [ tab_7 ] are illustrated in figures [ times ] and [ performance ] from 2 to 7.,title="fig : " ] from 2 to 7.,title="fig : " ] from 2 to 7.,title="fig : " ] from 2 to 7.,title="fig : " ] from 2 to 7.,title="fig : " ] from 2 to 7.,title="fig : " ] there are several observations that can be made based on data in tables [ tab_5 ] , [ tab_6 ] and [ tab_7 ] : observation 1 .
: : the algorithm of numerical integration for higher order finite elements has the potential for high performance execution on massively multi - core architectures .
the performance obtained , above 200 gflops , and for some orders of approximation and some variants of the algorithm above 350 gflops for gtx580 and above 60 gflops with several results above 150 gflops for hd5870 , can be considered high , especially for nvidia where it is in the range of 12%-23.5% of the peak single precision floating point performance .
observation 2 .
: : the performance of gtx580 gpu is much higher than that of hd5870 , despite the fact that radeon hd5870 theoretical peak performance is almost twice as big as that of geforce gtx580 .
this can be explained to certain degree by higher overhead due to driver software for radeon @xcite and the fact that basic execution units of hd5870 are vector units , the fact that we do not exploit in our kernels .
observation 3 .
: : for gpus , the times for global memory initialization and data transfers from the host memory are comparable to the time of kernel execution ( for lower values of @xmath33 being even several times larger ) .
this , once again , here for finite element numerical integration algorithm , confirms the fact that slow pcie connection is an obstacle on a way to get higher performance for general purpose codes executed on gpus . for the algorithm of numerical integration , with increasing order of approximation the ratio of operations to memory transfers grows and hence the influence of low pcie performance diminishes .
observation 4 .
: : the times of kernel execution , i.e. the times for numerical integration calculations are always much lower for gpus than for the reference cpu , however , the total execution times , including memory transfers are better for gpus only in the case of higher orders of approximation .
observation 5 .
: : there is a large difference between the performance of the gpu as perceived by an external user and as exhibited during actual calculations .
one reason for this , data transfer times , was already mentioned .
the second is the fact , that , in order to ensure proper mapping of calculations onto the hardware , the gpu had to perform more operations than the cpu .
observation 6 .
: : there is no single kernel being the best for all situations ( different orders of approximation , different hardware ) , although the _ reg_nojac _ variant is the fastest in most of cases .
different kernels exhibit similar performance , with the best results obtained for medium orders of approximation .
observation 7 .
: : the option of sending to gpus , not only reference element shape functions , but also jacobian terms for all real processed elements and all integration points ( _ nojac _ versions ) turned out to be more efficient in most cases than the option of calculating the values on gpus ( _ jac _ variants ) . despite the fact that _ nojac _ versions require much longer times for transferring data to gpu memory than _ jac _ variants , this time remains much shorter than gpu initialization time and time for transfer of output data .
observation 8 .
: : _ shm _ variants are usually slower than _ reg _ versions , however we believe it is worth having them both , since for different gpu architectures this situation may change .
numerical integration for higher order 3d finite element approximations is a complex problem due to the large variation in required processor and memory resources .
thanks to a suitable parametrization of the designed opencl code for integration and a proper management of code execution by the created host side procedures , we have obtained the code that is portable across different gpu architectures and different orders of approximation , in the range of moderate degrees of shape function polynomials , from 2 to 7 . in the computational experiments reported in the paper we run the gpu kernels and the host managing procedures for all presented cases without changing a single line in the source code .
we obtained , for nvidia geforce gtx580 and amd radeon hd5870 graphics processors , the performance of calculations always above 200 gflops and 60 gflops , respectively .
this , especially for hd5870 , falls short to theoretical peak performance of the hardware , however still for many cases reported , the results can be considered satisfactory as for general purpose complex scientific calculations . in particular , for the case of gtx580 and one of the variants of numerical integration ( _ shm_nojac _ ) the performance for all orders of approximation exceeded 300 gflops ( 19% of theoretical peek ) and for another variant ( _ reg_nojac _ ) it reached for three orders ( @xmath33=2 , 4 and 5 ) more than 365 gflops , that is 23% of the theoretical maximum .
the main factor limiting the practical advantages of gpu calculations is the slow pcie connection between the gpu and the host computer . despite much higher gpu performance of calculations , the overall execution times for gpus and standard processors are comparable , because of long times devoted by gpus to memory initialization and data transfers . taking this into account
, it can be concluded that in the context of numerical integration for higher order finite elements , gpus should not be considered as a replacement for cpus , with all calculations off - loaded to gpus . in particular , when gpus are calculating domain integrals , cpus can do some other useful work , such as performing integration of boundary terms , integration of separate sets of elements or assembly of already created element stiffness matrices . in such scenarios ,
gpus can form valuable accelerators increasing substantially the overall performance of finite element codes .
r. strzodka , m. doggett , a. kolb , scientific computation for simulations on programmable graphics hardware , simulation modelling practice and theory , special issue : programmable graphics hardware 13 ( 8) ( 2005 ) 667680 .
d. komatitsch , d. micha , g. erlebacher , porting a high - order finite - element earthquake modeling application to nvidia graphics cards using cuda , journal of parallel and distributed computing 69 ( 5 ) ( 2009 ) 451460 .
d. komatitsch , g. erlebacher , d. gddeke , d. micha , high - order finite - element seismic wave propagation modeling with mpi on a large gpu cluster , journal of computational physics 229 ( 20 ) ( 2010 ) 7692 7714 .
d. gddeke , h. wobker , r. strzodka , j. mohd - yusof , p. mccormick , s. turek , co processor acceleration of an unmodified parallel solid mechanics code with feastgpu , international journal of computational science and engineering 4 ( 4 ) ( 2009 ) 254269 .
d. gddeke , r. strzodka , j. mohd - yusof , p. mccormick , s. h. buijssen , m. grajewski , s. turek , exploring weak scalability for fem calculations on a gpu - enhanced cluster , parallel computing 33 ( 10 - 11 ) ( 2007 ) 685699 .
v. volkov , j. w. demmel , benchmarking gpus to tune dense linear algebra , in : proceedings of the 2008 acm / ieee conference on supercomputing , sc 08 , ieee press , piscataway , nj , usa , 2008 , pp . 31:131:11 .
l. demkowicz , j. kurtz , d. pardo , m. paszynski , w. rachowicz , a. zdunek , computing with _ hp_-adaptive finite elements , vol .
2 : frontiers three dimensional elliptic and maxwell problems with applications , chapman & hall / crc , 2007 . p.
e. j. vos , s. j. sherwin , r. m. kirby , from h to p efficiently : implementing finite and spectral / hp element methods to achieve optimal performance for low- and high - order discretisations , j. comput .
229 ( 2010 ) 51615181 .
a. dziekonski , p. sypek , a. lamecki , m. mrozowski , http://dx.doi.org/10.1002/nme.4452[generation of large finite - element matrices on multiple graphics processors ] , international journal for numerical methods in engineering ( 2012 ) c. cecka , a. j. lew , e. darve , http://dx.doi.org/10.1002/nme.2989[assembly of finite element methods on graphics processors ] , international journal for numerical methods in engineering 85 ( 5 ) ( 2011 ) 640669 . c. cecka , a. j. lew , e. darve , application of assembly of finite element methods on graphics processors for real - time elastodynamics , in : w. mei w. hwu ( ed . ) , gpu computing gems
. jade edition , morgan kaufmann , 2011 , pp . 187205 .
g. r. markall , a. slemmer , d. a. ham , p. h. j. kelly , c. d. cantwell , s. j. sherwin , http://dx.doi.org/10.1002/fld.3648[finite element assembly strategies on multi - core and many - core architectures ] , international journal for numerical methods in fluids 71 ( 1 ) ( 2013 ) 8097 .
d. gddeke , r. strzodka , s. turek , performance and accuracy of hardware - oriented native- , emulated- and mixed - precision solvers in fem simulations , international journal of parallel , emergent and distributed systems 22 ( 4 ) ( 2007 ) 221256 .
m. baboulin , a. buttari , j. dongarra , j. kurzak , j. langou , j. langou , p. luszczek , s. tomov , accelerating scientific computations with mixed precision algorithms , computer physics communications 180 ( 12 ) ( 2009 ) 2526 2533 .
s. rul , h. vandierendonck , j. dhaene , k. de bosschere , an experimental study on performance portability of opencl kernels , in : application accelerators in high performance computing , 2010 symposium , papers , knoxville , tn , usa , 2010 , p.
k. bana , a modular design for parallel adaptive finite element computational kernels , in : m. bubak , g. van albada , p. sloot , j. dongarra ( eds . ) , computational science iccs 2004 , 4th international conference , krakw , poland , june 2004 , proceedings , part ii , vol .
3037 of lecture notes in computer science , springer , 2004 , pp . 155162 .
k. bana , a model for parallel adaptive finite element software , in : r. kornhuber , r. hoppe , j. priaux , o. pironneau , o.widlund , j. xu ( eds . ) , domain decomposition methods in science and engineering , vol .
40 of lecture notes in computational science and engineering , springer , 2004 , pp . 159166 .
k. bana , agent architecture for mesh based simulation systems , in : v. alexandrov , g. van albada , p. sloot , j. dongarra ( eds . ) , computational science iccs 2006 , 6th international conference , reading , uk , may 28 - 31 , 2006 , proceedings , part iii , vol .
3993 of lecture notes in computer science , springer , 2006 , pp . 743750 .
k. bana , parallelization of large scale adaptive finite element computations , in : r. wyrzykowski , j. dongarra , m. paprzycki , j. waniewski ( eds . ) , parallel processing and applied mathematics , proceedings of vth international conference , ppam 2003 , czstochowa , poland , 2003 , vol .
3019 of lecture notes in computer science , springer , 2004 , pp . 431438 .
f. kruel , k. bana , finite element numerical integration on powerxcell processors , in : ppam09 : proceedings of the 8th international conference on parallel processing and applied mathematics , springer - verlag , berlin , heidelberg , 2010 , pp .
517524 . p.
paszewski , p. macio , k. bana , finite element numerical integration on gpus , in : ppam09 : proceedings of the 8th international conference on parallel processing and applied mathematics , springer - verlag , berlin , heidelberg , 2010 , pp .
411420 . p. macio , k. bana , testing tesla architecture for scientific computing : the performance of matrix - vector product , in : proceedings of international multiconference on computer science and information technology , wisa , poland , october , 2008 , polish information processing society , 2008 , pp
. 285291 .
|
the paper considers the problem of implementation on graphics processors of numerical integration routines for higher order finite element approximations .
the design of suitable gpu kernels is investigated in the context of general purpose integration procedures , as well as particular example applications .
the most important characteristic of the problem investigated is the large variation of required processor and memory resources associated with different degrees of approximating polynomials .
the questions that we try to answer are whether it is possible to design a single integration kernel for different gpus and different orders of approximation and what performance can be expected in such a case .
finite element method , higher order approximation , numerical integration , graphics processors , gpu , opencl
|
it is very well - known that light propagation in the limit of geometric optics can be described by fermat s principle @xcite .
it states that the trajectories of rays can be obtained by the extremization of their optical paths @xcite .
geodesics in a spacetime are obtained likewise , by extremizing the distance between two given spacetime events .
this means that these two issues are intimately related ( @xcite , page 127 , and references therein ) and aspects raising in one area should be obtained in the other .
this must be specially the case for the spacetime metric .
indeed , it has been already shown how this geometric structure steering light propagation always arise in the context of nonlinear theories of the electromagnetism @xcite , some linear @xcite and nonlinear media @xcite , condensed matter ( @xcite and references therein ) , etc .
approaches where effective metrics arise are generically called analogue models @xcite ( mainly in the optical and acoustic scopes ) and they have been studied in a variety of scenarios , such as hawking radiation coming from analogue apparent horizons @xcite and analogue quantum gravity ( see e.g. @xcite and references therein ) , only to cite some .
it should be stressed that in these models the concept of optical metrics ( in this work we shall not focus on acoustic metrics ) , effective metrics for the propagation of light rays , is an emergent aspect .
once established , they can be regarded as fundamental quantities , for they directly describe photon trajectories .
this is to be contrasted with the metric of the background spacetime , which is a fundamental quantity present ab initio in the lagrangian density of interest , related to what optical metrics shall also depend upon . with the use of the equivalence principle
, optical geometries can be defined in any background spacetime .
the control of light trajectories can be approached , for instance , with transformation optics ( see ref .
@xcite for a comprehensive review and applications ) .
it is important since it takes advantage of the symmetries of maxwell equations in vacuum and in continuous media to make a link between coordinate transformations ( or geometries ) and dielectric coefficients @xcite .
nonetheless , this analogy is only valid when the permittivity and permeability tensors are equal , which is very restrictive in terms of generic dielectric tensors and , in general , can only be attained with metamaterials @xcite . in the context of nonlinear theories of the electromagnetism
, it is always possible to obtain effective metrics out of propagation of light rays .
this means that given a lagrangian density dependent upon the invariants of the electromagnetism , one can always `` geometrize '' the propagation of photons in terms of it ; see @xcite and references therein . nevertheless , this is not generally the case for light propagation in material media and for linear ones it is already known that a closure relation should be satisfied by the dielectric coefficients ( permittivity and permeability ) such that emergent optical metrics be dependent only upon spacetime coordinates @xcite . a generalization to the closure relation
does not yet exist for nonlinear media , in spite of the fact that optical geometries arise for some nonlinear isotropic liquids @xcite . without the closure relation an optical metric would generically depend upon properties of photons , such as their wave - vectors .
this would be of interest , for instance , to investigate extensions of riemannian geometries , like finsler models , shedding some light into quantum gravity phenomenology , due to modified energy - momentum dispersion relations , or the dark matter problem ( @xcite and references therein ) .
one aspect related to effective geometries that is not usually recalled is that , due to the fact they depend upon external electromagnetic fields and dielectric coefficients , for instance , in principle it is possible to conceive scenarios where they are very simple ( such as minkowskian or conformally minkowskian ) , even in curved background spacetimes , when the aforesaid fields and dielectric tensors could be controlled or presented convenient properties .
this leads precisely to the novelty of this work : to show the significance , physical reasonableness and nontriviality of the cases where optical metrics are rendered or naturally are simple .
such an analysis could have far - reaching implications for material sciences , charged black hole spacetimes , compact systems , etc . for instance , as we shall underline and exemplify , flat ( minkowski ) optical metrics for material media would lead to the practical absence of reflection and refraction of light rays coming from / going to an almost vacuous region ( since our analyses will be done in the limit of geometric optics and for usual media , one can assume that their permeabilities are very close to that of vacuum [ nonmagnetic media ] @xcite , which in practical terms means that rays do not reflect if the medium they come from has the same index of refraction as the one they go to ) ; for compact systems , it would lead photons to propagate freely there .
we stress that we are not interested here in elaborating on specific analogue models of gravity , but rather to discuss the relevance and pertinence of cases more related to its absence when curved background spacetimes are taken into account .
this article is organized as follows . in section [ optical metric ]
we revisit how the geometric interpretation may arise and its limitations for light propagation in the limit of geometric optics .
section [ exploring ] is devoted to the study ( for the first time ) of some cases where optical metrics could be minkowski or almost minkowski in the contexts of charged black holes described by nonlinear lagrangians of the electromagnetism , nonlinear liquid media and in some regions of neutron stars .
this could happen if some of their parameters fulfilled some constraints , which we derive and discuss in detail , specially regarding their physical reasonableness .
in section [ conc ] we gather the key points raised and discuss possible far - reaching consequences of our analyses for material media and astrophysics .
units here are such that @xmath0 , unless otherwise stated .
in the limit of geometric optics , the important concept is that of rays , in very close resemblance with material particles @xcite .
nevertheless , wave properties are also important , since in the geometric optics limit wave packets could always be made , leading to ray description .
let us start with a monochromatic plane wave described by its wave four - vector @xmath1 , where @xmath2 is its frequency and @xmath3 its wave vector that defines the direction of propagation . in the limit of geometric optics
, such a four - vector must be orthogonal to a hypersurface [ the eikonal @xmath4 , thus @xmath5 .
let us assume that the background spacetime is arbitrary .
making use of a locally inertial frame , the equations governing the propagation of rays are @xcite @xmath6 where @xmath7 is the time recorded there and @xmath8 .
the latter relation is obtained by dint of the fresnel equation ( dispersion relation ) @xcite to the medium under interest . for media
whose dielectric coefficients are functions of the space - time coordinates or for those ones that are frequency dependent , but operating far from resonance or plasma frequencies , the fresnel equation can always be written as @xcite @xmath9 where @xmath10 is a rank-4 completely symmetric tensor depending only on the electromagnetic field and its continuous derivatives . in this work we shall be concerned with situations in which @xmath11 is decomposable in terms of a tensorial product of @xmath12s ( naturally occurring in nonlinear lagrangians of the electromagnetism @xcite and for some nonlinear isotropic media @xcite ) , such that eq .
( [ fresnelgen ] ) can be factorized in quadratic terms of the form @xmath13 we shall see that only in such situations one could make analogies of the propagation of light rays and geodesics in a curved manifold .
given eq .
( [ fresnelrestrict ] ) , it is simple to show that eqs .
( [ geoopteq ] ) can be cast as @xcite @xmath14 where we have defined @xmath15 as a parameter along the ray trajectories .
clearly eq .
( [ geooptcovariant ] ) is the covariant version of eqs .
( [ geoopteq ] ) and therefore is valid in any coordinate system . in order to obtain the ray paths ,
one has to solve eqs .
( [ geooptcovariant ] ) . for eq .
( [ fresnelrestrict ] ) this solution can be obtained generically , as we shall show now . from eqs .
( [ geooptcovariant ] ) and eq .
( [ fresnelrestrict ] ) , we have that @xmath16 where @xmath17 is defined such that @xmath18 . by differentiating @xmath19 in the above equation with respect to @xmath20 and using the second equation in eq .
( [ geooptcovariant ] ) , after some simplifications one has that @xmath21 therefore , if one assumes that rays propagate in an effective spacetime whose geometrical structure is encompassed by @xmath17 , then the second term of the left - hand side of eq .
( [ ddotx ] ) is proportional to the christoffel symbols .
based on this effective geometric structure , eq .
( [ ddotx ] ) can be cast as @xmath22 which makes it clear that rays follow along null geodesics in an effective spacetime metric @xmath17 [ according to the definition of @xmath23 and eqs .
( [ geooptcovariant ] ) ] : @xmath24 therefore , in order to know the ray trajectories , the effective metric structure is vital .
note that light - cones are determined by @xmath17 [ see the above equation ] .
out of the limit of geometric optics , the background metric is the responsible for the causal structure .
let us now say a few words about the case where @xmath25 .
in such a scenario eq .
( [ kintermsofraypath ] ) will be modified and by consequence the right - hand side of eq .
( [ ddotgeodesic ] ) will gain terms not proportional to @xmath26 .
therefore , only the case @xmath27 is connected with usual geodesics for light rays .
concerning the geometrization of light propagation in material media , we point out that the approach we made use of is slightly different from the one coming from transformation optics . the former deals with given dielectric coefficients (
so there is no induction of a material medium due to a coordinate transformation ) , which by means of the fresnel equation are mapped onto induced geometries for light rays ( either ordinary or extraordinary ) . for light propagation in a nonlinear theory of the electromagnetism
, one can indeed map it onto a continuous dielectric medium , but the associated permittivity and permeability tensors are generally nonlinear and different @xcite
. specific properties of light rays in our description arise from convenient optical metrics ( or effective geometries ) , similarly as in transformation optics .
contrary to the usual cases in which nonlinear theories of electromagnetism in flat space in the limit of geometric optics lead to the appearance of an effective curved metric for the propagation of perturbations , in this paper we wonder under which circumstances some nonlinear theories defined in a curved spacetime could cancel out the effects of gravity , such that perturbations would follow along light - like curves of minkowski spacetime .
similarly , one can also wonder the properties nonlinear media and fields in given curved backgrounds should have so that their optical metrics be flat .
thus , in this section we scrutinize the properties some well - known electromagnetic systems in the literature ( but very different in their nature ) should have in order that their optical metric structure be simple ( such as minkowskian ) . here
we shall investigate the conditions a nonlinear lagrangian , an electric field and a background metric ( all of them connected by the field equations and not fixed a priori ) should satisfy such that the associated optical metric be `` controllable '' .
it is already known that for any lagrangian @xmath28 ( valid for configurations where the magnetic field is absent ) , @xmath29 is the invariant constructed from the electromagnetic tensor @xmath30 , the following effective metric emerges for the propagation of disturbances in this context @xcite : @xmath31 where @xmath32 and @xmath33 is instead the second derivative of @xmath34 with respect to @xmath35 . to illustrate what may happen with nonlinear electromagnetic theories in curved backgrounds ,
let us analyze the case of a spherically symmetric nonlinear charged black hole related to a nonlinear lagrangian @xmath28 , in the realm of general relativity .
it is simple to show that its background metric is given by ( see for instance @xcite ) @xmath36 where @xmath37 is a constant of integration ( charge of the system ) and the electric field @xmath38 itself is determined by @xmath39 which is nothing more than the sole equation of the electromagnetism for the case under interest coming from an arbitrary @xmath28 , given that there @xmath40 . if the nonlinear lagrangian to the system is known , then the effective geometry is given and only its consequences could be probed . for the problem of a nonlinear black hole spacetime , in order to have a controllable effective metric
, strictly speaking , one must assume that the nonlinearities of the lagrangian are to be fixed , as well as the electric field of the system .
this case may be of relevance in assessing certain kinds of nonlinearities of the electromagnetism whose associated light propagation in the corresponding nonlinear black hole spacetimes are simple , irrespective of how involved the background metrics may be .
let us start our analyses by assuming that @xmath41 where @xmath42 is a given function of the radial coordinate that satisfies the requirement @xmath43 . from eq .
( [ hmununled ] ) one clearly sees this is the sole term related to the electrodynamics of the system that contributes to its effective geometry .
simplifications of eq .
( [ geffcontraint ] ) taking into account eq .
( [ ernled ] ) leads to the generic solution @xmath44\label{ersolution},\ ] ] where we considered that its asymptotic solution is minkowskian , as implied by @xmath43 .
let us write the lagrangian density as @xmath45 where @xmath46 is a nonlinear term and @xmath47 corresponds to the maxwellian contribution for the lagrangian . from eqs .
( [ ernled ] ) and ( [ l ] ) , one shows that @xmath46 must satisfy the condition @xmath48 with the constraint that @xmath49 .
therefore , the hierarchy of the problem under interest is the following : choose a @xmath42 such that @xmath43 ; then find the electric field @xmath38 by solving eq .
( [ ersolution ] ) ; afterwards solve eq .
( [ lnl ] ) and , finally , solve for @xmath50 according to eq .
( [ g00nled ] ) . for each choice of @xmath42 ,
the effective metric becomes @xmath51-r^2d\theta^2-r^2\sin^2\theta d\varphi^2 \label{ds2eff},\ ] ] while the background metric is @xmath52 for the case of black holes , it would be of interest to analyze the situation where @xmath53 , since it would lead to a simple optical metric structure . in this case , from eqs .
( [ g00nled ] ) , ( [ ersolution ] ) and ( [ lnl ] ) , one shows that for @xmath38 different from a constant @xmath54\nonumber \\&+ & re_{r , r^2}[(e_{r , r})^2 - 2e_r e_{r , r^2 } ] = 0\label{erfg00}\end{aligned}\ ] ] and ( background metric ) @xmath55 where we have defined @xmath56 . in what follows we shall not investigate eqs .
( [ erfg00 ] ) and ( [ g00fg00 ] ) but rather make generic analyses of photon propagation .
we start by focusing on the role played by @xmath42 on radial geodesics of the effective metric , in particular at the points where @xmath50 is null , which invalidate the conformal analogy between ( [ ds2eff ] ) and ( [ ds2 ] ) .
apparent singularities of the background metric ( [ ds2 ] ) can be removed by introducing the painlev - gullstrand ( pg ) coordinates , defined as @xmath57 .
the integration of the ingoing null radial geodesics related to eq .
( [ ds2eff ] ) in the aforesaid coordinates leads us to @xmath58 & \dot t={\frac{\textstyle{f}}{\textstyle{g_{00}}}}\left(1-\sqrt{1-g_{00}}\right),\label{tdot}\end{aligned}\ ] ] where @xmath59 and @xmath20 is an affine parameter . for the background spacetime [ @xmath60
, the above equations show that ingoing light rays cross its outer horizon ( denoted from now on by @xmath61 ) .
nothing similar happens with the effective metric when @xmath62 . from eqs .
( [ rdot ] ) and ( [ tdot ] ) , one can easily see that in this case @xmath63 at @xmath64 , indicating that light rays can not cross from outside to inside of the outer horizon , since external observer fields do not perceive any singularity there . in other words
, the outer horizon avoids that any eventual singularity in the background metric can be reached by photons .
a direct way of seeing that indeed photons can not cross @xmath64 is by introducing the new `` radial '' coordinate @xmath65 , which leads eq .
( [ ds2eff ] ) to @xmath66 by assuming that @xmath50 is a rational function that vanishes ( at least ) at @xmath64 , one has that the domain of @xmath67 is the whole real line , while its image @xmath68 is bound ( from below ) by @xmath64 .
this indicates that the region @xmath69 is excluded from the portion of the space - time accessible to the ingoing photons ; in this coordinate system , radial null geodesics are _ maximally extended_. note also from eq .
( [ ds2eff2 ] ) that in the case under investigation it is impossible to render the effective metric minkowskian .
we shall now analyze the case of nonlinear media in the context of curved backgrounds such that its optical metric be flat .
in particular , we shall use liquid media since under natural circumstances they are isotropic .
this simplifies tremendously the propagation of disturbances there @xcite .
when electromagnetic fields are present in these media , they might eventually become optically anisotropic ( leading to the phenomenon of birefringence ) , with the induced optical axis reflecting their symmetries @xcite . just in order to be simple and to evidence the physical ideas we want to convey ,
let us analyze kerr media ( see @xcite and references therein ) . for this case , we have that its permeability is a constant , defined by @xmath70 , and its permittivity is given by @xmath71 , where @xmath72 is the norm of the external electric field @xmath73 and @xmath74 the four - velocity of an observer , orthogonal to @xmath73 , who measures the electric field .
let us investigate here only extraordinary photons , related to the anisotropic solution of the fresnel equation in the presence of electric fields .
ordinary photons will be generically described at the end of this section .
the associated effective metric for this case can be obtained from its fresnel equation and the result , when generalized to an arbitrary coordinate system , is @xcite @xmath75v^{\mu}v^{\nu } - \frac{\epsilon'}{\epsilon}e^{\mu}e^{\nu}\label{kerrmetric},\ ] ] with @xmath76 .
observe that for the case @xmath77 is a constant , eq .
( [ kerrmetric ] ) gives us the well - known gordon metric ( see for instance refs .
it is simple to show that the inverse of eq .
( [ kerrmetric ] ) is @xmath78 assume a ( given ) background metric with spherical symmetry such that @xmath79 .
consider the same symmetry for the ( unknown ) electric field , that with respect to an observer at rest [ @xmath80 is then @xmath81 .
for this case , eq . ( [ hmunukerr ] )
admits the solution @xmath82 as far as @xmath83 and @xmath84 when eq .
( [ h11cond2 ] ) is inserted into eq .
( [ h00cond1 ] ) , one has that @xmath85 from the radial symmetry involved and eqs .
( [ h11cond2 ] ) and ( [ mu_epsilon ] ) , it follows that the generic solution to the norm of the electric field can be cast as @xmath86\label{er},\ ] ] where @xmath87 .
in order to exemplify the above formalism , let us investigate the case @xmath88 , valid , for instance , for exterior solutions to general relativity in the spherically symmetric case . here , eq .
( [ er ] ) simplifies to @xmath89 ^ 2\label{eg00}.\ ] ] for a concrete example , assume @xmath90 , where @xmath91 is a constant and @xmath92 .
then , eq . ( [ eg00 ] ) reads @xmath93 from maxwell equations in material media , given that @xmath94 , the associated charge density is @xmath95= \frac{e_0}{4\pi\mu_0}\frac{\left(r-{2m}\right)^{\frac{3}{2}}\left(4r-3m\right)}{r_0 ^ 2}\label{charge_density}.\ ] ] one clearly sees from eqs .
( [ erschw ] ) and ( [ charge_density ] ) that the electric field is never null and that the system has a nonzero net charge . for situations
where @xmath96 , for instance , one has from the above equations that @xmath97 , which means @xmath98 .
this is the dielectric response a kerr liquid medium must have in the presence of the electric field with norm given by eq .
( [ erschw ] ) for supporting straight light ray trajectories with speed @xmath99 there .
we finally remark that all the analyses in this section were related to the optical metric felt by extraordinary rays . the ordinary one is simply given by a gordon - like metric , obtained from eq .
( [ hmunukerr ] ) by assuming @xmath100 .
as it can be checked in the particular example of this section , it is not possible to render both ordinary and extraordinary optical metrics minkowskian at the same time , since they are very different in their functional forms .
generally , one could in principle only render the ordinary metric minkowskian in moving media @xcite , since further degrees of freedom should emerge there .
one could also imagine astrophysical consequences related to the flexibility of optical metrics .
consider for instance the photon propagation in the interior of compact objects .
such photons are present in material media , not in vacuum , and therefore in the limit of geometric optics it would also be necessary to know the dielectric properties of the astrophysical system in order for photons to be properly described . photon trajectories _ must _ be related to effective geometries and they carry the information of the underlying spacetime metric and medium
. naively speaking , one would expect that the presence of a material medium would considerably slow the photons and all the information we can get from the astrophysical system would come from the outermost regions .
notwithstanding , we have seen that whenever an optical metric becomes flat photons would not be trapped anymore and could freely escape the system .
this could constitute a mechanism to release energy from compact stars and coalescent systems that could take place intermittently , due generally to their nontrivial dynamical evolution .
let us analyze , only as a particular case of the above discussion , the conditions required for the extraordinary optical metric to be minkowskian in the vicinities of the origin of a static , charged and spherically symmetric compact system .
such an analysis is deemed to be important since it would give us insights into the physically interesting cases to pay attention in more precise investigations .
the main point here is not related to the strength of the nonlinearities in the dielectric coefficients ( obviously small for regular electric fields ) , but solely to their existence , since it is only in this case they would naturally be connected with electric fields , which must coincide with the ones generated by charge distributions in stars and , thus , be regular .
analyses of this sort are motivated by the possibility of allowing photons produced in the innermost regions of compact stars to propagate freely outwards , giving us hope to assess how such regions are .
in any other case this seems highly improbable . near the center of the aforesaid system
, the total energy density @xmath101 ( taking into account the contributions due to electromagnetic and matter fields ) could always be taylor - expanded , yielding @xmath102 where @xmath103 and @xmath104 are constants , and @xmath105 is the total energy density at the origin . note that some @xmath106 s are null , such as the ones related to odd @xmath107 , due to symmetry criteria . nevertheless , by making use of last subsection s results
, we shall show that their specification is irrelevant for the regularity of an electric field in a finite region around the origin of a star ( by choosing a convenient @xmath108 for so ) .
let us define , only for convenience , @xmath109 and @xmath110 .
in the case we are interested in , einstein equations lead to @xcite @xmath111 and @xmath112\label{derlog},\end{aligned}\ ] ] where @xmath113 is the total radial pressure at @xmath114 and the second term of eq. ( [ g11nearorigin ] ) was assumed to be much smaller than its first one .
assume now that the dielectric properties of the matter content in the central parts of a compact system can be effectively described by those of a transparent nonlinear liquid .
this is reasonable and natural since it should in principle behave similarly to a plasma , and as so it should be transparent and present low - losses for waves whose frequencies are higher than the plasma one .
thus , such a dispersive nonlinear medium would always behave as a transparent nonlinear liquid for any convenient frequency range above the plasma one such that its permittivity could be taken as only dependent upon the electric field there ( practically constant in terms of the frequency ) .
actually , when searching for flat optical metrics , one is exactly seeking to cancel it out . from eqs .
( [ er ] ) , ( [ g11nearorigin ] ) and ( [ derlog ] ) , one has that the norm of the electric field near the origin for a minkowskian optical metric would behave as @xmath115 being thus null there only when @xmath116 .
note that any term in eq .
( [ derlog ] ) whose @xmath117 s power is larger than the unity is irrelevant for the issue of the regularity of the electric field .
taking into account eq .
( [ mu_epsilon ] ) to the case under interest , one sees that its leading order term is @xmath118 , which fixes the frequency range where our analyses would be valid .
corrections to this value can be easily obtained from the integration of eq .
( [ derlog ] ) .
let us now qualitatively assess possible candidates for our formalism .
as we have stated previously , any reasonable and self - consistent candidate should be such that @xmath116 .
for this case , due to the regularity of the electric field at @xmath114 , one could safely approximate the total energy - momentum tensor there by its matter counterpart , which we shall assume to be a perfect one . in this regard
, it can be checked that the well - known @xmath119 model for dense hadronic matter within the relativistic mean - field approximation ( see for instance @xcite and references therein ) can satisfy such a constraint , since in its high baryon density limit ( when compared to the nuclear density , @xmath120 ) , @xmath121 from below .
for instance , for the nl3 parametrization to this model @xcite , one can verify that for @xmath122 , @xmath123 takes place .
let us see if this could also happen for systems that exhibit an innermost strange quark phase .
taking , as an example , a mit bag - like model to describe such a phase [ @xmath124 , with @xmath125 and @xmath126 ( see @xcite and references therein ) ] , one indeed has that @xmath127 also leads eq .
( [ ernearorigin ] ) to be null at @xmath114 . as we have already said , another obvious criterion that reasonable candidates for flat optical metrics must have is a nontrivial electric structure .
in this regard , we stress that there is not a conclusive observational answer to that yet .
it is known that the phenomenology of compact stars would evidence electric aspects to these systems only when their charge are of the order of @xmath128c ( related to electric fields of the order of @xmath129v / m ) @xcite .
naturally , our optical analyses would become meaningful for much lower charges , expected to be around @xmath130c or even smaller ( related to fields @xmath131v / m ) .
nevertheless , it is very likely that phase transitions lead to the appearance of large electric fields , at least in some regions of stratified stars @xcite . besides
, it is also believed that strange stars should have huge electric fields near their surfaces , around @xmath132v / m @xcite , which also point to interior electric structures ( surely of smaller magnitudes ) .
actually , the presence of these huge electric fields could even explain the observed high magnetic fields in compact systems @xcite , as well as masses of the order of two solar masses @xcite .
the opposite is also naturally true : since chiefly deutsch @xcite and goldreich and julian @xcite , it is well - known that high conducting magnetic stars must also present electric field structures [ internal and external ] .
therefore , our simple analyses suggest that stratified neutron stars with central densities of a few nuclear densities , as well as strange quark stars , could be possible candidates for minkowskian optical geometries , at least for some of their regions , more likely the outermost ones .
it also clearly shows that it is far from obvious that dense systems should be optically thick or opaque a priori . detailed studies in this direction will be conducted elsewhere .
we have seen that the flexibility of effective metrics could have far - reaching implications .
if it may be rendered or naturally becomes trivial ( minkowskian ) , then one ( or nature ) could totally cancel out light interaction with matter and fields , and light rays would propagate as if they were in vacuum .
besides , there the associated light rays would not experience anymore the phenomenon of refraction nor reflection when coming from or going to vacuum . we stress that if the medium is birefringent without the control of optical metrics , it will keep being so even when they are controlled .
the reason is simply because birefringent media have two optical metrics : an ordinary and an extraordinary , and generally they could only be controlled one at a time . only for completeness
, we recall that material particles , such as electrons , would feel normally the material structure and are bound to the geometry of the background spacetime . light propagation analyses in material media could also be imagined for assessing their equations of state if accurate measurements and dense systems are at play .
( for a proposal of assessment of the properties of spacetimes with scattering measurements , see ref .
one such example could be a quark - gluon plasma , which could shed a light into the issue of the microphysics of the innermost region of neutron stars , motivated by the already existent analyses for its dielectric coefficients ( see @xcite and references therein ) .
this would be done as follows : given the dielectric coefficients , electromagnetic fields and the optical metric of a medium , one could solve the equations for the background metric and therefore infer its equation of state . the simplest case to be analysed
would be when light trajectories follow straight lines , as it is supposed to be evidenced experimentally by the lack of reflection , refraction , etc , for light when it crosses the system .
for the case spherically symmetric spacetimes and electromagnetic fields are involved in kerr media , for instance , the aforesaid analysis is summarized by eqs .
( [ h11cond2 ] ) and ( [ mu_epsilon ] ) .
we stress that for obtaining an unique optical metric out of the propagation of rays , in general , one should make measurements of wave aspects of light , because this is a way to differentiate between a given effective geometry and its conformal one [ see @xmath133 in eq .
( [ kintermsofraypath ] ) ] .
the latter geometries are indistinguishable for the propagation of light rays alone since they follow null geodesics there ( see appendix d in ref .
@xcite for further details ) .
concerning minkowskian optical metrics for compact astrophysical objects , we point out that our analyses suggest that it would be of physical interest to investigate the situations where @xmath116 . naively speaking
, it seems that stars with central densities larger than a few nuclear densities could be interesting in the above scenario .
detailed analyses would require specific systems and we leave them to be done in future works .
we generally point out that the bending of light is different for nonlinear charged black holes since light rays follow null geodesics with respect to the effective geometry , not with the background one . for this case ,
light rays would be naturally more appropriate to scrutinize effective theories of the electromagnetism
. properties of the background spacetime could also be assessed , but in a more indirect way , given that the effective metric felt by photons also contains the geometry felt by massive particles .
summing up , in this work we have attempted to clarify the relevance and applicability of optical metrics in the limit of geometric optics for ray propagation ( either ordinary or extraordinary ) .
examples concerning light propagation in nonlinear liquid media , nonlinear charged black holes and compact astrophysical systems were given for the first time to stress and evidence the possibility ( which shows its flexibility , but not in a trivial way ) of having simple optical metrics ( such as minkowskian ) and their possible far - reaching implications in several scenarios .
we plan to enter deeply into the details on these issues elsewhere .
we are indebted to the participants of the _ seminario informale _ at sapienza university for their comments apropos of this work .
e.b . acknowledges the financial support provided by the capes - icranet program through the grant bex 13956/13 .
is likewise grateful to cnpq- conselho nacional de desenvolvimento cientfico e tecnolgico of the brazilian government within the postdoctoral program `` science without borders '' .
10 f. weber and r. negreiros . .
in a. kizilers and a. w. thomas , editors , _ american institute of physics conference series _ ,
volume 1354 of _ american institute of physics conference series _ ,
pages 1318 , may 2011 .
|
we firstly revisit the importance , naturalness and limitations of the so - called optical metrics for describing the propagation of light rays in the limit of geometric optics .
we then exemplify their flexibility and nontriviality in some nonlinear material media and in the context of nonlinear theories of the electromagnetism , both in presence of curved backgrounds , where optical metrics could be flat and inaccessible regions for the propagation of photons could be conceived , respectively .
finally , we underline and discuss the relevance and potential applications of our analyses in a broad sense , ranging from material media to compact astrophysical systems .
|
recent analyses of the helioseismic data , from both the michelson doppler imager ( mdi ) instrument on board the soho spacecraft ( toomre et al .
2000 ) and the global oscillation network group ( gong ) project ( antia & basu 2000 ) , have provided strong evidence that the previously observed torsional oscillations ( e.g. howard & labonte 1980 ; snodgrass , howard & webster 1985 ; kosovichev & schou 1997 ; schou et al .
1998 ) , with periods of about 11 years , penetrate into the convection zone to depths of at least 10 percent in radius .
further studies of these data have produced interesting , but rather inconsistent results .
in particular , howe et al .
( 2000b ) find evidence for the presence of similar oscillations near the tachocline situated close to the bottom of the convection zone , but with a markedly shorter period of about @xmath0 years , whereas antia & basu ( 2000 ) do not find such oscillations ( whilst allowing the possibility that more irregular variations may be present ) .
whatever the true dynamical behaviour at these lower levels turns out to be , both these results indicate that the variations in the differential rotation can have different dynamical behaviour at the top and the bottom of the solar convection zone , with the oscillations at the bottom having either significantly shorter periods or non periodic behaviour . recently , _
spatiotemporal fragmentation / bifurcation _ has been proposed as a dynamical mechanism to account for the possible multi - mode behaviour in different parts of the solar convection zone ( covas et al . 2000a , hereafter ctm1 ) .
evidence for this mechanism was found in the context of a two dimensional axisymmetric mean field dynamo model in a spherical shell , with a semi
open outer boundary condition , in which the only nonlinearity is the action of the azimuthal component of the lorentz force of the dynamo generated magnetic field on the solar angular velocity .
the underlying zero order angular velocity was chosen to be consistent with the recent helioseismic data .
our initial study ( ctm1 ) suffered from two major shortcomings .
the dynamo model employed inevitably contained major approximations , and the zeroth order rotation profile used is bound to include uncertainties , given the uncertainties in the inversion techniques , as well as the short extent in time over which data sets are so far available . to address the first shortcoming , an extensive study was made of the persistence of the fragmentation following plausible changes in the details of the dynamo model .
this study showed that the mechanism remains surprisingly robust in presence of a variety of rather severe changes to the model ( covas et al .
2001 , hereafter ctm2 ) .
regarding the second shortcoming , ctm2 also included a preliminary study , allowing the zero order rotational profile to be given by either the inversion of the mdi data reported by toomre et al ( 2000 ) or of the corresponding gong data ( howe et al 2000b ) .
this also showed the mechanism to be robust with respect to such changes to the rotation profile .
given the significant uncertainties that still remain in the helioseismic measurements , especially given the limited length of the data so far available , it is crucial to study the robustness of the spatiotemporal fragmentation mechanism to a wider range of the rotation profiles consistent with the observations .
new measurements of the solar internal rotation with soho mdi data have been reported recently by vorontsov et al ( 2001 ) , using novel inversion techniques that are quite different from those employed earlier .
results of a 2d inversion confirmed the previous findings that the torsional oscillations ( migrating zonal flows ) penetrate deep into the convection zone .
however , given the limited accuracy of the helioseismic data currently available , they did not allow the resolution of these oscillations all the way down to the bottom of the convection zone .
the so - called 1.5d inversion procedure has also been used , which consists in measuring separately the consecutive terms @xmath1 in the expansion @xmath2 where @xmath3 is the @xmath4th legendre polynomial , @xmath5 is the distance from the centre and @xmath6 is the co latitude .
the 1.5d inversion of vorontsov et al ( 2001 ) was principally targeted at improvement of the spatial resolution of the rotation profile near and just below the base of the convection zone , in the tachocline region the region where the dynamo mechanism is most sensitive to the rotation profile , since the largest gradients of the angular velocity appear to be localized there .
an important feature of the solar internal rotation as seen in the helioseismic measurements is that the rapid spatial variation of the angular velocity in the tachocline is governed principally by the first two or three terms in the expansion ( 1 ) ; the contribution of higher terms appears to be an order of magnitude smaller , being essentially undetectable in the helioseismic data available at present .
another important feature is that this dominant part of the variation of the angular velocity in the tachocline with depth and latitude , captured by the first few terms in the expansion ( 1 ) , does not show any changes with solar activity ( vorontsov et al 2001 ) . taken together , these two features suggest considering the torsional oscillations as a small - amplitude , rapidly - varying ( in @xmath6 ) , time - dependent component of the solar rotation , imposed on a large - amplitude , slowly - varying ( in @xmath6 ) , nearly - stationary component described by the first terms of the expansion ( 1 ) .
these considerations show that at least a dominant part of the time - independent ( background ) component of solar rotation can be evaluated by a simple truncation of the expansion ( 1 ) to the first two or three terms , making the truncated expansion a natural candidate for the zero - order rotation profile in the dynamo modelling .
this choice might be more consistent than using a time average of the inverted rotation profiles ( as in ctm2 ) , since the observational data used in the inversions do not yet cover enough ( ca .
5 yr ) of a complete solar cycle to allow proper averaging . with the zero - order profile derived from the time - independent part of the data we can also eliminate any possibility that torsional oscillations produced by the dynamo model originate from the time varying zonal flows already present ( in time - averaged form ) in the background rotation profile .
thus we here describe a comparative study of the effects of employing zero order rotation profiles , obtained by taking the first two and the first three terms of the expansion ( [ expansion ] ) .
these profiles , inferred from the 360d soho mdi data set , are shown in fig .
[ vorontsov135 ] . # 1#20.45#1 # 1#20.45#1
in order to test the sensitivity of the spatiotemporal fragmentation with respect to the uncertainties in the rotation profile , we used the following dynamo model , which was also used in ctm1,2 .
we assume that the gross features of the large scale solar magnetic field can be described by a mean field dynamo model , with the standard equation @xmath7 here @xmath8 , the term proportional to @xmath9 represents the effects of turbulent diamagnetism , and the velocity field is taken to be of the form @xmath10 , where @xmath11 , @xmath12 is a prescribed underlying rotation law and the component @xmath13 satisfies @xmath14 where @xmath15 is the operator @xmath16 and @xmath17 is the induction constant .
the sole nonlinearity in the dynamo equation is the feedback of the azimuthal component of the lorentz force ( eq . ( [ ns ] ) ) , which modifies only slightly the underlying imposed rotation law , but nevertheless limits the magnetic fields at finite amplitude .
the assumption of axisymmetry allows the field @xmath18 to be split simply into toroidal and poloidal parts , @xmath19 , and eq .
( [ mfe ] ) then yields two scalar equations for @xmath20 and @xmath21 . nondimensionalizing in terms of the solar radius @xmath22 and time @xmath23 , where @xmath24 is the maximum value of @xmath25 , and putting @xmath26 , @xmath27 , @xmath28 , @xmath29 and @xmath30 , results in a system of equations for @xmath31 and @xmath13 .
the dynamo parameters are the two magnetic reynolds numbers @xmath32 and @xmath33 , and the turbulent prandtl number @xmath34 .
@xmath35 is the solar surface equatorial angular velocity and @xmath36 .
thus @xmath37 and @xmath24 are the turbulent magnetic diffusivity and viscosity respectively , @xmath38 is fixed when @xmath24 is determined ( see sect .
[ res ] ) , but the value of @xmath39 is more uncertain .
the density @xmath40 is assumed to be uniform . given the ill determined nature of boundary conditions in astrophysical settings , we make physically motivated choices following ctm1,2 . at inner boundary we chose @xmath41 , ensuring angular momentum conservation , and an overshoot type condition on @xmath42 ( cf . moss & brooke 2000 ) . at the outer boundary
, we used an open boundary condition @xmath43 on @xmath21 and used vacuum boundary conditions for @xmath42 ( see kitchatinov et al .
( 2000 ) and ctm1,2 for motivation ) . equations ( [ mfe ] ) and ( [ ns ] ) were solved using the code described in moss & brooke ( 2000 ) ( where more details are given ; see also covas et al .
2000a ) together with the above boundary conditions , over the range @xmath44 , @xmath45 .
we set @xmath46 ; with the solar convection zone proper being thought to occupy the region @xmath47 , the region @xmath48 can be thought of as an overshoot region / tachocline .
we note that there is some evidence that the tachocline may be rather narrower than can be directly resolved by the usual inversion techniques .
the uncertainties associated with this possibility are difficult to assess , but we do note that the appearance of fragmentation appears to be quite robust with respect to changes in the model ( ctm2 ) . in the following simulations we used a mesh resolution of @xmath49 points , uniformly distributed in radius and latitude respectively . in this investigation , we took the zero order rotation profile @xmath50 in the region @xmath51 to be given successively by the first two and the first three terms of the series ( [ expansion ] ) given above . for the alpha - effect
, we took @xmath52 , where @xmath53 ( cf .
rdiger & brandenburg 1995 ) and @xmath54 for @xmath55 with cubic interpolation to zero at @xmath56 and @xmath57 , with the convention that @xmath58 and @xmath59 . also , in order to take into account the likely decrease in the turbulent diffusion coefficient @xmath25 in the overshoot region , we allowed a simple linear decrease from @xmath60 at @xmath61 to @xmath62 in @xmath63 .
we note , however , that as was shown in ctm2 , these latter details of the model are unlikely to remove the possibility of spatiotemporal fragmentations .
we calibrated our model so that near marginal excitation the cycle period was about 22 years .
this determined @xmath64 , corresponding to @xmath65 @xmath66 sec@xmath67 , given the known values of @xmath35 and @xmath22 . with the rotation law given by the first three terms of ( 1 ) , the first solutions to be excited in the linear theory are limit cycles with odd ( dipolar ) parity with respect to the equator , with marginal dynamo number @xmath68 .
the even parity ( quadrupolar ) solutions are also excited at a similar marginal dynamo number of @xmath69 .
these values did not change significantly when the first two terms of expansion ( 1 ) were used .
it is arguable that the turbulent prandtl number is of the order of unity , and we set @xmath70 . for the parameter range that we investigated
, the even parity solutions can be nonlinearly stable .
given that the sun is observed to be close to an odd ( dipolar ) parity state , and that previous experience shows that small changes in the physical model can cause a change between odd and even parities in the stable nonlinear solution , we chose to impose dipolar parity on our solutions . with these parameter values , each of the rotation profiles considered here produced butterfly diagrams which are in qualitative agreement with the observations .
the model also successfully produced torsional oscillations that penetrate into the convection zone , in all cases considered , similar to those deduced from recent helioseismic data ( howe et al .
2000a , vorontsov et al . 2001 ) , and studied in covas et al .
( 2000b ) # 1#20.42#1
# 1#20.42#1 both of the rotation profiles considered here produced spatiotemporal fragmentation .
as examples , the top and the bottom panels of fig . [ fragmentation_vorontsov ]
show the radial contours of the angular velocity residuals @xmath71 as a function of time for a cut at latitude @xmath72 , with the zero order rotation profile given by the first two and the first three terms in the expression ( [ expansion ] ) , respectively . to test further this robustness , we also used functions @xmath73 , obtained from an inversion with slightly different regularisation parameters and in a somewhat different frequency range of the input data ( vorontsov et al 2000 ) .
we found no qualitative changes in the spatiotemporal fragmentation . finally , as a comparison with our previous results obtained by employing the gong and mdi data , we have plotted in fig .
[ fragmentation.mdi.gong.vorontsov.13.135 ] the radius of the top of the fragmentation region .
this qualitative similarity of the fragmentation in all the cases is further evidence of robustness . # 1#20.4#1
we have shown that spatiotemporal fragmentation persists even when only the lowest order terms in the legendre expansion of the inversion for the solar rotation law are used .
this is important , as these terms appear to be approximately time independent ( over the scale of a solar cycle ) , and also dominate the expansion .
indeed , these terms really represent the rotation law that _ should _ be input to the dynamo code in principle solution of eq .
( [ ns ] ) should then _ determine _ the variations of @xmath74 over a cycle .
thus , these results allay worries that the inconsistency of averaging the observed rotation law over only a part of a solar cycle might be giving misleading results .
the results presented here , together with those of ctm1 and ctm2 , show spatiotemporal fragmentation to be a very persistent feature of our simulations .
this finding suggests that nonlinear dynamo modelling has a real potential predictive power , which could lead to a better understanding of the solar dynamo when longer data sets as well as improved analyses of them become available .
this in turn , in conjunction with the expected improvements in the helioseismological data , could lead to an improved determination of the properties of the solar interior , including the nature of the tachocline .
|
analyses of recent helioseismic data indicate that the dynamical regimes at the base of the convection zone can be different from those observed at the top , having either significantly shorter periods or non periodic behaviour .
recently spatiotemporal fragmentation / bifurcation has been proposed as a dynamical mechanism to account for the multi - mode behaviour that is possibly observed in the solar convection zone , without requiring separate physical mechanisms with different time scales at different depths . here
we study the robustness of this mechanism with respect to changes to the zero order rotation profile , motivated by the uncertainties of and differences between the various reductions of the helioseimic data .
we find that spatiotemporal fragmentation is a common feature of the reductions investigated .
|
phenomenological approaches to quantum gravity have recently started to develop in parallel with the more traditional attempts to construct such a theory from first principles . in particular , increasing attention has focused on the consideration that maybe lorentz invariance is not a fundamental law , but an effective low - energy symmetry which is broken at high energies ( see , e.g. , @xcite for a general introduction ) .
conceptually speaking , in quantum gravity theories from first principles , it is not really clear whether lorentz invariance is fundamental or effective , and in the latter case , how its breaking scale is related to the planck scale .
for example , while many string theory scenarios axiomatically incorporate lorentz invariance , it has been argued that in certain situations , violations of lorentz invariance may occur in a way consistent with world - sheet conformal invariance @xcite , thus leading to acceptable string theory backgrounds . in the context of loop quantum gravity , in @xcite
it has been argued that quantum effects should modify the relativistic dispersion relations , although the issue seems far from settled ( see @xcite for some general remarks ) . in scenarios of emergent gravity based on condensed matter analogies @xcite ,
the situation is clearer : lorentz invariance is a low - energy effective symmetry , and so it is expected to break at some scale , although not necessarily related to ( and therefore possibly much higher than ) the planck scale @xcite .
a simple way of modelling a wide range of lorentz violating effects ( and a quite natural one in the case of condensed matter analogies ) consists in modifying the dispersion relations at high energy @xcite .
this modification can be subluminal or superluminal , depending on whether high - frequency modes move slower or faster than low - energy ones .
we should mention that there also exist ways of modifying the dispersion relations , for example by introducing a minimum length or maximum energy , without violating lorentz invariance @xcite . here , however , we will be interested in modifications that explicitly break lorentz invariance . from extrapolation of current experiments , we know that there exist stringent bounds on the most commonly expected types of lorentz violations at the planck scale ( see , e.g. , @xcite and references therein ) .
nevertheless , even violations at much higher energy scales might still significantly affect black hole physics
. in this paper we will investigate the effects that a superluminal modification of the dispersion relation would have on the hawking radiation produced by collapsing configurations .
hawking s original prediction that black holes radiate thermally @xcite , rested on the implicit assumption that the low - energy laws of physics , and in particular lorentz invariance , are preserved up to arbitrarily large scales , much higher than the planck energy .
the question of the robustness of hawking s prediction to modifications of the transplanckian physics has been tackled principally by analyzing effective field theories such that at high energies a modification of the dispersion relation is incorporated ( see however @xcite for a different take on the problem ) .
historically , attention has mainly been given to subluminal dispersion relations .
important contributions such as @xcite , and more recently @xcite seem to have settled the robustness of hawking radiation with respect to subluminal modifications and truncation of the frequency spectrum , although this conclusion still rests on certain assumptions , usually related to the behaviour of the fields near the horizon . in any case , it is important to remember that this only solves part of the so - called ` transplanckian problem ' with regard to hawking radiation . indeed
, subluminal modifications gradually dampen the influence of ultra - high frequencies , and so they do not explore arbitrarily large frequencies . so even assuming that it has been demonstrated that hawking s result can be recovered without appeal to transplanckian frequencies , and that it is robust to subluminal modifications , the question remains whether it is also robust with respect to a ( non - dampened ) modification of the transplanckian physics .
superluminal modifications differ conceptually from subluminal ones , in that they gradually magnify the influence of ultra - high energies , and thereby offer an interesting test - case for the transplanckian robustness of hawking radiation .
an additional motivation for our study comes from a possible connection with experiment .
it is well - known ( see , e.g. , @xcite ) that hawking radiation is not a purely gravitational effect , but a characteristic of quantum field theory in curved spacetimes with a horizon .
condensed matter systems such as bose - einstein condensates ( becs ) can , under certain approximations , be described by a relativistic quantum field theory .
it has also been shown , again under certain idealizing approximations , that ( acoustic ) black hole horizons in becs are dynamically stable @xcite .
therefore , becs are expected to be good candidates for a possible experimental verification of hawking radiation @xcite .
but since the physics of becs automatically leads to a superluminal dispersion relation at high energies , the question is again which kind of modifications are to be expected in the laboratory realization of a bec black hole with respect to the standard hawking picture . of the above - mentioned works on the robustness of hawking radiation ,
a few have also tried to address superluminal modifications @xcite .
various problems make this case quite different from the subluminal one .
these problems can be related to the fact that the horizon becomes frequency - dependent when modifying the dispersion relations .
this is also true in the subluminal case , though qualitatively in a very different way . with superluminal modifications ,
the horizon lies ever closer to the singularity for increasing frequencies , and asymptotically coincides with it .
this causes the ` apparent ' interior of the black hole ( the interior of the zero - frequency horizon ) to be exposed to the outside world .
since it seems unreasonable to impose a condition arbitrarily close to the singularity as long as we do not have a solidly confirmed quantum theory of gravity , most of the approaches used for subluminal dispersion are invalid , or at least questionable , for superluminal dispersion . moreover
, since it seems reasonable to expect that quantum effects will remove the general relativistic singularity , a critical frequency might appear above which no horizon would be experienced at all .
in the analysis that follows we will try to avoid making any further assumptions about the physics near the singularity .
for instance we will analyze the characteristics of the radiation at retarded times at which the singularity has not yet formed .
our approach will be based on a derivation of hawking radiation through the relation between the asymptotically past ray trajectories and the asymptotically future ones in the case of a collapsing configuration .
the only assumption about these asymptotic extremities are the standard ones , namely a minkowski geometry in the asymptotic past , and flatness at spatial infinity also in the asymptotic future .
the language we will use is related to the fluid analogy for black holes , which provides the intuitive picture of the spacetime vacuum flowing into the black hole and getting shredded in its center .
our main results can be summarized as follows .
three crucial elements distinguish the late - time radiation with superluminal dispersion relations from standard relativistic ones .
first , at any instant there will be a critical frequency above which no horizon has yet been experienced .
this critical frequency will induce a finite limit in the modes contributing to the radiation , which will therefore have a lower intensity than in the standard case , even if the critical frequency is well above the planck scale .
second , due to the effective frequency - dependence of the horizon , the surface gravity will also become frequency - dependent and the radiation will depend on the physics inside the black hole . unless special conditions are imposed on the profile to ensure that the surface gravity is nonetheless the same for all frequencies below the critical one
, the radiation spectrum also undergoes a strong qualitative modification .
depending on the relation between the critical frequency and the planck scale , the radiation from high frequencies is no longer negligible compared to the low - frequency thermal part , but can even become dominant .
this effect becomes more important with increasing critical frequency .
finally , a third effect is that the radiation will extinguish as time advances .
the remainder of this article has the following structure . in section [ s : model ]
, we will describe and motivate the classical geometry of our model and the concrete kinds of profiles that we are interested in . in section [ s : sdr ]
, we will briefly review how hawking s standard result can be obtained in the formalism of our choice for the case of standard relativistic dispersion relations .
this calculation will be adapted in section [ s : mdr ] to superluminal dispersion relations , and we will analytically obtain a formula for the late - time radiation for this case . then , in section [ s : graphics ] , we will present graphics obtained from this analytic formula by numerical integration . these graphics will illustrate the results mentioned in the previous paragraph , which we will discuss in more depth and compare with other results in the recent literature in section [ s : conclusion ] .
we will study radiation effects in simple 1 + 1 dimensional collapsing configurations .
it is well known that in the hawking process , most of the radiation comes out in the @xmath0-wave sector .
hence , a spherically symmetric treatment , effectively 1 + 1 dimensional , suffices to capture the most relevant aspects of the process .
we will work in painlev - gullstrand coordinates , where @xmath1\d t^2 - 2v(t , x)\d t \d x + \d x^2~ , \label{acoustic - metric}\end{aligned}\]]and make the further simplification of taking a constant @xmath2 ( we will always write @xmath2 explicitly to differentiate it from the frequency / wave - number dependent @xmath3 that will show up in the presence of a superluminal dispersion relation ) . in this manner all the information about the configuration is encoded in the single function @xmath4 .
the painlev - gullstrand coordinates have the advantage , compared to the schwarzschild form , of being regular at the horizon .
moreover , they suggest a natural interpretation in terms of the language of acoustic models . in such a model , @xmath2 represents the speed of sound and @xmath5 is the velocity of the fluid ( which corresponds to the velocity of an observer free - falling into the black hole ) . throughout the paper we will consider the fluid
to be left - moving , @xmath6 , so that the outgoing particles of light / sound move towards the right .
the most relevant aspects of the analysis presented hereafter depend only on the qualitative features of the fluid profile .
however , to justify some specific calculations later on it is helpful to use the following concrete profiles , see figs .
[ fig : v - constant - kappa ] and [ fig : v - increasing - kappa ] . consider a velocity profile @xmath7 such that , and further decreasing monotonically as . in the fluid image ,
the fluid nearly stands still at large distances and accelerates inwards , with a sonic point or horizon at @xmath8 .
the further decrease in @xmath9 can either be linear until a constant limiting value is almost achieved , as in fig .
[ fig : v - constant - kappa ] , or of the form @xmath10again up to a constant limiting value , as in fig .
[ fig : v - increasing - kappa ] .
let us recall that this second velocity profile corresponds to the schwarzschild line element with @xmath11 the central mass , as can be seen @xcite by reparametrizing the time coordinate and using @xmath12 as radial coordinate .
velocity profile of a black hole with a linear slope of the velocity @xmath7 , and hence a constant surface gravity @xmath13 , from the classical horizon at @xmath8 down to some predefined limiting value .
the auxiliary function @xmath14 separates each instantaneous velocity profile @xmath4 into a dynamical : @xmath15 and a stationary region ii : @xmath16 . ]
velocity profile characteristic of a schwarzschild black hole .
the slope of @xmath7 increases from the classical horizon at @xmath8 leftwards up to a constant limiting value , leading to a surface gravity @xmath13 which increases with the frequency . ] up to here , we have described a stationary profile .
now , to incorporate the dynamics of the collapse , let us introduce a monotonically decreasing function @xmath14 , , and define @xmath4 as @xmath17 imagining the collapse of a homogeneous star , the function @xmath14 represents the distance from the star surface to its gravitational ( schwarzschild ) radius . the dynamical configuration that we obtain consists of a series of snapshots . in each snapshot ,
@xmath18 increases ( i.e. , the fluid accelerates ) from @xmath19 for @xmath20 , up to a point @xmath21 , and then remains constant as @xmath22 further decreases towards @xmath23 . for consecutive snapshots , the point @xmath21 moves leftwards
, so @xmath4 covers an ever larger part of @xmath7 . from the point of view of an outgoing ( right - moving ) particle of light / sound , the configuration is nicely split up into , first , a dynamical or @xmath24-dependent region , and then a stationary @xmath22-dependent region , as defined in .
the difference between the two types of profiles , figs .
[ fig : v - constant - kappa ] and [ fig : v - increasing - kappa ] , can best be explained in terms of the surface gravity . when modifying the dispersion relation , the horizon becomes a frequency - dependent concept : each frequency experiences a different horizon , as we will see explicitly in section [ s : mdr ] .
in particular , for superluminal modifications , the horizon forms later ( i.e. , at higher values of @xmath25 , or more negative values of @xmath22 ) for increasing frequencies . the surface gravity
will then also become frequency - dependent : @xmath26 where @xmath27 is the frequency - dependent location of the horizon
. this surface gravity will be seen in section [ s : mdr ] and [ s : graphics ] to play a crucial role .
this explains our choice of the two types of profiles : in our first profile , fig .
[ fig : v - constant - kappa ] , we have considered a linear velocity profile @xmath28 from the classical or zero - frequency horizon ( @xmath29 ) down to some given limiting value , @xmath30 , from where it stays constant .
this maximum velocity defines a critical frequency @xmath31 as we will discuss in more detail in section [ s : mdr ] .
then the surface gravity @xmath13 will be frequency - independent up to this critical frequency @xmath31 after which it will rapidly vanish . in the second profile ,
see fig .
[ fig : v - increasing - kappa ] , we have taken a velocity profile typical for a schwarzschild black hole , and therefore @xmath13 increases with @xmath32 , again up to a predefined limiting value corresponding to the horizon for a critical frequency @xmath31 . in the context of the fluid analogy , it seems obvious that some mechanism will avoid the formation of a singularity .
but in any case , we will also take into account the limit for an infinite critical frequency , which corresponds to a velocity profile with @xmath33 .
in this section we will briefly review a way of deriving hawking s formula for the radiation of a black hole with standard ( relativistic ) dispersion relations of the form . for the sake of simplicity
, we will only consider a massless scalar field .
we will summarize the main steps : from the klein - gordon inner product and the definition of the bogoliubov @xmath34 coefficients , over the relation between past and future null coordinates , to the black - body radiation in the wave - packet formulation .
our aim is to present the key points of the procedure in such a way that they can easily be adapted to superluminally modified dispersion relations
the subject of the next sections .
the dalembertian or wave equation for a massless scalar field in 1 + 1 dimensions for the metric ( [ acoustic - metric ] ) with constant @xmath2 can be written as @xmath35 this conformally invariant theory is equivalent to the dimensionally - reduced 3 + 1 spherically - symmetric theory if one neglects the backscattering due to the angular - momentum potential barrier ( responsible for the so - called grey - body factors ) . in the space of solutions of this wave equation
, we can define the klein - gordon pseudo - scalar product @xmath36which is independent of the choice of the spatial slice @xmath37 . for a slice , and in particular for this
becomes @xmath38~.\]]we can define future null coordinates @xmath39 and @xmath40 , such that , when , @xmath41 writing the inner product in terms of these null coordinates , gives , for the limit @xmath42 , @xmath43_{w=+\infty } \right .
\nonumber \\ & & \hspace{-10 mm } \left .
+ \int_{-\infty}^{+\infty } \d w ~ \left[\varphi_1 \partial_w\varphi_2^ * -\varphi_2^ * \partial_w\varphi_1\right]_{u=+\infty } \right\}~. \label{kg - product}\end{aligned}\]]note that , in the derivation of this formula , we have only made use of coordinate transformations , without any appeal to their null character or to geometrical tools such as conformal diagrams or the deformation of cauchy surfaces . this is important because in the case of modified dispersion relations , such geometrical concepts become problematic and actually can only be maintained in the context of a rainbow geometry
if at all .
similarly we could have used past null coordinates @xmath44 , which obey , when @xmath45 , @xmath46to calculate the inner product in the asymptotic past .
the right - moving positive - energy solutions associated with the asymptotic past and with the asymptotic future , normalized in the dirac - delta sense , can be expressed as @xmath47 respectively , where we use primes to indicate asymptotic past values .
the relevant bogoliubov @xmath34 coefficients encoding the production of radiation are defined as .
the mode mixing relevant for the hawking process occurs in the right - moving sector .
therefore , we only need to calculate the first term in the scalar product ( [ kg - product ] ) . then , plugging into the definition of @xmath34 and integrating by parts we obtain the simple expression @xmath48so that all the information about the produced radiation is contained in the relation @xmath49 . for a standard relativistic dispersion relation ,
it is well known that the relation between @xmath51 and @xmath52 for a configuration that forms a horizon ( in our case at @xmath8 ) can be expressed at late times as ( note that we use a subscript @xmath53 for all quantities associated with the horizon ) , where @xmath54 , @xmath55 and the surface gravity @xmath56 are constants .
we can define a threshold time @xmath57 at which an asymptotic observer will start to detect thermal radiation from the black hole .
this retarded time corresponds to the moment at which the function @xmath50 enters the exponential regime .
we can then rewrite the previous expression , valid for @xmath58 , as @xmath59 plugging this relation into and integrating in @xmath52 gives @xmath60 \exp[-i\frac{c\omega}{\kappa}\ln(\omega'a_0)]\nonumber\\ & & ~~\times \exp(-i\omega u_i)\exp(-{\pi c\omega \over 2 \kappa } ) \gamma(ic\omega/\kappa ) ~.\end{aligned}\ ] ] in order to obtain physically sensible results , it is a good precautionary measure to replace the monochromatic rays described until now by wave packets ( see e.g. @xcite or the discussion in @xcite ) .
positive - energy wave packets can be defined as @xmath61 where @xmath62 with @xmath63 an overall reference and @xmath64 an integer phase parameter .
the central frequencies of the wave packets are , with @xmath65 their width . then , from the expression and assuming that the wave packets are sufficiently narrow ( @xmath66 ) , we obtain @xmath67 where we have defined @xmath68 finally , integration in @xmath32 gives the number of particles with frequency @xmath69 detected at time @xmath70 by an asymptotic observer : @xmath71which reproduces hawking s formula ( in the absence of backscattering ) and corresponds to a planckian spectrum with temperature @xmath72 .
in this section we will indicate how the late - time radiation originating from the formation of a black hole from collapse can be calculated in the case of superluminal dispersion relations .
we introduce superluminally modified dispersion relations by adding a quartic term to the wave equation : @xmath73 where @xmath74 ( the ` planck scale ' ) is the scale at which non - relativistic deviations in the associated dispersion relation @xmath75become significant .
we use this relation for concreteness and because it is the one that shows up in bose einstein condensates , where @xmath76 , with @xmath77 the healing length of the condensate .
however , qualitatively , our results will not depend on the specific form of the deviations from the relativistic dispersion relation but on their superluminal character .
this dispersion relation leads to a modification in both the phase velocity @xmath78 and the group velocity @xmath79 .
behaviour of the effective phase @xmath80 and group @xmath81 speeds of light / sound with respect to the wave number @xmath82 . due to the non - relativistic dispersion relation ,
the effective velocities become ` superluminal ' for @xmath83 ( where @xmath74 is the planck scale ) . ] for a right - moving wave , we have @xmath84 where we have introduced the effective @xmath82-dependent phase and group speeds of light / sound @xmath85 respectively ( see fig . [
fig : c_k ] ) . both @xmath80 and @xmath81
become larger than @xmath2 ( i.e. , become ` superluminal ' ) as @xmath82 increases above @xmath74 . at first sight
, it seems obvious that the ray equation should be defined in terms of the group velocity @xmath79 .
nevertheless , the question of whether the velocity relevant for hawking radiation is the group or the phase velocity seems to be tricky @xcite .
for example , below suggests that the phase velocity might be relevant .
we limit ourselves to remark that @xmath81 and @xmath80 show the same qualitative behaviour .
hence our results are independent of this question and we will simply write @xmath3 ( or @xmath86 when wishing to emphasize the frequency - dependence ) . then , there will be a horizon , which becomes frequency - dependent , when @xmath87 , irrespectively of whether @xmath81 or @xmath80 is used for @xmath3 .
moreover , since @xmath3 becomes arbitrarily high for increasing wave number @xmath82 , given a certain @xmath30 at a particular instant of time , there will be a critical @xmath31 such that waves with an initial frequency @xmath88 do not experience a horizon at all .
the only exception to this rule occurs when the velocity profile ends in a singularity @xmath89 , which implies @xmath90 .
our aim is to calculate the black hole radiation with superluminally modified dispersion relations .
we will now repeat the main steps of section [ s : sdr ] , and point out where and how these modifications must be taken into account . the essential point with regard to the pseudo - scalar product
is the following .
its explicit form for @xmath91constant is not changed by the presence of the @xmath92 term in the wave equation ( see also the discussion in @xcite ) .
indeed , it is still a well - defined inner product , and in particular it is still a conserved quantity , since @xmath93 which can be seen , by repeated integration by parts , to vanish under the usual assumption that the fields die off asymptotically .
note that the modification of the dispersion relation singles out a preferred time frame : the ` laboratory ' time @xmath24 .
changing to another time @xmath94 will in general lead to a mixing between @xmath24 and @xmath22 , and hence the simple relations given here would no longer be valid . using the preferred time @xmath24 , and making exactly the same change of coordinates as in the case of standard dispersion relations
, we can again transform the inner product , evaluated at @xmath42 , into the expression in terms of @xmath52 and @xmath95 . as in the standard case ,
only the first term is relevant for the hawking process : @xmath96_{w=+\infty}~.\end{aligned}\ ] ] note that we are now using @xmath52 and @xmath95 merely as a perfectly good set of ( auxiliary ) coordinates , in order to cast the inner product into a useful form .
however , we can not associate them any null character such as they had in the case of standard dispersion relations .
let us define some sets of spacetime functions that will prove to be useful in what follows . given a fluid profile of the type described in section [ s : model ]
, we can integrate the ray equation @xmath97 starting from the past left infinity towards the right .
the ray initially has a frequency @xmath32 and an associated initial wave number @xmath98 , from which we can deduce the value of @xmath99 . in the left region , where the velocity profile is dynamic but position - independent ( @xmath100 , with @xmath14 the auxiliary function introduced in section [ s : model ] ) ,
@xmath101 can be considered as fixed while the frequency changes ( this is what happens in a mode solution of equation in this region ) .
then , we can define the function @xmath102 where @xmath103 is the instantaneous frequency of the particle at each time @xmath24 , defined through the dispersion relation and such that . when the ray reaches the kink it passes into a stationary region in which the velocity profile only depends on the position ( @xmath104 ) .
at the kink , the ray still has the initial wave number @xmath98 and a frequency @xmath105 . in its propagation towards the right this frequency now remains fixed while its wave number becomes a decreasing function @xmath106 of the position , such that @xmath107 , with @xmath108 the moment at which the kink is crossed . the final frequency of the ray
will then simply be @xmath105 and its final wave number @xmath109 . in this region ,
then , the function @xmath110 can be expressed as @xmath111 note here that @xmath105 and @xmath112 both depend on the initial @xmath32 . the same can be done by integrating the ray equation starting from the future . in this way we can define @xmath113 functions .
( the same procedure can be used to define @xmath114 and @xmath115 ) .
it is worth noting that @xmath116 and @xmath113 are not null coordinates in the usual geometric sense , since they are frequency - dependent , but they nevertheless share many properties with null coordinates .
the calculation of the inner product ( [ scalar_u ] ) involves the limit @xmath117 ( equivalently , @xmath118 ) .
so , the general form @xmath119 can be simplified to @xmath120 .
we can therefore change variables in the inner product from @xmath52 to @xmath121 .
the combination of the derivative and the integral means that the form of is preserved in the new integration variable @xmath121 .
then we can write the bogoliubov @xmath34 coefficients relevant for the hawking process as @xmath122_{w_{\omega}=+\infty}~. \label{kg - modified - product}\ ] ] now , assuming that the profiles vary slowly ( in scales much larger than the planck distance ) , the right - moving positive - energy modes associated with past and future infinity can be approximated by the following simple expressions : @xmath123 @xmath124 so that the bogoliubov coefficients read @xmath125 in analogy with the standard case , all information about the radiation is seen to be encoded in the relation @xmath126 . in this expression @xmath32 is the initial frequency of a ray at the past left infinity and @xmath127 its final frequency when reaching the future right infinity .
the previous approximation for the bogoliubov coefficients amounts to considering profiles that vary slowly both with @xmath22 and @xmath24 .
this is equivalent to considering large black holes .
in general the quartic term in the wave equation , or equivalently the quartic modification of the dispersion relation , introduce a new source of backscattering , on top of the usual angular - momentum potential barrier which we have already neglected . in our approximation , this additional backscattering ( beyond the standard grey - body factors ) has been neglected . for large black holes this contribution
will in any case be very small as has been observed in numerical simulations @xcite .
in addition we are also neglecting any reflection caused by the kink .
however , let us remark that given an approximative scheme for calculating @xmath126 for general profiles @xmath4 , the same @xmath126 obtained from a profile with a kink would be obtainable from one ( or several ) specific @xmath4 , this time perfectly smooth and thus causing no further backscattering .
our results , which rely only on the specific form of the relation @xmath126 , are therefore valid beyond the specific configurations with a kink presented in this paper .
our next task is to calculate the uniparametric family of functions @xmath128 and the relation between @xmath128 and @xmath121 for different configurations . as explained in section [ ss : rainbow ] , the relation @xmath126 is obtained by integrating the ray equation using the profiles discussed in section [ s : model ] .
these can be described by means of stationary profiles @xmath7 and an auxiliary function @xmath14 , see the definition of @xmath4 in .
we will use a straightforward extension of the procedure established in @xcite ( see also @xcite for a summary ) for a relativistic dispersion relation .
care must be taken , however , with the quantities that depend on the frequency @xmath32 .
in particular : * we will denote by @xmath27 and @xmath129 the position and the time at which the horizon associated with a particular initial frequency @xmath32 is formed . *
the surface gravity @xmath13 , defined in , allows to write , for all @xmath130 and for @xmath22 close to @xmath27 : @xmath131 up to higher - order terms in @xmath132 . *
the linearization of @xmath14 near @xmath129 also requires the introduction of an @xmath32-dependent parameter @xmath133 : @xmath134 again up to higher - order terms .
let us consider laboratory times @xmath135 such that @xmath14 has already crossed the classical horizon at : .
as we explained earlier we can define a critical frequency @xmath136 as the minimum initial frequency such that @xmath137 for all @xmath138 , or , in other words , the minimum frequency which at that particular time has not experienced any horizon yet . in the dynamical part of the profile , where @xmath139 , for rays crossing the kink just before the formation of the horizon , integration of the ray equation leads to @xmath140 for small values of @xmath141 ( see @xcite for details of the calculation in the case of a relativistic dispersion relation ) , where @xmath142 is the ( frequency - dependent ) time at which the kink separating the dynamical and the stationary regions is crossed , is the speed of light in the asymptotic past , and @xmath143 is the ray constituting the horizon associated with the frequency @xmath32 .
in the stationary part of the profile , where @xmath144 , we obtain , for , again using a method and notation based on @xcite , @xmath145~,\ ] ] where @xmath146 is the speed of light in the asymptotic future , and @xmath147 a bulk integration constant .
so @xmath148 both regions are connected by combining and . making use of
, this gives @xmath149 this relation is valid for frequencies for which a horizon is experienced , i.e. for @xmath130 and times @xmath150 , where @xmath151 is the threshold time defined in section [ ss : relation_u(u ) ] ( which , unsurprisingly , has become frequency - dependent ) .
again , as in the case of standard dispersion relations section [ ss : relation_u(u)]we can write @xmath152 where @xmath153 is essentially @xmath151 ( with possible higher - order corrections ) . assuming that the collapse takes place rapidly , it is a good approximation to replace @xmath153 by @xmath154 .
actually , as we will see shortly , this is a conservative estimate , in the sense that it slightly underestimates the superluminal correction to the radiation spectrum .
the relation @xmath155 we are considering interpolates between a linear behaviour at early times and an exponential behaviour at late times .
it has the same form as the relation @xmath50 for standard dispersion relations , and so we can continue following the steps of the standard case . in particular , the equivalent of is obtained by integrating out @xmath113 .
note that we use the subscript @xmath105 to emphasize that the @xmath113 are not the null coordinates of the standard case , but this should not be interpreted as an explicit function of @xmath105 and so does not complicate the integration steps in @xmath113 and @xmath105 .
however , when integrating @xmath156 in @xmath32 to obtain @xmath157 , see section [ ss : wave - standard ] , we must carefully consider the frequency - dependence of the relevant terms , i.e. , of @xmath158 , @xmath159 and @xmath13 .
the term carrying @xmath160 is moduloed away in , and we have replaced @xmath159 by @xmath161 , so the only relevant frequency - dependent factor that we are left with is the surface gravity @xmath13 .
moreover , because of the critical frequency @xmath31 in the horizon formation process , a finite upper boundary will also be induced in the integral .
indeed , frequencies @xmath88 do not contribute to the radiation at all , since they do not experience a horizon .
this is a delicate but crucial point .
it was already observed long ago by jacobson @xcite that trying to solve the transplanckian problem naively by imposing a cut - off frequency would seemingly extinguish hawking radiation on a relatively short time scale . in our case ,
however , this cut - off is not imposed _
ad hoc _ , but appears explicitly because of the superluminal character of the system at high frequencies .
moreover , the critical frequency , and hence the upper boundary induced in the integral , depend directly on the physics inside the horizon .
indeed , given a certain velocity profile , and in particular its behaviour near the center of the black hole , the critical frequency can be calculated by setting @xmath162 in eq . and extracting the corresponding critical frequency from the dispersion relation .
we will see this effect graphically in section [ s : graphics ] . in analogy with eqs . and
, we now obtain the number of particles detected for each frequency @xmath69 as @xmath163 \over \left[{\kappa_0 \over \kappa_{\omega'}}(z -z_{l,\omega'})\right]^2}\;{1 \over \exp({2\pi c\omega_j \over \kappa_{\omega'}})-1}\nonumber,\end{aligned}\ ] ] where now , with ( which corresponds to the standard @xmath164 of ) , and . changing the integration variable from @xmath32 to @xmath165 , we finally obtain the central expression in our analysis : @xmath166 \over \left[{\kappa_0 \over \kappa_{\omega'}}(z -z_{l,\omega'})\right]^2}\ ; { 1 \over \exp({2\pi c\omega_j \over \kappa_{\omega'}})-1}.\ ] ] note that , as we indicated earlier , the use of @xmath154 in the definition of @xmath167 is a conservative stance .
indeed , strictly speaking , we should write @xmath168 . for a fixed @xmath169 , a smaller value of @xmath167 , and hence a larger value of @xmath153 ,
means that a larger part of the central peak of the integrand will be integrated over .
since we are replacing @xmath153 by the upper bound @xmath154 , we are overestimating the resulting radiation ( i.e. , underestimating the modification with respect to the standard hawking radiation ) .
the expression brings out the two crucial factors mentioned earlier , and a third , corollary one . *
first , it shows the dependence of the total radiation on the critical frequency @xmath31 ( through the integration boundary @xmath169 induced by it ) , as discussed just before obtaining formula . * second , it shows the importance of the frequency - dependent @xmath13 ( to be compared with the fixed @xmath164 of the standard case ) . given a concrete profile @xmath4 , the frequency - dependence of @xmath13
can be derived explicitly , as we will discuss next .
* finally , as a corollary of the first point , it shows that , as @xmath70 ( and hence @xmath167 ) increases , a smaller part of the central peak of the integrand will be integrated over , so the radiation will die off as @xmath70 advances .
a careful analysis has shown that , so far , most formulas for the standard case could be adapted to superluminal dispersion relations by replacing the relevant magnitudes with their frequency - dependent counterparts . for example , @xmath170 was obtained by replacing @xmath51 and @xmath52 by @xmath116 and @xmath121 , and in particular @xmath54 , @xmath57 and @xmath164 by @xmath160 , @xmath159 and @xmath13 , respectively . given a concrete profile
, we can explicitly deduce the relation between @xmath13 and @xmath32 , with the planck scale @xmath74 as a parameter , as follows .
the horizon for a particular initial frequency @xmath32 is formed when @xmath171 i.e. , when @xmath172 , where we have used the phase velocity for concreteness . but
again , qualitatively , our results would be similar if taking the group velocity @xmath79 instead of @xmath78 .
taking into account that @xmath173 , the dispersion relation in the asymptotic past can be written as latexmath:[\[\begin{aligned } \label{omega - v_h } \omega'^2= given a concrete profile , @xmath27 can then be obtained and @xmath13 calculated explicitly . for the profiles of the first type , see fig .
[ fig : v - constant - kappa ] , the result is trivial . indeed , @xmath28 increases linearly between the horizon corresponding to @xmath175 and the one for @xmath31 , so we obtain a constant @xmath176 for all @xmath130 . for a schwarzschild profile as in fig .
[ fig : v - increasing - kappa ] , on the other hand , @xmath177 so we obtain @xmath178 note that @xmath13 remains nearly constant until frequencies of the order of magnitude of the planck scale are reached , and then starts to increase rapidly , see fig .
[ fig : kappa - omega ] . as we will see graphically in the next section
, this can have an important qualitative influence on the radiation spectrum .
surface gravity @xmath13 with respect to the ( logarithm of the ) frequency for a schwarzschild - type black hole as in fig .
[ fig : v - increasing - kappa ] , with the planck scale @xmath179 and the critical frequency @xmath180 ( where @xmath181 represents an arbitrary time unit . ) ] we now have all the tools necessary to compute and plot for different parameters of the profiles described in section [ s : model ] .
we have calculated eq . numerically using the gauss - chebyshev quadrature method .
the results are plotted in figs .
[ fig : cut - off][fig : u_l ] and perfectly illustrate the three important factors that we deduced theoretically in section [ ss : wave - mod ] . note that in all the figures we have plotted @xmath182 against @xmath105 to make visual comparison with the usual thermal spectra in 3 + 1 dimensions easier .
[ fig : cut - off ] shows the influence of the critical frequency @xmath31 for a profile with constant @xmath13 , as in fig .
[ fig : v - constant - kappa ] , and @xmath183 ( i.e. , immediately after the horizons have formed for all @xmath130 ) .
the constant @xmath13 guarantees that the form of the thermal spectrum is preserved . however , the intensity of the radiation decreases with decreasing critical frequency . actually , only for extremely high critical frequencies is the original hawking spectrum recovered . for a critical frequency still well above the planck scale
, the decrease can be significant .
for example , for @xmath184 , the peak intensity decreases by nearly 30% , while for @xmath185 , the decrease is approximately 40% . at the other end ,
note that extremely low critical frequencies still leave a significant amount of radiation .
for example , for @xmath186 , still 20% of the original peak radiation is obtained .
so hawking radiation receives significant contributions from an extremely wide range of frequencies .
influence of the critical frequency on the radiation spectrum for a black hole with velocity profile such that the surface gravity is constant , as in fig .
[ fig : v - constant - kappa ] , and different values of the critical frequency .
we have chosen @xmath187 , @xmath188 , @xmath189 , @xmath190 and the planck scale @xmath179 ( where @xmath181 represents an arbitrary time unit ) , which amounts to considering a solar - mass black hole . from top to bottom we have plotted @xmath191 .
note that the standard hawking spectrum coincides perfectly with the upper curve , which effectively corresponds to the absence of a critical frequency . ]
[ fig : increasing - kappa ] shows the radiation spectrum for a schwarzschild - like profile and hence increasing @xmath13 , as in fig .
[ fig : v - increasing - kappa ] , in particular for critical frequencies @xmath31 close to the planck scale . on top of the general decrease of the standard thermal part of the spectrum ( approximately 40% , as in the previous case ) due to the finite integration boundary induced by the critical frequency , the fact that the surface gravity is now frequency - dependent leads to an important qualitative change of the spectrum .
the high - frequency tail of the spectrum is totally transformed .
actually , if the critical frequency is sufficiently higher than the planck scale , the dominant source of radiation lies in the high - frequency region .
note that this effect is truly a consequence of the modification of the physics for frequencies above the planck scale .
this can be appreciated by noticing that , at @xmath192 , the whole tail - modifying effect has disappeared and the usual thermal form of the radiation spectrum is recovered ( although still with the quantitative decrease described above ) .
influence of a frequency - dependent surface gravity @xmath13 on the radiation spectrum for a black hole with schwarzschild - type velocity profile ( surface gravity increases with frequency ) , as in fig .
[ fig : v - increasing - kappa ] , and different values of the critical frequency around the planck scale @xmath74 : ( from top to bottom ) .
the standard hawking spectrum is depicted in dashed dotted line for comparison .
numerical values of @xmath193 , @xmath74 etc as in fig .
[ fig : cut - off ] . ] finally , fig .
[ fig : u_l ] shows the influence of the measuring time @xmath70 ( measured with respect to @xmath161 ) for a profile of the second type ( increasing @xmath13 ) .
it is clearly seen that the radiation dies off with time , and actually dies off rather fast . for an actual solar - mass black hole and an @xmath31 of the order of the planck scale ,
the radiation would last only a few tens of milliseconds .
note that this effect is a corollary of the existence of a critical frequency @xmath31 , since in its absence , the integration boundaries in would be infinite , and so the integral would be insensitive to a change @xmath194 . the combined effect of fig .
[ fig : increasing - kappa ] and fig .
[ fig : u_l ] leads to the following qualitative picture for the further collapse towards a singularity once the initial or classical horizon has formed .
as the interior gradually uncovers a larger portion of the schwarzschild geometry , two competing processes will take place .
on the one hand , the spectrum acquires ever larger contributions associated with higher and higher temperatures . on the other hand ,
the overall magnitude of the spectrum is damped with time .
the question of which process dominates would depend on the fine details of the dynamics of the collapse , and might be further complicated by backreaction effects , which we have not considered in our analysis .
influence of the measuring time @xmath70 on the radiation spectrum , for a schwarzschild - type velocity profile , see fig .
[ fig : v - increasing - kappa ] and @xmath195 , compare with fig .
[ fig : increasing - kappa ] .
different values of @xmath70 ( from top to bottom ) : @xmath196 with @xmath181 denoting an arbitrary time unit .
the standard hawking spectrum is depicted in dashed dotted line for comparison .
numerical values of @xmath193 , @xmath74 etc as in fig .
[ fig : cut - off ] . ]
we have discussed the hawking radiation for a collapsing configuration with superluminal dispersion relations .
modifications of the dispersion relation cause the horizon , and various associated quantities such as the surface gravity @xmath164 , to become frequency - dependent .
in particular , a critical frequency @xmath31 naturally appears such that frequencies higher than @xmath31 do not experience a horizon at all .
more generally , it also means that the standard geometric concepts traditionally used to study black holes must be handled with care .
nevertheless , through a detailed analysis , we have seen that the equations related to the late - time radiation can be adapted quite straightforwardly from standard ( relativistic ) to superluminal dispersion relations .
we analytically derived an approximate equation for the particle production at late times with superluminal modifications of the dispersion relations .
this equation clearly showed that important modifications in the late - time radiation should be expected , first , due to the existence of the critical frequency @xmath31 and the finite upper boundary it induces in the integration , and second , due to the frequency - dependence of @xmath13 .
we integrated numerically , and plotted the resulting spectrum , thereby confirming these expectations .
we have seen that the standard hawking spectrum is recovered only in a very particular case : in the limit when the critical frequency goes to infinity ( i.e. , when the profile for the velocity @xmath197 goes to a singularity ) and moreover the surface gravity @xmath13 is constant ( linear velocity profiles of the type of fig .
[ fig : v - constant - kappa ] ) . for lower critical frequencies ,
as long as @xmath164 remains constant , the thermal form of the hawking spectrum is maintained , but the intensity decreases rapidly with decreasing @xmath31 .
a non - negligible radiation persists , however , even for extremely low critical frequencies .
this clearly establishes the first point , namely the importance of the critical frequency @xmath31 , which illustrates the statement made in the introduction that with superluminal dispersion relations , the interior of the black hole is probed , which can significantly affect the hawking process . for a schwarzschild - like profile
, the velocity profile generally leads to a frequency - dependent surface gravity @xmath13 , increasing with the frequency @xmath32 .
this means that the standard hawking result can not be recovered for a schwarzschild black hole with superluminal dispersion relations , even when the velocity profile has a singularity . actually , when the velocity profile has a limiting value , the same quantitative decrease of the hawking part of the spectrum as before shows up .
moreover , if the critical frequency @xmath31 is above the planck scale , a drastic qualitative change of the radiation spectrum takes place , and for sufficiently high @xmath31 ( roughly a few times the planck scale ) the high - frequency part of the spectrum even becomes dominant .
this shows the importance of the surface gravity , which again illustrates the role played by the interior of the black hole .
finally , we have also seen that , as a corollary of the existence of a critical frequency and of the finite upper boundary induced by it in the thermal response function , the radiation spectrum dies off as time advances .
a few observations might be useful to connect our work with existing results on hawking radiation and its sensitivity to modified dispersion relations .
the most general observation is that the ` robustness ' of hawking radiation , which is often considered to be a well - established result , actually depends crucially on a series of assumptions .
these assumptions might be reasonable in the case of subluminal dispersion relations .
but for the case of superluminal dispersion relations , as we have shown explicitly , the assumptions needed to reproduce the standard hawking result depend on the physics inside the ( zero - frequency ) horizon , and moreover in a way which is not compatible with the schwarzschild geometry .
in particular , in @xcite , it was shown for a stationary scenario that hawking radiation is robust with respect to superluminal modifications of the dispersion relations , provided that positive free fall frequency modes were in their ground state just before crossing the horizon .
however , it was also admitted that it is not clear whether this is the physically correct quantum state condition . in @xcite
, three explicit assumptions were given for the previous condition to hold : freely falling frame , ground state and adiabatic evolution .
in the same article , it is similarly pointed out that these assumptions might fail for a superluminal dispersion relation , since ( as we also mentioned in the introduction ) for high - frequency modes this implies that one makes an assumption about the physics at the singularity .
rather than imposing any conditions on the state near the horizon , and hence ultimately at the singularity , we have fixed the _
initial _ geometry to be minkowskian and the quantum field to be in the natural associated vacuum state , and evolved this into a black hole configuration .
since our results are seemingly in contradiction with the ones obtained in @xcite and @xcite , it is worth examining in which sense the conditions stated by those authors are violated in our approach . if one considers a collapse scenario , for example of a bec , in a laboratory setting , then it is quite natural to assume that the `` freely falling frame '' condition will be violated .
indeed , the superluminal modification is in this case associated with the existence of a privileged external reference frame : the black hole rest frame or lab frame ( as we noted in sec .
[ ss : generalization_inner_product ] ) , and not the freely falling frame , as assumed in @xcite .
note that this violation of the free - fall frame condition automatically implies a violation of the ground state condition in the sense in which this condition is formulated in @xcite , namely as the ground state `` with respect to the freely falling frame '' .
indeed , our results suggest the following picture .
the low - energy modes experience the classical geometry , are therefore dragged along in the free - falling frame and so at the horizon they occupy the vacuum state associated with this free - falling frame , namely the unruh vacuum .
hence they contribute to the black hole radiation in the traditional hawking way .
however , the ultra - high - frequency modes ( above the critical frequency ) do not see the horizon .
hence they do not couple to the classical geometry of the collapse , but rather remain connected to the external or laboratory frame , and therefore pass through the black hole ( nearly ) undisturbed .
the ground state of these high - frequency modes is then not the vacuum associated with the freely falling frame , but the boulware vacuum associated with the original minkowski geometry , or in other words , with the stationary reference frame of the lab .
so these frequencies above the critical frequency do not contribute to the thermal output spectrum .
the overall radiation is a convolution of the contributions from all the different initial frequencies , where the surface gravity for each frequency can be interpreted as leading to an effective weighting factor .
depending on the internal physics of the black hole , this leads to either a spectrum of the traditional hawking form but with a reduced intensity ( in the case of a constant surface gravity ) , or a modified spectrum where the high - frequency contributions dominate ( as for a schwarzschild profile , provided that the critical frequency lies sufficiently above the planck scale ) . in any case , what our analysis shows with respect to the situation treated in @xcite and @xcite is the following .
the robustness of hawking radiation discussed by those authors is the robustness with respect to a superluminal modification under the precise conditions occurring in the standard case , in which all modes occupy the ground state associated with the free - falling frame just before crossing the horizon .
this assumption implies that all modes , regardless of their frequency , occupy the unruh vacuum .
our analysis shows precisely that the superluminality can lead to a spontaneous breaking of these conditions .
the natural evolution of a collapsing configuration is such that the frequencies above the critical frequency @xmath31 are not in the unruh vacuum associated with the freely falling frame , but remain in the boulware vacuum associated with the initial asymptotic minkowski condition .
it should then come as no surprise that this can indeed have an important impact on the resulting radiation spectrum .
we should mention that in a recent paper by schtzhold and unruh @xcite based on an alternative approximation method , a new potential problem with superluminal dispersion relations is highlighted .
this `` ultra - violet catastrophe '' is associated with a higher - than - linear growth of the group velocity at large wave numbers , and should therefore not occur for the dispersion relation used in our work .
however , we have found that a schwarzschild interior geometry could give place to the same sort of phenomenology .
a final observation concerns the recent article @xcite , in which the authors numerically simulate the formation of an acoustic horizon in a bec , and analyze the creation of correlated pairs of phonons through the so - called truncated wigner method .
our findings for the case of a constant surface gravity seem to be in qualitative disagreement with the discussion presented in @xcite , since the authors of this paper assert having observed a stationary hawking flux , while our analysis concludes that the radiation should quickly fade off .
the source of this apparent discrepancy might reside in the fact that the correlation function which they study is normalized , and therefore probes the form of the spectrum ( which we also find stationary with time ) , but does not provide details about the net amount of particle production .
we kindly thank stefano liberati and matt visser for enlightening comments and discussions , and jos luis jaramillo for very useful suggestions on the numerical aspects of this work .
g.j . would also like to express his gratitude to fernando barbero , daniel gmez vergel and lorenzo sindoni for useful discussions , and sissa ( trieste ) for hospitality .
financial support was provided by the spanish mec through the projects fis2005 - 05736-c03 - 01 , fis2005 - 05736-c03 - 02 and fis2006 - 26387-e .
99 t. jacobson , s. liberati and d. mattingly , `` lorentz violation at high energy : concepts , phenomena and astrophysical constraints , '' annals phys . * 321 * , 150 ( 2006 ) [ arxiv : astro - ph/0505267 ] .
n. e. mavromatos , `` lorentz invariance violation from string theory , '' pos qg - ph:027 ( 2007 ) [ arxiv:0708.2250 [ hep - th ] ] .
r. gambini and j. pullin , `` nonstandard optics from quantum spacetime , '' phys .
d * 59 * , 124021 ( 1999 ) [ arxiv : gr - qc/9809038 ] .
a. ashtekar , `` loop quantum gravity : four recent advances and a dozen frequently asked questions , '' arxiv:0705.2222 [ gr - qc ] . c. barcel , s. liberati and m. visser , `` analogue gravity , '' living rev .
* 8 * , 12 ( 2005 ) [ arxiv : gr - qc/0505065 ] .
g. e. volovik , `` the universe in a helium droplet , '' clarendon press , oxford ( 2003 ) . f. r. klinkhamer and g. e. volovik , `` merging gauge coupling constants without grand unification , '' pisma zh .
fiz . * 81 * , 683 ( 2005 ) [ jetp lett . *
81 * , 551 ( 2005 ) ] [ arxiv : hep - ph/0505033 ] .
d. mattingly , `` modern tests of lorentz invariance , '' living rev .
* 8 * , 5 ( 2005 ) [ arxiv : gr - qc/0502097 ] . j. magueijo and l. smolin , `` lorentz invariance with an invariant energy scale , '' phys .
lett . * 88 * , 190403 ( 2002 ) [ arxiv : hep - th/0112090 ] .
s. w. hawking , `` black hole explosions , '' nature * 248 * , 30 ( 1974 ) .
s. w. hawking , `` particle creation by black holes , '' commun .
phys . * 43 * 199 ( 1975 ) .
i. agullo , j. navarro - salas , g. j. olmo and l. parker , `` short - distance contribution to the spectrum of hawking radiation , '' phys .
d * 76 * , 044018 ( 2007 ) [ arxiv : hep - th/0611355 ] . w. g. unruh , `` experimental black hole evaporation , '' phys . rev .
* 46 * , 1351 ( 1981 ) .
t. jacobson , `` black hole evaporation and ultrashort distances , '' phys .
d * 44 * , 1731 ( 1991 ) . t. jacobson , `` black hole radiation in the presence of a short distance cutoff , '' phys .
d * 48 * , 728 ( 1993 ) [ arxiv : hep - th/9303103 ] .
w. g. unruh , `` sonic analog of black holes and the effects of high frequencies on black hole evaporation , '' phys .
d * 51 * , 2827 ( 1995 ) .
r. brout , s. massar , r. parentani and p. spindel , `` hawking radiation without transplanckian frequencies , '' phys .
d * 52 * , 4559 ( 1995 ) .
[ arxiv : hep - th/9506121 ] .
s. corley and t. jacobson , `` hawking spectrum and high frequency dispersion , '' phys . rev .
d * 54 * , 1568 ( 1996 ) .
[ arxiv : hep - th/9601073 ] .
s. corley , `` particle creation via high frequency dispersion , '' phys .
d * 55 * , 6155 ( 1997 ) .
s. corley , `` computing the spectrum of black hole radiation in the presence of high frequency dispersion : an analytical approach , '' phys .
d * 57 * , 6280 ( 1998 ) .
[ arxiv : hep - th/9710075 ] .
w. g. unruh and r. schtzhold , `` on the universality of the hawking effect , '' phys .
d * 71 * , 024028 ( 2005 ) .
[ arxiv : gr - qc/0408009 ] .
r. schtzhold and w. g. unruh , `` on the origin of the particles in black hole evaporation , '' phys .
d * 78 * , 041504 ( 2008 ) [ arxiv:0804.1686 [ gr - qc ] ] .
m. visser , `` essential and inessential features of hawking radiation , '' int .
j. mod .
d * 12 * , 649 ( 2003 ) .
[ arxiv : hep - th/0106111 ] . c. barcel
, a. cano , l. j. garay and g. jannes , `` stability analysis of sonic horizons in bose - einstein condensates , '' phys . rev .
d * 74 * , 024008 ( 2006 ) [ arxiv : gr - qc/0603089 ] .
s. corley and t. jacobson , `` black hole lasers , '' phys .
d * 59 * , 124011 ( 1999 ) [ arxiv : hep - th/9806203 ] .
l. j. garay , j. r. anglin , j. i. cirac and p. zoller , `` black holes in bose - einstein condensates , '' phys .
lett . * 85 * , 4643 ( 2000 ) [ arxiv : gr - qc/0002015 ]
. l. j. garay , j. r. anglin , j. i. cirac and p. zoller , `` sonic black holes in dilute bose - einstein condensates , '' phys .
a * 63 * , 023611 ( 2001 ) [ arxiv : gr - qc/0005131 ] . c. barcel , s. liberati and m. visser , `` analog gravity from bose - einstein condensates , '' class .
* 18 * , 1137 ( 2001 ) [ arxiv : gr - qc/0011026 ] . c. barcel , s. liberati and m. visser , `` towards the observation of hawking radiation in bose - einstein condensates , '' int .
j. mod .
a * 18 * , 3735 ( 2003 ) [ arxiv : gr - qc/0110036 ] .
r. schtzhold , `` on the detectability of quantum radiation in bose - einstein condensates , '' phys .
* 97 * , 190405 ( 2006 ) [ arxiv : quant - ph/0602180 ] .
s. wster and c. m. savage , `` limits to the analogue hawking temperature in a bose - einstein condensate , '' phys .
a * 76 * , 013608 ( 2007 ) [ arxiv : cond - mat/0702045 ] .
r. balbinot , a. fabbri and s. fagnocchi , `` non - local density correlations as signal of hawking radiation in bec acoustic black holes , '' phys . rev .
a * 78 * , 021603 ( 2008 ) [ arxiv:0711.4520 [ cond-mat.other ] ] .
i. carusotto , s. fagnocchi , a. recati , r. balbinot and a. fabbri , `` numerical observation of hawking radiation from acoustic black holes in atomic becs , '' new j. phys .
* 10 * , 103001 ( 2008 ) [ arxiv:0803.0507 [ cond-mat.other ] ] .
s. wster , `` phonon background versus analogue hawking radiation in bose - einstein condensates , '' phys .
a * 78 * , 021601(r ) ( 2008 ) [ arxiv:0805.1358 [ cond-mat.other ] ] .
m. visser , `` acoustic black holes : horizons , ergospheres , and hawking radiation , '' class .
grav . * 15 * , 1767 ( 1998 ) [ arxiv : gr - qc/9712010 ] .
a. fabbri and j. navarro - salas , `` modeling black hole evaporation , '' imperial college press , london ( 2005 ) .
m. visser and s. weinfurtner , `` analogue spacetimes : toy models for `` quantum gravity '' , '' pos qg - ph:042 ( 2007 ) [ arxiv:0712.0427 [ gr - qc ] ]
. m. visser , `` emergent rainbow spacetimes : two pedagogical examples , '' arxiv:0712.0810 [ gr - qc ] . c. barcel , s. liberati , s. sonego and m. visser , `` quasi - particle creation by analogue black holes , '' class .
* 23 * , 5341 ( 2006 ) . [ arxiv : gr - qc/0604058 ] .
c. barcel , s. liberati , s. sonego and m. visser , `` hawking - like radiation does not require a trapped region , '' phys .
lett . * 97 * , 171301 ( 2006 ) .
[ arxiv : gr - qc/0607008 ] .
|
we analyze the hawking radiation process due to collapsing configurations in the presence of superluminal modifications of the dispersion relation . with such superluminal dispersion relations ,
the horizon effectively becomes a frequency - dependent concept .
in particular , at every moment of the collapse , there is a critical frequency above which no horizon is experienced .
we show that , as a consequence , the late - time radiation suffers strong modifications , both quantitative and qualitative , compared to the standard hawking picture .
concretely , we show that the radiation spectrum becomes dependent on the measuring time , on the surface gravities associated with different frequencies , and on the critical frequency .
even if the critical frequency is well above the planck scale , important modifications still show up .
|
considerable attention has recently been paid to two - state nano - mechanical systems @xcite and the possibility of observing quantum effects in them . in the experiments of rueckes _ et al . _
@xcite crossed carbon nano - tubes were suspended between supports and the suspended element was electrostatically flexed between two states .
_ @xcite propose to use an electrostatically flexed cantilever to explore the possibility of macroscopic quantum tunnelling in a nano - mechanical system .
@xcite suggest using the two buckled states of a nanorod as the two states and investigate the possibility of observing quantum effects . here
we consider a suspended elastic rod of rectangular cross section under longitudinal compression .
the compression is used to adjust the potential energy for transverse displacements from the harmonic to the double well regime as shown in the fig.[figpotential ] as the compressional strain is increased to the buckling instability @xcite , the frequency of the fundamental vibrational mode drops continuously to zero . beyond the instability ,
the system has a double well potential for the transverse motion .
+ the two minima in the potential energy curve describe the two possible buckled states at a particular strain @xcite and the system can change from one state to the other , under thermal fluctuations or quantum tunneling . in our earlier publications
@xcite we have used multidimensional transition state theory ( tst ) to derive expressions for the transition rate from one potential well to the other .
we now include the effect of friction on the reaction rate . for this , we follow the procedure @xcite used for the study of barrier crossing and other dynamical problems in the presence of friction .
we consider the normal modes and associated quantum properties of an elastic rectangular rod of length @xmath0 , width @xmath1 and thickness @xmath2 ( satisfying @xmath3 ) @xcite .
we assume that @xmath2 is smaller than @xmath1 so that transverse displacements @xmath4 only occur in the `` @xmath2 '' direction .
@xmath5 is the linear modulus ( energy per unit length ) of the rod and is related to the elastic modulus @xmath6 of the material by @xmath7 @xmath1 @xmath2 .
the bending moment @xmath8 is given by @xmath9 for a rod of rectangular cross section and @xmath10 is the mass per unit length .
we take the length of the rod ( uncompressed ) to be @xmath11 . as in the fig .
[ figb2 ] , we apply compression on the two ends , reducing the separation between the two to @xmath0 . then if @xmath4 denotes the displacement of the rod in the ` @xmath2 ' direction , the total potential energy is given by @xcite @xmath12=\frac 12\int\limits_0^ldx(f\kappa ^2(y^{\prime \prime } ) ^2+f\varepsilon ( y^{\prime } ) ^2)+\frac f{8l_0}(\int\limits_0^ldx(y^{\prime } ) ^2)^2+\frac { fl_0}{2}\varepsilon ^2.\ ] ] in the above @xmath13 is the strain , negative if compressive . we imagine that each segment of the rod ( labelled by @xmath14 ) is coupled to a collection of harmonic oscillators which form the reservoir .
each section has its own independent collection of harmonic oscillators .
so the total system , rod plus the reservoir has the energy @xcite
@xmath15+\frac 12\sum_{\alpha=1}^n \int\limits_0^ldx\nonumber\\ & \times \left\ { m_\alpha ( x)\stackrel{.}{\xi } _
\alpha ^2 + m_\alpha ( x)\omega _ \alpha ^2(x)\left ( \xi _ \alpha ( x)-\frac{c_\alpha ( x)y(x)}{m_\alpha ( x)\omega _ \alpha ^2(x)}\right ) ^2\right\}.\end{aligned}\ ] ] in the above , @xmath16 denotes the position of @xmath17-th harmonic oscillator of the reservoir , coupled to @xmath4 , the displacement of the rod at location @xmath14 .
it has a frequency @xmath18 and mass @xmath19 , @xmath20 determines the coupling of the @xmath16 to the @xmath4 .
the way , the coupling has been chosen , the barrier height remains unchanged . here
@xmath21 varies over the collection of harmonic oscillators , and we let @xmath21 vary from @xmath22 to @xmath23 .
to derive the quantum rate we use the following methodology . the hamiltonian given in eq .
( [ b87 ] ) may be treated as a quantum hamiltonian . for a finite discrete set of oscillators one may evaluate the quantum rate using quantum transition state theory under harmonic approximations .
the quantum transition state theory rate expression is given by @xcite @xmath24 in the above @xmath25 and @xmath6 are the partition function at the saddle point and at the reactant , and @xmath26 is the frequency of the ` unstable mode ' . following pollak @xcite , the quantum partition function at the saddle point @xmath27 at the saddle point we have @xmath28 real oscillators and one imaginary oscillator .
@xmath29 is the ` bath ' frequency and @xmath30 is the ` system ' frequency at the saddle point .
note the well known divergence of @xmath31 at low temperatures .
the quantum partition function at the reactant geometry @xmath32 in the above @xmath33 is the ` bath ' frequency and @xmath34 is the ` system ' frequency at the reactant geometry .
to find the equilibrium state , we extremise potential energy functional with respect to @xmath4 . for this we put @xmath36}{\delta y(x)}=0 $ ] .
this leads to the differential equation @xmath37=0\ ] ] and the hinged end points have boundary conditions @xmath38 and @xmath39 the only solution to eq .
( [ c126 ] ) is @xmath40 if @xmath41 where @xmath42 but if @xmath43 two more solutions are possible .
they are given by @xmath44 on substituting eq .
( [ c127 ] ) into eq .
( [ c126 ] ) , we find a nonlinear equation for @xmath45 , which on solution gives @xmath46 these two are the buckled states which are minima of the potential energy surface for @xmath43 , as all the normal modes around this are stable .
the solution @xmath40 is now a saddle point ( see next section for details ) .
we find the potential energy for the buckled states @xmath47 .
@xmath40 is the straight rod configuration and is the saddle point for transition from one buckled state to the other .
its potential energy @xmath48 is zero .
one can calculate the barrier height for the process of going from one buckled state to the other as @xmath49 the kinetic energy of the rod is @xmath50 , where @xmath51 is the mass per unit length .
we now look at small amplitude vibrations around each extremum @xmath52 , which could be the buckled or the saddle point .
one gets @xmath53}{\delta y(x_1)\delta
y(x)}% \right ) _ { y_{ext}(x_1)}\delta y(x_1,t).\ ] ] on evaluation we get @xmath54 where the operator @xmath55 is defined by @xmath56 using the boundary conditions for the hinged end points , we find the normal modes of the rod , @xmath57 . at the saddle point , @xmath58 @xmath59 . using this in eq .
( [ c32 ] ) , we obtain @xmath60 with n=1,2,3 .... the normal mode frequencies at the saddle point are given by ( @xmath61 is used to denote the saddle point ) @xmath62 @xmath63 .
@xmath64 is the unstable mode and it has the imaginary frequency @xmath65 , where @xmath66 for small amplitude vibrations around the buckled state , one has to put @xmath67 .
the normal modes are the same as in the eq .
( [ b33 ] ) , but the normal mode frequencies are different .
they are @xmath68 while @xmath69 the reaction rate using quantum transition state theory may be written as @xcite @xmath70 where , we have used the following expression for @xmath71 @xcite @xmath72 and the expression for @xmath73 is given by @xmath74 in the above @xmath75 and @xmath76 represents frequencies of normal modes of the system at reactant geometry and at the saddle point respectively in absence of bath . in case
when the system is coupled to the bath @xmath77 is the new ` bath ' frequency and @xmath30 is the new ` system ' frequency at the saddle point , also @xmath33 is the new ` bath ' frequency and @xmath34 is the new ` system ' frequency at the reactant geometry . to evaluate @xmath73
, we use the following two identities @xcite @xmath78 and @xmath79 following pallak @xcite one can prove the following identity @xmath80 where @xmath81 is an arbitrary number .
also @xmath82 we use the notation @xmath83 .
insertion of eq .
( [ b113 ] ) into eq .
( [ b112 ] ) and use of eq .
( [ b104a ] ) and eq .
( [ b104b ] ) with @xmath84 gives @xmath85 interchanging the order of the products and using ratio of eq .
( [ b104a ] ) and eq .
( [ b104b])with the identification @xmath86 gives the desired result .
@xmath87 on simplification we get @xmath88
as @xmath89 , @xmath90 and the reaction rate diverges [ see eq .
( 18 ) ] . so the rate expression in eq . ( 18 ) is valid only if one is not too near @xmath91 .
this is due to the setting in of the second buckling instability .
as the rod is compressed , first the mode @xmath92 becomes unstable and this is the first buckling instability and the rod buckles as a result of this .
the length at which this occurs shall be denoted by @xmath93 .
if one supposes that the rod is compressed further keeping the straight rod configuration , then at a length @xmath94 , the mode @xmath95 too would become unstable and this is the second buckling instability .
what happens here is a reaction path bifurcation for the crossing from one buckled state to the other and is very interesting . for @xmath96 , there is only one saddle point but for @xmath97 , this saddle point bifurcates into two and consequently the calculation of rate near the bifurcation is a challenging problem . in a similar fashion one
can have the third instability at a length @xmath98 etc .
but these present no problem as far as rate calculation is concerned ( see below ) . in order to analyze the rate near and beyond the second buckling instability , we proceed as follows .
we assume that the eq .
( 12 ) has solutions of the form @xmath99 using this , the elastic potential energy is given by @xmath100= & \frac{f\pi ^4(a_1 ^ 2 + 4a_2 ^ 2)^2}{8l^4l_0}+\frac{f\pi ^2a_1 ^ 2(\varepsilon -\varepsilon _
c)}{2l^2}\nonumber\\&+\frac{2f\pi ^2a_2 ^ 2(\varepsilon -4\varepsilon _ c)}{l^2}.\end{aligned}\ ] ] the two buckling instabilities are clearly evident from this expression - as each the coefficient of @xmath101 or @xmath102 changes sign from positive to negative . finding the extrema of this potential leads to the following solutions for ( @xmath103 , @xmath104 ) \1 .
( @xmath105 , @xmath105 ) : this is the straight rod configuration . between first and second buckling ( i.e. @xmath106 ) ,
this is the saddle point .
but after the second buckling , it is no longer a saddle , but it becomes a hill top .
it has an energy @xmath107 .
( @xmath108 , @xmath105 ) : these are the buckled states and both of them have the same energy @xmath109 . \3 .
( @xmath105 , @xmath110 ) : these are the two new saddle points that arise from the bifurcation of the one that existed for @xmath111 .
these two have the same energy @xmath112 . at these saddle points ,
the rod has a bent ( s - shaped ) geometry . beyond the second buckling instability ,
the barrier height is given by @xmath113 beyond the second buckling instability , away from the instability , one can do a normal mode analysis near the vicinity of the new saddles - there are two of them , both making identical contributions to the reaction rate . near the buckled state ,
the normal modes have the frequencies given in the eq .
( [ c36 ] ) and eq .
( [ c36a ] ) , while near the saddle , the frequencies are given by @xmath114 @xmath115 @xmath116 for @xmath117 . in the above @xmath118 and
has an imaginary frequency with @xmath119 now the quantum rate beyond the second buckling instability can be calculated taking the saddle to be the bent configuration .
@xmath120 where , we have used the following expression for @xmath121 @xcite @xmath122 and the expression for @xmath123 is given by @xmath124 we have multiplied the quantum rate by a factor of 2 to account for the fact that there are two saddles of equal energy .
it is interesting that the normal modes for this saddle retain their stability , irrespective of what the compression is .
the first mode is always unstable and other modes always stable for all values of @xmath81 .
therefore , this rate expression is valid for all values of @xmath125 - that is even through the third buckling instability . near the second buckling instability ( @xmath126 ) , both @xmath127 and @xmath128 vanishes ,
causing the rate to diverge [ see eq .
( 18 ) and eq .
the cure for this divergence is to go beyond the harmonic approximations for the first two modes .
our discussion here follows that of weiss @xcite . for the model described by the hamiltonian @xmath129+\frac 12\sum_\alpha \int\limits_0^ldx\nonumber\\ & \times \left\ { m_\alpha ( x)\stackrel{.}{\xi } _
\alpha ^2 + m_\alpha ( x)\omega _ \alpha ^2(x)\left ( \xi _ \alpha ( x)-\frac{c_\alpha ( x)y(x)}{m_\alpha ( x)\omega _ \alpha ^2(x)}\right ) ^2\right\}.\end{aligned}\ ] ] we note that @xmath19 is independent of @xmath14 and may be written as @xmath130 , @xmath131 too , as well as @xmath20 .
so we can write the above expression as @xmath132 + \nonumber\\ & \frac 12\sum\limits_\alpha \int dx\left\ { m_\alpha \stackrel{.}{\xi } _ \alpha ^2(x)+m_\alpha \omega _
\alpha ^2\left ( \xi _ \alpha ( x)-\frac{c_\alpha y(x)}{m_\alpha \omega _ \alpha ^2}\right ) ^2\right\}.\end{aligned}\ ] ] now let @xmath133 and @xmath134 then , @xmath135\nonumber\\ & + \frac 12\sum\limits_{\alpha , n}\left\ { m_\alpha \stackrel{.}{\xi } _ { \alpha , n}^2+m_\alpha \omega _ \alpha ^2\left ( \xi _ { \alpha , n}-\frac{c_\alpha a_n}{m_\alpha \omega
_ \alpha ^2}\right ) ^2\right\},\end{aligned}\ ] ] which decouples all the modes .
the hamiltonian for the first two modes coupled with bath is given by @xmath136\nonumber\\ & + \frac 12\sum\limits_\alpha \{m_\alpha \stackrel{.}{\xi } _
{ \alpha , 1}^2+m_\alpha \stackrel{.}{\xi } _ { \alpha , 2}^2+m_\alpha \omega _ \alpha ^2\left ( \xi _ { \alpha , 1}-\frac{c_\alpha a_1}{m_\alpha \omega _ \alpha ^2}\right ) ^2\nonumber\\ & + m_\alpha \omega _ \alpha ^2\left ( \xi _ { \alpha , 2}-\frac{c_\alpha a_2}{m_\alpha \omega
_ \alpha ^2}\right ) ^2\}\end{aligned}\ ] ] in the euclidean action ( @xmath137 ) contains contributions from the system ( @xmath138 ) , the reservoir ( @xmath139 ) and the interaction ( @xmath140 ) , @xmath141 with @xmath142 in this optimized action two modes are decoupled .
@xmath144 now we write euclidean action for the two modes ( with bath ) separately . for the first mode @xmath145 and for the second mode @xmath146 the expression for @xmath147
is equivalent to the expression for an inverted parabolic potential coupled to @xmath23 harmonic oscillatore .
so we have a system of ( @xmath148 ) harmonic oscillators and we can calculate @xmath26 using the method discussed in the classical calculation @xcite .
now we will calculate the partition function using the ` action ' defined in eq .
( [ b116c ] ) .
following weiss @xcite we define `` influence action '' @xmath149 , which captures the influence of the environment on the equilibrium properties of the open system .
@xmath150=s_{r}^{(e)}[{\bm \xi}]+s_{i}^{(e)}[a_2,{\bm \xi}]\ ] ] the stationary paths of action , which we denote by @xmath151 and by @xmath152 , obey the euclidean classical equations of motion @xmath153 we choose for convenience to periodically continue the paths @xmath154 , @xmath155 outside the range @xmath156 by writing them as fourier series @xmath157 where @xmath158 , @xmath159 and @xmath160 is a bosonic matsubara frequency .
substituting eq .
( [ b118 ] ) into eq .
( [ b116a ] ) and eq .
( [ b116 ] ) , we obtain @xmath161= & \sum\limits_{\alpha = 1}^n { \beta \hbar } \sum\limits_{n=-\infty } ^{n=\infty } \frac{m_\alpha } 2\nonumber\\&(\nu _
n^2\left| \xi_{\alpha,2 , n}\right| ^2+\omega _
\alpha ^2\left| \xi_{\alpha,2 , n}-\frac{c_\alpha } { m_\alpha \omega _
\alpha ^2}a_{2,n}\right| ^2).\end{aligned}\ ] ] next , we decompose @xmath162 into classical term @xmath163 and a deviation @xmath164 describing quantum fluctuations , @xmath165 in the second form , we have used the solution of the oscillator mode @xmath162 following from eq .
( [ b117 ] ) .
since @xmath154 is a stationary point of action , the term linear in the deviation is eliminated and we find an expression in which the quadratic forms of @xmath166 and @xmath104 are decoupled , @xmath167=s_{r , i}^{(e)}[{\bm y}]+s_{\inf l}^{(e)}[a_2],\ ] ] @xmath168&= \sum\limits_{\alpha = 1}^n { { \beta \hbar } } \sum\limits_{n=-\infty } ^{n=\infty } \frac{m_\alpha } 2(\nu _ n^2+\omega _ \alpha ^2)\left| y_{\alpha,2 , n}\right| ^2 \nonumber\\ & = \sum\limits_{\alpha = 1}^n\int\limits_0^{\beta \hbar } d\tau \frac{m_\alpha } 2 ( \overset{.}{y}_{\alpha,2 } ^2+\omega _ \alpha ^2y_{\alpha,2}^2),\end{aligned}\ ] ] @xmath169=\sum\limits_{\alpha = 1}^n { { \beta \hbar } } \frac{c_\alpha ^2}{2m_\alpha } \sum\limits_{n=-\infty } ^{n=\infty } \left ( \frac{\left| a_{2,n}\right| ^2}{\omega _ \alpha
^2}-\frac{\left| a_{2,n}\right| ^2}{(\nu_n ^2+\omega _ \alpha ^2)}\right).\ ] ] in the fourier series representation , the influence action ( eq .
( [ b123 ] ) ) takes the complete form @xmath170={\mu}{\beta \hbar } \sum\limits_{n=-\infty } ^{n=\infty } ( \zeta _ n\left| a_{2,n}\right|^2),\ ] ] where @xmath171 assuming ohmic friction @xcite @xmath172 we want to calculate the ratio of partition functions for the second mode at the saddle and at the reactant geometry . as all other modes are harmonic , there contributions to the rate is calculated using the same procedure as discussed in the section 3.1 . in our calculation
we use the ` action ' for the second mode at the saddle point , given in the eq .
( [ b116c ] ) . expanding the kinetic term into its fourier components ,
we get @xmath173\end{aligned}\ ] ] so the partition function for the second mode at the saddle point is given by @xmath174 \nonumber\\ & \times \gamma [ \frac{\beta \hbar } { 4\pi } ( \frac{4\pi } { \beta \hbar } + \gamma + \sqrt{\gamma ^2 - 4\omega _ 2 ^ 2(\overline{a}_2)})]\times q_{bath}.\end{aligned}\ ] ] we have already determined @xmath175 , @xmath176 and @xmath177 variationally in the absence of reservoir . for the second mode at the reactant geometry , the expression for the partition function is given below @xmath178 \nonumber\\ & \times \gamma [ \frac{\beta \hbar } { 4\pi } ( \frac{4\pi } { \beta \hbar } + \gamma + \sqrt{\gamma ^2 - 4\omega _ 2 ^ 2})]\times q_{bath}.\end{aligned}\ ] ] in the regime where @xmath179 , the rate may be calculated using ( transition state is assumed to be straight rod ) @xmath180 in the regime where @xmath181 , the rate may be calculated using ( transition state is assumed to be bent rod ) @xmath182
we now use the above formulae to calculate the rate of passage from one buckled state to the other , for @xmath183 rod of dimensions @xmath184,@xmath185 , @xmath186 , the case considered by carr _
the young s modulus and density of @xmath183 are @xmath187 and @xmath188 .
the first and second buckling instabilities occur at lengths which we denote as @xmath93 and @xmath94 respectively and their values are given by @xmath189 , @xmath190 .
in the regime @xmath191 , the saddle point has straight rod configuration and in the regime @xmath181 , the saddle point has bent configuration . at @xmath192
we do not show the rate calculated near the second buckling instability - as it is extremely low ( fig .
[ figqfric50t001 ] ) .
the equations for rates have product over all the infinite modes , as would be normal for a continuum model .
this however is not realistic , and this can be easily modified to account for discreteness of these rod .
we do this , by restricting the product to contain just the same number of normal modes as the perpendicular degrees of freedom for the discrete model .
we have thus taken contributions from first @xmath193 normal modes of the rod .
[ figqfric50t30 ] shows the quantum rate against compression , made at a temperature of @xmath194 .
our calculation including friction shows that friction lowers rate of conversion from one buckled state to the other .
in this paper we have discussed the buckling of a nano rod under compression and rate of it s conversion from one buckled state to the other using quantum version of multidimensional transition state theory . as compression increases , the rod buckles ( first buckling instability ) , and has now two stable states .
from one stable state it can go over to the other by thermal fluctuations or quantum tunneling . using a continuum approach
, we have calculated the rate of conversion from one state to the other using system plus reservoir model . the saddle point for the change from one state to the other
is the straight rod configuration .
the rate expression , however , diverges at the second buckling instability . at this point ,
the straight rod configuration , which was a saddle till then , becomes hill top and two new saddles are generated .
the new saddles have bent configurations and as the rod goes through further instabilities , they remain stable and the rate calculated according to the harmonic approximation around the saddle point remains finite . however
, this rate too , diverges near the second buckling instability . in the quantum transition state theory calculation we have calculated centroid partition function for the second mode , and derived expressions that are well behaved through the second buckling instability . using these expressions
, we have calculated the rates for nano - rods of dimensions @xmath184,@xmath185 , @xmath186 .
the author is greatful to prof .
k. l. sebastian for his continuous guidance and for suggesting this very interesting problem .
the author also thanks prof .
eli pollak for his comments on the manuscript .
m. roukes , scientific american , september , _ 48 _ ( 2001 ) .
h. g. craighead , science , * 290 * ( 2000 ) _ 1532_. h. park , j. park , a. k. l. lim , e. h. anderson , a. p. alivisatos , and p. l. mceuen , nature , * 407 * ( 2000 ) _ 57_. a. erbe , r. h. blick , a. tilke , a. kriele , and j. p. kotthaus , appl .
* 73 * ( 1998 ) _
3751_. a. n. cleland and m. l. roukes , appl .
* 69 * ( 1996 ) _
2653_. t. rueckes , k. kim , e. joselevich , g. y. tsewng , c. l. cheung , and c. m. lieber , science , * 289 * ( 2000 ) _ 94_. s. m. carr , w. e. lawrence , and m. n. wybourne , phys
b * 64 * ( 2001 ) _ 220101(r)_. l. euler , in _ elastic curves _ , translated and annotated by w. a. oldfather , c. a. ellis , and d. m. brown , reprinted from isis no . 58 xx(1 ) , 1774 ( saint catherine press , bruges , belgium ) .
p. hanggi , p. talkner , and m. borkovec , rev .
phys . * 62 * ( 1990 ) _ 251_. a. chakraborty , s. bagchi and k. l. sebastian , j. comput .
. nanosci . * 4 * ( 2007 ) _ 1_. a. chakraborty ,
phys . * 107 * ( 2009 ) _
1777_. u. weiss , _ quantum dissipative systems , series in modern condensed matter physics _ ( world scientific , singapore , 1998 )
. s. m. carr , w. e. lawrence , and m. n. wybourne , physica b * 316 * ( 2002 ) _
464_. s. m. carr and m. n. wybourne , appl .
* 82 * ( 2003 ) _
709_. w. e. lawrence , physica b * 316 - 317 * ( 2002 ) _
448_. eli pollak , chem .
* 127 * ( 1986 ) _ 178_. eli pollak , phys .
rev . a * 33 * ( 1986 ) _ 4244_. i. s. gradshteyn and i. m. ryzhik , _ table of integrals , series and products _ 4th ed .
( academic press , new york , 1980 ) .
|
we consider a suspended elastic rod under longitudinal compression . the compression can be used to adjust potential energy for transverse displacements from harmonic to double well regime . as compressional strain
is increased to the buckling instability , the frequency of fundamental vibrational mode drops continuously to zero ( first buckling instability ) . as one tunes the separation between ends of a rod , the system remains stable beyond the instability and develops a double well potential for transverse motion .
the two minima in potential energy curve describe two possible buckled states at a particular strain . from one buckled state it can go over to the other by thermal fluctuations or quantum tunnelling . using a continuum approach and transition state theory ( tst )
one can calculate the rate of conversion from one state to other .
saddle point for the change from one state to other is the straight rod configuration .
the rate , however , diverges at the second buckling instability . at this point ,
the straight rod configuration , which was a saddle till then , becomes hill top and two new saddles are generated .
the new saddles have bent configurations and as rod goes through further instabilities , they remain stable and the rate calculated according to harmonic approximation around saddle point remains finite . in our earlier paper classical rate calculation including friction has been carried out [ j. comput .
theor . nanosci .
* 4 * ( 2007 ) _ 1 _ ] , by assuming that each segment of the rod is coupled to its own collection of harmonic oscillators - our rate expression is well behaved through the second buckling instability . in this paper
we have extended our method to calculate quantum rate using the same system plus reservoir model .
we find that friction lowers the rate of conversion .
|
problems of convection induced by an internal heat source have been experimentally ( e.g. [ 5 ] ) and numerically ( e.g. [ 4 ] ) studied by many researchers .
the investigations concerned the effects of the heating and cooling rate , many investigations were performed for different conditions imposed on the lower and upper boundaries . in these cases
it is difficult to compare the evaluations of the heat fluxes in the heated fluid layer with the ones from the classical convection . the extent of the convection cells when the rayleigh number increases are greater than the ones in the classical rayleigh - bnard convection problem .
these problems are very important for the study of the motion in the atmosphere , influencing all other processes that take place here . in this paper a horizontal layer of viscous incompressible fluid with constant viscosity and thermal conductivity coefficients @xmath0 and @xmath1 is considered [ 6 ] . in this context , the heat and hydrostatic transfer equations [ 6 ] are @xmath2 @xmath3 where @xmath4 is the heating rate , @xmath5 , @xmath6 and @xmath7 are the potential temperature , pressure and density in the basic state . in the fluid ,
the temperature at all point varies at the same rate as the boundary temperature , so the problem is characterized by a constant potential temperature difference between the lower and the upper boundaries @xmath8 .
taking into account ( [ eq : heat ] ) this leads to the following formula for the potential temperature distribution [ 6 ] @xmath9.\ ] ] in nondimensional variables the system of equations characterizing the problem is @xmath10 where @xmath11 is the velocity , @xmath12 and @xmath13 are the temperature and pressure deviations from the basic state [ 6 ] , @xmath14 is the grashof number , @xmath15 is the prandtl number and @xmath16 is a dimensionless parameter characterizing the heating ( cooling ) rate of the layer .
the boundaries are considered rigid and ideal heat conducting , so the boundary conditions read @xmath17
our analytical study has two major steps : we deduce the eigenvalue problem associated with ( [ eq : ecuatie_nond])-([eq : b_conditions ] ) and then , we solve this problem using a modified method based on fourier series expansions of chandrasekhar [ 1 ] type . in [ 6 ]
the numerical investigations concerned the vertical distribution of the total heat fluxes and their individual components for small and moderate supercritical rayleigh number in the presence of a uniform heat source . in order to deduce the eigenvalue problem
let us consider the viscous incompressible fluid confined into a rectangular box bounded by two rigid walls : @xmath18 , @xmath19 , @xmath20 .
first we deduce the eigenvalue problem governing the linear stability equivalent to this problem .
we assume that any unknown function in ( [ eq : ecuatie_nond ] ) is of the form from [ 3 ] @xmath21 @xmath22 , where @xmath23 , @xmath24 , @xmath25 and @xmath26 are the box sizes . here
@xmath27 and @xmath28 are the number of cells in the @xmath29 and the @xmath30 direction .
using ( [ eq : normal_modes ] ) we have @xmath31 applying to the equation @xmath32 the curl operator , the system ( [ eq : ecuatie_nond ] ) becomes @xmath33 where @xmath34 and the boundary conditions are @xmath35 from @xmath36 , taking into account the boundary conditions ( [ eq : eigen_bc1 ] ) , we get @xmath37 . replacing this expression in @xmath38
, we obtain @xmath39 . in this way ,
the system ( [ eq : eigen_1 ] ) and the boundary conditions ( [ eq : eigen_bc1 ] ) reduce to the following two - point problem @xmath40 @xmath41
taking into account the boundary conditions imposed for the normal component of the velocity and for the normal component of the derivative of the velocity , a method based on fourier series expansions upon complete sets of orthogonal trigonometric functions that satisfy all boundary conditions it is not possible to apply . in this way , at first , we expanded the unknown function @xmath42 upon a complete set of orthogonal chandrasekhar type functions ( see chandrasekhar [ 1 ] ) , but here these functions introduced an extraperiodicity to the problem leading to the lose of one of the physical parameters , e.g. the heating ( cooling ) rate @xmath16 . in order to avoid this problem , let us modify the system ( [ eq : eigen_2 ] ) by a translation of the variable @xmath43 , @xmath44 , such that the eigenvalue problem becomes @xmath45 where @xmath46 , with the boundary conditions @xmath47 eliminating the unknown function @xmath42 between the two equations from ( [ eq : eigen_3 ] ) , the following differential equation in the unknown function @xmath48 only @xmath49 is obtained , with @xmath50 using the same method described in detail in [ 1 ] , we introduce a new function @xmath51 and rewrite the equation ( [ eq : ecuatie_in_t ] ) as a new system in the unknown functions @xmath48 and @xmath51 @xmath52 where @xmath53 is the rayleigh number . the rayleigh number is an eigenvalue in ( [ eq : sistem_simplu ] ) and @xmath54 represent the corresponding eigenvector .
we are interested in finding the smallest eigenvalue in ( [ eq : sistem_simplu ] ) , defining the neutral manifold .
taking into account the boundary conditions for @xmath48 , we set @xmath55 @xmath56 where @xmath57 is found from the equation @xmath58 equation ( [ eq : ecuatie_neomogena ] ) has the solution @xmath59 , \end{array}\ ] ] the unknown coefficients @xmath60 , @xmath61 , @xmath62 , @xmath63 , @xmath64 , @xmath65 are obtained from the system followed by imposing to the function @xmath57 to satisfy the boundary conditions imposed on @xmath48 @xmath66 replacing the obtained expressions for the coefficients @xmath67 , @xmath68 , @xmath69 in the equation @xmath70 and imposing the condition that the obtained equation be orthogonal to the functions @xmath71 , @xmath72 with respect to the inner product of @xmath73 , @xmath74 , we obtain an infinite system of algebraic linear equations in @xmath75 .
the secular equation is obtained by imposing the condition that the determinant of this system to vanish , i.e. @xmath76\big\| , \end{array}\ ] ] where @xmath77}{\pi^{2}(l - k)^{2}(l+k)^{2}}\ \ \
\textrm { if } k\neq l \end{array } , \right.$ ] @xmath78}{\pi(l - k)(l+k ) } \textrm { if } k\neq l \end{array } \right.$ ] , @xmath79 from the secular equation , keeping some of the parameters fixed , we obtain the neutral manifolds , e.g neutral curves or neutral surfaces .
in this case , a first approximation for the rayleigh number @xmath80 is obtained by taking @xmath81 and imposing to the first - order minor in the matrix associated with the infinite algebraic system in @xmath75 to vanish [ 1 ] .
but for @xmath81 the expression of the rayleigh number following from the secular equation is given by @xmath82}.\ ] ] in this expression the parameter @xmath16 does not occur so , we can not use this approximation .
we performed numerical evaluations for the rayleigh number for the second and the third approximation , i.e. @xmath83 and @xmath84 respectively .
the numerical results are presented in table 1 . from the approximate numerical evaluations
it can be seen that when the wavenumber @xmath85 is increasing , the approximate rayleigh number is also increasing .
when the wavenumber is kept constant an increase in the heating ( cooling ) rate parameter leads to a decreasing of the rayleigh number .
moreover , for values of the parameter @xmath16 greater than a certain value ( stabilized around 10 ) , the third approximation is greater , yet not very different , from the one obtained in the second approximation . in fig.1 the approximate neutral curve for the parameter characterizing the heating rate @xmath86
is presented .
when both parameters are varying the approximate neutral surface is obtained ( fig.2 ) . [ cols="^,^,^,^,^",options="header " , ] * table 1 * numerical evaluations of the rayleigh number for various values of the parameters @xmath16 and @xmath85 . for @xmath86.,width=188,height=170 ] we also verified the numerical results with a variational method : the rayleigh quotient method .
the obtained numerical results are also presented in table 1 .
however the approximations are limited by the difficult evaluation of the associated matrix for a large number of functions in the expansion sets .
also , in this cases , the convergence it is not always assured .
the advantage in using this method is that the expressions of the neutral manifolds are easy to obtain .
.,width=188,height=170 ] when @xmath87 the problem reduces to the particular case of rayleigh - bnard convection and the numerical evaluation lead us to a critical value similar to the classical value for the rayleigh number , i.e. @xmath88 for @xmath89 .
in this paper we are concerned with the stability of the mechanical equilibrium of a fluid layer , namely a problem of convection with uniform internal heat source in a parallelipipedic box .
the eigenvalue problem is obtained and then investigated in order to obtain the secular equation which gives the approximate stability limits .
numerical investigations were performed in order to establish the influence of the two parameters , the wavenumber and the parameter characterizing the heating ( cooling ) rate on the approximate values of the rayleigh number : when the heating ( cooling ) rate is increasing , the domain of stability is decreasing and an increase in the wavenumber enlarge the domain of stability .
we pointed out that the first approximation does not work in this case and remarked that some other classical method do not work either in this case .
|
motion in the atmosphere or mantle convection are two among phenomena of natural convection induced by internal heat sources .
they bifurcate from the conduction state as a result of its loss of stability . in spite of their importance , due to the occurrence of variable coefficients in the nonlinear partial differential equations governing the evolution of the perturbations around the basic equilibrium ,
so far these phenomena were treated mostly numerically and experimentally .
no rigorous study is known . in this paper
we realize for the first time such a linear study for the eigenvalue problem associated with those equations for a convection problem with an uniform internal heat source in a horizontal fluid layer bounded by two rigid walls .
our method uses fourier series expansions for the unknown functions .
numerical results and graphs are given showing a destabilizing effect of the presence of the heat source .
|
let us consider a circular membrane of the radius @xmath0 of incompressible , isotropic and hyperelastic material having the small constant width @xmath1 .
let a hydrostatic load @xmath2 be applied to the membrane with an inflexibly fixed contour . to describe the geometry of the deformed membrane ,
let us introduce the cylindrical coordinate system @xmath3 , the axis @xmath4 of which coincides with the symmetry axis of the membrane .
it follows from the axial symmetry of the problem that principal deformation directions at any point will coincide with meridians , parallels and normals to the deformed surface and all the parameters of the strained deformed state will be functions of initial distances @xmath5 of points of the membrane from its symmetry axis only .
denote the principal degrees of lengthenings in these directions by @xmath6 , @xmath7 and @xmath8 , respectively .
then , according to the nonlinear theory of elastic membrane @xcite , interior stresses @xmath9 and @xmath10 of the deformed shell in the direction of its meridian and parallel can be determined from the relations @xmath11 here @xmath12 and @xmath13 are axial and radial components of the radius of the generatix vector of the deformed shell , @xmath14 is a deformation energy function for the material of the shell , @xmath15 and @xmath16 are deformation invariants , and @xmath17 is the width of the membrane in the deformed state .
let us choose the deformation energy function @xmath18 in the form of its four - term approximation proposed by v.l .
biderman @xcite , i.e. , @xmath19 where @xmath20 @xmath21 are physical constants that are determined experimentally . by using function ( [ trotsenko:1.2 ] )
one can sufficiently precisely describe the law of deforming for a number of types of elastic materials up to their break .
if the origin of the coordinate system coincides with a pole of the deformed membrane , then the hydrostatic pressure acting on the shell , is of the form @xmath22 here @xmath23 is the density of liquid , @xmath24 is the acceleration of gravity , @xmath25 is a constant component of pressure in liquid , and @xmath26 is the pressure of gas onto the shell . let us introduce the following dimensionless parameters : @xmath27 w^*=w / c_1 , \qquad q^*=qr_0/(2c_1h_0 ) .
{ \end{array}}\ ] ] in what follows , we shall use these parameters omitting asterisks for the sake of simplicity .
movements corresponding to the static equilibrium state are singled out from all admissible movements by the fact that stationary values of the functional @xcite @xmath28ds\ ] ] correspond to them .
having placed the origin of the coordinate system @xmath3 at the center of the undeformed membrane , we present the hydrostatic load as @xmath29 where difference between the constants @xmath30 and @xmath31 in formula ( [ trotsenko:1.3 ] ) is due to the shift of the origin of the coordinate system . in this case , we should seek a stationary value of functional ( [ trotsenko:1.4 ] ) within the set of functions satisfying the boundary conditions @xmath32 utilizing usual tools of the calculus of variations , one can show that the euler equations for functional ( [ trotsenko:1.4 ] ) are : @xmath33 here @xmath34 and @xmath35 are principle curvatures of the deformed median surface of the membrane calculated by the formulae @xmath36 following rietz to find an extremum of functional ( [ trotsenko:1.4 ] ) we present functions @xmath12 and @xmath13 in the form of the following expansions : @xmath37 where in view of ( [ trotsenko:1.5 ] ) , the coordinate functions @xmath38 and @xmath39 have to obey the conditions @xmath40 then constants @xmath41 @xmath42 forming @xmath43-component vector @xmath44 will be determined from the nonlinear algebraic system @xmath45 in this case , components of the vector - function @xmath46 are of the form @xmath47s\hspace{1.5pt}ds , \vspace{3 mm } \\ \displaystyle g_{i+m}=\int\limits_0 ^ 1\left[u(\lambda_1,\lambda_2)\frac{dr}{ds } \frac{dv_k}{ds}+\left(u(\lambda_2,\lambda_1 ) \frac{\lambda_2}{s}+q\lambda_2\frac{dz}{ds}\right)v_i\right ] s\hspace{1.5pt}ds .
{ \end{array}}\ ] ] here @xmath48 we solve algebraic system ( [ trotsenko:1.9 ] ) by using the newton method according to which correction of an approximate solution at each iteration step is performed by the scheme @xmath49 where @xmath50 is the symmetric jacobi matrix of functions @xmath51 , @xmath52 of variables @xmath53 , @xmath54 , whose elements are calculated as follows @xmath55 \frac{du_i}{ds}\frac{du_j}{ds}+d\lambda_2\frac{dr}{ds}u_iu_j\right\ } s\hspace{1.5pt}ds,\ ] ] @xmath56 \frac{dv_i}{ds}\frac{dv_j}{ds } \right.\ ] ] @xmath57\frac{v_iv_j}{s^2}\right\}s\hspace{1.5pt}ds,\ ] ] @xmath58 @xmath59 @xmath60 \frac{du_i}{ds}v_j - d\lambda_2\frac{dz}{ds}u_iv_j\right\ } s\hspace{1.5pt}ds \qquad ( i , j=1,\ldots , m).\ ] ] so starting from the potential energy stationarity principle , we reduce the problem of determining finite deformations of the circular membrane to calculating integrals forming algebraic system and elements of the jacobi matrix with subsequent solving problems of linear algebra ( [ trotsenko:1.11 ] ) at each step of the newton iteration process
. the algorithm may be effective if one chooses properly systems of coordinate functions @xmath61 and @xmath62 that would allow one to get solutions with high precision , provided the dimension of the nonlinear algebraic system is not large . to this end
, let us expand solutions to be found in a taylor series in a neighbourhood of the point @xmath63 , using sequential differentiation of equilibrium equations and geometric relationships of the shell in order to get coefficients of the expansion . with regard to symmetry conditions at the pole of the deformed shell @xmath64 we can show that solutions boundered under @xmath65 have the following structure : @xmath66 taking into account expansions ( [ trotsenko:1.12 ] ) , sequences of coordinate functions @xmath61 and @xmath62 satisfying boundary conditions ( [ trotsenko:1.8 ] ) take the form @xmath67 let us present some results of calculations according to the algorithm proposed . to illustrate convergence of the rietz method , we write out values of functions @xmath12 and @xmath13 and their first two derivatives , depending on the number @xmath68 of approximations in expansions ( [ trotsenko:1.7 ] ) at the point @xmath69 for @xmath70 , @xmath71 in table 1
. ratios of elastic constants in biderman s potential are chosen as follows : @xmath72 in the last column of the table we give the relative error of the obtained approximation to the solution of the second equilibrium equation @xmath73 * table 1 * + cccccccc & & & & & & & + ' '' '' 1 & 0.7016 & 0.2865 & 0.2923 & 1.3966 & 1.4617 & 0.5408 & @xmath74 + 2 & 0.7824 & 0.3026 & 0.3975 & 1.4644 & 1.9279 & 0.7252 & @xmath75 + 3 & 0.7913 & 0.3064 & 0.4255 & 1.4753 & 2.0248 & 0.8345 & @xmath76 + 4 & 0.7926 & 0.3068 & 0.4350 & 1.4759 & 2.0421 & 0.8560 & @xmath77 + 5 & 0.7926 & 0.3069 & 0.4362 & 1.4758 & 2.0415 & 0.8591 & @xmath78 + 6 & 0.7926 & 0.3069 & 0.4362 & 1.4757 & 2.0406 & 0.8596 & @xmath79 + solutions possess the similar convergence at other points of the interval of integration for these equations . lengthening of the generatrix of the deformed shell affects weakly on the degree of convergence of the variational method . for a chosen system of coordinate functions ,
solutions have the uniform convergence as well as their first two derivatives for sufficiently wide range of parameters .
figure 1 graphic dependence of the sag of the shell at its pole @xmath80 , upon the applied uniform pressure @xmath81 is shown in figure1 .
the characteristic indicates the possibility of existence of several equilibrium forms for the same value of the load .
it is clearly seen that the diagram `` load - sag '' for the material under consideration has two extreme points @xmath82 and @xmath83 .
if the parameter of the load increases monotonely and reaches the value of the upper critical load , then there will be an abrupt passage of the system into a new equilibrium state , after which sags of the shell will again increase smoothly .
the monotone decrease of the load @xmath30 ( unloading stage ) leads to the backward distortion of the shell to the initial equilibrium form at lower critical load .
equilibrium states described by a falling part of the curve from the point @xmath82 to the point @xmath83 are unstable and they are not realized at all in the way of loading under consideration . in conclusion
, we consider to the main principles of choosing an initial approximation of the newton iteration process that allows us to get solutions of nonlinear algebraic system ( [ trotsenko:1.9 ] ) .
if solutions of these equations are far from singular points @xmath82 and @xmath83 , then one can construct an initial approximation , taking advantage of rapid convergence of solutions . as we can see from table 1 , the first rietz approximation is a good solution to the initial problem . in this case , we have to solve a system of two nonlinear algebraic equations .
the initial approximation is determined from physically noncontradictory values of @xmath53 and @xmath84 ( to the positive load there corresponds the positive sag of the shell )
. then we use the obtained solution to construct @xmath85-th approximation , setting the first and @xmath86-th components of the vector @xmath44 equal to obtained values of @xmath53 and @xmath84 , and putting other components equal to zero . the initial approximation constructed in this way improved by employing the newton method with 45 iterations . in the case , when it is necessary to obtain a series of equilibrium states , when choosing initial values of the vector @xmath44 one can use different extrapolational formulae to prolong solutions in the parameter @xmath30 .
however , when implementing the procedure of prolonging a solution with approximation to critical points @xmath82 and @xmath83 , the convergence of the newton process sharply comes down , and then the divergence appears clearly . to go through these equilibrium states of the shell one should change prolongation parameters .
the main point of this approach is that starting from some moment , the parameter of the load is supposed to be unknown , and as a given parameter we choose the value of the sag of the shell at the pole .
artificial introduction of a new parameter provides that the uniqueness condition for a solution in a neighbourhood of a fixed parameter is fulfilled , and , therefore , jacobian is nondegenerate , which is a necessary condition for the newton method to be applicable .
the algorithm proposed is also effective for other rotation shells in the nondeformed state , the generatrix of which intersects the symmetry axis at the right angle and principle curvatures at its pole are equal .
let us consider the case of loading a membrane by a hydrostatic load for @xmath87 and @xmath88 ( heavy liquid ) .
for such parameters of the load , stresses in the median surface of the deformed shell will be small , i.e. , @xmath89 , and the shell will take a nearly plane form with the exception of a narrow domain near the fixed contour of the membrane .
the increase of the constant component of the load @xmath31 leads to the increase of stresses in the membrane , that begin to play the prevalent role in forming its deformed surface .
this is accompanied by increasing the width of a domain with rapidly changing functions of movements and stresses , and turning a plane part of the surface of the shell into a curvilinear surface . in the case under consideration , the presence of a small parameter at higher derivatives transfers the initial regular problem into a class of nonlinear singular perturbed problems . in this case
, solutions will have domains both with fast and slow variability . as a result , when approximating solutions by means of polynomial basis , the convergence of successive approximations in the rietz method sharply comes down , and the increase of the number of coordinate functions in expansions ( [ trotsenko:1.7 ] ) leads to a loss of calculation stability before the required precision is achieved .
the situation is also aggravate by that the singularity degree of the problem depends on a solution itself that is a priori unknown .
thus , there arise the principal problem of extending a class of admissible functions by functions that allow getting within the framework of a unique algorithm approximate solutions with the equally high precision both in the presence of a narrow area of great changes in solutions and without it .
the problem indicated plays a key role in realizing the method of variations when solving singularly perturbed boundary problems .
it is one of suffuciently difficult problems of mathematical physics and theory of function approximation .
apparently for this reason , variantional methods
( unlike finite difference methods @xcite ) are not developed properly for solving even linear boundary problems with a small parameter the higher derivative .
one of the most important merits of the method of variations is that when constructing approximate solutions , one can use an information about problem , that can be obtained by means of preliminary analysis of required solutions . on the other hand , methods of a small parameter
may be also used to determine a structure of a solution and characterize its degeneration with the parameter at the higher derivative tending to zero . with this objective in view ,
let us carry out an asymptotic analysis of the initial equations .
searching for a direct expansion of solutions to system ( [ trotsenko:2.1 ] ) in the form @xmath93 let us expand operators @xmath94 @xmath95 in a taylor series in the parameter @xmath96 @xmath97 having substituted ( [ trotsenko:2.2 ] ) and ( [ trotsenko:2.3 ] ) into ( [ trotsenko:2.1 ] ) and having equated coefficients of expansions of the same degrees of @xmath96 , we come to @xmath98 we can obtain a bounded solution of the nonlinear equation @xmath99 with respect to the function @xmath100 by expanding it in a taylor series in a neighbourhood of the point @xmath63 with regard to the symmetry at a pole of the deformed shell . in this case
, we can show that @xmath101 where @xmath102 is an arbitrary constant . collecting the results obtained ,
we come to the general form of a solution bounded for @xmath63 as a result of direct expansion ( [ trotsenko:2.2 ] ) : @xmath108 where @xmath109 is a function of @xmath96 ; @xmath30 is the constant component of the hydrostatic pressure , that differs from @xmath31 at the expense of that the origin of the coordinate system passes into the center of the nondeformed membrane .
proceeding from equations ( [ trotsenko:2.1 ] ) , it is not possible to establish the structure of solutions in the boundary area , since equations of the first approximation are nonlinear and they are not integrated in the explicit form .
for this reason , in what follows , using the results of the asymptotic analysis of the linearized equilibrium equations with respect to some equilibrium state that is described by an analytic function of the form ( [ trotsenko:1.12 ] ) , we propose to establish the structure of their solutions and after that to endow solutions of the nonlinear problem with these properties .
suppose that the membrane is in the deformed state as a result of applying the hydrostatic load @xmath110 .
let the shell have passed into a new state at the expense of changing the constant pressure component by a small value @xmath111 .
denote by @xmath112 and @xmath113 projections of a movement of the shell from this load onto directions of its generatrix and outer normal , respectively .
we shall characterize the main strained deformed state by stresses @xmath9 and @xmath10 , and principal degrees of lengthenings @xmath6 , @xmath7 and @xmath8 .
having performed linearization of elasticity relations ( [ trotsenko:1.1 ] ) , we can establish the following relation between additional stresses @xmath114 and @xmath115 and deformations of the shell : @xmath116 where @xmath117 @xmath118 @xmath119 @xmath120 @xmath121 @xmath122 @xmath123 linearized equilibrium equations in movements , describing the perturbed state of the shell are of the form @xmath124 + r\left(\frac{dr}{ds}w+\frac{dz}{ds}u\right)= \left(\varepsilon^2\delta c^*+\delta l\right)r\lambda_1 , { \end{array}}\ ] ] where @xmath125\frac{dr}{ds } , \qquad \alpha_4^{(2)}=-r\bigl[c_{11}k_1+(c_{12}+t_1)k_2\bigr ] , \vspace{3 mm } \\
{ \displaystyle}\beta_1=\frac{rt_1}{\lambda_1 } , \qquad \beta_2=r\lambda_1\left[(c_{11}-t_1)k_1 ^ 2+(c_{12}+c_{21})k_1k_2 + ( c_{22}-t_2)k_2 ^ 2\right ] , \vspace{3 mm } \\ { \displaystyle}\beta_3=-\alpha_3 , \qquad \beta_4=\beta_4^{(1)}+\frac{d\beta_4^{(2)}}{ds } , \qquad \beta_4^{(1)}=\bigl[c_{12}k_1+(c_{22}-t_2)k_2\bigr]\frac{dr}{ds } , \vspace{3 mm } \\ { \displaystyle}\beta_4^{(2)}=rt_1k_1 , \qquad \varepsilon^2=\mu , \qquad \delta l = w(0 )
. { \end{array}}\ ] ] one can show that the structure of a direct expansion of solutions to equations ( [ trotsenko:2.8 ] ) in the parameter @xmath126 corresponds to structure ( [ trotsenko:2.6 ] ) established for the system of initial nonlinear equations .
the subject of further research is to construct integrals with high variability of a system of linear equations with variable coefficients ( [ trotsenko:2.8 ] ) .
since a direct expansion is an approximation to some solution of initial equations ( [ trotsenko:2.8 ] ) , we shall seek boundary layer part of an asymptotics as an approximation to a solution of the relative homogeneous system of equations .
in so doing , we shall take into account that functions @xmath13 and @xmath12 characterizing the geometry of the shell in the main deformed state are analytic functions in a certain domain and according to ( [ trotsenko:1.12 ] ) they are expanded in power series in even and odd powers of @xmath5 , respectively . on the basis of this fact , we can rewrite the initial homogeneous system of equations as follows : @xmath127 where @xmath128 and @xmath129 are analytic functions for @xmath130 $ ] . proceeding from the form of system of equations ( [ trotsenko:2.9 ] ) ,
we conclude that this system has singularities both in the parameter and unknown variable .
the presence of a small parameter at the higher derivative results in existence of a narrow area in a neighbourhood of the point @xmath131 , in which solutions of the system have large gradients .
the point @xmath63 , in its turn , is a regular point for equations ( [ trotsenko:2.9 ] ) , that assumes the relative asymptotics of solutions when @xmath65 . since our final goal is to construct an algorithm for solving a problem with the uniform convergence in the parameter @xmath126 , in what follows it is necessary to take into account both of these facts .
we have not succeededin constructing solutions a singular perturbed equation , with a regular singular point , that belongs to the class of equations under consideration containing an exponent factor @xcite .
simplified equations obtained from the initial ones while only the first terms of expansion of their coefficients remain unchanged , may serve as a guiding line for an asymptotic representation of solutions to system ( [ trotsenko:2.9 ] ) . in this case
, we can establish that a regular solution of these equations is expressed by bessel functions of order zero and their first derivatives . according to what has been said , we seek solutions of homogeneous system of equations ( [ trotsenko:2.8 ] ) with a large variability index in the following form : @xmath132 here @xmath133 , @xmath134 @xmath95 and @xmath135 are functions to be determined . the advisability of representing solutions in the form ( [ trotsenko:2.10 ] ) will be warranted if functions @xmath133 and @xmath134 that we obtain are regular and can be presented as a direct expansion in the parameter @xmath126 .
having substituted expressions ( [ trotsenko:2.10 ] ) into homogeneous system of equations ( [ trotsenko:2.8 ] ) and set equal to zero coefficients of @xmath136 and @xmath137 , we get four equations for the functions @xmath133 , @xmath134 and @xmath138 : @xmath139 + rw_2r'+ru_2z ' \vspace{3 mm } \\
\displaystyle \qquad + \varepsilon\left[\beta_3u_1p-(\beta_1w_1p)'+\beta_1p \bigl(w_1p/\tau - w_1^\prime\bigr)\right]+\beta_1p^2w_2 \bigl(1-\varepsilon^2/\tau^2\bigr)=0 , \vspace{3 mm } \\
\displaystyle \varepsilon^2\bigl[l_{21}(u_1)+l_{22}(w_1)\bigr]+\varepsilon \left[(\beta_1w_2p)'+\beta_1p\bigl(w_2^\prime - w_2p/\tau\bigr)- \beta_3u_2p\right ] \vspace{3 mm } \\
\displaystyle \qquad + w_1p^2\beta_1+rw_1r'+ru_1z'=0 .
{ \end{array}}\ ] ] to determine functions @xmath141 , @xmath142 and @xmath138 we substitute expansions ( [ trotsenko:2.12 ] ) into equations ( [ trotsenko:2.11 ] ) and set coefficients of the same degrees of @xmath126 equal to zero . in the zeroth - order approximation ,
i.e. , for functions @xmath143 and @xmath144 , we have @xmath145 @xmath146 here @xmath147 is the angle formed by the outer normal to the deformed shell and its symmetry axis .
it follows from ( [ trotsenko:2.14 ] ) that @xmath138 is an even and pure imaginary function for @xmath148 .
the fourth equation with regard for ( [ trotsenko:2.13 ] ) takes the form @xmath149 therefore , after the zeroth - order approximation , the function @xmath152 remains a free function that will be determined at the second step of the asymptotic integration of equations ( [ trotsenko:2.11 ] ) . without to finding explicit expressions for terms in expansions ( [ trotsenko:2.12 ] ) of higher approximations
, one may conclude from the form of system ( [ trotsenko:2.11 ] ) with regard to expressions for its coefficients , that @xmath153 and @xmath154 are even functions , and @xmath155 and @xmath156 are odd ones .
specific expressions for these functions after the zeroth - order and first approximations show that they are regular for @xmath63 .
putting together the results obtained , we conclude that a solution to system ( [ trotsenko:2.8 ] ) , bounded for @xmath63 and having the large variability , can be presented in the form of the following formal expansions : @xmath157 where @xmath158 is a bessel function of the order zero of pure imaginary argument ; @xmath159 is a function regular and odd in argument @xmath5 .
the constructed asymptotic representations of solutions ( [ trotsenko:2.6 ] ) and ( [ trotsenko:2.16 ] ) quantitatively present the structure of a solution , that we use in a sequel to construct systems of coordinate functions when solving the initial problem by the method of variations .
taking into account the asymptotics of expansions ( [ trotsenko:2.16 ] ) for @xmath65 , and also the asymptotics of fucntions @xmath160 for large values of the argument , after passing from normal and tangent movements of the shell to its movements in axial and radial directions , the general form of solutions to the initial nonlinear problem can be presented as follows : @xmath161 where @xmath162 , @xmath163 and @xmath164 are arbitrary constants .
if constants @xmath164 are known , then solving nonlinear algebraic system ( [ trotsenko:1.9 ] ) by means of the iteration newton method , we can construct an approximate solution for the problem .
however , the question of specific values of parameters @xmath164 for the nonlinear boundary problem under consideration remains open .
hence , from the conditions @xmath166 , @xmath167 we can get @xmath168 additional equations with respect to parameters @xmath164 .
these equations are of the following form : @xmath169 s\hspace{1.5pt}ds .
{ \end{array}}\ ] ] it is convenient to look for a solution of system ( [ trotsenko:2.19 ] ) by the generalized secant method @xcite .
the whole algorithm is reduced to that at each step of changing parameters @xmath164 ( outer iteration process ) one needs to solve nonlinear algebraic system ( [ trotsenko:1.9 ] ) .
therefore , unlike the traditional rietz method of solving nonlinear boundary problems , a peculiarity of the given approach is that at the expense of choosing parameters @xmath164 that characterize the variability of required solutions , by using a computer , we construct a system of coordinate functions , that approximates solutions to the initial problem in a way optimal in some sense . in conclusion , let us present results of calculations , giving an idea of possibilities and effectiveness of the algorithm proposed . to illustrate the convergence of successive approximations , in table 2 we write out values of functions @xmath12 and @xmath13 and their first two derivatives depending on the number @xmath68 of terms in expansions ( [ trotsenko:1.7 ] ) at the point @xmath170 .
in the last column we give the relative error of an approximate solution for problem ( [ trotsenko:1.14 ] ) .
parameters of the hydrostatic load in the coordinate system associated with the center of the nondeformed membrane are @xmath171 , @xmath172
. ratios of constants in elastic potential ( [ trotsenko:1.2 ] ) are chosen as follows : @xmath173 all calculations are performed with regard for only one parameter @xmath174 in coordinate systems ( [ trotsenko:2.18 ] ) .
1 & 0.32896 & 0.18816 & 1.9342 & 0.90532 & 0.97550 & 0.38008 & @xmath175 + 2 & 0.36718 & 0.18415 & 2.4584 & 0.90694 & 0.99159 & 0.46372 & @xmath76 + 3 & 0.36533 & 0.18145 & 2.3754 & 0.90692 & 0.99250 & 0.42182 & @xmath76 + 4 & 0.36451 & 0.17825 & 2.3497 & 0.90693 & 0.99278 & 0.41433 & @xmath176 + 5 & 0.36450 & 0.17830 & 2.3498 & 0.90693 & 0.99276 & 0.41456 & @xmath177 + 6 & 0.36448 & 0.17841 & 2.3461 & 0.90693 & 0.99275 & 0.41404 & @xmath178 + solutions possess also the analogous convergence at other points from the interval of integration of the initial equations . here the high precision of an approximate solution is achieved at the expense of both increasing the dimension of algebraic system ( [ trotsenko:1.9 ] ) and choosing parameters of the optimal coordinate system .
it turns out that it is sufficient to restrict oneself to one parameter @xmath179 only .
calculations show that requirements to the precision of determining parameters @xmath164 concern with input data of the hydrostatic load with high singularity in solutions of the problem only . as the width of the boundary layer increases ,
the rate of convergence of the algorithm is preserved , but influence of the precision of determining parameters for the optimal coordinate system on the final result decreases .
there comes a moment when for some value of the hydrostatic load we do not need any more to organize the outer iteration process for seeking for values of parameters @xmath164 . in this case
, the rate of convergence of the given algorithm coincides with the rate of convergence of the algorithm based on the use of an exponent basis .
it is worth noting that a polynomial basis for the example considered above allows one to obtain the relative error @xmath180 for the first eight terms in expansions ( [ trotsenko:1.7 ] ) , that is equal to @xmath181 .
the further increase of the dimension of algebraic system ( [ trotsenko:1.9 ] ) in this case leads to the loss of the stability of calculations .
while performing the calculating experiment , it came out that functional ( [ trotsenko:1.4 ] ) in the space of parameters @xmath164 had only one minimum .
this means that for the approximation under consideration there exists only one optimal coordinate system , which considerably simplify its search by means of a computer .
therefore , the proposed approach allows one to obtain approximate solutions with a suffuciently high degree of precision for the small dimension of an algebraic system both in the presence of the singular perturbation of a problem and without it .
meridional cuts of the deformed shell for different values of the hydrostatic load @xmath30 are presented in figure2 .
the other input data coincide with those used for calculations presented in table 2 .
dashed lines correspond to paths of passage of points of the nondeformed membrane to points of its deformed surface .
curves of dependence of principal degrees of lengthenings @xmath6 and stresses @xmath9 ( continuous curves ) and also @xmath7 and stresses @xmath10 ( dashed curves ) upon the parameter @xmath5 for different values of the parameter @xmath30 are presented in figures3 and 4 . from figures
one can see that at the starting stage of deforming the membrane , the main part of its surface is nearly plane that is in the homogeneous biaxial strained state . only in a neighbourhood of the support contour of the shell its surface and intererior
stresses undergo considerable changes . as the constant component of the hydrostatic load @xmath30 increases , the plate part of the surface gradually decreases .
since it is convenient to investigate behaviour of material of elastomers for large deformations by means of experiments on homogeneous biaxial extension of a thin sheet @xcite ( that often causes some difficulties when realizing such a type of deformations for experimental tests ) , the presence of the homogeneous strained state in the central part of the deformed membrane under the load of heavy liquid may be used in establishing a functional dependence of an elastic potential upon deformation invariants .
gospodarikov a.b .
, terentev v.f . and chernyh k.f . ,
numerical investigation of some applied problems of elastic statics for axisymmetrical membrane of elastomers , _ mechanics of deformed media _ , 1979 , n 3 , 610 ( in russian ) .
trotsenko v.a . ,
axisymmetrical problem on equilibrium of a circle membrane under the hydrostatic load , in : physics and technical applications of boundary problems , kyiv , naukova dumka , 1978 , 126140 ( in russian ) .
lomov s.a . and
yudina a.s . , structure of the fundamental system of solutions for a singular perturbed equation with a regular singular point , _ izvestiya an sssr .
_ , 1982 , v.46 , n 5 , 11241134 ( in russian ) .
|
a number of important results of studying large deformations of hyper - elastic shells are obtained using discrete methods of mathematical physics @xcite@xcite . in the present paper , using the variational method for solving nonlinear boundary problems of statics of hyper - elastic membranes under the regular hydrostatic load , we investigate peculiarities of deformation of a circular membrane whose mechanical characteristics are described by the bidermann - type elastic potential .
we develop an algorithm for solving a singular perturbation of nonlinear problem for the case of membrane loaded by heavy liquid .
this algorithm enables us to obtain approximate solutions both in the presence of boundary layer and without it .
the class of admissible functions , on which the variational method is realized , is chosen with account of the structure of formal asymptotic expansion of solutions of the corresponding linearized equations that have singularities in a small parameter at higher derivatives and in the independent variable .
we give examples of calculations that illustrate possibilities of the method suggested for solving the problem under consideration .
journal of nonlinear mathematical physics 1999 , v.6 , n 1 , . [ multiblock footnote omitted ] * variational methods for solving + nonlinear boundary problems of + statics of hyper - elastic membranes * [ trotsenko - fp ] _ v.a .
trotsenko _ _ institute of mathematics of the national academy of sciences of ukraine , + 3 tereshchenkivska str . , 252601 kyiv-4 , ukraine _ _ received june 10 , 1998 ; accepted july 13 , 1998 _
|
the cooler synchrotron cosy @xcite at the forschungszentrum jlich in germany can accelerate protons and deuterons up to about 3.7 gev / c .
both , unpolarized and polarized beams are available .
excellent beam quality can be achieved using electron- and/or stochastic cooling .
cosy can be used as an accelerator for external target experiments and as storage ring for internal target experiments .
the strangeness production experiments have been performed at the internal spectrometer anke by the cosy - anke collaboration , at the internal cosy-11 spectrometer by the cosy-11 collaboration , at the external tof facility by the cosy - tof collaboration and at the external big karl spectrometer by the cosy - momo and cosy - hires collaborations .
extensive measurements of kaon - pair production have been performed at several cosy facilities . total and differential
cross sections are now available for a variety of reactions .
the world data set of total cross sections for kaon pair production is shown in fig .
[ kaon - pair ] , left panel , as a function of the excess energy @xmath5 . the @xmath6 reaction ( black symbols )
has been studied by the cosy-11 @xcite and cosy - anke @xcite collaborations at excitation energies between 3 and 108 mev . the closely related reaction @xmath7 ( green )
has also been studied by cosy - anke @xcite .
in addition the reaction @xmath8 ( red ) has been investigated by cosy - anke @xcite using a deuterium cluster - jet as neutron target .
the momentum of the non - observed proton spectator and the excess energy @xmath5 have been reconstructed from the four - momenta of the detected deuteron and kaon pair .
the @xmath9 ( pink ) reaction has been measured by the cosy - momo collaboration @xcite .
the @xmath10 reaction studied by cosy - anke is an ideal isospin zero filter
. it could be sensitive to the production of the scalar meson @xmath11 .
but the total cross section ( blue point ) amounts only to 5 pb @xcite .
the high energy @xmath6 result ( open circle ) has been measured at saturne @xcite .
the invariant mass distribution of @xmath12 pairs has been measured by cosy - anke @xcite at three excess energies @xmath13 and 108 mev .
the internal anke spectrometer detects simultaneously slow positive and negative particles in the side detectors and fast positive particles in the forward detector . the spectrum taken at @xmath14 mev ( @xmath15 gev ) is shown in fig .
[ kaon - pair ] , right panel .
all three spectra show a strong peak which is due to the production of the @xmath16 meson which decays with about 50 % into @xmath12
. the prompt @xmath12 distribution can be described by a 4-body phase - space distribution modified by the @xmath1 and @xmath17 final state interaction ( fsi ) .
all three distributions show a low mass enhancement over the fit curves .
this enhancement has been discussed as a coupled channel effect due to the @xmath18 channel whose threshold ( dashed vertical line ) is about 8 mev above the @xmath12 threshold @xcite .
the important @xmath17 fsi has been observed by comparing the invariant mass distributions of @xmath17 and @xmath19 @xcite .
the ratio of @xmath17 to @xmath19 production changes by an order of magnitude within 50 mev .
the same holds true if one compares @xmath20 and @xmath21 .
the @xmath17 fsi can be described by assuming an imaginary scattering length of 1.5 fm . in fig .
[ kaon - pair - phi - tot ] , the left panel shows the total cross sections for @xmath6 as a function of the excess energy @xmath5 .
the data are from cosy-11 @xcite , cosy - anke @xcite and saturne - disto @xcite .
the dashed curve represents a pure 4-body phase - space calculation .
the dot - dashed curve includes the @xmath1 fsi and the solid curve includes the @xmath17 fsi . the right panel shows the total cross section for @xmath0 production as a function of the corresponding excess energy @xmath5 . the data can not be described by a pure three - body phase - space behavior .
but the inclusion of the @xmath1 fsi is sufficient to reproduce the energy dependence @xcite .
another important result is the angular distribution of the @xmath0 decay into @xmath12 at @xmath22 mev which is a pure @xmath23 distribution in the jackson frame .
thus , the @xmath0 meson is tensor polarized with @xmath24 along the beam axis as expected near threshold due to conservation laws @xcite . comparing the total cross sections for @xmath0- and @xmath25-production at corresponding excess energies
@xcite yields a ratio @xmath26 which is about eight times larger than the prediction @xmath27 @xcite based on the okubo - zweig - iizuka ( ozi ) rule .
the @xmath12- and @xmath0-production was also studied by the cosy - momo collaboration at the magnetic spectrograph big karl using the @xmath28 reaction @xcite .
the invariant mass distributions measured at three excess energies , 35 , 40 and 55 mev , can be described by pure phase - space distributions .
effects due to a possible @xmath29 fsi are not visible . in comparison to the @xmath30 reaction @xcite
the @xmath0 peak is less pronounced .
the surprising result of the @xmath31 study is that the the @xmath0 meson is tensor polarized with @xmath32 along the beam axis @xcite .
in contrast , the @xmath25 meson is unpolarized in the corresponding reaction @xmath33 @xcite .
the ratio @xmath26 of total cross sections for @xmath34 and @xmath35 at corresponding excess energies is about a factor 20 larger @xcite than predicted by the ozi - rule @xcite .
exclusive measurements of hyperon production have been performed using the associated strangeness reactions @xmath36 , @xmath37 , @xmath38 and @xmath39 by the cosy-11 , cosy - anke and cosy - tof collaborations .
the aim of those measurements is to understand ( i ) the reaction mechanism , ( ii ) to study the effect of @xmath40 resonances and ( iii ) to study the fsi between the outgoing nucleon and hyperon .
the cosy - tof detector enables exclusive measurements with a large solid angle in the laboratory system and almost @xmath41 in the c.m .
recent measurements of the reaction @xmath36 at three bombarding energies @xcite are shown in fig .
[ dalitz - plot ] in the form of dalitz plots .
the dalitz plots show strong deviations from a homogeneous phase - space distribution .
the data can be described using the isobar model of sibirtsev ( concept outlined in @xcite ) which takes into account the @xmath4 fsi and the contribution of three @xmath40 resonances , @xmath42 , @xmath43 and @xmath44 .
the exclusive measurements of @xmath36 and @xmath37 by cosy - tof provide also differential cross sections @xcite in the cms , jackson and helicity frames , see fig .
[ ang - distr - lambda ] .
in addition , first measurements with polarized proton beams have been performed .
the cosy - tof data will be used for a detailed partial wave analysis by the bonn - gatchina pwa group . in fig .
[ canu - schulte - wissermann ] @xcite the total cross sections for @xmath36 and @xmath37 @xcite are shown as a function of the corresponding excess energy @xmath5 .
the dashed curves representing a three - body phase - space @xmath45 dependence describe the @xmath37 data and there is no evidence for a @xmath46 fsi .
however , the @xmath36 data can only be described by including the @xmath4 fsi .
the right panel shows the ratio of the @xmath36 to @xmath37 total cross sections as a function of the excess energy .
the new experimental values from cosy - tof @xcite confirm the general trend towards a ratio of 2.2 measured at high energies @xcite .
the large increase of the ratio towards low energies is well described by the @xmath4 fsi .
the first measurement of the reaction @xmath38 close to threshold has been performed by cosy-11 @xcite , by detecting the @xmath47 in coincidence with the neutron .
the total cross sections at @xmath48 and 60 mev were astonishingly high compared to those for @xmath37 .
therefore , the energy dependence of the reaction @xmath38 close to threshold has been studied recently by cosy - anke @xcite by detecting @xmath47 in coincidence with @xmath49 from the decay of the @xmath50 and by detecting @xmath51 coincidences .
in contrast to the cosy-11 results @xcite the total cross sections for @xmath38 are very much smaller ( factor 50 at @xmath52 mev ) and slightly smaller than those for @xmath37 ( @xmath53 ) . the energy dependence can be described by a three - body phase - space calculation @xcite , with no evidence for a @xmath54 fsi .
details can be found in the contribution of yu .
valdau to these proceedings .
a recent determination of the total cross section for @xmath38 at @xmath55 mev from inclusive @xmath47 data by cosy - hires @xcite yielded also a value smaller than expected from the cosy-11 results .
but the result is in conflict with the exclusive measurement by cosy - anke @xcite .
possible reasons for this discrepancy are discussed in @xcite .
the @xmath4 fsi in the reaction @xmath36 has been studied with a high invariant mass resolution by the cosy - hires collaboration @xcite .
the kaons were detected at @xmath58 using the magnetic spectrograph big karl @xcite .
the double differential cross section is shown in fig .
[ hires - spectrum ] as a function of the missing mass , i.e. the invariant mass of the @xmath4 system .
the spectrum is characterized by a huge fsi enhancement near the @xmath4 threshold .
a narrow @xmath59 dibaryon resonance predicted near 2100 mev @xcite is not visible .
the data can be described by factorizing the reaction amplitude in terms of a production amplitude and a fsi enhancement factor ( solid curve in fig .
[ hires - spectrum ] ) .
taking for the enhancement the inverse jost function yields the spin - averaged effective range parameters of the @xmath4 interaction , i.e. the scattering length @xmath60 fm and the effective range @xmath61 fm .
the dashed curve is the corresponding phase - space distribution without fsi enhancement .
taking the free @xmath4 scattering data @xcite in a combined fit into account yields the possibility to disentangle the spin singlet and triplet effective range parameters of the @xmath4 system @xcite .
however , this result should be considered with respect to the theoretical uncertainties of the jost - function approach @xcite .
a direct determination of the spin singlet and triplet effective range parameters requires polarization measurements as proposed by gasparyan et al .
@xcite and planned by cosy - anke and cosy - tof .
the results presented here are due to the efforts of the collaborations cosy - anke , cosy-11 , cosy - hires , cosy - momo and cosy - tof and the cosy machine crew .
the authors thank k .- th .
brinkmann , h. clement , w. eyrich , h. freiesleben , m. hartmann , j. ritman , e. roderburg , m. schulte - wissermann , a. sibirtsev , h. strher and yu .
valdau for their help and clarifying discussions .
r. maier et al . , _ nucl .
instr . and
meth . phys .
a _ * 390 * , 1 ( 1997 ) . m. wolke , `` production of associated strangeness in the reaction @xmath6 close to threshold '' _ jl-3532 _ ( 1998 ) . c. quentmeier et al . , _ phys
b _ * 515 * , 276 ( 2001 ) [ arxiv : nucl - ex/0103001 ] .
p. winter et al . , _ phys .
b _ * 635 * , 23 ( 2006 ) [ arxiv : hep - ex/0602030 ] . m. hartmann et al . , _ phys
lett . _ * 96 * , 242301 ( 2006 ) ibid . * 97 * , 029901 ( 2006 ) [ arxiv : hep - ex/0604010 ] .
y. maeda et al . , _ phys .
c _ * 77 * , 015204 ( 2008 ) [ arxiv:0710.1755 ( nucl - ex ) ] . v. kleber et al .
lett . _ * 91 * , 172304 ( 2003 ) [ arxiv : nucl - ex/0304020 ] .
a. dzyuba et al . , _ eur .
j. a _ * 29 * , 245 ( 2006 ) [ arxiv : nucl - ex/0605030 ] .
y. maeda et al . , _ phys .
lett . _ * 97 * , 142301 ( 2006 ) [ arxiv : nucl - ex/0607001 ] .
y. maeda et al . , _ phys .
c _ * 79 * , 018201 ( 2009 ) [ arxiv:0811.4303 ( nucl - ex ) ] .
f. bellemann et al .
c _ * 75 * , 015204 ( 2007 ) [ arxiv : nucl - ex/0608047 ] .
x. yuan et al .
j. a _ * 42 * , 1 ( 2009 ) [ arxiv:0905.0979 ( nucl - ex ) ] .
f. balestra et al .
c _ * 63 * , 024004 ( 2001 ) .
a. dzyuba et al . , _ phys .
b _ * 668 * , 315 ( 2008 ) .
m. silarski et al .
c _ * 80 * , 045202 ( 2009 ) . m. abdel - bary et al . , _ phys .
b _ * 647 * , 351 ( 2007 ) [ arxiv : nucl - ex/0702059 ] . h.j .
lipkin , _ phys .
b _ * 60 * , 371 ( 1976 ) .
k. schnning et al .
b _ * 668 * , 258 ( 2008 ) .
s. abd el - samad et al . , _ phys .
b _ * 688 * , 142 ( 2010 ) .
a. sibirtsev et al . , _ eur .
j. a _ * 27 * , 269 ( 2006 ) [ arxiv : nucl - th/0512059 ] .
s. abdel - bary et al . , _ eur .
j. a _ * 46 * , 27 ( 2010 ) .
d. grzonka and k. kilian , _ nucl .
a _ * 626 * , 41c ( 1997 ) .
balewski et al . , _ nucl .
a _ * 626 * , 85c ( 1997 ) .
balewski et al . , _ phys .
lett . b _ * 420 * , 211 ( 1998 ) .
s. severin et al . , _ phys .
* 83 * , 682 ( 1999 ) .
p. kowina et al . , _ eur .
j. a _ * 22 * , 293 ( 2004 ) .
m. abd el - samad et al . , _ phys .
b _ * 632 * , 27 ( 2006 )
valdau et al . , _ phys .
b _ * 652 * , 245 ( 2007 ) [ arxiv : nucl - ex/0703044 ] . w.j .
fickinger et al . , _ phys .
rev . _ * 125 * , 2082 ( 1962 ) .
a. baldini et al . , _ landolt - brnstein _ , new series , i/12b ( 1988 ) .
t. roek et al . , _ phys .
b _ * 643 * , 251 ( 2006 ) .
valdau et al . , _ phys .
c _ * 81 * , 045208 ( 2010 ) .
a. budzanowski et al . , _ phys .
b _ * 692 * , 10 ( 2010 ) . yu .
valdau and c. wilkin , submitted to _ phys .
b _ ( 2010 ) [ arxiv:1008.3465v1 ( nucl - ex ) ] .
a. budzanowski et al . , _ phys .
b _ * 687 * , 31 ( 2010 ) .
m. drochner et al .
a _ * 643 * , 55 ( 1998 ) .
a.t.m . aerts and c.b .
b _ * 253 * , 116 ( 1985 ) .
g. alexander et al . , _ phys .
_ * 173 * , 1452 ( 1968 ) .
b. sechi - zorn et al .
* 175 * , 1753 ( 1968 ) .
a. gasparyan et al .
c _ * 72 * , 034006 ( 2005 ) .
a. gasparyan et al . ,
c _ * 69 * , 034006 ( 2004 ) .
|
the paper gives an overview of strangeness - production experiments at the cooler synchrotron cosy .
results on kaon - pair and @xmath0 meson production in @xmath1 , @xmath2 and @xmath3 collisions , hyperon - production experiments and @xmath4 final - state interaction studies are presented .
address = hiskp , universitt bonn , nussallee 14 - 16 , d 53115 bonn , germany address = fakultt fr physik , universitt duisburg - essen , lotharstr . 1 ,
d 47048 duisburg , germany address = inst .
nucl .
physics , pan , ul .
radzikowskiego 152 , 31 - 342 krakw , poland
|
in this paper we consider two classical geometric control problems @xcite : the problem of euler s elasticae and the problem of sub - riemannian ( sr ) geodesics on @xmath0 .
solution curves to both problems have many applications in mechanics @xcite , robotics @xcite , image processing @xcite and modelling of human visual system @xcite .
although solutions are well known in geometric control community , the authors have noticed a common confusion in applied societies where people sometimes mix these two curves .
the reason is that the formulation of the problems is very similar , and at first sight one can wrongly deduce that euler s elasticae are obtained via reparametrization of sr - geodesics by arclength . to prevent this possible confusion we clarify that these two curves coincide only in the case of straight line .
the structure of the paper is the following .
first we briefly expose the history of the problems in introduction .
afterwards , in section [ sec : statement ] , we formulate both problems as optimal control problems on the lie group @xmath0 and discuss two natural parameterizations for the solution curves . in section [ sec : pmp ] we apply pontryagin maximum principle for both problems and prove the main result in theorem [ thm1 ] .
finally , in section [ sec : comparison ] , we present several simulations with comparison of elasticae and sr - geodesics , which support our theoretical result . in 1744
leonhard euler considered the problem on stationary configurations of an elastic rod with fixed endpoints and tangents at the endpoints @xcite .
euler obtained differential equations for stationary configurations of the rod and described their possible qualitative types .
these configurations are called euler s elasticae .
euler s elasticae are critical points of the elastic energy functional .
the question as to which of the critical points are minima ( local or global ) was answered in @xcite .
afterwards the software to the boundary value problem ( bvp ) was proposed in @xcite , and thus a numerical realization of the optimal synthesis was obtained .
the sub - riemannian problem on @xmath0 is formulated as follows . by
given two unit vectors @xmath1 and @xmath2 attached respectively at two given points @xmath3 and @xmath4 in the plane , to find an optimal motion in the plane that transfers @xmath5 to @xmath6 such that the vector @xmath7 is transferred to the vector @xmath8 .
the vector can move forward or backward and rotate simultaneously .
the required motion should be optimal in the sense of minimal length in the space @xmath9 , so called sr - length , where @xmath10 is the slope of the moving vector .
the problem can be seen as an optimal motion planning problem for the reeds - shepp car , which can move forward and backward and rotate on a place @xcite .
it has important relations to vision @xcite and image processing @xcite .
sr - geodesics are critical points of the sr - length functional .
explicit parametrization of the geodesics by jacobi elliptic functions and elliptic integrals was obtained in @xcite .
a study of optimality of the geodesics was done in @xcite , where the optimal synthesis was obtained . here and below in the text by
geodesic we mean a sub - riemannian geodesic on @xmath0 .
both curves ( elasticae and geodesics ) are determined by optimal control problems on the lie group @xmath0 , the group of euclidean motions of a plane .
the group @xmath0 is represented by @xmath11 matrices @xcite @xmath12 the group can be naturally identified with the coupled space of positions and orientations @xmath13 by identifying @xmath14 .
then for each @xmath15 one has the left multiplication @xmath16 via the push - forward @xmath17 of the left - multiplication one gets the left - invariant frame @xmath18 from the lie - algebra basis @xmath19 at the unity @xmath20 : @xmath21 for a curve @xmath22 on @xmath0 its projection @xmath23 on @xmath24 is called the _ spatial projection_. in this note we consider only so - called horizontal curves , that satisfy the horizontality condition @xmath25 .
there exist two natural parameterizations of horizontal curves on @xmath0 suitable for different needs ( see @xcite ) : * sub - riemannian ( sr ) arclength parameter @xmath26 is defined by @xmath27 , where @xmath28 for a given constant @xmath29 . here and
further in the text by dot we denote the derivative @xmath30 . *
spatial arclength parameter @xmath31 is defined by @xmath32 , where @xmath33 . here and further in the text
we denote by prime the derivative @xmath34 .
the spatial projection @xmath35 of a horizontal curve @xmath36 can have singularities cusp points , which appear when the velocity vector @xmath37 has nonzero component only in @xmath10 direction .
the spatial arclength parameterization is well - defined only on segments of @xmath36 whose spatial projection does not have cusps . for details
see @xcite .
euler s elastica problem is to find a @xmath38 $ ] curve @xmath39 \to { \operatorname{se(2)}\nolimits}$ ] , s.t . @xmath40
\gamma(0 ) = g_0 , \quad \gamma(s ) = g_1 , \end{array } \quad \int_0^{s } \frac{u(s)^2}{2 } \ , \d s \to \min,\ ] ] where the control @xmath41
\to \r$ ] is @xmath42 function , the initial point @xmath43 and the terminal point @xmath44 are given , and the terminal time @xmath45 is fixed and sufficiently large to guarantee that @xmath46 can be connected with @xmath47 .
notice that minimization of @xmath48 in the elastica problem is equivalent to minimization of @xmath49 for any @xmath50 .
thanks to left - invariance we can set @xmath51 without loss of generality . thus this problem is equivalent to @xmath52 \gamma(0 ) = e , \quad \gamma(s ) = g_1 , \end{array } \quad \int_0^{s } \frac{\xi^2 + u(s)^2}{2 } \ , \d s \to \min.\ ] ] sr - problem on @xmath0 is to find a lipschizian curve @xmath53 \to { \operatorname{se(2)}\nolimits}$ ] , s.t . @xmath54
\tgamma(0 ) = g_0 , \quad \tgamma(t ) = g_1 , \end{array } \quad \int_0^{t}\sqrt{\xi^2 u_1(t)^2 + u_2(t)^2 } \ , \d t \to \min,\ ] ] where the controls @xmath55 \to \r$ ] are @xmath56 functions , the initial point @xmath46 and the terminal point @xmath47 are given , @xmath50 is a constant and the terminal time @xmath57 is free . by cauchy - schwarz inequality and left - invariance
the problem is equivalent to @xmath58 \tgamma(0 ) = e , \quad \tgamma(t ) = g_1 , \end{array } \quad \int_0^{t}\frac{\xi^2 u_1(t)^2 + u_2(t)^2}{2 } \ , dt \to \min,\ ] ] where terminal time @xmath57 now is fixed .
[ remark : controls ] the control parameter @xmath59 represents the curvature of the spatial projection @xmath23 of a trajectory @xmath36 in ( [ eq : elasticasyst ] ) , while the control @xmath60 and @xmath61 represent the `` spatial '' and `` angular '' components of velocity vector @xmath62 in ( [ eq : geodcontsystsquared ] ) .
notice that @xmath35 is parameterized by spatial arclength @xmath31 , i.e. , @xmath32 for @xmath63 $ ] , while the spatial projection @xmath64 of a trajectory @xmath65 in ( [ eq : geodcontsystsquared ] ) satisfies @xmath66 for @xmath67 $ ] . comparing ( [ eq : elasticasyst ] ) and ( [ eq : geodcontsystsquared ] ) with account of remark [ remark : controls ] one may wrongly deduce , that if the spatial projection @xmath68 is reparameterized by spatial arclength @xmath31 ( corresponding to the control @xmath69 ) and the terminal time @xmath70 is chosen such that @xmath71 , then @xmath35 coincides with @xmath68 . although the substitution @xmath72 , @xmath73 , @xmath74 and @xmath75 in ( [ eq : geodcontsystsquared ] ) indeed gives ( [ eq : elasticasyst ] ) , we will show that @xmath35 coincides with @xmath68 only when @xmath76 , i.e. , when the both curves are segments of a straight line .
here we apply pontryagin maximum principle ( pmp ) @xcite to ( [ eq : elasticasyst ] ) and ( [ eq : geodcontsystsquared ] ) .
it can be shown , that abnormal extremals in ( [ eq : elasticasyst ] ) are given by straight lines , and they are contained in the set of normal extremals ( see @xcite ) . in ( [ eq : geodcontsystsquared ] ) abnormal extremals do not exist @xcite .
thus we consider only the normal case .
the control dependent hamiltonian of pmp in problem ( [ eq : elasticasyst ] ) reads as @xmath77 where @xmath78 denotes the action of a covector on a vector , and @xmath79 are basis one forms .
the maximization condition of pmp reads as @xmath80 where @xmath81 denotes an extremal control and @xmath82 is an extremal .
the maximization condition gives the expression for the extremal control @xmath83 then the maximized hamiltonian reads as @xmath84 the hamiltonian system with the hamiltonian @xmath85 is defined as @xmath86 thus , the hamiltonian system for problem ( [ eq : elasticasyst ] ) reads as @xmath87 \text{--- the vertical part , } & \text{--- the horizontal part , } \end{array}\ ] ] with the boundary conditions @xmath88 the control dependent hamiltonian of pmp in problem ( [ eq : geodcontsystsquared ] ) reads as @xmath89 the maximization condition of pmp reads as @xmath90 the maximization condition gives the expression for the extremal controls @xmath91 then the maximized hamiltonian reads as @xmath92 the hamiltonian system for problem ( [ eq : geodcontsystsquared ] ) reads as @xmath93 \text{--- the vertical part , } & \text{--- the horizontal part , } \end{array}\ ] ] with the boundary conditions @xmath94 [ rm : hamsysgeodspar ] switching to spatial arclength parameter @xmath95 ( well - defined before the first cusp ) leads to the following hamiltonian system : @xmath96 \text{--- the vertical part , } & \text{--- the horizontal part . }
\end{array}\ ] ] by analyzing ( [ eq : hamsys ] ) and ( [ eq : hamsysgeod ] ) one gets the following result .
[ thm1 ] let @xmath97\}$ ] be an elastica , and let @xmath98\}$ ] be a sr - geodesic such that @xmath99 for all @xmath100 $ ] .
further , let @xmath101 , and let @xmath102 , @xmath103 $ ] , @xmath104 , be the inverse function . if @xmath105 for some interval @xmath106 \cap [ 0,s]$ ] , @xmath107 , then the both curves @xmath108 and @xmath109\}$ ] are straight line segments .
\b ) for the maximally continued sr - geodesic @xmath123 , @xmath124 , define the first cusp time after @xmath125 : @xmath126 the function @xmath127 is real analytic and increasing for @xmath128 , thus one can define an inverse function @xmath129 , @xmath130 , @xmath131 so there exists a real analytic vector function @xmath132 , @xmath133 .
its image on @xmath24 is a curve @xmath134 , whose curvature is given by @xmath135 notice that at the cusp point @xmath136 and @xmath137 , thus @xmath138 . introducing a polar angle
@xmath139 $ ] in the plane @xmath140 , from the geodesic system one gets ( see @xcite for details ) : @xmath141 let @xmath142 , @xmath143 .
if @xmath144 then @xmath145 , thus identity ( [ eq : twocurves ] ) is impossible . if @xmath146 then the curves @xmath108 and @xmath134 are straight line segments . if @xmath147 then @xmath148 and if @xmath149 , then either ( [ eq : tcuspc1c2 ] ) or @xmath150 thus @xmath151 consequently , we can assume inequalities ( [ eq : tcuspc1c2 ] ) , possibly , after time reversal on @xmath152 .
\c ) suppose we have @xmath153 for all @xmath154 \subset [ 0,s_{cusp})$ ] , @xmath107 .
all the functions in ( [ eq : twocurvest ] ) are real analytic for @xmath155 and are analytically continued to the interval @xmath156 . by the uniqueness theorem for analytic functions , identity ( [ eq : twocurvest ] ) holds for @xmath157 .
thus @xmath158 , @xmath157 . but @xmath120 is bounded on @xmath159 , while @xmath138 , a contradiction .
in this section we support our theoretical result by series of simulations , where we compare elasticae and sr - geodesics . in the first simulation we fix the initial momentum @xmath160 ,
integrate the hamiltonian systems ( [ eq : hamsys ] ) and ( [ eq : hamsysgeodspar ] ) with this initial momentum , and plot the projection in the plane of the corresponding trajectories , see top row in figure [ fig : examp1 ] .
the experiment shows that sr - geodesic provides a good local approximation for the elastica in a neighborhood of the origin , when the same initial momentum was used for both systems . although , these two curves coincide one with another only in the case of straight line . in the second simulation
we show the difference between optimal elasticae and sr - minimizers ( optimal geodesics ) in solution of the boundary value problem ( bvp ) .
we organize the experiment as follows .
fix the initial point @xmath161 and the terminal point @xmath47 ; compute the sr - minimizer @xmath162 departing from @xmath46 and arriving at @xmath47 ; compute the length @xmath163 of the spatial projection @xmath164 of the minimizer @xmath162 ; compute the optimal elastica @xmath165 that connects @xmath46 with @xmath47 and has length @xmath163 ; plot the projection in the plane of both curves .
see the bottom row in figure [ fig : examp1 ] .
the experiment clearly shows the difference between these two curves , and again they coincide one with another only in the case of a segment of a straight line .
subsequently we show that the result is stable with respect to change of parameter @xmath166 that balance penalization of spatial and angular displacement of sr - geodesics , see the left column of the bottom row in figure [ fig : examp1 ] . ) and ( [ eq : hamsysgeodspar ] ) with the same initial momenta @xmath167 ( from left to right ) . *
bottom : * obtained as a solution to the boundary value problem with @xmath161 and @xmath168 ( from left to right ) , where the length of elastica is taken the same as the spatial length of sr - geodesic . in the left plot the parameter @xmath166 for sr - geodesics is varied .
this corresponds to different length of elasticae . ] ) and ( [ eq : hamsysgeodspar ] ) with the same initial momenta @xmath167 ( from left to right ) . *
bottom : * obtained as a solution to the boundary value problem with @xmath161 and @xmath168 ( from left to right ) , where the length of elastica is taken the same as the spatial length of sr - geodesic . in the left plot the parameter @xmath166 for sr - geodesics is varied .
this corresponds to different length of elasticae . ] ) and ( [ eq : hamsysgeodspar ] ) with the same initial momenta @xmath167 ( from left to right ) . *
bottom : * obtained as a solution to the boundary value problem with @xmath161 and @xmath168 ( from left to right ) , where the length of elastica is taken the same as the spatial length of sr - geodesic . in the left plot the parameter @xmath166 for sr - geodesics is varied .
this corresponds to different length of elasticae . ] ) and ( [ eq : hamsysgeodspar ] ) with the same initial momenta @xmath167 ( from left to right ) . *
bottom : * obtained as a solution to the boundary value problem with @xmath161 and @xmath168 ( from left to right ) , where the length of elastica is taken the same as the spatial length of sr - geodesic . in the left plot the parameter @xmath166 for sr - geodesics is varied .
this corresponds to different length of elasticae . ] ) and ( [ eq : hamsysgeodspar ] ) with the same initial momenta @xmath167 ( from left to right ) .
* bottom : * obtained as a solution to the boundary value problem with @xmath161 and @xmath168 ( from left to right ) , where the length of elastica is taken the same as the spatial length of sr - geodesic . in the left plot the parameter @xmath166 for sr - geodesics is varied .
this corresponds to different length of elasticae . ] ) and ( [ eq : hamsysgeodspar ] ) with the same initial momenta @xmath167 ( from left to right ) . *
bottom : * obtained as a solution to the boundary value problem with @xmath161 and @xmath168 ( from left to right ) , where the length of elastica is taken the same as the spatial length of sr - geodesic . in the left plot the parameter @xmath166 for sr - geodesics is varied .
this corresponds to different length of elasticae . ]
references v. jurdjevic , geometric control theory , cambridge univesrity press , cambridge , 1997 .
agrachev , yu.l .
sachkov , control theory from the geometric viewpoint , springer - verlag , 2004 .
l. euler , methodus inveniendi lineas curves maximi minimive proprietate gaudentes , sive solutioproblematis isoperimitrici latissimo sensu accepti , appendix i , `` de curvis elasticis '' ( lausanne , geneva , 1773 ; gtti , moscow , 1934 ) . y.l .
sachkov , optimality of euler s elasticae , doklady mathematics , vol .
76 ( 2007 ) , no .
3 , 817819 . a. ardentov and yu .
sachkov , solution to euler s elastic problem ( in russian ) , avtomatika i telemekhanika , 4 , 2009 pp .
( english translation in automation and remote control , 70(4 ) , pp . 633 - 643 . ) j.p .
laumond , nonholonomic motion planning for mobile robots , lecture notes in control and information sciences 229 . springer ( 1998 ) .
sachkov , i. moiseev , maxwell strata in sub - riemannian problem on the group of motions of a plane .
esaim control optim .
16 , 380 - 399 ( 2010 ) .
sachkov , cut locus and optimal synthesis in the sub - riemannian problem on the group of motions of a plane .
esaim control optim .
17 , 2 , 293 - 321 ( 2011 ) .
d. mumford , elastica and computer vision , in : algebraic geometry and its applications , c. bajaj ( ed ) , springer , berlin , 1994 pp 491 - 506 .
da chen , j .-
m . mirebeau and l.d .
cohen , global minimum for curvature penalized minimal path method . in xianghua xie ,
mark w. jones , and gary k. l. tam , editors , proceedings of the british machine vision conference ( bmvc ) , pages 86.1 - 86.12 .
bmva press , september 2015 .
mashtakov , a.a .
ardentov and y.l .
sachkov , parallel algorithm and software for image inpainting via sub - riemannian minimizers on the group of rototranslations .
numerical mathematics : theory , methods and applications . 2013 .
pp 95 - 115 .
bekkers , r. duits , a. mashtakov and g.r .
sanguinetti , a pde approach to data - driven sub - riemannian geodesics in se(2 ) .
siam journal on imaging sciences , 2015 , 8:4 , pp 2740 - 2770 .
g. citti and a. sarti , a cortical based model of perceptual completion in the roto - translation space , j. math .
imaging vis .
, 24 ( 2006 ) , 307326 .
r. duits , u. boscain , f. rossi and y.l .
sachkov , association fields via cuspless sub - riemannian geodesics in se(2 ) .
jmiv , 49 ( 2 ) , ( 2014 ) , 384417 .
sachkov , maxwell strata in the euler elastic problem , jdcs , 2008 , vol .
2 , 169234 . yu.l .
sachkov , conjugate points in euler s elastic problem , jdcs , 2008 , vol .
3 , 409439 . yu.l .
sachkov and e.l .
sachkova , exponential mapping in euler s elastic problem , jdcs , 2014 , vol .
20 , 125 .
sachkov , conjugate and cut time in sub - riemannian problem on the group of motions of a plane , esaim : cocv , 2010 , vol .
16 , 10181039 .
pontryagin , v.g .
boltyanskii , r.v .
gamkrelidze and e.f .
mishchenko , the mathematical theory of optimal processes , pergamon press , oxford ( 1964 ) .
|
in this note we describe a relation between euler s elasticae and sub - riemannian geodesics on @xmath0 . analyzing the hamiltonian system of pontryagin maximum principle
we show that these two curves coincide only in the case when they are segments of a straight line . +
* keywords : * elastica , sub - riemannian geodesic , group of rototranslations .
|
the mission of the terrestrial planet finder ( tpf ) is to find earth - like planets orbiting other stars by imaging , and to determine their atmospheric constituents by spectral analysis of their reflected light ( beichman 1999 ) .
imaging provides separation of the light reflected from the planet from that of the star . because the star is so bright compared to the planet ,
the dynamic range required is of order @xmath1 at visible wavelengths .
the `` book '' approach proposed for tpf is to operate in the infrared ( 7 to 14 microns ) where the dynamic range required is `` only '' @xmath2 . however , working in the ir increases the resolution requirement of the telescope or interferometer needed to separate the image of the planet from the image of the star .
a 20 meter interferometer baseline or telescope diameter may be sufficient in the visible band , but in the ir , a 200 meter telescope or 200 meter baseline interferometer is required .
such a large telescope is out of the question , and a 200 meter interferometer can not be built on a single structure , so free flyers are needed , making for very difficult construction and operation .
the proposed interferometer would be used in a mode where the starlight is `` nulled '' out in a narrow fringe .
this fringe is then swept around by rotating the interferometer , allowing the much fainter planet to be detected .
an optical telescope produces an image of an unresolved star that forms the familiar airy pattern which consists of a bright central peak surrounded by concentric rings of ever decreasing brightness .
these rings are due to diffraction from the sharp circular aperture of the telecope .
apodization is a technique that can be used to decrease brightness of the diffraction rings .
it can be accomplished by either altering the shape or by modifying the transmission of the aperture , or both . in this paper
we consider a simpler approach to tpf that achieves the dynamic range required to separate the light reflected from an earth - like planet from that of its star .
we suggest that apodizing a telescope with a filter having an _ optimized shape and transmission _ would sufficiently reduce the diffraction lobes of the telescope to allow detection of earth - like planets orbiting solar - like stars .
techniques for reducing optical diffraction lobes have been in use since the 19th century .
analagous algorithms are commonly used for sidebandreduction in radio astronomy and electronic signal processing applications ( harris 1978 ) and an excellent review of optical techniques is given in ( jacquinot 1964 ) .
oliver ( 1975 ) suggested apodizing the still - to - be - built large space telescope to allow that telescope to detect jupiter - sized exo - planets . in 1980 ,
black ( 1980 ) carried out laboratory experiments , showing that apodizing could be used to separate two point sources that had a @xmath3 ratio and a 1 arcsecond separation , comparable to the magnitude ratio and separation of sirius and its white dwarf companion .
watson et al ( 1991 ) suggest that apodization could be combined with a coronagraph in a space - based telescope to directly detect exo - planets .
we propose to use an approach that does not require nulling interferometry to achieve the dynamic range required for planet detection .
this approach provides sufficient dynamic range at separations of a few times the diffraction limit of a telescope at visible or near - ir wavelengths , so telescope size ( or interferometer baseline ) is greatly reduced .
our approach is elegantly simple .
we place a square , transmission weighted apodizing aperture in the telescope pupil or at a relay image of the telescope pupil .
the advantage of a square aperture was suggested by zanoni and hill ( 1965 ) who were interested in measuring the gravitational deflection of starlight by the sun without having to wait for a solar eclipse .
the square aperture results in most of its extended diffraction occuring along the x and y axes perpendicular to the edges of the square and strong suppression around the diagonal directions .
we then apodize the aperture with a transmission function that has peak transmission in the center of the aperture and drops off toward the edges of the aperture .
this produces extended regions around the diagonals of the point spread function ( psf ) that drop precipitously in intensity . in those regions ,
the intensity is less than @xmath4 of the psf peak , even for realistic levels of random wavefront error in the pupil ( 1/60 wave of random roughness in the surface finish ) and bandpasses ( 3000 angstroms ) .
however , it is obvious that any telescope that can directly image exo - planets will have to be of extremely high quality .
ripple and imperfections in the mirror , spiders , or mirror supports will all diffract light at angles comparable to the expected angular separation of a planet ( angel et al , 1984 , breckenridge et al , 1984 ) .
trauger et al ( 2000 ) suggest that high order , very small throw adaptive optics may be needed to correct imperfections in the mirror .
the format of the transmission function is critical to achieving the full dynamic range . that transmission function is formed by taking the product of two functions , one aligned with the vertical edges of the square aperture , and the other aligned with the horizontal edges .
using these crossed 1-d functions provides 3 to 4 orders of magnitude better suppression of the diffraction sidelobes than using a circular symmetric apodization , even when it is combined with a square aperture .
the intensity of the psf of a diffraction - limited telescope is just the square modulus of the fourier transform of the pupil function of the telescope .
a circular aperture having a constant transmission inside the pupil and zero outside has a psf in the familiar airy pattern of a bright central peak surrounded by alternate dark and bright rings . for an unapodized rectangular aperture of width @xmath5 and height @xmath6 , @xmath7 where @xmath8 , @xmath9 , and @xmath10 distance from pupil to image .
along the diagonal ( @xmath11 ) of a square aperture ( @xmath12 ) we get @xmath13 the intensity in the diffraction pattern drops rapidly along the diagonal directions . if we also apodize the square with crossed transmission functions perpendicular to the edges of the square , the intensity in the diagonal drops even faster with increasing distance from the central peak
the transmission of the aperture must have the general form @xmath14 , providing rapid attenuation along the square aperture diagonal .
we have numerically tested the combination of a square aperture and several different transmission functions , and the results appear to be quite insensitive to the functional form of the apodization .
the ( slightly ) best transmission functions tested so far are the so - called sonine functions suggested by oliver ( 1975 ) .
crossed sonine functions have a transmission function of the form @xmath15 where @xmath16 is a small integer such as 3,4 , or 5 .
one tradeoff with apodization is that the overall aperture transmission is reduced below 20% , so the optimum apodizer is one that combines the highest transmission with the best reduction of diffraction sidelobes .
we tested the case of combining a square aperture and a circularly symmetric transmission sonine function , but this results in @xmath17 less dynamic range .
we also tried various values for @xmath16 with the sonine function and two other transmission functions : @xmath18 and @xmath19 ( evaluated from @xmath20 to @xmath21 ) . comparisons of using the different shaped functions are shown in our simulation results in the next section .
the dynamic range achieved by using a square aperture combined with crossed transmission functions is demonstrated by results shown in figures 1 - 6 .
figure [ fig1 ] shows the square aperture embedded in a larger field ( upper left ) : the image of two point sources ( 100:1 ratio ) through the square aperture with no apodization ( upper right ) ; the ( sonine , @xmath16 = 4 ) apodized aperture ( lower left ) ; and the image of the same two point sources through the apodized aperture ( lower right ) .
the images of the point sources have been expanded around the region where the sources are located .
the point source images all are displayed using the 4th root of the intensity , scaled up with a gamma of 0.25 , using an idl color display table .
obviously , the apodization has dramatically reduced the diffraction sidelobes .
for the cases shown here , the `` planet '' is separated from the star by 6 times the diffraction - limited element set by the aperture size .
figure [ fig2 ] shows four images through the sonine apodized aperture ( @xmath16 = 5 ) , again expanding the region around the point sources .
the upper left is for @xmath2 brightness ratio of star to planet ; upper right is @xmath1 ratio of star to planet , no wavefront aberration ; lower left has random error of 1/72 wave rms in the aperture with @xmath1 ratio ; and lower right is 1/50 wave rms random error , again with a @xmath1 ratio of star to planet . very high optical quality is obviously required to detect earth - like planets . for comparison , in figure [ fig3 ] , we show the equivalent imaging of a pair of point sources through a circular aperture , with and without sonine ( @xmath22 ) apodization .
the upper two images are for a 100:1 ratio of `` star '' to `` planet '' .
the lower images are for @xmath1 ratio of `` star '' to planet .
obviously , apodization of a circular aperture does not come close to providing the dynamic range required for earth - like planet detection .
figure [ fig4 ] consists of four contour plots of the region around the point sources , for different apodizing functions .
all plots are for the case where the point source ratio is @xmath1 and the aberration is about 1/72 wave .
the contours are on a logarithmic scale with the lowest contour at @xmath23 and the peak of the `` star '' at @xmath24 .
the two upper plots use crossed sonine functions for apodization with @xmath25 ( upper left ) and @xmath22 ( upper right ) .
the lower plots use crossed @xmath26 apodization ( lower left ) and @xmath27 ( lower right ) .
there are small differences in the results .
the @xmath22 and @xmath27 images are slightly cleaner but the transmission of the apertures is less , as noted on the plots .
these results show that the dynamic range is only a weak function of the specific apodizing shape .
they also show that there is a very large region outside the central diffraction spikes that has sufficient dynamic range to detect planets , even terrestrial - sized planets .
figure [ fig5 ] shows diagonal cuts through the monochromatic psf for four cases : 1 ) circular aperture , no apodization ; 2 ) square aperture , no apodization ; 3 ) circular aperture with sonine apodization ; 4 ) square aperture with crossed sonine apodization . the results in figures 1 - 5 were all calculated for monochromatic light . in figure
[ fig6 ] , we show cuts through the diagonals of the psf s for the case where we have integrated over a 3000 @xmath28 bandpass . this has the effect of reducing the deep nulls in the diffraction patterns .
the vertical dashed line shows the position of a point in the cuts that is separated from the central peak by three diffraction - limited elements .
curve 1 shows a cut through the psf for an unapodized circular aperture .
curve 2 shows the effect of sonine apodization on the psf for the circular aperture .
curve 3 shows the apodized square aperture ( sonine apodization ) without the planet ( down @xmath1 from the star ) and curve 4 has a planet down @xmath1 from the star .
again , these are calculated with about 1/72 wave rms aberration in the pupil .
broadening the bandpass seems to enhance , rather than degrade , the performance of the apodizers .
these results clearly demonstrate the power of the square aperture when combined with crossed transmission function apodization .
one can apodize in a relay plane of the pupil , allowing use of any shape telescope aperture initially .
the apodizing mask can then be rotated to sweep out a full field , rather than having to rotate the telescope or telescope array .
the technique could be combined with the hypertelescope ( labeyrie 1996 , guyon 1999 ) approach to building a large telescope and apodizing the densified pupil .
because of the simplicity of the technique , one could easily imagine that it could be used on a precursor mission to tpf
. it might be possible to detect earth - like planets around the nearest stars with only a modest aperture , operating at visible or near infrared wavelengths . at a minimum
, giant planets could be imaged and their spectra measured .
for all of the examples that we have shown , we have located the `` planet '' six diffraction - limited elements from the `` star '' , along the diagonal , where it is well separated .
if we move the `` planet '' in closer to the `` star '' , the minimum separation where the `` planet '' signal is just above the wings of the `` star '' diffraction is about three diffraction limited elements ( this may be deduced from the plots in figure 6 ) .
one can extrapolate this scale to the minimum sizes of telescopes needed to separate out planets from stars for tpf .
for example , if we wish to image in the wavelength region around @xmath29 where the oxygen `` a '' line is located , then the diffraction- limit of a two meter aperture telescope with a square aperture ( given by @xmath30 , where d is the telescope diameter ) is 0.076 arcsec .
since we can just detect an earth - like planet spaced three times the diffraction limit from the star , this would allow detection of an earth in a 1 au orbit at a little over 4 parsecs .
obviously , planets in larger radius orbits or larger planets could be detected around more distant stars .
using larger apodized telescopes would increase the distance and therefore the number of sun - like stars that one could survey for planets in 1 au or larger orbits .
the nearby ( 3.2 pc ) star epsilon eridani has recently ( hatzes et al ) been reported to have a jupiter - like planet orbiting it .
an earth - like planet in a 1 au orbit around epsilon eridani imaged by a 2 meter apodized telescope would be separated by .3 arcsecond .
2200 photons would be detected from the planet in one hour for a 3000 @xmath28 bandwidth , 90 % q.e .
detector and 15 % throughput for the apodized telescope . in the same integration time , the jupiter - like planet , orbiting at 3.3 au would give 275000 detected photons ( assuming it has a similar albedo to jupiter ) .
there are many additional areas to be explored .
can we combine square apodization with nulling , or with coronagraphy to allow detection of planets in closer to the stellar peak , thereby reducing the required aperture size for detecting planets around more distant stars ? what is the optimum apodization function , making the narrowest diffraction spikes or maximizing the transmission ?
can we combine apodization with dilute array ( interferometric ) telescope concepts , in place of a large monolithic telescope ? what is the best layout for a telescope array ?
can we build a single structure to rigidly support the array ?
what are the detailed specifications on optical quality , at all spatial scales ? finally ,
if apodization allows operation at shorter wavelengths , can spectra at these wavelengths definitively detect the signatures of life ?
we plan to carry out lab tests to demonstrate the effectiveness of apodization techniques .
relatively straightforward experiments using superpolished spherical mirrors and pairs of point sources with a controlled flux ratio should allow feasibility testing of this approach .
the authors would like to acknowledge very helpful suggestions and discussions with steven ridgway , robert stachnik , daniel gezari and rick lyon .
we also wish to acknowledge support for the work from an svs subcontract ( svs inc sub-00 - 095 ) for tpf studies to the smithsonian astrophysical observatory .
guyon , o. and roddier f. 1999 , `` working on the fringe : optical and ir interferometry from ground and space .
'' proceedings from asp conference vol . 194 . edited by stephen unwin and robert stachnik , p.201 trauger , j. , stapelfeldt , k. , backman , d. , brown , r.,a
. burg , r. , burrows , a. , ftaclas , c. , kirkpatrick , j. d. , kulkarni , s. , lunine , j. , marcy , g. , redding , d. , sahai , r. and woodgate , b. 2000 , aas dps meeting 32 , 26.08 .
|
the mission of nasa s terrestrial planet finder ( tpf ) is to find earth - like planets orbiting other stars and characterize the atmospheres of these planets using spectroscopy .
because of the enormous brightness ratio between the star and the reflected light from the planet , techniques must be found to reduce the brightness of the star . the current favorite approach to doing
this is with interferometry : interfering the light from two or more separated telescopes with a @xmath0 phase shift , nulling out the starlight .
while this technique can , in principle , achieve the required dynamic range , building a space interferometer that has the necessary characteristics poses immense technical difficulties . in this paper
, we suggest a much simpler approach to achieving the required dynamic range . by simply adjusting the transmissive shape of a telescope aperture
, the intensity in large regions around the stellar image can be reduced nearly to zero .
this approach could lead to construction of a tpf using conventional technologies , requiring space optics on a much smaller scale than the current tpf approach .
subject headings : astronomical instrumentation instrumentation : interferometers techniques : interferometric stars stars : imaging stars : planetary systems
|
when formulating the criteria for what defines _
portable _ source code , one tends to draw a distinction between two types of portability , namely that associated with functionality and that associated with performance . while the former of these covers
what is perhaps usually implied when referring to code portability , i.e. , the ability of a single code to run anywhere ( rebuild the code on various architectures , the code will run and produce correct results ) , the latter , the existence of which have often been declared a fallacy , relates to the ability of a single code to run productively anywhere , i.e. , achieving , say , @xmath1 of the hand - tuned performance on any given architecture . as such
, making a code portable will often result in a compromise between superior performance on a single platform and decent performance on all potential platforms .
one possible way of circumventing this problem is by cluttering the source code with multiple conditional preprocessor directives and retain architecture - specific versions of key kernels with vendor - specific intrinsics .
however , pursuing this strategy is bound to severely hamper two of the other main best practices of code development , namely those which are concerned with securing code _
maintainability _ and testability @xcite . + for execution on standard multi - core central processing unit ( cpu ) architectures , the _ openmp standard _
@xcite first released in the second half of the 1990 s as a means to parallelize loops across compute threads
has nowadays manifested itself as the de facto programming model for intranode worksharing and , in particular , single - instruction multiple - data ( simd ) parallel instructions . the advantages of employing openmp compiler directives are manifold ; the standard is open and supported by most vendors , the compiler directives ( or pragmas ) are ignored entirely by those few compilers that do not offer support , and the syntax is relatively simple to comprehend , even to the uninitiated , making the learning barrier less steep than , e.g. , that for intra- and internode message passing using the mpi protocol .
recently , with the advent of openmp versions 4.0 and 4.5 , the standard now also allows for targeting other types of architectures such as many - core x86 processors , e.g. , intel^^ xeon phi^^ products , and graphics processing units ( gpus ) , e.g. , nvidia^^ or amd^^ gpus .
these feature enhancements from homo- to heterogenenous platforms allow for offloading of compute intensive code regions to coprocessors via so - called _ device _ constructs ( in this terminology , the cpu , to which the coprocessor is connected , is denoted the _ host _ ) .
however , there exist intrinsic difficulties in transferring the worksharing capabilities of openmp on the host to a corresponding worksharing on any attached device , in particular gpus , as the architectural details of these are fundamentally different from those of the host .
specifically , the traditional openmp model of spawning a single team of threads , which then divide loop iterations among them , will not be suitable per se for execution on gpus , as these are composed of multiple processing elements ( pes ) that run in parallel ( with no synchronization options between the threads in different pes ) and where each individual pe has the ability to efficiently perform vector - like operations .
there are ways to bypass this issue , though , for instance by creating multiple teams of threads , across which the workload may be distributed .
however , this is almost guaranteed to result in _ accelerated _ code that will not execute satisfactorily on the host ( thereby losing performance portability ) , which will reintroduce the need for * ( i ) * preprocessing of different code blocks in addition to * ( ii ) * code duplication with the required maintenance that inevitably arises as a result of this .
+ instead , an alternative standard the _ openacc standard _ @xcite , recently founded by a consortium consisting of nvidia^^ and several key vendors , including cray^^ and pgi^^has been established in an attempt to facilitate such portable and maintainable parallelism across various architectures , that is , provide the ability to expose to the compiler in a simple way what parts of the code can be run in parallel , and then leave it to the compiler to build the accelerated code _ on a given platform _ , be that the host ( as a real alternative to openmp ) or the device ( as an inherently different approach to openmp ) . since openacc is designed with heterogenous programming in mind
, it is not biased towards multi - core cpu architectures in the same way as openmp is , while still being able to advantage from past endeavours made by compiler developers that tuned these towards vectorization . in particular , as opposed to adhering to a strictly prescriptive programming model of simple thread teams , the openacc standard extends this by adding an additional layer of flexibility ; for instance , when targeting gpus , on which it is reasonable to think of a pe as a streaming multiprocessor ( sm ) , the standard offers the possibility of parallelizing over threadblocks in a grid , i.e. , individual sms , warps ( a group of 32 threads comprising the minimum execution unit ) , as well as individual cuda threads within a warp . in the openacc terminology ,
these three layers are known as gangs , workers , and vectors .
+ the openacc programming model is a so - called host - directed sequential model capable of leveraging the parallel potential of one or even multiple accelerator devices , be that gpus or multi - core coprocessors .
the constructs , which the standard is designed to target , are the same as those the openmp standard are aimed at , namely nested loop structures , which themselves are suitable for shared memory multiprocessing .
the translator between the input program and the individual simd units on the target architecture is the compiler , and the vectorization inhibitors are thus the same as for openmp , e.g. , loop carried dependencies , prohibitive complexity ( e.g. , function calls ) , as well as indirect addressing and striding . as the compiler is directed by pragmas used to expose the parallel regions in the code , the code developer may approach the problem of offloading a certain kernel in a descriptive and , importantly , incremental manner , which makes for a less cumbersome workflow than if the entire code has to be implemented from scratch using some low - level approach unrelated to the language in which the actual code base is written ( e.g. , cuda or opencl , in a gpu context ) . furthermore , as the accelerated code will be based on the original source , this warrants , in turn , a functionally and performance portable , intuitively transparent , and easily extendable and maintainable implementation .
+ in the present work , it will be demonstrated how to make use of openacc compiler directives in the acceleration of electronic structure theories , with an illustrative application to two popular many - body methods , namely the second - order mller - plesset ( mp2 ) model @xcite in its resolution - of - the - identity ( ri ) approximated form and the preeminent perturbative ( t ) triples correction to the coupled cluster @xcite ( cc ) singles and doubles ( ccsd ) model @xcite , compositely known as the ccsd(t ) model @xcite . to focus the discussion
, the present work will be solely concerned with offloading to gpu accelerators , although we stress once again how the openacc standard , and hence the present approach , is completely general and , as such , not limited to nor biased against accelerators of a specific kind .
on that note , we will start by briefly alluding to the existing literature , which despite the use of general - purpose gpus in natural sciences being a rather recent topic is rich with work devoted to gpu - accelerated quantum chemistry , concerned with diverse topics ranging from the generation of electron repulsion integrals ( eris ) @xcite , over self - consistent hartree - fock ( hf ) , complete active space ( cas ) , and density functional theory ( dft ) methods @xcite , to solvent models @xcite , force fields @xcite , and semi - empirical @xcite as well as many - body methods @xcite . adding to this quantum , several program packages are today released with full or partial gpu - enabled features , e.g. , terachem @xcite , nwchem @xcite , among others .
+ that said , the adaption of gpus and other types of coprocessors in the acceleration of quantum chemical and physical methods and applications as a whole has often been criticized for being too difficult a task to undertake implementation - wise and hence not worth the effort , in part also because of the uncertainty of ( the shape of / general consensus on ) future architectures that have always permeated the community ( hesitation towards vectorization , massive parallelism , high - performance computing , etc . ) @xcite .
on top of that , it is important to note how the clear majority of all quantum chemical implementations in existence today have been developed and are being maintained by domain scientists , many of which hold temporary positions working ( simultaneously ) on various theoretical and/or application - oriented projects @xcite .
thus , while said scientists might decide to implement a given accelerated method from scratch ( again , presumably making use of some low - level approach ) , they might not be responsible for extending it with new features and capabilities in the future nor for the routine maintenance required with the emergence of novel architectures , not to mention the finer architectural diversities between different generations of , for instance , gpu accelerators .
add to that the typical requirements of a single code base and , in particular , platform - independence , which most codes are subject to in the sense that any addition of accelerated code must not interfere with the standard compilation process ( cf .
the discussion in the opening paragraph of the present section ) , and one will potentially arrive at more reasons against investing the necessary efforts than actual reasons in favor of doing so .
+ however , a recent paper from the martnez group at stanford university has outlined an alternative strategy for avoiding the above issues , albeit one targeted exclusively at gpu hardware @xcite .
the paper , which succeeds earlier work on so - called meta - programming from within the same group @xcite , proposes the application of an automatic code generator capable of employing mathematical knowledge of the underlying algorithms ( for the efficient generation of eris , in this case ) to create a wide spectrum of possible code variants that explore , in their own words , the full space of all possible program implementations .
this type of empirical , so - called profile - guided optimization subsequently tests the ( multiple ) candidate implementations on the target architecture at hand before deciding upon a final version , and the combination of a graph - based code generator , a tester , and an optimizer hence aims at producing the optimal implementation for previous and current generations of gpus as well as , at least in principle , future generations .
+ these numerous code variants , which may be automatically produced using the engine in ref .
, will each make up different compromises between a number of parameters and factors ; for instance , the efficiency of a given kernel will be tightly bound to its utilization of the complex memory hierarchy on the gpus , the latency of which increases drastically upon traversing up through it , and the algorithmic complexity of the individual kernels trivially increases with respect to the maximum angular momentum of the one - electron basis set . furthermore , the choice of resulting implementation will also depend on the choice of floating point precision , since lowering this from double to single precision
granted that the requested accuracy allows for it may give preference to other variants .
importantly , however , the use of single precision arithmetics will result in an overall speed - up of the calculation , for execution on the host cpu and particularly so for execution on gpus .
now , in comparison with the generation of eris , the computation of correlation energies using ( perturbative ) many - body methods poses rather different requirements on the final implementation , as we will return to in [ implementations_section ] .
that being said , it is the intent of the present author to illustrate , through openacc implementations of the ri - mp2 and ccsd(t ) methods , how simple and transparent such methods may be accelerated .
it should be noted , however , that the adaption of an existing cpu eri code to gpus ( or other types of accelerators , _ vide supra _ ) by means of openacc directives will no doubt be more challenging , but it is still probably easier for the developer than using , e.g. , cuda and writing the code from scratch . in particular , the programmer may start by incrementally porting one kernel after another by exposing the inherent parallelism ( loop structures ) of the algorithm using directives , thereby verifying the correctness of the implementation along the way prior to subjecting the initial implementation to a subsequent optimization . in addition , the final implementation will require a minimum of maintenance , as the optimizer with respect to novel architectures and compute capabilities remains the compiler itself , and since the directives are treated as mere comments to a non - accelerating compiler , this applies for the efforts that will have to be invested into platform - independence of the source as well .
+ the present work is organized as follows .
following a brief introduction to the ri - mp2 and ccsd(t ) methods in [ theory_section ] , the actual openacc implementations of these using a combination of single and double precision arithmetics are discussed in [ implementations_section ] . in [ com_details_section ] , we provide some computational details on the calculations to follow in [ results_section ] , in which the performance of the implementations is assessed through calculations on test systems consisting of alanine amino acids in @xmath2-helix arrangements using one - electron basis sets of increasing size ( ranging from double- to pentuple-@xmath0 quality ) . finally , a short summary alongside some conclusive remarks are presented in [ conclusion_section ] .
in electronic structure theory , cc and many - body perturbation theory @xcite ( mbpt ) both offer a systematic approach towards the full configuration interaction @xcite ( fci ) wave function the exact solution to the time - independent , non - relativistic electronic schrdinger equation within a given one - electron basis set . in both hierarchies of methods , the mean - field hf solution acts as the reference to which correlated corrections are made by including excited configurations in the wave function . moving beyond the hf approximation , not only fermi correlation
is considered , as also the correlated motion of electrons of opposite spin start being described .
this implies that upon traversing up through either of the cc or mbpt hierarchies , increasingly more of such dynamical correlation is included in the wave function , with a complete description met at the target fci limit .
+ in the mbpt series , the lowest - order perturbative correction to the hf energy for the isolated effect of connected double excitations is that of the non - iterative mp2 model @xcite . as an improvement
, the cc hierarchy offers the iterative ccsd model @xcite , which accounts not only for connected double excitations , but does so to infinite order in the space of all single and double excitations out of the hf reference .
however , in order to achieve relative energies of chemical accuracy ( @xmath3 kcal / mol ) , the ccsd energy has to be augmented by correction terms built from higher - level excitations .
to a lowest approximation , corrections for effects due to connected triple excitations have to be considered . among such higher - level approaches ,
the ccsd(t ) model @xcite is by far the most prominent .
unfortunately , the additional accuracy of the ccsd and , in particular , the ccsd(t ) models over the mp2 model comes at a price , with the evaluation of the rate - determining step scaling non - iteratively as @xmath4 in the mp2 model ( where @xmath5 is a composite measure of the total system size ) , iteratively as @xmath6 in the ccsd model , and non - iteratively as @xmath7 for the ( t ) correction to the converged ccsd energy .
thus , whereas all three methods have nowadays become standard tools to computational quantum chemists , the current application range of the mp2 model considerably exceeds that of the ccsd(t ) model , which are the two models that we will limit our focus to herein , due to the practical differences in associated computational cost . in spite of the fact that the mp2 model is formally far less expensive than the ccsd(t ) model , the time - to - solution for the average mp2 calculation may still be substantial when targeting increasingly larger systems , e.g. , in comparison with cheaper , albeit less rigorous and less systematic methods such as those of semi - empirical nature or those that calculate the energy from a functional of the one - electron density ( dft methods ) .
for this reason , most efficient implementations of the mp2 model invoke an ri approximation for reducing the computational cost ( the prefactor ) as well as lowering the memory constraints @xcite . despite being an approximation to the mp2 model ,
the focused development of optimized auxiliary basis sets has managed to significantly reduce the inherent ri error , and for most reasonable choices of the intrinsic thresholds of a modern implementation , the error affiliated with the actual approximation will be negligible . however , while the computational cost is notably reduced in an ri - mp2 calculation with respect to its mp2 counterpart , its formal @xmath4 scaling with the size of the system will often still deem it demanding if no fundamental algorithmic changes are made @xcite . + in the closed - shell ( canonical ) mp2 model , the correlation energy is given by the following expression @xmath8 in terms of two - electron eris , @xmath9 ( mulliken notation ) , over spatial ( spin - free ) hf virtual orbitals @xmath10 and occupied orbitals @xmath11 , as well as the difference in energy between these @xmath12 besides the final evaluation of the mp2 energy in [ mp2_energy ] , the dominant part of an mp2 calculation is the construction of the four - index eris . in the ri - mp2 method , these are approximated by products of two- , @xmath13 , and three - index , @xmath14 , integrals by means of the following symmetric decomposition @xmath15 in [ eri_sym_decomp ] , greek indices denote atomic orbitals within an auxiliary fitting basis set used for spanning the ri , and the fitting coefficients , @xmath16 , are defined as @xmath17_{\alpha\gamma } \ .
\label{fitting_coef}\end{aligned}\ ] ] in terms of computational cost , the evaluation of two- and three - index integrals and/or the calculation of the fitting coefficients will dominate the overall calculation for small systems ( @xmath18- and @xmath19-scaling processes , respectively ) .
however , upon an increase in system size , the final @xmath4-scaling assembly of the two - electron integrals in [ eri_sym_decomp ] will start to dominate , and this process is thus an ideal candidate for accelerating the ri - mp2 model , cf .
[ implementations_section ] .
as touched upon in the opening paragraph of the present section , the ( t ) correction scales non - iterative as @xmath7 , as opposed to the @xmath6-scaling iterative evaluation of the energy and cluster amplitudes in the preceding ccsd calculation .
thus , for all but the smallest systems , the evaluation of the ( t ) correction , at least in its conventional canonical formulation , will dominate . in a basis of local hf orbitals , however , this balance might shift , making the underlying evaluation of the ccsd energy ( and amplitudes ) the computational bottleneck @xcite . + for closed - shell systems , the ccsd(t ) energy is defined by a triples correction , @xmath20 , to the ccsd energy , in turn defined by two energy contributions rationalized from mbpt : @xmath21}$ ] , a fourth - order term involving ccsd doubles amplitudes , @xmath22 , and @xmath23}$ ] , a fifth - order term involving ccsd singles amplitudes , @xmath24 .
in a basis of canonical hf spatial orbitals , these may be expressed as @xcite [ piecuch_4th_and_5th_order ] @xmath25 } & = \sum_{abc}\sum_{ijk}\tilde{t}^{abc}_{ijk}t^{abc}_{ijk}\epsilon^{abc}_{ijk } \label{piecuch_4th_order } \\
e^{[5 ] } & = \sum_{abc}\sum_{ijk}\tilde{z}^{abc}_{ijk}t^{abc}_{ijk}\epsilon^{abc}_{ijk } \ .
\label{piecuch_5th_order } \end{aligned}\ ] ] in [ piecuch_4th_and_5th_order ] , @xmath26 is defined on par with @xmath27 in [ e_orb_diff ] , and the triples amplitudes , @xmath28 , are defined as @xmath29 where @xmath30 is a symmetrization operator @xmath31 the @xmath32 coefficients in [ piecuch_5th_order ] are given as @xmath33 and an arbitrary six - index quantity @xmath34 in [ piecuch_4th_and_5th_order ] is defined as @xmath35 the rate - determining step is now identified as being the construction of the triples amplitudes in [ analytical_triples_ampl ] , whereas the final evaluation of @xmath20 only scales as @xmath6 . as opposed to the evaluation of the ri - mp2 energy in [ rimp2_theory_subsection ] , for which only parts of the algorithm will be accelerated ( [ mp2_energy ] and [ eri_sym_decomp ] combined ) , all of the ( t ) kernels will be offloaded to the accelerator(s ) , as will be detailed in the following [ implementations_section ] .
as is obvious from [ theory_section ] , the evaluation of both the ri - mp2 energy and the ( t ) correction relies on a vast number of matrix - matrix multiplications ( mmms ) , which in all modern implementations of the methods are handled by making calls to the ` dgemm ` routine of an optimized level-3 basic linear algebra subprograms @xcite ( blas-3 ) library .
the fact that both methods may be recasted such that they utilize blas libraries optimized for modern multi - core cpu architectures ( e.g. , intel^^ mkl , atlas @xcite , or gotoblas @xcite , to name a few ) implies that they are amenable to accelerators as well ( gpus , in the present context ) , for which optimized blas libraries too exist , e.g. , magma @xcite and cublas @xcite .
thus , in contrast to the generation of eris , cf .
the discussion in [ intro_section ] , which is comprised of extensive parallel regions , each with numerous tightly nested loop structures , this forces somewhat different requirements on an accelerated implementation for coprocessors ; in particular , aspects such as efficient data movement patterns and data reuse become highly important , since the tensors involved in many - body methods tend to grow large ( in excess of what might be afforded by the available device memory ) and the current interconnects between a host and its attached accelerator(s ) ( e.g. , a pcie bus between a cpu host and a gpu ) generally impede unnecessary ( or redundant ) movement of data . the few remaining loop constructs ( _ vide infra _ ) are then translated by descriptive directives into kernels that can run in parallel on the accelerator . by letting the host processor orchestrate the execution of all parallel kernels , those cores on the host that do not control an accelerator are free to participate in the workload .
this is an important feature ; by allowing for the final code to execute in a hybrid manner on both the host cpu as well as the attached accelerator device(s ) , the present approach is not biased towards neither the former nor the latter of the two platforms ( making for a fair comparison of the relative cpu vs. gpu throughput @xcite ) . if the implementation was instead formulated in terms of some low - level language such as cuda , such a bias would be present . in addition , the openacc standard further provides features for circumventing the default host - directed behavior , in which a sequence of operations gets scheduled for sequential execution on the device , by instead allowing for these to overlap in different pipelines , i.e. , executing operations ( computations and data transfers ) asynchronously , given that all potential race conditions have been appropriately mapped out .
+ throughout the years , elaborate attempts have been made at pushing the limits for the ri - mp2 @xcite and ccsd(t ) @xcite methods by devising massively parallel implementations of the standard canonical formulations in [ theory_section ] .
in addition , the methods have previously been ported to gpus within the groups of aln aspuru - guzik @xcite ( ri - mp2 ) , nakajima @xcite ( ri - mp2 ) , mark gordon @xcite ( ccsd(t ) ) , and karol kowalski @xcite ( ccsd(t ) ) , as , e.g. , reviewed in recent contributions to a book devoted to the subject of electronic structure calculations on graphics units @xcite . in
most , if not all of these implementations , the main driver routine , regardless of the method , consists of an outer nested loops , most often over occupied orbital indices with restricted summations to incorporate permutational symmetry of the two - electron eris @xcite . for both methods ,
each loop cycle then involves ( numerous ) large mmms , possibly intertwined by tensor ( index ) permutations and the division by orbital energy differences .
since each loop cycle is data - independent of all others , these may be trivially distributed over multiple cpu nodes using mpi .
+ in a likeminded fashion , openmp worksharing directives may be used to parallelize the individual loop cycles over the available ( @xmath36 , or more generally , ` omp\_num\_threads ` ) cores on the host cpu node . in this case ,
@xmath36 single - threaded mmms are executed concurrently , which will generally be more efficient than running a single multi - threaded mmm ( over @xmath36 threads ) sequentially @xmath36 times , albeit at the expense of having to store multiple private copies of some of the involved intermediates . as a prerequisite for using openmp this way
, one needs to flatten ( or collapse ) the two ( in the case of the ri - mp2 method , cf . [ mp2_energy ] ) or three ( in the case of the ( t ) correction , cf . [ piecuch_4th_and_5th_order ] ) outer loops into one large , composite outer loop with loop cycles of identical weight .
this way , the evaluation of the ri - mp2 energy or ( t ) correction is programmed to execute in a hybrid heterogeneous manner on both the cpu ( host ) cores and one or even multiple accelerator devices ; combined , the accelerators may be regarded as augmenting the host cpu by additional cores ( one for each attached accelerator ) , albeit some that operate at a significantly higher clock rate .
as far as the use of the gpus is concerned , all of the required data blocks ( tiles of the fitting coefficient tensor in the ri - mp2 code , and tiles of the involved ccsd doubles amplitudes and eris in the ccsd(t ) code ) can be transferred asynchronously , thereby overlapping all data movement between the host and the device(s ) with computations on the latter , and for the actual tensors , we allocate these directly into page - locked ( or pinned ) memory in order to lower the cost of the data transfers .
+ recently , in an invited chapter to a new book devoted to parallel programming with openacc , the present author outlined in some detail how to incrementally accelerate an openmp - parallelized cpu - only version of the ri - mp2 @xmath4-scaling kernel ( [ mp2_energy ] and [ eri_sym_decomp ] combined ) by means of openacc compiler directives @xcite .
herein , the ( t ) correction in [ ccsdpt_theory_subsection ] is accelerated in a similar fashion , and the source code for both implementations ( ri - mp2 and the ( t ) correction ) accompany the present work under an mit license @xcite .
furthermore , we will here assess the potential of lowering the numerical precision in certain parts of the ri - mp2 and ( t ) kernels . in standard implementations of the two methods , all of the involved steps ( evaluation of eris , amplitudes , fitting coefficients , etc .
, as well as the final energy assemblies ) are performed using double precision arithmetics exclusively . however , motivated by two recent papers @xcite , which claim that single precision arithmetics can be applied in the construction of the four - index eris ( [ eri_sym_decomp ] ) used for ri - mp2 as well as in the construction of the triples amplitudes ( [ analytical_triples_ampl ] ) for ccsd(t ) , we will do exactly this by substituting selective calls to ` dgemm ` with corresponding calls to ` sgemm ` in the codes .
not only is the performance bound to improve from the use of single precision arithmetics , but also the storage requirements on the gpu main memory will be lowered .
for the purpose of evaluating the implementations of the two methods in double or mixed floating point precision , we will conduct performance tests on chains ( @xmath2-helix arrangements ) of alanine amino acids ( [ ala]-@xmath37 , where @xmath37 is the number of residues ) in cc - pv@xmath38z ( @xmath39 d , t , q , and 5 ) one - electron basis sets @xcite and corresponding cc - pv@xmath38z - ri auxiliary basis sets @xcite ( in the case of ri - mp2 ) . due to the difference in application range , the ri - mp2 and ccsd(t ) implementations will be tested for different problems sizes ( a composite measure of the number of electrons and number of basis functions ) ; specifically , the ri - mp2 implementations are tested for the [ ala]-6 system in the cc - pv@xmath38z ( @xmath39 t , q , and 5 ) basis sets as well as the systems [ ala]-7 to [ ala]-10 in a cc - pvqz basis , whereas the ccsd(t ) implementations are tested for the [ ala]-1 system in the cc - pv@xmath38z ( @xmath39 d , t , and q ) basis sets as well as the systems [ ala]-2 to [ ala]-6 in a cc - pvdz basis . + in terms of computational hardware , the accelerators used are nvidia^^ kepler k40 gpus ( 2880 processor cores @xmath40 745 mhz ( gpu boost @xmath40 875 mhz enabled for all calculations ) and 12 gb main memory , from here on simply abbreviated as ` k40s ' ) and the host nodes are intel^^ ivy bridge e5 - 2690 v2 , dual socket 10-core cpus ( 20 cores @xmath40 3.00 ghz and 128 gb main memory ) , i.e. , ` omp\_num\_threads ` @xmath41 if not otherwise noted .
the host math library is intel^^ mkl ( version 11.2 ) and the corresponding device math library is cublas ( cuda 7.5 ) . all calculations are serial ( non - mpi ) , and the openmp-/openacc - compatible fortran compiler used is that of the pgi compiler suite ( version 16.4 ) .
+ in all implementations , the tensors containing orbital energies , fitting coefficients , ccsd singles and doubles amplitudes , as well as eris are initialized with unique , yet arbitrary single and/or double precision numbers ( depending on the context ) , as the calculation of these is not the objective of the present study , cf .
the git hash of the code in ref . used for the actual production runs is ` 42f76337 ` .
in [ figure_1_fig ] and [ figure_2_fig ] , results for the total speed - up over the threaded cpu - only implementations ( using a total of 20 openmp threads ) are reported for fixed system size varying basis sets ( [ figure_1_fig ] ) and varying system sizes fixed basis set ( [ figure_2_fig ] ) as a function of the number of k40s involved in the calculations . focusing first on the double precision ri - mp2 results , we note from [ figure_1_fig ] how the [ ala]-6/cc - pvtz problem ( by far the smallest of all the systems in the present ri - mp2 test ) is too small for the use of the k40s to be optimal , while for all other combinations of system sizes and basis sets in [ figure_1_fig ] and [ figure_2_fig ] , the use indeed appears so ( _ vide infra _ ) . from the double precision ccsd(t ) results , on the other hand , a performance improvement is observed in both cases , with no obvious sign of saturation visible neither with respect to the size of the employed basis set in [ figure_1_fig ] nor with respect to system size in [ figure_2_fig ] . moving from double to mixed precision arithmetics , the multi - gpu scaling of the ri - mp2 implementation
is observed to be unaffected for the [ ala]-6/cc - pvtz problem , but significantly improved for all other tests ; in particular , the performance is even improved in the transition from a quadruple- to a pentuple-@xmath0 basis set in [ figure_1_fig ] . for the mixed precision ccsd(t )
implementation , however , the picture is much the same as for the equivalent implementation in double precision , although the improvement in scaling with increasing system size is now not as prominent . in explaining why this is so , we recall how only parts of the overall calculation of the ( t ) correction is performed using single precision arithmetics ( the construction of the triples amplitudes ) , as opposed to the ri - mp2 implementation , in which only the final ( computationally insignificant ) energy assembly is performed in double precision numbers .
thus , the speed - up from calling ` sgemm ` over ` dgemm ` will be smaller .
also , the dimensions of the involved mmms for the tested ccsd(t ) problem sizes are greatly reduced with respect to those of the ri - mp2 problem sizes a practical necessity , due to the larger memory and cost requirements of the former of the two methods which further impedes ( or disfavors ) the scaling of the single precision implementation in the case of ccsd(t ) .
the use of single precision does , however , allow for even larger problem sizes to be addressed , as we will see later on , since the overall cpu and gpu memory requirements are reduced . -6 and ccsd(t)/[ala]-1 ) and varying basis sets ( cc - pv@xmath38z where @xmath39 d , t , q , and 5 ) . for details ,
please see the text . ]
-@xmath37 where @xmath42 ) and a fixed basis set ( ri - mp2/cc - pvqz and ccsd(t)/cc - pvdz ) . for details
, please see the text . ]
having assessed the accumulated speed - up over the ordinary cpu - only implementations , we next report results for the actual scaling with the number of k40s . in [ figure_3_fig ] and [ figure_4_fig ] ,
these scalings are presented , with the relative deviation from ideal behavior written out in the limit of six gpus . as opposed to if the calculations were executed exclusively on the gpus , i.e. , in the non - hybrid limit where @xmath43 , for which the ideal scaling is the trivial proportionality with the number of gpus , @xmath44 ( performance doubling , tripling , etc .
, on two , three , etc . , gpus ) , this is not the case for hybrid cpu / gpu execution , i.e. , whenever @xmath45 , as each cpu core is now treated as an accelerator on its own .
thus , the ideal speed - up for the latter , heterogeneous case is defined as @xmath46 in the definition of @xmath47 in [ r_speed_up ] , the constant factor @xmath48 , where @xmath49 is the time ratio between a cpu - only calculation ( @xmath50 ` omp\_num\_threads ` ; @xmath51 ) and a gpu - only calculation using a single gpu ( @xmath52 ) , accounts for the relative difference in processing power between a single cpu core ( assuming ideal openmp parallelization on the host ) and a single gpu . from the results in [ figure_3_fig ] and [ figure_4_fig ] , we note how the scaling becomes ( near-)ideal for all but the smallest problem sizes , a statement valid for both models in both double and mixed precision , regardless of the fact that these dimensions are significantly reduced in the ccsd(t ) tests , cf .
the discussion above .
/mp-@xmath37 where @xmath53 ) using the implementations in either double ( dp ) or mixed precision ( mp ) numbers . ] in addition , we may monitor how large a percentage of the actual computations is being handled by the gpus in the hybrid ri - mp2 and ccsd(t ) implementations by noting how these involve a total of @xmath54 and @xmath55 tasks of equal weight @xmath56 where @xmath57 denotes the number of occupied orbitals . through a dynamic openmp schedule , these individual tasks are distributed among the cpu cores and k40s in either of the implementations , so the scaling results in [ figure_3_fig ] and [ figure_4_fig ] may be complemented by corresponding results for the relative gpu workload . from the results , presented in [ figure_5_fig ] as the accumulated workload in percent , we note how for the present problem sizes and cpu / gpu hardware , the actual utilization of the host node is minor ( less than @xmath58 ) when , say , three or more gpus are attached to said node , regardless of the model .
still , the hybrid schemes are always faster than non - hybrid analogues , i.e. , when @xmath43 .
finally , we compare the total time - to - solution for the cpu - only and hybrid cpu / gpu implementations of the ri - mp2 and ccsd(t ) models in [ figure_6_fig ] . from these results , using six k40s , as in the present study , is seen to reduce the total time - to - solution over the cpu - only implementations in either double or mixed precision by at least an order of magnitude for all but the smallest possible problem sizes .
this is indeed a noteworthy acceleration of both models . in particular , we note how for the ri - mp2 model , the accelerated calculation on the [ ala]-10 system in a cc - pvqz basis ( @xmath59 , @xmath60 , and @xmath61 where @xmath62 and @xmath63 are the number of virtual orbitals and auxiliary basis functions , respectively ) took less time than the corresponding cpu - only calculation on the significantly smaller [ ala]-6 system within the same basis ( @xmath64 , @xmath65 , and @xmath66 ) , while for the calculation of the ( t ) correction , an accelerated calculation on the [ ala]-6/cc - pvdz system ( @xmath64 and @xmath67 , cf . the note in ref . )
terminates in less time than a corresponding cpu - only calculation on the [ ala]-4 system within the same basis ( @xmath68 and @xmath69 ) . on the basis of these results
, it may be argued that the use of a combination of openmp and openacc compiler directives as long as the complete fitting coefficients , integrals , ccsd amplitudes , etc .
, fit into main memory on the host makes it somewhat unnecessary to explicitly parallelize the code using mpi with the complications that inevitably arise from this .
in the present work , openacc compiler directives have been used to compactly and efficiently accelerate the @xmath4- and @xmath7-scaling rate - determining steps of the evaluation of the ri - mp2 energy and ( t ) triples correction , respectively . in the accelerated implementations ,
the kernels have all been offloaded to gpus , and an asynchronous pipelining model has been introduced for the involved computations and data traffic . due to their minimal memory footprints and efficient dependence on optimized math libraries , the implementations using either double or a combination of both single and double precision arithmetics are practically capable of scaling to as large systems as allowed for by the capacity of the host main memory .
, it was argued that one can not completely rely on a general - purpose compiler in the search for better program transformations since such an imaginary object as an ideal compiler can not exist , with reference to the somewhat demoralizing full employment theorem for compiler writers which states that for any optimizing compiler there will always exist a superior @xcite . while this is perfectly true , instead of viewing it as a motivation for attempting to _ beat the compiler _ by creating additional compilation layers , such as code generators , testers , etc .
, one might take an alternative stand by obeying to a sort of ad hoc , yet generalizable philosophy of _ embracing the compiler _ ; namely , given that an optimal coprocessor has not yet been mutually decided upon ( for obvious reasons , as different vendors seek to promote their own variant ) , and given that consensus is yet to be reached on whether or not there is indeed a need for accelerators in the field of electronic structure theory , one might instead try to make the most out of what hardware is currently available ( as well as near - future prospective hardware ) by investing only the least possible amount of effort into porting one s code to this .
the present work is intended to argue the case that the compiler directives of the openacc standard serve exactly this purpose by providing the means for targeting various coprocessors ( e.g. , gpus or many - core x86 processors ) in addition to the multi - core host node itself in an _ efficient _ , _ transparent _ , and _ portable _ high - level manner .
+ while the performance degradation of an openacc implementation ( with respect to a hand - tuned low - level implementation ) is bound to be larger for more complex electronic structure algorithms , such as the generation of eris or mean - field hf and dft methods , than in the present case of mbpt methods , the application to the evaluation of the ri - mp2 energy and ( t ) triples correction is intended to illustrate a number of key advantages of the use of compiler directives over a reformulation of an optimized implementation for cpus in terms of , e.g. , cuda or opencl for execution on gpus .
first and foremost , it is the opinion of the present author that accelerated code needs to be relatively easy and fast to implement , as new bottlenecks are bound to appear as soon as one part of a complex algorithm has been ported to accelerators ( cf .
law @xcite ) .
second , the use of compiler directives guarantees on par with the use of openmp worksharing directives for simd instructions on standard cpu architectures that the final code remains functionally portable , i.e. , the addition of accelerated code does not interfere with the standard compilation of the code on commodity hardware using standard non - accelerating compilers .
third , since the ri - mp2 and ccsd(t ) methods alongside other , more advanced non - iterative cc many - body methods alike @xcite are intrinsically reliant on large matrix - vector and matrix - matrix operations , the main need for accelerators in this context is for offloading exactly these .
thus , besides a number of small generic kernels , e.g. , tensor index permutations or energy summations , compiler directives are primarily used for optimizing the data transfers between the host and the device(s ) , for instance by overlapping these with device computations .
hopefully , the generality of the discussion in the present work will encourage and possibly aid others to accelerate similar codes of their own . as a means to facilitate exactly this , the present implementations come distributed alongside this work for others to reuse in part or even in full @xcite .
+ finally , one particular potential area of application for the present implementations deserves a dedicated mentioning .
while the discussion of the methods herein has been exclusively concerned with their standard canonical formulations for full molecular systems , we note how both methods have also been formulated within a number of so - called local correlation schemes , of which one branch relies on either a physical @xcite or orbital - based @xcite fragmentation of the molecular system . in these schemes ,
standard ( pseudo-)canonical calculations are performed for each of the fragments before the energy for the full system is assembled at the end of the total calculation .
thus , by accelerating each of the individual fragment ( and possible pair fragment ) calculations , the total calculation will be accelerated as well without the need for investing additional efforts , and the resulting reduction in time - to - solution hence has the potential to help increase the range of application even further for these various schemes .
the author gratefully acknowledges support from the nvidia^^ corporation , in particular for granting access to an internal test cluster , the resources of which were used in the preparation of the present work .
the author furthermore wishes to thank mark berger , jeff larkin , roberto gomperts , michael wolfe , and brent leback of the nvidia^^ corporation for general support and discussions , as well as prof .
t. daniel crawford of virginia tech , dr . filippo lipparini of johannes gutenberg - universitt mainz , and dr .
radovan bast of uit the arctic university of norway for providing fruitful comments to the manuscript .
see , for instance , the talk by jeff larkin on performance portability from the sc15 conference : http://images.nvidia.com/events/sc15/pdfs/sc15-performance-portability-openacc-larkin.pdf ( accessed ) wilson , g. ; aruliah , d. a. ; brown , c. t. ; hong , n. p. c. ; davis , m. ; guy , r. t. ; haddock , s. h. d. ; huff , k. d. ; mitchell , i. m. ; plumbley , m. d. ; waugh , b. ; white , e. p. ; wilson , p. _ plos biol . _ * 2014 * , _ 12 _ , e1001745 ( accessed ) see , for instance , the talk by james beyer and jeff larkin on openmp vs. openacc from the gtc2016 conference : + http://on-demand.gputechconf.com/gtc/2016/presentation/s6410-jeff-larkin-beyer-comparing-open-acc-openmp.pdf ( accessed ) ( accessed ) mller , c. ; plesset , m. s. _ phys .
rev . _ * 1934 * , _ 46 _ , 618ek , j. _ j. chem . phys . _ * 1966 * , _ 45 _ , 4256ek , j. _ adv .
* 1969 * , _ 14 _ , 35 paldus , j. ; ek , j. ; shavitt , i. _ phys .
a _ * 1972 * , _ 5 _ , 50 purvis , g. d. ; bartlett , r. j. _ j. chem . phys . _ * 1982 * , _ 76 _ , 1910 raghavachari , k. ; trucks , g. w. ; pople , j. a. ; head - gordon , m. _ chem . phys .
lett . _ * 1989 * , _ 157 _ , 479 stanton , j. f. _ chem .
lett . _ * 1997 * , _ 281 _ , 130 ufimtsev , i. s. ; martnez , t. j. _ j. chem . theory comp .
_ * 2008 * , _ 4 _ , 222 luehr , n. ; ufimtsev , i. s. ; martnez , t. j. _ j. chem .
theory comp . _
* 2011 * , _ 7 _ , 949 asadchev , a. ; allada , v. ; felder , j. ; bode , b. m. ; gordon , m. s. ; windus , t. l. _ j. chem .
theory comp . _ * 2010 * , _ 6 _ , 696 wilkinson , k. a. ; sherwood , p. ; guest , m. f. ; naidoo , k. j. _ j. comp .
chem . _ * 2011 * , _ 32 _ , 2313 miao , y. ; merz jr .
, k. m. _ j. chem .
theory comp . _ * 2013 * , _ 9 _ , 965 miao , y. ; merz jr .
, k. m. _ j. chem .
theory comp . _ * 2015 * , _ 11 _ , 1449 toivanen , e. ; losilla , s. a. ; sundholm , d. _ phys .
_ * 2015 * , _ 17 _ , 31480 yasuda , k. _ j. chem . theory comp . _ * 2008 * ,
_ 4 _ , 1230 ufimtsev , i. s. ; martnez , t. j. _ j. chem . theory comp . _ * 2009 * , _ 5 _ , 1004 ufimtsev , i. s. ; martnez , t. j. _ j. chem . theory comp . _ * 2009 * , _ 5 _ , 2619 asadchev , a. ; gordon , m. s. _ j. chem .
theory comp . _ * 2012 * , _ 8 _ , 4166 hohenstein , e. g. ; luehr , n. ; ufimtsev , i. s. ; martnez , t. j. _ j. chem . phys . _ * 2015 * , _ 142 _ , 224103 hohenstein , e. g. ; bouduban , m. e. f. ; song , c. ; luehr , n. ; ufimtsev , i. s. ; martnez , t. j. _ j. chem .
* 2015 * , _ 143 _ , 014111 snyder jr . , j. w. ; curchod , b. f. e. ; martnez , t. j. _ j. phys . chem .
lett . _ * 2016 * , _ 7 _ , 2444 yoshikawa , t. ; nakai , h. _ j. comp . chem . _ * 2015 * , _ 36 _ , 164 andrade , x. ; aspuru - guzik , a. _ j. chem .
theory comp . _ * 2013 * , _ 9 _ , 4360 cawkwell , m. j. ; wood , m. a. ; niklasson , a. m. n. ; mniszewski , s. m. _ j. chem .
theory comp . _ * 2014 * , _ 10 _ , 5391 losilla , s. a. ; watson , m. a. ; aspuru - guzik , a. ; sundholm , d. _ j. chem .
theory comp . _
* 2015 * , _ 11 _ , 2053 kussmann , j. ; ochsenfeld , c. _ j. chem .
theory comp . _ * 2015 * , _ 11 _ , 918 liu , f. ; luehr , n. ; kulik , h. j. ; martnez , t. j. _ j. chem .
theory comp . _
* 2015 * , _ 11 _ , 3131 nitsche , m. a. ; ferreria , m. ; mocskos , e. e. ; lebrero , m. c. g. _ j. chem . theory comp . _ * 2014 * , _ 10 _ , 959 gtz , a. w. ; williamson , m. j. ; xu , d. ; poole , d. ; le grand , s. ; walker , r. c. _ j. chem .
theory comp . _ * 2012 * , _ 8 _ , 1542 salomon - ferrer , r. ; gtz , a. w. ; poole , d. ; le grand , s. ; walker , r. c. _ j. chem .
theory comp .
_ * 2013 * , _ 9 _ , 3878 wu , x. ; koslowski , a. ; thiel , w. _ j. chem .
theory comp .
_ * 2012 * ,
_ 8 _ , 2272 vogt , l. ; olivares - amaya , r. ; kermes , s. ; shao , y. ; amador - bedolla , c. ; aspuru - guzik , a. _ j. phys . chem .
* 2008 * , _ 112 _ , 2049 olivares - amaya , r. ; watson , m. a. ; edgar , r. g. ; vogt , l. ; shao , y. ; aspuru - guzik , a. _ j. chem .
theory comp . _ * 2010 * , _ 6 _ , 135 watson , m. a. ; olivares - amaya , r. ; edgar , r. g. ; aspuru - guzik , a. _ comput .
eng . _ * 2010 * , _ 12 _ , 40 katouda , m. ; naruse , a. ; hirano , y. ; nakajima , t. _ j. comp .
chem . _ * 2016 * , doran , a. e. ; hirata , s. _ j. chem .
theory comp . _ * 2016 * , deprince iii , a. e. ; hammond , j. r. _ j. chem .
theory comp . _
* 2011 * , _ 7 _ , 1287 deprince iii , a. e. ; kennedy , m. r. ; sumpter , b. g. ; sherrill , c. d. _ mol .
phys . _ * 2014 * , _ 112 _ , 844 asadchev , a. ; gordon , m. s. _ j. chem . theory comp .
_ * 2013 * , _ 9 _ , 3385 leang , s. s. ; rendell , a. p. ; gordon , m. s. _ j. chem .
theory comp .
_ * 2014 * , _ 10 _ , 908 ma , w. ; krishnamoorthy , s. ; villa , o. ; kowalski , k. ; agrawal , g. _ cluster comput . _ * 2013 * , _ 16 _ , 131 ma , w. ; krishnamoorthy , s. ; villa , o. ; kowalski , k. _ j. chem .
theory comp . _ * 2011 * , _ 7 _ , 1316 bhaskaran - nair , k. ; ma , w. ; krishnamoorthy , s. ; villa , o. ; van dam , h. j. j. ; apr , e. ; kowalski , k. _ j. chem .
theory comp . _ * 2013 * , _ 9 _ , 1949 fales , b. s. ; levine , b. g. _ j. chem .
theory comp . _ * 2015 * , _ 11 _ , 4708 ufimtsev , i. s. ; martnez , t. j. _ comput .
eng . _ * 2008 * , _ 10 _ , 26 valiev , m. ; bylaska , e. j. ; govind , n. ; kowalski , k. ; straatsma , t. p. ; van dam , h. j. j. ; wang , d. ; nieplocha , j. ; apr , e. ; windus , t. l. ; de jong , w. a. _ comput . phys .
commun . _ * 2010 * , _ 181 _ , 1477 pratx , g. ; xing , l. _ med .
phys . _ * 2011 * , _ 38 _ , 2685 jacob , c. r. _ j. phys . chem .
lett . _ * 2016 * , _ 7 _ , 351 song , c. ; wang , l .- p .
; martnez , t. j. _ j. chem .
theory comp . _ * 2016 * , _ 12 _ , 92 titov , a. ; ufimtsev , i. s. ; luehr , n. ; martnez , t. j. _ j. chem .
theory comp . _ * 2013 * , _ 9 _ , 213 shavitt , i. ; bartlett , r. j. _ many - body methods in chemistry and physics : many - body perturbation theory and coupled - cluster theory _ ; cambridge university press : cambridge , uk , 2009 helgaker , t. ; jrgensen , p. ; olsen , j. _ molecular electronic - structure theory _
, 1st ed . ;
wiley & sons , ltd . : west sussex , uk , 2000 whitten , j. l. _ j. chem .
phys . _ * 1973 * , _ 58 _ , 4496 feyereisen , m. w. ; fitzgerald , g. ; komornicki , a. _ chem . phys .
lett . _ * 1993 * , _ 208 _ , 359 vahtras , o. ; almlf , j. ; feyereisen , m. w. _ chem .
lett . _ * 1993 * , _ 213 _ , 514 weigend , f. ; hser , m. _ theor .
acc . _ * 1997 * , _ 97 _ , 331 almlf , j. _ chem .
lett . _ * 1991 * , _ 181 _ , 319 ayala , p. y. ; scuseria , g. e. _ j. chem .
* 1999 * , _ 110 _ , 3660 werner , h .- j . ;
manby , f. r. ; knowles , p. j. _ j. chem . phys . _ * 2003 * , _ 118 _ , 8149 doser , b. ; lambrecht , d. s. ; kussmann , j. ; ochsenfeld , c. _ j. chem .
phys . _ * 2009 * , _ 130 _ , 064107 doser , b. ; zienau , j. ; clin , l. ; lambrecht , d. s. ; ochsenfeld , c. _ z. phys . chem . _ * 2010 * , _ 224 _ , 397 baudin , p. ; ettenhuber , p. ; reine , s. ; kristensen , k. ; kjrgaard , t. _ j. chem .
* 2016 * , _ 144 _ , 054102 schtz , m. ; werner , h .- j . _
_ * 2000 * , _ 318 _ , 370 schtz , m. _ j. chem .
_ * 2000 * , _ 113 _ , 9986 werner , h .- j . ;
schtz , m. _ j. chem . phys . _ * 2011 * , _ 135 _ , 144116 friedrich , j. ; dolg , m. _ j. chem . theory comput . _ * 2009 * , _ 5 _ , 287 kobayashi , m. ; nakai , h. _ j. chem . phys . _ * 2009 * , _ 131 _ , 114108 li , w. ; piecuch , p. ; gour , j. r. ; li , s. _ j. chem . phys . _ * 2009 * , _ 131 _ , 114109 rolik , z. ; kllay , m. _ j. chem .
phys . _ * 2011 * , _ 135 _ , 104111 rolik , z. ; szegedy , l. ; ladjnszki , i. ; ladczki , b. ; kllay , m. _ j. chem . phys . _ * 2013 * , _ 139 _ , 094105 schtz , m. ; yang , j. ; chan , g. k .- l . ; manby , f. r. ; werner , h .- j .
_ j. chem .
phys . _ * 2013 * , _ 138 _ , 054109 riplinger , c. ; sandhoefer , b. ; hansen , a. ; neese , f. _ j. chem .
phys . _ * 2013 * , _ 139 _ , 134101 eriksen , j. j. ; baudin , p. ; ettenhuber , p. ; kristensen , k. ; kjrgaard , t. ; jrgensen , p. _ j. chem .
theory comput .
_ * 2015 * , _ 11 _ , 2984 piecuch , p. ; kucharski , s. a. ; kowalski , k. ; musia , m. _ comp
* 2002 * , _ 149 _ , 71 lawson , c. l. ; hanson , r. j. ; kincaid , d. ; krogh , f. t. _ acm trans . math .
* 1979 * , _ 5 _ , 308 dongarra , j. j. ; ducroz , j. ; hammarling , s. ; hanson , r. _ acm trans . math
_ * 1988 * , _ 14 _ , 1 dongarra , j. j. ; duff , i. ; ducroz , j. ; hammarling , s. _ acm trans . math . soft . _ * 1990 * , _ 16 _ , 1 whaley , r. c. ; petitet , a. ; dongarra , j. j. _ parallel comput . _ * 2001 * , _ 27 _ , 3 goto , k. ; van de geijn , r. _ acm trans .
* 2008 * , _ 35 _ , 1 tomov , s. ; dongarra , j. j. ; baboulin , m. _ parallel comput . _ * 2010 * , _ 36 _ , 232 nath , r. ; tomov , s. ; dongarra , j. j. _ int . j. high perform .
appl . _ * 2010 * , _ 24 _ , 511 ( accessed ) lee , v. w. ; kim , c. ; chhugani , j. ; deisher , m. ; kim , d. ; nguyen , a. d. ; satish , n. ; smelyanskiy , m. ; chennupaty , s. ; hammarlund , p. ; singhal , r. ; dubey , p. debunking the 100x gpu vs. cpu myth : an evaluation of throughput computing on cpu and gpu .
2010 bernholdt , d. e. ; harrison , r. j. _ chem . phys .
lett . _ * 1996 * , _ 250 _ , 477 katouda , m. ; nagase , s. _ int . j. quantum chem . _ * 2009 * , _ 109 _ , 2121 katouda , m. ; nakajima , t. _ j. chem .
theory comput . _ * 2013 * , _ 9 _ , 5373 maschio , l. _ j. chem .
theory comput . _ * 2011 * , _ 7 _ , 2818 werner , h .- j . ;
knowles , p. j. ; knizia , g. ; manby , f. ; schutz , m. _ wires comput . mol .
sci . _ * 2012 * , _ 2 _ , 242 lotrich , v. ; flocke , n. ; ponton , m. ; yau , a. d. ; perera , a. ; deumens , e. ; bartlett , r. j. _ j. chem .
phys . _ * 2008 * , _ 128 _ , 15 deumens , e. ; lotrich , v. ; perera , a. ; ponton , m. j. ; sanders , b. a. ; bartlett , r. j. _ wires comput . mol .
sci . _ * 2012 * , _ 1 _ , 895 prochnow , e. ; harding , m. e. ; gauss , j. _ j. chem .
theory comput . _ * 2010 * , _ 6 _ , 2339 janowski , t. ; pulay , p. _ j. chem .
theory comput .
_ * 2008 * , _ 4 _ , 1585 baker , j. ; janowski , t. ; wolinski , k. ; pulay , p. _ wires comput .
sci . _ * 2012 * , _ 2 _ , 63 bentz , j. l. ; olson , r. m. ; gordon , m. s. ; schmidt , m. w. ; kendall , r. a. _ comp .
comm . _ * 2007 * , _ 176 _ , 589 olson , r. m. ; bentz , j. l. ; kendall , r. a. ; schmidt , m. w. ; gordon , m. s. _ j. chem .
theory comput . _
* 2007 * , _ 3 _ , 1312 kobayashi , r. ; rendell , a. p. _ chem .
lett . _ * 1997 * , _ 265 _ , 1 valiev , m. ; bylaska , e. j. ; govind , n. ; kowalski , k. ; straatsma , t. p. ; van dam , h. j. j. ; wang , d. ; nieplocha , j. ; apr , e. ; windus , t. l. ; de jong , w. a. _ comput .
commun . _ * 2010 * , _ 181 _ , 1477 apr , e. ; harrison , r. j. ; shelton , w. a. ; tipparaju , v. ; vzquez - mayagoitia , a. _ journal of physics : conference series _ * 2009 * , _ 180 _ , 012027 olivares - amaya , r. ; jinich , a. ; watson , m. a. ; aspuru - guzik , a. in _ electronic structure calculations on graphics processing units : from quantum chemistry to condensed matter physics _ ; walker , r. c. , gtz , a. w. , eds . ; john wiley & sons , ltd .
: chichester , uk , 2016 ma , w. ; bhaskaran - nair , k. ; villa , o. ; apr , e. ; tumeo , a. ; krishnamoorthy , s. ; kowalski , k. in _ electronic structure calculations on graphics processing units : from quantum chemistry to condensed matter physics _ ; walker , r. c. , gtz , a. w. , eds . ; john wiley & sons , ltd .
: chichester , uk , 2016 rendell , a. p. ; lee , t. j. ; komornicki , a. _ chem .
lett . _ * 1991 * , _ 178 _ , 462 rendell , a. p. ; lee , t. j. ; komornicki , a. ; wilson , s. _ theor . chim .
acc . _ * 1992 * , _ 84 _ , 271 eriksen , j. j. in _ parallel programming with openacc _ ; farber , r. , ed . ; morgan kaufmann : burlington , usa , 2016 openacc - accelerated coupled cluster codes : + https://gitlab.com/januseriksen/openacc_cc_codes ( accessed ) vysotskiy , v. p. ; cederbaum , l. s. _ j. chem .
theory comp . _ * 2011 * , _ 7 _ , 320 knizia , g. ; li , w. ; simon , s. ; werner , h .- j .
_ j. chem .
theory comp . _ * 2011 * , _ 7 _ , 2387 dunning jr .
, t. h. _ j. chem .
phys . _ * 1989 * , _ 90 _ , 1007 weigend , f. ; khn , a. ; httig , c. _ j. chem .
phys . _ * 2002 * , _ 116 _ , 3175 httig , c. _ phys . chem .
phys . _ * 2005 * , _ 7 _ , 59 due to a limited amount of memory on the host node ( 128 gb ) , the accelerated ccsd(t ) calculation on the [ ala]-6/cc - pvdz problem only spawned a total of 14 openmp threads instead of the default 20 threads .
out of these fourteen threads , six were devoted to the six k40s , while the remaining eight threads did computations on the host node .
finally , due to timelimit restrictions , the mixed precision cpu - only ccsd(t)/[ala]-6 calculation ( using 14 openmp threads ) was not performed , as were none of the corresponding [ ala]-6 calculations in double precision since the memory requirements for these exceeded the size of the host node memory ( i.e. , the calculations required more than 128 gb ) .
appel , a. w. ; ginsburg , m. _ modern compiler implementation in c _ ; cambridge university press : cambridge , uk , 2004 amdahl , g. _ afips conf .
proc . _ * 1967 * , _ 30 _ , 483 kucharski , s. a. ; bartlett , r. j. _ j. chem .
phys . _ * 1998 * , _ 108 _ , 5243 kucharski , s. a. ; bartlett , r. j. _ j. chem . phys . _ * 1998 * , _ 108 _ , 9221 crawford , t. d. ; stanton , j. f. _ int . j. quant .
chem . _ * 1998 * , _ 70 _ , 601 kowalski , k. ; piecuch , p. _ j. chem .
_ * 2000 * , _ 113 _ , 18 piecuch , p. ; wloch , m. _ j. chem . phys . _
* 2005 * , _ 123 _ , 224105 piecuch , p. ; wloch , m. ; gour , j. r. ; kinal , a. _ chem . phys .
_ * 2006 * , _ 418 _ , 467 kllay , m. ; gauss , j. _ j. chem .
phys . _ * 2005 * , _ 123 _ , 214105 kllay , m. ; gauss , j. _ j. chem . phys . _ * 2008 * , _ 129 _ , 144101 eriksen , j. j. ; kristensen , k. ; kjrgaard , t. ; jrgensen , p. ; gauss , j. _ j. chem .
phys . _ * 2014 * ,
_ 140 _ , 064108 eriksen , j. j. ; jrgensen , p. ; olsen , j. ; gauss , j. _ j. chem . phys . _ * 2014 * , _ 140 _ , 174114 eriksen , j. j. ; jrgensen , p. ; gauss , j. _ j. chem . phys . _ * 2015 * , _ 142 _ , 014102 eriksen , j. j. ; matthews , d. a. ; jrgensen , p. ; gauss , j. _ j. chem . phys . _ * 2015 * , _ 143 _ , 041101 kristensen , k. ; eriksen , j. j. ; matthews , d. a. ; olsen , j. ; jrgensen , p. _ j. chem . phys . _ * 2016 * , _ 144 _ , 064103 eriksen , j. j. ; matthews , d. a. ; jrgensen , p. ; gauss , j. _ j. chem . phys . _ * 2016 * , _ 144 _ , 194102 eriksen , j. j. ; matthews , d. a. ; jrgensen , p. ; gauss , j. _ j. chem . phys . _ * 2016 * , _ 144 _ , 194103 kobayashi , m. ; imamura , y. ; nakai , h. _ j. chem . phys . _ * 2007 * , _ 127 _ , 074103 mochizuki , y. ; yamashita , k. ; murase , t. ; nakano , t. ; fukuzawa , k. ; takematsu , k. ; watanabe , h. ; tanaka , s. _ chem . phys . lett . _ * 2008 * , _ 457 _ , 396 ishikawa , t. ; kuwata , k. _ chem . phys . lett . _ * 2009 * , _ 474 _ , 195 guo , y. ; li , w. ; li , s. _ j. phys . chem .
_ 118 _ , 8996 guo , y. ; li , w. ; yuan , d. ; li , s. _ sci . china chem . _ * 2014 * , _ 57 _ , 1393 friedrich , j. _ j. chem .
phys . _ * 2007 * , _ 126 _ , 154110 friedrich , j. ; dolg , m. _ j. chem .
theory comput . _ * 2009 * , _ 5 _ , 287 zikowski , m. ; jansk , b. ; kjrgaard , t. ; jrgensen , p. _ j. chem .
phys . _ * 2010 * , _ 133 _ , 014107 kristensen , k. ; zikowski , m. ; jansk , b. ; kjrgaard , t. ; jrgensen , p. _ j. chem
. theory comput . _ * 2011 * , _ 7 _ , 1677 ettenhuber , p. ; baudin , p. ; kjrgaard , t. ; jrgensen , p. ; kristensen , k. _ j. chem .
phys . _ * 2016 * , _ 144 _ , 164116
|
it is demonstrated how the non - proprietary openacc standard of compiler directives may be used to compactly and efficiently accelerate the rate - determining steps of two of the most routinely applied many - body methods of electronic structure theory , namely the second - order mller - plesset ( mp2 ) model in its resolution - of - the - identity ( ri ) approximated form and the ( t ) triples correction to the coupled cluster singles and doubles model ( ccsd(t ) ) . by means of compute directives as well as the use of optimized device math libraries , the operations involved in the energy kernels have been ported to graphics processing unit ( gpu ) accelerators , and the associated data transfers correspondingly optimized to such a degree that the final implementations ( using either double and/or single precision arithmetics ) are capable of scaling to as large systems as allowed for by the capacity of the host central processing unit ( cpu ) main memory .
the performance of the hybrid cpu / gpu implementations is assessed through calculations on test systems of alanine amino acid chains using one - electron basis sets of increasing size ( ranging from double- to pentuple-@xmath0 quality ) . for all but the smallest problem sizes of the present study ,
the optimized accelerated codes ( using a single multi - core cpu host node in conjunction with six gpus ) are found to be capable of reducing the total time - to - solution by at least an order of magnitude over optimized , openmp - threaded cpu - only reference implementations .
|
quasars are believed to be powered by supermassive black holes ( smbhs ) rapidly growing by accreting matter at a large fraction of their eddington rate which , assuming the typical quasar sed of @xcite , corresponds to @xmath30 . we will therefore also refer to obscured agns as obscured quasars if they have @xmath30 . ] .
the accreted matter forms a disk , is heated to temperatures of @xmath31 k , and emits thermal emission at optical and ultraviolet wavelengths .
clouds of gas ionised by this radiation will emit line radiation , and lines which are relatively close to the black hole have rapid orbits which will show broad velocity dispersions ( @xmath322000 km s@xmath1 ) , while clouds further away will show narrower lines .
dust can survive when the equilibrium temperature with the ultraviolet photons is @xmath332000 k. this dust absorbs optical and ultraviolet radiation and re - emits it at infrared wavelengths characteristic of the dust temperature .
if the line of sight of an observer to the central region is blocked by dust , the optical and ultraviolet continuum , as well as the broad lines , will not be observable .
such a source is known as an obscured quasar . while radio - loud obscured quasars have been known for a long time , in the form of high - excitation narrow - line radio galaxies ( see * ? ? ? * for a review ) , the population of radio - quiet obscured quasars had remained elusive , with only a few individual objects known .
this has changed recently , when large numbers of these sources have been found in the spectroscopic database of the sloan digital sky survey @xcite .
samples of spectroscopically - confirmed obscured quasars have now also been identified in surveys using x - ray selection ( e.g. * ? ? ?
* ; * ? ? ?
* ) as well as mid - infrared , sometimes combined with radio , selection ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
it seems that the obscured quasars are at least as common as the the unobscured quasars ( e.g. * ? ? ?
* ) , and probably outnumber these by a @xmath82:1 ratio ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
there is a large body of evidence supporting obscuration of quasars as an orientation effect , where obscured and unobscured quasars are intrinsically identical sources , only dust around the accretion disk covers certain lines of sight so that the broad emission lines and the optical and ultraviolet continuum are not visible ( the dusty torus of the unified schemes , e.g. * ? ? ? * ) .
however , dust distributed within the host galaxy can also cause obscuration , particularly if the galaxy is at an early evolutionary phase and is rich in gas and dust ( see e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
the importance of galaxy - scale dust is evident in recent studies of quasars where obscuration can not be solely assigned to dust from a torus around the accretion disk : in some objects even the narrow - lines are obscured @xcite .
the jets emanating from some obscured quasars are face - on @xcite , something that is not expected for quasars obscured by the torus , but can be explained if the obscuring dust is on a larger ( @xmath8kpc ) scale .
the deep silicate absorption features and spectral energy distributions ( seds ) suggest foreground extinction is sometimes present @xcite .
this all provides indirect evidence suggesting obscuring dust on @xmath8 kpc scales , something which has been spectroscopically confirmed through the measurement of the h@xmath34 to h@xmath35 balmer decrement @xcite .
dust present on kpc scales is expected to be relatively cool ( @xmath3650 k ) and will emit thermally in the far - infrared ( @xmath3740 @xmath11 m ) , regardless of whether it is heated by young stars or by the central active galactic nucleus .
observing this dust around the peak of its emission ( around rest - frame @xmath865 @xmath11 m ) is not optimal , since warmer dust can still contribute significantly there .
the rayleigh - jeans tail of the emission , however , will have a much smaller contribution from warm dust , and has the additional advantage of being observable from the ground , for example in the atmospheric windows at 850 @xmath11 m or 1.2 mm .
in this paper we present continuum observations of a sample of 21 high - redshift ( @xmath382 ) radio - intermediate obscured quasars at 1.2 mm , combined with other infrared and submillimetre data , as well as a search for co in two of the sources .
the sample was selected to approximately match the break in the unobscured luminosity function at @xmath12 ( @xmath39 - 25.7 , * ? ? ?
* ) , so in terms of radiation these quasars represent the energetically - dominant population around the peak of the quasar activity .
the sources are , however , radio - intermediate in that their radio luminosities are slightly higher than those expected from radio - quiet quasars ( with @xmath40 @xmath41 w hz@xmath1 sr@xmath1 * ? ? ?
nevertheless , observations with very large baseline interferometry suggest the relative importance of the compact cores makes them more similar to the genuinely radio - quiet population than to the radio galaxies and radio - loud quasars ( see * ? ? ?
section [ sec : obs ] summarises the observations and data reduction . in section [ sec : res ]
we show the resulting fluxes , inferred luminosities , star - formation rates , dust masses , characteristic scale of the cool dust and discuss the possible obscuration by kpc - scale dust .
section [ sec : broad ] shows the results of broad - band sed fitting and section [ sec : mean ] presents the mean sed for our sample . in section
[ sec : co ] we discuss the molecular gas observations .
we compare our sample to other millimetre or submillimetre observations of @xmath12 quasars in section [ sec : comp ] .
the discussion and summary are found in section [ sec : disc ] .
the sample of obscured quasars was observed at a wavelength of 1.2 mm ( 250 ghz ) using the 117-element max - planck millimetre bolometer array ( mambo - ii * ? ? ?
* hereafter simply mambo ) , at the iram pico de veleta 30 m telescope .
the observations were carried out in `` on - off '' mode during the mambo pool at the iram 30 m telescope throughout various semesters .
ams16 was observed in winter 2005 - 2006 , while the remaining objects were observed between summer 2007 ( sum07 ) and winter 2007 - 2008 ( win07 ) . due to a warning issued by iram ( see below ) , the objects detected during the win07 pool were re - observed in summer 2008 ( sum08 ) and winter 2008 - 2009 ( win08 ) . to minimise the effect of atmospheric absorption , the science targets were observed whenever possible at elevations @xmath4245 deg , and often at around 60 deg .
observing at elevations higher than 75 deg is , however , not possible with the 30 m telescope .
for the first - order removal of the sky , the observations make use of a secondary mirror to chop ( or `` wobble '' ) , with a typical azimuthal throw of 32 to 45 arcsec and a chopping frequency of 2 hz . the on - off observations are done by nodding , so that the sky - only position sampled by the chopping is alternatively on the right or on the left of the sky - plus - source position . each nod position ( on or off )
is observed for one minute , and to minimise overheads an on - off - off - on sequence is used .
hence integration blocks ( `` scans '' ) are multiples of 4 minutes .
the on - off observations were carried out centred on the most sensitive pixel , number 20 , with the atmospheric opacity @xmath43 and medium - to - low sky noise ( @xmath33200 mjy beam@xmath1 ) . each source
was typically observed in several blocks of 20 minute integrations .
the `` sky - only '' position can sometimes fall on a real astronomical source .
given that high - redshift agns are often found to have ( sub)millimetre - bright galaxies nearby @xcite , there is a real risk of ruining the sky subtraction by chopping onto a bright source . to minimise the risk of this happening ( and the impact on the final data ) , the chop angle ( `` wobbler throw '' )
was alternatively varied between 32 , 35 and 42 arcsec , and observations were carried out on different dates and at different times of the day .
this means the chop in azimuth corresponds to different physical offsets in ra and dec , so the chance of repeatedly falling on the same astronomical source is minimised . throughout the observing
runs , gain calibration was performed by observing neptune , uranus or mars , and the flux calibration was always found to be accurate to @xmath4420% .
the gain calibration was also monitored regularly by observing other sources with known bright millimetre fluxes ( @xmath425 jy ) .
total power measurements at different elevations were made to infer the atmospheric opacity ( `` skydip '' ) .
these were performed typically every 2 hours under stable weather conditions , and at shorter intervals for less stable conditions .
the skydips were generally done with the telescope azimuthal angle matched to that of the science targets .
the focus of the telescope was also regularly monitored on bright sources . before the first science target and every 40 minutes the accuracy of the telescope pointing
was checked using the nearby source j1638@xmath45573 ( 0.6 jy ) .
observations of j1638@xmath45573 and other bright sources were also used to check for anomalous refraction , in which case no science observations were made . during the observations , the sensitivity of mambo in a 1 second integration was typically between 35 and 45 mjy beam@xmath1 . including typically 60% overheads , the observations of this sample of 21 obscured quasars totalled approximately 65 hours .
the data were reduced using the mopsic pipeline ( developed by r. zylka ) .
mopsic removes the atmospheric emission using the two chopper positions .
the sky noise is measured by taking the weighted mean of the correlated signal from adjacent pixels , and subsequently removed .
the mambo pixels are separated in the array by two beams and weak point sources ( @xmath25150 mjy ) will therefore not be detected in adjacent pixels . using the measurements of the atmospheric opacity from the skydips , and the elevation at which objects
were observed , the extinction due to atmospheric water vapour is then corrected .
gain calibration converts counts to flux density , and finally , for each science target the weighted mean of all the scans is taken .
these are the flux densities quoted in table [ tab : obs ] : a relatively uniform root mean square ( rms ) noise @xmath80.55 mjy beam@xmath1 was achieved for most of the sources , with only a few sources having an rms @xmath80.7 - 0.9 mjy beam@xmath1 . in september
2008 iram issued a warning to mambo observers : several technical problems had ocurred during the win07 semester , which included thermal and electric malfunctions in the preamplifier , heightened sensitivity of the bolometer to microphotonics and spikes in the detected signal .
these problems were expected to affect a fraction of the scans , and the warning suggested that any @xmath2510 mjy detections obtained during this semester should be checked .
four of sources from this sample were detected during win07 : ams07 , ams08 ams12 and ams17 . all four were observed again during the win08 period , together with extra scans of ams10 , ams11 and ams20 . the data for win07 and win08 were reduced independently , with the results shown in table [ tab:07_08 ] .
the win07 flux densities are systematically higher , so only sources detected in the sum08/win08 period are considered safe detections .
however , the source ams17 was independently detected by ( * ? ? ?
* their source 22558 ) , so their flux density is used in table [ tab : obs ] .
the source ams07 is not considered a safe detection .
the potential worry was detecting sources spuriously so for the sources that are clearly not detected the combined data can be used to yield deeper limits .
table [ tab:07_08 ] marks in bold the flux densities that are used in table [ tab : obs ] .
two objects were observed in continuum at 350 @xmath11 m , ams13 and ams16 , on june 16th and 17th 2005 , using the submillimeter high angular resolution camera ii ( sharc - ii ) at the 10.4 m telescope of the caltech submillimeter observatory ( cso ) at mauna kea , hawaii @xcite .
sharc - ii is a background - limited camera utilizing a `` ccd - style '' bolometer array with 12 @xmath46 32 pixels . at 350 @xmath11 m
, the beam size is 8.5 arcsec , with a 2.59 @xmath46 0.97 arcmin@xmath19 field of view .
the observations were taken during very dry weather : ams16 was observed during 2 hours with an opacity at 225 ghz of @xmath470.05 , and ams13 was observed for 1 hour with @xmath470.06
. the dish surface optimization system ( dsos , * ? ? ?
* ) was used during the observation to correct the dish surface figure for imperfections and gravitational deformations as the dish moves in elevation .
the raw data were reduced with the comprehensive reduction utility for sharc - ii , crush - version 1.40a9 - 2 ( kovcs 2006 ) .
we used the sweep mode of sharc - ii to observe the sources : in this mode the telescope moves in a lissajous pattern that keeps the central regions of the maps fully sampled , but causes the edges to be much noisier than the central regions .
pointing was checked with uranus and secondary object crl_16293 - 2422 several times during the observations .
the pointing accuracy is better than 3 arcsec .
uranus was also used as the flux calibrator .
we used starlink s `` stats '' package to measure the flux of the calibrator and estimate the noise of the image .
the resulting root - mean - square noise values for ams13 and ams16 are 7.0 mjy beam@xmath1 and 5.5 mjy beam@xmath1 , respectively , and neither source is detected at the 3@xmath48 level .
two sources , ams12 and ams16 were observed using the plateau de bure interferometer ( pdbi ) to search for molecular gas in the host galaxies , by using carbon monoxide ( co ) as a tracer of the molecular gas .
ams16 was the first object observed with mambo , in 2005 . at the redshift of the source , @xmath49 , the co ( 4 - 3 ) rotational transition falls in the 3 mm atmospheric window [ where ( 4 - 3 ) stands for the transition from the @xmath50 to the @xmath51 angular quantum numbers ] .
the observations were carried out on 28 august and 01 september 2007 , with 5 antennae in the most compact configuration ( d ) .
the source 3c 345 was used as amplitude and bandpass calibrator , mwc 349 as amplitude calibrator and the phase calibrators were 1823@xmath45568 ( 28 aug . ) and 1637@xmath45574 ( 01 sep . ) .
the correlator was configured with both polarizations observing at a central frequency of 89.2002 ghz , with a bandwidth of 1 ghz . for the co ( 4 - 3 ) transition at @xmath49
, this corresponds to @xmath52 km s@xmath1 , and should suffice to avoid missing the line given the accuracy of the redshift .
the night of 28 august had moderately good conditions , and 25% of the data were flagged automatically .
the night of 01 september had excellent conditions ( with only 1% of the data flagged ) although for one antenna , 1 spectral unit out of the 8 ( l07 in horizontal polarization for antenna 6 ) had to be flagged for the entire night .
the total usable time from both nights amounted to 12.9 hours on - source .
the data were reduced using the gildas software .
after amplitude , phase and bandpass calibration , data cubes of the phase calibrators and of ams16 were created .
the phase calibrators came out perfectly centred at the phase center , suggesting excellent phase calibration , but ams16 was not detected either in continuum or in line .
a spectrum was extracted at the position of the source , with the shape of the beam . the root - mean square noise achieved in
the spectrum was 0.65 mjy beam@xmath1 per 10 mhz bin .
assuming a box - car co line shape with full width to zero intensity of 400 km s@xmath1 , this corresponds to 74 mjy beam@xmath1 km s@xmath1 .
the continuum sensitivity achieved over the entire bandwidth was 0.065 mjy beam@xmath1 rms .
ams12 was also observed using the pdbi , to look for the co ( 3 - 2 ) transition .
the observations , carried out during the 30 april and 13 may 2009 , were centred on 91.796 ghz using a 1 ghz bandwidth .
depending on the day , the calibrators 3c 345 , mwc 349 , 2145@xmath45067 and 0923@xmath45392 were used as flux or bandpass calibrators , while the source 1637@xmath45574 was used as a phase calibrator on both dates . on the 30 april ,
only 5 antennae were available and in addition approximately 50% of the data were flagged . for both dates the water vapour radiometer for antenna 2 did not work ,
so no atmospheric phase corrections were applied to the data from this antenna . in total ,
7.3 hours of usable data were obtained and an rms noise of 0.70 mjy beam@xmath1 per 10 mhz channel was reached .
the co ( 3 - 2 ) transition was clearly detected in ams12 , with an integrated flux of 630 mjy beam@xmath1 km s@xmath1 .
the continuum rms is 0.1 mjy beam@xmath1 , and no continuum is detected at 3 mm , so that we obtain a 3@xmath48 limit of @xmath530.3 mjy . in the case of ams16 ,
the lack of continuum sets an 3@xmath48 upper limit on the 89 ghz ( 3 mm ) flux density of @xmath530.2 mjy .
the millimetre continuum observations provide us with a wealth of information about the far - infrared luminosities and cool - dust masses of our sources .
they can also be used to test whether there is enough dust along the host galaxy to cause the obscuration in the quasars . in the following section ,
we derive these physical properties from our observations . the data uniformly available for this sample includes the following wavelengths : 3.6 , 4.5 , 5.8 , 8.0 , 24 , 70 and 160 @xmath11 m , as well as 1.2 mm .
for a small number of sources data is also available at 350 or 850 @xmath11 m .
with such excellent wavelength coverage , one might expect the far - infrared luminosities and cool - dust masses to be very accurately constrained .
however , the wavelengths corresponding to the rest - frame mid - infrared are dominated by hot dust from the warm torus around the central engine , and can not be used to constrain the cool - dust component as we explain below .
the far - infrared luminosity is dominated by cool dust , expected to be characterised by temperatures @xmath835 - 50 k. this means that emission is expected to peak around 65 - 90 @xmath11 m and drop exponentially at shorter wavelengths .
the emission from the warm torus is caused by dust at a range of temperatures , typically @xmath8100 - 2000 k , whose emission will peak at shorter wavelengths than that of the cool dust .
if the agn is powerful enough , as is the case in the sources making up this sample , the dust emission from the torus will dominate at mid - infrared wavelengths .
hence , the mid - infrared emission is not useful in constraining the cool dust mass and luminosity . to illustrate this , figure [ fig : illu ] shows a model rest - frame sed consisiting of cool dust modelled by a gray body with temperature @xmath54 k and emissivity @xmath55 together with a model agn torus from the radiative transfer models of @xcite .
the spitzer mips bands at 24 , 70 and 160 @xmath11 m , as well as the mambo band at 1.2 mm have been marked in gray , assuming this source to be at @xmath5 .
the emission seen at 1.2 mm will be completely dominated by the cool dust , while the emission at 24 @xmath11 m will be completely dominated by the warm dust of the torus .
the 70 @xmath11 m band is expected to be dominated by the torus , with some contribution from the cooler dust , while the situation for the emission observed at 160 @xmath11 m is expected to be reverse : the cool dust dominates with some contribution from the torus .
this is only an illustration , and the relative importance of each component will vary depending on the actual temperature of the cool dust , the actual sed of the torus and their relative luminosities .
however , figure [ fig : illu ] does show that when a powerful agn is present , the data at 24 and 70 @xmath11 m and the mid - infrared spectra will not provide very powerful constraints on the cool - dust emission .
it is the data at 1.2 mm that will provide the best measurement of this emission . the data at 160 @xmath11 m should provide useful constraints but for most sources in our sample it is too shallow to provide detections . data at 350 or 850 @xmath11 m is extremely useful but is only available for 3 sources .
hence , we will determine the far - infrared luminosities based on the flux densities at 1.2 mm alone for most of the sources in this sample . for sources where the data at 160 , 350 or 850 @xmath11 m provide a useful constrain
, we will make use of this data too .
however , the 24 @xmath11 m data will not be used to constrain the cool dust emission ( and in most cases , neither will the 70 @xmath11 m data ) .
the mambo observations yield detections at the @xmath423@xmath48 level for 5 out of 21 sources observed ( 24% ) .
the results are summarised in table [ tab : obs ] and figure [ fig : s_z ] ( where sources with no spectroscopic redshift were placed at @xmath56 , the median for sources with a spectroscopic redshift ) .
the mean noise achieved is 0.56 mjy beam@xmath1 , the median is 0.54 mjy beam@xmath1 .
hence , for the non detections , typical 3@xmath48 limits @xmath81.65 mjy are obtained .
with 76% of the sources undetected , the median flux density does not provide a useful characteristic number for the sample .
a more meaningful number is the mean flux density at 1.2 mm for the entire sample .
the flux densities of all sources were therefore combined using a weighted mean : @xmath57 where @xmath58 .
this yielded a statistical detection , with @xmath59 mjy .
a potential worry is that a few bright sources might be causing the statistical detection , while the rest of the sources are intrinsically faint .
stacking only the non - detections still yields a detection at the 3.9@xmath48 level : @xmath60 mjy . stacking all the detections
leads to a mean flux density of @xmath61 mjy .
figure [ fig : dbn_s ] shows the distribution of flux densities ( from table [ tab : obs ] ) , which is centred on a positive value and skewed towards the higher flux densities .
this figure supports the statement that the typical flux density of an object from this sample is @xmath80.5 - 1.0 mjy .
if the actual flux densities of the sources were so low that the distribution were dominated by the noise , figure [ fig : dbn_s ] would be centred on 0 , with a symmetric characteristic width @xmath80.55 mjy , and only a slight skew towards the high fluxes ( the detections with @[email protected] mjy ) .
we now consider whether the detection at 1.2 mm is related to the presence or lack of narrow lines in the optical spectra .
of the 5 detections at 1.2 mm , 4 show narrow lines ( 80% ) and only 1 ( 20% ) has a blank optical spectrum ( ams19 ) .
alternatively , of the 9 objects with narrow lines , 4 have detections at 1.2 mm ( 44% ) , while of the 12 with no narrow lines , only 1 is detected ( 8% ) . due to the small numbers , no trend is significant , but at first sight the data hints that sources with narrow lines are more likely to be detected at 1.2 mm .
looking at the stacked flux densities for narrow - line and for blank objects ( no narrow lines ) suggests the dust content to be similar : the weighted mean for narrow - line sources is [email protected] mjy , while for blank sources it is [email protected] mjy , so both are statistically detected .
although at first sight the difference appears significant , the narrow - line objects include ams12 with a flux of 3.7 mjy which is affecting the mean : the mean of the other 8 narrow line objects excluding ams12 is 0.76 mjy , very similar to that of the objects with no narrow lines .
hence , the mean mm flux densities for both narrow line and blank objects appear similar . given a flux density ,
the infrared luminosity of a gray body is given by : @xmath64 where @xmath65 is planck s constant , @xmath66 is boltzmann s constant , @xmath67 is the stefan - boltzmann constant , and @xmath68 is the luminosity distance at redshift @xmath69 .
the terms @xmath70 and @xmath71 refer to observed and rest - frame frequency respectively , @xmath72 .
the term @xmath73 is the planck radiation function : @xmath74 figure [ fig : lir ] and table [ tab : infer ] show the inferred values of @xmath6 assuming two different gray bodies , one with @xmath75 k and @xmath76 ( representative of unobscured quasars at high redshift , see * ? ? ?
* ) , the other with @xmath77 k and @xmath55 ( representative of submillimetre - selected galaxies , smgs , see * ? ? ?
the value of @xmath78 , also assuming @xmath56 , corresponds to @[email protected]@xmath80 ( @xmath75 k and @xmath76 ) or [email protected]@xmath80 ( for @xmath77 k and @xmath55 ) .
hence , the obscured quasars are typically ultra - luminous in the far - infrared , with a characteristic luminosity of @xmath7 ( the mean of both values ) . despite the range in spectroscopic redshifts ( 1.6 @xmath814.2 ) , the relatively flat selection function at 1.2
mm means that the detection of sources is primarily determined by their far - infrared luminosities , and is only weakly dependent on redshift ( see figure [ fig : lir ] ) .
in fact , for a given flux density , the corresponding far - infrared luminosity is actually smaller at high redshift .
hence intrinsically less luminous objects are easier to detect at higher redshift .
this probably explains why in our sample , the higher redshift sources ( @xmath382 ) have a slightly higher detection rate than the lower redshift sources ( @xmath822 ) .
the luminosity arising in the far - infrared is a considerable fraction of the agn bolometric luminosity ( @xmath83 , between infrared and x - ray energies , so between @xmath84 and @xmath85 hz ) which is typically @xmath86 for our sample ( see section [ sec : broad ] ) .
hence , @xmath6 is typically @xmath80.1 - 0.3@xmath83 depending on the assumed gray body .
this is slightly higher than what is found in low - redshift unobscured quasars , where @[email protected]@xmath83 ( e.g. * ? ?
* ; * ? ? ?
* ; * ? ? ?
* and see also section [ sec : comp ] ) . for smgs , @xmath6@xmath87@xmath83 since the far - infrared luminosity consitutes the bulk of the bolometric emission ( e.g. * ? ? ?
in addition to the data at 1.2 mm , we have access to data at 70 and 160 @xmath11 m for all 21 sources ( from * ? ? ? * ) ,
data at 350 for 2 sources ( ams13 and ams16 , see section [ sec : sharc ] ) and data at 850 @xmath11 m for one source ( ams11 , data from * ? ? ?
in most of the cases , the two fiducial gray bodies provide an acceptable fit ( see figure 5 ) . for the sources where
neither the @xmath75 k and @xmath76 or the @xmath77 k and @xmath55 gray bodies are appropriate , we use the data to find better fits .
we only fit sources that have been detected at 1.2 mm . for ams08
, figure 5 suggests that the 70 @xmath11 m data are dominated by the cool dust component and not by the warm torus , so we use the data at 70 and 160 @xmath11 m , as well as 1.2 mm , and allow @xmath35 , @xmath88 and @xmath6 to vary . for ams12 and ams19 , the 70 @xmath11 m data do not provide tight constraints , so @xmath35 is kept fixed at a value of 1.5 , and only @xmath88 and @xmath6 are allowed to vary .
the results are summarised in table [ tab : gb_fit ] and overplotted in figure 5 . under the assumption that the far - infrared luminosity arises solely from dust heated by young stars , and not by the agn itself , star - formation rates ( sfrs )
can be estimated from @xmath6 . since the dust is primarily heated by young stars , the conversion to total sfr involves assuming an initial mass function ( imf ) .
we use the conversion from , @xmath89 where @xmath90 refers to the total infrared luminosity .
this conversion assumes a @xcite imf .
given that at mid - infrared wavelengths our sources are dominated by emission from warm dust ( @xmath91100 k ) heated by the agn and not by stars ( see * ? ? ?
* ) , the mid - infrared luminosities due to star - formation can not be easily derived .
we therefore approximate the total infrared luminosity to the far - infrared luminosity , @xmath90@xmath87@xmath6 .
@xmath6 is determined by integrating analytically the fiducial gray - bodies at all wavelengths , so there is no wavelength at which the integration of the gray body is cut off .
however , any other components which would contribute to @xmath90 and are not part of the gray body ( e.g. pahs or mid - infrared continuum ) are neglected .
table [ tab : infer ] summarises the results . for the sources detected at 1.2 mm ,
the inferred values of @xmath6 and sfrs are in the range 300 - 3000 , comparable to those of smgs ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
the sfr corresponding to @xmath92 suggests that the characteristic sfr for the sample is @xmath8700 .
@xcite detected polycyclic aromatic hydrocarbon ( pah ) emission in three of the sources in this sample ( see table [ tab : obs ] ) , and found the luminosity of the detected 7.7 @xmath11 m pahs to be comparable to the typical luminosities of pahs found in the sample of smgs studied by @xcite .
the far - infrared luminosities inferred here are slightly lower but still comparable to those of smgs .
if a significant fraction of @xmath6 is actually due to agn - heated dust , these values for the sfrs will be overestimated .
this could happen if dust at @xmath63kpc scales is being heated by the @xmath83@xmath93 quasar .
in addition to the infrared data , radio observations are also available , which could in principle yield further constraints on the sfrs . the multi - frequency and high - resolution radio data , however , suggest the radio emission is completely agn - dominated , and therefore does not provide information on the sfrs ( see * ?
* ; * ? ? ?
it is not possible to discriminate whether the far - infrared luminosity arises from agn- or star - formation - heated dust : pahs were only detected in 3 objects in our sample yet the upper limits obtained for the other sources correspond to very high pah luminosities ( and hence are not very stringent limits , see * ? ? ?
the sfrs inferred for our sample are extremely high but physically - possible .
the mass of cool dust responsible for the far - infrared emission can be estimated as : @xmath94 where @xmath95 is the mass absorption coefficent and the main source of uncertainty , so we will consider two different fiducial values to illustrate the estimated values of @xmath9 .
we follow @xcite and assume @xmath960.04 m@xmath19 kg@xmath1 at 1.2 mm . the values for individual objects
are listed in table [ tab : infer ] . using @xmath97 and
again assuming @xmath56 as the typical redshift , we estimate the characteristic dust mass to be @xmath98@xmath99 and 5.1@xmath100 ( for @xmath75 k , @xmath76 and @xmath77 k , @xmath55 respectively ) .
if instead we had followed @xcite in assuming a value of @xmath962.64 m@xmath19 kg@xmath1 at 125 @xmath11 m , the resulting estimates of @xmath101 would have been @xmath102 and 2.3@xmath100 ( for @xmath75 k , @xmath76 and for @xmath77 k , @xmath55 ) , which makes little difference to our arguments . in any case , the values of @xmath9 quoted should be taken as indicative only , and we take @xmath103 as the fiducial cool - dust mass for this sample ( the mean of the four values quoted above ) .
we now estimate the characteristic distance of the dust under two assumptions : that it is heated by young stars , and that it is heated by the agn .
we note that if the dust is heated by stars , then the inferred sfrs are similar to those of smgs .
interferometric observations of smgs suggest a characteristic scale of @xmath82 kpc @xcite , so if the cool dust around our obscured quasars is heated by stars , the scale of the dust is expected to be similar to this value .
to estimate the scale of agn - heated dust , we equate the grain rate absorption of energy from ultraviolet photons to the radiated energy ( see e.g. * ? ? ?
* for a derivation ) :
@xmath104 where @xmath105 is the ultraviolet luminosity of the central source , and @xmath106 is the opacity at ultraviolet wavelengths .
the symbols @xmath107 and @xmath108 stand for the grain size and material density , and @xmath109 is the emissivity . following we assume @xmath1100.1 @xmath11 m and @xmath1113000 kg m@xmath112 as fiducial values , which yields : @xmath113 the typical extinction - corrected bolometric luminosity of our sample of quasars is @xmath86 ( see section [ sec : broad ] ) . assuming @xmath114 @xmath83 ( e.g. * ? ? ?
* ) the typical @xmath105 is about 5@xmath80 . if we assume the ultraviolet photons travel unhindered up to the characteristic radius of the dust , so that @xmath115 , and assume @xmath75 k and @xmath76 , then the dust responsible for the 1.2 mm emission is at a distance from the agn of @xmath810 kpc ( 27 kpc for @xmath77 k and @xmath55 ) . given that @xmath106 is likely to be @xmath116 at kpc scales , since the cool dust itself can absorb the uv photons from the agn and shield the dust on larger scales ( see section [ sec : host ] ) , these scales should be taken as upper limits .
hence the cool dust is expected to be distributed on scales @xmath3310 kpc .
the characteristic cool - dust mass derived from our observations can also be used to estimate how much obscuration the host galaxy can cause .
we are considering the cool dust only ( @xmath11750 k ) , which emits in the far - infrared .
we are interested in the amount of extinction it can cause at optical and mid - infrared wavelengths , where the emission from the cool dust is negligible .
hence , a simple screen approximation neglecting the emission from the obscuring dust is appropriate , so that a full radiative - transfer calculation can be avoided . in order to do this ,
we use the dust extinction models of @xcite for the milky way ( mw ) and small magellanic cloud ( smc ) .
these laws , however , correspond to @xmath118 for @xmath119200 @xmath11 m , so extrapolating directly from @xmath95 at 1.2 mm to the visual band ( 0.55 @xmath11 m ) using these extinction models would not be consistent with the values of @xmath120 we have assumed .
we therefore estimate the opacity at 200 @xmath11 m , given a typical dust mass of 3@xmath100 and assuming @xmath76 . assuming a spherically - symmetric smooth distribution of dust with a characteristic radius @xmath121 and volume @xmath122 , the column density and hence the oppacity can be estimated : @xmath123 then , using an extinction law , the corresponding extinction in the visual band , @xmath124 , can be derived ( in units of magnitudes ) . assuming a characteristic radius of 10 kpc , @xmath960.04 m@xmath19 kg@xmath1 at 1.2 mm , we estimate @xmath125 , which requires @xmath124@xmath1264 for the mw model and 1.5 for smc .
if instead we assume @xmath962.64 m@xmath19 kg@xmath1 at 125 @xmath11 m and the same radius , the opacity increases to @xmath127 and requires @[email protected] ( mw ) or 2.5 ( smc ) .
assuming instead a radius of 2 kpc , characteristic of the gas and dust in smgs @xcite , the inferred value of @xmath128 is @xmath129 , corresponding to values of @xmath124@xmath12635 - 65 ( smc dust and depending on @xmath95 ) or even @xmath130 ( mw ) .
a radius of 20 kpc leads instead to low values of @xmath124 @xmath80.3 - 1.5 ( using both mw and smc extinction curves ) . given that the condition for a quasar to be considered obscured is @xmath124@xmath425 ( see * ? ? ?
* ) , and comparing to the estimates of @xmath124 from section [ sec : broad ] shown in table [ tab : infer ] , the main conclusion from this estimate is that there is probably enough cool dust in the host galaxies of many of these sources to obscure the quasars , regardless of the orientation of the torus with our line of sight .
there are , however , examples of high - redshift unobscured quasars with similar ( sub)millimeter fluxes and hence similarly large dust masses : the presence of large dust masses does not guarantee heavy obscuration .
this is also the case for our sample ( see section [ sec : fluxes ] ) .
a likely explanation for this is that the dust in the host galaxy has a very irregular or clumpy distribution , so that only some lines of sight will be obscured . rather than orientation - dependent obscuration , as is the case for orientation caused by the torus , the obscuration ( or lack of ) by dust in the host galaxy is likely to be dependent on luck , although samples of obscured quasars are expected to be biased towards `` bad luck '' in the form of blocked lines of sight .
in this section , we make use of broad - band data to characterise the seds of our sources .
we use the irac and mips-24 @xmath11 m data to obtain an estimate of @xmath83 and @xmath124 , from broad - band sed fitting .
the approximation we use is that an obscured quasar sed can be parametrised as an unobscured quasar sed , with bolometric luminosity @xmath83 , with a foreground screen of dust causing an amount @xmath124 of extinction .
the values for @xmath83 and @xmath124 obtained in this way are only a parametrisation of the obscured sed .
the @xmath124 is only the apparent extinction , assuming the unobscured sed to be the underlying sed .
this is an accurate description at optical and ultraviolet wavelengths : the intrinsic illuminating source is the accretion disk around the smbh , which emits thermally with a range of temperatures @xmath31 k , and the dust will simply absorb this and re - emit it at longer wavelenghts . the mid - infrared emission from unobscured quasars , however , is already reprocessed light from warm dust ( @xmath36100 - 2000 k ) .
the dust responsible for this emission must be directly illuminated by the accretion disk , and must be relatively close to the source of ultraviolet photons : it is likely to originate in the inner region of the torus .
for example , following section [ sec : scale ] and using equation [ eq : rad ] , @xmath36300 k corresponds to @xmath850 pc . for an unobscured quasar
, the mid - infrared emission will travel almost unhindered until it reaches the observer , while in the case of an obscured quasar , it still has to go through a considerable amount of dust
. most of this dust will be further away ( and thus colder ) and might not be directly illuminated by the central engine anyway .
thus , the dust obscuring the mid - infrared emission is once again unlikely to emit very significantly at the same wavelengths .
the approximation of the obscured sed as an unobscured sed with a foreground screen of dust is still defendable at mid - infrared wavelengths , although it will be a poorer approximation than in the optical regime . in any case , it is a simplification of the situation .
this approximation breaks down completely in the far - infrared , so we do not attempt to model the far - infrared seds in this way . instead ,
the far - infrared component of the seds was modelled as a gray body , as described in section [ sec : firlum ] . to model the seds in this way
, we use the 24-@xmath11 m data from the catalogue of @xcite , and the irac data from @xcite as well as deeper observations around 4 objects from our program pid 20705 ( pi m. lacy , the objects are ams05 , ams12 , ams16 and ams17 ) .
we use the @xcite median quasar sed .
the quasar sed is then obscured by a screen of mw - type dust , using the models of @xcite .
it is this dust law that will determine the depth of the silicate feature ( around 9.7 @xmath11 m ) in the models .
an elliptical galaxy sed ( from * ? ? ?
* ) is also used , to represent the old stellar population , and this stellar light is not subject to any obscuration .
the quasar bolometric luminosity , @xmath124 , and the luminosity of the galaxy are all allowed to vary , and the best fit is kept when no redshift is available , while the blue component is irrelevant in the wavelength range discussed here ( 3.6 - 24 @xmath11 m ) .
in addition , we are not selecting models here , only fitting parameters , so for each object we simply keep the values with the lowest @xmath131 .
finally , due to the narrower wavelength range here , we allow the errors in photometry to be @xmath132% ( c.f . * ? ? ?
] . for sources with many limits ( namely ams08 , ams10 , ams15 , ams18 ) ,
the marginalised posterior distribution function ( pdf ) was flat for a range of values of @xmath124 ( i.e. the same @xmath131 was obtained for all values of @xmath124 above a certain value ) . given that low - luminosity objects are always more common than high - luminosity objects , and given that the inferred value of @xmath83 will correlate with @xmath124 will cause the inferred value of @xmath83 to increase .
] , we chose the lowest value of @xmath124 ( to the closest multiple of 5 ) and associated @xmath83 within the flat region of the posterior pdf .
the results are summarised in table [ tab : infer ] and all seds are presented in figure 5 .
the best - fit seds are shown ( blue solid line ) , as well as the individual components ( quasar and elliptical galaxy , dashed lines ) .
the broad - band data at 3.6 , 4.5 , 5.8 , 8.0 , 24 , 70 and 160 @xmath11 m as well as the irs spectra are superimposed . on the far - infrared end , the 1.2 mm point , and the two gray bodies are also shown .
for the sources ams13 and ams16 , the 350 @xmath11 m data from sharc are also shown , and for ams11 a data point at 850 @xmath11 m from @xcite is also included .
for source ams05 , additional data at 1.2 , 1.7 and 2.2 @xmath11 m are also shown ( from * ? ? ?
the irs spectra were not used in the fit , but used to provide instead an independent quality check on the fits . in most sources the best fit to the
broad - band sed is in reasonable agreement with the irs spectrum . in a few cases
the irs spectra show deeper silicate absorption features than those resulting from the best - fit @xmath124 ( e.g. ams06 , ams19 ) , in other cases the agreement is very good ( e.g. ams05 , ams13 ) .
the overall agreement between the irs spectra and the best - fit broad - band seds is reasonably good .
two sources have very low values of @xmath124 from the broad - band sed fitting : ams02 ( @[email protected] ) and ams20 ( @[email protected] ) .
this would suggest both objects are unobscured quasars , and should not make it into the sample .
however , both objects showed blank optical spectra , despite long integrations ( see * ? ? ?
* ; * ? ? ?
the irs spectrum of ams02 shows the silicate feature in absorption while the spectrum of ams20 shows a hint of silicate absorption ( although in * ? ? ?
* it was not considered secure due to the low signal - to - noise ratio ) .
both sources have 3.6 @xmath11 m flux densities at the edge of the selection criteria , and ams02 also has a 24 @xmath11 m density at the edge of the criteria ( see table 1 and figure 1 of * ? ? ?
ams02 is certainly a heavily obscured source , the flat sed might be due to a relatively bright host galaxy compared to the agn emission .
ams20 is less clear , the flat sed could also be due to the host galaxy contribution , and in any case the blank optical spectrum shows it is certainly not an unobscured quasar .
it might be a reddened quasar ( 0@xmath25@xmath124@xmath335 ) , or an obscured quasar with a peculiar mid - infrared sed . for our sample , we find no significant correlations between @xmath6 and @xmath83 or @xmath6 and @xmath124 .
we now derive a typical dust sed for our sample of obscured quasars and compare it to clumpy models of agn tori .
as mentioned earlier , due to the large number of upper limits the median flux density at 1.2 mm is not a very useful quantity ( since it yields an upper limit ) , yet we find the weighted mean yields a statistical detection . to be consistent , we will therefore consider mean quantities for the mid - infrared too .
we note that within our sample there is no correlation between @xmath83 and @xmath6 , so we treat the ( rest - frame ) near - infrared , mid - infrared and far - infrared components independently .
the mean sed of our spectroscopically - confirmed objects is shown as a solid red line in figure [ fig : mean_sed ] .
we do warn about several caveats : the far - infrared , mid - infrared and near - infrared components have been determined in different ways .
the far - infrared component is an analytic gray body to the mean flux density at 1.2 mm , with an empirical quasar sed ( from the mean @xmath83 ) .
it is still ill constrained , since only some objects have been detected at 70 or 160 @xmath11 m .
the mid - infrared component results from stacking irs spectra , and is probably the best determined part , although it uses only sources with spectroscopic redshifts and irs spectra . for the entire sed , only sources with spectroscopic redshifts are used . for sources with no redshift from optical spectroscopy , redshifts can only be determined if the silicate feature is deep enough to be noticeable in noisy spectra .
the mean mid - infrared sed is therefore partially biased towards sources with deep absorption features at 9.7 @xmath11 m .
the near - infrared component uses the data from all sources , but because these sources have a variety of redshifts , of values of @xmath124 , and of host galaxy luminosities , it is really a mixture of heavily - extinct agn light and stellar light .
in addition , the sources in this sample probably include examples of both torus- and host - obscured quasars , meaning that the mean sed will represent a mixture of such objects . despite all these caveats
, we consider the mean sed to be a useful fiducial representation of obscured quasars .
we parametrise the far - infrared component as being described by the two fiducial gray bodies .
the far - infrared luminosity is determined by the weighted mean flux density , @xmath97 mjy at @xmath56 . assuming the gray body with @xmath13347 k and @xmath1341.6 results in @xmath135 , while assuming @xmath13335 k and @xmath1341.5 , we obtain @xmath136 .
the mid - infrared component was obtained by shifting the irs spectra to rest - frame wavelengths , interpolating to a common wavelength grid and taking the mean luminosity density of the 14 sources with spectroscopic redshifts and mid - infrared spectra .
different wavelength bins have a different total number of sources contributing , and it was decided to cut below 4.5 @xmath11 m and above 12.5 @xmath11 m , where 3 or less sources were contributing .
figure [ fig : mean_sed ] shows the mean irs spectrum smoothed by a box - car average of 0.7 @xmath11 m .
the depth of the silicate absorption feature at 9.7 @xmath11 m , @xmath137 is usually measured by interpolating the continuum at 9.7 @xmath11 m using the continuum bluewards and redwards .
the difference in flux densities between the interpolated continuum ( @xmath138 ) and the observed feature ( @xmath139 ) then yields @xmath137 : @xmath140 this is difficult to measure from the mean irs spectrum , since there are no data points redwards of the absorption feature , so no real interpolation can be made .
however , comparing the flux densities of the 9.7 @xmath11 m feature and the 8 @xmath11 m continuum in figure [ fig : mean_sed ] , the depth of the feature is likely to be in the range @xmath141 .
figure [ fig : mean_sed ] also shows that this is deeper than what is expected from the unobscured sed with the screen of dust ( using the mean @xmath124 : dashed line ) , which otherwise fits the region of 4.5 - 8.5 @xmath11 m quite well .
we note that the mean @xmath124 from all sources with spectroscopic redshifts ( @xmath124@xmath12632 is very similar to the mean of the 14 sources used for the mid - infrared component , @xmath124@xmath12634 ) .
this small difference leads to a practically indistinguishable sed in the mid - infrared and can not explain the discrepancy in the depth of the silicate feature . at this stage ,
the discrepancy with the silicate feature is only due to the dust extinction model used : an extinction law with deeper silicate absorption ( such as the small magellanic cloud dust model of pei et al .
1992 ) can plausibly reproduce the observed feature .
however , a typical silicate feature as deep as @xmath142 suggests some degree of geometrical obscuration caused by cold dust ( see * ? ? ? * for a discussion ) .
this lends further support for some degree of host obscuration .
finally , the mean spectrum from the irs shows an absorption feature at @xmath87 @xmath11 m probably due to hydrogenated amorphous carbons . for the near - infrared component , for each object
the flux densities of the irac bands were converted to luminosities at the rest - frame wavelength .
they were then binned in a grid ranging between 0.5 and 4.5 @xmath11 m .
looking at the seds in figure 5 it seems that the upper limits are generally close to the detections in adjacent bands , so we treat limits like detections . in the bins with contributions from several sources ,
the mean luminosity density was computed .
however , some of the bins have a contribution from only one object and some from none ( these last bins are not shown in the plots ) . the solid red line in figure [ fig : mean_sed ]
shows the mean photometry at rest - frame near - infrared , smoothed by a box - car average of 0.6 @xmath11 m . the line is irregular , because of the vastly different individual seds seen in figure 5 : in some objects , all irac bands are likely to be dominated by the agn ( e.g. ams19 ) , in others by the host galaxy ( e.g. ams16 ) .
if instead of using limits as data , we do not use them at all , the resulting sed is very similar .
this part of the sed should clearly be treated with caution .
however , an estimate of the luminosity of the typical host galaxy can be attempted .
overplotted on figure [ fig : mean_sed ] is a @xmath143 elliptical galaxy from @xcite , normalised to be 12.6@xmath46 brighter than an @xmath144 galaxy in the local k - band luminosity function @xcite .
assuming @xmath82 magnitudes of passive evolution between @xmath12 and @xmath143 , it is the equivalent to a present day 2@xmath144 galaxy ( this is partly by selection , as is described in detail by * ? ? ?
* ; * ? ? ?
we can thus see that the luminosity of the near - infrared component of the typical obscured quasar sed corresponds approximately to the progenitor of a 2@xmath144 galaxy , although it might also have an important contribution from agn light ( in which case the galaxy luminosity will be overestimated ) .
it is of particular interest to compare the mean sed to clumpy models of the agn torus .
this is shown in figure [ fig : mean_clumpy ] ( left panel ) , where the mean sed is overplotted on the models of @xcite .
the models can not reproduce simultaneously the relatively flat sed at wavelengths shorter than 8 @xmath11 m , the sharp drop in the near - infrared and the deep silicate absorption .
the left panel of figure [ fig : mean_clumpy ] also shows that the models with deeper silicate features still struggle to match the observed silicate depth ( @xmath145 ) yet will overpredict the far - infrared luminosity .
the right panel of figure [ fig : mean_clumpy ] shows the two flattest clumpy seds from the right panel , now with additional extinction of @xmath124@xmath12680 added due to foreground dust .
we see that the clumpy models , with the additional extinction and re - emission from the cool dust can reproduce reasonably well the observed sed .
since the dust is cold , it re - emits in the far - infrared , meaning it does not `` fill - up '' the silicate absorption feature , unlike dust at @xmath8300 k would ( again , see * ? ? ?
* ) . to reproduce the deep silicate feature , a value of @xmath124@xmath880 is required ( see figure [ fig : mean_clumpy ] ) .
the typical dust mass of @xmath103 can cause this extinction of @xmath124@xmath12680 provided that the dust is on scales @xmath252.3 - 3.0 kpc ( from equation [ eq : tau_m ] in section [ sec : host ] and assuming mw dust and the two values of @xmath146 ) .
this is in excellent agreement with the inferred scales of the cool dust emission in smgs ( typically @xmath82 kpc * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
@xcite studied in detail the mid - infrared spectra and mid- to near - infrared seds of 21 high - redshift heavily obscured quasars , selected to be bright ( @xmath421 mjy ) at 24 @xmath11 m , optically faint ( typically @xmath147 ) and with mid - infrared spectra characteristic of heavily obscured agn ( see their section 2 for more details ) .
they found that 9 out of 21 sources did not fit clumpy models well and required an additional foreground screen of dust that did not re - emit at mid- or near - infrared wavelengths .
our findings for the mean sed are consistent with their results , and in addition , we are able to estimate the @xmath124 in a consistent way from @xmath9 , which was derived from our observations at 1.2 mm .
the resulting dust size required for @xmath9 to cause the @xmath124 ( @xmath82.3 - 3.0 kpc ) is very close to the sizes estimated for the far - infrared emission of smgs .
again , our results provide evidence for obscuration due to dust on kpc scales , dust that has lower temperatures and is distributed on larger scales than what is expected for the dust of the torus .
figure [ fig : mean_comp ] shows the mid - infrared component of the mean sed with the spectra of four local ultra - luminous infrared galaxies overplotted ( ulirgs , with @xmath148 l@xmath149 ) .
the spectra for these sources were obtained from @xcite and @xcite .
these ulirgs represent one pure - agn , iras 12514@xmath451027 ( hereafter iras 12514 ) , two agn - starburst composites , mrk 231 and mrk 273 , and one starburst - dominated source , arp 220 .
all spectra have been normalised to match our mean sed between 7.0 and 7.8 @xmath11 m ( around 4@xmath150 hz ) .
compared to these sources , our mean sed is relatively flat , similarly to mrk 231 or iras 12514 , yet the silicate absorption feature is significantly deeper , more similar to that of mrk 273 or even arp 220 .
once again , we see the peculiar spectral shape of this mean sed , where the silicate feature is suprisingly deep given the relative flatness of the continuum .
the most similar nearby ulirg is iras 12514 , which is known to be an obscured quasar @xcite , although this source has a shallower feature at 9.7 @xmath11 m .
two sources , ams12 and ams16 , were observed with the iram pdbi to search for the co ( 3 - 2 ) or ( 4 - 3 ) rotational transitions , tracers of molecular gas .
the co ( 3 - 2 ) line in ams12 is clearly detected , and is well described by a gaussian with a peak flux of 2.14 mjy and a full - width half - maximum ( fwhm ) of 275 km s@xmath1 ( see figure [ fig : coams12 ] ) .
the integrated line flux of this gaussian is @xmath14630@xmath1550 mjy km s@xmath1 , where @xmath151 is the velocity - integrated co line flux .
this corresponds to a ( frequency - integrated ) luminosity of @xmath163.2@xmath17 , or a brightness - temperature luminosity of @xmath152 k km s@xmath1 pc@xmath19 .
the central frequency of the line is at -16 km s@xmath1 , corresponding to a frequency shift of 4.9 mhz ( so a central frequency of 91.801 ghz ) , so the redshift of the co line is @xmath153 . at our signal to noise ,
the line shape is well described by a gaussian , there is no double profile visible , which could suggest a disk .
there is also no hint of two separate lines which could suggest that the molecular gas is undergoing a major merger . in the case of ams16 ,
no line was detected ( see figure [ fig : coams16 ] ) .
the full data cube was searched by looking for line emission at various spectral resolutions , using a routine provided by roberto neri , but no significant emission was detected at the position of the source .
a signal at the 4.5 @xmath48 level can be seen at 17 19 44 @xmath4558 47 00 . if real this would correspond to a line offset from the narrow - line redshift by @xmath8600 km s@xmath1 and with a full width at zero intensity of @xmath8850 km s@xmath1 . however , the position does not correspond to any source detectable in the r - band , 3.6 @xmath11 m or 1.4 ghz image .
this position corresponds to a 17 arcsec offset , or 116 kpc at the redshift of the source .
we do not consider this emission any further
. the lack of detection of the co ( 4 - 3 ) line can either be due to the line being shifted outside of the observed bandwidth ( @xmath1541680 km s@xmath1 shift ) , or the line being within the bandwidth but too faint .
the redshift was determined from three narrow lines , ly @xmath34 ( 1216 ) , n iv ( 1240 ) and c iv ( 1549 ) with individual redshifts of @xmath155 , 4.166 and 4.170 respectively ( see martnez - sansigre et al . , 2006a ) .
thus the maximum redshift offset of any individual line , with respect to the average @xmath156 is that of the n iv line with @xmath157 , which only corresponds to a 175 km s@xmath1 offset . given the bandwidth used , we consider the possibility that the co ( 4 - 3 ) line falls outside of the observed range unlikely , so detection of co ( 4 - 3 ) in ams16 can be ruled out down to a 3@xmath48 limit of @xmath1580.23 jy km s@xmath1 , assuming a box - car line shape with full width to zero intensity of 400 km s@xmath1 . from these limits , we now estimate the limits on the co ( 4 - 3 ) luminosities and derived molecular gas properties .
figure [ fig : coams16 ] shows the spectrum extracted at the position of ams16 . at @xmath49
, @xmath159230 mjy km s@xmath1 corresponds to a frequency - integrated line luminosity of @xmath1603@xmath17 , or a brightness - temperature luminosity of @xmath161 k km s@xmath1 pc@xmath19 . if we assume the molecular gas responsible for the co emission is dense enough and warm enough for the ( 3 - 2 ) or ( 4 - 3 ) transitions to be thermalised , then the brightness - temperature luminosities of these high - j transitions are the same as that of the ( 1 - 0 ) transition : @xmath162 and @xmath163 .
following @xcite , we can then estimate an upper limit on the mass of molecular hydrogen , by using @xmath164 where @xmath1650.8
( k km s@xmath1 ) @xmath1 is the conversion factor for local ulirgs @xcite .
thus , under the assumption that @xmath162 , we infer a total gas mass of @xmath166 for ams12 . assuming that @xmath167
, we can derive a limit for the total h@xmath168 mass in ams16 of @xmath169 .
the far - infared luminosity of ams12 of @xmath170 ( see table [ tab : gb_fit ] ) , suggests a sfr of 4300 .
the ratio of far to co luminosity is log@xmath171=3.0 $ ] , the gas - to - dust ratio , @xmath26@xmath27@xmath9 , is 19 , and the gas depletion timescale is estimated to be @xmath1724 myr . for ams16 , we can only infer limits for these values : if we assume the values from the @xmath77 k , @xmath55 gray body , we infer log@xmath171\geq$]2.4 , @xmath26@xmath27@xmath9@xmath448 and @xmath17316 myr .
assuming instead the @xmath75 k , @xmath76 gray body yields the following values : log@xmath174\geq$]2.9 , @xmath26@xmath27@xmath9@xmath4420 and @xmath1755 myr .
for both ams12 and ams16 , the estimated values of log@xmath174 $ ] are at the low end but consistent with the observed values for other high - redshift sources with molecular gas measurements .
the gas - to - dust ratios are much lower than the values @xmath8100 - 150 found for local spiral galaxies .
the values of @xmath176 are ultimately derived from the quantity @xmath177 and are thus also at the low end of the observed values .
a possible explanation for the high values of @xmath177 is that a fraction of the far - infrared luminosity is due to agn heated dust , rather than by dust heated by young stars .
we warn that if the assumption that the ( 4 - 3 ) transition is thermalised is inappropriate , @xmath178 or @xmath179 , we will have underestimated @xmath26 , and hence @xmath176 and the gas - to - dust ratio . for a discussion about whether this assumption is appropriate , see @xcite and @xcite , although for the high - redshift agn studied in detail it seems to be a reasonable assumption . from the fwhm of the co(3 - 2 ) line in ams12 , @xmath180275 km s@xmath1 , and assuming a characteristic radius , we can estimate the dynamical mass given ( see * ? ? ?
* ) : @xmath181 the co is unresolved in the @xmath85 arcsecond resolution observations with the pdbi observations , yielding a size limit of @xmath18240 kpc .
a better estimate for @xmath121 is 2 kpc @xcite .
the inferred mass is then @xmath183 , and is obviously dependent on the unknown inclination angle @xmath184 .
the ratio of gas to dynamical mass is then @xmath185 .
thus the estimated gas mass is larger than the estimated dynamical mass , unless the inclination angle @xmath184 is @xmath186 .
it seems likely that the gas mass represents a high fraction of the total dynamical mass , as found for other high - redshift galaxies observed in co @xcite .
other samples of @xmath12 quasars have been observed at millimetre or submillimetre wavelengths , allowing us to compare the far - infrared luminosities inferred for the different samples .
far - infrared emission being isotropic , the properties of obscured and unobscured quasars should be similar if obscuration is only an orientation effect . to test this
, we consider here three other studies at approximately the same redshift : * the sample of , consisting of optically - selected unobscured quasars observed at 1.2 mm .
these sources are intrinsically very bright with @xmath187 , corresponding to @xmath83 in the range @xmath188 , and at @xmath189 .
they have a detection rate of 9 out of 26 sources ( 35% ) . * the sample presented by ( * ? ? ?
* p04 ) , which consists of optically unobscured quasars known also to be x - ray unabsorbed .
these sources have values of @xmath83@xmath93 , similar to our sample , with @xmath190 .
they were observed at 850 @xmath11 m , with a detection rate of 1 out of 20 ( 5% ) . * the optically unobscured but x - ray absorbed quasars from the sample of ( * ? ? ?
* s05 ) , with similar values of @xmath83 to our sample , @xmath191 .
they were observed at 850 @xmath11 m with 6 out of 19 sources detected at the @xmath2 level ( 32% ) . in all samples ,
the bolometric luminosities have been estimated consistently using the @xcite sed , and correcting for absorption at x - ray energies ( s05 ) or dust - extinction at optical wavelengths ( this sample ) .
the far - infrared luminosities have all been computed assuming a gray body with @xmath75 k , @xmath76 .
changing these parameters will affect very similarly the far - infrared luminosities of all samples .
figure [ fig : lfir_lbol1 ] shows @xmath6 vs @xmath83 for the four samples of @xmath12 quasars .
the coloured symbols represent the individual sources , the black filled symbols the mean values for each sample , and the black empty symbols the mean of the non - detections .
the dashed line shows a constant far - infrared fraction of @xmath192 , characteristic of the low - redshift quasars studied by @xcite . from the available data ,
the o03 and p04 samples as a whole have a ratio of @xmath193 similar to the low - redshift quasars . however , our sample and the s05 samples both seem to have larger far - infrared fractions than the o03 and p04 samples .
if all obscuration were orientation dependent then the host galaxies of obscured and unobscured quasars would be expected to have , on average , the same properties , such as star - formation rates and cool dust content . if , however , some obscured quasars are at a different evolutionary stage , then their host galaxy can be reasonably expected to have a larger gas and dust content , which might be causing the obscuration , or at least contributing to it
. obscured quasars might therefore be in a different evolutionary phase to unobscured quasars .
the extinction due to dust is negligible in the far - infrared ( fir ) , so that at wavelengths @xmath194 @xmath11 m the emission becomes essentially isotropic .
hence , if orientation is the only difference between obscured and unobscured quasars , they should have virtually identical cool dust properties , irrespective of whether this emission is due to dust heated by the quasar or by young stars in the host galaxy .
although it is difficult to quantify the significance of the difference between samples , it seems that optically obscured quasars ( and optically - unobscured but x - ray absorbed quasars ) have , on average , higher far - infrared luminosities than unobscured quasars . in turn
, this implies that the cool - dust content of the host galaxies are different so that , on average , obscured quasars are hosted by dustier galaxies .
this is , of course , partly a result of selection .
the samples selected to contain obscured sources will of course be biased towards dustier objects , but such a bias will only show in the far - infrared observations if obscuring dust on kpc scales is important ( as opposed to dust in the torus only , which will emit mostly at mid - infrared wavelengths ) . the s05 sample is particularly biased towards such host obscuration , since it contains optically - bright quasars that are not obscured by the torus , but which show absorption at x - ray energies .
hence , the difference in far - infrared luminosities between obscured and unobscured quasars lends strong support to the scenario that some of the obscuring dust is in the host galaxy , not only the torus , and there is a hint that many obscured quasars might represent a different evolutionary phase to unobscured quasars .
we have observed a sample of @xmath0 radio - intermediate obscured quasars at 1.2 mm ( 250 ghz ) using mambo . the detection rate is 5 out of 21 ( 24% ) .
sources with narrow lines seem to have a higher detection rate than sources with no narrow lines , but with such small numbers the difference is not significant .
larger samples will be required to study any differences .
stacking leads to a statistical detection of @xmath195 .
even if only the non - detections are stacked , they still yield a statistical detection , with @xmath196 .
thus , the typical flux density of this sample is @xmath80.5 - 1.0 mjy , and this corresponds to a far - infrared luminosity @xmath7 .
if the far - infrared luminosity is powered entirely by star - formation , and not by agn - heated dust , then the typical inferred star - formation rate is @xmath8700 .
this large star - formation rate is comparable to those inferred in ulirgs and smgs , the most powerful starburst galaxies known .
the observations at 1.2 mm also allow us to estimate the mass of the cool dust , and we find a typical mass of @xmath10 .
if heated by young stars , this dust is expected to be distributed on @xmath82 kpc scales ( the scale found for smgs ) .
if it is heated by the central agn , with @xmath83@xmath93 , then it is expected to be distributed on scales @xmath3310 kpc .
combining our observations at mid - infrared and millimetre wavelengths , we present dust seds for our sample , and derive a typical sed for our sample of high - redshift obscured quasars .
the clumpy tori models can not reproduce the sed .
however , clumpy tori models with an additional screen of cold dust ( and the corresponding far - infrared emission ) do reproduce this mean sed .
the amount of extinction from this additional screen is derived from the cool dust mass , from the observations at 1.2 mm .
the depth of the silicate feature can be consistently achieved by the inferred cool dust mass provided that the dust is on scales @xmath332 kpc , which is in excellent agreement with the typical radius of the far - infrared emission in smgs .
again , this lends support to kpc - scale dust along the host galaxy playing an important role in the obscuration of these sources .
obscuration by dust on kpc scales would also explain why in about half of the sample the narrow emission lines from the central agn are not detectable , as well as why in some cases the jets seem to be pointing towards us , something unexpected if they are only obscured by the torus of the unified schemes .
however , we remind the reader that unobscured quasars have also been found to have similar cool - dust masses , and that in our sample there is not a one - to - one correspondence between the presence ( or lack of ) narrow lines and detection at 1.2 mm .
the presence of a large mass of dust does not guarantee obscuration .
a clumpy dusty interstellar medium where different lines of sight have vastly different optical depths is a likely explanation for why some quasars with large dust masses are obscured while others are not . when comparing to other samples of high - redshift quasars , we find that the obscured quasars probably have a higher fraction of their luminosity emerging at far - infrared wavelengths , compared to unobscured quasars .
the far - infrared emission is isotropic , so that this difference can not be ascribed to orientation - dependent obscuration : obscured quasars are likely to have higher far - infrared luminosities and cool - dust masses than unobscured quasars , suggesting the host galaxies are in a dustier , presumably earlier , phase .
additionally , we have searched for molecular gas in two sources at @xmath13 and 4.169 , using the co ( 3 - 2 ) and ( 4 - 3 ) transitions . in the @xmath13 source
we detect a line with @xmath1973.2@xmath17 ( equivalent to a brightness - temperature luminosity of @xmath152 k km s@xmath1 pc@xmath19 ) . in the other source
, the lack of detection suggests a line luminosity @xmath1603@xmath17 ( @xmath161 k km s@xmath1 pc@xmath198 . under the assumption that in these objects the ( 3 - 2 ) and ( 4 - 3 ) transitions are thermalised , we can estimate the molecular gas contents to be @xmath166 and @xmath23 , respectively .
the estimated gas depletion timescales are @xmath199 and @xmath2516 myr , and low gas - to - dust mass ratios of @xmath26@xmath27@xmath9@xmath28 and @xmath200 are inferred .
a dynamical mass of @xmath201 is estimated from the co(3 - 2 ) detection .
we thank the iram staff for help with the observations and data reduction for this program .
we are particularly grateful to stphane leon for extensive help with the observations at the 30 m , robert zylka for making the mopsic package publicly available , jan martin winters for help with the pdbi observations , philippe salom for help reducing the pdbi data on ams16 and roberto neri for use of his software .
we also thank javier rodn , veronica roccatagliata and aurora sicilia - aguilar for help with software .
we thank the support and assistance provided by the cso staff and sharc - ii team at caltech during the observations and data reduction .
we are also grateful to the clumpy group for making their models publicly available .
this manuscript was improved by the suggestions of an anonymous referee .
the work was partially supported by grants associated with spitzer programs go-20705 and go-30634 , and is based on observations made with the spitzer space telescope , which is operated by the jet propulsion laboratory , california institute of technology under a contract with nasa .
|
we present observations at 1.2 mm with mambo - ii of a sample of @xmath0 radio - intermediate obscured quasars , as well as co observations of two sources with the plateau de bure interferometer .
the typical rms noise achieved by the mambo observations is 0.55 mjy beam@xmath1 and 5 out of 21 sources ( 24% ) are detected at a significance of @xmath2 . stacking all sources leads to a statistical detection of @xmath3 mjy and stacking only the non - detections also yields a statistical detection , with @xmath4 mjy . at the typical redshift of the sample , @xmath5 , 1 mjy corresponds to a far - infrared luminosity @xmath6@xmath7 .
if the far - infrared luminosity is powered entirely by star - formation , and not by agn - heated dust , then the characteristic inferred star - formation rate is @xmath8700 .
this far - infrared luminosity implies a dust mass of @xmath9@xmath10 , which is expected to be distributed on @xmath8kpc scales .
we estimate that such large dust masses on kpc scales can plausibly cause the obscuration of the quasars . combining our observations at 1.2 mm with mid- and far - infrared data , and additional observations for two objects at 350 @xmath11 m using sharc - ii
, we present dust seds for our sample and derive a mean sed for our sample .
this mean sed is not well fitted by clumpy torus models , unless additional extinction and far - infrared re - emission due to cool dust are included .
this additional extinction can be consistently achieved by the mass of cool dust responsible for the far - infrared emission , provided the bulk of the dust is within a radius @xmath82 - 3 kpc .
comparison of our sample to other samples of @xmath12 quasars suggests that obscured quasars have , on average , higher far - infrared luminosities than unobscured quasars . there is a hint that the host galaxies of obscured quasars must have higher cool - dust masses and are therefore often found at an earlier evolutionary phase than those of unobscured quasars . for one source at @xmath13
, we detect the co(3 - 2 ) transition , with @xmath14630@xmath1550 mjy km s@xmath1 , corresponding to @xmath163.2@xmath17 , or a brightness - temperature luminosity of @xmath18 k km s@xmath1 pc@xmath19 . for another source at @xmath20 ,
the lack of detection of the co(4 - 3 ) line suggests the line to have a brightness - temperature luminosity @xmath21 k km s@xmath1 pc@xmath19 . under the assumption that in these objects the high - j transitions are thermalised
, we can estimate the molecular gas contents to be @xmath22 and @xmath23 , respectively .
the estimated gas depletion timescales are @xmath24 myr and @xmath2516 myr , and low gas - to - dust mass ratios of @xmath26@xmath27@xmath9@xmath28 and @xmath29 are inferred .
these values are at the low end but consistent with those of other high - redshift galaxies .
|
in this two - part work we are interested in the way a solid immersed in a viscous incompressible fluid ( in dimension 2 or 3 ) can deform itself and then interact with the environing fluid in order to stabilize exponentially to zero the velocity of the fluid and also its own velocities . the domain occupied by the solid at time @xmath0
is denoted by @xmath1 .
we assume that @xmath2 , where @xmath3 is a bounded smooth domain .
the fluid surrounding the solid occupies the domain @xmath4 .
( 0,0)(-100,-90 ) ( 0,0)(0,0)[lb ] ( 80,-30)(0,0)[lb ] ( -20,25)(0,0)[lb ] the movement of the solid in the inertial frame of reference is described through the time by a lagrangian mapping denoted by @xmath5 , so we have @xmath6 the mapping @xmath7 can be decomposed as follows @xmath8 where the vector @xmath9 describes the position of the center of mass and @xmath10 is the rotation associated with the angular velocity of the solid . in dimension 3
the angular velocity is a vector field whereas it is only a scalar function in dimension 2 .
however , @xmath11 can be immersed in @xmath12 and this scalar function can be read on the third component of a 3d - vector . more generally in this work , since all the calculations made in dimension 3 make sense in dimension 2 , we will consider only vector fields of @xmath12 and matrix fields of @xmath13 .
for instance , @xmath14 and @xmath15 are related to each other through the following cauchy problem @xmath16 will represent the identity matrix of @xmath13.]where in dimension 2 we have @xmath17 and @xmath18 .
+ the couple @xmath19 describes the position of the solid and is unknown , whereas the mapping @xmath20 can be imposed .
this latter represents the deformation of the solid in its own frame of reference and is considered as the control function on which we can act physically .
when this lagrangian mapping @xmath20 is invertible , we can link to it an eulerian velocity @xmath21 through the following cauchy problem @xmath22 without loss of generality , we assume that @xmath23 , for a sake of simplicity . if @xmath24 denotes the inverse of @xmath20 , we have @xmath25 the fluid flow is described by its velocity @xmath26 and its pressure @xmath27 which are assumed to satisfy the incompressible navier - stokes equations . for @xmath28 satisfying a set of hypotheses given further , the system which governs the dynamics between the fluid and the solid is the following @xmath29 @xmath30 @xmath31 @xmath32 where @xmath33 and where the velocity @xmath34 is defined by the following change of frame @xmath35 the symbol @xmath36 denotes the cross product in @xmath12 .
the linear map @xmath37 can be represented by the matrix @xmath38 . in equations and , the mass of the solid @xmath39 is constant , whereas the inertia moment depends _ a priori _ on time . in dimension 2
the inertia moment is a scalar function which can be read on the inertia matrix given by @xmath40 in dimension 3 it is a tensor written as @xmath41 the quantity @xmath42 denotes the density of the solid , and obeys the principle of mass conservation @xmath43 where @xmath44 denotes the jacobian matrix of the mapping @xmath5 . for a sake of simplicity
we assume that the solid is homogeneous at time @xmath45 : @xmath46 in system , @xmath47 is the kinematic viscosity of the fluid and the normalized vector @xmath48 is the normal at @xmath49 exterior to @xmath50 .
it is a coupled system between the incompressible navier - stokes equations and the differential equations - given by the newton s laws .
the coupling is in particular made in the fluid - structure interface , through the equality of velocities and through the cauchy stress tensor given by @xmath51 indeed , the dirichlet condition partially imposed by the deformation of the solid ( through the velocity @xmath34 ) influences the behavior of the fluid whose the response is the quantity @xmath52 in the fluid - solid interface . it represents the force that the fluid applies on the solid , and then it determines the global dynamics of the solid ( through equations et ) and thus its position .
+ the problem is the following : what is the deformation @xmath28 of the solid we have to impose in order to stabilize the environing fluid and thus induce a behavior of the fluid which stabilizes the velocities of the solid ?
we shall assume a set of hypotheses on the control function @xmath28 , that we state as follows : h1 : : for all @xmath53 , @xmath20 is a @xmath54-diffeomorphism from @xmath55 onto @xmath56 .
+ h2 : : in order to respect the incompressibility condition given by , the volume of the whole solid has to be preserved through the time . that is equivalent to assume that @xmath57 h3 : : , which leads us to assume that @xmath58 h4 : : , which leads us to assume that @xmath59 imposing constraints and enables us to get the two following constraints on the undulatory velocity @xmath34 @xmath60 as equations and are written , the constraints and are implicitly satisfied in system . hypotheses * h3 * and * h4 * are made to guarantee the _ self - propelled _ nature of the motion of the solid , that means no other help than its own deformation enables it to interact and to move in the surrounding fluid . +
the existence of global - in - time strong solutions for system has been studied in @xcite in dimension 2 and more recently in @xcite in dimension 3 . in particular , this existence in dimension 3 is conditioned by the smallness of the data , namely the initial condition @xmath61 and the displacement of the solid @xmath62 will represent the identity mapping of @xmath63 . ] ( in some sobolev spaces ) . for the full nonlinear system , the equations are written in the eulerian configuration , and thus we are lead to think that the eulerian velocity @xmath21 is the more suitable quantity to be chosen as a control function ( instead of @xmath28 ) .
but such a mapping is defined on the domain @xmath64 , which is itself defined by @xmath20 .
moreover , the study of such a nonlinear system is based on the preliminary study of the corresponding linearized system which is @xmath65 @xmath66 @xmath67 @xmath68 and where the more suitable control to be chosen is the function @xmath69 , related to the lagrangian velocity @xmath70 by @xmath71 notice that the constraints and are nonlinear with respect to the mapping @xmath28 .
we linearize them when we consider the linear system . for this linear system , the constraint induced by hypothesis * h1 * can be relaxed , since we only consider mappings @xmath72 continuous in time and such that @xmath73 .
thus the notion of _ admissible control _ for this linearized problem is made precise in definition [ deflincontrol ] .
the main result of this first part is theorem [ thstablinx ] , which is equivalent to the following one : assume that @xmath74 . for all @xmath61 satisfying @xmath75 and the following compatibility conditions @xmath76 system is stabilizable with an arbitrary exponential decay rate @xmath77 , that is to say that for all @xmath78 there exists a boundary control @xmath79 and a positive constant c , @xmath80 or @xmath81 will denote some generic positive constants which do not depend on time or on the unknowns .
] depending only on @xmath61 such that for all @xmath82 the solution @xmath83 of system satisfies @xmath84 for proving this theorem we study the system that has to satisfy the functions @xmath85 and the goal is then to prove that there exists a control @xmath69 such that this system admits a solution @xmath86 bounded in some infinite time horizon space .
the strategy we follow is globally the same as the one used in @xcite , at least for the linearized problem .
it first consists in rewriting the full nonlinear system in space domains which do not depend on time anymore , by using a change of variables and a change of unknowns
. then we can make appear all the nonlinearities ( specially those which are due to the variations of the geometry through the time ) and we can set properly the linearized system .
the second step of the proof consists in formulating the linearized system in terms of operators where the pressure is actually eliminated and encodes a _ mass - added effect_. this writing enables us to define an analytic semigroup of contraction generated by an operator which presents interesting spectral properties : indeed the unstable modes that we have to stabilize are actually countable and in finite number . besides , we prove by a unique continuation argument the approximate controllability of this linearized system .
thus , in order to define the boundary control that stabilizes the full linearized system , it is sufficient to consider a finite - dimension linear system for which the approximate controllability is equivalent to the feedback stabilizability .
the aforementioned control can be defined on a finite - dimension space in a feedback operator form , what will be useful for proving the stabilization of the full nonlinear system . + with regards to the methods , a novelty is the means provided in a last section which enables us to define from a boundary control an internal deformation satisfying the linearized constraints .
this result too will be useful for the definition of a deformation of the solid - which has to satisfy the nonlinear constraints - that stabilizes the nonlinear system : considering a deformation which satisfies in a first time the linearized constraints is not necessarily from a mathematical point of view , but the method with which we obtain it will be important for defining a deformation satisfying the nonlinear constraints ( see section 5 of part ii ) ; besides , from a physical point of view , considering for the linearized system a deformation which satisfies the linear constraints is relevant since it ensures the conservation of the momenta for the whole fluid - solid linear system .
+ the idea of considering first the linearized problem relies on the fact that for small perturbations ( that is to say for small initial conditions @xmath87 , @xmath88 and @xmath89 ) the behavior of the nonlinear system is close to the one of the linearized system .
+ + thus the same statement will be proven in the second part of this work for the unknowns of the full nonlinear system . the result is nonintuitive : it says somewhat that all the fluid in which the solid swims can be stabilized just by the help of this swimmer , at an intermediate reynolds number .
this kind of problem has been investigated in @xcite for instance , where it is considered other types of fluid - swimmer systems .
the same kind of purpose has been also investigated at a low reynolds number in @xcite and more recently in @xcite , for the stokes system , and in @xcite in the case of a perfect fluid .
the control of the motion of a boat at a high reynolds number has been recently studied in @xcite . besides
, the same kind of techniques that we use in this work have been used for other coupled systems involving the incompressible navier - stokes equations ; let us cite @xcite and @xcite for instance , where the stabilization of other fluid - structure problems is proven .
definitions and notation are given in section [ secdef ] .
the linearized system is studied in section [ linearsec ] where it is rewritten through an operator formulation for which we prove useful properties .
the approximate controllability of this linearized system is proven in section [ secapproxcont ] .
it leads to the main result of this paper , namely the feedback stabilization of the linearized system in section [ secfeedback ] .
finally in section [ secahah ] we provide a means to recover an internal deformation of the solid from a feedback boundary control which lives only in the fluid - solid interface .
this final section is the transition to the second part of this work where the considered control is an internal deformation of the solid .
we denote by @xmath90 the domain occupied by the fluid at time @xmath45 , and by @xmath91 the domain occupied by the solid at @xmath45 .
we assume that @xmath63 is smooth enough .
we set @xmath92 let us introduce some functional spaces .
, we use the notation @xmath93^d$ ] or @xmath94^{d^k}$ ] , for some positive integer @xmath95 , for all bounded domain @xmath96 of @xmath11 or @xmath12 .
we classically define @xmath97 let us keep in mind the continuous embedding : @xmath98 we finally set the spaces dealing with compatibility conditions @xmath99 for more simplicity , we assume that the density @xmath42 at time @xmath45 is constant with respect to the space : @xmath100 we assume without loss of generality that @xmath101 .
this implies in particular @xmath102 let us state the conditions we shall assume on mappings chosen as control functions . [ deflincontrol ]
a deformation @xmath28 is said admissible for the linear system if @xmath103 and if for all @xmath82 it satisfies the following hypotheses @xmath104
@xmath105 @xmath106 @xmath107 @xmath108 for what follows , we need to introduce some operators .
let us first remind the notation of the stress tensor of some vector field @xmath26 @xmath109 and let us denote the hessian matrix operator as @xmath110 for @xmath111 and @xmath112 , we define @xmath113 and @xmath114 as being the respective solutions @xmath115 and @xmath116 of the neumann problems @xmath117 for @xmath118 , we define @xmath119 as being the solution of the stokes problem @xmath120 similarly we define @xmath121 as being the solution of @xmath122 the operators @xmath123 and @xmath124 are called _ lifting operators_. we also define the following integration operators @xmath125 we denote by @xmath126 the so - called leray or helmholtz operator , which is the orthogonal projection induced by the decomposition @xmath127 then we denote in @xmath128 the classical stokes operator by @xmath129 with domain @xmath130 .
let us first consider system only when @xmath131 ; in that case we denote by @xmath132 the unknowns . but this system can be transformed into a system whose unknowns are only @xmath133 . indeed , by following the method which is used in @xcite or @xcite for instance ,
the pressure @xmath27 can be eliminated in the equations , and . by this
means we obtain that @xmath27 can be written @xmath134 where @xmath135 is solution of the following neumann problem @xmath136 and @xmath115 is solution of this other neumann problem which involves the boundary conditions @xmath137 moreover , @xmath138 can be expressed through a lifting ; more precisely , we have @xmath139 thus , we can split system into two systems , one satisfied by @xmath140 , and the other one satisfied by @xmath141 . explicitly , by denoting @xmath142 , system
can be rewritten ( for @xmath143 ) as follows @xmath144 with @xmath145 , & \quad & \mathbb{m}_{add } = \left [ \begin{matrix } 0 & 0 & 0 \\ 0 & \mathcal{c}n & \mathcal{c}\hat{n } \\ 0 & \hat{\mathcal{c}}n & \hat{\mathcal{c}}\hat{n } \end{matrix } \right],\end{aligned}\ ] ] and @xmath146.\end{aligned}\ ] ] the operator @xmath147 represents a _ mass - added _ effect .
we are going to see that it contributes to making the `` effective mass operator '' - namely @xmath148 - self - adjoint and positive .
let us set @xmath150 @xmath151 is self - adjoint and positive .
observe that @xmath152 is self - adjoint and positive .
then it is sufficient to show that @xmath147 is self - adjoint and non - negative .
let us begin with noticing that @xmath153 is self - adjoint . indeed ,
if @xmath154 and @xmath155 denote respectively @xmath156 and @xmath157 , by using twice the green formula we get @xmath158 we can also see that @xmath159 is self - adjoint .
if @xmath160 and @xmath161 denote respectively @xmath162 and @xmath163 , by using twice the green formula we get @xmath164 likewise , let us show that @xmath165 .
first , we denote @xmath166 , and by using twice the green formula , we have @xmath167 on the other hand , if we denote @xmath168 , let us notice that @xmath169 . and then @xmath170 then we can conclude that @xmath147 is self - adjoint . in order to prove that @xmath147 is non - negative , let us see that the corresponding quadratic term can be written @xmath171 for @xmath172 , @xmath173 and @xmath166 . in the following
, we will denote @xmath174 [ propsa ] the domain of @xmath175 is @xmath176 and @xmath177 is self - adjoint .
let @xmath178 and @xmath179 lie in @xmath180 .
we set @xmath181 and @xmath182 , that is to say that for @xmath183 we have @xmath184 @xmath185 @xmath186 with @xmath187 we calculate @xmath188 and by integration by parts we get @xmath189 then , by swapping the roles of @xmath190 and @xmath191 , it is easy to see that @xmath192 it shows that @xmath193 is self - adjoint . since @xmath151 is self - adjoint , the proof is complete .
+ the resolvent of @xmath175 is compact . for @xmath194
, we consider the system @xmath195 where @xmath196 is the unknown .
this system can be rewritten as @xmath197 @xmath198 @xmath199 with @xmath200 by the same kind of calculations as the ones made in the proof of proposition [ propsa ] , this problem is equivalent to the following variational problem : @xmath201 with @xmath196 , @xmath202 , and @xmath203 by choosing @xmath204 , we use the lax - milgram theorem and prove that problem has a unique solution .
thus @xmath205 is invertible , and since @xmath151 is positive , @xmath206 is also invertible .
the results of the two last propositions yield the following theorem .
[ thanalsg ] the operator @xmath207 is the infinitesimal generator of an analytic semigroup on @xmath208 , and the resolvent of @xmath175 is compact .
[ propoperator ] the triplet @xmath209 is solution of system if and only if @xmath210 and @xmath211 satisfy the operator formulation @xmath212 with @xmath213 , @xmath214 , and @xmath215.\end{aligned}\ ] ] let us first remind that @xmath69 must obey an incompressibility constraint given by @xmath216 thus we can formally extend @xmath69 in the whole domain @xmath3 while assuming that @xmath217 in @xmath218 .
that is why we can make the control appear only in the equation , the one which deals with @xmath219 .
let us remind the linear system
: @xmath220 @xmath106 @xmath221 @xmath222 the boundary condition on @xmath223 can be tackled by using a lifting method , like in @xcite for instance .
it consists in splitting the velocity @xmath224 and the pressure @xmath225 , so that we have @xmath226 and @xmath227 @xmath228 @xmath229 @xmath230 this system can be formulated as follows @xmath231 with @xmath232 , @xmath233 and @xmath234 , & \quad & \mathbb{b}_{\lambda}^{(0 ) } = \left [ \begin{array } { c } \lambda l_0 \\ \displaystyle \mathcal{c}(-2\nu d + na_0)l_0
\\ \displaystyle \hat{\mathcal{c}}(-2\nu d + na_0)l_0 \end{array } \right ] .\end{aligned}\ ] ] notice that @xmath235 .
the duhamel s formula gives @xmath236 and an integration by parts leads to @xmath237 notice that @xmath238 and that @xmath239.\end{aligned}\ ] ] thus we have @xmath240 and finally system can be expressed formally in the form given by . for @xmath241 , @xmath242 and @xmath79 system admits a unique solution @xmath83 such that @xmath243 with regards to the formulation and the properties of the operator @xmath175 , this result can be deduced from proposition 3.3 of @xcite .
in order to prove that system is exponentially stabilizable , let us first show that it is approximatively controllable .
[ thapproxcont ] system is approximately controllable , in the space @xmath244 by boundary velocities @xmath79 satisfying @xmath216 note that the operator formulation given by proposition [ propoperator ] does not enable us to write system as an evolution equation .
instead of exploiting this operator formulation , we directly use the writing and the definition of approximate controllability , as it is done in @xcite .
+ let us show that if @xmath245 then the reachable set @xmath246 at time @xmath247 , when the control @xmath69 describes @xmath248 with the compatibility condition @xmath249 is dense in the space @xmath250 that we endow with the scalar product @xmath251 for that , let @xmath252 be in @xmath253 .
we want to show that @xmath254 .
+ let us introduce the adjoint system @xmath255 @xmath256 @xmath257 @xmath258 by integrations by parts from systems and , we obtain @xmath259 thus we deduce that if @xmath260 , then we have @xmath261 for all @xmath79 such that @xmath262 .
this is equivalent to say that there exists a constant @xmath263 such that @xmath264 we can consider @xmath265 instead of @xmath266 , which does not modify the system . thus we get @xmath267 in @xmath268
. then equations and become @xmath269 now , let us look at @xmath270 and @xmath271 with the condition @xmath272 and the following homogeneous dirichlet condition which is deduced from and @xmath273 we can use an expansion of the solution @xmath274 of system @xmath275 in terms of the eigenfunctions of the stokes operator , similarly as it is done in @xcite thus the approximate controllability problem is reduced to showing that if @xmath276 with @xmath277 , then @xmath278 in @xmath218
. then we get the following unique continuation result ( see @xcite ) @xmath279 then we have only @xmath280 an energy estimate leads us to @xmath281 combined to , we get @xmath282 and thus @xmath283 . using a result from @xcite
, this implies @xmath284 the condition enables us to conclude @xmath285 then the proof is completed .
the adjoint system introduced in - can be written in terms of operators ; indeed , by denoting @xmath286 , we can formulate this system as follows @xmath287 where @xmath175 is self - adjoint , so that we can write @xmath288 the adjoint operator of @xmath289 ( whose expression is given in proposition [ propoperator ] ) can be expressed as @xmath290 , where @xmath291 is defined as the solution of @xmath292 ( see lemma a.4 of @xcite for more details ) .
then , proving the result stated above by using the classical characterization of approximate controllability instead of using directly the definition would lead to other difficulties .
given a control @xmath293 with the compatibility condition @xmath249 ( which can be chosen in a feedback form , as explained in the previous section ) , we want to define an internal deformation of the solid , denoted by @xmath294 , which is _ admissible _ in the sense of definition [ deflincontrol ] .
in particular we look for @xmath294 which satisfies the linearized constraints given by , and such that @xmath295 we also want the norm of @xmath296 in the space @xmath297 to be controlled by the norm of @xmath69 in @xmath298 .
[ lemmaxistence ] for @xmath303 satisfying @xmath249 system admits a unique solution @xmath304 in @xmath305 , for @xmath306 large enough .
moreover , there exists a positive constant @xmath307 such that @xmath308 besides , if @xmath309 , and if @xmath310 and @xmath311 denote the solutions associated with @xmath312 and @xmath313 respectively , then @xmath314 the compatibility condition assumed for the datum @xmath69 is useless for the proof of proposition [ lemmaxistence ] , but contributes to making the mapping @xmath294 so obtained an _ admissible _ control ( in the sense of definition [ deflincontrol ] ) .
the constraints imposed in definition [ deflincontrol ] are equivalent to the following ones expressed in term of @xmath304 : @xmath315 thus we have to verify that the mapping @xmath304 solution of satisfies these constraints .
the first constraint , which corresponds to , is satisfied thanks to the compatibility condition assumed for @xmath69 .
for the two other ones , let us remind that we have @xmath316 and since the tensor @xmath317 is symmetric , we also have @xmath318 then equation leads us to @xmath319 and thus by using the divergence formula we get the two other constraints . instead of solving directly system ,
let us first consider a lifting of the nonhomogeneous dirichlet condition .
we set @xmath320 the solution of the following dirichlet problem @xmath321 with the classical estimates ( see @xcite , the nonhomogeneous dirichlet condition can be lifted by theorem 3.4 of chapter ii , and the resolution made by using exercise 3.4 and theorem 3.2 of chapter iii ) : @xmath322 then by setting @xmath323 , we are interested in solving the following system @xmath324 for some @xmath325 large enough , in the space @xmath326 in a first time . a solution of this system can be obtained as a fixed point of the following mapping @xmath327 where @xmath328 is the solution of the classical elliptic system @xmath329 for proving that this mapping is well - defined , let us give some preliminary estimates .
the equality @xmath330 leads in @xmath331 to @xmath332 and then the poincar inequality provides a positive constant @xmath333 such that @xmath334 we can estimate the norm @xmath326 as follows @xmath335 the trace of @xmath336 on @xmath223 which appears in the expression of @xmath337 can be estimated as follows @xmath338 with @xmath339 , for some @xmath340 which can be chosen small enough .
thus , by taking the inner product of the equality by @xmath341 , we obtain @xmath342 by using in the left - hand - side , it gives @xmath343 we now use the young inequality on the right - hand - side , by introducing some @xmath344 which can be chosen as small as desired , as follows @xmath345 with @xmath346 and @xmath347 .
then we have @xmath348 thus , by setting @xmath349 and by choosing @xmath350 small enough , and @xmath351 and @xmath352 large enough , we can see that the ball @xmath353 is stable by the mapping @xmath354 . by the same inequalities we can see that @xmath354 is a contraction in @xmath353 , and
thus @xmath354 admits a unique fixed point in @xmath326 .
+ the same method can be applied in order to prove the regularity in @xmath355 . indeed ,
since we have the equality @xmath356 , the gradient satisfies a similar equality , as follows @xmath357 so that we can show that @xmath358 lies in @xmath326 .
then we have the estimate @xmath359 and since @xmath360 , we have @xmath361 the estimate which deals with the time - derivative of @xmath328 can be obtained easily . indeed , by taking the inner scalar product of the equality @xmath362 by @xmath274 , we notice that the contribution of the right - hand - side force vanishes , as follows @xmath363 because @xmath274 satisfies the constraints @xmath364 by this means we get easily @xmath365 ( mr1387461 ) [ 10.1080/03605309608821198 ] c. fabre and g. lebeau , _ prolongement unique des solutions de lquation de stokes _ , ( french ) [ unique continuation of the solutions of the stokes equation ] , comm .
partial differential equations , * 21 * ( 1996 ) , no . 3 - 4 , 573596 .
( mr2371113 ) [ 10.1016/j.anihpc.2006.06.008 ] j. p. raymond , _ stokes and navier - stokes equations with nonhomogeneous boundary conditions _ , ann .
h. poincar anal .
non linaire , * 24 * ( 2007 ) , no . 6 , 921951 . ( mr2393436 )
[ 10.1007/s00205 - 007 - 0092 - 2 ] j. san martn , j. f. scheid , t. takahashi and m. tucsnak , _ an initial and boundary value problem modeling of fish - like swimming _ , arch .
anal . , * 188 * ( 2008 ) , no . 3 , 429455 .
( mr2027754 ) [ 10.1007/s00021 - 003 - 0083 - 4 ] t. takahashi and m. tucsnak , _ global strong solutions for the two - dimensional motion of an infinite cylinder in a viscous fluid _ , j. math .
fluid mech . ,
* 6 * ( 2004 ) , 5377 .
|
this paper is the first part of a work which consists in proving the stabilization to zero of a fluid - solid system , in dimension 2 and 3 . the considered system couples a deformable solid and a viscous incompressible fluid which satisfies the incompressible navier - stokes equations . by deforming itself
, the solid can interact with the environing fluid and then move itself .
the control function represents nothing else than the deformation of the solid in its own frame of reference .
we there prove that the velocities of the linearized system are stabilizable to zero with an arbitrary exponential decay rate , by a boundary deformation velocity which can be chosen in the form of a feedback operator .
we then show that this boundary feedback operator can be obtained from an internal deformation of the solid which satisfies the linearized physical constraints that a _ self - propelled _ solid has to satisfy . sbastien court ( communicated by the associate editor name )
|
Subsets and Splits
Data Mining Abstracts
This query searches for and retrieves 2000 documents containing the phrase 'data mining' in the abstract, providing limited but specific document samples.